(David Ingram) First Language Acquisition Method

First Language Acquisition
Method, Description, and Explanation
*c;
David Ingram
Department of Linguistics
The University of British Columbia
llK Unrvcrw has
Cambridge University Pres

Cambridge
New York New Rochelle Melbourne Sydney
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 1RP
40 West 20th Street, New York, NY 10011, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
0Cambridge University Press 1989

First published 1989
Printed at The Bath Press, Avon

#
British Library cataloguing in publication data

Ingram, David.
First language acquisition: method,
description and explanation.
1. Children. Language skills. Acquisition.
I. Title.
401l.9
Library of Congress cataloguing in publication data

Ingram, David, 1944- .
First language acquisition: method, description, and explanation I
David Ingram.
p. cm.
Bibliography.
Includes index.
ISBN 0 521 34109 4. ISBN 0 521 34916 8 (pbk)
1. Language acquisition. I. Title.
P118.144 1988.
401’.9 - dc19 88-16215 CIP
ISBN 0 521 34109 4 hard covers

ISBN 0 521 34916 8 paperback
CE
For my mother,
Mary Strailman
Contents
Preface Page ix
1 Introduction 1
Further reading 4
PART I FOUNDATIONS 5
The history of child language studies 7
2.1 The period of diary studies (18761926) 7
2.2 The period of large sample studies (1926-1957) 11
2.3 The period of longitudinal language sampling (1957 to
present) 21
Further reading 30
Stages of language acquisition 32
3.1 Some possible meanings of ‘stage’ 32
3.2 Some proposals on stages of acquisition 38
3.3 Descriptive vs. explanatory stage 54
Further reading 57
Explanation and language acquisition 59
4.1 Introduction 59
4.2 Child Language vs. Language Acquisition 60
4.3 A theory of acquisition 63
4.4 Theoretical assumptions about language acquisition 69
4.5 Sources of variation among children 77
Further reading 80
PART I1 MILESTONES 81
5 The period of prelinguistic development 83
5.1 Introduction 83
5.2 Infant speech perception 84
5.3 Infant speech production 96
vii
viii CONTENTS
5.4 Early cognitive development 115
5.5 The linguistic environment 127
Further reading 137
6 The period of single-word utterances 139
6.1 The definition of word acquisition 139
6.2 Early word comprehension and production 140
6.3 The explanation of early word meaning 155
6.4 Pragmatic and grammatical development 160
6.5 The onset of phonemic perception and production 178
Further reading 231
7 The period of the first word combinations 234
7.1 Defining the period of the first word combinations 234
7.2 Grammatical development: an overview 236
7.3 The grammatical analysis of early word combinations 261
7.4 Current theoretical approaches 302
7.5 The methodology of grammatical analyses of children 332
Further reading 337
8 The period of simple sentences: phonological and semantic
acquisition 340
8.1 Introduction 340
8.2 The phonological acquisition of single morphemes 341
8.3 The further development of word meaning 394
Further reading 432
9 The period of simple sentences: acquisition of grammatical
morphemes 435
9.2 Morphological acquisition in English: a descriptive overview 439
9.3 The acquisition of Aux in English questions 454
9.4 Other aspects of English grammatical acquisition 465
9.5 Cross-linguistic morphological acquisition 493
9.6 The explanation of morphological acquisition 499
9.7 Linguistic input and grammatical acquisition 506
Further reading 513
10 Concluding remarks 516
Bibliography 519
Author index 549

General index 561
Preface
This enterprise officially began in April, 1984, in the Dallas-Fort Worth

airport while I was undergoing a seven-hour flight delay. The original idea,
conceived a year earlier, was to present an exhaustive encyclopedic review
of child language. My feeling at the time was that the field contained a
massive literature which was being lost in recent years. In that year,
however, my orientation changed rather dramatically. For one thing, my
own work was concentrating on methodological issues in analyzing phono-
logical and grammatical samples from children. Methodology took on a
greater importance for me than it had ever done before. At the same time,
the field began a self-assessment, which still continues, of its theoretical
underpinnings. The collection of data for its own sake came under particlar
criticism, and new theoretical issues came to the forefront regarding the
learnability of linguistic constructions and the relevance of acquisition data
for linguistic theory. By the onset of writing, therefore, my goals had
changed considerably. I now saw my purpose as one of providing a more
balanced view of methodology, descriptive review, and theory.
As should have been expected, this ambitious enterprise took a long
time, slightly over three years, to complete. One problem, of course, was
trying to maintain the balance between these three areas. With any
particular study, I have tried to discuss the issues at stake, the relevant
methodological decisions, and the most appropriate and insightful literature
available. One consequence of this was a sacrifice in the area of descriptive
review. It was impossible to provide exhaustive literature reviews while
simultaneously giving in-depth methodological and theoretical information.
Instead, I opted for a selection of representative studies in the areas
covered. The belief was that the deeper coverage of selected studies would
better prepare the reader to undergo his or her own reading of other works
than would a superficial and incomplete treatment of a larger number of
studies.
Despite the decision to restrict the literature review, the book still ended
up being much longer than expected. Even so, it still only covers the core
areas of phonology, morphology, syntax, and semantics. The potential
ix
X PREFACE
effects of parental speech are discussed, but only in relation to these core
areas. Relatively little is said about areas such as pragmatics and conver-
sational patterns, and nothing appears on related areas such as second
language acquisition, bilingualism, twin speech, language disorders, or
spelling. Their inclusion would have led me into additional volumes. In
addition, material originally intended for the text has since been eliminated,
to be included instead in a separate accompanying Workbook. This
material includes information on language sampling, bibliographies, review
questions, and exercises.
This brings me to the potential audience. The book is directed towards
those interested in the acquisition of the structure of language, rather than
its use in everyday conversation. Further, it is for those who are relatively
serious about the field of child language acquisition. Students who would
use it are primarily those who are in either upper-level undergraduate or
graduate courses. I have, for example, used it during the last three years in
an undergraduate course for linguistics students which requires a previous
course in linguistics. I also hope it will be of value to professionals in the
field. One group might be researchers who would like a more detailed
review of areas outside their primary research interests. Another potential
reader is the practicing language clinician who would like to update his or
her knowledge of normal language acquisition.
To acknowledge all those who contributed either directly or indirectly to
the book would require pages. Certainly all the authors of the numerous
works I have read over the last twenty years deserve recognition for their
contributions to our field. I would like, however, to single out a few
individuals who have played an important role in my thinking in some of the
areas of language acquisition. Part of my continued interest in phonological
acquisition is no doubt motivated by the wonderful colleagues I have in this
area. These include Mary Louise Edwards, Larry Leonard, John Locke,
Marcy Macken, Lise Menn, Richard Schwartz, Larry Shriberg, and Carol
Stoel-Gammon. My understanding of grammatical and semantic acqui-
sition has been greatly assisted by the opportunity to discuss the area, either
briefly or at length, with people such as Elizabeth Bates, Lois Bloom,
Martin Braine, Robin Chapman, Richard Cromer, David Crystal, Bruce
Denving, Paul Fletcher, Alan Kamhi, Judith Johnston, Stan Kuczaj, Brian
MacWhinney, Jon Miller, and Catherine Snow. My knowledge of the more
recent research from a nativist perspective owes a debt to discussions with,
and the opportunity to hear presentations by, Stephen Crain, Helen
Goodluck, William O’Grady, Marianne Phinney, Janet Randall, Tom
Poeper, Larry Solan, and Ken Wexler.
I would like to single out four people for a special intellectual acknow-
ledgement. During most of the writing of the book, I have had the unique
Preface xi
opportunity to have regular interactions with four of the finest minds I have
ever met - Guy Carden, Henry Davis, Heather Goad, and Cliff Pye. I
cannot conceive of a richer enviroment in which to attempt such an
undertaking. Guy, a theoretical linguist with an appreciation of the value of
acquisition data, was always ready to drop what he was doing to go over a
draft or discuss a particular difficult theoretical point. I have never met
anyone so willing to give of his time to aid the work of others. Henry has a
better combined grasp of theoretical linguistics and the child language
literature than anyone I have ever met. I have enjoyed watching him play
the devil’s advocate, both with theoretical linguists and data-oriented
researchers in child language. His 1987 doctoral dissertation is an outstand-
ing theoretical contribution on language acquisitioa-Heather is a former
undergraduate student in our department who is now studying for her
doctorate in linguistics at the University of Southern California. She has a
keen eye for detail and was very quick to push for evidence when theoretical
claims were discussed in our seminars on language acquisition. Cliff came to
UBC as a postdoctoral fellow after completing his excellent dissertation on
the acquisition of QuichC. For two years I had the opportunity to discuss the
text with him on a daily basis. More than to anyone, I owe him a special
thanks for his constant input. When he left in 1986 to take up a position at
the University of Kansas, he left as both a colleague and a close friend.
I would like to make a personal acknowledgement to my children Jennika
and Daniel. During the last three ye&s I have tried to allot my writing time
so that it would not take away too much from time that I spent with them.
Even so, there were occasions when things had to get done, or when the
stress and preoccupation with the task no doubt left me less than energetic.
They were always supportive, however, and their occasional arm around
the shoulder, as we peered into the computer screen, gave me more support
at times than they ever realized. Lastly, I want to acknowledge the lifelong
support of my mother, Mary Strailman, to whom this book is dedicated.
1 Introduction
The study of children’s language acquisition is a field that comprises a large

body of literature, dating back well over one hundred years. To make sense
of these works, however, is no easy task. One reason for this is that the topic
is one without a discipline. There are virtually no university departments of
language acquisition (or child language as it is often called), and only one
major journal is devoted to it (The Journal of Child Language, since 1974).
Instead, we have language acquisition subfields of other areas such as
linguistics, psychology, education, and communication disorders. As may
be expected, these disciplines approach the topic from very different
perspectives. The result has been a handful of general texts from the view of
one or another of these disciplines, often to the virtual exclusion of others.
This book is consistent with this trend in one sense - it will use as its
foundation the current interests of the field of linguistics. It will differ,
however, in that it will attempt to transcend a limited linguistic view, and
discuss works and issues that have been produced in the other disciplines
above. To do this, it will deal with three central aspects of the study of
language acquisition: methodology, description, and explanation.
By METHODOLOGY I mean the ways in which one decides to approach the
data of language acquisition. If there is any single void in the booklength
treatment of language acquisition, it is the discussion of how to do or
practice it. This dimension covers several topics, from data collection to the
reasons for data collection. Here, the following methodological areas will
be covered:
1, techniques of data collection - from spontaneous language sampling

to experimental procedures such as elicited imitation, comprehension
testing, and metalinguistic judgements;
2. techniques of linguistic analysis, focussing on grammatical and phono-
logical analysis;
3. aspects of measurement - that is, how to decide on appropriate
measures for the analysis of data from both experimental and natural-
istic studies;
1
2 INTRODUCTION
4. approaches to data - from the formulation of testable hypotheses from
linguistic theory to the establishment of inductive generalizations
from the analysis of children’s language.
These are the various procedures needed in order to collect reliable
information on the language of young children.
A second characteristic of child language is that over the years it has
accumulated a great deal of information on children’s linguistic behavior.
For example, we know a lot about when certain behaviors appear, such as
the first words, the kinds of things children say, and the kinds of errors they
make. This body of knowledge does not pretend to explain anything, in the
sense of providing a theory of development, but it constitutes the facts to be
explained. The presentation of the current state of what we know children
do constitutes the DESCRIITION of language acquisition. This book attempts
to provide an in-depth review of these facts, and tries to separate facts from
explanations. For example, to say that a child is at the two-word stage does
not explain anything, but provides something to be explained.
The discussion of the descriptive aspect of language acquisition will be
conducted by a presentation of the following traditional periods of acqui-
sition:
1. prelinguistic development - birth to end of first year;
2. single-word utterances - from around 1 year to 14 years of age;
3. the first word combinations - from around 14 to 2 years of age;
4. simple and complex sentences - the third year of life.
By this point the normal child can be said to have mastered the rudimentary
aspects of most aspects of language, and I do not discuss in any detail the
subsequent development, from age 4 up to adulthood. That period is
represented by a substantial literature that would require another book-
length treatment of its own.
The above periods are described in terms of grammatical developments,
but this is done only to provide a frame of reference. For each period,
information will be provided on phonological, semantic, and syntactic
development, and the child’s linguistic environment. The discussion in
these sections looks carefully at reported findings and selects just those that
appear to result from reliable research. To do this, the focus will be on
depth instead of breadth. Rather than just reporting superficially on the
results of several studies, I will concentrate in more detail on selected
studies of significance. This approach is based on the assumption that
careful scrutiny of the studies on language acquisition is necessary to
establish what is actually known.
The last and most difficult area of language acquisition is EXPLANATION,
Introduction 3
that is, the construction of a theory of language acquisition. Such a theory
will account for how the child acquires language, based on the linguistic
input. This discussion will include the following:
1. an examination of behaviorist, maturationist, and constructionist
views of language acquisition;
2. the discussion of the notion of ‘explanatory’ stage, as contrasted with
‘descriptive’ stage;
3. the proposal of several assumptions about language acquisition that
will allow us to formulate testable hypotheses;
4. the concentration on the relation between the child’s linguistic com-
petence and performance.
Very early in the book I lay out the theoretical assumptions underlying
much of the methodological and descriptive information. It is taken for
granted that the latter two pursuits are only interesting to the extent that
they lead to principles which help explain how the child acquires language.
While these three aspects provide the focus of the book, they are not
presented as separate sections, but are intertwined throughout. The actual
structure of the book is as follows. Part I, ‘Foundations’, attempts to initiate
the reader into the major methodological, descriptive, and explanatory
issues. I discuss the primary methods used to study children’s language, and
concentrate especially on the method of the spontaneous language sample.
The reader is encouraged to carry out an actual language sample if she or he
has not yet done so in order to bring to the studies discussed later the
necessary critical appreciation. This overview also includes a historical
review of the field and of how various methodological and explanatory
developments have occurred. The notion of ‘stage’ is carefully considered,
since it is a term widely used in the literature.
Part 11, ‘Milestones’, provides an introductory treatment of the main
descriptive periods. As far as possible, within the limits of current know-
ledge, I give what is known about the child’s receptive ability, followed by a
discussion of expressive ability at the same point in development. This is
done to show over and over how the child’s receptive ability precedes and
influences expression. In other books this aspect is often ignored or
underemphasized, with the emphasis instead on expressive language only.
This part also repeatedly examines the extent to which we can make claims
about the child’s rule system, based on linguistic behavior. As far as
possible, we will assume that the child’s behavior reflects competence,
unless explicit performance factors can be isolated.
Part I1 also provides a detailed discussion of how to analyze children’s
language. Explicit suggestions are made on how to do phonological,
morphological, and syntactic analyses. A recurrent theme will be the issue
4 INTRODUCTION
of ‘productivity’. This is the concern with separating the child‘s linguistic

forms which are rule-based from those which are imitated or rote-learned.
It is the rule-based or ‘productive’ forms which reveal the most about the
child’s internal linguistic system.
A major goal of the book is to provide the reader with the ability to
practice the field of language acquisition, not just learn about it. Such an
ability requires awareness of all three areas discussed above. We need to
establish principles that explain language acquisition, principles that are
falsifiable through the description of children’s behavior, behavior that
reliably represents the child’s linguistic knowledge. The course of the
relation between these three can begin with any one and be traced to the
others.
Further reading
There have been several introductory texts on the study of language
acquisition of children. The two most used ones appear to be Dale (1976)
Language development: structure and function, and de Villiers & de Villiers
(1978) Language acquisition. A recent book of chapters contributed by
different authors is Language acquisition, edited by Fletcher and Garman
(1979’2nd edn 1986). Still another recent book, Oksaar (1983) Language
acquisition in the early years, has been translated into English from the
original German of 1977. A recent text from a Piagetian perspective is
Language development from birth to three by Anisfeld (1984). Earlier
efforts at a text include McNeill (1970a), Menyuk (1971), Cazden (1972),
and Bloom & Lahey (1978). There is also a highly readable introduction
contained in Clark & Clark (1977).
Part I
Foundations
2 The history of child language studies
The field of child language acquisition is one that has gone through several
changes over the years in both the methods and the theoretical orientation
used. It is my belief that an understanding of the field requires an initial
exposure to its history. My impression is that the field has lost this
perspective and that many investigators are unaware of its rich literature.
This chapter, then, is designed to provide a historical overview that will
discuss the evolution of current methods, descriptive findings, and theoreti-
cal perspectives. It will proceed by discussing three major periods of child
language studies. These periods are identified by the dominant method in
each. They are:
1. The period of diary studies (1876-1926)
2. The period of large sample studies (1926-1957)
3. The period of longitudinal studies (1957-present)
For each period we will look at the major studies, the methodological
approach, the most general findings, and the theoretical orientation.
2.1 The period of diary studies (1876-1926)

2.1.1 Methodological orientation
The first studies on language acquisition began to appear over one hundred
years ago. These were a part of a general interest in child development that
occurred at that time, led in many respects by the work of G. Stanley Hall in
North America and William Preyer in Europe. For the first time in history,
children became the focus of study to determine the way in which they
develop in general. Child language was just one branch of this work.
The method selected for this enterprise was that of the parental diary.
The linguist or psychologist parent would keep a diary of hidher child’s
learning over some period of time. These observations could be specifically
on language, but often they were more generally on everything from motor
development to musical awareness, e.g. Preyer (1889). These diaries were
7
8 THE HISTORY OF CHILD LANGUAGE STUDIES
eventually published, sometimes in their raw form with little commentary

(e.g. Roussey 1899-1900), but more commonly with selected entries and
interpretations (e.g. Vinson 1915). These works came to be known as ‘baby
biographies’, and they provide a rich descriptive foundation for the field.
The diaries show some common characteristics, though it is dangerous to
think of them as homogeneous. They do vary tremendously in their quality
and detail. One feature, already alluded to, is that they are usually the
record of a parent observer. This feature has been considered as both a
strength and a weakness of the approach. An undeniable strength is that
the observer clearly knows the child well; behaviors noted, consequently,
are not idiosyncratic, but presumably either common ones or ones that
mark a new development. In language acquisition, for example, major
milestones may appear and be acquired over a matter of days; the parent
observer will be able to spot these and note their characteristics. By the
same token, diaries have been criticized for being biased in that the parent
observer will only record what s/he sees to be an important development -
other important behaviors may go unnoticed. Without a theoretical orien-
tation, the argument goes, how does one know what to record? One does
not, so one randomly puts down what is noticed. Further, randomness is
introduced by the fact that the diarist records events when possible. Many
published diaries are full of gaps when either the parent or child was away
for various reasons. Data on development during these periods, unfortu-
nately, is gone forever. Despite these problems, many of the diaries of this
period remain as some of our most detailed reports on language acqui-
sition.
2.1.2 Major diary studies

While sporadic observations on children’s language appear throughout the
nineteenth century, the active publication of baby biographies can be dated
from 1876 with the publication of H. Taine’s paper that appeared translated
into English the next year in Mind. Taine reported on his daughter’s
linguistic development from birth to the end of the second year of life, and
his paper stimulated Charles Darwin to return to some notes he had made
on his own son’s language years earlier, as the basis for a report which he too
then published in the following number of Mind. These papers led to much
better and more intensive diaries in both English and other languages. In
Europe, the most extensive general diary was by Preyer (1889) on the
development of his son Axel, a work which contains excellent linguistic
information and which was later translated into English. It was followed by
the first classic work devoted exclusively to child language, Clara and
Wilhelm Stern’s Die Kinderspruche (1907). This book, which unfortunately
has never been published in English, reports on the language development
2.1 The period of diary studies (1876-1926) 9
of their two children, Hilde and Gunter. In English, we have only hints of
this work through the authors’ translated books dealing with general
development, e.g. Stern (1924). Die Kindersprache is perhaps best known
as the firstly widely accepted statement on the stages of language acquisition
which all children are proposed to follow. This question of ‘stage’ will be
returned to in the next chapter.
In North America, the main force of the work on language acquisition
was stimulated by G. Stanley Hall of Clark University. While Hall wrote
little on language acquisition, he stimulated work through his encourage-
ment of baby biographies, such as the one done by Hogan (1898). Hall
contributed also in two other ways. First, as editor of Pedagogical Seminary,
he encouraged the publication of articles on language acquisition. During
this period, the journal was constantly publishing reports on children’s early
language, e.g. Bateman (1916), Brandenburg (1915), Chamberlain &
Chamberlain (1904,1905,1909), Nice (1917,1920), Pelsma (1910). Lukens
(1894) remains a very readable review of the findings to that date on
children’s language acquisition. These early works grew in number so
rapidly that the first bibliography in Pedagogical Seminary by Wilson (1898)
listed 641 entries, many dealing with language. Second, Hall planned to
build at Clark University a Child Study Institute, with an entire floor
devoted to child language. He had already begun a pruject to archive as
much child language data as possible from other languages as well as
English. Unfortunately, both the onset of the World War and a shift in
theoretical focus aborted this ambitious plan.
The goal of the large majority of these works was descriptive. That is, the
biographies were by and large concerned with plotting the facts of language
acquisition, with little concern for theory construction. The result was an
enormous descriptive literature, which unfortunately has been overlooked
by many modern researchers. It would be unfair, however, to say that these
studies were void of theoretical assumptions. When the question arose, the
most frequent view was that the child brought a great deal of internal
linguistic organization to the task. The child was perceived as being very
creative, and capable of discovering the structure of language from its
environment. Taine (1877:258) captures this feeling:
We only help it [the child: DI] to catch them [general ideas: DI] by the
suggestion of our words. It attaches to them ideas that we do not
expect and spontaneously generalizes outside and beyond our cadres.
At times it invents not only the meaning of the word but the word itself
. . . In short, it learns a ready-made language as a true musician learns
counterpoint or a true poet prosody; it is an original genius adapting
itself to a form constructed bit by bit by a succession of original
geniuses; if language were wanting, the child would recover it little by
little or would discover an equivalent.
While there was this view of the child, little effort went toward specifying
the actual principles used by the child to construct a language system.
While the dominant method of this period was the parental diary, this
does not mean that diary studies ended with 1926. Rather, other methods
came along and joined the diary approach. There have been numerous
diary studies conducted since 1926, and the best ones in English have
appeared since then. Clearly, the most cited is the four-volume work by
Leopold (1939-49) on his daughter Hildegard from birth to age 2 years.
While Hildegard acquired both English and German, most discussion has
centered on Hildegard’s acquisition of English. Perhaps the next most
detailed book is that of Lewis (1936, 2nd edn 1951) who reported on the
early language of a boy referred to as ‘K’. Both these books are full of
detailed observations. The most recent major diary is probably that of
Smith (1973) on his son A’s phonological development. The most novel
diary is that of Weir (1962), who tape-recorded the presleep monologues of
her son Anthony from 2;2 to 2;4. Rich diary studies exist for other languages
as well, e.g. French (GrCgoire 1937, 1947), Russian (Gvozdev 1949), and
Polish (Zarqbina 1965).
2.1.3 The value of diary studies

The value of diary studies is enormous in providing a database for the field.
The role these data play are quite diverse. In some cases, the data are rich
enough to provide the basis for a study. In other cases, the data may be
supplemental, that is, used in conjunction with newly collected data to
substantiate a point. Also, the data may be merely suggestive, i.e. providing
an interesting observation that can be followed up with original research.
I .
The studies that are reported on later in the book will sometimes show all
these uses of diary data.
In any research project, the first step is an in-depth review of the
literature. In acquisition, we have one additional step, the careful evalu-
ation of available diaries for relevant data. One of the skills needed for the
study of language acquisition, then, is the ability to extract data from diary
studies. To do it requires careful reading and evaluation of the diaries
relevant to one’s interest. For example, we need to assess the parent
observer’s qualifications to observe certain aspects of language. It is difficult
to get reliable phonological data, for instance, from a diarist with no
phonetic skills; and the semantic study of early word meaning is of dubious
value when we are only given word lists, with little mention of errors of
usage. We are constantly rejecting or overlooking certain aspects in search
of those points of importance.
There is no doubt that the difficulty of pulling out relevant data from a
Table 2.1 A summary of Axel Preyer’s language at 1;5, taken from
Table VI in Ingram (1978), itselfa summary of Preyer (1889).
Production: only two words since 1;l; ‘atta’ which meant ‘going’ and ‘heiss’ (hot).
Comprehension: (a) since 1;1, responds correctly to: ‘where is papa? Mama?’
(b) consistently understands the following words: ‘clock, ear, shoe, chair,
shoulder, foot, forehead, chin, nose, blow, beard, hair, hat, meat, eye, arm,
hand, cheek, head, mouth, table, light, cupboard, flower’.
(c) obeys the following verbs: ‘run, kick, lie down, cough, blow, bring, give,
come, kiss’.
diary has led to either total avoidance of the approach or emphasis on just
one or two highly reliable sources. We would like to argue, however, that
the gains are worth the effort. For one thing, the diaries usually provide a
comprehensiveness that is impossible to replicate starting from scratch. The
collection of daily observations for one or more children for up to several
years is a tremendous undertaking. One can benefit from a parent
observer’s years of effort in a few hours of reading. Second, children vary
greatly and it may only be the occasional child who shows a particular
pattern. An examination of a cross-section of diaries may reveal one such
child, and provide clues for what to look for in seeking another for careful
study.
Let us look at an example of how useful data can be extracted from a diary
record. It is a common observation that children seem to comprehend more
than they produce. What is the database for this observation? Preyer (1889)
is an early study that provides explicit data on this point. In Ingram (1978),
I extracted linguistic data from this diary on Preyer’s son Axel as part of a
general discussion of the relation of language and cognition. Table 2.1 here
summarizes the data from Table VI of Ingram (1978) on Axel’s comprehen-
sion and production at age 1;5.
We can see that Axel has around 50 items in his comprehension while
only producing two words. As far as his production goes, Preyer observes
(p. 131): ‘Characteristic for this period is the precision with which the
various moods of feeling are expressed, without articulate sounds, by means
of the voice. . .’ The general findings that might be classified as the conven-
tional wisdom on language acquisition are based on data such as these.
2.2 The period of large sample studies (19261957)

While diary studies continued on after the First World War, a major shift in
the field occurred around this time. It was the result of the emerging form of
psychology that has come to be known as ‘behaviorism’. Just as the earlier

period was traced to G. Stanley Hall, this has been associated with the work
of Thomas B. Watson. The immediate meaning of behaviorism is simple
enough - a focus on behavior. Isn’t this what has been going on in the baby
biographies?
Behaviorism differed from the previous observations of behavior in two
respects: the role of the child in the learning of language, and the
measurement of observable behavior. Behaviorists wanted to develop a
theory of learning where the child’s changes in behavior were traced back
to, or explained by, observable conditions of the child’s environment. The
emphasis was on observable events in the interaction of the child and its
surrounding linguistic community. Within this view, the child is seen as
passively controlled by the environment; this is in contrast to the belief of
the earlier diarists that the active spontaneous behavior of the child is
central. Taine and others supplied the internal structure and abilities of the
child, i.e. its ‘genius’, that were rejected by the behaviorists as
unmeasurable.
Examining the actual work done on language acquisition during this
period, we find that most of it was devoted to measurement of language
change rather than to a study of the controlling factors of the child’s
environment. That is, as in the period of diary studies, most of the emphasis
was on description rather than explanation. The way this description
proceeded, however, was much different. This difference is captured
starkly by the following remark by McCarthy on diary studies:
Although this wealth of observational material proved stimulating
and suggestive for later research workers, it has little scientific merit,
for each of the studies employed a different method; the observations
were for the most part conducted on single children who were usually
either precocious or markedly retarded in their language develop-
ment; the records were made under varying conditions; and most of
the studies were subject to the unreliability of parents’ reports.
(McCarthy 1954: 494)
The earlier works were of little value because they were unsystematic,
focussed on single subjects, and provided little measurement of the child’s
behavior.
The result of the new emphasis of behaviorism was a series of studies
which, though as varied as the diary studies, attempted to remedy their
perceived deficiencies. The new accounts shared several characteristics. For
one, researchers were interested in determining what could be described as
normal behavior. To do this requires establishing norms through the
observation of large numbers ofchildren. It is the aspect of collecting data
on a large number of subjects that has led me to call this the period of large
sample studies. Here ‘large sample’ refers to the size of the sample of
children used. (This needs to be kept distinct from the use of ‘large sample’
to refer to the amount of language selected from a single child. In this latter
sense, several diary studies are ‘large sample’ studies.)
These studies also controlled for possible environmental influences by
selecting subjects very carefully. The children came from similar socio-
economic classes, and there were equal numbers of boys and girls. Whereas
diary studies tend to be longitudinal studies, i.e. studies of single children
changing over time, the large sample studies tend to be cross-sectional
studies, i.e. studies of different children at distinct ages. If enough subjects
are selected for each age group, presumably typical behavior is observed.
We can then make inferences about the change of behavior over time
without actually observing any one child change its behavior. There would
usually be similar sample sizes for each age group, and the age intervals
were usually separated by equal time periods, e.g. 2-year-olds vs. 3-year-
olds vs. 4-year-olds.
Other factors distinguished these studies, besides the focus on large
samples. A second one was the systematic observation of behavior, as
contrasted with the unsystematic observations of diaries. All subjects would
be studied for the same amount of time for the same behaviors. If a
language sample were collected, it would be of some predetermined size,
e.g. 100 utterances. If it were some specific behavior, a test would be
constructed to be given to all the subjects. Further, the testing or data
collection would be done by the same experimenter so that variations
between children could not be traced to different experimenter styles.
A third difference between these and previous diary studies was the way
in which the data were analyzed. Under the new experimental methodo-
logy, there was an emphasis on measurement. While statistics was still being
developed as a discipline, these studies are full of quantified results, usually
in the form of numerous tables full of proportions and percentages. These
measures were given mostly for grouped data rather than individual
children, since the amount of data collected from any single child was
usually quite small.
The content of these descriptive studies can best be seen by looking at a
historical overview of the major work of the period. Table 2.2 gives a brief
overview of some of the major studies in terms of sample size, the kind of
data, and the topic under study. We date this period as beginning in 1926
because this is the date of the first study of this kind, by M. Smith (1926).
Smith’s study is typical of most. They concentrated on very superficial
aspects of language. Most generally, there were three areas of concentra-
tion: vocabulary growth, sentence length, and correctness of articulation.
Table 2.2 Some general characteristics of some of the major large sample
studies conducted between 1926 and 1957
Study Sample characteristics Topic
Smith (1926) 124 children between 2 and 5; Length of sentences and general
one-hour conversations aspects of sentence development
McCarthy (1930) 140 children between 1;6 and 4;6 Length of sentences and general
50 sentences each aspects of sentence development
Day (1932) 160 children between 2;O and 5;O Study of language in twins
50 sentences each
Fisher (1934) 72 children between 1;6 and 4;6 Study of gifted children
three-hour samples
Davis (1937) .
173 singletons, 166 twins, alI‘ - Comparison of twins with
between 5;6 and 6;6 singletons
50 sentences each
Young (1941) 74 children between 2;6 and 5;5 Comparison of lower- and
six hours of conversation middle-class children
Templin (1957) 430 children between 3;O and 8;O Length of sentences and general
50 sentences each aspects of sentence development
Smith’s (1926) results on vocabulary growth, for example, are still cited
today (e.g. Dale 1976:174) as the norms for English, and Templin (1957)
remains the most common reference for norms of articulatory development
(e.g. Ingram 1976a).
The study of sentence length is of particular interest. In the period of
diary studies, Stern & Stern (1907) had developed stages of acquisition that
were based on the observation that children’s sentences get increasingly
longer. Nice (1925) developed her own stages of sentence length (see
further Chapter 3) that used an explicit measure which she called the
‘mean sentence length’. This was calculated by counting the number of
words in each sentence of a child’s language sample and calculating the
average number of words per sentence. This measure was used in virtually
every study done during this period. In fact, in McCarthy’s (1954) major
review of this period, there is an extensive table that gives the mean
sentence length from several studies across several ages.
The other emphasis of these studies was to apply the results on these
three general areas to different groups of children. The first studies were
concerned with normal children, e.g. Smith (1926), McCarthy (1930), and
Wellman ef al. (1931). Soon, however, other groups were analyzed, e.g.
twins (Day 1932; Davis 1937), gifted children (Fisher 1934), and lower-class
children (Young 1941). This natural development can be traced to the fact
that most of these studies were conducted by students or colleagues of Smith
and McCarthy. Madorah Smith was at the University of Iowa and one series
of research came out of there, published in the University oflowa Studies in
Child Welfare. Dorothea McCarthy was at the University of Minnesota and
Table 2.3 A summary of the methodological design of Templin (1957)
Subjects: 430 total

eight subgroups of 30 boys and 30 girls at each of the following age intervals:
3;O 3;5 4;O 4;6 5;O 6;O 7;O 8;0
Tests: (1) Articulation - 176 sounds tested in selected words ages 3;O-S;0,with separate
word test for those 6;O-8;0.
(a) 3;O-5;0 repeated words after the examiner or spontaneously identified
pictures.
(b) 6;O-3;0 read the text words or repeated them after the examiner.
(2) Discrimination - pairs of syllables were spoken by the examiner and children
had to judge them as ‘same’ or ‘different’.
(a) 3;O-5;0used pairs of real words of objects that were identifiable (59 pairs,
e.g. ‘keys’ to ‘peas’).
(b) 6;O-8;0 used pairs of nonsense syllables (50 pairs, e.g. [sa vs. za].
(3) Sentence development- 50 utterances were elicited and transcribed on site using
toys and picture books.
(4) Vocabulary - receptive vocabulary was assessed through standardized tests.
(a) 3;O-5;0Ammons Full-Range Vocabulary Test
(b) 6;O-8;0 Seashore-Eckerson English Recognition Vocabulary Test.
directed several studies that appeared in the Monograph Series of the

University of Minnesota Institute of Child Welfare.
The character of these studies can be captured by looking at one of the
best of them, Templin (1957), which also, coincidentally, marks the end of
these studies as conducted over this 31-year period. Table 2.3 summarizes
its basic methodology. The subject size is the largest of all the studies, and
suggests that it was designed to be the definitive one of its type. The four
aspects examined include the three mentioned above plus a discrimination
task. The results are given, by and large, through extensively quantified
data. There are, for example, 71 tables and nine figures. Here is just a
sample of the table headings for the analysis of sentence structure:
(2.1) a. Table 39. Mean number of words per remark of boys and girls,
upper and lower socioeconomic status groups, and total subsam-
ples by age (p. 79).
b. Table 44. Mean percentages of total remarks in each sentence
construction category for boys and girls by age as found by
McCarthy, Davis, and Templin (pp. 86-7).
C. Table 48. Number and percentage of various types of subordinate
clauses used by boys and girls, upper and lower socioeconomic
status groups, and total subsamples, by age (p. 93).
Just as diary studies continued on after 1926, large sample studies have
continued beyond 1957. The change, however, has been a change in focus to
a new methodology. In the case of large sample studies, there has been a
noticeable drop in their number since 1957. Perhaps the major one since
then has been Olmsted’s (1971) study on the phonological development of
100 children.
2.2.2 Strengths and weaknesses

To benefit from the extensive data collected by these studies, we need to be
aware of their strengths and weaknesses. They have three glaring weak-
nesses that have led to their virtual rejection by modern linguists. One is
their lack of linguistic sophistication, a fact that led to superficiality of
content. Language, for example, is much more than vocabulary, sentence
length, and speech sounds. It is a system of rules, and insights into the
acquisition of these rules is at the core of the study of language acquisition.
An example of this linguistic naivetC is found in the work of Irwin, who
between 1941 and 1952 published numerous articles on the development of
speech sounds in infants from birth to 2 years (see McCarthy 1954 for
references). In many of these papers, Irwin refers to the ‘phonemes’ of the
young infant between birth and 1year; the term phoneme, however, refers
to a linguistic unit that is meaningless when applied to the prelinguistic
vocalizations of infants.
A second weakness is the focus on grouped data rather than on the
patterns of individual children. Language functions as a system of rules or
units that interact with each other. It is usually impossible to see these
interactions in grouped data unless it was planned for ahead of time. For
example, data on what auxiliary verbs appear at what age do not tell us
much about how rules that affect auxiliaries, such as Subject-Auxiliary
Inversion, are acquired; and norms of sound acquisition do not reveal much
about how the individual child acquires a system of phonological rules.
A third weakness, ironically, is a methodological one. Most of these
studies, like the diary ones, were done without the aid of modern recording
equipment. Instead, most language samples were done by someone writing
the child’s sentences down as quickly as possible. One should be highly
worried about data gathered in this fashion. The problem is worse when we
consider phonetic transcription. Templin transcribed all of her data herself
with no other transcribers, and without the use of a tape-recorder. She says
(1957:19): ‘The use of recording equipment is not efficient when recording
must be done in many places under varying and often unsatisfactory
acoustic conditions.’ With diary studies, at least, we are dealing, in the
better ones, with a linguistically trained person who is familiar with the
child. It is for these three weaknesses that modern investigators tend to
reject these descriptive studies, and if they look at any data, it is from one or
two selected accounts.
I
As with the earlier diary studies, the weaknesses in these accounts are not
sufficient to lead us to disregard them outright. Rather, they need to be
examined with these concerns in mind, for despite the above problems they
have strengths that need to be considered. Normative data can be helpful in
several ways. For professionals who need to identify children as either
precocious or delayed, the norms are a useful step in that decision. The
norms help give researchers an idea of just how typical a particular child is.
It is easy in a diary study to assume that a child is usual when in fact it is not.
Let me give two examples of this. In Leopold’s famous diary of Hildegard,
he observed that Hildegard replaced adult initial /f-/ in words like ‘fall’ with
a [w] or [VI. We could assume this is a typical pattern in English acquisition.
Ingram et al. (1980), however, did not find a single instance of this in their
study of 73 children! Velten (1943) observed that his daughter Joan
produced only two vowels, [a] and [u], for a very long time. An examination
of any reasonable number of children (e.g. Ingram & Mitchell to appear)
will show that this is highly unusual. Normative data will place any
individual child within a larger picture of development. Normative data also
give researchers an idea of where to begin in the study of a particular topic.
For example, we may wish to study the acquisition of auxiliary verbs. The
large sample studies can tell us what ages will be the most fruitful for our
study.
Besides norms, large sample studies provide data that themselves can be
used for analysis. One approach, which has been sparingly used, is to go to
the original data. Take, for example, the data collected by Young (1941).
Recall from Table 2.2 that she collected six hours of language samples from
74 children between 2;6 and 5 3 . Despite our reservations about the data
themselves, they could be used to yield initial results on a tremendous range
of questions.
Another way to use these studies is to reanalyze the data to suit one’s own
goals. Throughout this book, we will emphasize the need to work with data
and to manipulate them in various ways. Even though a particular diarist or
experimentalist presents and interprets data in a certain way, we are not
obliged to accept hidher analysis or interpretation. In some cases, we may
find new results for these data by trying out new measures. In other cases,
we may even conclude that the interpretations do not follow from the data
as presented. We can return to the large sample studies and view them in
relation to issues not actually considered by the original researcher.
Examples of this process will appear throughout Part 11.
The emphasis of the large sample studies on methods is one that is
important and valid. We can learn from their concerns an awareness of the
kinds of problems we face when we do not consider questions of subject
selection and research design. In particular, they have made an invaluable
contribution with their emphasis on measurement. A continuing weakness

of the linguistic analysis of a single child is that the analyses will vary
tremendously depending on which sentences we decide are produced by
rules. While measures do not tell us the nature of a rule, they will allow us to
decide when a rule exists in some form. It is a basic premise of this book that
some form of measurement is a prerequisite to our linguistic analysis of a
child. The ways in which the large sample studies scrutinized data through
various measures are important in showing how new patterns can be
isolated. Once identified, the linguistic patterns can be subjected to
linguistic analysis.
2.2.3 Theoretical orientation

While most work during this period was descriptive, it would be unfair to
leave it without a discussion of some attempts to explain acquisition from the
behaviorist viewpoint. The behaviorists wanted to explain acquisition by
assigning to the childvery little innate behavior. On thisview, the child would
be born with a few very general skills; let us consider what these may be. First,
it would be able to vocalize. Second, it would be able to process auditorily the
vocalizations of others, and presumably be able to recognize similarities
between these and its own. Thirdly, it would be able to pair a vocalization
with some context, for example, it would have the ability to relate the vocali-
zation ‘mommy’ with the context of the parent. Let us call this the ability to
form associations. Fourth, the child is born with basic drives that will moti-
vate it to form associations. For example, one of these drives is the drive to
nourish itself. The actual associations that are formed, such as the one
between the word ‘mommy’,the internal state of hunger, and the person who
is the child’s mother, are the result of its experience. They can occur because
the child has one further ability, the ability to be conditioned, i.e. to build up
associations by being exposed to events juxtaposed to internal drives.
To explain language acquisition for the behaviorist, then, is to determine
the set of environmental conditions that lead the child to identify and
associate events with internal states. There were a few attempts to do this
during the period of large sample studies. Probably the first to receive
attention was the attempt of Bloomfield in his classic book Language (1933:
ch. 2 ) . Table 2.4 gives the five steps Bloomfield proposes to account for the
child’s acquisition of word meaning. The child recognizes similarities
between its vocalizations and the adult’s, and then associates a particular
speech event with a context. These five steps center on the child’s pro-
duction of speech. Bloomfield points out that the child also develops the
role of a hearer. The child will become reinforced to act in certain ways
when he hears certain speech forms, such as ‘Wave your hand to Daddy.’
Table 2.4 Five steps in the acquisition of language, according to
Bloomfield (1933: 29-31)
1. Child innately vocalizes, and can recognize sounds which are similar and which are
different. He develops the habit of repeating a familiar speech sound, e.g. [da].
2. When someone such as the mother produces a word similar to one of his babblings, e.g.
‘doll’, he will imitate it with his closest speech form, e.g. [da].
3. The mother’s use of ‘doll’ in the context of dolls will lead the child to associate the sounds
with the event of seeing the doll. The sight of the doll becomes a stimulus for saying [da].
4. The habit of saying [da] in specificcontexts, e.g. seeing one’s doll after one’s bath, will lead
the child to say [da] in that context when the doll is absent. That is, speech becomes
displaced.
5 . His successful attempts at speech are reinforced, leading him to adult-like pronunciation.
His imperfect attempts are lost.
This account of acquiring early word meaning attempts to combine the

role of speaking and listening. Bloomfield states (p. 31):
After he has learned a number of such twofold sets, he develops a
habit by which one type always involves the other: as soon as he learns
to speak a new word, he is also able to respond to it when he hears
others speak it, and, vice versa, as soon as he learns how to respond to
some new word, he is usually able, also, to speak it on proper
occasion. The latter transference seems to be the more difficult of the
two.
He sees the child as first acquiring a word separately in comprehension and
production, and only later connecting the two.
There are several aspects of this view of learning that should be noted.
For one, it certainly accounts for some part of early word acquisition. The
child needs to imitate adult words in some sense of the word ‘imitate’. Also,
children do repeat speech a great deal, suggestive of some form of practice
on the part of the child. Furthermore, it is common to see children who have
been conditioned to act to language, for example, taught to wave bye-bye or
say thank you. The crucial question, to which we will return, is whether or
not it is sufficient. For example, Bloomfield says (p. 30) that children do not
ever invent words, but only have imperfect exposure to the correct range of
meanings. Is the process described above sufficient to explain the complex
set of meanings that the adult speaker of a language possesses? As we will
see later, others have argued that much more innate knowledge needs to be
proposed to account for this.
A second aspect of this theory is the important fact that it is testable: that
is, it makes predictions about acquisition that can be falsified. We can
observe young infants and their mothers to see if children imitate adult
words with utterances from their own phonetic repertoire. Or, we can see if
there is a one-to-one relation between the conditioning of the child and its
word acquisition. We can see if the child acquires words in comprehension
and production separately at first and only later simultaneously. Or, for
word meaning, we can compare the use of a child’s word to the contexts in
which it was first taught. Throughout the book, we will emphasize the
importance of developing testable theories of acquisition.
A third aspect of Bloomfield’s view is that it appears to be focussed on
pronunciation and early word use. How does he account for the child’s
acquisition of grammar? Here Bloomfield has less to say. He has no theory
that explains how the syntax of the language is acquired, although he does
make mention of the role of analogy in historical change (p. 275). He does,
however, emphasize the role of correction by the parent. For example, he
says (p. 31): ‘if he says Daddy bringed it, he merely gets a disappointing
answer such as No! You must say “Daddy brought it”.’ This does not tell us
how the child establishes the grammar for the correct utterance, but it
proposes that parents must do a tremendous amount of language teaching.
As we shall see later, this prediction is not borne out by current evidence.
The major attempt during this period to explain language acquisition is
that of Skinner (1957). As pointed out in its Preface, the book covers his
research from 1934 to 1955. It is the culmination of his theoretical work
during these years and it is ironic that it was published the same year as
Templin’s study. These two works constitute two of the reasons I place the
end of this period in 1957. Skinner’s proposals are an elaborate develop-
ment of the basic notions of stimulus, reinforcement, and association,
introduced above. They do include, though, some attempts to account for
the learning of syntax. To oversimplify: the structure of a sentence consists
of a chain of associations between the words in the sentence. Suppose, for
example, that the child knows the words ‘dog’ and ‘run’, and hears them in
the sentence ‘The dog is running.’ The child may imitate this as ‘dog run’ and
be positively reinforced or rewarded. The word ‘dog’ becomes a response
by the child to some stimulation, say seeing a dog run; and ‘dog’ in turn
becomes a conditioned stimulus for the word ‘run’. The development of
grammar for the child, then, is the learning of a set of associations between
words that can lead to classes of words. A grammar will be a set of classes
that occur in a predicted serial order, based on these associations.
As is apparent from even as brief an account as this, Skinner allows for
very little innate language structure. This point of view, so prevalent during
this period, was attacked heavily by linguists in the next period who
emphasized the hierarchical complexity of language. Nonetheless, behav-
iorism has continued on through today although most of the research has
been on associations between words apart from syntax. There is a major
work by Mowrer (1960) to account for language in behaviorist terms, and an
2.3 The period of longitudinal language sampling (1957 to present) 21
early attempt by Jenkins & Palermo (1964) to develop Skinner’s ideas on
how syntax could be acquired without resorting to innate linguistic prin-
ciples. An important question, which we will return to in later chapters, is
the following. The attempt to account for early two-word utterances by
imitation and word association is possible, although it becomes subject to
massive difficulties as an attempt to explain more complex, adult-like
language. When, in the development of language, does the child show
language structure that is too complex to be explained by simple principles
of association and imitation?
2.3 The period of longitudinal language sampling (1957 to present)

The third method used in collecting data on language acquisition is what we
call longitudinal language sampling. In longitudinal language sampling, the
child is visited at predetermined intervals for a reasonable length of time
with the purpose of collecting a representative sample. It can be seen as a
natural outgrowth of the two methods discussed. Diary studies are longi-
tudinal, but they usually consist of notes rather than complete language
samples for some predetermined length of time. (It should be added,
though, that some one-day language samples were conducted, e.g. Bran-
denburg 1915.) Large sample studies took language samples, but they were
normally quite short, and not longitudinal.
In the late 1950s, three independent groups of investigators developed a
renewed interest in studying language acquisition in children. These were
Martin Braine at Walter Reed Hospital in Bethesda, Maryland, Susan
Ervin and Wick Miller at the University of California, Berkeley, and Roger
Brown at Harvard University. They each developed their own method of
longitudinal language sampling. Their work was followed by that of Lois
Bloom of Columbia University. We will begin by considering the shared
features of their methods of data collection, then discuss each individually.
The longitudinal language samples differ from the diary studies in that the
subjects are usually not the offspring of the investigators. Instead, they are
children who are selected specifically because they meet predetermined
criteria; for example, they are usually on the verge of beginning multiword
utterances and are talkative. Also, more than one child is studied; in three
of the four studies above, three children were selected. This number is used
as an absolute minimum necessary to determine general features of
acquisition: if one is chosen, we do not know if the child is typical or not; if
two, we do not know which of the two is typical and which is unusual; with
three, we at least have a majority that can be used to make such a decision.
Table 2.5 General information on four major studies using longitudinal

language sampling
Children (age
Investigator range in months) Sampling schedule
Braine (1963a) Andrew (19-23) parental diary of all multi-word utterances pro-
Gregory (19-22) duced. For Steven, there were tape-recordings for
Steven (23-24) four hours over a four-week period (12 sessions.)
Miller & Ervin (1964) Susan (21- ) initially weekly in 45-minute sessions; later every
Lisa (24- ) two months for 2 or 3 sessions for 4-5 hours.
Christy (24- ) Sampling over a two-year period.
Harlan (24- )
Carl (24- )
Brown (1973) Adam

Eve
[$-{)
two hours every two weeks; two observers present.
Sarah (27-44) half-hour every week.
Bloom (1970) Eric (19-26) eight hours over three or four days, every six
Gia (19-27) weeks.
Kathryn (21-24)
The longitudinal language samples incorporate some of the experimental

concerns of the large sample studies. The children are visited on a regular
schedule at predetermined times for a predetermined amount of time.
There are sometimes two visitors, one to take notes and one to interact with
the child. All sessions are tape-recorded for later transcription.
There is one additional feature which separates these studies from the
large sample ones. In the latter, sampling was often very short, up to the
collection of 50 or 100 utterances (see Table 2.2). With longitudinal
sampling, however, much larger samples are sought, so that a more
representative sample of the child's general language ability is obtained.
Table 2.5 presents some general information on the four studies men-
tioned above. Braine (1963) used a method that is between the diary study
and the collection of longitudinal samples as described. He asked the
mothers of his three subjects, Andrew, Gregory, and Steven, to write down
every unique two-word utterance they produced. This was done during the
time when two-word utterances are not very common. Later, he initiated
systematic tape-recording. The use of a parental diary combined with
longitudinal sampling is a sensible technique which, interestingly enough,
has not been used very frequently.
The studies of Miller & Ervin (1964), Brown (1973), and Bloom (1970) all
involved systematic visits. Miller and Ervin started with short, frequent
visits in the early months, then switched to longer, infrequent visits. The
latter schedule is similar to the decision by Bloom. A six- to eight-week
interval presumably allows the child to show considerable change. It misses,
on the other hand, possible crucial periods of change which may occur over
a day or two: such changes can only be captured by having a flexible
schedule that allows emergency visits when alerted by the parent of
changes.
These studies provided a database that is quite unlike that of the other
two periods. We can understand this shift in the method when we examine
the motivation for these studies. In 1961 there was a famous conference on
first language acquisition in Dedham, Massachusetts. The papers from that
meeting, including ones by Bloom and Miller & Ervin, were published in
1964, edited by Ursula Bellugi and Roger Brown. The editors say the
following about this new wave of studies (p. 5):
Quite recently, investigators in several parts of the United States have
begun research on the acquisition of language as-it-is-described-by-
linguistic-science. This new work is concerned, as the earlier work
usually was not, with phonemes, derivational affixes, inflections,
syntactic classes, immediate constituents, and grammatical trans-
formations.
Many of the previous diary and large sample studies were not only
descriptive, but often looked at only superficial issues, such as when
particular features appeared in the child’s speech. In these later studies
researchers wanted to look for the emergence of rules, and to describe the
developing grammar of the child. They wanted, for example, to go beyond
dating the appearance of two-word utterances to writing rules of their
structural properties.
The shift of interest in language acquisition toward rule-based descrip-
tions of child language coincides with a paradigmatic change in linguistics
stimulated by the publication of Noam Chomsky’s Syntactic structures
(1957). In that book, Chomsky redefined the goal of linguistics from one of
description to one of explanation, with a theory of grammar which placed
syntax at its center. This new field of transformational grammar stimulated
research on syntax where previous linguistic research had concentrated
more on phonemics and morphology. Chomsky defined grammar as a set of
rules which generates the grammatical sentences of the language. In
language acquisition, the goal became to establish how the child acquires
rules of sentence formation. The primary focus of the Dedham meeting was
the nature of the child’s early grammar, and it was the primary interest of all
of the investigators listed in Table 2.5.
It is of interest that Chomsky was invited to attend the Dedham
conference as a discussant. His brief comments on the meeting (Chomsky
1964) remain an important comment on the field of language acquisition. In

his discussion, Chomsky pointed out the important distinction between
performance and competence, one elaborated upon in Chomsky (1965).
The role of the linguist is to describe linguistic competence, that is, the
underlying rule system that every native speaker of a language has. This rule
system manifests itself in performance, i.e. when a speaker talks and listens
to language. He states: ‘It should be clearly recognized that a grammar is
not a description of the performance of a speaker, but rather of his linguistic
competence’ (p. 35). Language samples, of course, are examples not only
of performance, but of performance through speaking only. Because of
this, Chomsky was highly critical of the use of only language samples for
writing grammars for children.
It is important to look at some length at his comments on methodology.
Here is a paragraph which sums up his general suggestions (p. 36):
. . . it seems to me that, if anything far-reaching and real is to be
discovered about the actual grammar of the child, then rather devious
kinds of observations of his performance, his abilities, and his
comprehension in many different kinds of circumstances will have to
be obtained, so that a variety of evidence may be brought to bear on
the attempt to determine what is in fact his underlying linguistic
competence at each stage of development. Direct description of the
child’s actual verbal output is no more likely to provide an account of
the real underlying competence in the case of child language than in
the case of adult language . . . not that one shouldn’t start here
[emphasis mine: DI], perhaps, but one surely shouldn’t end here. . .
There are two central points made by Chomsky that need to be emphasized
and which constitute central themes of the present book. One has to do with
the kind of data used to make claims about the child’s language. I emphasize
the last remark about language sampling because I believe it is an important
and necessary starting point. It is with this in mind that the above studies
were conducted. Both Brown (1973) and Miller & Ervin (1964) talk about
using other means of testing to supplement the language sample. One can
start with language samples of various kinds to get an initial insight into
some linguistic topic. Then, most likely, it will be necessary to switch to
another procedure such as comprehension testing or elicited imitation.
Unfortunately, this has not always been the practice. One reason is that it
can take so long to transcribe data that the child has often changed its
grammar by the time we return; this is a practical problem, though, that can
be resolved by doing shorter initial transcriptions. Alternatively, we can use
completed transcripts to get an initial idea of where to look, and then begin
original elicitation that focusses on the issue being studied. As we will see in
Part 11, there are a number of creative techniques that have been developed
in recent years that go beyond language sampling.
The second point Chomsky makes in this extract concerns the difference
between a child’scompetence and its performance. To what extent can we say
that a child’s performance, as reflected in a language sample, reflects a child’s
competence? If there is a tremendous difference, then child language
research is left to study the performance factors that limit the child in speak-
ing, such as memory, attention, processing time, etc. In this case, language
acquisitionisverymuch abranchof psychology. If they areclose together, i.e.
if we can determine agreat deal about the child’s competence or rules from its
performance, then it is much more a branch of linguistics, in that we can talk
about the form of the child’s rules, the way they become more adult-like, etc.
Chomskyis rather pessimisticon this point and thus, not surprisingly, has seen
little that language acquisition can contribute to linguistictheory (see Ingram
1985a). In Chapter 4,however, we will argue that language acquisition can
contribute to linguistics if we assume a closer relation between competence
and performance. Throughout Part 11, this will be demonstrated in the dis-
cussion of the acquisition of particular linguistic subsystems.
2.3.2 Theoretical orientation

It has already been mentioned that the studies of this period have been
driven by an interest in the development of linguistic rules. The assumption
that the child has systematic behavior, however, still leaves a great deal of
latitude regarding the extent to which linguistic structure is learned and the
extent to which it is innate. Here we give a simplified overview of the major
theoretical positions that can be taken.
So far, we have considered briefly behaviorism which assigns very little
internal structure to the child except general abilities such as the ability to
form associations and be conditioned. In Aspects of the Theory of Syntax
(1965) Chomsky expressed a completely different position, which we will
refer to as nativism. In this view, language is seen as an extremely rich and
complex system, certainly more than a series of associations between words
that are linearly ordered. Language consists of hierarchical structure and at
least two levels of representation, a deep or underlying structure (or
D-structure, Chomsky 1981) that is mapped by transformations onto a
surface structure (or S-structure) . Universal principles of language apply to
restrict what can be a possible grammar of any language. The collection of
these universal principles which determine the form of any humanly
possible language is referred to as Universal Grammar (UG). Nativism
argues that these universal principles (or UG) are innate, that is, they are
part of the genetic program the child is born with.
There are two major arguments put forth for this position. One follows
from the fact that language is highly creative. We continually produce
sentences that we have never heard before, and our grammar can
potentially generate an infinite number of sentences. If the child learned
language using rote learning and imitation, we could never account for the
rate of acquisition that occurs. By a very young age, children are capable of
producing a range of sentence structures that is nearly comparable to that of
the adult. It is argued that this ability could not be attained without highly
complex innate ability. Second, the language the child hears is ‘degenerate’
in the sense that it does not provide the child with the information necessary
to acquire linguistic structure. Take, for example, the sentence ‘Is the boy
who left happy?’ It consists, among other things, of a moved auxiliary ‘is’
that agrees in number with a noun phrase ‘the boy’, which in turn is
modified by a relative clause ‘who left’. There is nothing in the phonetic
string of the sentence, [1~6abxhulcfthaepi],that tells the child about the
structure, or the nature of possible rules that move constituents. This kind
of knowledge about the possible structure of the language is part of the
child’s Universal Grammar.
Another example of this can be given in reference to our earlier
discussion of the acquisition of the meaning of the word ‘doll’ (cf. Table
2.4). Bloomfield proposed that the child learns this by pairing the vocali-
zation of ‘doll’ in the context of the object doll. What stops the child from
thinking ‘doll’ means ‘a doll seated’, or ‘a pink thing’, etc., in fact an almost
infinite range of possibilities? There will need to be some innate limitations
on what ‘doll’ could mean in that context, these being restrictions on
possible relations between cognitive and semantic categories (cf. Jacken-
doff 1983).
Within nativism, we suggest that there are two possible positions, the
maturationist vs. the constructionist. These two views have to do with the
way in which the principles of Universal Grammar became available to the
child. The maturationist view, which can be assigned to Chomsky, states
that the principles are released or become available to the child at some
genetically determined time. They could come quite early, like the ability to
walk, or later, like puberty. The constructionist view, on the other hand,
does not allow linguistic behavior to change due to maturation, but rather
accounts for all changes by a building up of structure. What occurs later
does not replace what occurred earlier, but rather builds upon it. The
constructionist view, then, is a more restricted view of development in that
it allows less change over time than a maturationist view. The point will be
returned to in detail in Chapter 4.
The constructionist view has been most clearly articulated in psychology
by Piaget (e.g. Piaget 1971). It focusses on the importance of determining
sequential stages of development. Piaget (1971) has discussed the fact that
his view is nativist, but emphasizes that it is nativism with a focus on
development. For example, he states (p. 16): ‘one still needs to explain in
detail how, in the field of knowledge as in that of organic epigenesis, this
collaboration between the genome and the environment actually works’.
Much of the research on language acquisition done in this period seems to
have a constructionist slant, although it is not always explicitly articulated.
The distinction between these two views can be used to isolate what
appear to be two distinct fields of language acquisition that have evolved in
the last decade or so. So far, we have used the terms ‘child language’ and
‘language acquisition’ interchangeably. Wasow (1983:191), however, has
preferred the following distinction:
There has been, for some years now, a fairly sharp split in the field of
developmental psycholinguistics between what I will call researchers
in ‘child language’ versus those in ‘language acquisition’. Child
language research is concerned primarily with what children say; that
is, it focusses on the data. The central concern of most child language
research is on data collection and classification, with correspondingly
close attention to data collection techniques, and relatively little
concern for abstract theoretical issues. Language acquisition special-
ists . . . regard child language data as interesting only to the extent that
it bears on questions of linguistic theory.
Henceforth, I’ll use this distinction by Wasow, and use capitals when doing
so, Le. Child Language vs. Language Acquisition.
It is now possible to characterize the theoretical approach of the
longitudinal studies cited above. They can be called part of Child Language
in that they focus on the data, and are concerned with data collection
techniques. They also, however, have a theoretical orientation that tends to
be constructionist, that is, they attempt to determine stages of acquisition
that follow from the analysis of data. Wasow (1983) comments on this in a
footnote (p. 91), stating that this kind of theorizing tends to be inductive,
Le. ‘the hypotheses are generated from patterns observed in the corpora’.
Wasow makes one further point about the theoretical work in Child Lan-
guage. He states that its hypotheses are local, i.e. investigators ‘posit
strategies for dealing with particular phenomena at a specific stage of
development, rarely attempting to relate to general issues in linguistic
theory’ (p. 191). For purposes of reference, I’ll call those people who
pursue Child Language in this fashion psycholinguists.
The field of Language Acquisition, on the other hand, starts with
linguistic theory and then turns to questions of language acquisition. It
considers language development in two ways. One, it views theories of adult
language and isolates predictions those theories make about language

acquisition. For example, a theory might say in one area that one construc-
tion or kind of rule is more basic or unmarked than another. This leads to a
prediction that the child will acquire the unmarked case earlier than the
marked case. For example, a theory might say that adjective + noun
constructions like ‘red apple’ are more basic than relative clauses like ‘the
apple that’s red’. The fact that children acquire adjective + noun construc-
tions early can be said to be confirming data from acquisition. As currently
articulated, however, Language Acquisition does not allow data from
acquisition to disconfirm a linguistic theory. As an illustration, suppose that
our theory states that relative clauses are less marked or simpler than
+ +
adjective noun constructions. The fact that adjective noun construc-
tions are acquired first does not disconfirm this, because one can always say
that the principle which states that relative clauses are less marked has not
matured yet. The maturationist view of Language Acquisition will always
allow one to reject acquisition data as counterevidence. In Chapter 4 I will
argue for a constructionist view that will allow us to use acquisition data as
counterevidence.
The second way in which Language Acquisition discusses acquisition is in
its consideration of the logical problem of learnability. Again, we begin with
theories of grammar and concern about how we may restrict them. This
approach may be demonstrated by looking at the sentences below (2.2a-d).
The (d) sentence has an asterisk in front of it because it is ungrammatical in
English.
(2.2) a. I gave a book to Mary
b. I gave Mary a book
c. I reported the crime to the police
d. * I reported the police the crime
We can propose a theory that says there is a rule of English called Dative
which moves an indirect object next to the verb, producing a sentence like
(2.2b). Dative has not applied in sentences (2.2a) and (2.2~).We need to
restrict the Dative rule so that it does not apply in (2.2d). To do this, we can
mark the verb ‘report’ as an exception to our Dative.
This is a possible analysis for these English sentences. The problem,
however, is this. How does the child ever learn that (2.2d) is an ungramma-
tical sentence? The child cannot learn it from positive evidence, Le. the
sentences that it hears, because it will only hear sentences like (2.2a-c).
These three provide the child with the evidence for the rule. The next
possibility is that the child learns it from direct negative evidence, i.e. the
parent corrects the child. This was, in fact, the proposal discussed earlier by
Bloomfield. There is, however, no evidence that this occurs, and Braine
(1971a) has produced several arguments against it: (i) transcripts like those
collected by the studies in Table 2.5 indicate that parents correct mostly for
semantic appropriateness, not grammaticality; (ii) even if we did correct
the child, how does the child know in some cases what was wrong? Also
(iii) transcripts studied to date indicate that children seem to be oblivious to
corrections when they are made (see also Zwicky 1970). Below is an
example of the last point from Braine (1971a: 161):
Child: Want other one spoon, Daddy.
Father: You mean, you want THE OTHER SPOON
Child: Yes, I want other one spoon, please, Daddy.
Father: Can you say ‘the other spoon’?
Child: Other.. . o n e . . . spoon.
Father: Say.. . ‘other’.
Child: Other
Father: Spoon
Child: Spoon
Father: Other . . . spoon
Child: Other . . . spoon. Now give me the other one spoon.
The third possibility is that the child may learn that (2.2d) is ungrammati-
cal by what is called indirect negative evidence. This is knowledge that is
obtained by noting that something has never occurred. For example, the
child would need to calculate the number of times it had heard sentences
like (2.2a-c) and decide that (2.2d) must be ungrammatical because it has
never been heard. This seems an implausible possibility because it would
require the child to do a tremendous amount of calculating. Further, it goes
contrary to the creative dimension of the grammar where the child is
predicted to generate new utterances never heard.
Baker (1979) has proposed that we only consider linguistic analyses that
are learnable from positive evidence. If so, then our Dative analysis is
unlearnable and therefore rejected. The analysis he proposes in its place is
one in which the child must learn the structure underlying (2.2a) and (2.2b)
separately, and not relate them by rule.
The issue of learnability, as formulated and used in Language Acqui-
sition, does not require looking at acquisition data. We return to this in
Chapter 4, however, and show the relevance of acquisition data in Lan-
guage Acquisition, reconciling the two divergent enterprises identified by
Wasow within a field we can call Child Language Acquisition.
Further reading
Bibliographies
The first bibliography on language acquisition was that of Wilson (1898)
which covers much of the period of diary studies. The next is that of Leopold
(1952). This is a very useful reference source which gives brief descriptions of
many of the studies. For example, on Taine (1877), he states (p. 94): ‘The
famous philosopher’s observations of his two children . . . originality of the
child emphasized too much . . .’; and on Chamberlain & Chamberlain (1904,
1905, 1909): ‘interesting case study of daughter to 3;11, mostly third year.
Careful observers; selected observations, not systematic, but thorough.
“Phonetic” transcription unsatisfactory, but improves somewhat as the study
progresses’ (p. 15). Slobin (1972) is an updating of Leopold’s bibliography.
Slobin deletes Leopold’s brief descriptions, and also some of the earlier refer-
ences. It is important, therefore, to double-check Leopold (1952) with regard
to the earlier literature. Besides adding new references, Slobin also provides
appendices specifically on non-English studies. He also provides three valu-
able indices: 1, according to language spoken by child (44 language groupings
are given); 2, according to content of reference (e.g. babbling, phonology,
syntax); and 3, content of reference according to language spoken by child,
that is, cross-listing the first two. The most recent bibliography is Abraham-
sen (1977). Abrahamsen has selected over 1500 entries in the literature that
she felt were relevant to child language research as of 1977. The organization
is topical, with the main divisions as follows: I: general resources; 11: syntac-
tic development; 111: semantic development; IV: grammar; and V: phono-
logy and orthography. All references are for English (so that Slobin 1972
remains the primary reference for non-English studies) and the studies
thought to be most important are set apart with asterisks.
Readers
There are currently three books of selected readings. Bar-Adon & Leopold
(1971), Ferguson & Slobin (1973), and Bloom (1978). Bar-Adon & Leopold
is the richest for selectionsfrom early researchers. It contains a selection from
Taine (1877) and Darwin (1877) as well as items from other diary works.
Ferguson & Slobin concentrate on non-English studies and provide the only
translations available for some of these, particularly for some important ones
in Russian. Bloom (1978) concentrates on recent work in English although
there are selections from the work of M. Smith (e.g. Smith 1933).
Suggested reading
A feel for the nature of the research done during each of these three periods
can be obtained by reading a sampling of each. A minimum introduction
Further reading 31
would be the following: Taine (1877), Templin (1957:3-18) ‘I: Introduction,
11: The experiment’, Brown (197351-9) ‘The study of Adam, Eve, and
Sarah’. The earliest form of the behaviorist view can be found in Bloomfield
(1933:2&37). Bar-Adon & Leopold’s book (1971) Child language. A book
of readings contains selections from several early works; each selection is
preceded by a brief explanation which helps to set the work in its place in
history.
3 Stages of language acquisition
The word ‘stage’ is possibly the most used term in language acquisition.
The literature is rich with its use, and an article or book rarely comes
along without its appearance. Unfortunately, few people attempt to define
the way in which they are using it. In one respect this is understandable.
Unlike other technical terms, stage is part of our everyday vocabulary,
and we have grown up using it. It is easy to use it with the misleading
assumption that our hearer understands the word in the same way that we
do.
In a scientific enterprise such as language acquisition, however, we need
to be careful to define what requirements we set in order for a stage to be
proposed. As will be shown, there are various ways in which the term can
be used, each with different implications. We need to define ‘stage’ in
order to describe the facts of child language, and these facts are a
necessary data base for the explanation of different stages. Careful analy-
sis of the use of ‘stage’ is crucial to the constructionist view, since we
ultimately want to use stages of acquisition as evidence for a theory of
language acquisition.
This chapter will discuss ‘stage’ by first looking into some of the ways in
which it can be used. Next, it will provide a general review of the stages of
acquisition that have been put forward in the literature. Lastly, it will return
to ‘stage’ and define it in two ways, distinguishing between descriptive and
explanatory stages of acquisition.
3.1 Some possible meanings of ‘stage’

We will see throughout the Child Language literature a multiplicity of uses
of ‘stage’. This section discusses just a few ways in which it may appear, with
the goal of raising our consciousness about this word. In reading research, it
will usually be up to the reader to determine any particular author’s use of
the term, since it is rarely defined. Later in this volume more precise
definitions for its use will be developed.
32
3.1.1 Single behaviors

One way in which we may use ‘stage’, and probably the simplest, is in the
sense of a point on a continuum. This will be referred to as the continuity
requirement for the use of ‘stage’. A continuous stage is one where a single
dimension of behavior is being observed, and the difference between stages
is only between the points along the continuum that are being selected for
observation. We can provide simple examples from time and space. For
time, we can say a child is at the ‘2-year stage’ or ‘3-year stage’ where the
only change being described is one of changing time. Sometimes we see that
these labels say little more than that there is a difference. Or, during one’s
work day, we can say we are at the ‘noon’ stage, or the ‘453 pm’ stage.
Spatially, children can be placed at different stages based on their height:
we can say, for example, that a child is at the ‘2 foot’ stage, or the ‘3 foot’
stage.
The continuity stage can also be proposed for the measurement of some
behavior that is either on the increase or decrease. We can look at both
non-linguistic and linguistic behaviors. Suppose, for example, that a child’s
use of thumb-sucking increases from 20 percent at age 0;2 to 50 percent at
0;4. We could decide that 20 percent is not much thumb-sucking, but that 50
percent is a lot. Then we conclude that the child was not at the thumb-
sucking stage at 0;2, but is at the thumb-sucking stage at 0;4. Notice that this
claim does not sound as bad as our earlier ‘4:53 pm’ stage of working, but it
really does not claim much more: it only focusses at some point on a
continuum. A linguistic example can be given in relation to single-word
usage by young children. Suppose that a child at l ; O uses one-word
utterances such as ‘mama’ or ‘bye-bye’ as 100 percent of its meaningful
vocalizations. At 1;6 the child may use one-word utterances as 70 percent of
its meaningful vocalizations, since it is also now using two-word utterances.
We could conclude that the child at l;O is at the ‘one-word’ stage, but the
child at 1;6 is not. This use of stage is only based on continuity; as might be
anticipated, we would suggest that it does not tell us much about the child’s
organization.
A second use of ‘stage’ is one where we build upon the continuity
requirement and add a new requirement that the continuity has been
halted. We will refer to this as the plateau requirement. Let us first examine
some obvious examples, and then some less obvious ones. We can look at
death as a stage of this sort. For example, we can say that someone who has
died is in his ‘death’ stage. A behavior that was under change, i.e. life
activity, has been halted, in this case permanently. Or, we can look at
someone’s growth which has now ceased, where the person will remain at a
34 STAGES OF LANGUAGE ACQUISITION
certain height, say 5 feet 6 inches. We could then say that she is at the ‘5 foot
6 inch’ stage, meaning she will not change any more. There are less obvious
linguistic examples where we can use stage in this sense. For one, we could
describe a child who is severely retarded, and whose language is perman-
ently arrested at a point where only one-word utterances are used. We could
say that the child is at the ‘one-word’ stage where ‘stage’ refers to a behavior
that is permanently halted at a point on the continuum. Another linguistic
example, which is common, is to use ‘stage’ to refer to the attainment of the
adult behavior on some linguistic feature, such as the use of relative clauses.
This use of ‘stage’ sounds insightful, but does not say much more than does
the continuity requirement above.
A third use of ‘stage’, one that approaches common use, is to add a third
restriction to the two already mentioned, i.e. a transition requirement. The
transition requirement restricts ‘stage’ to cases where the behavior that has
plateaued is expected to change again at some later time. A child who is at
the thumb-sucking stage under this use is a child who has reached some
maximal use of thumb sucking, and will continue on for some time until it
drops. A drop in rate, of course, is not the only direction in which the
change may occur, although it is the most common use. We could have a
plateau, and then an increase. It is common in learning a new skill, for
example, that there is an increase in ability, followed by a plateau, followed
by an increase. The period of the plateau can be referred to as a stage. In
learning to juggle, we may eventually move on to four balls, but our ability
is limited to three balls for the time being. A linguistic example would be the
mean length of the child’s utterances. We might find that the child’s
utterances have been increasing steadily, but that for several weeks their
mean length has been constant, say around three words per utterance. We
could then say the child is at the ‘three-word’ stage, meaning that there is a
temporary halt at this point.
Our third definition of ‘stage’ is a static one, in that it describes a time
when no change has occurred. We can shift this perspective, and instead
limit ‘stage’ to a point where there is active change. In learning, there are
occasions when the changes that occur seem more rapid than others. In
physical growth, these are those periods when the child is said to be
‘shooting up’; in language, these are those periods when the child seems to
.have made a breakthrough and makes rapid progress. In these changes,
there is a sudden increase in use that then remains constant. This can be
captured by adding an acceleration requirement. A ‘stage’, then, is a period
of rapid acceleration in the development of a linguistic ability that will end
in a plateau, i.e. a steady rate of use (possibly final acquisition) afterwards.
A linguistic example is the way children acquire words in production. At
first there is a slow period of growth from around l ; O to 1;6, and then a rapid
continuous stage plateau stage

3.1 3.2
transition stage acceleration stage

3.3 3.4
Figures 3.1-3.4 Graphic representations of four definitions of stage over one
behavior, where the vertical axis measures rate and the horizontal axis measures
time. Arrows indicate the stages.
spurt in the size of the child’s vocabulary. We could refer to this spurt as a
stage of acquisition. Figures 3.1-4 show each of these possible uses of
‘stage’ in terms of how they would appear upon. measurement.
3.1.2 Multiple behaviors

These four kinds of ‘stages’ all refer to a single behavior. We may, of course,
wish to stop here, and say that stages can only be proposed for single
behaviors, and that we do not wish to relate behaviors to each other, i.e.
we may not wish to countenance general stages of development. In
developmental psychology, for example, Jean Piaget has suggested the
Sensorimotor Period from birth to around age 2 years as consisting of six
general stages. Each stage consists of a cluster of behaviors that are related
by general principles. Uzgiris & Hunt (1975), however, have argued against
these general stages, and say that we can only discuss stages of specific
behaviors such as imitation or knowledge of the spatial relations of objects.
In language, we could reject the idea of general stages of acquisition and
concentrate on the stages of specific topics, like the stages of acquiring the
meaning of ‘cat’, or the stages in learning how to pronounce a sound like /t/.
It is common, however, to see researchers go beyond single behaviors
and to use ‘stage’ to refer to the relations between behaviors. The simplest
relation that can hold would be where one behavior has succeeded another.
Let us call this a succession requirement and such a stage a ‘succession
stage’. Imagine someone who likes to dabble in crafts and spends one year
doing painting, and the next year doing pottery. H e has gone through two
such stages, a ‘painting stage’ and a ‘pottery stage’. Or, there are certain
reflexes that infants show for a short time and then lose: these could each be
considered a stage. In language acquisition there is a period of time where
children will overextend the meanings of their words, for example calling all
four-legged animals ‘dog’; it lasts from around 1;6 to 2;6. This behavior
could be isolated and referred to as the ‘overextension stage’.
While some behaviors may drop out, many appear for a first time,
develop over time, and stay in the child’s repertoire of behaviors. We can
then add a co-occurrence requirement, Le. a requirement that a stage is
defined as the point at which at least two behaviors exist together. For
example, there is a time from around l ; O to 2;O when children use
single-word utterances such as ‘mama’. From 1;6 to 2;O these occur along
with some usage of two-word combinations such as ‘see mama’, ‘my
cookie’, etc. We can talk about a ‘two-word stage’ as the period when one-
and two-word utterances co-occur. This kind of stage can be called a
‘co-occurrence stage’. The name of each stage can be assigned by labeling it
according to the dominant or new behavior that appears. In the above
example, the label ‘two-word stage’ is taken from the new behavior.
Suppose that one-word utterances are dominant (Le. more frequent) from
1;6 to 1;9, and that two-word utterances are dominant from 1;9 to 2;O. If we
used the dominant behavior as the label, then only the ldtter time period
would be called the ‘two-word stage’.
There are two further kinds of use of ‘stage’ that can be distinguished. We
will add a third requirement to succession and co-occurrence, which we will
call the principle requirement. This requirement states that a stage only
exists if we propose a principle that accounts for the co-occurrence of the
behaviors. For example, children walk and pronounce their first words
around the same time. It is not likely, however, that we would want to refer
to these two as a stage, since there seems to be no causal relation between
succession stage co-occurrence stage

3.5 (new behavior)
3.6
E
t Principle A
co-occurrence stage principle or implicational stage

(dominant behavior) 3.8
3.7
Figures 3.5-3.8 Graphic representations of four definitions of stage that relate
two or more behaviors, where the vertical axis measures rate and the horizontal
axis measures time. Arrows indicate the stages.
them, that is, their co-occurrence is correlational. Other behaviors may look
similar enough that we may propose a principle to relate them.
The addition of the principle requirement leads to two further kinds of
stage. One kind is what we shall call an ‘implicational stage’. This stage
occurs when the principle we propose to relate behaviors is such that the
occurrence of one behavior necessarily implies the occurrence of the other.
For example, we could propose that children acquire general phonological
features before more specific ones. This claim implies by logical necessity
that the stage in which the child acquires the difference between /p/ and /t/
will follow the stage in which children distinguish between /p/ and /1/
(assuming that class features like [k sonorant] are more general than place
features like [k alveolar]).
The second is what we shall call a ‘principle stage’. This occurs when the
behaviors being explained are such that one does not by necessity imply the
other. For example, children appear to begin to play symbolically with
objects, as in pretending that a block of wood is a car, around the same time
that they begin to acquire language (see Part 11, Chapter 6). Piaget has
proposed that these two behaviors are related by a general principle which is
that the child has developed the symbolic function. This could be referred to
as the ‘symbolic stage’ of development. The stage would be named for the
principle that underlies the behaviors. Figures 3.5-3.8 present a graphic
representation of these possible definitions of stage.
As hopefully is obvious by now, there is a multiplicity of ways in which
‘stage’ can be used. The student of language acquisition will probably find
all the above uses and more when reading works in the field. The above
discussion was provided to present the difficultieswith the term. Section 3.2
will provide an overview of the better-known attempts to describe stages of
language acquisition, and section 3.3 will attempt to limit our use of the
term in future work.
3.2 Some proposals on stages of acquisition

There are numerous proposals for stages of acquisition in the Child
Language literature. Here we will select just three, one from each of the
periods described in Chapter 3, to give an idea how they have developed
over time. As can be seen, they tend to focus on characteristics of
grammatical development. This section provides an overview to the more
detailed treatment in Part 11.
3.2.1 Stern (1924)

Stern & Stern (1907) is generally considered to be the first real classic in child
language. Oksaar (1983:8), for example, states: ‘Child language research
done by psychologists reached its climax with the work of Stern & Stern
(1907). It initiated modern paedolinguistics.’ Though their book Kinder-
spruche was never translated into English, Stern (1924) provides a com-
prehensive statement of their view of the major stages of language acqui-
sition. Table 3.1 gives an overview of these. Stern varies between the terms
‘stage’and ‘period’. He sees development as scanning five time periods. One
he calls a Preliminary stage, the other four he calls periods and numbers them
1to 4. He starts his periods of acquisition at 1;0, with the onset of the first
word, because he feels that speech really begins ‘from the moment in which
the child, for the first time, utters a sound with full consciousness of its
meaning and for the purpose of communication’ (p. 143).
Table 3.1 Summary of Stern’s (1924) Preliminary stage and four periods
of language acquisition
Preliminary stage (first year) This stage consists of three behaviors: (i) babbling, (ii)
unintelligible imitation, (iii) preliminary understanding.
First period (1;O-1;6) The child acquires a small number of sounds with special meanings,
which express the ideas of an entire sentence. There is, however, no evidence that the child
understands grammar.
Second period (1;6-2;O) The child realizes that everything has a name, with: (i) a sub-
sequent spurt in word acquisition, (ii) questions about the names of things. Soon after, the first
multi-word utterances appear, first hesitantly, then more fluently.
There are three stages of vocabulary growth:
(a) substance - increase in nouns;
(b) action - increase in verbs;
(c) relation and distinction - increase in qualifying and relational words.
Third period (2;O-2;6) Sentences become well-formed in that they contain words for the
major grammatical relations, like ‘subject and object’. The acquisition of inflections begins,
and may last for years. Syntax consists of the loose linking of words together where word order
may vary. The range of questions expands.
Fourth period (2;6 on) The simple juxtaposition of words in syntax is replaced by
hierarchical structure and the acquisition of embedded or subordinate sentences. The
acquisition of some grammatical morphemes still continues. The child’s questions now include
those of time and causality.
While there is no language during the first year, Stern sees it as an

important precursor for later acquisition. There are three major activities
during this year: babbling, imitation or ‘echo babbling’, and rudimentary
understanding. At first, these three are independent of one another, but
gradually merge together. When they do, around 1;0, real language begins.
We can see this stage as one of the co-occurrence of these three behaviors,
where the conclusion is marked by their coordination.
Babbling occurs because the child is innately disposed or programmed to
babble. Stern does not use these exact words, but talks about ‘impulses’ and
‘inner energies’. There are two inner drives behind babbling, one the drive
to express oneself, and the other the drive to socialize with others. The drive
to express oneself leads to practice which is necessary for later facility with
words. It also allows the child to establish associations ‘between certain
states of feeling and the sound-forms belonging to them’ (p. 143). The
child’s imitations or ‘echo babbles’ are also the result of an ‘impulse’ or
innate disposition. This activity also allows the child to practice speech.
Rudimentary understanding refers to the child’s responses to certain
utterances such as the clapping to ‘pat-a-cake’ or waving ‘bye-bye’. These
acts do not show understanding of the structure of language, but only an
association of a sound or set of sounds with certain actions and contexts.

Stern is very explicit in his belief that the beginning of language is with the
first words produced, not with comprehension: ‘This first understanding of
speech has at first nothing to do with an intellectual grasp of the logical
significance of words, in fact months pass before this stage is reached’
(p. 143, fn. 4).
The first period begins when the child consciously produces a word with
meaning, around 1;O. For this stage to begin, then, active production is
required. The main behavior of this period is the slow growth of one-word
sentences. Stern emphasizes the gradualness of the development of the
child’s productive vocabulary, and defines the end of the period by a spurt in
vocabulary usage (p. 145):
The child’s power of speech, however, after it has begun, does not
then increase at equal speed from day to day . . . Periods of stagnation
. . . are suddenly replaced by periods of rapid development, to be
followed in their turn by very slow development.
As an example, he cites Axel Preyer who only acquired two words in his
production, ‘atta’ and ‘ta’ over six months. This stage can be seen as a
‘transition stage’ (Figure 3.3) between no words and rapid development.
The selection of words during this period is not random, but is deter-
mined by several factors. First, many words used around the child are
beyond its current cognitive level of understanding. They are more likely to
come within its understanding if they consist of sounds that are part of the
child’s babbling repertoire. There is thus the influence of the child’s
articulation capabilities. Secondly, the words selected tend to be onomato-
poeic, that is, phonetically similar to the sound of things, e.g. ‘moo-moo’,
‘meow’, etc. Stern addresses the interplay of the words presented to the
child by its environment and the desire of the child to express inner needs
and ideas with its own phonetic repertoire. He denies the exclusive
operation of either factor and introduces the notion of convergence:
Here again our idea of ‘convergence’ must come to our aid. Early
speech is the combined result of aptitudes and speech impulses,
inherent in the child’s nature, and of stimulus offered for his imitation
from without. It seldom occurs that a child quite independently raises
a descriptive sound or a natural sound-expression to a real designation
in speech, for he cannot, as a rule, reach such a stage because his
environment deprives him of any opportunity, by offering him ready-
made expressions (wow-wow, etc.). (P. 149)
He allows the child, however, to be quite creative with its first words. For
example, his son Gunter at 1;5 used ‘fff’ to refer to all lights and lamps,
presumably the result of the imitation of blowing them out.
Stern presents a common view of the time when he calls the one-word
utterances of this first period one-word sentences. These utterances have
certain properties: (a) they convey more than just a word - they convey a
whole idea; (b) one word may be used to express several ideas, e.g. ‘mama’
may mean ‘mother, come here’, ‘mother, give me’, or ‘mother, set me on
the chair’ (p. 151); (c) they do not have grammatical structure; (d) they are
not purely statements of propositions but are the inseparable union of the
expression of a concept and the child’s internal needs; (e) they are not
members of classes such as specific nominals (e.g. ‘mama’ for one’s mother)
or general nominals (e.g. ‘dog’ for all dogs), because the child is not yet
cognitively able to generalize from its experiences; and thus, (f) their use is
determined in large part by associative reactions to some present experi-
ence. While he uses the term ‘sentence’, Stern nonetheless denies the
existence of any grammatical structure. This point is in agreement with
others who studied language acquisition during this historical period, e.g.
Jespersen (1922).
The first period ends and the second begins with the child’s rapid
development of vocabulary. In the secund period, vocabulary growth goes
through an ‘acceleration stage’ (Figure 3.4). Around the same time,
another behavior occurs, an interest in learning the names of things by
asking for names with a question like ‘that?’ (p. 164). Sometimes these two
events may occur together as they did with his daughter Hilde around 1;6,
or they may occur quite separately. His son Gunther asked for the names of
things around 1;7 but took several more months before his word spurt. At
first glance this appears to be just a co-occurrence stage (Figure 3.6), but
Stern wants to go further than this. He offers a principle to relate these two;
the child has realized that ‘every thing has a name’. He does not label this
principle, so let us call it the Principle of the Linguistic Sign, Le. the
realization that phonetic events can be paired in a conventional way with
concepts. The rapid acquisition of vocabulary and the desire to seek names
are both the result of this underlying principle.
One feature of this principle is that it is an ‘awakening’ in Stern’s
terminology. This suggests that it is an internal ability that has been
triggered by the previous experience of the earlier first period which was
presented as a transition stage. The child goes beyond the limits of its earlier
associative links between sounds and real-world events.
The insight into the relation between sign and import which the child
gains here is something fundamentally different from the simple
dealing with perceptions and their associations. And the demand that
some name must belong to every object, whatever its nature, we may
consider as a real - perhaps the child’s first - general thought.
(P. 165)
Another feature of the period is that the word spurt may occur first in
comprehension, only later to be evidenced in production. Recall that
Gunther asked for the names of things several months before his productive
word spurt. Presumably his receptive vocabulary was increasing greatly
during this time, although Stern does not say much about it.
Stern goes on to analyze the nature or categorization of the vocabulary
developed during this period and beyond. He proposes three stages:
(a) substance stage
(b) action stage
(c) relation and attribute stage
Since he did not attribute classes of this type to the first period (at least not
in their adult form), we can presume that the Principle of the Linguistic Sign
also contains some guidelines on how to classify. These three stages appear
to be ‘co-occurrence stages’ (Figure 3.6) focussing on new behaviors that
enter the child’s language. In the substantive stage, most of the words are
referring to nominals, i.e. the names of things. Later, in the action stage,
terms for actions enter the vocabulary and grow. These are a new behavior
that co-occur with the nominals. Unfortunately, Stern does not give us
much information on how the stages progress, or what principle or
principles determine them. Nor does he tell us how to classify words into
these categories - a potential problem, given the holistic meaning of the
child’s utterances.
The above developments all refer to vocabulary. A second major
characteristic of the second period is the onset of multiword utterances.
Although he is not very explicit, Stern appears to want to separate these
from the milestones in vocabulary, i.e. the vocabulary spurt and the onset of
multiword utterances co-occur, but they are not caused by the same
principle. The latter appears to occur later than the former, so that we could
interpret from this sequence that the Principle of the Linguistic Sign is a
necessary, but not sufficient, cause of multiword speech. Since they bear
this relation, we can do one of two things, either treat the two separately, so
that the vocabulary is in the second period, when the syntax (Le. one-word
sentences) is still in the first, or treat them as both in the same period, but
define the second period as one composed of two principles, the Principle of
the Linguistic Sign followed by the principle that leads to early word
combinations. We will assume that Stern takes the latter course, although
this is not obvious from his writing.
We can look at the cause of the early multiword combinations and their
structural properties, but as for the cause, again we have to search a bit for
Stern’s position. Three potential causes can be located. One is that the
child’s cognitive advances are so great that the child needs more language
structure: ‘The general cause of all this working-up is a certain “language
famine”. For normally the wealth of experiences and the need of expression
grows far more quickly than the supply of forms of words. . .’ (p. 161). That
is, the child’s needs to express itself lead it to seek or develop new, more
complex expression. Another possible cause could be the child’s com-
prehension, which is in advance of production. Stern comments that his
children understood multiword utterances during the last months of pro-
ducing one-word sentences. For example, Hilde at 1;4showed understand-
ing of ‘Touch your nose with your foot’ one month before her first
multiword utterances in production. The third possible cause is the child’s
growing imitative ability. Stern distinguishes between direct imitation,
which comes immediately after hearing, and indirect imitation, which is
delayed or internal. He states (p. 155): ‘with increasing age direct imitation
. . . falls more and more into the background and leaves indirect imitation to
do most of the work’. If these three aspects come to be coordinated, as the
three factors at the end of Stern’s preliminary stage, then we can say that
their interaction leads to early multiword speech.
Stern is very reluctant to give these first multiword utterances any
structure. Rather, he sees them as loosely juxtaposed, initially in the form
of successive one-word sentences:
It is true that these word collections were not all so fluently spoken as
are our sentences; their characteristic rather was a jerky utterance
(often interspersed by pauses) of the isolated words; in such a case it is
really more a question of a short chain of one-word sentences as: Hilde
- cocaa (come Hilde, the cocoa is here), ater - dolly (Father look, I
have a dolly). It is only by degrees that such loose juxtapositions pass
into the firm union of a really coherent sentence. (P. 166)
Stern has little specific to propose on the structure of these non-syntactic
combinations, except that they are somewhat limited in their creativity. For
example, he states at one point: ‘For a year the words the child uses are fixed
forms only’ (p. 168).
The first major gains in syntax occur in the thirdperiod around 2;O. There
are two major grammatical changes: (i) the onset of the acquisition of
inflections, and (ii) the combination of words by syntactic rules, not just
juxtaposition. He does not cite a specific cause, but implies it is a principle
that leads the child to syntax: ‘The two-year-old child begins to acquire all
this in the different forms of inflection (declension, conjugation, com-
parison) fairly simultaneously, so that, in reality, we are dealing with united
psychic progress along the whole line’ (p. 168). Let us call this the Principle
of Syntactic Structure, one which directs the child to formulate rules of
sentence structure within the restrictions of Universal Grammar and some
theory of acquisition. The fact that Stern believes the source of this principle
is internal is clear from the following remarks:
Now, it can be readily understood that, in spite of great individual
differences in speech-development, certain regularities in the sequence
ofspeech-capabilities are to be found in the child, for general laws of
mental growth are at work here, and it is the consequence of these that
every phase acquires the necessary supply of words and phrases . . .
Since many of these laws act more or less independently of the kind of
speech that influences the child or of any other conditions of environ-
ment, it is evident how strong must be the part played by the inner
faculty and its gradual growth in the development of the child’s speech.
(P. 158)
Despite these innate abilities, the child’s development is still quite slow.
It takes children several years to complete the acquisition of inflections.
Also, the increase in sentence length with greater productivity around age
2-3 years still goes through a great deal of non-fluent production. He gives
the following sequence from Hilde (1;lO) as a request for her mother to go
get some pickles (p. 169):
(3.1) Mama. Want pickies. Room. Wanty pickies. Back. Dada, mama fetch.
Also, the child will move certain words to the sentence-initial or final
position in order to give them prominence. Regarding inflectional develop-
ment, there is just one mechanism that is proposed to account for the child’s
development, the ability to use analogy. This mechanism shows up in other
discussions around this time, as in Jespersen (1922). The child will notice
the regular inflections and apply them to irregular forms, e.g. ‘drinked’ for
‘drunk’ or ‘badder’ for ‘worse’. There is no discussion of how the child
unlearns these, or of how the child’s syntactic rules change over time. Given
the lack of a major syntactic theory at the time, however, this is not
surprising.
Stern’s fourth period starts from around 2;6. It is marked by three
changes. The first is the appearance of subordinate sentences. At first, the
complementizers, relative pronouns, and adverbials that mark the subord-
inate clause may be absent. For example, at 2;6 Gunter said the sentence
in (3.2a) with the meaning of (3.2b) (as translated from German):
(3.2) a. Mother say, builded has Gunter
b. I want to tell mother what Gunter has been building
At 3;O Hilde used each of the sentences in (3.3), again translated from
German (p. 171):
(3.3) a. indirect question I will look in the kitchen and ask if she is coming
here
b. temporal clause Will whip the doll till it hurts her
c. causal clause That moves so today because it is broken
d. conditionalclause You’ll get no bread and butter if you’re so
naughty
e. final clause You must take away the beds so that I can get
out
f. consecutiveclause Dolly has disturbed me, so that I could not
sleep
While these may appear around 2;6, their acquisition will vary from type to
type and may take up to 5;O to be acquired. For example, he cites hypo-
thetical conditional clauses as quite late. He gives the following example
(p. 171) from Gunter, which was used at 4;8 in reference to a friend who
had an operation: ‘If he were (not) cut at once, then he had died.’
Stern, as earlier, does not discuss the principle behind this, nor its relation
to the earlier syntactic development of the third period. Since it is given as a
separate account, we will assume it is a separate principle, and shall call it
the Principle ofSubordination. Further, we will assume that the Principle of
Syntactic Structure is a necessary precursor. In this sense, Stern’s fourth
period is a ‘co-occurrence stage’ labeled for the new behavior that has
appeared (Figure 3.6).
A second and apparently independent change of the fourth period con-
cerns the use of questions. Up until now, the child has asked concrete ques-
tions like ‘what?’ and ‘where?’. Now, the child begins to use the more
abstract questions of ‘why?’ and ‘when?’. These early ‘why’ questions can
be quite persistent, as is evidenced by the following dialogue between Hilde
and her mother at 3;7 (p. 172):
(3.4)
Child: What is he eating?
Mother: Fish.
Child: Why does he eat fish?
Mother: Because he is hungry.
Child: Why doesn’t he eat rolls?
Mother: Because we don’t give him any.
Child: Why don’t we give him any?
Mother: Because bakers only make rolls for people.
Child: Why not for fishes?
Mother: Because they haven’t enough flour.
Child: Why then haven’t they enough?
Mother: Because not enough corn is grown.
Table 3.2 A summary of the major characteristics of Nice’s (1925) stages

of acquisition
~ ~~~
I. Single word stage The child uses single-word utterances.

2. Early sentence stage (1;5.5 at onset). The first multiword utterances appear, although
single-word utterances are still most frequent utterances at onset. Most sentences are
incomplete, and consist of mostly nouns, verbs, adjectives, and adverbs.
3. Short sentence stage (commonly around age 3;O). The acquisition of inflections and
grammatical words begins. The ALS ranges from 3.5 to 4.5. Incomplete sentences become
less frequent, yet compound and complex sentences are rare. The ratio between word
classes becomes stable.
4. Transition stage A period of change where the child moves from incomplete to complete
sentences. The ALS is around 5.0. Little is known about this stage.
5 . Complete sentence stage (around age 4;O). Most sentences are well-formed, and complex
and compound sentences become more common.
This change is apparently a separate co-occurring behavior in the area of

cognition that is not related to the onset of subordinate clauses by any
unifying principle.
A third major behavior that appears in this stage is the creative construc-
tion of new words by the child. Stern emphasizes that this varies greatly
from child to child, with some hardly doing it at all. German is known for its
extensive use of compounds, and Stern found his children each creating
40-50 of their own by age 5 years. Some examples of these compounds from
his children, as translated, are given in (3.5):
(3.5) a. fire box [a fireproof casserole over a stove]

b. child-soldier [an officer who looked short]
c. grey-red [in reference to a color]
d. muffiner [someone who sells muffins]
e. the smoke [a cigar]
f. raggy [torn]
g. splashiness [result of splashing water]
Stern does not discuss the principle behind these changes either, We can
isolate it as separate one from those discussed earlier and for reference
purposes call it the Principle of Derivational Morphology.
Stern’s stages capture behavioral milestones that are described by other
diarists during this historical period. There are some attempts to relate
these behaviors to each other through the proposal of unifying principles,
although this is not done consistently. Some behaviors are related and
others simply co-occur.
Table 3.3 The basic measures used by Nice (1925)for Jive stages of
language acqukition
Stages
Measures
1 2 3 4 5
Average Length of Sentence (words) 1.00 3.5-3.5 5.0

Number of subjects
(a) English speaking 18 20 13 2 15
(b) nowEnglish speaking 17 20
Range of duration (in months)
(a) English speaking 4-9 4-1
(b) non-English speaking 4-12
Average duration (in months)
(a) English speaking 6
(b) non-English speaking 6.6
Range of ages at onset (in months)
(a) English speaking 13-27
(b) nowEnglish speaking 12-23
Average age at onset
(a) English speaking 17.5 mos 3 yrs 4 yrs
(b) nowEnglish speaking 18.3 mos
Vocabulary size at time of first utterances 14-180 475-1135
Proportion of incomplete sentences 1.OO below 0.6 0.2-0.6 below 0.2
Proportion of vocabulary
(a) nouns 0.65+ 0.5-0.6
(b) verbs 0.2-0.24
Proportion of complex sentences very low more adult-like
3.2.2 Nice (1925)

In 1925 Margaret Nice introduced the measure of the Average Length of
Sentence (ALS) as a means for dividing acquisition into stages. The ALS is
the mean number of words used by the child in its spontaneous language.
Nice wanted to build upon the traditional approach to stages, as exemplified
by Stern, by adding a quantified measure to them. She also added a name to
each stage, typically that for the new behavior which occurs (see figure 3.6).
Table 3.2 provides a summary of Nice’s stages.
Nice’s stages are typical of the kind of work done during the period of
large sample studies. Instead of intensively studying one or two children as
Stern did, she seeks data from several children. The work is by and large
descriptive, with virtually no attempt at hypothesizing general principles.
Further, there is the emphasis on the development of measures of super-

ficial linguistic behaviors. Table 3.3 gives a listing of the measures reported
for the five stages; they are not consistently given for each, but are randomly
presented.
There is nothing given us on Nice’s single word stage except figures on its
average duration. For both English and non-English children, it lasts
around six months. This stage is comparable to Stern’s first period, which
begins around the end of the first year of life.
The second stage, the early sentence stage, looks similar to Stern’s second
period. Nice states that it normally starts around 1;6. At the onset, the child
uses mostly single-word utterances even though multiword utterances are
appearing. For example, a 1042-utterancesample of Nice’s fourth child, H,
at 2;O revealed an ALS of 1.35, consisting of 683 single-word utterances,
352 two-word utterances, and only seven three-word utterances. The size of
the child’s vocabulary may vary greatly at the time of the onset of the first
multiword utterances. For the 19 children she sampled, the range of
vocabulary at the start of this stage went from as low as 14 to as high as 180.
This range is interesting because it is suggestive of a very loose relation
between the word spurt and the onset of multiword utterances. Recall in our
review of Stern that we suggested different principles behind each. Nice has
less data on the typical duration of this stage, but states the lengths for three
of her children as three, seven, and 13 months respectively. Curiously, she
has not told us how she measures its end! Lastly, Nice talks about the
vocabulary of the child in this stage. She uses the term baby ratio to refer to
the fact that 65 percent or more of the child’s words are nouns. This is
consistent with Stern’s proposal of a substance stage.
The short sentence stage looks like Stern’s third period, although Nice’s
data suggest that it may occur later than suggested by Stern’s dates. She
assigns to it an ALS of 3.5 to 4.5. Nice describes this stage as one in which
the child begins to develop inflections and grammatical words. There are as
yet virtually no complex or compound sentences. She provides three
measures for these developments. One is the proportion of incomplete
sentences. An incomplete sentence is one which is lacking either a subject or
verb. At the onset of the stage this will be as high as 0.60, and at the end, as
low as 0.20. Stated differently, the ratio of incomplete sentences drops from
3/5 at the onset to 1/5 at the end. It could be viewed as a stage in which the
child acquires well-formed simple sentences. The other measures are the
proportion of complex sentences and the proportion of compound sentences.
As with several of her measures, they are not defined. At this stage, they are
expected to be quite low. She cites the example of her child R who at 4;O
with an ALS of 3.9 had the proportion of 0.01 and 0.00 respectively for a
350-sentence sample.
A potentially important new development happens around this time in
regard to vocabulary. When the child has acquired around 400 to 600 words,
it shows what Nice calls the stable ratio. That is, there is now a stable ratio
between the proportions of the various word classes in the vocabulary.
Nouns comprise 50-60 percent of the vocabulary, and verbs around 20-24
percent. The event presumably marks the end of the three vocabulary
stages proposed by Stern for the second period (see Table 3.1).
The next stage of any consequence is the complete sentence stage, with an
average length of sentences of 6.0-8.0. Between it and the earlier short
sentence stage there is a transition stage that is barely worth mentioning. It is
apparently a stage with an ALS around 5.0. Nice, however, had not been
able to find enough children with that ALS to say anything of much
consequence. The complete sentence stage is one where most sentences are
well-formed, i.e. the proportion of incomplete sentences drops below 0.20.
Further, complex and compound sentences are more common, although
little normative data are provided. This stage looks similar to Stern’s fourth
period. In fact, if we ignore the transition stage, the two sets of stages line up
relatively neatly.
In examining Nice’s stages in comparison to Stern’s, we see a loss of
interest in explanation, and in certain respects, description. There are no
principles proposed to account for changes that occur. For most of Nice’s
observations, the stages are merely ‘continuous’ (Figure 3.1), with the
onset of the stable ratio revealing a ‘plateau’ stage (Figure 3.2). Nice also
has observed fewer behaviors than Stern, with no discussion of the
prelinguistic stage, word spurts, comprehension, or questions. Nice does
contribute, however, some useful notions of measurement. It is useful to
know, for example, the typical age when a behavior begins, as well as its
normal duration. Also, when we talk about the onset of subordination, a
measure of its distribution will give us some ideas about its productivity.
The introduction of measurement is an important contribution of this
historical period that has been improved upon in subsequent research.
3.2.3 Roger Brown (1973)

The most frequently cited general stages of grammatical development in
recent years have been those proposed by Brown (1973). Brown’s work has
already been mentioned in Chapter 2 (see Table 2.5) in relation to the
longitudinal development of Adam, Eve, and Sarah. As mentioned, he
collected longitudinal language samples on the three children from the
beginning of multiword utterances until their language was quite advanced.
Given the enormous amount of data that was eventually collected, he
wanted to select in a rational way some portion of the data for more detailed
Table 3.4 A summary of Brown’s (1973)five stages of grammatical

development
Stage Range of MLU Upper Midpoint Stage name and description

(morphemes) Bound
1.oo The period of single-word utterances

The use of single words without any
grammatical knowledge
I 1.00-1.99 Semantic roles and syntactic relations
early 1.00-1.49 The onset and acquisition of the basic semantic
late 1.50-1.99 5 1.75 relations used in language like Agent, Patient.
Word order is first syntactic device acquired.
I1 2.00-2.49 7 2.2s Modulation of meaning
The child begins to acquire inflections and
grammatical morphemes. Most are actually
acquired in subsequent stages.
I11 2.50-2.99 9 2.75 Modalities of the simple sentence
The active acquisition of the English auxiliary
as it appears in yes-no questions, wh-
questions, imperatives, and negative
questions.
IV 3.00-3.99 11 3.50 Embedding of one sentence within another
Complex sentences appear with object noun
phrase complements, embedded wh-
questions, and relative clauses.
V 4,OO and up 13 4.00 Coordination of simple sentences and
propositional relations
The active development of sentence, noun
phrase, and verb phrase coordination with the
use of conjunctions.
analysis. It was with this selective process in mind that he originally

proposed his stages.
Like Nice, Brown decided to use the average length of sentences as a
means to divide up the developmental continuum. To calculate this, he
chose to count the number of morphemes in sentences, instead of words,
because he felt that it would be a more sensitive measure. For example, the
utterances ‘boy play dog’ and ‘boys playing dogs’ have the same number of
words, but the latter has three more morphemes. Nice had just counted
words as had most of the previous researchers. Brown also developed a set
of criteria (p. 54) to use to determine what gets counted as a morpheme.
This way of measuring sentence length was referred to by Brown as the
child’s Mean Length of Utterance (or MLU).
Next, Brown arbitrarily divided up the MLU continuum into five stages
(see Table 3.4), by a range of MLUs. For example Stage I is when the
child’s MLU falls between 1.0 and 2.0. Each stage is labeled by a successive
Roman numeral. At this point, then, these are an example of continuous
stages (Figure 3.1). Brown was aware of this and did not at this point claim
any more than that. To study each of these arbitrary stages, Brown selected
an MLU that represented the midpoint of each stage. He believed that
samples from the children at those midpoints would be typical of the kinds
of linguistic behaviors shown for that range of MLUs. The midpoints are far
enough apart that the separate samples should also show change from one
another. Since many of his samples tended to fall into the latter part of
Stage I, he subsequently divided Stage I into two substages, early and late
Stage I. The midpoint for Stage I for his analysis was thus the midpoint for
late Stage I. Lastly, he selected all the samples for Adam, Eve, and Sarah
that were around the midpoints selected. The smallest of these samples
contained 713 utterances. He selected this number and subsequently
analyzed 713 utterances from each of the three children at each of the five
midpoints.
Upon analyzing his data, Brown found distinct linguistic behaviors for
each stage. He states (p. 59):
A stage is named . . , either for a process that is the major new
development occurring in that interval or for an exceptionally elabor-
ate development of a process at that stage. However, the whole
development of any one of the major constructional processes is not
contained within a given stage interval.
The stages with this addition become co-occurrence stages in the sense of
either Figure 3.6 or Figure 3.7. It also suggests that Brown may wish to
relate these behaviors by principles, moving then to principle stages and a
theory of acquisition. Brown stops short of doing this. Instead, he retreats
to a more descriptive position (p. 58): ‘The stages are not known to be true
stages in Piaget’s sense; that is they may not be qualitative changes of
organization forced on the investigator by the data themselves.’ The use of
‘may’ here, though, suggests that he is optimistic that they may turn out to
be so.
Stage Z is one which is marked by the onset of multiword utterances.
Brown emphasizes two properties of these early word combinations:
(i) they show the consistent use of word order, and (ii) the sentences reveal
+
a basic set of ten semantic relations, such as Agent Action or Action +
Object. Word order is proposed as the child’s first syntactic device. The
relations are semantic, however, rather than syntactic. While this stage
looks like Stern’s second period, it differs in that it assigns more grammati-
cal structure to the child. Stern described these early sentences as loosely
juxtaposed with little structure. In Part I1 we: shall return to this discrep-
ancy, and focus on the possible differences between German and English
acquisition.
In Stage ZZ, the child begins to acquire grammatical morphemes and

inflections, just as it does in Stern’s third period and Nice’s short sentence
stage. The stage is marked by the onset of these, with most acquired later.
To study grammatical morphemes, Brown developed a new measure of
acquisition, the percentage of obligatory occurrerzce. This measures the
number of times that a morpheme occurs in sentences where it would be
required by adult rules of grammar. For example, if a child says ‘two shoe’
and ‘two shoes’, the percentage of obligatory occurrence for the plural
morpheme would be 50 percent since it occurs in one of two obligatory
contexts. Brown reports on 14 grammatical morphemes that can be scored
by this measure. He finds that they occur in a regular order of acquisition,
and attempts to explain the order by considering the syntactic, semantic,
and phonological complexity of the individual morphemes. His study of
these first two stages is more detailed than either Stern’s or Nice’s, both in
terms of descriptive information and in the attempt to explain their
development,
The next stages are only briefly described in Brown (1973), and one must
look elsewhere for details on them. Stage ZZZ appears to be a substage of
Stern’s third period and Nice’s short sentence stage. It focusses on the
acquisition of the English auxiliary in different sentence types. In par-
ticular, it draws upon work done on the acquisition of questions and
negation by Klima & Bellugi-Klima (1966). Stages N and V , both dealing
with sentences with more than one clause, are substages of Stern’s fourth
period and Nice’s complete sentence stage. While Nice proposes that these
two constructions occur together, Brown finds that sentence embedding at
least begins before sentence coordination with conjunctions. While he gives
examples of these constructions in the grammar of English, he gives no data
at all on their pattern of acquisition.
In outlining these five stages, Brown purposely does not attempt to
construct a theory of acquisition. He does, though, discuss two basic
theoretical points that are part of such a theory. One, he looks at the form of
the child’s grammar in some detail. This is done by looking at both the
t categories used and the rules of combination. The first categories are
proposed to be semantic, linked by rules of word ordering. The semantic
grammar of Stage I will in Stage I1 begin to become a syntactic one. All
three of the first stages deal with simple sentences. A second theoretical
concern is the relation of the child’s performance, as evidenced in spon-
taneous speech, to its actual competence. I propose that we need to
consider the two to be relatively close together (and this will be discussed in
greater detail in Chapter 4). Brown, on the other hand, gives a rich
interpretation to the child’s sentences, supporting the view that the child’s
grammar is much better than indicated by its performance. Given this
Table 3.5 Five major periods of language acquisition and their relation to
proposals by Stern, Nice, and Brown
Period
Stern’s term Nice’s term Brown’s term
The period of prelinguistic development ( C n ; O )

Preliminary stage
The period of single-word utterances (1;0-1;6)
First period Single word stage Period of single-word utterances
The period of first word combinations (1;6-2;0)
Second period Early sentence stage Stage I: semantic roles and syntactic
relations
The period of simple sentences
Third period Short sentence stage Stage 11: modulation of meaning
Stage 111: modalities of the simple sentence
The period of complex sentences
Fourth period Complete sentence stage Stage IV: embedding of one sentence within
another
Stage V: coordination of simple sentences
and propositional relations
position, he discusses the performance factors which may restrict the child
from showing this ability. He also looks into the performance factors that
may help the child discover the structure of the adult language, for example,
guidelines such as paying attention to the ends of words. There is, however,
little attention to the principles that would be part of Universal Grammar
and used to construct a grammar, i.e. a set of linguistic rules.
Before leaving Brown, there is one last point that needs to be addressed,
i.e. his view of development before his five stages. He does discuss briefly
his ideas on the period of single-word utterances which precedes the five
stages. He sees this period as pre-grammatical, i.e. as not having any of the
structural properties of Stage 1. Thus, Brown gives this period’s utterances
a lean interpretation, as does Stern. Since Brown switches to a rich
interpretation for Stage I, however, he needs to propose a marked qualita-
tive change between the two. He needs to give single-word utterances a lean
interpretation because he bases much of his inference on early structure on
the child’s use of word order.
3.2.4 Summary
The three proposals of stages of acquisition given above provide some
preliminary facts about the child’s early linguistic behavior. They also
reflect the style of language acquisition research reported in Chapter 2. The
stages proposed define ‘stage’ in a variety of ways, including those outlined

in Figures 3.1-3.8. When general stages are given, they tend to describe
co-occurrence rather than principles, although some attempts have been
made at the latter.
It is not clear that any one approach is superior to the others; despite
variation, they overlap a great deal in their general features. Until such a
time as we have a principled set of stages, I propose that we emphasize as
much as possible the limited theoretical importance of our current pro-
posals about stages of acquisition. Rather, we should recognize that these
general stages do little more than isolate co-occurring linguistic behaviors
with a focus on the newest or most prominent. Table 3.5 gives five ‘periods’
I have derived from a comparison of the work of Stern, Nice, and Brown, to
be used as descriptive labels for the more elaborate discussion of language
acquisition to be given in Part 11. To help separate the multiple use of
‘stage’, we restrict its use henceforth to cases where some cause is proposed
for the stage, as in the case of the ‘principle stage’ of Figure 3.8.
3.3 Descriptive vs. explanatory stage

We can substitute various terms for ‘stage’ in Figures 3.1-7: ‘points’,
‘periods’, ‘plateaus’, etc. My concern is to narrow the use of ‘stage’to those
cases where we are referring to behaviors that are being explained in some
way. To explain a behavior is to offer a cause for its occurrence. In
Figures 3.1-8 only the last introduces some notion of explanation. In fact,
we can add explanation to any of the uses of ‘stage’ in Figures 3.1-7 if we
propose some cause for the designated stage. For example, we could talk
about the ‘word spurt stage’ of acquisition when the child’s vocabulary
suddenly increases around 1;6: as such, it is only an ‘acceleration stage’ as in
Figure 3.4. In our discussion of Stern’s stages, however, we have proposed
the Principle of the Linguistic Sign as the cause of this change, and this
proposed cause allows us to call this change a ‘stage’. Without this proposed
cause, we can only call it ‘the word spurt’, ‘the word spurt period’, ‘the time
of the word spurt’, etc., i.e. we must use some term other than ‘stage’.
Brainerd (1978) provides an important treatment of the use of ‘stage’ in
developmental research. He makes a distinction between two kinds of
stages. The first is a descriptive stage which has two characteristics: (i) it
consists of behaviors that undergo change, and (ii) there are antecedent
variables proposed that are responsible for the change. The first character-
istic is comparable to the continuity requirement that we discussed at the
onset of section 3.2 which underlies all eight stages in Figures 3.1-8. The
second characteristic is similar to the requirement we have just made to
restrict the use of ‘stage’ to cases where a principle or cause is proposed.
3.3 Descriptive vs. explanatory stage 55
Henceforth, we will use the term ‘descriptive stage’ in the way just defined.
The ‘word spurt stage’, consisting of the behavior of rapid vocabulary
acquisition and the Principle of the Linguistic Sign, is a descriptive stage in
this sense.
The second kind of stage that Brainerd discusses is the explanatory stage.
An explanatory stage is one which has the two characteristics above, plus
(iii) the antecedent variables or causes must be subject to independent
measurement. As Brainerd sees it, most of the proposed stages in the
psychological literature, such as those of Freud or Piaget, are descriptive
rather than explanatory. Without some independent measure, the so-called
explanations are nothing more than circular definitions. Brainerd gives as
an example Freud’s oral stage of development. The behavior being
observed in this case is thumb-sucking. Its proposed cause is the ‘oral stage’,
i.e. a stage when the child centers on oral gratification. The child sucks its
thumb because it is in the oral stage. Our evidence for the oral stage is that
the child sucks its thumb. As Brainerd says (p. 174): “‘he sucks his thumb
because he is in the oral stage” is simply a paraphrase of “he sucks his
thumb”’. We can say the same thing about our descriptive stage ‘the word
spurt stage’: the word spurt is caused by the Principle of the Linguistic Sign.
How do we know the latter exists? Because the child is undergoing a word
spurt. There is no independent measure of the principle.
The postulation of descriptive stages is a first step in eventually reaching
explanatory stages. It is important, however, to develop ways to provide
independent measurement of our proposed causes. Since Brainerd is a
behaviorist, he uses terms like ‘variables’ and ‘measurement’. For our study
of language acquisition, we would like to replace these words with principles
and evidence. That is, our hypothetical principles behind the child’s specific
linguistic changes should be supported by independent evidence. The kinds
of evidence could be behavioral, but they often will not be. For example,
suppose that we propose the symbolic function, which is a more genera1
ability to deal with symbols. Independent evidence for this principle would
be provided if all children show symbolic play with objects, e.g. pretending
that a block of wood is a car, at the same time as the word spurt.
Independent evidence for our proposals for language acquisition can come
from some other branches of linguistics such as historical change, language
disorders, or linguistic theory. Our principles may lead to claims about
complexity that predict certain linguistic behaviors in other domains. Or,
our principles may predict that certain kinds of linguistic analyses may be
better than others. The most important advances will be those made at the
higher levels when we talk about the structure of language, not just the
occurrence of some linguistic event.
Brainerd is very careful in his consideration of what might constitute
independent evidence. One potential form of independent evidence is the

kind of predictions a stage makes about invariant sequences of behavior.
Suppose, for example, that we propose a stage of acquisition that is like
Brown’s Stage 11. This hypothetical stage is one where the child builds upon
a previous development with a new one. Let us say the old behavior is the
development of basic semantic relations like Agent + Action as in ‘daddy
go’ in Stage I. In Stage 11, the child adds inflectional morphology such as in
‘daddy going’ where the progressive morpheme ‘-ing’ is added. Stage I1
consists of the old behavior plus a new one. The stage predicts that children
will always show an invariant sequence, i.e. that two-word lexical combin-
ations will always precede the use of lexical morphemes with inflectional
marking.
To the extent that such predictions are testable (Le. measurable), the
invariant sequence will constitute independent evidence. That is, we may
find out, in studying children, that some children acquire some inflectional
morphology in the period of single-word utterances. Such a case, in fact, has
been observed (see Munson & Ingram 1985). If so, then these cases are
counterevidence to the stage being proposed. Brainerd, in fact, is reluctant
to allow such data to be considered, on the grounds that we may make the
mistake of letting the predictions of the stage determine our analyses of the
relevant data. For example, we could say that the child who says ‘going’ as a
single-word utterance before two-word combinations like ‘daddy go’ does
not have an inflectional suffix because the theory does not allow it. In such a
situation, we are letting our hypotheses determine our analyses of the
observed behavior. Brainerd refers to such cases as a ‘measurement
sequence’. He states (p. 176): ‘A measurement sequence consists of the
immediately preceding item plus some new things. When behaviors are
related in this manner, the only way they can be acquired is in an invariant
sequence.’ I suggest we call these logical measurement sequences.
Measurement sequences may result from the form of our theory, or they
can result directly from the measures we select for the analysis of data. For
example, we can say that there will be three divisions to our analyses:
(i) children who never do some adult-like behavior, (ii) children who
sometimes do some adult-like behavior, and (iii) children who always do
the adult behavior. This measure yields three stages of acquisition that are
totally determined by the measure selected. Thus, they have no evidence
for their existence apart from the measure used. Let us call these quantita-
tive measurement sequences.
We have to be careful because of the problems mentioned above in
evaluating evidence for explanatory stages of acquisition. When we propose
analyses of data that are logical measurement sequences, we cannot take
these as evidence for an explanatory stage. This does not mean that such
Further reading 57
analyses are of no value, because they support the descriptive stage under
consideration. It may be that in some cases our construction of a logical
measurement sequence is not possible without being totally ad hoc. A
serious problem, however, is that logical measurement sequences can lead
researchers to use quantitative measurement sequences. That is, they select
measures that support their predictions, while ignoring those that do not;
we will refer to such mistakes as stage errors, i.e. stages that result
exclusively from a quantitative measurement sequence. A specific example
of such a case will be discussed later in Part I1 in relation to the acquisition of
before and after. (This example is also discussed in Ingram 1981b.)
While we have focussed on stages here, a similar problem arises when we
compare children with the purpose of distinguishing types of children. We
can suggest that children at some stage will fall into two types, so that the
child’s language acquisition system will allow it two options. One such
typology might be that some children focus on learning mostly substance
words in the period of single-word utterances, while others concentrate on
social words like ‘hi’, ‘bye’, etc. In fact, such a suggestion appears in Nelson
(1973). We will call such a position, in parallel with the terminology above,
a logical measurement typology. This can lead to the selection of a measure
that automatically measures children into the two groups. This I will call a
quantitative measurement typology, again in parallel with the above terms.
In our example, one such measure would be to find the average number of
social words used by children, then divide those using more than the
average as ‘expressive’ learners, and those that use fewer as ‘referential’
learners. The measure determines the typology because the children are
just showing a normal distribution of behavior. Another possibility is that
the children differ because they are at different ‘stages’ of acquisition. We
shall call decisions like these type errors, and point out actual cases in
Part 11.
Further reading
General literature
The literature on general stages is surprisingly restricted, with the efforts of
Stern, Nice, and Brown constituting three of the major attempts. The
proposals of all three are in a form that is readable for the beginning
student; specific readings are: Stern (1924: 143-70), Nice (1925), and
Brown (1973:632). In Part 11, we will consider one other approach for
early grammar proposed by Halliday (1975). Two other general attempts
are not in English. One is that of Kaczmarek (1953) in Polish; the only
discussion of this in English is a brief treatment in Bar-Adon & Leopold
(1971:133-4). Interestingly, there appears to be a rich literature on acqui-
sition in Polish (e.g. Zarebina 1965) which is virtually unknown to English

researchers. Other attempts at stages are those of Malrieu (1973) in French,
and Guillaume (1927). The latter article has been translated and appears in
English in Bloom (1978). Ingram (1976a: 10-15) compares grammatical
stages as discussed in this chapter with phonological and cognitive stages. A
more extensive discussion of cognitive stages in relation to linguistic stages
during prelinguistic development can be found in Anisfeld (1984). A critical
treatment of stages can also be found in Brainerd (1978), while an earlier
version of my discussion of the topic is in Ingram (1981b).
4 Explanation and language acquisition
4.1 Introduction
In Chapter 1it was stressed that language acquisition is an area of study that
crosses a variety of disciplines. Because of this, the goals of investigators
vary tremendously, and, thus, the extent and domain of explanation sought.
As already noted, much of the research in the field has led to descriptive
observations. This has been the consequence in some cases of an atheoreti-
cal orientation, or else of a belief that theorizing must follow from a strong
data base. Much of the work of those in education or in language disorders
is descriptive in nature, with the primary goal of establishing norms of
acquisition. This is a useful practical goal, yet it does not take us very far in
understanding why acquisition proceeds as it does. Much of the data from
the diaries of linguists also lacks attempts at extensive explanation. As
N. V. Smith (1982) has said somewhat facetiously (p. 471): ‘by providing a
plethora of factual observations, it [child language: DI] offers a rich field for
the taxonomist and anecdotalists’.
So far I have addressed the issue of explanation in two contexts. In
Chapter 2, I presented the theoretical orientation of the three major
historical periods in the study of child language. It was pointed out that each
of these three periods had its own theoretical orientation, with the nativism
of the last period being divided into constructionism and maturationism.
These two were contrasted with behaviorism, which assigned very limited
innate ability to the child. In the last chapter, I dealt with the question of
stages of acquisition and finished with the suggestion that our ultimate goal
in determining stages is to limit ourselves to explanatory stages, i.e. changes
in the child’s ability that can be accounted for by principles that themselves
are testable.
In the present chapter, I treat explanation in more depth, as a framework
for the rest of the book. First, returning again to the distinction between
Child Language and Language Acquisition, I discuss the way each differs in
its approach to explanation in language learning. I will argue that each
needs to adjust its approach in the direction of the other to improve upon
59
60 EXPLANATION AND LANGUAGE ACQUISITION
some current difficulties in orientation. Once adjusted, the two approaches

are less distinct, and become for all intents one discipline of Child Language
Acquisition. This unified field has the following properties: (i) it requires
the study of children’s language; (ii) it examines children’s data with
well-defined theories of grammar; and (iii) it concentrates on methods that
establish when a child’s linguistic behavior is rule-based.
The goals of Child Language Acquisition are therefore two-fold: (i) to
provide a testing ground for current theories of grammar; (ii) to develop a
theory of acquisition. To pursue both these goals requires a careful look at
the information obtainable from data from children. The feasibility of these
two goals is discussed through the examination of the relation between the
child’s linguistic performance and its linguistic competence. To accomplish
these goals we need to be able to make statements about a child’s linguistic
competence from its performance. We also need to relate the child’s
comprehension to its production within a theory of performance. That is,
what are the factors which limit the structure of language? This chapter
presents a set of assumptions that enable us to pursue these goals. Without
these assumptions, there would not be any potential for explanation in the
first place, and the field would need to restrict its goals tremendously.
4.2 Child Language vs. Language Acquisition

At present, these two approaches to the study of children’s language are
going in very different directions. On the one hand, we have those who
pursue Language Acquisition who, for referential purposes, we will refer to
as linguists. (It does seem that, with many exceptions, those who do
Language Acquisition tend to be trained in linguistics, while those who do
Child Language tend to be trained in psychology. As we will suggest,
however, their goals should be the same, basically the development of a
theory of grammar and a theory of learning.) They are concerned almost
exclusively with developing a theory of grammar, and tend to minimize the
role of data from children in this enterprise. Further, they tend to view
Child Language as primarily descriptively oriented with little interest in
theory (see comments by Wasow in Chapter 2, on page 27). On the other
side, there are those who pursue Child Language, who we will refer to as
psycholinguists. They focus on children’s data, propose various stages of
acquisition, tend toward inductive statements about the nature of language,
and focus on performance factors in acquisition (e.g. constraints on
memory, processing limitations, etc.). They tend also to be sceptical of
linguistic theorizing about children’s language as being too removed from
the data. Here I will point out some problems with these approaches and
suggest changes for each.
4.2 Child Language vs. Language Acquisition 61
Linguists are rather dubious about the role of language acquisition data in
the process of developing a theory of grammar. One reason for this is that
language acquisition data are often not used in ways that are of interest to
the linguist. It is not clear, for example, that many linguists are aware of the
rich database that exists in the field. When theoretical issues are raised
where acquisition data may be relevant, it is usually assumed that no data
exist when in fact they do.
Most psycholinguists, however, seem to have misunderstood the goal of
linguistics, and often are unaware of current linguistic issues. The linguist is
concerned with testing hypotheses about language, so that a theory of
grammar, be it Chomsky’s work (1965, 1981), or that of others, such as
Fillmore’s case grammar (Fillmore 1968), is just an interim step along the
way. Theories of grammar are constantly changing as we come to understand
better the structure of language. In this context, child language data are
only interesting to the extent that they may be used to prove or disprove a
particular theory. For example, suppose we have two distinct theories that
explain a structure, let us say relative clauses, in very different ways. They
may each lead to very different predictions about how a child acquires
relative clauses. Acquisition data on relative clauses, then, will only be of
interest to the linguist to the extent that they provide a means to help choose
between the two theories. The linguist will only be interested in acquisition
data when he is convinced that this contribution is possible.
Psycholinguists have not seen their enterprise as a contribution to
linguistics in this sense. Rather, they have seen linguistic theory as a field
that provides Child Language with a method of description that can be
applied to children’s language. That is, linguistics is expected to provide the
‘correct’ theory for the psycholinguist to use to understand the child. This
perspective, however, can only lead to frustration. One of the best
statements of this is given by Roger Brown (1970a: vii, viii):
I used to think that psychology took from linguistics because linguis-
tics was the more advanced field, ‘advanced’ perhaps meaning here
nearer the truth about its subject matter . . . I do not think this
anymore. Linguistics has, after all, solved very few practical problems
. . . The fact that linguistic theory changes, and does so at a rapid clip,
poses real difficulties for the psychologist who wants to use linguistic
theory in his own work.
This misunderstanding of the goal of linguistics led to the virtual cessation of
grammar writing in the 1970s. Most of the work, in the view of the linguist,
was consequently descriptive and taxonomic. It did not contribute to our
understanding of the nature of language.
It is important for the psycholinguist to recognize the goals of the linguist,
and to understand the potential contribution of acquisition data to them.

There are two ways the psycholinguist may proceed. First, he can become
conversant with a linguistic theory of one kind or another, and examine
children’s language for evidence that may support or suggest revisions in
that theory. Lightfoot (1979:18) makes this point when he says: ‘Data from
acquisition, say, will simply constitute one more argument for or against the
theory, one more area where the theory makes predictions, and will help to
reduce the indeterminacy of the linguistic description.’
An alternative way to proceed is for the psycholinguist to construct a
theory of acquisition, again within some general theoretical guidelines.
This approach would start with a careful analysis of stages of acqui-
sition. It would construct a theory that consists of principles which are
proposed to explain the child’s language at each stage, and also the
changes from stage to stage. For example, it is well known that children
at some point overgeneralize the plural morpheme and say ‘foots’
instead of ‘feet’. A theory of acquisition (or learning) will be needed to
explain the above form. Much of the theorizing, in fact, has been along
these lines.
In summary, the linguist has yet to be convinced of the role of acquisition
data in theory construction. When acquisition is discussed, as in the area of
learnability (see 2.3), actual data from children are not seen as relevant.
The psycholinguist has not always understood this role, and has looked
instead for a final theory that can be used as a tool to understand the child.
A reconciliation will require more awareness of the goals of linguistic
theory, and the issues in theory testing. One way is to develop evidence for
the role of acquisition data in theory construction; the other is to develop a
theory of acquisition which will yield information on how the child
formulates rules.
While Child Language focusses on data, it is unfair to suggest that it is
atheoretical. There has been a good deal of theorizing, although it tends to
vary from domain to domain, and often does not relate to findings in other
areas. It is a discipline, however, which is rather impatient with theorizing
that ignores and inaccurately cites data. For example, Baker (1979) talks
about the possibility of a Dative Movement rule in English to relate
sentences (4.la) and (4.lb). (See Section 2.3.)
(4.1) a. I gave a book to Mary
b. I gave Mary a book
c. I reported the crime to the police
d. * I reported the police the crime
He points out that if children acquire such a rule, it should result in
sentences like (4.ld). Baker goes on to say (p. 570): ‘What little evidence I
have seen on the performance of children suggests that they do not make
errors of the expected types in connection with these putative rules.’ He
goes on to assume that children do not make such errors and subsequently
rejects the Dative Movement rule as unlearnable.
This is a peculiar research strategy indeed! The psycholinguist doing
Child Language has developed techniques that allow him or her to answer
questions like this one. The approach would be to resolvefirst what children
actually do, then go on to discuss the rule under consideration: Ingram
(1985c), for example, has found that children do make mistakes like
sentence (4.ld), suggesting that there is a Dative Movement rule. (See also
Mazurkewich 1982, 1984, Mazurkewich & White 1984.) Evidence like this
alters one’s analysis tremendously. Much of the research on learnability in
Language Acquisition is of this kind, where claims are made about what
children do, with little, if any, confirming evidence. It is from this perspec-
tive, then, that psycholinguists criticize linguists for a lack of awareness of
the data.
Various techniques that can be used to see what children do or do not do
have, as we will see in Part 11, been effectively used to examine how children
acquire particular structures of English. In particular, elicited production
and comprehension tasks, metalinguistic judgements, and elicited imitation
have contributed to our findings on the young child’s knowledge of grammar.
The linguist, then, needs to become more aware of the data base of Child
Language and its methodological developments over the years.
If these two branches of language acquisition move in the directions
suggested, it becomes clear that we are discussing one field with two
different aspects emphasized. One works from the theory down, the other
from the data up. Both, however, share two major features. First, each
works from a theoretical basis, with the goal to explain the nature of
language, both in its adult and developing state. Second, both recognize the
need to examine children’s data to ascertain what children actually learn.
The latter will only be possible through the use of well-developed tech-
niques of data collection or analysis. With these shared assumptions, we
have a unified endeavor of Child Language Acquisition with two sub-
branches that are capable of dialogue.
4.3 A theory of acquisition

We have been using the term ‘theory of acquisition’, as distinct from ‘theory
of language’ (or ‘theory of grammar’). Chomsky, in various places (e.g.
Chomsky 1965, 1981), has described the theory of language as one which
consists of a set of principles (or Universal Grammar, UG) that underlie
language:
64 EXPLANATION A N D LANGUAGE ACQUISITION
What we expect to find, then, is a highly structured theory of UG

based on a number of fundamental principles that sharply restrict the
class of attainable grammars and narrowly constrain their form, but
with parameters that have to be fixed by experience.
(Chomsky 1981: 4)
A theory of language is affected by language learning in that the grammar

must be learnable, Le. be obtainable by listening to the surrounding
language. The child hears a certain pattern and ‘fixes a parameter’, i.e. it
decides that the language operates one way rather than another.
This theory of language, however, does not deal specifically with the
possibility that the final grammar may result only after a series of stages of
acquisition. The above view sees acquisition as essentially ‘instantaneous’
(Chomsky 1975) in that the child’s grammar is adult-like once the parameter
is set. We use theory of acquisition as a set of principles, distinct from those
of UG, that account for the stages the child goes through to reach the adult
grammar. If there are only two stages, pre- and post-parameter setting, then
the change is virtually uninteresting. In such a case, the theory of learning is
rather simple - fix the parameter. If there is a series of stages (and data at
this point suggest this to be true), then the theory of acquisition will explain
how each stage is structured and how it develops from the previous stage
and into the next one. Historically, Language Acquisition has concentrated
on the theory of language, while Child Language has concentrated on the
theory of acquisition.
The theory of acquisition will have two distinct components. One will be
the set of principles that lead to the construction of the grammar, i.e. those
that concern the child’s grammar or linguistic competence. These principles
will deal with how the child constructs a rule of grammar and changes it over
time. The focus is on the nature of the child’s rule system; it is concerned
with competence factors. The second component looks at the psychological
processes the child uses in learning the language. These are what we shall
call performance factors. Performance factors enter into the child’s com-
prehension and production of language. In comprehension, performance
factors deal with how the child establishes meaning in the language input, as
well as with the cognitive restrictions that temporarily retard development.
In production, these factors describe the reasons why the child’s spoken
language may not reflect its linguistic competence. They also describe
mechanisms the child may use to achieve the expression of their com-
prehension. Much of the Child Language literature is on the specification of
performance factors in acquisition.
Let us look at some examples of the kinds of principles that will be part of
the theory of acquisition, beginning with competence factors. I will present
two possible principles that have been proposed, one dealing with the
acquisition of morphology, the other with the acquisition of syntax.
Earlier, I mentioned the long-observed fact that children overgeneralize
morphological inflectional suffixes in English to irregular forms, saying
things like ‘foots’ and ‘breaked’. Stern, as pointed out, referred to these as
‘child etymologies’. We also mentioned that these overgeneralizations are
not accounted for by a theory of language, which only accounts for the form
of the rule. The occurrence of these overgeneralizations and their sub-
sequent loss must be explained by a theory of acquisition.
Dresher (1981) has proposed a principle to account for these overgener-
alizations as part of a detailed analysis of the learnability of Old English
morphology. He states (p. 192): ‘a learner adopts the most highly valued
rules (i.e. rules requiring the fewest features) consistent with, and some-
times overriding the available data’. A rule that states that the plural of
‘foot’is ‘feet’is more complex, i.e. has more features, than the rule that says
that the plural is made by affixing ‘-s’,since ‘foot’is exceptional and requires
more marking than nouns like ‘cat’, ‘dog’, etc. I will call this the generuli-
zation principle and leave its exact formulation open. This principle
accounts for cases like ‘foots’which occur even though the child never hears
‘foots’. As discussed by Dresher, this principle is part of the child’s program
to acquire rules of grammar. It restricts the form of the child’s rules at some
intermediate point in development.
There are some features of this principle that deserve consideration.
First, it makes predictions about the child’s grammar for comprehension as
well as production. If the child’s rule is general, then the young child should
at the time of producing ‘foots’ also have ‘foots’ in comprehension as an
acceptable utterance. Should the young child reject ‘foots’ as ungrammati-
cal, yet produce ‘foots’, we would need to adjust our theory of acquisition to
account for this, possibly accounting for the latter case by a separate
performance factor. (The testing of these predictions, of course, will be no
easy matter!)
Another point is that the generalization principle is not sufficient to
account for the acquisition of morphological endings. As we shall see in
Part 11, children do not produce overgeneralizations like ‘foots’ at the onset
of plural acquisition, but instead, either omit the plural or else use it
correctly. To account for this, we need another principle which elsewhere
(Ingram 1985a) I have called the lexical principle: learn individual
paradigmatic alternations as separate lexical items. This principle claims
that the child first acquires paradigmatic variants like ‘cat, cats’, ‘dog, dogs’
as separate words, and only later realizes that there is a separable plural
morpheme ‘-s’. The lexical principle predicts that the child will initially get
‘foot, feet’ correctly, and only later change the latter to ‘foots’. (This is in
fact, as we will see, true.) If so, our theory of acquisition will contain the
lexical principle and the generalization principle, and it will need to specify
the conditions under which the one yields to the other.
These two principles are still not sufficient to account for the acquisition
of the adult rule of plural formation. The child could conclude that there
are two forms of the plural for ‘foot’, these being ‘foots’ and ‘feet’. Wexler
& Culicover (1980), who have led the research into principles of language
acquisition, propose the uniqueness principle to make this last step. The
principle, in its most general form, will tell the child to select only one of
the above forms, the one that is used in the child’s linguistic environment.
As with the other principles, we will leave its exact formulation open.
These three principles, the lexical principle, the generalization principle,
and the uniqueness principle, are examples of the kinds of principles that
will be part of the theory of acquisition that attempts to explain morpholo-
gical acquisition.
An example of a principle of acquisition in the area of syntax is the
principle discussed in Chomsky (1975:32). There, he presents the sen-
tences in (4.2), where sentence (4.2b) is the correct question form of
(4.2a) :
(4.2) a. The man who is tall is in the room
b. Is the man who is tall in the room?
c. * Is the man who tall is in the room?
(4.2b) is the result of the rule in English of Subject-Auxiliary Inversion
which moves the auxiliary to the front of questions. Chomsky refers to this
rule as a structure-dependent rule, meaning that it requires a structural
analysis of the sentence to apply it. It moves the auxiliary of the main
clause. The rule would be structure-independent if it ignored structure, and
simply said ‘move the first auxiliary’. If the rule were such, it would
produce the ungrammatical (4.2c), where the first auxiliary is the ‘is’ in the
relative clause ‘who is tall’.
Chomsky goes on to claim (with no evidence cited) that children never
form questions like (4.2~).The reason is that the child has a principle
which we shall call the principle of structure-dependent rules which blocks
the child from ever having a rule that is structure-independent.
The only reasonable conclusion is that UG contains the principle that
all such rules must be structure-dependent. That is, the child’s mind
. . . contains the instruction: Construct a structure-dependent rule
ignoring all structure-independent rules. The principle of structure-
dependence is not learned, but forms part of the conditions for
language learning. (Chomsky 1975:32-3)
We see this as a principle of language acquisition. The corresponding
principle of UG is that all rules in language are structure-dependent. Such a
principle restricts a possible rule of language, but it does not restrict the
child from using structure-independent rules at some early point in acqui-
sition. To do this, we need the separate principle of structure-dependent
rules.
As with our earlier principles, this one will need to be revised in the face
of acquisition data. Like Baker (1979), Chomsky has no data to support his
claim about what children do, apart from experience in hearing children
speak. Elsewhere (Ingram 1985a) we have pointed out examples where
children do seem to form a structure-independent rule before a structure-
dependent one. One example concerns the acquisition of the English rule of
Subject-Verb Agreement. This rule operates in English to make a verb
agree in number with its subject, as shown in (4.3):
(4.3) a. The man is in the room
b. The men are in the room
We could formulate two forms to the rule, one structure-independent, the
other structure-dependent. These two are stated in (4.4):
(4.4) a. structure-independent rule: a verb agrees in number with a pre-
ceding noun
b. structure-dependent rule: a verb agrees with its subject
Sentences like (4.5) show why the rule for adults is structure-dependent, in
that (4.a) and (4.b) are ungrammatical because the verb is agreeing with the
preceding noun instead of with its subject:
(4.5) a. * Which balls are the boy throwing?
b. * Which ball is the boys throwing?
c. Which balls is the boy throwing?
d. Which balls are the boys throwing?
I am currently studying the acquisition of Subject-Verb Agreement in
English, using sentences like those in ( 4 3 , asking young children to judge
sentences as either ‘good’ or ‘silly’, using the technique of metalinguistic
awareness. Interestingly, early results indicate that younger children will
accept sentences like (4.5~)and (4.5d) as bad. That is, the younger children
are using the structure-independent rule of (4.4a), rather than the adult
structure-dependent rule. In Ingram (1985a), I suggest that we need
another principle to precede the principle of structure-dependent rules.
This principle, which I shall call here the principle of linear sequence, states
that the child should first establish rules based on structure-independent
linear order, such as ‘precede’ or ‘follow’. This principle yields to the one
proposed by Chomsky when the child eventually correctly processes

sentences like (4.5) which trigger the later principle.
Our theory of acquisition will aim at finding a set of principles that allow
the child to acquire rules of language in stages before acquiring the
complete adult grammar. The above examples have led to five possible
principles so far: the lexical principle, the generalization principle, the
uniqueness principle, the principle of linear sequence, the principle of
structure-dependent rules. At this point, they are only being described,
though I will attempt to specify them and others more accurately in later
chapters. They are seen as competence factors in language acquisition.
While some research has sought competence factors like the above, much
of Child Language has concentrated on performance factors. These are
factors that affect the processing of language in either comprehension or
production. They may be facilitative, i.e. aiding in accessing the relevant
data necessary to activate competence factors, or prohibitive, Le. tempo-
rarily masking the child’s competence. Let us look at some examples of
each.
Slobin (1973) has proposed several performance factors in the acquisition
of language under the label of ‘operating principles’. In comprehension,
one such principle, which he calls Principle A, is ‘Pay attention to the ends
of words’ (p. 412). We will refer to such principles as performance prin-
ciples, to distinguish them from competence factors. Slobin uses Prin-
ciple A to account for the fact that the acquisition of suffixes is easier than
the acquisition of prefixed morphemes. A principle like this may be peculiar
to language, or it may also be a general cognitive operation that describes a
recency effect in auditory processing. Another such principle would be one
which instructs the child to pay attention to stressed words. Presumably,
recency or prominence will enhance the child’s ability to process the
incoming signal for analysis.
It is possible to see such principles as prohibitive. For example, a child
may know several words, but may have trouble in retrieving them.
Problems with word retrieval may give the impression that the child has a
smaller vocabulary than it actually has. Some prohibitive performance
factors may be beneficial, in that they may allow the child to avoid some
features of the input so that they may concentrate on others. For example,
the principle ‘Ignore unstressed syllables’ could be a reverse of the
prominence principle which would allow the child to concentrate on
stressed material. The unstressed material could be returned to later when
better understanding is achieved.
Performance factors will also be crucial in our understanding of the
child’s produced utterances. At the time the child is producing two-word
utterances, it may in fact understand longer utterances, possibly with three
or four words in them. If so, we would need to consider the operation of a
restriction on sentence length. Memory factors and planning factors such as
false starts are also possible performance factors. These factors will mask
the child’s underlying knowledge of grammar. In some cases, the factors
may be facilitative. For example, young children often repeat themselves
over and over again, as if practicing language; such self-repetitions are not
particularly adult-like, but they seem an important part of the child’s
acquisition of language. Another facilitative performance factor is ‘chain-
ing’, where the child starts a new sentence with the last word of the previous
sentence, e.g. ‘see cat. cat big.’ Francis (1969) has discussed several factors
of this kind that facilitate acquisition.
While it is important to isolate performance factors like those above, it is
also necessary to remember that they are only part of the language
acquisition process. They provide us with insight into how the child obtains
information and expresses itself, but they do not tell us about the structure
of language. Even though they reveal a lot about the processing of
language, they do not reveal information about the child’s rule system. A
description of only a child’s performance, then, is not in itself a theory of
acquisition.
4.4 Theoretical assumptions about language acquisition

Earlier, we talked about three positions on the nature of language acqui-
sition: behaviorism, constructionism, and maturationism. These three
positions differ in the extent to which they assign innate structure. So far,
we have implicitly assumed that the child has a good deal of innate structure
beyond that necessary to form the associations and reinforcements
described in behaviorism. This is apparent in the discussion of inherent
principles that occurred in the previous section. Also, by concentrating on
stages of acquisition, as in Chapter 3 , we have implied a constructionist
perspective. In this section, we shall discuss this position more directly and
express some assumptions about the method of inquiry into language
acquisition that will be pursued through the remainder of the book.
In section 4.2 we characterized Child Language as an area that tends to
look at the child’s language at some point in time, with relatively little
interest in the final adult state, that is, the child data are not always seen as
evidence for or against a theory of grammar. Further, the child’s grammar
at some point in time may be quite different from the adult’s. Language
Acquisition was described as a field which looks at theories of grammar and
questions of learnability, rather than data from children. It leads to a tacit
assumption that the child’s language is not particularly different from the
adult’s, except where the difference is due to the effects of performance
factors. We subsequently proposed a reconciliation into Child Language

Acquisition, a field which works from acquisition data and a theoretical
perspective. On the one side, it is a part of the effort to construct a theory of
grammar; on the other, it is concerned with developing a theory of
acquisition. We need to look more carefully now at how child language data
can be used in both these endeavors. We need to see how acquisition data
can be used to: (i) provide evidence for or against adult theories of
grammar, and (ii) provide insight into the child’s linguistic competence.
4.4.1 Acquisition data and linguistic theory

To accomplish the first task, we need to examine carefully the relation
between the child’s grammar and the adult’s grammar. Specifically, we want
to develop a theory which defines the extent to which the child may change
or restructure its language system. If the child can change its grammar quite
drastically, then acquisition data will be of little value to linguistic theory. If,
however, we can limit restructuring, we will be providing a role of
acquisition in linguistic theory. These points will be explained below.
The three theories described in Chapter 2 differ greatly in their claims
about the child’s restructuring of its language. Behaviorism sees learning as
incremental, in that habits are established gradually over time. Changes
that occur are gradual so that restructuring is highly restricted and subject to
the environmental conditions. Once a habit is established, it is fixed and can
only be lost bit by bit. While behaviorism restricts change greatly, it
provides no linguistic principles that are part of the changes that take place.
More interesting are the views on restructuring from the two nativist
positions. Maturationism, oddly enough, predicts two extreme possibili-
ties. One of these states that the principles of UG appear early, possibly as
early as the period of one-word utterances, so that restructuring is virtually
non-existent. The child’s grammar from the onset is seen as essentially
adult-like, with only performance factors accounting for the differences
between it and adult language. This was referred to years ago by Watt
(1970) as the strong inclusion hypothesis.
The other possible maturationist position, which I will call the restructur-
ing hypothesis, allows restructuring under two conditions. In one case,
restructuring occurs late in acquisition because the sentences that the child
needs to hear to trigger the appropriate principle are not heard. There are
several structures of English that are rare in spoken language and virtually
non-existent in the speech to children. A few examples are given in (4.6):
(4.6) a. Backwards pronominalization: When hei arrived, Mickeyi was

hungry
b. Subject complement clauses: That he is late is possible
c. Passives with ‘by’-phrases: The cat was chased by the dog
Restructuring would occur later in development when the child experienced

the appropriate sentences. The other possibility is that some principles
simply mature late, regardless of the input - for example, (4.6~)above.
Carol Chomsky (1969) studied the acquisition of pronominalization and
found that it consistently was acquired around 5;6. She could not find a
non-linguistic explanation for the uniformity of the date of acquisition and
concluded that the necessary principles mature around 5;6.
Whether one takes one of these two positions, or a combination of both,
their acceptance renders acquisition data irrelevant to linguistic theory. The
reason for this is different in each case. If we take the strong inclusion
hypothesis, then all changes are due to changes in performance, not
competence. The child’s grammar is adult-like from the onset; thus, no
interesting insights into the adult grammar will be available from the child.
If we take the latter position, which allows extensive restructuring later in
development, it also leads us to the rejection of child language data, since
the change in the child’s grammar is not due to the nature of the child’s
earlier grammar, but rather to other factors, i.e. new input or maturation.
Acquisition data will only be relevant if we can relate later changes to the
child’s system.
The latter point can be exemplified by looking at the acquisition of
English adjectives within noun phrases. One common analysis of English
prenominal adjectives, as in (4.7a), is to say that they are moved into
prenominal position from a relative clause. Let us call this rule Adjective
Preposing. If the rule of Adjective Preposing does not apply, then we get a
relative clause as in (4.7b):
(4.7) a. the red dog
b. the dog which is red
Data from child language indicate that children acquire prenominal adjec-
tive structures like (4.7a) before they acquire relative clause structures as in
(4.7b). The analysis above, however, claims that the preposed adjective
structure is more complex, Le. it involves all the structure necessary for
(4.7b) plus the rule of Adjective Preposing. Do the child language data,
then, constitute evidence against the above analysis of English adjectives?
Both the maturationist positions can argue against the child language
data. The strong inclusion hypothesis would say that the child who produces
‘the red dog’ already has the relative clause structure, and the rule of
Adjective Preposing, but does not produce relative clauses due to perform-
ance factors. One such factor could be a restriction on sentence length. The
other position can argue that the child restructures later, so that the first
adjective+noun structures are different from the later ones. The child will
only reanalyze ‘the red dog’ later on. One reason why Adjective Preposing
does not exist might be that the child has not yet heard enough relative
clauses to formulate the correct version of the rule. Another possibility is
that the ability to embed sentences has not yet matured, and needs time to
do so. Major restructuring will occur, whether as a result of hearing the
appropriate input or as a result of a maturational event. In all cases, the
child language data are explained, and hence are irrelevant to the adult
analysis.
Under the maturationist positions, the goals of Language Acquisition are
limited. Under the strong inclusion hypothesis, it becomes a field con-
cerned with performance factors. Under the restructuring hypothesis, it
focusses on one of two things. For one, it can document when children hear
specific sentence structures in acquisition. Part of this task also involves
determining the amount of time necessary with a particular structure before
a principle necessary to acquire it is triggered. That is, what is the necessary
trigger experience? Some research on this question can be found in Nelson
(1981). The other area is to determine the relative times when principles
mature. The strong inclusion hypothesis, of course, assumes early release of
the principles, as its program of research will be concerned to prove the
existence of adult principles as early as possible. The restructuring hypo-
thesis, however, will allow for later appearance, so it will be concerned with
relating the appearance of each of the principles in time.
It may turn out, ultimately, that the maturationist view is correct, and
that we need to limit the goals of language acquisition in the ways just
described. It is possible, however, to restrict the extent of restructuring in
ways that will allow acquisition data to play a role in linguistic theory. This is
possible if we take a constructionist view of acquisition. In its simplest form,
constructionism proposes that a rule at any particular stage always adds to
or builds upon the earlier forms of a rule. The final form of the rule in the
adult grammar will result from the final addition to the rule. This view
restricts change, in that the rule at stage n+l can never be qualitatively
completely different from stage n.
This position can be exemplified in our discussion of Adjective Prepos-
ing. Acquisition data indicate that the first structures are of the form
+
adjective noun as in ‘red dog’. If there is evidence that this is productive (a
point to be returned to), we can propose a phrase structure rule to produce
it. Once proposed, however, it will always be a part of the child’s grammar.
The position does not allow the restructuring necessary to have Adjective
Preposing once the child is exposed to relative clauses. It does allow,
however, three other analyses of the relation between the two structures in
(4.6): (i) the two structures are generated separately, Le. there are no
principles of UG to relate them in the syntax; (ii) the adjective+noun
structures are of two kinds, those generated by a phrase structure rule only,
and those that result from Adjective Preposing; or (iii) the underlying
+ +
structure for all proposed adjectives is adjective noun relative clause,
and the Adjective Preposing rule moves an adjective from the relative
clause into an empty adjective category. The rule would be structure-
preserving in the sense of Emonds (1976). The adjective data in this case do
not dictate a simple solution, but they limit the range of possibilities.
There has been some extensive debate as to whether the correct approach
is constructionism or maturationism (e.g. Piattelli-Palmarini 1980),
although the debate has never been formulated quite in the form just given.
Here I would like to give two arguments for the selection of constructionism
over maturationism as the correct underlying theory of acquisition. The first
is an argument based on the potential contribution of each. If we select
maturationism, our view of acquisition research is limited to the goals
mentioned above. That is, it restricts the field to a highly descriptive
discipline. If we select constructionism, however, it expands the role of
acquisition to include a part of a theory of linguistic competence. All things
being equal, i.e. the internal arguments for one do not settle the issue ahead
of time, then the position that expands the scope of a discipline is preferable
to one that limits it. If constructionism is proven false, i.e. that it does not
lead to generalizations about the nature of linguistic competence, then we
will be no worse off than if we had chosen maturationism. If true, we will
add new data to linguistic theory.
The second argument concerns the testability of the two positions. As has
been argued before, the maturationist view is not testable, in that acquisition
data cannot be used either toprove or disprove it. Constructionism, however,
predicts that the principles it develops should lead to generalizations about
language that will be consistent with independent analyses of historical
change and linguistictheory. The inductive generalizations should play a role
much the same as that discussed for historical change in Lightfoot (1979).
The selection of a constructionist view leads us to our first assumption
about language acquisition that will underlie the discussion throughout the
book. It is summarized as follows:
The Constructionist Assumption: The form of the child’s grammar at
any point of change which we shall call stage n will consist of
+
everything at stage n plus the new feature(s) of stage n I .
This assumption adds to the other features of ‘stage’discussed at the end of
Chapter 3, namely that there has to be a change, and that a principle be
proposed to cause the change.
4.4.2 The determination of the child’s linguistic competence

So far, we have selected a view of acquisition where acquisition data may
play a role in linguistic theory. This will only be possible, however, to the
extent that we can establish something about the child’s linguistic com-
petence or rule system. We need to be able to make statements about the
psychological reality of our analyses of the child’s language.
The importance of this fact can be exemplified by discussing the emer-
gence of a phrase structure grammar. In (4.8) I give five hypothetical
productions of a young child over two years. Let us assume that these were
all produced in exactly the same context as requests for a cookie:
(4.8) a. time 1: (grunt)
b. time 2: cookie
c. time 3: want cookie
d. time 4: want eat cookie
e. time 5: I want to eat a cookie
For purposes of discussion, let us assume, like Chomsky (1981), that the
underlying phrase structure of the equivalent sentence in adult English is as
given in (4.9):
(4.9) I want [SCOMP [sPRO [Vpto eat a cookie]]]
Here COMP is the category filled by ‘that’ in the sentence ‘I know that Bill
ate a cookie’. In (4.9) the COMP is empty, i.e. there is no lexical item for it.
The PRO is also an empty category that will be interpreted in the grammar
to be the same referent as the subject of the main clause, ‘1’.An ungramma-
tical English sentence that is close to (4.9) would be ‘I want that I eat a
cookie’. The question to be posed is this: at which time, if any, in (4.8) can
we say that the child has the phrase structure of (4.9)?
If we return temporarily to the maturationist view, we can see that the
issue is untestable. We could say that it has matured at any one of the times,
and that performance factors restrict evidence in the output for it. One may
be intuitively less comfortable with the claim at time 1 than at time 4, but
the claim is the same for both times. One’s discomfort (if it exists at all!)
stems from a tendency to assume such structures around the time the child
appears to use relevant structures. If we wish to pursue analyses of child
language, however, we will need to make this assumption more precise.
Let us call the structure bracketed by S in (4.9) a tenseless clause, and say
it consists of a COMP category, a PRO, and a VP. Our dilemma, stated
differently, is to decide when we can conclude from the child’s language use
that a tenseless clause has been acquired. As an initial attempt, we can say
that the child does not have PRO until sentences are produced with overt
embedded subjects, as in (4.10a). Further, he or she does not have COMP
until sentences like (4.10b) are produced with overt COMPs:
(4.10) a. I want Bill eat cookie
b. I want that Bill eat cookie
Such a proposal, in fact, has been made by Maratsos (1978).
The facts from acquisition studies (e.g. Ingram 1972) indicate that
sentences like ‘want eat cookie’, i.e. time 4 in (4.8d), appear earlier than
those with overt embedded subjects and COMPs, as in (4.10). Maratsos
(1978) has argued from these facts that the child at time 4 then cannot be
said to have the structure in (4.9). Instead, he proposes that the underlying
structure of ‘want eat cookie’ is that of aux +verb + noun, parallel to adult
structures like ‘can eat cookie’ or ‘do eat cookie’. Chomsky (1981:9),
however, has argued for the maturationist position for time 4, stating that
the tenseless clause is the unmarked case. Presumably, this means that it
either matures earlier, or else is less affected by performance factors. He
does not explicitly state which is the case.
These alternative analyses exemplify the two views discussed. Chomsky’s
view is essentially untestable, in that child language data cannot be used for
or against it. Without an independent theory of markedness, what is
unmarked will be what the child uses first. This circularity is comparable to
that pointed out by Brainerd (see Chapter 2) about some stages proposed as
explanatory. Maratsos’s analysis, however, is testable. As pointed out by
Davis (1983), it predicts that children should produce sentences like (4.11a’
c) in parallel with other sentences with auxiliaries like (4.11b, d):
(4.11) a. * want I eat cookie?
b. can I eat cookie?
c. * want not go now
d. can not go now
Examples (4.11a, c) are asterisked to indicate that children do not seem to
produce these (see Davis 1983 for details). Davis (1983) concludes that
Maratsos’s analysis of ‘want’ in ‘want eat cookie’ as an auxiliary is thus
wrong. Elsewhere (Ingram 1985a) I have argued it is a serial verb construc-
tion, as in English sentences like ‘Go get a cookie’ and ‘Come see me’.
There are no studies on the acquisition of these, but serial verb analyses
predict they should occur around the time of ‘want eat cookie’.
This extended discussion of the sentences in (4.8) demonstrates the need
to include some restrictions on when we can say that a child has a particular
category or construction. If we allow the child to have rich underlying
structures in relation to its output, then we are again in the area of seeking
out performance factors. If we limit the distance between the child’s
76 EXPLANATION AND LANGUAGE ACQUlSlTlON
performance (in either comprehension or production) and its competence,

then we can make testable claims about the child’s rule system. Toward this
end, I propose the Competence Assumption, as stated below:
Competence Assumption: Assume that the child’s linguistic perform-
ance is relatively close to the child’s linguistic competence. That is, do
not propose a linguistic construct until there is evidence for it in the
child’s performance.
This is, of course, a very preliminary formulation of this assumption. To
make it more precise, we will ultimately need a more detailed understand-
ing of the performance factors operating in child language, and of the
relation between comprehension and production.
The Competence Assumption helps resolve the problem of how we can
go about making claims about the child’s underlying linguistic competence.
There is, however, one further problem to resolve before we can do this
with some degree of confidence. The Competence Assumption will isolate
the child’s data that potentially could have a particular structure, but it does
not allow us to decide if the structure is rule-based for the child. For
example, the child who says ‘I know that Bill ate’ provides data suggestive
that there is a COMP category, since ‘that’ was produced. However, the
child could have memorized part or all of the sentence. Therefore, we need
a further assumption about when a child has a productive rule of grammar.
In Ingram (1981d), I discuss the point that children’s utterances only
gradually become rule-based or productive. There, we found four steps
towards productivity brought about by rule (see also MacWhinney 1982):
step 1: unanalyzed whole utterance
step 2: analyzed utterance without productivity
step 3: partially productive utterance
step 4: productive utterance
An unanalyzed whole utterance is one where the child simply learns the
utterance without any knowledge of its internal structure. An example with
adults would be learning phrases in a foreign language for a vacation with no
knowledge of the structure of the language. In the next step to understand-
ing, the child may know the words as individual items, but still memorize
their combination without rules. Idiomatic expressions for adults are often
like this. For example, ‘What’s up?’ is an expression where we know the
individual words, but it is probably not generated by rules. A partially
productive utterance is one where only part of the utterance appears to
allow replacement or novelty. In the game ‘Simon says’, ‘Simon says’ in a
sense is non-productive in that it cannot change, even though the speaker
may. As will be presented in Part 11, children will learn some specific word
and combine others with it. The fourth step, or productive utterance, is one
produced by a rule which combines categories of grammar. It is the
productive utterance that we want to analyze to establish principles about
the child’s linguistic competence.
Unfortunately, establishing productivity has not always been a goal in the
use of acquisition data. It is possible to find discussions where claims about
the child’s competence are made on the basis of a small number of
utterances, as few as one or two. In such cases, it may be that the utterances
were not rule-based, but rather were partially memorized speech or even
speech errors. To avoid the suggestion, we add the Productivity Assump-
tion to our assumptions about children’s data:
Productivity Assumption: Assume that the child’s utterance was pro-
duced by a rule only when there is evidence that the rule is productive,
i.e. that it creates new instances of the structure under discussion.
Ways to establish the productivity of children’s utterances are discussed in
Part 11.
These three assumptions, the Constructionist Assumption, the Com-
petence Assumption and the Productivity Assumption, will restrict the
extent of restructuring and also the data available to make claims about the
child’s rules. They allow, however, for acquisition data to be used in
linguistic theory. Further, they restrict us to analyses leading to predictions
about the child’s language that are falsifiable.
4.5 Sources of variation among children

It does not take much exposure to children to realize that they can vary
greatly in their language use. This individual variation among children is a
feature that is emphasized by some and minimized by others. Chila
Language with its extensive experience with child data tends to emphasize
the extent of variation, often without explicit theoretical explanation. It is
common in Child Language, therefore, to see taxonomies of types of
language learners. Language Acquisition, on the other hand, with its focus
on a highly restricted theory of UG, tends to minimize variation. Since all
children are born with the same set of universal principles, they will be
similar in their linguistic competence. The variation that does occur will
need to be explained by other factors.
Elsewhere (Ingram 1981e) I have proposed three sources for variation
between children:
Performance variation: variation due to biologically determined indi-
vidual capacities or abilities of the child that lead to preferences for, or
better skill at, particular linguistic subsystems.
Environmental variation: variation due to environmental effects, from

obvious ones such as the need to hear a language to speak it, to more
subtle ones such as the effect of frequency on specific language forms.
Linguistic variation: variation due to the range of structural possi-
bilities allowed by Universal Grammar.
The theory of acquisition we construct will need to incorporate all three

kinds of variation.
Performance variation is a general category to cover all the kinds of vari-
ation that result from individual capacities and preferences. A non-linguistic
example would be running, where we can all do it, but some may do it much
better than others. While we can talk about runners vs. non-runners, we do
not wish to imply that running itself is only genetically determined for some
of us. Rather, it is the degree of ability to perform that leads to the typology.
Or, we can call people runners who like to run, even though they may not be
very good at it. Fun runs are filled with runners who prefer to run, i.e. they
enjoy it, even though they may not be gifted. These individual capacities and
preferences can lead to tremendous variation between people.
It should not be surprising that similar factors are at work in language
acquisition. Some children may simply be better at language learning than
others, and acquire it faster. In some cases, a child may be better at one
aspect of language, e.g. phonology, and subsequently have a more
advanced phonology in relation to syntax than another child. We might
expect to find some children more cautious than others, so that their
production of language has few errors. Other children may be more daring
and show several errors indicative of an incomplete mastering of some
aspect of language. The latter children will show their production closer to
their comprehension than others, and probably be better subjects for
acquisition studies. Differences can also result from specific preferences in
the child, for example, for certain sounds, sentence structures, or perform-
ance principles. These various differences are important to isolate, but at
the same time they will not tell us anything about linguistic competence.
They will not, therefore, contribute to our knowledge of UG.
Environmental variation will occur as a result of the differing linguistic
environments of the child. The most obvious effect is that the child only
learns the language that it hears. Ochs (1982a), for example, has shown that
children in Samoa will not learn certain grammatical morphemes if their
dialect of Samoan does not contain them. Maturationists often lose interest
in environmental effects at this point. For maturationists who accept the
restructuring hypothesis, there is an added interest in when certain structures
appear in the input language, and how often they need to be used. Presum-
ably, variation will appear if principles, can be triggered independently
of each other. Research in Child Language has indicated that there may
be other effects, such as earlier or more frequent use of a structure by the
child due to more frequent use by the parent. The extent of such effects is
still being argued, but we need to acknowledge their possibility in the
analysis of variation that we find.
The last kind of variation, linguistic variation, is the kind that is predicted
by Chomsky and others to be limited. We know from linguistic studies that
languages can vary greatly from one to another. Chomsky (1981) has
proposed a theory of UG whereby language will differ on certain para-
meters, e.g. those that require subjects, as in English, and those that can
drop their subjects, as in Italian. The latter are called pro drop languages.
Once the child notices this parameter, an entire set of assumptions about
the language follow. The child, in this theory, does not discover each fact
about a language independently, but would only need to discover a small set
of important parameters. Within any aspect, there will be a typical or
unmarked way to do things, and an unusual or marked way. The unmarked
way is expected by UG and thus will be the first guess by the child about the
language. The child will only reject the unmarked case when the language
the child hears presents evidence that it does things in the ‘marked’ way.
The extent of linguistic variation allowed by this theory is very limited.
Suppose that on the pro drop parameter, pro drop languages are the
unmarked case and therefore English is the marked case. We could envision
instances where one child has realized English is different or marked,
whereas another has not. These two children would appear quite different
from each other since one will be assigning a different set of principles to its
grammar from the other. This possible kind of variation does not seem
likely, however, given the extensive amount of exposure to language that
children have by the time they begin to form sentences. Given that children
are predicted to establish parameters from minimal exposure to the
language, the thousands of sentences heard by the normal &year-old
should be more than ample to set such a basic parameter as pro drop.
Another possibility is that the child may be learning an unmarked
language, but mistake it as the marked language, that is, the child would set
the wrong parameter. This possible variation is also questionable, on the
grounds that the unmarked language would not be likely to show aspects
that look like the marked alternative. Also, it would require an additional
ability in the child to retreat from a wrong parameter. It is not clear how this
could happen: more primitive kinds of innate release mechanisms do not
show such reversals, e.g. new goslings following the first moving object.
The current theory about UG predicts rather limited variation between
children. The assumptions about language acquisition presented in the
previous section will allow us to see if either of the above possible linguistic
variations occur. If so, it will increase our knowledge of the parameters

presented by UG and how they are set. If not, they will reduce the study of
individual variation to one that concentrates on linguistic performance and
the effects of the environment.
Further reading
General literature
The basic ideas in this chapter were first expressed in Ingram (1985a). White
(1982) discusses explanation from a different perspective. Her major
assumption is that the child’s language at any stage of development is a
possible human language. This is a weaker assumption than the Construc-
tionist Assumption, but it forces linguistic theory to account for acquisition
data. Chomsky’s position can be found in Chomsky (1975). It was first
adapted to issues of learnability by Wexler & Culicover (1980). An
extremely critical discussion of the attempts to explain acquisition to date is
Atkinson (1982).
The following are suggested excerpts from the above: Chomsky
(1975:3-35), White (1982: chs. 1-3), Wexler & Culicover (198O:l-29),
Atkinson (1982:1-26).
Part 11
Milestones
5 The period of prelinguistic development
5.1 Introduction
Most generally, this is the period of the infant’s development that extends
from birth to the onset of the first word. The discussion of this period begins
with an attempt to define its boundaries more carefully.
At first glance, the selection of birth as the beginning of the study of the
infant appears uncontroversial. Even here, however, it is necessary to
realize that an arbitrary point on the continuum has been imposed. In a
more extensive study, it would also be desirable to explore the nature of the
infant’s prenatal development, for there are important issues concerning
the fetus’s physiological development, particularly regarding the auditory
system and the vocal tract. There are also the possible effects of in utero
auditory experiences. We assume, however, that birth represents the
infant’s readiness to begin the experiences we normally associate with
human development.
The other boundary of the period needs to be set by narrowing down
what we mean by the acquisition of the first word. As will be seen in
section 6.1, defining this milestone is no easy matter. Recall from Chapter 3
that Stern defined the end of prelinguistic development as the onset of the
first word in production. Throughout Part 11, however, I will begin
whenever possible with the child’s receptive ability. The end of this period,
therefore, will be marked by the onset of the first words in comprehension.
The description of prelinguistic development will concentrate on three
major areas of development. The first is the infant’s ability to perceive
linguistic stimuli. This is the area of infant speech perception, which is
defined as the ability to perceive speech before the recognition that such
speech conveys meaning. This ability will later be compared to child speech
perception, which refers to the perception of meaningful speech, or lan-
guage. Child speech perception, then, begins with the period of single-word
utterances. (I will use the term ‘infant’ for the first year of life, and ‘child‘
thereafter; this appears to be a common practice, with no theoretical
implications.)
83
84 THE PERIOD OF PRELINGUISTIC DEVELOPMENT
A second area is that of infant speech production. Similar to the above,

this refers to the infant’s ability to produce speech-like vocalizations before
they are given linguistic meaning by the infant. Infant speech production is
commonly referred to by terms such as ‘cooing’ or ‘babbling’. This chapter
will make these terms more precise. In the period of single-word utterances,
infant speech production is followed by child speech production, which is
the production of meaningful speech, or words.
The third area to be covered is infant-adult interaction. This concerns the
kind of language addressed to the infant, as well as the infant’s response.
This area is of importance in that it focusses on the nature of the infant’s
linguistic environment.
Given that the goal is the study of the child’s language acquisition, why
would we even want to look at development before language begins?
The reason is that we need to examine this period to see what precursors to
language may exist. If there is a discontinuity between prelinguistic and
linguistic development, then the importance of looking at prelinguistic
development is minimized. If, however, important continuities occur, then
we will need to identify them and incorporate them into our theory of
acquisition. A constructionist theory, of course, will predict the existence of
such continuities.
5.2 Infant speech perception

Until recent years, the young child was considered to have little innate
ability. This certainly appears to be true to the casual observer: the young
infant responds to sound, but shows no recognition that it is in any way
meaningful. As will be shown, however, the infant has been found to have
an ability to perceive speech beyond anyone’s original expectations. Before
examining this ability, however, we will look at the theoretical implications
of studying the infant’s speech perception, and also some of the methods
used to investigate it.
5.2.1 The explanation of speech perception

The general theories of behaviorism and nativism, with the latter divided
into constructionism and maturationism, were discussed in Part I. These I
approaches can be examined in relation to the infant’s ability to perceive

speech. Each theory makes very specific predictions about the infant’s
ability.
Aslin & Pisoni (1980), developing the ideas of Gottlieb (1976), propose
four possible theories of speech perception. The first of these, the percep-
tual learning theory, is basically a behaviorist theory. It proposes that the
child’s perception is undeveloped at birth. During the first year of life,
however, the infant’s experience with auditory input will lead him to
recognize or perceive individual speech sounds. This theory, then, predicts
that the young infant will be very poor at speech perception, and that he will
only be able to perceive the distinctions of the language he is learning.
Nativist theories, on the other hand, propose that the child has innate
perceptual abilities. Aslin & Pisoni provide three distinct nativist theories.
One of these is the attunement theory, which is similar to a constructionist
theory. The attunement theory proposes that the infant is born with the
ability to perceive some of the basic sounds of language: other sounds,
however, will develop as the result of experience with the language being
acquired. There will be three directions to the infant’s development. The
sounds that the infant can perceive at birth will either be maintained, if they
are in the language being acquired, or lost, if they are not in the language.
Other sounds will be added if they are in the language but not part of the
innate set. This theory predicts that all infants will have the same initial
ability, and that linguistic experience will play a role in subsequent
development. The fact that the infant uses experience to develop new
perceptions, built upon a preliminary innate set, makes this theory a
constructionist one.
The other two nativist theories are versions of the maturationist view.
One of these is the universal theory, which claims that the infant is born with
the ability to perceive all the speech sounds found in human languages. This
theory claims that the young child maintains those sounds that occur in the
native language, and eventually loses the ability to perceive the others. The
universal theory is roughly comparable to the strong inclusion hypothesis
proposed for syntax in Chapter 4. Both of these positions suggest that the
child’s ability is very adult-like from an early age.
The last theory discussed by Aslin & Pisoni is the maturational theory. It
claims that the infant’s perceptual ability appears according to a biologically
determined schedule that is virtually unaffected by experience. This theory
predicts that all children will show these new perceptual abilities at
approximately the same times. The maturational theory is similar to a
version of the restructuring hypothesis of Chapter 4. The latter position
proposes that grammatical principles will simply appear at some consistent
point in grammatical development, again unaffected by the linguistic
environment. Both approaches allow the child’s abilities to be quite
different from one point to another.
An important point about these four theories is that they make empiric-
ally testable predictions; that is, they make distinct predictions about the
infant’s perceptual ability. These four theories and their predictions are
summarized in Table 5.1.
THEORIES
Speechsounds Perceptual learning theory Attunement theory Universal theory Maturational theory
Ability at birth none basic sounds all human speech sounds some or all sounds
Ability to perceive will never be perceived will only perceive them if will perceive all of them will perceive them at point
non-native speech sounds without experience part of basic sounds at which ability matures
during 1st year of life
Effects of linguistic determines entire course of determines the acquisition no role, therefore no no role, therefore no
experience child’s ability of non-basic sounds cross-linguistic differences cross-linguistic differences
in infant’s ability in infant’s ability
Until recent years, however, the testing of these predictions was only
logically possible. How is one to find out if a 4-month old infant, for
example, can hear or perceive the English fricative [J]? We certainly cannot
ask them, as we do adults. The ability to test these predictions requires new
and creative methods of investigation for use with infants. In the last 15
years there have been several major developments in the methods used to
test young infants that have enabled scientists to begin to test these theories.
The next section presents an overview of the most important methods that
have been developed.
5.2.2 Methods to test infant speech perception

Eilers (1980) describes three general procedures to study the speech per-
ception ability of infants. These are the high-amplitude sucking paradigm,
the heart-rate paradigm, and the visually reinforced infant speech discrimi-
nation paradigm. Each seems to be suitable for children at certain ages only.
The high-amplitude sucking paradigm (henceforth HAS) was first used
successfully by Eimas, Siqueland, Jusczyk & Vigorito (1971). In this task,
the infant is given a pacifier which contains a wire attached to a computer
which measures the child’s rate of sucking. As the infant sucks the pacifier,
he hears auditory stimuli, usually in the form of single syllables such as [ba],
[ba], [ba], etc. During an acquisition phase, the infant learns that he may
increase the amplitude (or loudness) of the auditory stimulus by increasing
the rate of sucking. When an increased rate of sucking occurs, the infant
enters the habituation phase of the experiment. After hearing the same
stimulus for several minutes, the infant gets used to (or habituates to) the
sound and subsequently decreases the rate of sucking. When this happens,
the infant is presented with one of two conditions. If the infant is in the
control group, he continues to hear the same stimulus item while his sucking
rate is recorded. If the infant is in the experimental group, however, he will
be presented with a new stimulus item. When the infant perceives that a
new stimulus has appeared, he will increase his sucking rate to make the
stimulus more audible. The new stimulus interrupts the habituation to the
old stimulus, and hence its appearance marks the dishabituation phase of
the experiment. If the rate of sucking for the experimental group is
significantly greater than that of the control group at this point, then it is
concluded that the infants perceived the difference between the stimuli.
Eilers (1980) suggests that this approach works primarily with infants from
birth to 3 or 4 months of age.
The heart-rate paradigm (henceforth HR) was developed around the
same time as the HAS (see Moffitt 1971). It differs from the HAS primarily
in that the infant does not control the presentation of stimuli. The infant is
monitored through the attachment of electrodes which record his heart

rate. When the experimenter is assured that the rate is normal she presents
the child with a series of auditory stimuli. e.g. [pa], [pa], [pa] etc. The
infant’s heart rate will decelerate as he listens to the stimuli. The stimuli are
halted then repeated several times until deceleration is minimal. At that
point, a new stimulus is introduced and the heart rate is followed to see if
deceleration occurs. If so, the experimenter infers that the child heard the
difference between the stimuli. The age range for use of the H R is greater
than for the HAS, being applicable from birth to 6-8 months. The upper
range occurs because the increased mobility of the older infant makes the
procedures difficult to complete.
The third technique, the visually reinforced infant speech discrimination
paradigm (referred to as the VRISD by Eilers, Wilson & Moore, 1976), is
the most recent, and was first reported in that publication. It represents a
major breakthrough as it can be used with older infants from 6 months up to
18 months of age. The description presented here is adapted from Kuhl
(1981) who has refined the technique.
The infant is seated on the parent’s lap and faces a research assistant; the
latter holds the attention of the infant by presenting a variety of toys.
Meanwhile, to the other side there is a visual reinforcer, which consists of a
toy animal in a dark plexiglass box. The toy animal is not visible to the
infant. Both the mother and assistant wear earphones so that they do not
hear the auditory stimuli and unconsciously prompt the infant. In the
background, the infant hears a series of auditory stimuli, e.g. [pa], [pa],
[pa], etc. Initially, the stimuli are changed and the toy animal (visual
reinforcer) is lit up by lights inside the box. The infant eventually learns
to turn his head to look at the box each time the stimuli change. The
experimenter and assistant both record all instances of the infant’s head
turns. A correct score would be if the infant turned his head at the time the
stimulus was changed. Eilers, Wilson & Moore (1977) used a criterion of at
least five out of six correct responses during stimulus changes to conclude
that a discrimination had occurred.
These three techniques have been adapted in a variety of ways to
investigate infant speech perception. As will be seen, the result has been
much better grasp of the infant’s ability. Unfortunately, the field has also
had its share of methodological difficulties and debates; nonetheless, there
are findings that are worthy of serious consideration.
5.2.3 Experimental results

To study an infant’s speech perception, we need to decide at the onset the
nature of the speech stimuli to be tested. Ideally, we would like to present
Table 5.2 The voice onset times for Spanish and English labial stop
consonants
LANGUAGE DIFFERENCES Phonetic

VOT English Spanish symbol
Fully voiced: voicing begins through does not occur occurs as /b/ [b]
entire production of the consonant.
Partially voiced: voicing begins during occurs as lbl does not occur [b]
the articulation of the consonant.
Voiceless unaspirated: voicing begins at occurs as stop occurs as /p/ [p]
release of consonant. after lsl, e.g.
‘spa’
Voiceless aspirated: voicing occurs after occurs as /p/ does not occur [p”]
release of consonant. , e.g. ‘pa’
subjects with a string of speech and have them tell us what they hear,
perhaps through making phonetic transcriptions and providing phonetic
commentary. While difficult to do with adults, this is of course impossible to
do with infants. Consequently, we must select a highly restricted set of
auditory stimuli, usually single syllables differing in just one phonetic
dimension, for example [pa] vs. [ba]. The techniques outlined above will
enable us to see if infants hear the two sounds as the same or as different.
Given the need to control for the interference of other acoustic cues, most
of the stimuli used have been artificially made.
Probably the most frequently studied acoustic feature in infant speech
perception has been voice onset time, or VOT. VOT refers to the moment at
which voicing occurs in relation to the release of a prevocalic consonant.
Table 5.2 demonstrates the V0.T for labial stop consonants in English and
Spanish. While both of these languages contain consonants that are written
as ‘b’ and ‘p’, their phonetic properties are quite distinct. The Spanish
distinction is between a fully voiced and voiceless unaspirated stop whereas
the English one is between a partially voiced and a voiceless aspirated
consonant. In addition, English has the voiceless unaspirated allophone [p]
that only occurs after /s/ (see Ladefoged 1975 for details).
One reason that stop consonants have been studied is that VOT can be
manipulated in the phonetics laboratory. Another, more important, reason
is that we have clear evidence which shows how adults perceive these
consonants categorially. To see this, we need to look for a moment at how
VOT is measured. We can assign a zero at the point at which a consonant is
released, and measure a consonant’s VOT from that point in milliseconds
(msec). The voiceless unaspirated consonants in Table 5.2 will have a zero
VOT, since voicing in these begins at the point of release. Those consonants
that have prevoicing (for example, fully voiced or partially voiced) will be
assigned minus VOT values, and those with voicing lag, e.g. voiceless
aspirated, will be assigned positive values. Experimentally, we can con-
struct artificial speech stimuli that differ systematically in their VOT values,
e.g. a [ba] will have a VOT of -10 msec, or a [pa] a VOT of +10 msec. We
can then play these artificial (or synthetic) consonants to adults and ask
them if they hear a /b/ or a /p/. If perception were continuous, speakers
should gradually begin to hear a /b/ as the minus values increase. Studies
like those of Lisker & Abramson (1967) show that this is not the case.
Around a VOT of +25, English speakers suddenly hear the stimuli as /pa/.
Before this, they hear them as/ba/. This result, which has been replicated, is
referred to as categorical perception.
We can study infants to see what specific acoustic parameters they can
perceive. One such parameter is VOT. Another aspect we can examine is
whether or not infants can perceive categorically as adults do. In 1971,
Eimas et al. published the results of an initial inquiry into both of these
issues. It has since become a classic study that initiated the field of infant
speech perception.
Eimas, Siqueland, Jusczyk & Vigorito (1971) This study was done to
examine two issues: (i) if infants can hear the difference between voiced
and voiceless stop consonants, and if so, (ii) if their discrimination is
categorical. To do this, they used the HAS with two groups of infants:
1-month olds and 4-month olds. There were 26 infants in each group, with
equal numbers of boys and girls.
The stimuli they used were synthetic speech sounds prepared at Haskins
Laboratories, the primary research center for much of the research on adult
perception of VOT differences. The stimuli were even prepared by Lisker
and Abramson, the investigators who reported categorical perception in
adults. There were six speech stimuli used, with VOT values of -20, 0,
+20, +40, +60 and +80 msec. The consonants in the first three of these are
perceived by adults as /b/, i.e. an English voiced consonant. The consonants
in the last three speech stimuli are perceived by adults as /p/. Recall from
above that English adults place the boundary for English /b/ vs. /p/
around +25 msec. The infants in both age groups were assigned to
experience one of three experimental conditions. In one condition, called
the 20 D condition, the infants were to hear the speech stimuli that had
VOT values of +20 and +40 msec. The ‘20’ represents the fact that the
stimuli differ by 20 msec in VOT and the ‘D’ refers to the fact that the
difference between the two crosses the adult boundary for categorical
perception, and thus the sounds are different. Another group of infants
was to experience two stimuli that were 20 msec apart in VOT, but which
20 D I
I
I
I
I
I 1
0
1‘- Y I
II
I
I
I
1
I
i\
I
-
I
I I
I I
1 I I
I I I
I I I
I I I
I I I I I Ill I I I hLLuhLu
B S 4 3 2 1 1 2 3 4 B 5 4 3 2 1 1 2 3 4 B S 4 3 2 1 1 2 3 4
Time (minutes)
Figure 5.1 The mean number of sucking responses per minute for three groups
of 4-month-old infants, taken from Eimas, Siqueland, Jusczyk & Vigorito (1971:
fig. 2).
were within an adult category. That is, if the infant heard these pairs as
adults do, they would sound the same. Thus, this was called the 20 S
condition, with ‘S’ for ‘same’. The two stimuli used for /b/ were -20 and 0,
and the two for /p/ were +60 and +80. Half the infants in condition 20 S
would hear the /b/ stimuli, and the other half would hear the /p/ stimuli. A
third group of infants was assigned to the control condition. These infants
would hear only one of the six speech stimuli, instead of two. This group was
needed to check that infants did not increase or change their sucking rate
even when no changes in stimuli were presented. There were eight infants
from each age group in each of the 20 D and 20 S conditions, and ten of
each in the control condition.
When the infants entered the laboratory, they were given a pacifier which
would record their sucking rate. At first, their sucking rate was recorded to
determine their normal (or baseline) rate. This was generally about 25 sucks
per minute. Next, the child would hear a stimulus item. The infant’s
increased sucking would result in an increase in the loudness and rate of the
stimulus. The infant would hear repeated instances of the sound until his
interest wavered, as measured by a decrease in the sucking rate. For the

infants in the 20 D and 20 S conditions, it was necessary to decide when
habituation to the first stimulus was sufficient to introduce the second
stimulus. The experimenters decided to shift to the second stimulus when
the decrease in sucking rate was at least 20 percent for two consecutive
minutes when compared to the rate for the immediately preceding minute.
This is an arbitrary measure that proved to be effective. The infant would
then hear the second stimulus item for four minutes, after which the
experiment was terminated. The control group infants, of course, would
hear only one stimulus over the entire experiment.
Figure 5.1 gives the results of the study (Figure 2 in Eimas et al., 1971).
The numbers along the vertical axis give the mean number of sucking
responses (or sucks) per minute at different times during the experiment.
The vertical line in dashes for each of the groups represents when the
sucking response had decreased according to the measure described in the
previous paragraph. At this point, the infants in the 20 D and 20 S
conditions were presented with a new stimulus. The control infants,
however, continued to receive a single stimulus sound. First observe the
control group responses to the far right, indicated with a zero. The sucking
rate continues to drop for this group, which is expected since this group is
still hearing the same sound. For the 20 D group (far left), however, there is
a sharp increase at this point, indicating that they have heard the change of
sounds. What about the 20 S group? As shown in the center of Figure 5.1,
they show a slight increase at this point. Statistical analysis of the data
revealed that only the 20 D infants showed a significant increase at the point
of stimulus change. The 20 S infants, however, did not show a significant
increase, that is, the slight increase they showed could have been due to
chance variation. These infants did not appear to hear the within-category
change of VOT. The experimenters conclude from this that infants have
categorical perception the same as adults do.
As you might imagine, the effects of these results were enormous. They
initiated a wide range of studies to examine the other speech sounds that
infants can hear. The results of these studies are impressive (see the reviews
mentioned at the end of the chapter). For example, it was found that
English children could discriminate a variety of non-English sounds at an
early age, sounds that even adult English speakers could not hear. At first it
was felt that this was a unique characteristic of the human auditory system
until Kuhl & Miller (1975) found similar results after testing chinchillas. It
now appears that the mammalian auditory system in general is able at birth
to perceive categorically the range of human speech sounds.
Eimas and his colleagues concluded from this and subsequent studies that
children are born with the perceptual abilities to perceive speech. The role
of experience, therefore, is primarily to help children maintain the distinc-
tions that occur in their language. Their research, then, is in support of the
universal theory of Table 5.1. During the mid 1970s, the bulk of research
expanded the list of the infant’s discriminable speech sounds (see the review
in Butterfield & Cairns 1974). Curiously, little thought has been given as to
when the infant loses the ability to make distinctions that are not in the
native language.
While the universal theory is still held by many investigators, it has been
challenged more recently by a series of studies that argue for an attunement
theory. Recall that this theory does not deny the impressive early abilities of
infants, but proposes that some more peripheral distinctions in language
require linguistic experience. To prove it, one needs to demonstrate that
certain distinctions are acquired only after linguistic experience with that
language. Some of the first studies to argue this position were Lasky,
Syrdal-Lasky & Klein (1975) and Streeter (1976). Here we will take a closer
look at one of the more recent of these studies by Eilers, Gavin & Wilson
(1979).
Eilers, Gavin & Wilson (1979) This study was undertaken to compare the
speech discrimination ability of English and Spanish infants on pairs of
synthetic stops that differ in VOT. Recall from Table 5.2 that English and
Spanish differ in the VOT values for the voiced vs. voiceless distinction:
Spanish places the distinction in the prevoicing range, while English places
it later (around +25 msec). The researchers decided to present the infants
with pairs representative of the contrasts in both languages. If the Universal
Theory is true, the infants should hear both distinctions. If the Attunement
Theory is true, then the infants should do better on just that distinction in
their native language. For the latter situation to occur, however, one of the
distinctions will need to be a basic one, in the sense of Table 5.1, and the
other to be a non-basic one. The non-basic one should show the effects of
experience.
Like Eimas et al. (1971), Eilers, Gavin & Wilson used synthetic speech
stimuli prepared at the Haskins Laboratory. The pair selected to test the
English contrast of ha/vs. /pa/ have VOT values of + 10 and +40 msec. The
Spanish pair VOT values were -20 and +10 msec. The subjects were eight
Spanish and eight English infants, all between 6 and 8 months of age. They
selected older children than did Eimas et al. because they wanted to allow
for the possibility of experience having an effect.
Since the infants were older, they used the VRISD already discussed. The
infant sat on the mother’s lap and was entertained by the assistant while
speech stimuli were presented over a speaker. The infant was first con-
ditioned (see 2.2.2) to turn to look at the visual reinforcer (a toy animal)
that would light up when the stimulus changed. In this study, the initial
conditioning was done with the pair [bit], [b~t].All the infants were able
to hear this pair and turned their heads toward the visual reinforcer when
there was a change from one to the other. Then they were presented with
the test stimuli. The infants were tested six times for each pair. Three
times the stimulus item changed, and three times it did not. A head turn
was scored as ‘correct’ when it coincided with a change in stimulus. The
infant was scored ‘wrong’ if he turned his head when the stimuli did not
change, or if he did not turn his head when the stimuli did change. An
infant was judged able to perceive a discrimination if he was correct five
out of six times.
There are two further details of the study that are worth emphasizing.
First, there is always the possibility that the mother or assistant might
unconsciously alert the infant to a sound change. To avoid this, both the
mother and assistant wore earphones and listened to music. Second, there
is the problem of judging what constitutes a head turn. The assistant and
experimenter, who watched through a one-way window, would judge
independently. If they felt that a head turn had occurred, they would push
a button. If both pushed their buttons, the electronic equipment would
then cause the visual reinforcer to light up. In this study, the assistant and
experimenter agreed 96 percent of the time on what constituted a head turn.
The measure used in the analysis was the percentage of correct head turns
on each pair of stimuli for each group of infants. The English infants were
correct 92 percent for the English pair, but only 46 percent for the Spanish
pair. When subjected to statistical analysis this difference proved to be
significant, i.e. greater than what would be expected due to chance
variation. The English infants, therefore, were better on the English
contrast. The Spanish infants, on the other hand, did well on both contrasts.
They got 80 percent correct on the Spanish pair and 86 percent on the
English pair, i.e. they did equally well on both pairs.
These results support the attunement theory if certain assumptions are
made. One is that the boundary for the English contrast needs to be a basic
contrast in languages, that is, one which all infants are born to discriminate.
Eilers, Gavin & Wilson (1979) suggest this, arguing that there are acoustic
reasons for this contrast being more salient. If so, this explains why the
Spanish infants were able to make this discrimination without any experi-
ence of hearing it. Another assumption is that the Spanish contrast is not a
basic contrast, and is therefore one which requires linguistic experience.
This will explain why only the Spanish infants were able to perceive this
contrast, while the English infants were not. This study, then, provides
suggestive evidence that experience may be necessary for certain less salient
speech discriminations.
5.2.4 Some controversial issues

The two studies described above represent the universal theory and the
attunement theory respectively. Both views assign the infant excellent
perceptual ability at birth, with the latter allowing some development over
time. Over the last few years, the disagreement between proponents of
these two views has polarized into a vitriolic debate. Possibly the best place
to view this is the Journal of Child Language, one issue of which offers a
direct confrontation between the two views (see Jusczyk, Shea & Aslin 1984
vs. Eilers, Oller, Bull & Gavin 1984). The basis of the controversy has been
on method and interpretation. For example, Eilers, Gavin & Wilson have
been criticized for accepting five out of six correct responses as sufficient to
conclude perception. Or, the claim is made that certain results are not
supportive of the interpretation they are given. It is important, therefore, to
realize that there is no universal consensus on the results of speech
perception studies to date.
Some of the controversial issues can be summarized as follows. (i) The
role of experience: the universal theory concludes from numerous studies
that the infant’s ability is excellent at birth and only some fine tuning is
necessary. The work of Eilers and her colleagues, however, suggests that
some contrasts may only be perceived after some linguistic experience.
(ii) Method ofinquiry: each of the three methods used to date has been
criticized. For example, it has been argued that heart rate responses may be
too gross to use to test certain contrasts. The measures used to conclude
accurate perception have also been criticized. (iii) The level of analysis: the
fact that infants can make fine distinctions was initially interpreted as the
result of a unique human ability for speech perception. Kuhl & Miller
(1975), however, have shown that chinchillas can also categorically perceive
speech. The latter study suggests that this ability is not uniquely human, but
a characteristic of the mammalian auditory system. (iv) The loss ofdiscrirni-
natory ability: both of the above theories will need to account for the loss of
innate speech discriminations. One position is that these are lost for ever,
and that adults, therefore, will not be able to perceive contrasts that do not
occur in their language. An opposing position is that the contrasts are not
lost, but that they will only be more difficult to discriminate. Recent work
by Werker & Tees (1984) indicate that some loss may occur by as early as 10
months of age. (v) The relation between physiological maturation and
speech perception: while much of the auditory system is physiologically
developed at birth, some changes continue to occur over the first year of life
(see Eilers & Gavin 1981: 192-4 for a review). The issue is whether or not
this development may contribute to a possible increased ability to perceive
certain speech discriminations.
96 THE PERIOD OF PRELINGWISTIC DEVELOPMENT
Despite these controversies, there are some general conclusions we can

draw about the infant’s speech perception. The major one is that the
young infant is born with much greater ability than was ever thought just a
few years ago. This fact makes the child’s rapid linguistic development a
year later less difficult to understand (though no less impressive!). It
appears that this innate ability combined with a year’s listening experience
is sufficient for the young child to begin to recognize language-specific
words around the end of the first year. Further, it appears that these
perceptions are categorical in that the discriminations are more abrupt at
specific acoustic parameters than at others. These two findings make the
young infant’s speech perception much more adult-like than was ever
anticipated.
5.3 Infant speech production

At first glance it might appear that infant speech production should be
easier to study than perception, since it is more overt. This, however, is not
necessarily the case. Lewis (1951), in his classic study of the child K, states
in very human terms the difficulties this area poses (p. xi):
The sounds that a child makes during his first few months are so
elusive and apparently so remote from anything that might be called
language that any observer - however interested in speech - might
well be pardoned for waiting until the noises become, at any rate, a
little more obviously human. To persist in making observations one
must be interested in the variety of human sounds merely as sounds,
one must have faith in the continuity of growth, and in addition,
perhaps, one must have something of that insensitiveness to ridicule
which is found at its highest level in the truly devoted parent.
The difficulty of recording the young infant’s sounds is such that it is an area
with a very small, but devoted group of researchers. As in the preceding
section, the discussion here will be concerned with the theoretical expla-
nations available, methods used, and lastly the findings to date.
5.3.1 The explanation of speech production

Using the theoretical possibilities from section 5.2.1, adapted to speech
production, the following theories (at least) can be considered:
Universal theory: the infant begins with the ability to articulate all
human speech sounds, then loses those that do not occur in the
Articulatory learning theory: the infant is born with virtually no
articulatory ability. Early speech sounds will consist of those heard in
the environment.
Maturational theory: the onset of human speech sounds will be
gradual, that is, according to a biological predetermined program.
Infants in all linguistic environments will show the appearance of
specific sounds at the same approximate ages.
Refinement (vs. attunement) theory: the infant begins with a prelimi-
nary or basic set of speech sounds to build upon. He then acquires or
adds other less basic sounds from the linguistic environment.
These four theories make predictions that are testable. It should be

possible (assuming adequate methodologies) to collect the infant’s vocali-
zations across language groups and see which of these theories is supported.
The primary evidence will consist of two bodies of data: the infant’s
prelinguistic vocalizations and the child’s early word productions. In this
chapter we will examine the current evidence from the former, in order to
draw some preliminary conclusions. The topic will be returned to in the next
chapter when the structure of the child’s early words is examined.
To pursue these theoretical possibilities, it is important to specify what
will be considered to be the onset of speech production. I selected birth as
our starting point with regard to speech perception. This was done despite
the fact that there may be perceptual development prenatally, and despite
the fact that the auditory system does undergo some physiological change
during the first year. Birth, however, proves to be a point where we can
assume that the physical structure is sufficiently intact to get clear evidence
of processing. Further, it is a point where our methods of inquiry are
applicable. We need to select a similar point from which to begin to test
theories of speech production.
Most investigators begin their study of speech production at birth. While
I will also look at the infant’s productions from that point, I suggest that the
above theories can only begin to be appropriately examined with data from
infants at around 6-8 months of age. The reason for this is the tremendous
physiological development that takes place in the infant’s speech apparatus
after birth, especially during the first year. These changes will continue in
fact up to 14 years of age. It is at approximately 6 months, however, that the
child’s vocal tract begins to approximate to its adult shape, and that the
vocal behavior generally referred to as ‘babbling’ begins.
The infant’s vocal tract is quite different from the adult’s; Lieberman
(1975), for example, has noted that the infant vocal tract is actually more
similar to that of the adult chimpanzee than it is to that of the adult human.
98 THE PERIOD OF PRELINGVISTIC DEVELOPMENT
Kent (1981: 106-7), citing the work of Goldstein (1979), gives four major
differences between the two: ‘(1)the infant’s tract is appreciably shorter,
(2) the pharynx is relatively shorter in the infant, (3) the infant’s tract is
wider in relation to its length, and (4) because the infant does not have
erupted dentition, the oral cavity is flatter than in the adult’.
One consequence of these structural differences is that the resonant cavi-
ties are quite different for speech production. The size of the oral cavity is
much more restricted than the adult’s, with the tongue filling most of it.
Further, the tongue’s movement is also limited by incomplete maturation of
its complex set of muscles. Besides a narrow oral cavity, the infant does not
yet have a pharyngeal cavity (the area behind the back of the tongue in the
adult). Lastly, the operation of the velum is designed so that the infant can
breathe predominantly through the nasal cavity. Kent points out that the
transition to possible oral breathing occurs around 6 months of age. Before
this, the infant’s vocalizations are thus highly nasalized.
These differences combine to demonstrate that the infant’s vocalizations
will be heavily influenced in the first six months by physiological properties
of the developing vocal tract. While these changes do not stop at age 6
months, the infant develops enough by then to produce sounds that are
more easily recognizable to adults as human speech sounds. I propose,
then, that the development of these sounds over the last half of the first year
of life will constitute the testing ground for the theories just outlined.
There have been proponents for variants of each of the possible theories
of early speech production. Possibly the first to gain popularity during the
period of diary studies was a form of the universal theory. Due to the restric-
ted capacities of the infant’s vocal tract (as just described), the infant will
begin with many velar-like vocalizations. Early diarists saw in these sounds
the exotic sounds of the diverse languages of the world being described by
the new discipline of descriptive linguistics. Ignorant of their physiological
determinants, the diarists would hear an assortment of sounds not heard in
their native languages. As these diverse sounds were compiled from more
and more diaries, they came to be an extensive list of non-native forms.
The expression of the universal theory is most often identified with the
Russian linguist Roman Jakobson. In 1941, Jakobson published in German
a book entitled Kindersprache, Aphasie, und allgemeine Lautgesetze, which
was translated into English in 1968 as Child language, aphasia, and phono-
logical universals. In this book, and in several of his other writings, Jakob-
son made the proposal that infants babble the wide range of sounds that
may occur in the languages of the world. This position resulted from his
review of a large number of diary studies on children acquiring different
languages - Jakobson himself never actually collected data from children
directly.
The following lengthy extract, taken from his presentation at the Fifth
International Congress of Linguists in 1939, presents Jakobson’s basic views
on babbling:
During the babbling period the child easily produces the widest
variety of sounds (for example, clicks, palatalized, rounded or phar-
yngealized consonants, affricates, sibilants, etc.) almost all of which
he eliminates upon passing to the ‘few words’ stage . . . It is true that
some of these disappearing sounds are not maintained by the child
when they do not occur in the speech of those around him, but there
are other sounds which suffer the same fate despite their presence in
the adult speech, and the baby reacquires them only after much effort.
Such is frequently the case with velars, sibilants, and liquids.
An important feature of Jakobson’s version of the universal theory is his
proposal for the transition from babbling to words. The universal theory as
stated earlier simply allows the infant to maintain the sounds that will occur
in the language, while dropping others. Jakobson, however, restricts these
sounds further by suggesting an abrupt discontinuity between babbling and
the first words. The latter will be constrained by a theory that predicts the
order of appearance of phonological oppositions. Only the sounds that
participate in the first phonological oppositions (or contrasts) will be
maintained. Other sounds that are part of the language being acquired,
however, will be lost temporarily only to be reacquired later. The details of
Jakobson’s theory of the emergence of phonological oppositions will be
presented in Chapter 6. Here the important point is that the infant is given
the ability to babble or produce all possible speech sounds. Jakobson’s
theory had a tremendous impact on the field of child language and has been
frequently cited as a fact about the infant’s babbling, even in recent years.
Since the advent of behaviorism, there have been several theoretical
proposals made that could be described as variants of an articulatory
learning theory. Examples of these can be seen in Skinner (1957), Mowrer
(1960), and Winitz (1969). The last contains a particularly in-depth review
and elaboration of such theories. Most generally, acquisition proceeds in
the following way. First, since the mother will occasionally vocalize while
feeding the infant, her vocalizations become associated with a primary drive
(feeding). Since the infant’s vocalizations sound similar to the mother’s,
they too acquire reinforcing properties. Further, the infant’s vocalizations
will be reinforced or rewarded by the mother, especially if they sound like
the mother’s. That is, the mother (or parents) will reward or encourage the
infant to produce vocalizations like the adult language (e.g. ‘mama’,
‘papa’). This leads the infant to imitate actively the speech he hears.
While such theories rarely discuss the acquisition of different languages,
it seems that they predict potential linguistic variation. The infant’s need to
be reinforced for sounding adult-like suggests that the infant’s babbling will
soon begin to sound like the adult language, if only at first in general
features like rhythm and timing. If this is so, the infant’s babbling should
sound different from language to language. Further, there should not be
any use of non-native sounds in the infant’s later babbling or first sounds
since these would not be reinforced by adults. For these theories to work,
the infant’s motor speech development at the end of the first year should be
sufficient to allow the child to imitate specific adult sounds.
A version of the maturational theory can be found in Locke (1983). Locke
proposes that infants in all linguistic communities will show the appearance
of certain basic, or what he calls ‘repertoire’, sounds at biologically predeter-
mined times. There is, therefore, no effect of the linguistic community on
the infant’s vocalizations. He states (p. 84):
I will suggest that no genuine accommodations to the adult system will
be evident until the child reaches the systemic stage of phonological
acquisition, which probably occurs at some time after the first 50
words are in use.
He therefore predicts that the babbling of infants in different linguistic
communities will all be the same.
Locke’s theory, however, is not completely maturational. At the point
when phonological acquisition begins, presumably around the 50th word
produced, the child shifts to a different procedure. Locke says that ‘it seems
reasonable to ascribe the child’s articulatory progress for nonrepertoire
sounds to an environmental interactive process we might inelegantly term
learning’ (p. 87). I know of no one who has proposed a complete
maturational theory of speech production. We can of course hypothesize
one, which predicts that all sounds, not just basic or repertoire ones, occur
at approximately the same time for all children.
The refinement theory is basically a constructionist theory. That is, it
proposes that the infant will go through well-defined stages of vocalization
where the earlier stages provide the building blocks of later ones. To be
consistent with constructionism, these stages will also require interaction
with the environment for their development. A theory of this kind can be
found in the work of Oller (1980,1981). We will go through the details of his
theory in section 5.3.3. Generally, he outlines five stages of infant vocali-
zation which each contribute to the infant developing adult-like speech
production. He proposes four major parameters that develop: pitch and
voice quality, resonance, timing, and amplitude. The emergence of adult-
like speech does not appear reflexively, however, but as a result of the
infant’s exploration of his speech ability and selective imitation. Oller’s
model, however, does not have a strong interactional component. He
believes that the infant in the first year is primarily focussing on the basic
inventory of speech abilities. Infants in different linguistic communities,
therefore, are all coming toward the same point and will sound similar to
one another. In this sense, Oller’s viewpoint is similar to Locke’s in regard
to predictions about cross-linguistic babbling. A more constructionist
model would predict some cross-linguistic differences, either in the later
babbling or at least in the structure of the early words.
5.3.2 Methods to study infant speech production

Initially, the traditional method of phonetic transcription was used to record
infants’ vocalizations. As in recording a new language, the International
Phonetic Alphabet (IPA) would be used. There have been numerous
studies which have used this method. As was pointed out in 5.2.1, however,
the vocalizations of the very young infant, up to around 6 months at least,
are not very adult-like. Many of the infant’s early sounds, unfortunately,
cannot be transcribed by the IPA. This method, therefore, is more suited to
studies with older infants. It does allow, however, some preliminary
observations on the infant’s earlier vocalizations.
In recent years, attempts have been made to analyze the infant’s
vocalizations by submitting tape-recordings to spectrographic analysis. The
spectrograph is a machine which provides a visual copy of the primary
acoustic features of utterances. The first major discussion of the attempt to
do this can be found in Lynip (1951). Unfortunately, this approach does not
always make the study of infant vocalizations any easier. Often the
spectrograms are so difficult to read that it is impossible to judge certain
acoustic parameters, at least with very young infants. Lieberman (1980) has
been able to do spectrographic analysis of infant vowel productions with
infants only as young as 4 months old.
With infants below 3 or 4 months of age, the most efficient technique has
been developed by Stark, Rose & McLagan (1975). They made spectro-
grams of selected vocalizations, then used multiple interpretations. Each
analyst would read the spectrogram and hear the vocalizations, then
describe whether or not certain features were present. Table 5.3 gives a
summary of the features they looked for. Stark, Rose & McLagan refer to
this as an auditory judgement-spectral feature system, and argue that it is the
most effective approach currently available for studying the earliest vocali-
zations of infants. In using this system, the analyst does not need to identify
a segment as a recognizable one of some human language, but instead only
needs to identify some more noticeable features. Even with this amount of
caution, Stark (1980) points out that certain features cannot be identified in
Table 5.3 The auditory and spectral features used by Stark, Rose &
McLagan (1975) to study the vocalizations of two infants from 1 to 8
weeks of age
Category Feature
Features of breath direction, voicing, egressive vs. ingressive airflow

pitch and loudness silence in mid-segment
voicing: voiced, voiceless or breathy
voicing change
forceful expulsion of air
pitch: absent, normal, high
pitch contour: flat, rise, fall or combination
raised pitch glide
loudness: faint, normal to loud
Degree of constriction of vocal tract open (vowel-like) or closed (consonant-like)
above glottis
Features of vowel-like sounds glide
glottal stop
voice quality features: harsh, vocal fry,’
pharyngeal friction, subharmonic break2
Features of consonant-like sounds friction noise, trill, nasal, liquid, stops, click,
fricative
sometimes referred to as ‘creaky voice’

* a doubling of the harmonics of the fundamental frequency, giving a roughness of the voice
without a change in pitch
these first vocalizations with any inter-judge reliability. These difficult

features are place of articulation, for consonants, tongue height and
position for vowels, and nasality of vowels.
Besides recording techniques such as the above, some studies have used
recognition techniques. I use the latter term to refer to cases where
listeners are played segments of infants’ vocalizations and asked to
identify them in some manner. This kind of technique is often used in
studies trying to see if infants in different linguistic communities babble
the same or not. Typically, the listener will hear two segments of babbling
and judge whether the infants come from the same language or different
ones. Indeed, most of the studies which have tried to investigate the
effects of the linguistic environment on babbling have used a version of
this technique.
5.3.3 Studies on infant speech production

The studies on infant speech production range from diary studies using IPA
to modern studies applying spectrographic analyses. Here we will look at
Table 5.4 M . M. Lewis’s results on the vocalizations produced by infants
during the first year of life, taken from Lewis (1936151: Table 1 )
Sound class Sounds uttered in discomfort Sounds uttered in comfort
I. Vowels (i) Onset: immediately after (i) Onset: when the discomfort
birth cries have already begun to
(ii) Limited mainly to sounds a, appear
a, e, e (ii) a wider range, much less well
(iii) Often nasalized defined in quality
(iii) very rarely nasalized
11. Early consonants The semi-consonant ti appears The back consonants y, x, g, k, r
early, followed by h, I, and 0
111. Later consonants The front consonants, almost The front consonants, both nasal
exclusively nasal: m, n and oral: m, n, p, b, t , d
two classic studies that constituted the primary references in the area for
several years, and then turn our attention to more recent work.
Lewis (1936/51) Lewis’s classic work, originally a Ph.D. dissertation at the

University of London, remains one of the richest diaries ever recorded on
development during the first two years of life. It is also the only English
source for the primary data reported by Stern & Stern (1907) on the
acquisition of German by their two children. Lewis combines his observa-
tions on a young boy, K, with those collected by other diarists. Much of
these data are reported in a rich collection of appendices.
Lewis kept a record in IPA of K’s vocalizations during the first year, and
subsequently compared it with diary data from three other children: Axel
Preyer (Preyer 1889), Gunter Stern (Stern & Stern 1907), and Hoyer
(Hoyer & Hoyer 1924). Lewis divided the sounds of these four children into
comfort and discomfort sounds, since he considered that they were different
from each other. Table 5.4 gives his summary of his analysis.
One of the major findings was that the infants’ early consonants were
back or velar sounds, with more typical labial and dental consonants not
occurring until later. It is because of these early velar articulations that the
early months of the infant’s development have been referred to as a
‘cooing’ stage. (Stark 1980, in fact, retains this traditional term in her
typology.) Another important finding was that the vowel-like vocali-
zations were difficult to identify, and appeared to occur as lower or more
fronted vowels. Given what we now know about the physiological prop-
erties of the infant’s vocal tract, these results are not surprising. They
were seen, however, as the first serious attempt to identify the young
infant’s vocalizations.
Table 5.5 A summary of the major characteristics of the series of studies
on infant speech production done by 0 . Irwin between 1941 and 1949,
based on McCarthy (1954)
Subjects: Earlier studies (1941), based on 40 infants during first 10 days of life; small groups of
10 to 15 infants in the first six months.
Later studies (1946-1948) 95 infants in the first 30 months, originally to be visited
twice a month longitudinally; there was a significant attrition rate.
Method: Vocalizations were tape-recorded and transcribed. The sample for each child
consisted of all the vocalizations that occurred within 30 breath groups. Multiple
transcribers were used and reliability was claimed to be over 90%.
Measures: 1 . Simple frequencies' of the occurrence of vowel and consonant tokens
2. Simple frequencies' of vowel and consonant types
3. Ratios of token frequencies of vowels to consonants
4. Ratios of type frequencies of vowels to consonants
5. Type to token ratios, Le. how often were specific types used
6. Comparison of adult and child frequencies of occurrence.
~ ~ ~ ~
I All frequencies are given for each child's sample for 30 breath groups.
Irwin (194149) As might be expected, the behaviorist oriented research-

ers in the United States were not satisfied with a diary account of this topic.
A researcher at the University of Iowa, Orvis Irwin, initiated a series of
large sample studies on infant vocalizations, and his studies remain the most
extensive ever undertaken to study early infant speech production (an
insightful treatment of his work can be found in McCarthy 1954: 507-11).
Irwin began his research by taking every phonetics course available at
the University of Iowa, and then extensively studying with several phoneti-
cians. He subsequently trained two assistants, Curry and Chen, the latter
eventually completing both a master's thesis and doctoral dissertation on
the topic. Either alone or with one of the above, Irwin published approxi-
mately 30 articles on his research from 1941 to 1949.
Table 5.5 gives a general overview of Irwin's studies. His subjects fall into
two groups, one followed for the earlier studies, the other for the later ones.
The later group of 95 infants was originally to be visited twice a month for 24
years. For various reasons, not all of the infants were visited for the entire
period. The data from these latter infants were analyzed in seven major
reports from 1946 to 1949. These studies are the ones most often cited (e.g.
in Winitz 1969). Irwin is not always clear on the method by which data were
collected and transcribed. In one early report, he does claim that observer
reliability was above 90 percent. Given the difficulties found by Stark, Rose
& McLagan (1975), however, one needs to view these studies with a certain
amount of caution. The sample size for each infant consisted of all the
vocalizations produced within 30 breath groups. These breath groups could
Table 5.6 Major results of 0 . Irwin on infant vocalizations during the first
year of life, adapted from McCarthy (1954: Table 2)
Age (in months)

Measures 1-2 3-4 5-6 7-8 9-10 11-12
No. of subjects 62 80 75 64 62 62
No. of records 125 181 166 170 147 149
1 . Mean number (tokens)’ of
vowels 49 53 56 55 56 58
consonants 11 14 18 21 25 34
2. Mean number (types)’ of
vowels 5 7 7 8 8 9
consonants 3 5 5 7 8 1 0
Percentage of consonants
glottal, velar 98 90 87 75 53 41
dental, alveolar 1 6 5 12 26 36
labial 1 4 8 13 21 32
Percentage of vowels
front 72 61 62 63 62 62
central 25 26 24 20 17 16
back 2 14 14 17 21 22
Per 30 breath groups per subject.
include both crying and non-crying sounds: presumably the younger the
infant, the greater number of crying vocalizations were included. The
measures most frequently reported were frequency of vowels and con-
sonants, for both types and tokens, and various ratios between these
frequencies.
In Table 5.6 I give some of the data from the seven major studies between
1946 and 1949 that are behind the findings in Irwin’s work. This table
reflects four findings. First, vowel-like vocalizations are much more fre-
quent than consonant-like ones at the onset, and the number of vowel-like
segments remains constant over the year while the number of consonants
nearly triples. Second, the differences between vowels and consonants are
not so great when types are observed. In fact, by 7 to 8 months of age, there
are as many types of consonants as vowels in the infant’s vocalizations.
Third, the distribution of consonants reveals that virtually all of the
consonants up to 6 months are either glottal or velar. A shift begins at 6
months, however, so that by 12 months the majority of consonants are
either labial or dental/alveolar. Fourth, the data on vowels indicate that
back vowels are infrequent in the first months, but increase gradually
throughout the year.
Despite methodological problems, Irwin’s studies provide an extensive
body of data that in essence support and add to Lewis’s results. Together,
they present a picture of the young infant using a predominance of vowel-like
vocalizations that sound like front or low vowels. Further, the early
consonantal productions are either velar or glottal. A shift towards more
back vowels and front consonants begins around 6 months of age. These
results were commonly cited in the literature over the subsequent years. Of
some consequence is the fact that these available data were quite in
opposition to the claim by, Jakobson (1941/68) that children babble all
possible sounds. Despite this, Jakobson’s position continued to be stated, in
the face of available findings.
Oller (1980) Sporadic diary studies continued to appear after Lewis’s

work (e.g. Cruttenden 1970; Stockman, Woods & Tishman 1981). Major
projects, however, were initiated in the 1970s by two investigators, Rachel
Stark at Johns Hopkins University in Baltimore and Kim Oller, initially at
the University of Washington and subsequently at the University of
Miami. Both have proposed stages of development based on their
findings. Here we will focus on Oller’s results and his subsequent theoreti-
cal interpretations.
Oller (1980) presents an analysis of infant speech production which
divides the infant’s development into five stages. His overview of this
analysis is presented in Table 5.7. First, we will look at the descriptive
characteristics of his stages.
The first stage, the phonation stage, is one characterized primarily by
crying and generally reflexive vocalization (e.g. grunts, burps, etc.). During
these first few weeks, however, some non-reflexive vocalizations do appear.
Oller (p. 95) labels these first sounds as QRNs, or ‘quasi-resonant nuclei’:
these elements have been categorized by previous authors as vowels,
syllabic consonants, andlor ‘small throaty sounds.’ Quasi-resonant
nuclei include normal phonation (not vocal fry, not breathy voice,
etc.) but do not seem to involve any systematic contrast between
opening and closing of the vocal tract, and do not make use of the full
potential of the vocal cavity to function as a resonating tube.
Oller points out that these often sound like syllabic nasals or nasalized
vowels. When producing QRNs, the infant will often have the mouth
closed, or nearly closed.
The second stage, at 2 to 3 months, is the GOO stage, and is the one when
the velar-like sounds reported by Irwin appear, and occur along with
QRNs. GOOs are back consonant-like sounds, often velars, and often
voiced fricatives. They comprise the sounds [y, x, k, g] proposed by Lewis
(see Table 5.4). Oller points out that the occurrence of GOOs in his subject
L at 2 months ranged from 13 percent to 49 percent of the infant’s
non-reflexive sounds. They may combine with QRNs to produce
Table 5.7 Oller’s stages of infant speech production, taken from Oller (1980: Table 2)
Normal infant’s age Characteristic vocalization types Metaphonological characteristic of mature languages
~~ ~
0-1 month QRN (quasi-resonant nucleus) Normal phonation in non-reflexive vocalizations

phonation stage
2-3 months GOO (QRN plus velar or uvular consonant-like Vocalizations with closure: alternation between opening
GOO stage element) and closure of the vocal tract
4-6 months FRN (fully resonant nuclei) Use of resonance capacity providing possibility for
expansion stage contrasts of resonance types
RSP (raspberry) Front as opposed to back (GOO) closures
Further manipulation of vocalizations during closure
SQ (squeal) Pitch contrasts
GRL (growl)
YEL (yell) Amplitude contrasts
IES (ingressive-egressive sequence) Further control of vocal breath stream
MB (marginal babble) Alternation of full opening and closure of the vocal tract
7-10 months BB (canonical babbling) Syllabic timing constraints on relationship of openings and
canonical stage closures (vocalic transitions)
11-12 months VAR (variegated babbling) Contrasts of consonantal and vocalic type
variegated babbling
stage
GIB (gibberish) Contrasts of stress
syllable-like sounds, but the timing of the opening and closing of these
syllables is irregular and not particularly adult-like.
Around 4 months of age the expansion stage begins, a period char-
acterized by the addition of a range of vocalizations to the infant’s phonetic
inventory. These do not appear in any particular order, and the infant may
go through periods of concentration on one or the other. Table 5.7 lists
seven such sounds. Four of these are obvious from their labels: RSP
(raspberry), bilabial or labiolingual trills; SQ (squeal), high-pitched sounds;
GRL (growl), low-pitched creaky sounds (the opposite of squeals); YEL
(yell), high-amplitude or loud vocalizations. In any particular sample from an
infant at this stage, the occurrence of any of these may vary tremendously.
This is also true for another of the new sounds, IES, or ingressive-
egressive sequences. These are vocalizations that concentrate on varying
ingressive and egressive breathing.
The two other new vocalizations merit special attention. FRNs, or
fully-resonant nuclei, are vowel-like vocalizations that sound more like adult
vowels. They have stronger resonances than the QRNs. For Oller’s subject
L, they comprised less than 10 percent of the vocalizations between 4 and 5
months. However, when adjacent to consonant-like elements, they
accounted for 59 percent of L’s vocalizations. MB, or marginal babbling,
begins in this stage, but is infrequent, less than 5 percent of L’s vocalizations
between 4 and 8 months. Marginal babbling consists of ‘sequences in which
a closure of the vocal tract is opposed with an FRN’ (p. 98). These
sequences are not reduplicated like later babblings, and they do not have
regular timing to the syllables.
The canonical stage around 7 months is the one usually identified with the
onset of babbling. To distinguish the babbling at this time from the earlier
marginal babbling, Oller refers to canonical babbling, or BB. ‘BB includes
both consonant-like units and FRNs (vowels) in a timing relationship that
conforms to mature language restrictions’ (p. 98). Oller points out that the
onset of the stage is relatively sudden, and that many of these units are
reduplicated, e.g. [bababa], [dadada]. He distinguishes two kinds of
canonical babbling: reduplicated babbling (RB) and single-consonant bab-
bling (SCB), e.g. [ba]. At this stage the latter is as frequent as reduplicated
babbling.
The last stage of infant vocalization for Oller is the variegated babbling
stage. There are two new types of vocalizations that appear: VAR (varie-
gated babbling) and GIB (gibberish). Variegated babbling consists of
sequences of consonants and vowels that differ. These vocalizations appear
to overcome the restrictions on the earlier canonical babbling. Gibberish
refers to ‘phonetic sequences with contrasts of syllabic stress. In GIB, fully
canonical syllables occur in the same utterances with low-stressed, often
slurred syllables that seem to have QRNs as nuclei’ (p. 99). These
sequences sound as though the infant is producing an adult sentence with no
recognizable words.
As just described, Oller’s stages are only co-occurrence stages in the
sense of Chapter 3, Figure 3.6, defined by the occurrence of new vocal
behaviors. Oller wants, however, to go beyond this and to describe these
stages as construction stages. That is, he wants the stages to build upon each
other toward adult-like speech. To do this, he outlines what he calls
‘metaphonological parameters’. There are four of these general parameters
that need to be developed and coordinated before adult-like vocalizations
can be achieved:
Pitch and voice quality parameter: the development of the control of
pitch is practiced through squeals and growls; the emerging ability to
control pitch is seen later in development through the pitch variations
found in variegated babbling and gibberish.
Resonance parameter: the development of the articulation of fully
resonant vowels. The frequent quasi-resonant nuclei of the first
months give way to fully-resonant nuclei in the later stages.
Timing parameter: the development of control over the breathing
cycle in order to produce smooth consonant-vowel transitions. The
emergence of reduplicated babblings initiates adult-like timing of
syllables.
Amplitude parameter: the development of the control of amplitude or
loudness. The concentrated use of yells is an example of manipulation
of this parameter.
Table 5.7 lists the development of these parameters over Oller’s five stages.
Oller’s stages are constructionist in that the earlier developments of these
four parameters build upon each other toward more adult-like speech. To
make these stages descriptive, in the sense of section 3.3, we also need to
propose causes for the changes the infant undergoes. Here we will consider
three possible causes: physiological change, maturational growth, and
intentional exploration.
We have already discussed the extensive physiological or structural
changes undergone during the first year of life. Kent (1981) has explored the
possibility that these changes may account for the stages described by Oller.
Table 5.8 gives the anatomical developments suggested by Kent as possible
causes of the infant’s vocalizations from birth to 6 months of age. Kent is
careful, however, not to conclude that these are the causes. He states
(p. 111):
Table 5.8 Parallels between Oller's first three stages of phonetic development and physiological development, from
Kent (1981: Table 1)
1 Age of infant Phonetic development Anatomic-physiologic correlate

I
' 0-1 month quasi-resonant nucleus Nasal breathing and nasalized vocalization because of engagement of
phonation stage larynx and nasopharynx. Tongue has mostly back-and-forth motions and
nearly fills the oral cavity.
2-3 months quasi-resonant nucleus plus velar or Some change in shape of oral cavity and an increase in mobility of tongue;
GOO stage uvular constrictions but tongue motion is still constrained by larynx-nasopharynx engagement.
4-6 months fully resonant nuclei Disengagement of larynx and nasopharynx allows increased separation of
expansion stage oral and nasal cavities, so that non-nasal vowels are readily produced.
raspberry (labial) The intra-oral air pressure necessary for fricative-like productions can be
developed with some regularity because of larynx-nasopharynx
disengagement. Raspberry results from forcing air through lips, which
close after each air burst because of natural restoring forces.
squeal and growl Contrasts in vocal pitch are heightened perhaps because descent of larynx
into neck makes the vocal folds more vulnerable to forces of
supralaryngeal muscles.
yelling Better coordination of respiratory system and larynx, together with
prolonged oral radiation of sound, permit loud voice.
marginal babble Alternation of full opening and closure of vocal tract is enhanced by
larynx-nasopharynx disengagement.
Although a causal relationship has not been firmly established, the
evidence certainly invites the tentative conclusion that major disconti-
nuities in vocal behavior in the first year are related to significant
remodeling of the oropharyngeal anatomy. This is not to argue for a
physiologic-anatomic determinism of early vocalization, but merely
to stress the importance of physiologic and anatomic factors in
evaluating early vocal behavior.
Oller (1980) acknowledges these changes as a contributing factor, but

wishes to go beyond them, pointing out that they do not account for certain
developments. For example, a purely physiological explanation would
predict a gradual increase in fully-resonant nuclei as the vocal tract
develops. Their occurrence, however, is relatively abrupt. Oller argues
(p. 108): ‘Such a result would suggest that an infant’s introduction of FRNs
is motivated in part by central (nonanatomical) factors. One possibility is
that the infant is attempting to produce sounds like those heard in his or her
environment.’
A second possibility is that these stages are the result of a biologically
predetermined maturational schedule, a position argued for in Locke
(1983), and presented earlier in this section. There are two major character-
istics of a maturational model of development. One is that the infant is seen
as a passive participant in the process, in the sense that it does not explore,
discover, seek out or otherwise operate on the environment to achieve the
skill involved; instead, there is a natural unfolding from within. The second
is that the role of the environment or experience is minimal. These two
characteristics are interrelated. As long as the infant’s environment is
normal, the sequence of development will be the same for all infants, with
no observable effects of linguistic experience.
Oller (1981) discusses the possibility that the infant’s stage of production
could be explained by a maturational or biological model. As with the
physiological explanation, he does not deny that maturation plays some
role. He does not, however, want these to be the only two factors. Given
that maturational models are not always easy to test (see the discussion in
Chapter 4), it is important to look carefully at Oller’s discussion of this. His
arguments are relevant not just here but also to later chapters where this
model is raised to explain other areas of language acquisition.
Oller (1981) discusses a biological explanation in which the infant’s
vocalizations are innate reflexes which, when they are ready (Le. matured),
appear in response to some environmental stimulus. He proposes three
components to this model: (i) the reflexive behavior should be elicited
whenever the appropriate stimulus occurs, (ii) any stimulus within the class
of appropriate stimuli should elicit the behavior; (iii) the reflexive behavior
does not change its physical form across time. He argues that infant
vocalizations do not have any of these properties. There does not appear to
be any environmental stimulus that will consistently elicit infant speech.
Further, infants are very selective in the adults they are willing to vocalize
with. Lastly, as shown in Table 5.7, the infant’s vocalizations change
structure. Oller concludes that the infant’s vocalizations cannot be reflexive
behaviors of a passive being. Rather, the infant is said to play an active role
in development.
Locke’s model (1983) does not have all the properties of the biological
model described above. He allows the infant’s vocalizations to change, and
he considers that the vocalizations do not occur as a result of external
stimuli (that is, through hearing adults), but of internal stimuli. His
evidence for this is data suggesting that deaf infants babble the same way as
hearing infants, even though they do not hear speech. Importantly, he
argues that deaf children babble the same sounds as normal children, and
cites this as evidence that linguistic experience plays no role in devel-
opment.
Oller (1981) offers the third possible explanation for infant production,
that the infant is intentionally exploring his speech capacities. In this
proposal, the infant is an active participant in the process, and he is sensitive
to the linguistic environment. He gives the following arguments in support
of his position, granting that current research is insufficient to confirm them
at this point in time. First, young infants appear to engage in selective vocal
imitation, that is, they appear to enter into prelinguistic dialogues with
adults. Second, their production of specific types of vocalizations goes
through active cycles that look like practice or vocalplay. For example, the
infant may produce a series of vowels, like [i] [a] [u] as if practicing the
extreme capacities of the vocal tract (Ingram 1981d:lOl). ‘Such a systematic
alternation behavior suggests that the child is literally engaged in practice of
phonetic contrasts and of combinatorial possibilities of an emerging vocalic
system’ (Oller 1981:96).
These arguments concentrate on the claim that the infant is an active
participant in learning. Experience, however, also plays a role. He argues,
therefore, against Locke’s claim that deaf children babble the same as
normal children. Gilbert (1982) has pointed out that this finding, originally
given in Lenneberg, Rebelsky & Nichols (1965), was based on scanty
evidence. In a detailed study of one deaf child, Oller et al. (1985) have found
that the deaf infant’s babbling never became canonical, that is, she never
developed beyond the expansion stage. They conclude that linguistic input
is necessary for the infant to develop fully resonant vowels. If this claim can
be substantiated by further research, it would constitute a major argument
in favor of Oller’s constructionist-like view of development.
Even stronger evidence would exist if we could demonstrate that the
infant’s active involvement in its own speech and that of those around it
leads to differential babbling. That is, cross-linguistic differences in the
babbling of infants would constitute major support of Oller’s model, and
counterevidence to the maturational explanation. To date, attempts to
demonstrate cross-linguistic differences have led to contradictory results.
Locke (1983) concludes from his review of this research that all infants
babble alike. Oller & Eilers (1982) compared the babbling of English and
Spanish infants at 12 months and found no differences. The latter study is
particularly important since these authors were looking for differences to
support their model.
A recent study by de Boysson-Bardies, Sagart & Durand (1984),
however, suggests that the issue is not yet settled. They argue that the
comparison of segmental inventories of infants may not be the best way to
see cross-linguistic differences, given that there is a basic set of segments
across languages. They suggest that linguistic differences may be more
obvious if there is focussing on the acquisition of supralinguistic features
(such as intonational variations). Further, they feel that more attention
needs to be given to selecting for comparison languages that are so different
as to lead to predictable differences between the infants learning them.
They selected French, Arabic, and Cantonese Chinese, since these lan-
guages differ on specific non-segmental parameters: for example, Arabic
differs from French in having (i) an emphatic voice quality, (ii) a more
posterior tongue position, and (iii) more sharp contrasts in pitch. They then
collected samples of babbling from infants exposed to each of the three
languages. The infants were 6,8, and 10 months old. There were 15-second
samples selected from all the samples to be used in the experiment. They
used the recognition technique described in 5.3.2. Adult speakers of French
were played pairs of babbled sequences and asked to identify the French
infants. Their results suggest that the adults could do this for infants at 8
months of age. For example, when paired with samples of Arabic infants,
the French infants were correctly identified, 76 percent for the 8-month-old
samples and 75 percent for the 10-month-old samples. While this is only a
preliminary finding, it suggests that more subtle research may yield cross-
linguistic differences in infant babbling.
5.3.4 The relation of speech perception and production

We saw that two views are currently being proposed concerning speech
perception: there is the universal theory, which claims that the infant can
make all the necessary perceptual distinctions at birth, and the attunement
theory, which grants extensive ability at birth, but proposes that non-core or
peripheral perceptual features require linguistic experience to be acquired.

As for speech production, there is the maturational theory which sees the
cause of speech production as totally determined by biological factors, and
the refinement theory which sees the infant’s active interaction with his
linguistic environment as an important additional factor.
When we compare these positions, there are two natural alignments that
occur, if we assume that any theory in one domain should be logically
consistent with one in another domain. One alignment is what I shall call the
universal-maturationistposition. It proposes that the infant is born with the
basic capacity to perceive and produce speech. The perceptual maturation
is ahead of production, so it is basically complete at birth. Production skills
require the first year of maturational growth. The infant’s role in the process
is minimal, as is the role of experience. The second alignment is that of an
attunement-refinement theory. This theory is constructionist in essence,
seeing the child as an active participant in his development. The child begins
with some basic abilities, then builds upon these as he interacts with the
linguistic environment. Experience will play a role in development, and the
stages that occur will be related to the structure of the earlier stages.
The evidence to date does not clearly support one of these theories over
the other. The key research will be tQ demonstrate the importance of
linguistic experience, as well as to develop a logically consistent theory of
the stages that the infant follows. Some research indicates that experience
plays a role in perceptual development, but we need to know more about
what constitutes the salient and non-salient acoustic parameters of develop-
ment. Also, we have yet to see any proposals of stages of infant speech
perception that look anything like those for speech production. The stages
for the latter proposed by Oller are constructionist stages that are descrip-
tive in the sense of Chapter 4.There are behaviors under change, and the
causes are physiological growth and active exploration by the infant. To
make them explanatory, we will need independent evidence that the child is
an active explorer. Support can come from the speech perception research,
and from what we know of the infant’s cognitive development, a topic we
turn to in the next section. The evidence for the role of experience in speech
production is also only suggestive at this time, at least for the infant’s
babbling. We shall argue in the next chapter that there is evidence that the
cross-linguistic differences predicted by the attunement theory are more
evident in the sound system of the child’s early words.
In Chapter 4 we discussed certain assumptions about how we may
proceed to develop a theory of language acquisition. One of these, the
Constructionist Assumption, can be adjusted for our discussion here. This
assumption states that theories that are constructionist are in the long run
more subject to being falsified than maturational theories. Based on this
assumption, we proposed that we start with such theories and examine their
predictions. The attunement-refinement theory of infant speech develop-
ment makes specific predictions about the role of experience that have been
partially supported by research. Using the Constructionist Assumption, we
will continue to evaluate its viability in the next chapter when we examine
its predictions about the young child’s ability to perceive and produce his
first words.
5.4 Early cognitive development

An aspect of the infant’s prelinguistic development which is no easier to
study than speech is that of cognitive development. By cognitive develop-
ment we mean the infant’s growing knowledge of the world around him. In
order to understand the meanings of the child’s first words and subsequent
sentences it is important to have an awareness of the child’s knowledge of
the world in general. The study of the infant’s cognition is important iI: two
specific ways: to understand what the child means by its words, and to
explore the role that cognitive growth may play in language development.
For example, must certain cognitive developments occur before certain
linguistic ones may occur? Stated differently, does the child’s cognitive
ability limit or constrain semantic development just as early physiological
structure limits the infant’s speech production?
We could discuss early cognitive development as we did speech, by
exploring various theoretical approaches and their predictions for early
behaviors. This, however, could take us far away from our main concern of
language acquisition. Consequently, we will concentrate on one theoretical
approach to early cognition, that of the Swiss psychologist, Jean Piaget.
Piaget’s work on cognitive development constitutes an enormous under-
taking which holds a central place in child development research.
In the early 1900s, Piaget followed the development of his three children,
Jacqueline, Lucienne, and Laurent. His study of his children was in the
spirit of the early baby biographies of his day, as were discussed in
Chapter 2. He reported the results of his observations in three books,
originally published in French and only years later in English. (The dates
given below for Piaget’s books will be those of the English translations.)
The first book, The origins of intelligence in children (1952), presented a
general theory of what he called ‘sensorimotor intelligence’, his term for the
child’s cognitive ability during the first two years of life. In it, he outlined the
characteristics of six stages of sensorimotor development. This was followed
by The construction of reality in the child (1954). In this work Piaget
presents his observations on four areas of cognitive development: the object
concept, the spatial field, causality, and time. He shows how the child’s
knowledge of each changes through the six sensorimotor stages. The third
book is Play, dreams and imitation in childhood (1948) which is the most
important of the three for the study of language acquisition - its original
French title, La formation du symbole, is more directly indicative of this.
The book follows three major developments that lead to the ability to
have symbols (and hence language): imitation, play, and mental repre-
sentation.
Piaget’s work is important for the study of language acquisition for three
reasons in particular. First, it is an impressive body offacts about the child’s
early development. Like other baby biographies, his observations tell us a
great deal about what infants do. Second, his work is important because he
goes beyond description and attempts to explain early development. In
particular, he developed a constructionist view of development that has
been the model for several subsequent theories of language acquisition.
Here it is important to emphasize that we are talking about theform of the
theory, not its content. By form I mean that stages are actively constructed
by the child, building upon earlier structures, and influenced by the
environment. Third, Piaget’s theory is important for its content. That is,
Piaget’s theory offers one explanation of language acquisition, proposing
that it is the result of general symbolic growth. His claims about the relation
of language to cognition have led to a larger body of studies in recent years.
We will eventually look at all three of these aspects of Piaget’s work.
Here, however, I will restrict myself to a general introduction to his theory,
and a brief description of his observations on cognitive development before
the first words.
5.4.1 General aspects of Piaget’s theory

When Piaget talks about infants, he uses the term sensorimotor intelligence.
The infant does not have the ability to represent objects, that is, to recall
them in their absence. Rather, the infant’s knowledge is restricted to the
ways his senses have responded to objects or to his own manipulations of
them. As Anisfeld (1984: 15) has described this: ‘the knowledge that young
infants have of objects is in terms of the sensorimotor impressions the
objects have on them and the sensory and motor adjustments the objects
require’. The infant will not begin to have concepts as adults do until the
second year of life, around the time language begins. Until then, he only
knows objects primarily through his actions on them. The young infant does
not have abstract concepts but sensorimotor schemata, which are the action
patterns associated with objects. The development of the first year is a
gradual move away from schemes to eventual abstract classification.
The infant begins knowledge with basic reflexes and two functional
processes: assimilation and accommodation. Assimilation occurs when the
infant processes incoming events by adjusting those events to its internal
structures. An example from speech perception would be to perceive a
sound that we are not familiar with to be the same as one with which we are
familiar. We assimilate it to, or treat it the same as, a unit in our internal
system. Also, assimilation will allow the infant to realize a familiar sound or
event. Accommodation is the opposite process. In accommodation the
infant changes the internal structure to become more like the external
event. For example, the Spanish infant who learns to perceive a difference
between [b] and [ I4has accommodated his perceptual system to the
linguistic input he learns. Development, then, is seen as the constant
interaction of assimilation and accommodation to the sensorimotor input.
Once that change becomes part of our system, we assimilate other events to
it. For example, the child who acquires the English past tense ‘-ed’ has
accommodated to the linguistic input. When he proceeds to say ‘eated’
and ‘breaked’, he is assimilating the past tense of these verbs to his new
structure. Only after a period of assimilation or exercise is the new structure
sufficiently established to begin a new accommodation. Eventually these
two processes lead to a third one, organization. These processes allow the
child to ‘construct’ its reality, hence the term ‘constructionism’.
Piaget’s theory emphasizes the innate nature of the three processes of
assimilation, accommodation and organization, and how they build upon
innate reflexes to construct knowledge. It is in this emphasis on the
processes that he differs from stronger nativist or maturationist positions.
Maturationists have argued that there is no difference in the positions, in
that what has the potential to exist must be said to be innate. Piaget (1971:
16) has conceded this point to a degree, but argues that one still has to
account for how the collaboration between the innate potential and the
environment actually works. This point was described in somewhat differ-
ent terms in Chapter 4. The above processes will prove useful when later
other milestones of language acquisition are considered.
5.4.2 The first four stages of sensorimotor development

Piaget’s six stages of sensorimotor development are proposed to cover
development from birth to around 18months to 2 years. (There is a problem
in determining the end of the period in relation to the onset of language, and
that will be returned to in the next chapter.) For the period we are
examining, that is, birth up to the first word understood, only the first four
stages are relevant. The description of the fifth and sixth stages will be
considered in the next chapter.
The first stage is characterized by the appearance of reflexes. These
provide the basis upon which subsequent developments are built. In speech
perception, the reflex would be the innate ability activated by the initial
exposure to sound. In speech production, these are the vegetative sounds
associated with the process of eating discussed by Stark, Rose & McLagan
(1975). Most of Piaget’s actual discussion focusses on reflexes of sucking
and vision.
The first adaptations or accommodations appear in sensorimotor stage 2.
The following selection from Piaget’s observation 40 in Piaget (195278) on
his daughter Jacqueline discusses this for crying behavior:
Around 0;l it seems as though crying stops simply expressing hunger
or physical discomfort (espectially intestinal pains) to become slightly
differentiated. The cries cease, for example, when the child is taken
out of the crib and resume more vigorously when he is set down for a
moment before the meal. Or again, real cries of rage may be observed
if the feeding is interrupted. It seems evident, in these two examples,
that crying is connected with behavior patterns of expectation and
disappointment which imply acquired adaptation.
Piaget describes at this stage the appearance of the primary circular
reaction. This term refers to basic repetitive behavior by the child to
exercise newly acquired behaviors. Observation 40 also points out some
instances of these regarding crying and babbling (p. 78):
It is then that the first ‘circular reactions’ related to phonation may
definitely be observed. Sometimes, for instance, the wail which
precedes or prolongs the crying is kept up for its own sake because it is
an interesting sound: 0;1(22) . . . At 0;2(15) the crying is transformed
into playing with the voice, ‘aha’, ‘ahi’, etc. At 0;2(15) she even
interrupts her meal to resume her babbling.
This stage coincides with the GOO stage in Oller’s framework (Table 5.7).
Circular reactions are also referred to as ‘reproductive’ assimilations, that
is, the reproduction or repetition of the new structure. These are com-
plementary with recognitory assimilation, that is, the recognition of a
known or acquired schema.
The primary circular reactions of stage 2 are activities of the infant that
focus on his body, such as thumb-sucking, babbling, grasping things to
exercise the grasping activity, or shaking his arms or body. In all of these
activities, the end or goal of the activity is the same as the means to achieve
it. An initial separation between ends and means occurs in stage 3 with the
onset of the secondary circular reaction (around age 0;5). These activities
are secondary in that the action of the infant’s body is used as a means which
achieves a result outside of the activity itself. The infant’s shaking his arms,
8
for example, may cause the mobile attached to the crib to move. The infant
repeats the body act in order to see the outside event once more. Piaget
states (1952:154): ‘after reproducing the interesting results discovered by
chance on his own body, the child tries sooner or later to conserve also those
which his action bears on the external environment’. Because of this
preliminary separation of ends and means, Piaget emphasizes that these are
the first instances of intentional behavior by the infant. He acknowledges
the difficulty of defining intention, and emphasizes that these early behav-
iors are still far from being like the intentions of the adult. He states
(p. 148): ‘intentional adaptation begins as soon as the child transcends the
level of simple corporeal activities (sucking, listening and making sounds,
looking and grasping) and acts upon things and uses the interrelationships
of objects’.
Recall that the infant does not have concepts at this time, but rather
sensorimotor schemata. Primary circular reactions help establish primary
schemata (e.g. grasping, shaking, vocalizing, etc.). The secondary circular
reactions lead to secondary schemata, which are sensorimotor impressions
of the properties of objects. These are the precursors to later classes or
concepts: ‘The secondary schemata constitute the first outline of what will
become “cla~ses’~ or concepts in reflective intelligence: perceiving an object
as being something “to shake”, “to rub”, etc.’ (Piaget 1952:183). These
schemata will become even more complex in the next stage. We see at stage
3 the onset of possible intentional behavior and the first roots of later
classification.
During stage 3, the infant’s use of a circular reaction is to prolong an
event which he discovered or caused by chance. These acts consequently
only involve the continuation of a primary circular reaction (e.g. shaking).
In stage 4, however, the distinction between the ends and means becomes
more pronounced. The infant appears to want to cause some activity, and
will try out his available circular reaction to accomplish this. Suppose, for
example, that the infant wants to grasp an object, but that object is not
accessible because of some obstacle - let’s say the infant wants to grasp a
box which the parent is holding. The infant can use one of its other
schemata, such as pushing objects, or striking objects, to achieve the end,
which is to grasp the box. Importantly, the infant will not attempt to
discover new means; this will not occur until the next stage.
Piaget has used these stages to follow the development of the major
conceptual domains of intelligence. In his second book (1954), he presents
his findings on the development of the object concept, spatial displacement
of objects, causality, and time. This was followed up in his third book (1948)
with a treatment of imitation and play behavior. His observations on the
major behaviors in these areas through the sensorimotor stages constitute
an important database on the infant’s cognitive development. Table 5.9
presents a summary of some of these behaviors for stages 3 and 4. These
observations give a general picture of the infant around the time of the first
words.
The infant’s knowledge of objects is becoming advanced by the end of
sensorimotor stage 4 (or the end of the first year). The infant indicates that
he recognizes people and objects, and will explore the properties of new
objects. He even shows an awareness that certain events anticipate others,
for example, that standing up indicates someone is about to leave. The
coordination of actions, e.g. removing an obstacle to be able to hold an
object, provides the infant with an early form of causality.
The area of object permanence is probably the most known area of
Piaget’s observations on infants. The infant only gradually indicates
awareness that an object exists when it is no longer in sight. In stage 4,the
infant for the first time will actively search for an object that is suddenly
removed from its sight. This development of object permanence along with
the growing awareness of objects appears to occur just before the advent of
word comprehension.
Lastly, imitative ability is also increasing. In stage 3, the imitation is
limited to acts already done previously by the infant. Piaget states that the
stage 3 infant will only imitate sounds that he recognizes as ones that he has
already produced (or babbled). Attempts at novel sounds do not appear
until stage 4.The latter, of course, will be essential for the acquisition of the
adult language.
5.4.3 The development of meaning

We might expect the discussion of meaning to wait until the infant begins to
acquire words. It is crucial in Piaget’s theory, however, for later develop-
ments to have their roots in earlier developments. This is true for the growth
of the awareness of meaning, where Piaget outlines prelinguistic develop-
ments that prepare the infant for word acquisition.
We can simplify the infant’s task by emphasizing two major goals: to
develop the notion of ‘word’, and to acquire the meanings of the words in
the native language. Logically, the former development necessarily pre-
cedes the latter. In our preliminary discussion of the acquisition of meaning,
we will concentrate on the first of these two. That is, how does the infant
arrive at the realization that vocalizations may be meaningful?
To begin, Piaget discusses the notion of the linguistic sign, proposed
earlier by the French linguist Ferdinand de Saussure. The linguistic sign is
the result of the sign$er-sign$ed relation where the signifier is some group
of sounds and the signified is the concept these sounds stand for or
Table 5.9 Some behaviors of young infants which are typical during stages 3 and 4 of sensorimotor development
from selected areas of cognition, selected from appendix in lngram (1978)
Cognitive domain Stage 3 Stage 4
Knowledge of objects reproductive assimilation (secondary circular coordination of secondary schemas:

reactions): removal of objects that are obstacles, e.g. moving
shaking objects another's hand to grasp object
stacking objects with hand intentional rejection of object
hitting two objects together prevision
putting object in mouth recognizes that someone who gets up plans to leave
recognitory assimilation active exploration for properties of new objects
fear of strangers visual and tactile exploration
recognizes footstep, door opening, bottle
knows presence of someone when touched
Object permanence attempts to follow trajectory of a fallen object for first time, will search for an object that is hidden
recognizes object by seeing only part of it. from him, e.g. under a blanket
Spatial field has interrupted prehension, i.e. will grasp for object has reversible operations, i.e. hands objects back to
that has been dropped or taken away adult
lacks ability to follow trajectory of an object or to studies the constant shape of objects by moving them
rotate objects to and from eyes
Imitation systematic imitation of sounds that he has already imitation of new sounds
made imitation of non-visible movements already made by
systematic imitation of movements seen by infant, e.g. child, e.g. opening and closing of mouth, putting out
clapping hands, opening and closing of hands, waving tongue
bye-bye imitation (with effort) of new visible models, e.g.
bending and straightening finger
represent. For example, the English word ‘tree’ is a phonetic string [tri]
paired with the concept TREE that we recall when we hear these sounds. (I’ll
indicate the concepts referred to, as distinguished from the word, in
capitals, e.g. TREE vs. ‘tree’.) The linguistic sign has two important prop-
erties in particular. One is that the relation is arbitrary. There is nothing
about the signified that requires a particular signifier. The French sound
[larbr] for ‘l’arbre’, for example, does just as well as the English sounds to
represent the concept TREE. Second, the concepts are abstract represen-
tations. We do not see the tree to understand the word, and some of the
concepts can be quite abstract, e.g. LOVE. The young child, then, needs to
develop the signified-signifier relation, abstract concepts, and the arbitrary
nature of the relation. (Additional developments include the ability to
recognize the perceptual form of the adult words.)
Piaget argues that the signifier-signified relation of the linguistic sign has
its roots in more primitive relations of signifiers and signifieds of the
sensorimotor period. Piaget (1952: 185-96) proposes steps that the signifier-
signified relation goes through from birth to the time when the linguistic sign
is developed. The first three sensorimotor stages are each characterized by
their own type of signal or elementary indication. Stage 4 is marked by the
appearance of the indication (or index, Piaget 1948; the translations for this
term vary across Piaget’s books). Stage 5 shows the advent of symbols.
Arbitrary signs do not occur until the end of the sensorimotor period. A
discussion of the emergence of symbols and signs will be postponed until the
next chapter. Here, we will briefly look at the development of signals and
indices (or indications).
It is important to realize that the signifier-signified relation can apply to
sensory impressions. To use one of Piaget’s examples, we may observe a
tree while sitting in a park. We may not see all of the tree and certainly we
do not observe it from all angles or perspectives, yet we know that we are
watching a tree. The signified, then, is TREE, the intellectual concept, not the
sensory attributes. The latter are the signifiers that allow us to recognize the
concept that represents the object. Piaget states (1952:190):
The signified of a perception - that is to say, the object itself - is
therefore essentially intellectual . . . In order to perceive their indi-
vidual realities as real objects it is essential to complete what one sees
by what one knows. Concerning the ‘signifier’, it is nothing other than
the few perceptive qualities recorded simultaneously and at the
present time by my sensory organs. . .
In the infant, we can not yet speak of concepts, but instead of sensorimotor
schemata. Nonetheless, we can describe the relation between the infant’s
perceptions and the early schemata as one of signifier-signified. These are
what Piaget means by ‘indications’ (of which ‘elementary’ indications are a
subcategory): ‘In a general way we shall call indication every sensory
impression or directly perceived quality whose signification (the ‘signified’)
is an object or a sensorimotor schema’ (Piaget 1952:191).
Elementary indications characterize the first three stages of sensorimotor
development. Stage 1 elementary signification refers to the connection
between a reflex act (e.g. sucking) and the sensory feeling the infant has
while exercising it: ‘In such a case the signifier is the elementary sensory
impression accompanying the play of the reflex . . . and the signified is the
sucking schema’ (Piaget 1952: 192). In reflexive crying, the signifier would
be the motor impression of crying as well as its auditory impression. The
signified is the crying schema itself.
The elementary indication of stage 2 is not far removed from that of
stage 1. At stage 2, the first habits are being acquired. Sensorimotor
impressions associated with a schema will set the schema in motion. For
example, being placed in the position for nursing may be a signifier to the
infant to begin to suck: ‘The consciousness of the position for nursing does
not signify anything more . . . than the awaiting and the beginning of the
sensory images connected with sucking’ (p. 193). In such cases the relation
is not arbitrary, as it will be later for the linguistic sign. Rather, the signifier
is one aspect of the real-world events associated with the schemata.
Another example is the mother’s voice as a signifier of the presence of the
mother.
The stage 3 elementary indication is the last of the three types of
elementary indications. It is the one which is part of the infant’s secondary
circular reactions. Just as the secondary circular reaction entails an initial
development of the infant’s own actions for their own sake toward their use
to cause other events, so too the signifier will now suggest a series of
events:
In effect, the significations of the second type remain essentially
functional and related to the subject’s own activity. That which the
sensory signals announce is that a certain thing is to be seen, heard,
grasped, etc. On the contrary, the significations of this third type
comprise from the beginning an element of foresight related to the
things themselves: the string hanging from the bassinet is not only to
be seen, grasped, and pulled, it serves to swing objects from a
distance, etc. (1952: 194)
An important restriction of this new foresight is that it is limited to activities
of objects or people tied to the infant’s activities; that is, they are always
part of the infant’s secondary circular reactions.
Here are some examples of stage 3 elementary indications (or signals):
At 0;7(10) he [Laurent: DI] cries in the morning as soon as he hears his
mother’s bed creak. Until then, although awake, he did not show his
hunger. But at the slightest creak, he moans and thus demands his
bottle. . .
At 0;8(3) Jacqueline smiles and says aa as soon as the door to her room
opens, before seeing the person who enters. She therefore under-
stands by this sign that someone will enter. (1952:195)
Piaget refers to the significations (or signals) of the first three stages as
primary. He reserves the term ‘indication’ (or ‘index’ in the translation of
Piaget 1948) for stage 4. Remember that in stage 4 the infant for the first
time coordinates schemata with separate objects. As such, the separation
between means and ends becomes more pronounced:
. . . a fourth type of sign is now constituted . . . which permits the child
to foresee, not only an event connected with his action, but also any
event conceived as being independent and connected with the activity
of the object. (1952:248)
He calls this prevision,meaning that the child can anticipate the next event
which is separate from the one in progress.
Table 5.9 gives one example of prevision, the recognition that the act of
standing up indicates another action, which is the act of leaving. Jacque-
line, for example, at 0;9(15) would cry if the person next to her moved or
stood up, since this gave the impression of leaving (p. 249). Piaget states
(p. 251): ‘prevision becomes possible in connection with facts rarely or
very recently observed, or even in connection with the actions of other
people. Thus to foresee the departure of a person when he rises or even
turns away is a prevision which is already very much detached from the
action in progress.’
By the end of sensorimotor stage 4, around age 1 year, the infant has
made some noteworthy gains. He has acquired a variety of sensorimotor
schemata, auditory as well as visual and motor, in a variety of areas.
There is an impressive growing knowledge about actions and objects, with
increasing ability to operate upon people and objects. Importantly for
language, activity is becoming more clearly intentional, as the gradual
separation between means and ends occurs. Vocalizations will become in
the next stages one of the means to achieve ends, just as secondary
schemata are at stage 4. In addition, the separation between signifiers and
their signifieds is occurring. This separation will eventually lead to the
acquisition of the arbitrary linguistic signs characteristic of human
language.
5.4.4 Piaget: pro and con

Piaget’s work stands out as the major effort to date on cognitive develop-
ment in both the areas of description and explanation. As stated earlier, the
work is substantial even if our only interest is to obtain an extensive body of
facts on what infants and children do. A small sampling of these facts has
been presented above. In addition, it provides a theoretical statement of
tremendous breadth on the cause of development. Indeed, his theory has
had such an impact that its vocabulary characterizes much of the discussion
in child development.
Both of the above observations can be used to justify the choice here of
Piaget’s theory as a vehicle for discussing cognitive development in the first
year of life. There is, however, a more substantive reason for its selection:
his work in a very general way provides the model for the approach taken in
this text toward language acquisition, that is, his research on development is
characterized by the constant interaction of method, description, and
explanation, with an awareness of the importance of each.
Piaget’s method is usually referred to in his own terminology as a clinical
method. It is defined as an approach where there is experimentation, in that
the experimenter has a hypothesis to pursue, and may vary conditions to
test it, but there is also careful observation of the child’s actions in the
context of the child’s total ability or mental states. One is constantly
attempting to understand the nature of the child’s behavior. Here the
important point is that Piaget was always discussing the importance of
method and what the best method might be to study a particular point. For
example, as early as Piaget (1929) he made the point that the study of
children is a difficult enterprise that requires extensive experience. He
states (pp. 8-9): ‘it is our opinion that in child psychology as in pathological
psychology, at least a year of daily practice is necessary before passing
beyond the inevitable fumbling stage of the beginner’. His is a far cry from
the rather casual treatment of method discussed in 4.3, when Chomsky
(1975) concludes that children do not produce a particular construction
because he has never heard them produce it.
A second characteristic of Piaget’s work is its descriptive force: he has
provided a plethora of facts to be explained about cognitive development.
This is the obvious consequence of a concern to work extensively with data.
The observations cited in his three books on sensorimotor development are
striking in the careful detail taken. They probably represent the most
careful diary ever done. While these books followed three children in detail,
others show his awareness of the occasional need to verify results with larger
populations: Inhelder & Piaget (1964), for example, presents results from
2,159 children. His appreciation of the need to work closely with children’s
data reflects the general orientation of the Child Language field. For
example, in Piaget (195520) he states: ‘for teachers and all those whose
work calls for an exact knowledge of the child’s mind, facts take precedence
over theory’.
Despite the flavor of this last quotation, it is nonetheless obvious that
Piaget was very much committed to the development of a core explanation
of child development. Indeed, it is his theory, rather than his method and
descriptive work, that has proven the central focus of both supporters and
critics. The assumptions I presented in Chapter 4 in regard to the explor-
ation of language acquisition are very heavily influenced by specific features
of Piaget’s theory. Most central is the Constructionist Assumption, which
was made with the hope it could yield a developmental theory for language
as extensive as Piaget’s for cognitive development. Piaget’s constructionism
is a position that deserves careful consideration. It is one that has been
presented in depth in several places, especially in Piaget (1971). It is also a
position, unfortunately, that has not even been considered by those who
approach language from a maturationist perspective. Recent books by
Lightfoot (1982), Chomsky (1975) and Jackendoff (1983), for example,
treat issues that have been discussed by Piaget at great length - none,
however, contains a single reference to Piaget. To discuss cognition without
a serious treatment of Piaget is as fallacious as attempts to study modern
syntactic theory without ever reading Chomsky. The Competence Assump-
tion is also implicit in all of Piaget’s work. He consistently only attempts to
assign to the child knowledge that can be ascertained from his behavior. The
Productivity Assumption is also implicit, in that Piaget is very cautious in
assigning a rule or schema to the child until there is consistent evidence for it
in the child’s behavior. This concern for productivity is a permanent part of
the clinical method.
When discussing Piaget’s work, it is also necessary to mention the place of
this work from one’s own viewpoint. Piaget has managed to create very
strong reactions to his theory. There are those who support it carte blanche,
and others who attack it vehemently. It is important for the student to be
aware of this and to weigh writings by each group accordingly. In this book,
Piaget’s work plays the following roles: (i) it provides the model for our
general orientation to language acquisition (in the sense of providing an
integration of method, description, and explanation); (ii) the form of his
theory leads to three assumptions that have been proposed here in
Chapter 4 for the study of child language acquisition; and (iii) his discussion
of the emergence and structure of meaning will be used here as a
demonstration of a constructionist attempt to account for the emergence of
meaning. This volume is not, however, an attempt to present and defend
Piaget’s view of the relation of cognition and language. In fact, as will be
discussed in Chapter 7, it will present a position more in support of
autonomous syntax.
Before concluding this section, it is important to present the major
criticism of Piaget’s description of sensorimotor development as presented
here. This concerns his description of general stages of sensorimotor
development. In probably the first extensive experimental attempt to
replicate Piaget, Uzgiris & Hunt (1975) found that infants did follow the
invariant sequences of behaviors found by Piaget, but not in well-defined
general stages. For example, the infant would show the stage 4 behaviors of
object permanence in the right sequence to other object permanence
behaviors, but he would not necessarily also show stage 4 behaviors in
causality. Their work, as well as that of several others, has led many investi-
gators to reject the idea of general sensorimotor stages. This is coupled with
the problem, pointed out by Brainerd (1978), that Piaget never explained
how to place a child into a general sensorimotor stage.
This issue is important for our study of language acquisition when we turn
to the relation between language and cognition. If there are general stages,
then we can see if those stages correlate with specific linguistic milestones. If
not, we need to concentrate exclusively on the relation of specific domaitls
to language, such as imitation, causality, etc. The general criticism above, in
fact, has led to work of the latter kind. It is my impression, however, that
this general critique is based on a misunderstanding of Piaget’s view on how
general stages are attained. The following statement of Piaget, in relation to
Laurent’s application of familiar schemata to new situations, suggests that
the attainment of a sensorimotor stage is no simple matter:
. . . it is perfectly normal that these first behavior patterns of the fourth
stage are constituted sporadically from the middle of the third stage
except that these episodic productions are only systematized and
consolidated one or two months later. (1952:214)
That is, there is a spiralling effect that is not going to be evident from strictly
experimental results like those in Uzgiris & Hunt (1975). Because of this, I
have maintained the original Piagetian descriptions in the text above.
5.5 The linguistic environment

So far, we have considered the infant from the point of view of the
development of its internal abilities. The infant, of course, does not acquire
language in a vacuum. In order to appreciate the language acquisition
process, it is also necessary to look at the infant’s interaction with the
linguistic environment. We need to examine the nature of the language
presented to the infant, or the linguistic input, and the characteristics of
adult-infant interaction.
5.5.1 The cultural influence

To begin, we need to be aware of the fact that cultures vary from each other
in their view of the infant and the way he should be treated. These cultural
attitudes can influence the nature of infant-adult interaction and thus the
language heard by the infant. Much of the work done in recent years on how
adults speak to infants has been on English-learning infants. In our culture,
to oversimplify, an infant is seen as someone to interact with actively, and
thus he can receive a great deal of linguistic input. Further, it is common
to see parents actively involved in linguistic stimulation of their infants as if
teaching them. This attitude is captured very nicely in the following excerpt
from Piaget (1971:19-20):
. . . very rarely have I been able, in America, to expound any aspect of
my stage theory without being asked, ‘How can you speed up their
development?’ And that excellent psychologist, J. Bruner, has gone as
far as to state that you can teach anything to any child at any age if you
go about it the right way.
We have to be careful not to generalize too quickly from our own cultural
biases to universal claims about how parents speak to their infants.
Pye (1983b) has recently discussed cultural attitudes that may exist
towards the young infant. Table 5.10 gives a brief summary of examples he
has found. The bulk of Pye’s discussion, however, concentrates on the
attitudes of the Quiche, a language group located in the western highland
region of Guatemala. The Quiche infant is seen as being born with a soul
that must be protected from being lost. The infant is treated with great care
so that the soul does not leave the body. Two of the possible ways to lose
one’s soul are excessive crying and falling down. Such views then lead to
expectations in the parents as to what constitutes appropriate infant
behavior.
In the QuichC case, the parental attitude is to ignore the infant until it is
producing recognizable adult-like language, as if addressing the infant
would jeopardize its soul. Pye states that Quiche parents direct very little
speech to their babies, and only begin to engage in conversational-style
interaction when their children are around 18 months to 2 years. Even then,
the parents continue to protect their child from outsiders. Pye found that
the mothers would even attempt to speak for their children when addressed
by an outsider. For example, one case he observed was when Pye’s assistant
said to a Quiche child ‘Do you know what you are holding?’ Before the child
Table 5.10 Some examples of cultural attitudes toward infants, based on
the summary in Pye (1983b)
Language
group Cultural attitude Reference
Mundugumor Infants are seen as a threat, causing loathing Mead (1963)

and jealousy; they are consequently kept in
isolation.
Mohave The fetus is born capable of understanding the Devereux (1949)
adult language.
Japanese Infants are extensions of the parents; their Caudill & Weinstein (1969)
needs are obvious and do not require extensive
interpersonal communication.
Samoan Rules of etiquette restrict parents from Ochs (1982b)
speaking to young infants.
Javanese Infants are not yet human. Goertz (1973)
could answer, the mother responded (translations are mine): ‘He says
“girl”.’ To child: ’You say “little horsie” ’ (2 times). To experimenter: ‘Play.
He says “little horsie plays”.’ In a later paper, Pye (1986) has found similar
‘to say’ routines in other cultures, but with different purposes. For example,
they occur in Kaluli (Schiefflin 1979) as teaching devices for developing
language.
Another effect of the QuichC attitude toward their infants is that the
culture has no developed linguistic routines for interacting with infants. Pye
says (1983b:18): ‘QuichC parents spend their time working, not entertaining
their children. I did not observe (nor could I elicit) any traditional games or
songs which parents engaged in with their young children.’ This is in marked
contrast with English parents, for whom such routines abound. Several of
these routines have been subjected to careful study:
give-and-take games (Ratner & Bruner 1978)
peekaboo (Bruner & Sherwood 1976)
book reading (Ninio & Bruner 1978; Snow & Goldfeld 1983)
Given this contrast in cultural attitudes, it is striking that the studies just
cited seek to explain the onset of later language milestones by the
occurrence of these play routines.
In the rest of this section, I will be reviewing data primarily on the
language addressed to English infants. At this point, however, we should be
reminded that the results are for English-learning infants. Thus, generali-
zations will need to await more careful studies in other cultures.
130 THE PERIOD OF PRELINGVISTIC DEVELOPMENT
5.5.2 Possible effects of the infant’s linguistic environment

We could, of course, be content with describing how parents speak to
children. This is, however, only one of the reasons for looking at this issue.
The major concern is to determine the possible effects of the linguistic
environment on the child. That is, to what extent can we say that the
environment has caused a particular milestone for the child? This issue
brings us back to the theoretical possibilities discussed in Chapter 4;here I
want to try to separate the predictions each makes.
The behaviorist view, with its emphasis on the role of the environment
and incremental learning, would expect a tremendous influence from
parental interaction with the child, beginning with the period of prelinguis-
tic development. In fact, the work on early infant-adult interaction by
Bruner (e.g. Bruner 1975; Ninio & Bruner 1978) is very behavioristic. It
sees the parental interaction with the infant as the basis of later linguistic
developments. The cross-cultural facts described above, however, create
some difficulties for this position, though it could be maintained by moving
the environmental influences to a later stage.
The maturational view places much less importance on input, especially
for the prelinguistic child. In this period, the child would need to hear
language, but the parent would not need to be involved in elaborate
simplifications of his or her own language. This view predicts a range of
possibilities, from no speech adjustments to children, to elaborate ones due
to cultural factors. The crucial point would be that parental speech
adjustments, should they occur, would not be done to teach language. The
possible cultural effects discussed above are not a problem for the matur-
ationist view, and could even be cited as some support for it. For example,
Quiche children do eventually begin to talk, despite what appears to be a
minimal amount of input.
The constructionist position is difficult to separate from the maturationist
one in this regard. In Piagetian work, the child is seen as developing his
cognitive capacities through interaction with the environment. Nowhere,
however, does Piaget propose that the environment needs to adjust itself to
the child. In fact, it is the child who needs to be flexible in order to deal with
a changing environment (Piaget 1971). In this sense, then, the child needs to
be able to acquire language within a range of varying environments. There
seems to be very little difference between the maturational and construc-
tionist views regarding the kind of language addressed to children.
Why, then, study the infant’s linguistic input, given the nativist orienta-
tion of the text? The reasons are two-fold. First, as recorded in Chapter 4,
nativism, either in the form of Chomsky’s or Piaget’s theory, has never
denied some role in acquisition to behaviorist principles. It is important to
know the extent to which parental reinforcement and shaping influences
language acquisition. The second reason is the one expressed in section 4.5:
there may be differences in the way children differ within and across
languages in their language acquisition that can be traced back to the
environment, not to the child’s internal programming. If we are going to
understand the nature of the child’s developing linguistic competence, we
need to separate out the non-linguistic factors that lead to variation. Our
Quiche discussion can provide a trivial example. Based on the ages in Pye
(1983a), QuichC children appear to acquire language slower than their
English counterparts. The fact that Quiche parents direct relatively little
speech to their children can account for this cross-linguistic variation.
5.5.3 Baby talk in English

For reference purposes, the language addressed to children is often called
motherese or baby talk. Neither term is particularly helpful. Motherese is
not adequate, since the infant is also addressed by fathers, other adults, and
children. Baby talk is also inadequate because it is used in the literature in a
negative sense as a form of language which uses a restricted set of features
such as changing [r] to [w], e.g. ‘wabbit’ for ‘rabbit’. Here I will use the
following broad definition of baby talk:
Baby talk: the language used by anyone in the linguistic community
when addressing a child
Thus, even if an adult spoke to a child in the same way as to an adult, we
would call this language baby talk. The set of features that distinguish the
person’s baby talk from non-baby talk in this case would be zero. Also, baby
talk as defined may vary from speaker to speaker and language to language.
The purpose initially will be to see if individuals speak differently to
children and, if so, what the distinguishing features are. These can then be
examined for their possible effects on the child’s language. This section will
begin the inquiry into baby talk by looking at how adults address infants.
There will be further discussion in subsequent chapters.
While there are numerous studies of language addressed to English
children, there are relatively few on the language addressed to infants. The
studies on baby talk to children have found features that are characteristic
of it. Kaye (1980a) divides these features into five general categories,
summarized here in Table 5.11. While the details behind them will be given
in subsequent chapters, they are presented here in order to provide an
initial general picture of what English baby talk looks like.
Here we will look at the language used by English parents to address their
prelinguistic infants. One possibility is that they do not talk to their infants -
Table 5.11 Five kinds of characteristics of English baby talk, taken from
Kaye (1980a: 489-90)“
Characteristics Specific examples and references
1. prosodic features higher pitch, greater range of frequencies, more-varied

intonation (Garnica 1977; Sachs 1977)
2. lexical features special forms like potty and nana (Ferguson 1964)
3. complexity features shorter utterances, fewer embedded clauses, fewer verb
auxiliaries, etc. (Snow 1977a; Furrow, Nelson & Benedict 1979)
4. redundancy features more immediate repetition and more repetition of the same words
or phrases over a period of time (Snow 1977a)
5 . content features restriction to topics in the child’s world (Snow 1977b)
a This is a direct quotation from Kaye (1980a: 489-90) reorganized into the form of a table.
as Pye has claimed for QuichC parents. As we will see, English parents have
a very different view of their infant from QuichC parents. As you might
expect, English mothers (the parents in the studies to be discussed) do talk
quite a lot to their infants. That being so, the next question is the purpose of
this language. One possibility is that they are directly teaching them in the
sense of behaviorism, as discussed in section 2.2.3 in relation to Bloomfield.
Or, they could be involved in indirect teaching by presenting a model of
how the infants should talk. If so, we would expect that parents’ language
should be quite simple during the period of prelinguistic development,
possibly dominated by single-word utterances. This would be supportive
evidence of a claim that the child’s single-word utterances follow from such
modeling. Still another possibility is that parents reserve attempts to teach
children language until they show some evidence of acquiring language
during the period of single-word utterances. If this is so, then the language
used to the prelinguistic infant may look quite different from that used later.
To study these possibilities requires observing parents addressing chil-
dren at two time periods and comparing their language at each. Here we will
review two studies which have used this methodology: Snow (1977b) and
Kaye (1980a).
Snow (1977b) Snow conducted her study with the hypothesis that English
baby talk would begin to show features like those in Table 5.11 when
children were around 1year and beginning to use language. For example, it
was predicted that tutorial types of questions like ‘What’s that?’ or requests
would not be used before that time since they would serve no teaching
purpose.
To study this hypothesis, Snow collected longitudinal samples using
video- and audio-tapes from two mother-child dyads. The infants, Ann and
Mary, were firstborn daughters of middle-class English parents. The tapes
were 20 minutes long, and included a feeding session and a play session. The
ages of the samples were as follows (in weeks):
1st year (in quarters) 2nd year (in quarters)

1st 2nd 3rd 4th 1st 2nd 3rd
Ann: 12 (none) 29 49 (none) (none) 79,87

Mary: 13 22 29 38,44,52 (none) 75 81
As can be seen, the samples are not exactly comparable, but they are
sufficiently close.
Snow’s quantitative analysis applied several measures to the mother’s
language in each of the above sessions. The main measures were MLU and
percentage of interrogative sentences. The results of both of these are
presented in figures, so no specific numbers are available. Neither figure
showed any abrupt changes during the second year of life. For both
mothers, MLU remained constantly around 4.0 throughout the samples.
This short MLU indicates that the complexity features of Table 5.11 are
probably part of baby talk even during the period of prelinguistic develop-
ment. The occurrence of interrogatives was also relatively constant, around
30 percent. Interestingly, the most frequent occurrence of interrogatives
was during the first session where these were around 40 percent for Mary’s
mother, and 60 percent for Ann’s. Snow concludes from this that the data
do not support the hypothesis that the features of baby talk in English are
used in response to the child’s needs for a simplified model of language.
There were other measures applied by Snow which did show change over
the study period. These were percentages of declaratives, imperatives, and
contentless utterances (for example, verses, songs, sound-play). Table 5.12
gives these percentages for those times when samples were available for
both girls. At first glance, these appear to support the hypothesis that baby
talk changes around 1 year. For example, we would expect a mother to use
more contentless utterances during the period of prelinguistic development
if she felt that the infant was not learning language at that time. Or, we
would expect the increase in imperatives when the child becomes able to
understand and follow them. These data, then, appear to provide some
support for the claim that English parents adjust their speech as their infants
begin to acquire language. Unfortunately, the data also lend themselves to
other interpretations. For example, we could claim that they show that the
Table 5.12 Percentages“ of Ann’s and Mary’s mothers’ utterances

that were imperatives, declaratives, or contentless utterances, at
four sessions, extracted from Snow (1977b: table 2)
Age in weeks
Type of utterance 12-13 29 49-52 79-81
Imperatives
Ann 10 6 25 23
Mary 15 7 28 5b
Declaratives
Ann 9 22 29 35
Mary 19 13 29 38
Contentless
Ann 22 38 26 16
Mary 26 32 18 17
a Percentages are rounded off.

This is a rather mysterious figure in light of the other data.
adult adjusts in order to provide a simplified model for language learning.

Or, they could show that the adult changes in response to the child’s level as
allowed by restrictions of the grammar of English. This adjustment would
be solely for the purpose of communication. In all the possible interpreta-
tions of the data, the independent evidence is usually lacking.
Snow’s explanation of her data is neither of the above. Instead, through a
qualitative analysis, Snow concludes that mothers treat their children as
conversational partners. Stated differently in the perspective of this section,
our culture perceives children as potential conversational partners, even
though they cannot speak or understand. Indeed, our culture allows such a
view of various animals, cars, lucky charms, and televisions. The changes in
Table 5.12, then, result from the content changes that occur as the infant
matures. For the youngest infants, conversation will be limited to grunt,
burps, and smiles. For older children, the content may even include motor
acts and some meaningless vocalizations.
We can appreciate Snow’s perspective by looking at some of the examples
she provides. Here is an example of an exchange between Ann and her
mother at 3 months (p. 12):
Mother
(Ann smiles)
Oh, what a nice little smile!
Yes, isn’t that nice?
There.
There’s a nice little smile.
(Ann burps)
What a nice wind as well!
Yes, that’s better, isn’t it?
Yes.
Yes.
(Ann vocalizes)
Yes!
There’s a nice noise.
Snow comments on this period (p. 13) ‘. . . it is possible to identify a fairly
restricted class of infant behaviors which at 0;3 qualify as unit-types: smiles,
laughs, burps, yawns, sneezes, coughs, coo-vocalizations, and looking
attentively at something.’ While all cultures will need to recognize the
potential for children to be conversational partners, ours apparently
assumes this to be the case from the very onset of life. (It would probably be
possible to find such conversations preceding birth, with the topic being
kicking activity in utero.)
Snow observes that mothers can be quite insistent on the child taking a
turn in the conversation. Here is an example (p. 16) of one such episode:
Mother
What can you see?
What are you looking at?
What are you looking at, hmm?
Hmmm?
(Ann says ‘haaa’)
Haaa.
These earlier conversations lead eventually to real-looking conversations
around 18 months, a topic we shall return to in the next chapter. For the
data for the present period, Snow concludes (p. 21): ‘The way mothers talk
to their babies is one reflection of their belief that the babies are capable of
reciprocal communication.’
Kaye (1980a) The preliminary data on baby talk to infants in Snow’s study
has been expanded by a much larger study by Kaye (1980a). Kaye examined
the language addressed to 37 infants by 36 mothers, one mother having
fraternal twins. Samples of 4-7 minutes were videotaped when the children
were 6,13 and 26 weeks of age; samples of 22 of the mothers were compared
to later samples collected at 26 months and 30 months as reported in Kaye &
Charney (1981). While the infant sample sessions were small, they still
produced a large total sample of 13,574 utterances from the mothers.
136 T H E P E R I O D OF PRELINGUISTIC D E V E L O P M E N T
Kaye used six measures to analyze the mothers’ utterances. Here, we
shall look at four of them:
number of utterances per minute
words per utterance (an MLU measure)
percentage of phatic utterances
percentage of exact repetitions
The term phatic utterance requires an explanation. These are (p. 493): ‘a
class of one-word greetings, consisting of the following utterances: Huh,
Uhhuh, Right, Sure, OK, Yes, Yeah, Yeh, Yep, Yay, Hello, Howdy,
Hm, What, Well, Ah, Oh, and Ooh.’
Like Snow, Kaye does not find that speech adjustment waits until the
child begins to acquire language. His data indicate, however, that adults
speak even shorter utterances to infants than to children acquiring lan-
guage. The short MLU to infants (2.76), high percentage of phatic
utterances (21 percent) and exact repetitions (16 percent) show even
simpler language than that used later. To the 2-year-olds the mothers’ MLU
increased to 3.68, and though the percentage of phatic utterances is not
given, exact repetitions reduced to 3.8 percent. Kaye states (p. 497):
‘. . .the mothers said many two-word utterances, as well as one-word
utterances not in the phatic category (e.g. No, Hey, Lookit, and the baby’s
name)’. While the mothers’ speech adjustments to infants may not be
intended to provide a teaching model of language, the highly repetitive and
simple language presented does provide what should be a helpful model for
acquisition.
Kaye also noticed that there were important individual differences
between mothers. He does not give a detailed analysis, but does provide an
example of two mothers who were very different from each other. Here are
partial samples of the language of these two mothers (p. 503):
Mother 1 Mother 2
1. Comeon. 1. Is that a burp?
2. Talk. 2. Huh?
3. Talk to me. 3. Or are you going to get the hiccups?
4. Can you talk to me? 4. Huh?
5. (laugh) say something. 5. You going to get the hiccups?
6. Comeon. 6. Huh?
7. Talk. 7. Yeah.
8. Can you talk? 8. Hi there.
9. Can you say something? 9. You look like you’re just
10. Well, talk. concentrating too hard.
10. Roseann.
Further reading 137
Mother 1 was highly directive, and used a lot of imperatives and direct
requests; for example, she used 44 percent requests in the sample as
compared to 0 percent for Mother 2. She also had over twice as many exact
repetitions (15 percent vs. 6 percent for Mother 2). Mother 2, on the other
hand, had a more narrative style with few requests made to the infant. She
also had twice as many fragments like ‘huh?’ (utterances 2, 4, 6) than
Mother 1 (56 percent vs. 24 percent). Also, she was more apt to drop an
auxiliary verb in questions, as in utterance 5 , than Mother 1 (50 percent vs.
20 percent of all questions). When we return to the role of the linguistic
environment in later chapters we will examine more carefully the possible
effects of these individual differences.
Kaye concludes that the language addressed to infants reflects the
mother’s expectations of the infant. He agrees, then, with Snow that
English mothers perceive their infants as possible conversational partners.
He takes this one step further, however, by allowing for mothers to vary in
their perceptions of their infants. He states (pp. 504-5):
The individual differences in maternal language input to the infants

and children in this study were less a matter of Iinguistic differences
among the mothers, and still less of different attitudes about linguistic
instruction, than they were differences in expectations of the baby as a
person and expectations about how one ought to relate to a baby
(Kaye 1980 [= 1980b, DI]). Authoritarian vs. ego-building vs. egalita-
rian attitudes have consequences in the complexity and structure of
the child’s linguistic environment from birth.
There are two points to be emphasized here. First, there are linguistic
differences that result from these different attitudes, as exemplified by
Mothers 1 and 2. The effects of these differences need to be observed.
Second, the kinds of perception will vary from culture to culture, as
mentioned in 5.5.1. We need to be careful not to draw universal conclusions
in this area of research based solely on work in English. Both of these issues
will be returned to in the next chapter.
Further reading
Infant speech perception
There is quite a large, and controversial, literature that has appeared since
1971. The bibliography in this chapter is a selection of some of the more
frequently cited works. Periodically, books have appeared containing
review papers of the field; the major ones are: Schiefelbusch & Lloyd (1974)
with reviews by Morse, Eimas, and Butterfield & Cairns; Yeni-Komshian,
138 T H E P E R I O D OF PRELINGUISTIC D E V E L O P M E N T
Kavanaugh & Ferguson (1980b) with reviews by Eilers, Kuhl, and A s h &
Pisoni; Schiefelbusch & Bricker (1981) with a review by Trehub, Bull, and
Schneider.
Znfant speech production

Overviews of stages of articulatory development can be found in the articles
by Oller and Stark in Yeni-Komshian, Kavanaugh & Ferguson (1980a).
Descriptive data are provided in Lewis (1936/51), Cruttenden (1970),
Blount (1972), the studies by 0. Irwin, and Oller et al. (1976). Summaries
are available on behaviorist views (Winitz 1969), maturationist views
(Locke 1983), and constructionist views (Oller 1981; Elbers 1982).
Cognitive development
There are numerous introductions to Piaget’s theory; a simple one can be
found in Ginsberg & Opper (1969), also in Piaget & Inhelder (1969). An
in-depth coverage is provided by Flavell(l963). The serious student should
take a look at Piaget’s own researches on sensorimotor development which
are contained in his three books on the topic (Piaget 1948,1952,1954). The
best treatment of Piaget in a language acquisition text is Anisfeld (1984).
Linguistic input
Like infant speech perception, infant-adult interaction has been an area of
active research in recent years. While there are several papers on the topic,
few focus on the linguistic aspects of the mother in relation to later
development and the theories discussed in 5.5. The cultural differences that
may occur can be seen in Schiefflin (1979). The summary of the English
view of the infant as a conversationalist can be seen in Snow (1977b).
Bruner (1975) presents a strong statement on the role of the parent.
6 The period of single-word utterances
6.1 The definition of word acquisition

As we have just seen, the young infant has progressed through several
developments during the first year of life. In a constructionist perspective,
these advances lay the foundation for later growth. In one sense, then, there
is no prelinguistic development, since the origins of the infant’s language
rest in its first reflexes. This point was raised several years ago by Kaplan &
Kaplan (1971) in a paper entitled ‘Is there such a thing as a prelinguistic
child?’. Despite this, language is most commonly seen as beginning with the
acquisition of the first words.
At first glance, it may appear that marking the onset of language by the
child’s first words should be an easy task. In her classic review of child
language, however, McCarthy (1954) pointed out how difficult it is to
discuss the acquisition of words. Adapting her discussion, we can propose
that a word is acquired in any one of the following definitions:
(i) a word of the adult language that is understood with some meaning,
however variable, by the child;
(ii) a word of the adult language that is understood in approximately its
adult meaning;
(iii) any vocalization of the child that is used in a consistent context;
(iv) a word of the adult language that is produced in a consistent context;
(v) a word of the adult language that is understood and used in an
adult-like manner;
(vi) a word of the adult language that is understood and used in an
adult-like manner, and is pronounced correctly.
There are several factors that need to be considered. First, there is the issue
of comprehension vs. production. Second, we need to decide between adult
and non-adult vocalizations as in definition (iii). Third, there is the issue of
correct pronunciation. If we take definition (i), then word acquisition
occurs around the end of the first year. If we take definition (vi), the first
words will not be acquired until much later.
139
140 T H E PERIOD O F S I N G L E -W O R D UTTERANCES
Recall in Stern’s stages, the first period begins with the onset of single
word utterances in production, presumably through definition (iv) above.
Here, however, this period will be shown as beginning with the onset of
consistent understanding of adult words. Thus, I will use definition (i) to
establish the boundary between what is prelinguistic and what is linguistic.
Nothing of import should be made of it, however, since it is primarily a
descriptive decision. It will only be important within a theory of acquisition
which attempts to propose principles for the transition of one infant ability
to another.
6.2 Early word comprehension and production

6.2.1 Onset and rate
Most of our information on word comprehension until recently came from
diary studies. The general picture was that children begin to respond
meaningfully to words around 1 year of age, and increase their receptive
vocabulary for several months before any noticeable gains in production.
This led to a general conclusion that comprehension precedes production. In
Table 2.1, for example, we gave data on Axel Preyer from Preyer (1889)
indicating that Axel at 17 months had a receptive vocabulary of 35 words but
a productive vocabulary of only two words. This impression was confirmed
by several other diarists, despite discrepancies in their methodologies.
Given the current findings on the infant’s perceptual abilities, it is not
surprising that the child should begin to identify adult vocalizations by 10
months or so. This ability, combined with its growing knowledge of the
world, enables it to pair the recognizable vocalizations with consistent
contexts, and eventually with concepts. The observation that the child’s
abilities lead the child to do this, of course, does not reduce the amazing
nature of the event.
Several studies during the period of large sample studies attempted to go
beyond the individual data of diaries to provide norms for the acquisition of
words. Most of this work focussed on production, to the neglect of
comprehension. In one of the most cited studies, Charlotte Buhler (1931)
looked at the age of the onset of the first word in production for 46 German
children. She found the following age distribution:
Age of first word acquisition No. of children
0;8-0;9 10
0;lO 19
1;O-l;l 10
1;2-1;3 5
1;4-1;5 2
From this, she concluded that the first words in production occur around 10
months.
Onset, of course, provides little information about the cause of develop-
ment. In the first, and for a long time most widely accepted, study on
English children, M. Smith (1926) looked at the average vocabulary size for
children at several age periods. I give here a simplification of her data for
the early months of development:
Age No. of words in vocabulary
0;8 0
0;lO 1
l;o 3
1;3 19
1;6 22
1;9 118
These simple figures show that the onset of a productive vocabulary at 10
months does not lead to rapid growth. Instead, there is a period of several
months before the word spurt in production sometime around 18 months.
We were left with these rather general findings until Benedict (1979)
provided the first controlled study comparing the onset of comprehension
and production.
Benedict (1979) Benedict followed the vocabulary acquisition of eight

children longitudinally for approximately a six-month period. By compar-
ing several children through the same method of inquiry, she was able to
overcome the difficulties inherent in trying to compare different diary
studies. At the same time, by limiting her subjects to eight, she was able to
do a more in-depth study of individual children than previous large sample
studies.
To score the acquisition of her subjects’ vocabularies, Benedict deter-
mined the ages at which each child acquired predetermined numbers of
words for comprehension and production. She presented her results for
each child in a way which allows for some rearranging of the data. Table 6.1
presents her general results for all her subjects. The advanced development
of comprehension is quite striking. First, the onset of comprehension was
nearly four months in advance of production. Second, the rate of acquiring
the first 50 words in comprehension was twice as fast as that for production.
The children needed approximately two weeks to acquire 10 words in
comprehension, but four weeks in production. These results in a carefully
controlled study confirm the initial findings of diary studies. There is one
discrepancy with the earlier findings in that the onset of the first words was
later for Benedict’s subjects than for Buhler’s, although the vocabulary
142 THE PERIOD OFSINGLE-WORD UTTERANCES
Table 6.1 The mean age at which Benedict’s (1979) eight children acquired
their 20th, 40th, and 50th words in comprehension and production
Comprehension Production
No. of words acquired Mean age No. of words acquired
0 0;10(14)
20 0;11(15)
30 1 ;0(3)
40 1;0(19)
50 1;W
1;1(21) 0
1;3(6) 20
1;4(14) 30
1;5(16) 40
1;9(15) 50
sizes for Benedict’s subjects at 18 months were larger than those found by
Smith. Given differences of method and Benedict’s small sample, however,
one should not make too much of these differences.
Benedict’s results are particularly interesting when we go beyond the
general findings and look at individual children. Table 6.2 provides a
simplified view of her results for four of the eight subjects, comparing the
age at which specific numbers of words were acquired in comprehension and
production. It shows two aspects of the data: the relation between the sizes
of the comprehension and production vocabularies, and the rate of acqui-
sition in each.
The results show four apparently very different children. Michael shows a
relatively small gap between comprehension and production. When he has
100 words comprehended, he already has 20 words produced. David shows
an even smaller gap, with 40 words produced with only 80 comprehended.
Diana and Elizabeth have a wider gap, with the latter showing 150 words
understood before any words produced. Clearly, the gap between the sizes
of the child’s receptive and productive vocabularies can vary tremendously.
Even though Michael and David have small gaps between the vocabu-
laries comprehended and produced, their patterns of development are by
no means the same. Looking at the rate of acquisition, based only on the
first 50 words in each vocabulary, Michael acquires his first 50 words in
comprehension and production at about the same rate, which is approxi-
mately two days per word. David, on the other hand, acquired the first 50
words in comprehension five times faster than in production, with rates of
one day per word vs. five days per word respectively. Stated differently,
David acquired his comprehension vocabulary twice as fast as Michael did,
but his productive vocabulary twice as slowly. Regarding rate, Elizabeth
Table 6.2 The number of words acquired in comprehension (C) and
production (P), and the number of days per word for the first 50 words in
comprehension and production, for four subjects taken from Benedict
(1979)“
Michael Elizabeth David Diana

Age C P C P C P C P
~~
0;9 0
0;lO 20 0 0 0
0;11 30 30 20 0 30
1;o 80 0 60 40 1 40 0
1;l 80 50 20 50
1;2 100 20 150 0 30 80
1;3 30 20 80 40 100 20
1;4 50 30 30
1;5 40 50
1;6 50 40
1;7 50
Rate b 2.3 2.0 1.6 2.0 1.o 5.0 1.8 3.7
Vocabularies at specific ages are approximate in that the data were not presented in this
manner originally.
Mean number of days needed to acquire a new word.
is very similar to Michael, in that her rate of acquisition for the two
vocabularies is similar, around two days per word for each. Diana, on the
other hand, is more like David in that her rate of acquisition of receptive
vocabulary is twice as fast as her acquisition of productive vocabulary.
These results show that we have to be very cautious in applying the norms
of Table 6.1 to individual children. Further, the results in Table 6.2 reveal
that rate of acquisition and the gap between comprehension and production
are, to a certain extent, independent. Two general conclusions are: (i) the
rate of comprehension acquisition is the same as or greater than that for
production; and (ii) the gap between the two varies greatly, with a norm of
approximately 100 words understood at the time of the first words
produced.
6.2.2 General semantic categories

In the period of large sample studies, investigators looked at the general
meanings of the child’s first words by classifying words into syntactic
categories such as noun, verb, adjective, etc. While these are categories of
the adult’s language, we have to be cautious about assigning such categories
to the child’s. Our Competence Assumption, in fact, requires us to do so
only if there is evidence in the child’s linguistic behavior for such classes.
144 THE PERIOD OF SINGLE-WORD UTTERANCES
Table 6.3 Nelson’s (1973) semantic categories for early word meaning, as
defined and adapted by Benedict (1979:192-3)
1. Specific nominals: words that refer to only one exemplar of a category, but are not
necessarily limited to proper names, e.g. ‘Daddy’, ‘Coppy’ (name of pet).
2. General nominals: words which refer to all members of a category. It includes inanimate
and animate objects, and pronouns like ‘this’, ‘that’, ‘he’.
3 . Action words: words that elicit specific actions from the child or that accompany actions of
the child. It includes social-action games, e.g. ‘peekaboo’ and ‘what does doggie say?’;
event words, e.g. ‘eat’; locatives, e.g. ‘where’s -?’; general actions, e.g. ‘give’; and
action inhibitors, e.g. ‘no’, ‘don’t touch’.
4. Modifiers: words that refer to properties or qualities of things or events. It includes
attributes, e.g. ‘big’; states, e.g. ‘allgone’, ‘hot’; locatives, e.g. ‘there’; possessives, e.g.
‘mine’.
5 . Personal-social: words that express affective states and social relationships. It includes
assertions, e.g. ‘yes’, ‘no’, ‘want’; and social-expressive actions, e.g. ‘bye-bye’, ‘hi’,
‘nite-nite’.
In the very important monograph entitled Structure and strategy in

learning to talk, Katherine Nelson (1973) developed a more semantically
oriented taxonomy to replace the earlier syntactic ones. Table 6.3 provides
Nelson’s taxonomy, as modified by Benedict (1979), who was a student of
Nelson’s. Benedict’s adaptation was done to make the categories more
compatible for analyzing comprehension,
Nelson followed 18 children longitudinally, starting when they were
around 1 year. Below I give the three arbitrary age groups she divided her
subjects into:
Group (Age) No. of children
I 1;2-1;3 7
I1 1;o-l;l 5
I11 0;l~;ll 6
She asked the parents of each child to keep a parental diary of their word
productions. In addition, she visited each child once a month. We saw
above that Benedict altered this methodology by visiting every two weeks,
and also by studying comprehension vocabulary. Nelson followed her
children until they had acquired their 50th word in production.
Nelson analyzed the words acquired by her 18 children according to the
categories in Table 6.3. Table 6.4 gives a summary of her major findings.
First, for all children, general nominals were the largest class of words, with
51 percent of the children’s words falling into this category. This confirms
Stern’s earlier claim of a substance stage at the onset of vocabulary
development. She also found that the acquisition of the nominals changed
Table 6.4 Percentage of vocabulary that fell into six categories for 18
subjects in Nelson (1973) and for two groups of expressive and referential
children
Category Referential (Y O ) Expressive (YO) Combined (Y O )
Specific nominals 3 7 14
General nominals 38 17 51
Action words 4 6 13
Modifiers 2 6 9
Personal-social 1 12 8
Other 1 2 4
over time; general nominals increased in class size, while specific nominals
decreased:
Category No. of words acquired
1-10 21-30 41-50
Specific nominals (%) 24 14 9
General nominals (YO) 41 46 62
In addition, Nelson divided her children into two groups which differed in
their acquisition of these categories. These were called expressive and
referential children. As seen in Table 6.4, the expressive children had half as
many general nominals as the referential ones, but had many more
personal-social words - the referential children had very few of these.
There are at least two points that need to be discussed about this finding.
One is that while the children could be placed into these groups, their use of
these categories appeared to fall on a continuum. That is, except for a few
children at the extremes, children only tended toward one type or the other.
It may be that children may follow a normal distribution in the extent to
which they use general nominals or personal-social words. This is important
to state, since otherwise one may conclude that children fall neatly into one
category or another. Observe that we could set an arbitrary measure such as
saying that all children with 6 percent or more personal-social words are
expressive. If so, we have imposed a measurement sequence in the sense of
Chapter 4.
A second issue concerns the cause for this variation. In Chapter 4 we
explored three possible causes for variation. First, it could result from
performance differences, Le. that some children focus on naming while
others focus on self-expression. Nelson, in fact, seems to prefer an
explanation of this kind. A second cause could be the effect of the linguistic
environment. It is possible, for example, that some parents concentrate on
getting their children to learn names, while others allow them to concen-
Table 6.5 Percentages of occurrence of words in three semantic categories
for the first 80 words in comprehension (C), and the first 50 words in
production (P), for eight subjects studied by Benedict (1979)
No. of words acquired

0-10 0-30 0-50 0-80
Category P(Y0) C(%) P(Y0) C(Y0) P(O/O) C(X) C(Y0)
Specific nominals 24 30 17 19 11 17 15
General nominals 38 14 41 33 50 39 43
Action words 22 53 26 44 19 36 36
trate on self-expression. Nelson does not support this interpretation, in that

she describes cases where the adult may have one style while the child has
the other. A third possibility is that the difference reflects two distinct ways
children may choose to acquire language. We could say that the referential
child selects to acquire nominals and will use these as cues to sentence
structure, while the expressive child is concentrating on the pragmatics of
utterances. I am unaware of anyone who has interpreted the results in this
latter fashion. If Nelson is correct, we have found an interesting pattern of
individual variation in lexical development that tells us little about the
child's language acquisition system.
Benedict (1979) wanted to replicate the developmental changes sug-
gested by the data on acquisition of general and specific nominals and to
expand it to the emerging comprehension vocabulary. Table 6.5 provides
information on the three most frequent categories for her eight subjects
across the acquisition of the first 80 words in comprehension, and the first
50 words in production. Benedict noticed that the distribution of these
vocabularies was quite different for comprehension than for production. In
comprehension, the most frequent class at first was action words, followed
by specific nominals. In production, on the other hand, general nominals
are more frequent at the onset. She concludes from this that comprehension
and production are different in their propensity to focus on action words.
There is, however, a much simpler explanation. Figure 6.1 presents my
reanalysis of Benedict's data in Table 6.5. The distribution of the three
categories in comprehension and production is shown along the time axis.
As can be seen, the distribution of the three classes at the onset of
productive vocabulary around 15 months is the same as the distribution of
the comprehension vocabulary at that time. The production data, therefore,
simply reflect the fact that the children are producing lexical types in a
comparable proportion to their receptive vocabulary. If so, the data do not
support qualitative similarities between production and comprehension. In
-
KEY
Comprehension o Specific nominals
.......... Production A General nominals
m Action words
. .**A.......-.A
-5a
m
50
....
9 A ..............
-)L
&.e-
40 ,
-$
m
c)
3 30
’CI
0 \ 0. .....
...............E
....‘-m.........
Y%
C
20
. ............
e,
2 ....““..Q........0
0..
$ 10
I I 1 I I , I
0;11 1;0 1;l 1;2 1;3 1;4 1;5 1;6 1;7

Age
Figure 6.1 Percentages of occurrences for specific nominals, general nominals,
and action words, in comprehension and production, in the vocabularies in
Benedict (1979).
a more recent simple subject study, Hoek, Ingram & Gibson (1986) have
found a similar result.
6.2.3 Acquisition of individual lexical items

I began this chapter by defining word acquisition in a very loose way in
reference to the assignment of some meaning. Ultimately, of course, the
child will need not only to use a word with meaning, but also to use it with
the same meaning as that assigned by the adult language. That is, we will
eventually need to push our definition of acquisition to correct usage.
At the onset of acquisition, the words first acquired in comprehension
have a strong tendency towards actions. Table 6.6 provides Benedict’s data
for the most frequent early words understood. Here frequency refers to the
percentage of subjects who understood a word as one of their first 50 words.
As can be seen, several of these are part of social-action games that are very
limited in their meanings. The young child does not have a wide range of
activities when it responds to terms like ‘pat-a-cake’ and ‘peekaboo’.
‘Dance’, ‘give kisses’, and even ‘bye-bye’ are for the young child very
restricted acts. This limited context for words is characteristic of the first
meanings of the child’s words. This observation has been supported in a
Table 6.6 The most frequent action words found in the first 50 words
comprehended and produced by eight children, based on the percentage of
subjects who acquired each word, from Benedict (1979)
Categories Comprehension (YO) Production (”/.I

~
Social-action games bye-bye 88 bye-bye 63

pat-a-cake 88 woof-woof 63
dance 63 nite-nite 50
peekaboo 63 peekaboo 50
do nice 63
give kisses 50
General actions give 100 no 75
get 88 see 63
no 83 go 38
kiss 63 out 38
Locatives where’s 100 getdown 38
come here 88
look at 75
put in 63
get down 50
study by Snyder, Bates & Betherington (1981). As part of a more elab-

orate study, they analyzed the contextual flexibility of the first 50 words
comprehended for 32 children who had a mean age of 1;1(7). They
collected their data through the use of parental diaries. They devised
an elaborate set of rules (their Table 1) to score words as either contex-
tually restricted or contextually variable. Their children had vocabularies
that averaged 45 words in comprehension and 11 words in production.
Of these, 60 percent and 48 percent respectively were defined as con-
textually restricted. I will refer to cases where the child uses a word
for a more limited range of meanings than does the adult as underex-
tensions.
We can get an idea of the more specific categories that children use in
their early words by returning to Nelson’s study. Table 6.7 gives my
summary of what appear to be the most common words used by her
subjects. Nelson broke down the general nominals into several classes,
which are listed in Table 6.7 in order of descending class size. In other
words, foodldrink words were the largest class, followed by animals, etc. Of
course, these are descriptive labels, and we cannot assume from these
figures that the children had such general classes.
If we examine Table 6.7 carefully, we see that the words acquired refer to
inter-categories of meaning. For example, they use the term ‘dog’ rather
than the more general term ‘animal’, or a more specific one like ‘collie’ or
‘poodle’. This observation was first pointed out by Brown (1958b), and has
been studied in depth with older children by Anglin (1977).
Table 6.7 A composite of the most frequent words to occur across selected
semantic categories, taken from Nelson (1973). Percentages are for the
number of subjects who acquired a word in its first50 words in production
Category % of subjects with word production
Specific nominals“ ‘daddy’, ‘mommy’ (at least one proper name) 100
General nominals
Human” ‘baby’ 63
Non-humanh
food/drink ‘juice’ 67, ‘milk’, ‘cookie’ 56
‘water’ 44, ‘toast’ 39, ‘apple’, ‘cake’ 28
animals ‘dog’ 89, ‘cat’ 78, ‘duck 44, ‘horse’ 28
clothes ‘shoes’ 61, ‘hat’ 28
toys ‘ball’ 7 2 , ‘blocks’ 39
ve hicles ’car’ 72, ‘boat’, ‘truck’ 33
furniture ‘clock’ 39, ‘light’ 33
other ‘bottle’ 44, ‘key’ 33, ‘book’ 28
Action words“ ‘up’ 50, ‘sit’, ‘see’ 38
‘eat’, ‘down’, ‘go’ 25
Modifiers“ ‘hot’ 75, ‘allgone’, ‘more’ 38
‘dirty’, ‘cold’, ‘here’, ‘there’ 25
Personal-social words“ ‘hi’ 88, ‘bye(bye)’ 63, ‘no’, ‘yes(yeah)’ 50
‘please’, ‘thank-you’ 38
These are taken from Appendix A where Nelson gives the first 50 words for eight children.
These are taken from Table 8 where Nelson gives data for all 18 children.
Information like that in Table 6.7 is the first step in trying to determine
the nature of word meaning in young children. Besides knowing the specific
words children acquire, we need to look carefully at their contexts of usage.
This is, of course, no easy task. Most of the work in this area comes from
diaries and from the more recent studies which combine parental diaries
with naturalistic language sampling - even then our data are often incom-
plete. Some generalizable findings, however, have emerged.
As cited above, the earliest uses of words appear to be cases of
underextensions. Shortly after, however, there is evidence that the child’s
use of the word becomes generalized, not only to new appropriate
instances, but to inappropriate ones. For example, Braunwald (1978), in
her diary of her daughter Laura, provides the following data on the
production of ‘ball’:
(6.1) 1;0(9) picture of a ball in a book
1;0(9)-1;4 (i) a ball
(ii) round objects, e.g. grapefuit, orange, seedpod, door-
bell buzzer
(iii) request for the first and second servings of liquid in a
CUP
What started out as an underextension at 1;0(9) became overextended for

several months after. These instances of inappropriate usage have come to
be known as overextensions.
Overextensions have intrigued investigators for years, especially because
the word’s meanings can spread so far from the original context. Braunwald
gives a particularly striking example of Laura’s use of ‘cookie’:
(6.2) 1;0(9) used for cookies
1;0(9)-1;4 (i) novel round foods, e.g. cheerios, cucumber
(ii) ‘record players’ and/or ‘music’ on hi-fi or car radio
(iii) rocking and/or rocking chair
(iv) ice cream
The first overextension was to round new food items. Next, the perceptual
feature of roundness took over, where the round record turntable led to its
reference as a ‘cookie’. Soon she associated the music of the record player e
with the word. Eventually the rocking chair motion was associated to the
movements made in response to music. Vygotsky (1962) years ago com- 1
mented on these occurrences, which have come to be referred to as

associative complexes. As the above example has demonstrated, they can
get quite far from the original meaning.
The first systematic study of overextensions was done by Eve Clark
(1973a). Aware of the rich data contained in diary studies, Clark found as
many cases as she could in a survey of several early diaries. Her analysis of
these anecdotal examples found that most were due to perceptual similari-
ties between the objects. The perceptual features she found were shape,
size, sound, movement, taste, and texture. Her study shows the value of
using a review of diary studies as a first step in the study of a particular topic.
In this case, however, additional study is needed since the diaries did not
report on comprehension, and of course one cannot test the diary data to
find the full range of possible overextensions. A more comprehensive
treatment of the topic is Rescorla (1980).
Rescorla (1980) Rescorla, another student of Katherine Nelson’s, followed

six children longitudinally through the acquisition of the first 75 words in
production. Like Nelson (1973) and Benedict (1979) she relied heavily upon
diaries kept by the parents. The children were also visited for a one-or
two-hour session every two weeks. The children were approximately 1year
at the onset and were followed for at least six months. I
Since six children each acquired 75 words, there was a total of 445 words
acquired. Of these, 149 or 33 percent were observed as being overextended.
The 149 words overextended occurred so 190 times, indicating that many
words were only overextended once. These data are quite important in
I
Table 6.8 Three types of overextensions, as defined by Rescorla (1980)
~~ ~
Category Definition ExampIes

Categorical ‘the use of a word to label a referent ‘dada’ for mother
overextensions close to the word‘s standard referent ‘truck’ for bus
in some clear higher-order category of ‘apple’ for oranges
adult usage’ (pp. 325-6)
Analogical a non-adult referent for a word where 1. ‘ticktock’ for the sound of
overextensions it bore no clear categorical relation to dripping water
what would have been an appropriate 2. ‘hat’ for basket on child‘s
referent. The similarity could be: head
1. perceptual, 2. functional, or 3. ‘hot’ for glassware that the
3. affective (see examples) child was prohibited from
touching
Predicate when the child would ‘convey some ‘doll’ said in absence of doll to
statements information about the relationship refer to the place in the crib
between the immediate referent and where the doll was normally
some absent person, object, property found
or state, rather than labelling the
referent itself’ (p. 326)
showing that overextensions only occur for a subset of the child’s vocabu-
lary, despite the fact that they are noticeable when they do happen. They
also reveal the difficulty in using unsystematic diaries to study over-
extensions, since such diaries tend to concentrate on unique or novel events
with no comments on more regular behaviors.
The analysis of semantic categories indicated that certain categories were
more likely to have words overextended than others. Words within cate-
gories that were overextended were as follows: letters 100 percent; vehicles
76 percent; clothing 62 percent. Only 28 percent of animal words were
overextended, which is interesting in that many of the examples of
overextensions in the diary literature are animal terms. There were 12
words in particular which were overextended: ‘baby’, ‘apple’, ‘car’, ‘truck’,
‘shoe’, ‘hat’, ‘dada’, ‘cheese’, ‘ball’, ‘cat’, ‘dog’, ‘hot’. These were words
which (i) were used by at least four children, and (ii) were overextended by
at least half of the children who used them. All but ‘cheese’ appear among
the first words acquired by English children in Table 6.7. These 12 words
were overextended 55 times, and thus represent 29 percent of the 190
overextended cases. Overextension, then, appears to be a salient character-
istic of a small part of the child’s lexicon.
Rescorla next divided overextensions into three types. These types were
labeled and defined as shown in Table 6.8. The categorical overextensions
are what we typically think of when we think of overextensions. These
occurred in 105 cases or 55 percent of the total number of overextensions.
Table 6.9 Four patterns of associative complexes found by Rescorla (1980)
Pattern No. of words Description
1 18 The associative complex occurred only for about one month, e.g., one
child used ‘baby’ for himself, his toys, other children, and for mother
and father when they pretended to cry. It only happened during the
first month of the word’s acquisition.
2 14 Use of the associative complex for a longer period of 2 to 4 months.
71% of these occurred after 1;3.
3 18 The associative complex developed gradually over time, with new
meanings added sequentially. For example, one child’s use of ‘daddy’
went through these stages:
(1;l-1;2) for her mother;
(1;3-1;4) for all fathers;
(1;41;5) pictures of men or animal fathers.
4 8 The associative pattern showed a mixed or complex temporal pattern.
People words were most frequently overextended, followed by vehicles and

animals. Analogical overextensions were much less frequent, with 36 cases
or 19 percent of all overextensions. The most common example was the use
of ‘ball’ for round objects, which was done by five of the children. Lastly,
predicate statements occurred 47 times or 25 percent of the overextensions.
These are particularly interesting in that they are suggestive of later
two-word sentences. This point will be returned to in section 6.3.
Our characterization of overextensions has been that they follow earlier
underextended applications. Rescorla’s longitudinal analysis supports
this. The percentages of words overextended during the last seven months
of her study were 11, 9, 24, 29, 28, 28, 24. The rate of overextensions
increased over the first four months and was still 24 percent at the end.
Importantly, all the children were beginning to produce multiword sen-
tences by the last month of the study. Thus, overextensions continue on into
the next period and are not unique to the period of single-word utterances.
Further, Rescorla found that the words that were overextended tended to
be the earliest ones acquired:
Words acquired YOof words overextended
1-25 45
26-50 35
51-75 20
That is, nearly half of the words that were overextended in the study were
among the first 25 words acquired. Rescorla states (p. 329): ‘Most of these
words denoted some highly valued, familiar and salient object .’
Lastly, Rescorla analyzed the occurrence of associative complexes. She
defines the word (p. 330) as ‘a collection in which each referent shares some
common attribute with the standard but no single feature characterizes all
exemplars of the complex’. A word which showed an associative complex
would be one with overextensions that combined the categories of Table
6.8. There were 58 words, or 30 percent of the overextended words, which
were used as associative complexes. At least three children used the words
‘daddy’, ‘key’, ‘hot’, ‘mommy’, ‘hat’ and ‘cheese’ as associative complexes.
Rescorla found four patterns of associative complexes, summarized in
Table 6.9. We should not place too much importance on these, since they
look very much like a measurement sequence. The fourth pattern, for
example, appears to be just a category for everything that will not fit into the
first three. Nonetheless, the patterns show the variability that characterizes
associative complexes.
While data on overextensions in production are numerous, the facts on
overextensions in comprehension are less clear. Huttenlocher (1974)’ in a
naturalistic study on four children at around 1year, was the first to observe
that words overextended in production were nonetheless understood cor-
rectly. She states (p. 357): ‘With respect to comprehension, however, I
have noticed no overgeneralization. The children typically did not respond
to a word unless they knew its referent.’ This comment was made after
following her children over a six-month period. More recently, several
studies have indicated that some comprehension overextensions do occur,
but they are less frequent and obvious (Thomson & Chapman 1977; Kuczaj
1982a; Mervis & Canada 1983; Hoek, Ingram & Gibson 1986). The claim is
that the overextensions in comprehension will only be apparent if the child
is presented with a sufficient array of objects. For example, if the child
thinks ‘bear’ means four-legged animal, he will still perform correctly if
offered a choice of a bear, a bird, and an apple.
The first systematic study on comprehension and overextensions was
done by Thomson & Chapman (1977). They tested five English children
with mean age of 22.4 months, and with an average MLU of 1.55. Thus,
their subjects were beyond the period of single-word utterances. Since data
from other studies with younger children are less robust (e.g. Mervis &
Canada 1983; Hoek, Ingram & Gibson 1986)’ I will discuss Thomson &
Chapman’s study here rather than returning to it in the next chapter. The
results from Rescorla indicate that overextensions continue into the next
period, and that no qualitative changes occur. I assume then, without direct
evidence, that Thomson & Chapman’s results on slightly more advanced
children are generalizeable to children at least at the latter part of the period
of single-word utterances.
Thomson & Chapman selected four words that were documented as
being overextended from each diary report of the five children. General
information about the subjects and their test words is given in Table 6.10.
Thomson & Chapman collected a wide range of photographs and pictures to
use in testing for overextensions, as, for example, those used to test the
subject K:
(6.3) Test word Test materials
- 5 Polaroid pictures of mother
‘Mommy’ - 5 Polaroid pictures of father
- several Polaroid pictures of familiar men
- several Polaroid pictures of unfamiliar
people, varied by sex, age, facial features,
and clothing
Total of 42 pictures
‘woof’ - 10 pictures of different kinds of dogs
- 1 picture of K’s toy dog
- pictures of familiar animals (2 of bears, cat,
duck, cow, lamb, horse, toy horse, bird, toy
rabbit)
- pictures of less familiar animals (buffalo,
reindeer, wolf, donkey, kangaroo, giraffe,
elephant, hippopotamus, fox, koala bear,
penguin, tiger, llama, sea horse)
Total of 35 pictures
‘apple’ - 11 pictures of different kinds of apples

- pictures of familiar objects (pumpkin, 2 of
tomatoes, peas, orange, strawberry, 3 of
balls, 2 of footballs, lamp, lemon, banana,
carrot)
- pictures of less familiar objects (whole
onions, 2 of soap, plain green oval paper)
First there was a production task in which the child was shown pictures
until he labeled at least five pictures correctly and ten pictures incorrectly.
Next, there was a comprehension task in which each of the four words was
tested 16 times. On ten trials, the child was shown an appropriate picture
along with an inappropriate one. On five trials, an overextended picture was
shown with an inappropriate one. The last trial had two unrelated pictures
together. The child had to respond to either ‘Show me the X or ‘Where is
X?’. A word was scored as overextended in comprehension if a correct
picture was selected on the first ten trials seven times or Zess. Stated
differently, a child was credited with having the word understood correctly
if it got eight or more correct on the first ten trials.
Table 6.10 Subject, age, MLU and test words for five children studied by
Thomson (e Chapman (1977)
Subject Age MLU Test words
D 1;11 2.46 *doggie, *cow, *fish, *ketchup

F 2;3 1.66 Daddy, Joey, *kitty-cat, apple
I 1;9 1.32 Daddy, bow-wow, banana, ball
J 1;8 1.09 Daddy, *dog, apple, *ball
K 1;9 1.20 Daddy, Mommy, woof, *apple
* Words overextended in comprehension.
Table 6.10 indicates the words that were overextended in comprehension

according to Thomson & Chapman’s score. Only eight of the 20 words
tested were overextended. Curiously, the one child who overextended all
four words was the most advanced child, D. The other three children
showed much less evidence for overextensions. The results indicate that
some overextension may occur in comprehension, but that it is much less
widespread than in production.
6.3 The explanation of early word meaning

The presentation so far of the acquisition of word meaning has been
primarily descriptive. I have tried to present what appears to be some of the
more reliable information on how the child’s early vocabulary develops.
There was no attempt, however, to offer an explanation of what underlies
these behavioral changes. There are now two separate theoretical questions
that need to be addressed: what is the theory of semantics that accounts for
the adult’s language, and what is the theory of acquisition that accounts for
the child’s errors, Le. overextensions?
It is my impression that linguists are in much less agreement about what
constitutes a theory of semantics than they are about what constitutes a
grammar. For example, in syntax we can propose that the child must
acquire at least a phrase structure grammar that contains certain properties,
e.g. categories such as NP, VP, and rules that relate antecedents to
pronouns or moved NPs to their initial positions. What are the parallels to
these examples in semantic theory? Current theories make quite distinct
claims from each other. For example, feature theories assign word
meanings binary features such as [k human] while prototype theories
assign meanings according to proximity to some typical exemplar of the
word’s meaning. I suspect the beginning student could become quite
bogged down just understanding these theoretical alternatives without
learning anything more about the child’s language acquisition. Stated
differently, it appears that child data can be manipulated to support any of
the current theories, in the sense that the independent evidence is difficult
to find.
In this section I will focus on the second question raised above concerning
our theory of acquisition, using this discussion to set up a later discussion on
theories of semantic acquisition, to be presented in Chapter 7. I am taking,
therefore, an inductive approach, beginning with an analysis of what
aspects of the child’s linguistic behavior constitute evidence relevant to a
theory of semantics.
The Competence Assumption instructs us to consider the child’s per-
formance to be indicative of its competence. In the case of word over-
extension, the assumption will apply when we have evidence that the child
overextends (or underextends) a word in the same way in both comprehen-
sion and production. For example, in (6.1), we gave the example where
Laura Braunwald used ‘ball’ for the meaning of round objects. Suppose that
she also understood ‘ball’ as meaning round objects. If so, her performance
leads us to assume that ‘ball’ means round objects for Laura. It is data like
these that are directly relevant for the construction of theories of acqui-
sition. Certain theories, in fact (e.g. Clark 1973) have been constructed on
this assumption. The mechanism for such linguistic behaviors will be the
child’s linguistic (or semantic) system.
Unfortunately, such data are rare in language acquisition. Those few
studies that have followed overextensions in both so far suggest that the
majority of overextensions in production do not occur in comprehension.
For example, Huttenlocher (1974: 339) states in her important study: ‘I
don’t know how to test whether the child knows that “hi” is said as people
arrive and “byebye” as they leave.’ It appears then that the data from word
acquisition during the current period of acquisition do not easily lend
themselves as evidence for the nature of the child’s semantic system. This
state of affairs is further complicated by the influence of the child’s cognitive
development, to which we now return.
6.3.1 Cognitive factors in word acquisition

In section 5.4 we outlined the first four of Piaget’s six sensorimotor stages.
These four stages cover the period of development between birth and
approximately 1year of age. The next six months (1year to 18 months) are
described by Piaget as sensorimotor stages 5 and 6. As presented in 5.4.3,
the onset of concepts of the kind that underlie word meaning as in the adult
language do not appear until around 18 months or after the sensorimotor
period. The signifier-signified relationship develops through several stages.
Up to sensorimotor stage 5, the first relation that occurs is that of the signal
or elementary indication. During sensorimotor stage 5 , the first identifiable
word-like vocalizations appear, but Piaget argues that they are not yet
words or linguistic signs. That is, they do not yet pair vocalizations with
concepts. Instead, they are symbols, a relation that exists between signals
and signs. Here we will take a brief look at sensorimotor stages 5 and 6,
and examine the notion of symbol.
The major gain around 1 year is the appearance of the tertiary circular
reaction. In the fourth stage, the child is restricted to coordinating familiar
schemata to solve a novel problem. In stage 5, the child begins to discover
new ways that were not used before, although it happens through trial and
error. Piaget states (p. 264) that the fifth stage ‘is characterized, in effect,
by the formation of new schemata which are due no longer to a simple
reproduction of fortuitous results but to a sort of experimentation or search
for novelty as such’. Here the ability to control the means to reach an end is
increased. The child can begin to use objects as means, e.g. a stick to move
or get an object. As we will see in section 6.4, this achievement coincides
with the pragmatic use of vocalization, that is, language becomes one of
the means the child can use to obtain ends, either through understanding
or speaking.
Another development is more active exploration of objects, as well as
better ability to locate hidden objects. The child becomes more interested
in aspects of objects such as the sounds they make and their trajectory
when they fall. When he sees an object hidden in one place, such as under
a cup, and then in another, such as under a plate, he will look directly
under the plate. During stage 4, he would go to the cup if it were the pre-
vious place where the object was hidden. At stage 5 , however, the child
still cannot succeed in an invisible displacement. For example, if we hide
the ball in the cup, then put the cup behind our back, leaving the ball
behind us, then return the cup, the child does not look behind us for the
ball. The child’s involvement with objects and its interest in seeking objects
has been suggested as a prerequisite for establishing the signifier-signified
relation. I will return to this point below. If it is necessary, the child’s
incomplete achievement of object permanence would be a further argu-
ment for the lack of the linguistic sign at this period.
A third development in stage 5 is the appearance of symbols. Piaget
(1948) proposes that these occur in stage 6, but his own data (cf. Ingram
1978) suggests their appearance in stage 5, with a continuation into stage 6.
In fact, Piaget (1948: 53) states in reference to the growth of imitation: ‘It
was, as a matter of fact, during this fifth stage that J., L., and T. began to
make their first clumsy efforts to reproduce the words of adults.’ He sees
these words, however, as symbols or ‘motivated signifiers, i.e. those which
are related to the signified by some resemblance, as distinct from “signs”,
which are “arbitrary”’ (Piaget 1948: 68). They are still associated with the
contexts they occur in, and even represent for the child part of that
context.
(6.4) is an example from Piaget (1948: 216) of one such symbol used by
Jacqueline:
(6.4) ‘bow-wow’l;l(20) - to indicate dogs
- to indicate landlord’s dog
- to geometrical pattern on a rug
- to two horses
- to baby in a pram
- to hens
- at sight of dogs, horses, prams, and cyclists
- everything seen from balcony - animals, cars,
landlord, and people in general
- trucks being pulled by porters
- pattern on rug
- only for dogs
Piaget comments (p. 218): ‘. . . these first verbal schemas are intermediary
between the schemas of sensory-motor intelligence and conceptual
schemas’, or ‘the words applied by the child to these schemas are them-
selves intermediary between symbolic or imitative signifiers and true
signs’. This description of ‘bow-wow’ is typical of what was called earlier
an associative complex. Piaget’s position, then, is that associative com-
plexes are the result of the incompletely established relation between
signifiers and signifieds.
As evidence for his position, Piaget cites examples of children’s sym-
bolic play behavior. The first instances of symbolic play occur during
stage 6, but they are very much restricted. The child is not yet ready to
pretend that any object can be any other, and behaviors are very much
tied to the activity being symbolized, for example focussing on the child’s
activities, such as pretending to sleep, or pretending to wash oneself. It is
only gradually that the child reaches the more advanced play behavior of
having one object represent another, such as having a block of wood be a
truck.
Regarding overextensions then, Piaget would say that they represent, in
their earliest occurrences, the underdeveloped relation between signifiers
and signifieds. They do not so much reveal information about the child’s
concepts, however, as about the child’s sensorimotor schemata. There are,
however, some problems with the incomplete linguistic sign argument. One
is the striking finding of Rescorla that only a small percentage of the child’s
words overextend, even in the earliest records. If the relation is supposed to
be only partially developed, one would expect that all words would be
uniformly affected. Second, Thomson & Chapman’s data suggest that some
overextensions occur in comprehension in slightly older children. Even
more extensive data to this effect can be found in Anglin (1977). Thus, at
some point the data do become revealing of the child’s system. Further, our
evidence on word comprehension (Benedict 1979) suggests something like a
word spurt in comprehension before word production begins. This word
spurt is hard to explain without some reference to the Principle of the
Linguistic Sign. These factors make it difficult to eliminate the possibility
that some of the overextensions during the period of single-word utterances
may directly reflect the child’s organization of its semantic system.
Even if a child has reached the level of linguistic signs, it is still possible to
offer an alternative to the claim that overextensions result from the child’s
incomplete semantic system. As suggested in Bloom (1973) and Thomson &
Chapman (1977), it may be that the child overextends due to limited
vocabulary. That is, the child is presented with an unknown object, and calls
it by a word that is known. This may result purely from perceptual
similarities between the known and unknown objects. In Hoek, Ingram &
Gibson (1986), we tested for this in a diary study of a young girl, Claire, by
seeing if overextensions in production would occur for cases where the
correct word was not part of the child’s receptive vocabulary. There indeed
proved to be several cases of this kind. When this factor operates, we could
conclude that the child has adult-like knowledge for the words it has
acquired, but must resort to other means when presented with new objects.
Even adults probably have to resort to this strategy at one time or another.
6.3.2 Other factors behind overextensions

There is a reasonable amount of data which shows that for at least some
words children will overextend in production but not in comprehension. For
example, Rescorla (1980) refers to this situation for one of her subjects,
Rachel (p. 230):
During the period when she overextended car to a wide range of
vehicles, she was able to pick out all the same objects in response to
their correct name; these included motorcycle, bike, truck, plane, and
helicopter. Once she acquired productive labels for these concepts,
they began to emerge from the car cluster. Indicative of how the
process operated, her first label for airplane was sky car.
What, then, is the cause of such cases?
One possibility, as stated in Thomson & Chapman (1977), is that there
may be a retrieval problem, Le. the child may be unable to recall the correct
word, so he uses a more familiar one. Some evidence for this can be found in
Rescorla (1980) where she observes that overextended words tend to be
frequently used ones that are acquired early. In Hoek, Ingram & Gibson
(1986), we operationalized this into a prediction that overextensions in
production would more likely be an earlier acquired word used for a
recently acquired word instead of vice versa. Our analysis of the parental
diary of a young girl bore out this prediction. Lastly, there may be
phonological factors at work. That is, the child may use a phonologically
simpler word for one that is more complex. Schwartz & Leonard (1982)
have found that children tend to avoid temporarily words that are beyond
their present phonological system (see section 6.5). In Hoek, Ingram &
Gibson (1986), there was some evidence for this in our analysis in which
Claire was taught nonsense words that were either in or outside her
phonological system. Overextensions occurred in the direction of ‘in’ words
for ‘out’ words.
Summary It appears, then, that several factors may be at work when

children overextend their early words. It is therefore simplistic to say that
overextensions during the acquisition of the first words reveal a great deal
about the child’s semantic system. If anything, given the limited data on
overextensions in comprehension, the data indicate that children are
reasonably good at getting the meanings correct for their early words, at
least in regard to simple lexical items. The overextensions that do occur
may be the result of the child’s incomplete vocabulary.
6.4 Pragmatic and grammatical development

So far, we have discussed the meaning that underlies the child’s earliest
words. Besides the child’s word meaning, it is also possible to study the
child’s intentions when it speaks. What is the child’s intended effect when it
utters a word? Also, we will consider the possibility that the beginnings of
grammatical structure may have their origins during the period of single-
word utterances. Discussion of these topics begins with the child’s com-
prehension of multiword utterances during this period, and will be followed
by a discussion of the characteristics of the child’s early one-word sentences.
6.4.1 The comprehension of multiword sentences

We know from Benedict’s (1979) study that the child has a number of words
understood before the first word produced, possibly as many as 100.
Further, we know that the acquisition of the first 50 words in production can
take some time. Most attempts to follow receptive vocabulary after the first
100 words are given up because acquisition is so rapid. By the end of the
Table 6.11 Estimated size of receptive vocabulary for Craig and Amy in
Benedict (1979) at the time of the acquisition of the 50th word in
production
Measures Craig Amy
Comprehension
(a) Age at 100th word comprehended 1;O(17) 1;0(26)
(b) Age at 200th word comprehended 1;3(24) 1;3(16)
(c) No. of days to acquire 200th word after 100th word
comprehended" 97 80
(d) Rate of acquisition (days per word) 0.97 0.80
Production
(e) Age at 50th word produced 1;5(2) 1;6(29)
( f ) No. of days between 50th word produced and 200th word
comprehended 38 103
(g) Estimated size of receptive vocabulary at time of 50 words
produced 239 words329 words
This calculation assumes 30 days per month.
period of single-word utterances, then, the child has a relatively large

receptive vocabulary. Benedict, for example, gives data on two subjects for
the first 200 words in comprehension. Table 6.11 estimates the size of their
receptive vocabulary at the time of the 50th word produced. (We know from
Nelson 1973 that the latter figure tends to coincide with the first multiword
utterances in production.) The table assumes that the rate of acquisition is
that of the development between the 100th and 200th word understood,
probably a conservative estimate, and the estimated receptive vocabularies
are thus 239 and 329 words. Given this size of vocabulary, is it possible that
the children are also beginning to be able to comprehend utterances beyond
a single word? We know, for example, from section 5.5 that mothers speak
sentences, not words, to their young children. We will look into this issue by
first discussing methods of testing comprehension and then presenting some
research findings.
Method of testing comprehension It cannot be overemphasized that testing

comprehension, even of single words, in children around 1year to 18months
is quite difficult. How do we know, for example, that the child understands
'dog' in the way we do? The typical procedure is to present the child with
alternatives that are systematically varied. This was done in the study by
Thomson & Chapman (1977). Even so, children at this age are often not
obliging: they may not attend, crawl off, or start to cry. When they do
attend, we still have problems of scoring a response. They may touch one
object, then another, or only look at the object. Careful scoring procedures
need to be developed and consistently followed. It is important, therefore,
if possible to have several scorers.
These problems are even greater when we investigate the understanding
of sentences. For example, suppose we wish to test the child’s understand-
ing of the Possessor-Possessed structure as in ‘Mommy’sshoe’. How do we
do this? If we say ‘Give me Mommy’s shoe’ we are actually testing a greater
comprehension, which includes the verb ‘give’. Just saying ‘Mommy’s shoe’
does not direct the child to any scorable response. Suppose, however, that
we teach the child a game, so that we name something, and the child hands it
to us. We know from Benedict (1979) (see Table 6.6) that social-action
games are among the child’s earliest acquired Action words. If we say
‘Mommy’s shoe’ and the child hands us its mother’s shoe, does this mean
that the child understands the structure of the sentence? Not necessarily.
For one thing, it could have been the only shoe around. We need to be sure
that there are alternatives to choose from. Suppose the choice is Mommy’s
shoe and baby’s shoe. This tells us that the child recognizes two words in the
sentence, but that’s all. Of course, recognizing the two words is an
important start in grammatical development. We still do not know,
however, if the child recognizes the importance of word order or the
grammatical inflection “s’. It also does not tell us if the child can generalize
the pattern to novel situations like ‘Mommy’s glove’ or ‘Mommy’s sock’.
There are ways that are available to try to get around some of these
problems. One is to use novel stimuli that are unlikely to have been heard
previously, e.g. ‘Kiss the apple’ or ‘truck’sshoe’. To test for grammaticality,
we can give the child ungrammatical structures, e.g. ‘Mommy shoe’ or ‘shoe
Mommy’ to see if the child responds differently. We can not ask directly for
the child’s grammatical intuitions as we can with adults. Particularly with
children at 18 months, we can only try to get responses in controlled
circumstances, and interpret them with great caution.
Studies on sentence comprehension There has been a small number of
creative studies in recent years on the receptive ability of children who are
only producing single-word utterances. This section will look at four of
them which are suggestive of some comprehension of multiword utterances
before the production of multiword utterances.
Shipley, Smith & Gleitman (1969) This study, as the authors stressed
later (Gleitman, Shipley & Smith 1978), was not directly a study of
comprehension. Rather, it tested whether children would prefer to respond
to language from their parents that was at the child’s level of production, or
in advance of it. That is, it was indirectly a study of the young child’s notions
of grammaticality or well-formedness. Nonetheless, their data suggest some
information about children’s understanding of structure.
Shipley, Smith & Gleitman studied two groups of children which they
called the holophrastic group and the telegraphic group. Here we will look
just at their data in regard to the holophrastic group. There were four
subjects in the group:
Name Age MLU
Mike 1;6 1.06
Karen 1;8 1.10
Linus 2;o 1.09
Jeremy 2;o 1.16
These children are older than 18 months, but they were still using mostly
one-word utterances. Shipley, Smith & Gleitman state (p. 325): ‘Although
there were instances of two-word utterances in these children’s speech,
these were so rare as to suggest that they may represent merely benefit-of-
the-doubt decisions by the transcriber.’ I assume that they were at the very
end of the period of single-word utterances.
The children were tested individually in a room where they were allowed
to play freely while the mother and an experimenter talked to each other.
Then, periodically, the mother would turn and direct a command to her
child. There were three main types of commands directed to the child
(where V = verb, N = noun, F = functors):
(6.5) a. well-formed (VFN) ‘Throw me the ball!’
b. telegraphic (VN) ‘Throw ball!’
c. holophrastic (N) ‘~aii!7
The object was to see if the children would prefer one of these three kinds of
structures over another.
Shipley, Smith & Gleitman used several measures to score the children’s
response. Here I will discuss just one, which they called touch. This was
when the child came into physical contact with the toy named in the
command. They say (p. 328): ‘. . .we took this behavior as an indication
that the child had accepted the utterance as a “good” command and was
making the natural response’. Table 6.12 gives the results for the four
children on the three sentence types.
Shipley, Smith & Gleitman analyze their data statistically and draw the
following conclusions (p. 329): ‘. . .all holophrastic speakers obey more
often with single-word commands than with well-formed commands’. Their
statistics include other response measures such as looking at the object and
providing a verbal response. Even so, one would expect touch to show this
same pattern since it was probably the most reliable measure. They also claim
that the responses on VN were less frequent than with N. They conclude that
holophrastic children prefer commands at their own level of production.
Table 6.12 Percentage of times each of four children in Shipley, Smith &
Gleitman (1969) would touch an appropriate toy in response to each of
three constructions
Structures
Subjects N (Yo) V N (Yo) VFN (Yo)
Mike 33 50 16
Karen 80 75 83
Linus 46 16 42
Jeremy 16 33 0
Mean 52 44 35
The data in Table 6.12, however, invite closer examination. Only one of
the children, Linus, shows the pattern for preferring N over VN (though
Karen has a slight preference), but he also prefers VFN over VN. Two
children, Mike and Jeremy, show a clear preference for VN over N. Despite
statistical tendencies, these data suggest that children at the end of the
period of single-word utterances are beginning to respond to VN and VFN
structures. That is, the data are suggestive of an emerging ability to process
VN structures. The following studies more directly address this possibility.
Huttenlocher (1974) Huttenlocher reported preliminary results on her

longitudinal study of four children over a six-month period. The children
were between 10 and 13 months of age when the study began. They were
visited every few weeks and observed on their development of receptive and
productive language. Some of Huttenlocher’s remarks on lexical develop-
ment were discussed in section 6.3.
Huttenlocher not only recorded lexical development but wanted to
determine the children’s grammatical comprehension. For Wendy, only
two frequent words occur in her production at 1;2 (17), these being ‘see’ and
‘hi’. Huttenlocher says (p. 345): ‘Wendy does not at present, at 17 months,
give any evidence of comprehension of relational meaning based on syntax,
nor indeed of any comprehension of more than one word at a time.’ Her
receptive vocabulary, however, is still quite small, around a dozen items.
She is still, therefore, in a very early period of word acquisition. Similar
results occurred with Thomas, who had even less language at 17 months.
The other two children, however, were more advanced in their com-
prehension. The most extensive and revealing data are presented for Craig.
At 1;4(2) Craig had two words in production: ‘di’ and ‘uhuh’. His
comprehension vocabulary was larger, with at least a dozen words. At this
point, Huttenlocher makes the following observation (p. 347):
. . . he follows commands involving three differentiable elements. The
first time this increase occurs is when I notice that Craig will follow a
command of the form ‘Show the X to Mommy’ vs. ‘to Jane’ vs. ‘to
Candy’ vs. ‘to Julie’. Within a week he also differentiates ‘give’ from
‘show’ so that by 16 months he can follow the command ‘Give Jane the
cookie’ vs. ‘Show Jane the cookie’. I test this capacity many times,
contrasting each of the three critical elements.
This ability supports the Shipley, Smith & Gleitman (1969) data on VN, and
adds a third lexical item to the child’s receptive ability.
At 1;5(3), Huttenlocher found differential responses to the Possessor-
Possessed relation in ‘your bottle’ vs. ‘baby’s bottle’, and ‘your diaper’ vs.
‘baby’s diaper’. She tested this further with Craig at 1;5(30) by giving him
the eight commands, listed here:
(6.6) a. Give me baby’s bottle
b. Give me your bottle
c. Give Mommy baby’s bottle
d. Give Mommy your bottle
e. Show me baby’s bottle
f. Show me your bottle
g. Show Mommy baby’s bottle
h. Show Mommy your bottle
A further distinction occurred at 1;6(16) when Craig would respond
differently for ‘Get diaper’ and ‘Where’s diaper?’, where the latter question
elicited pointing. Other indirect evidence for questions occurred at 1;3(5)
where ‘uhuh’ would be made in response to ‘Do you want X?’but not to
‘Where’s the X?’. The data from Kirsten are less extensive, but show
responses to Possessor-Possessed structures like ‘your’ vs. ‘Mommy’s’
‘shoe’ or ‘nose’. This was at a time when no words were being produced.
These results indicate that Craig, at least, was able to recognize four
lexical words in a sentence and carry out a command in an appropriate
context. They do not indicate, however, syntactic processing. We do not
know if a scrambled word order would have produced similar responses or
not. We also do not know if the possessive suffix ‘s’ was being processed.
Even so, they indicate early processing of multiple words in the input. Craig
still seemed far from the point where he would be producing multiword
utterances.
Sachs & Truswell(1978) Sachs & Truswell(l978) examined in more detail

the ability of children to process two semantically contrasting items in a
sentence. They tested 12 children between 16 months and 2 years who were
still only producing single-word utterances. The children were visited in the
home and presented with Action + Object sentences that consisted of words
found in their own receptive vocabulary. The test items, then, varied from
child to child. The sentences were constructed in a way that they would
describe novel or unique situations which the child probably would not have
experienced. Eleven of the 12 children received four-way minimal contrasts
as exemplified in (6.7); the twelfth child was uncooperative and only
received two-way contrasts.
(6.7) a. Smell truck (Leslie 1;6)
b. Smell dolly
c. Kiss truck
d. Kiss dolly
They scored the children’s responses as either ‘1’ correct (the child carried
out the command); ‘2’ incorrect or partially correct (wrong action and/or
object), or ‘3’ no response. They were able to present each child an average
of 16 commands.
Sachs & Truswell found that 58 percent of the responses were correct, 16
percent incorrect or partially correct, and 6 percent elicited no response.
Further, of the 11 children who received four-way minimal contrasts as in
(6.7), ten got at least one such set correct. The one who did not was the
youngest child (1;4), who did get a two-way contrast correct (‘Kiss horsey’
vs. ‘Kiss teddy’). They conclude that children who use single-word pro-
ductions can at least understand novel Action + Object commands.
Miller, Chapman, Bronston & Reichle (1980) This account provides some
indirect evidence about the emergence of the understanding of two- and
three-term sentences by children. This study is summarized in Chapman
(1981a). As part of a general study on the development of cognition and
comprehension, Miller et al. tested comprehension on eight items, which
are given in Table 6.13. There were 12 children in each of the following age
groups: 10-12 months, 13-15 months, 16-18 months, and 19-21 months.
While there is no mention of the children’s language production, we will
assume that at least the first three groups were primarily children in the
period of single-word utterances.
Table 6.14 gives the number of children who passed a particular com-
prehension test item at least once. We can see at least four children in the
16-18 months group passed the items Possessor-Possession and Action-
Object. These results are consistent with results of Shipley, Smith &
Gleitman (1969), Huttenlocher (1974) and Sachs & Truswell (1978).
Interestingly, even in the 19-21 month group, only one child got an item
correct on the Agent-Action-Object structures. If Huttenlocher’s results
Table 6.13 Eight comprehension items tested in Miller, Chapman,
Bronston & Reichle (1980) for 48 children between 10 and 21 months of
age, based on Chapman (1981a: Table 4)
Item and examples Passing response
1. Person name e.g. ‘Where’s Mama?’ Child indicates correct person in

response to question.
2. Object name e.g. ‘Where’s X?’, ‘Go get Child looks at, gets, shows, or gives the
X , ‘Give me X (where X = words appropriate object among several
supplied by mothers) present in visual field.
3. Absent person or object (item passed Child searches for a person or object
from no. 1 or no. 2) when it is out of view.
4. Action verb, ‘V it; wanna V it’ or ‘can Child complies by carrying out action. If
you V it?’ (V = verb supplied by child is already attending to object, then
mother) action should be one not conventionally
associated with it.
5 . Possessor-Possession e.g. ‘Where’s Child appropriately locates the correct
Mama’s shoes?’ (items are ones passed person’s objects twice.
in no. 1 or no. 2 )
6. Action-Object e.g. ‘Kiss the shoe’, ‘Can Child complies for the appropriate
you hug the ball?’, ‘Wanna pat the object among several present. He cannot
book?’ already be attending to the object.
7. Agent (other than child)-Action e.g. Child selects toy among several present
‘Horsey eat’, ‘Make the doggie kiss’, and demonstrates action with the toy
‘Wanna make the horsey eat?’. serving as agent.
8. Agent (other than child)-Action-Object Child selects appropriate toy and object
e.g. ‘Horsey kiss the ball’, ‘Can doggie among those present and demonstrates
eat the diaper?’ appropriately.
are correct, these data suggest that it is not the number of words which is
important, but the function of these. Recall that Craig’s early success was
with structures of the form Action-Object-Recipient. Unfortunately,
Miller et af. did not test items of this kind.
Data suggest that children toward the end of the period of single-word
utterances begin to understand some structures which contain two or
three lexical items. Those that appear understood around this time are
Action-Object and Possessor-Possessed, and possibly Action-Object-
Recipient. No study yet has demonstrated sensitivity to word order or
grammatical morphemes. One general problem is the difficulty of testing
for children’s comprehension at this age. One possible methodology,
which has not yet been applied to this area, is the VRISD paradigm
discussed in section 6.2.2.
Table 6.14 Number of subjectsa in Miller, Chapman, Bronston & Reichle
(I 980) to pass a comprehension item at least once, based on Chapman
(1981a: Table 5)
Age group (in monthsy

Comprehension item 10-12 13-15 1618 19-21
1. Person name 12 11 11
2. Object name 12 12 12
3. Action verb 4 9 10
4. Possessor-Possession 1 5 10
5. Absent person or object 2 4 8
6. Action-Object 1 5 8
7. Agent-Action 0 1 7
8. Agent-Action-Object 0 0 1
Number of subjects in each group was 12.
6.4.2 Some characteristics of single-word utterances

Since the onset of diary studies Child Language has predominantly defined
word acquisition in production as beginning when some vocalization was
produced that sounded like an attempt at a word that was in the adult
language. Typically this would be ‘mama’ and ‘dada’. Some of the better
diaries avoided this, and looked for words, defined as child vocalizations
that appeared to have a consistent meaning despite not sounding like the
adult words to be acquired later. For example, Leopold (1939) observed
that his daughter Hildegard used a vocalization [?a] that coupled with a
pointing gesture served as a kind of demonstrative. A month later, Leopold
interpreted this vocalization as a volitional term, as if it meant ‘There is
something, and I want it.’ Still later, it acquired a rising intonation which
made its use similar to that for a ‘what’s that?’ question. It appeared that a
small number of vocalizations would come to be used in a meaningful way
around 1 year of age in a variety of pragmatic contexts. They appeared to
vary according to the child’s intentions in speaking.
Recall from our discussion on cognitive development that the first words
appear to be used around sensorimotor stages 5 and 6. This observation has
been supported in studies by Bates (1979), Harding & Golinkoff (1979) and
others. The child in these stages is beginning to develop its ability to achieve
ends through novel means. Vocalization appears to be one such means to
achieve ends such as obtaining objects, food, etc. Bates, Camaioni &
Volterra (1975), for example, refer to these vocalizations as ‘performa-
tives’, meaning they are direct reflections of the child’s intentions rather
than the child’s concepts. This, of course, is consistent with the Piagetian
claim that the child does not have concepts at that point. Bates, Camaioni &
Table 6.15 A summary of some of the major studies on the acquisition of
single-word utterances
Investigator Subject(s) Orientation

(Age in months)
Bloom (1973) Allison (9-22) Diary and four 40-minute video

tape-recordings of daughter. Argues
against existence of syntax before
multiword utterances.
Halliday (1975) Nigel (9-18) Diary study of son, divided into six
sample periods, and two phases of
development; proposes taxonomy of six
functions for Phase I.
Carter (1975a,b; David (12-16) Ten one-hour play sessions. Proposes
1978a,b) taxonomy of eight communicative
schemata.
Dore, Franklin, Miller 4 children (11-16 at Monthly videotapes for an eight-month
& Ramer (1976) onset) period; proposes stages of development
of grammatical structure, and a
taxonomy of pragmatic functions.
Greenfield & Smith Matthew (7-22), Diary notes plus nine (Matthew) and
(1976) Nicky (8-21) eight (Nicky) formal sessions; proposes a
taxonomy of 12 semantic functions.
Volterra go on to divide the performatives into two types: proto-declara-

tives, as in the Hildegard demonstrative at 8 months, and proto-
imperatives, i.e. vocalizations intended to get the adult to satisfy some need
on the child’s part. These kinds of vocalizations are small in number and
precede the more adult-like words acquired shortly after, as described in
Table 6.7.
Since the early 1970s there have been several detailed studies on the
acquisition of single-word utterances during the period of single-word
utterances. Table 6.15 gives a summary of some of the more extensive ones,
resulting in a set of observations and proposed taxonomies of the speech of
children during this period. To present these works descriptively and
critically would take us far beyond the scope of this text. Instead, we will
select some of the salient observations, proposals, and issues, beginning
with an outline of two of the proposed semantic taxonomies of this period.
These are the proposals by Halliday (1975) and Greenfield & Smith (1976).
Halliday (1W5) Halliday kept a detailed diary of his son Nigel from 9 to 18
months. To study these data, he divided Nigel’s development into six-week
intervals that he labeled as Nigel (NL) 0 through 6. The actual ages for these
are as follows, in months:
Table 6.16 Definitions for six functional concepts that underlie early word
use, according to Halliday (1975) and the number of words belonging to
each for samples NL 1 to 5
(1) Instrumental function: 2 3 5 5 1 0

used to satisfy the child’s needs to obtain goods or
services, the ‘I want’ function
(2) Regulatory function: 2 2 6 6 7
used to ccntrol the behavior of others; the ‘do as I
tell you’ function
( 3 ) Interactional function: 3 7 7 7 1 5
used to interact with those around child; the ‘me and
you’ function
(4) Personal function: 5 9 9 11 16
used to express the child’s own uniqueness; the ‘here
I am’ function
( 5 ) Heuristic function: - - - ? ?
used to explore the environment; the ‘tell me why’
function
( 6 ) Imaginative function: - - 2 3 4
used by child to create its own environment; the ’let’s
pretend’ function
Totals 12 21 29 32 52
NL 0:9 NL 3:13f NL 6:18

NL 1:la NL 4:15
NL 2:12 NL 5:164
The notes combined phonetic transcription with careful contextual obser-
vations.
Halliday divides Nigel’s development into three phases. Phase I refers to
the period from NL 0 to NL 5 , or development up to 164 months of age.
During Phase I, Nigel used only single-word utterances, and acquired a
small set of words (approximately 50). Halliday argues that Nigel had no
syntax during this phase, but only a direct mapping from sound to meanings.
The meanings in this case are ones that Nigel created for his utterances, not
ones he acquired from the linguistic environment. He sees the child, much
as Taine did in his (1877) classic diary study, as creating language at this
point. Halliday refers to the language created by the child during Phase I as
a proto-language, and the vocalizations proto-words. A proto-word is
basically the third definition of word in McCarthy (1954), given on page 139,
with the meaning coming from the child’s internal ability to express
intentions.
Halliday discusses the emergence of six functions during Phase I for
Nigel. These are briefly defined in Table 6.16 along with the number of
words that belonged to each function for NL 1 through 5. Four of these
functions, instrumental, regulatory, interactional, and personal, are fre-
quent during Phase I. The other two, heuristic and imaginative, develop
more during Phase 11. Table 6.17 gives Halliday’s analysis for NL 3 for the
29 proto-words acquired around 134 months of age.
Halliday wants to argue that these functions are the foundation upon
which later development is based. While the descriptions of each are
straightforward enough, there are no operational definitions of how anyone
can apply these to data from another child. The only method available is an
understanding of the definitions combined with a very careful reading of his
actual analyses from Nigel, focussing on the meanings as in Table 6.17.
A look at the data at NL 3 should provide insight into the basic features of
Halliday’s taxonomy. First, I have italicized under ‘meanings’ in Table 6.17
those words in Halliday’s description that look like possible models for
Nigel’s productions.
‘Powder’ is questionable and is thus indicated with a question mark.
‘What’s that?’ is my own guess for utterance numbers 16, 17 and 18 and is
therefore not included. Were this a diary by a less careful, nonlinguistic
observer, the claim would probably be that Nigel had acquired nine or ten
words. Probably over 60 percent of the words noted by Halliday would not
be proposed. This reflects Halliday’s claim that these are proto-words, that
is, created by the child to express itself, not adult words acquired from
English. Even the italicized words in the table, Halliday would claim, are
taken by the child and given their meaning. The meanings of proto-words
do not come from the adult language. A micro-analysis is highly suggestive
of intentional communication earlier than normally expected.
Another feature of Halliday’s analysis is that a few vocalizations serve
more than one function at NL 3. Examples are utterance numbers 18 and
19, and 27 and 28. Also, we have pairs that differ in intonation, e.g.
utterances 7 and 25, 4 and 26. These data represent two claims: that the
function for a limited number of proto-words may vary, and that the use of
tone may be contrastive at a very early age.
In NL 6, just six weeks after NL 5 , Nigel entered what Halliday labels
Phase ZI. This phase has the following features: (i) there is a word spurt -
Nigel’s vocabulary grew from 52 words to 145 words in this short period;
(ii) Nigel begins to enter into dialogue; (iii) shortly after, there is the onset
of syntax in the form of multiword utterances; and (iv) there is a change in
Table 6.17 A summary of Nigel’sproto-language at N L 3 (or 13imonths), based on Halliday (1975: Fig. 3)
Utterance no.
Proto-word (and tone) Gloss Meaning
Instrumental
1. %--- (mid) ‘give me that’ initiation of a general demand
2. yi (high level) ‘yes I want that’ response to a general demand with object present
3. a: (high rise-fall) ‘yes I want what you just offered’ response to a general demand for a service or
nonvisible object
4. bwga(-) (mid) ‘I want some powder’ specific demand for powder
bug(-) (mid)
5 . tRa(-) (mid) ‘I want (to go and get) clock’ specific demand for clock
tR3(-)
Regulatory
6. a;3;J (mid) ‘do that (again)’ normal imitation of a general command
7. mi
j (wide; ff) ‘do that right now!’ intensified imitation of a general command
8. 5-- (low) ‘yes (let’s) do that’ positive response to a general command
9. 2 2 (midon both) ‘no don’t (let’s) do that’ negative response to a general command
10. 3--- ‘let’s go for a walk’ specific command to go for a walk
11. pwi---; pew (high level) ‘let me play with the cat’ specific command to play with cat
Interactional
12. na; an:a (high level) ‘Anna!’ greeting Anna
13. da; dada (high level) ‘Daddy!’ greeting Daddy
14. ?e: (longlow) ‘yes it’s me’ response to interaction
‘yes I see’
15. 0 (low) ‘don’t be cross with me’ response to regulation or reproof
16. a::da (high rise and mid fall) ‘look, a picture; you say what it is’ normal initiation of object-oriented interaction;
[what’s that?]
17. a::da (high rise and mid fall) ‘another picture; now you say what that subsequent initiation of object-oriented interaction
one is’ [what’s that?]
18. Z(de-)dZ: ‘nice to see you, shall we look at this?’ initiation of person-oriented interaction [what’s
proclitc and (high level and high fall) that?]
Personal
19. z(de--)dZ: (mid) ‘look, that’s interesting!’ expression of personal interest
20. da (midlow) ‘a dog!’ specific interest in a dog
21. ba (mid low) ‘birds!’ specific interest in a bird
22. ba (midlow) ‘a bus!’ specific interest in a bus
23. 0e”cz (mid low for both) ‘an aeroplane’ specific interest in an aeroplane
24. eYi: (mid) ‘that’s nice’ expression of pleasure
ZYi:
25. m5i (high rise-fall) ‘that’s funny (look where it’s gone!)’ expression of surprise
26. bwga(-) (low fall) ‘a lot of talk’ expression of disgust
bug“(-)
27. gwyi--- (narrow low) ‘I’msleepy’ expression of desire to withdraw
Imaginative
28. (same as no. 27) ‘let’s pretend to go to sleep’ pretend play
29. bwe--- (high sung) ‘tra la la’ song
--- indicates repeated syllable; (-) (--) indicates number of optional repetitions; [ ] indicates my own guess.
functions. The six functions of Phase I merge into two general functions,
the mathetic and the pragmatic.
Some of the features of Phase I1 are part of the period of early word
combinations, so we will leave a discussion of them until the next chapter.
We can see, however, that Phase I contains very specific claims about the
nature of single-word utterances. To summarize, they express a set of basic
communicative functions that have the child as their source. Further, the
phonetic form of these is taken from the child’s vocalization, not borrowed
from the adult language. These words are referred to as proto-words, and
the child’s language during this period as a proto-language.
Several of the investigators cited in Table 6.14 have noted some of the
same features of the child’s early intentions. Carter (1979) refers to
proto-words as sensorimotor communicative schemas, preferring to use
terms that map the child’s phonetics onto schemata, the Piagetian term for
the child’s structures at this stage. Dore et al. (1976) prefer to focus on their
unique articulatory structure. They call them phonetically consistentforms
(PCFs), i.e. child vocalizations that are not adult words in the normal sense,
but which are fairly consistently produced, with some phonetic variation,
and which appear to have intentional meaning. With some minor discrep-
ancies, these terms appear to be interchangeable. The real differences, it
seems, are in the theories that underlie each.
Greenfield & Smith (1976) The study by Greenfield & Smith presents a
slightly different description of the child’s early vocalizations, and demon-
strates how susceptible these utterances are to variable interpretations. To
understand their proposals, it is important to review the data which they
analyzed. They studied two children, Matthew and Nicky, over several
months. The data consisted of diary notes and several formal sessions with
each child during which language sampling took place. Table 6.18 gives the
formal sessions, ages, and distribution of single-word and multimorphemic
utterances for both children during those sessions.
In examining Table 6.18, we can see that the data on single-word
utterances are taken from both the period of single-word utterances and the
period of early word combinations. In fact, most of the data on the earlier
period is from Matthew, sessions I-V. Only Nicky’s first session has no
instances of multimorphemic utterances - it is important to realize this in
comparing their analyses to others like Halliday’s, who was looking only at
the period of single-word utterances in proposing his functions for Phase I.
Further, given that at least Halliday sees Phase I1 (or roughly speaking, the
period of early word combinations) as qualitatively distinct, we need to be
cautious in expecting single-word utterances after the onset of multiword
utterances to be the same as those before them. I have argued elsewhere
Table 6.18 The formal sessions for Nicky and Matthew, giving their ages and
number of single-word and multimorphemic utterances at each, based on
Greenfield & Smith (1976: Tables 2, 3, 5)
~~
Matthew Nicky
No. of utterance9 No. of utterancesa
Session and age Single-word Multimorphemic Session and age Single-word Multimorphemic
I 1;0(15)(22) 13
I1 1;2(10)(18) 25
111 1;3(5)(17) 66
IV 1;4(2)
V 1;5(13) 83
I 1;6(4) 66
VI 1;6(18) 99 12
I1 1;6(27) 130 I
VI1 1;7(21) 91 17
111 1;7(29) 96 38
IV 1;8(23) 98 42
VI11 1;8(26) 108 72
V 1;9(17) 118 28
IX 1;10(1) 32 87
VI 1;10(21) 170 80
VI1 1;11(21) 119 119
VI11 2;0(23) 42 32 1
‘These are token frequencies; no information is given on types of words within these categories.
(Ingram 1979b) that overlooking this fact has led to a misunderstanding in

some places of some of the claims in the literature on the nature of
single-word utterances.
A second feature of the data in Table 6.17 is that the numbers represent
number of utterances, not vocabulary, which is presumably much smaller.
Further, frequent use of individual words may accentuate the numbers in
any session. Unfortunately, Greenfield & Smith do not provide any
overviews of the children’s receptive and productive vocabularies.
Greenfield & Smith propose 12 semantic functions for the children’s
one-word utterances. They also give an order of acquisition for these, based
on both the diary data and the formal sessions. The latter point is important
since some of these (e.g. Agent and Dative) appear early in their diary data
but are not attested in the formal sessions until much later. Here, I will only
present those relations which occur in at least five utterances during the
sessions when the children are in the period of single-word utterances (i.e.
sessions I-V for Matthew, and I, I1 for Nicky. Session I1 is added to expand
Nicky’s data for this period even though a small number of multiword
utterances occur).
Greenfield & Smith find that their 12 semantic relations are acquired in a
Table 6.19 The first semantic relations acquired by Nicky and Matthew
during the period of single-word utterances, adapted from the data in
Greenfield & Smith (1976: Tables 8, 9, 10, 11) and the subsequent text
PERFORMATIVES
Performative: use of nonstandard sounds to accompany actions; e.g. Matthew
used ‘dat’ at 0;8(6) to accompany clapping and playing pat-a-cake.
Later, more recognizable words would be used in similar ways.
Indicative Objects: calling attention to objects by naming them, usually accompanied
by pointing; e.g. Nicky at 1;0(7)would use ’dada’ as he went around
the house and pointed to various objects.
Volition: appears after the above function. It occurs when the child wants
something, and focusses on the desired object; e.g. Nicky’s use of
‘mama’ at 1;1(19) for general requests; Nicky’s use of ‘no’ to
express desire not to go to the backyard at 1;6(4).
Volitional Object: expression of desire for an object by naming the object; e.g.
Matthew at 1;5(13)says ‘bottle’with a whine, looking for his bottle.
ACTION-ENTITY RELATIONS
Action or State of Agent: reference to an action or state that requires an agent; e.g. Nicky at
1;6(27) says ‘down’ as he gets down from a chair; Matthew says
‘night-night’ at 1;2(5) while getting ready for bed.
Object: reference to an object which is directly involved in an action or
change of state; e.g. Nicky at 1;4(19)says ‘fan’ wanting the fan to be
turned on; at 1;1(0) Matthew says ‘ball’ after throwing a ball.
Action or State of Object: focus on the change of the object, rather than the initial or resultant
state; e.g. Nicky at 1;6(27)says ‘down’ after pulling his train down,
where the focus is on the state of the train; Matthew at 1;4(17) says
‘dirty’ in reference to milk in a dirty baby bottle.
sequence consisting of four general groupings. The first two of these groups
are acquired by Matthew and Nicky while they are still in the period of
single-word utterances. The first group of semantic functions acquired are
performatives. The second group are functions that show relations between
Actions and Entities. Table 6.19 gives definitions and examples of the
semantic relations that fit in each of these two general groupings, subject to
the restrictions cited at the end of the previous paragraph.
The first group of performatives look similar to Halliday’s categories in
the sense that they may consist of nonstandard vocalizations, and that they
are pragmatic, i.e. they focus on the purpose of the speech act. An exact
pairing of the two systems, however, is difficult. The Action-Entity
relations move from speech acts to the content of the child’s utterances. The
four relations shown are based on the assumption that the child has an
awareness of the concepts of Agent, Action, State, and Object. Given that
these categories follow the Performatives, their appearance is occurring
around the end of the period of single-word utterances, and it continues into
the next period.
The Action-Entity relations resemble to an extent the semantic relations
that appear to be understood at the end of this period, based on the
evidence presented in the previous section. The Object and Action or State
of an Object function could be related to the child’s ability to understand
Agent+Action sequences, yet Greenfield & Smith (1976) have an Action
or State of Agent. When we look at the examples used by Greenfield &
Smith, however, we find that the words the child uses are usually Action or
State words. One could relate these single-word productions to the com-
prehension of Action words, as defined in Table 6.12.
The second discrepancy concerns the Possessor-Possessed relation. Both
Huttenlocher and Miller et al. found evidence for the comprehension of this
semantic relation during the period of single-word utterances. Greenfield &
Smith, however, do not include it among their 12 semantic relations. There
are, however, two relations that look comparable. These are:
(a) Object associated with another Object or Location: this involves
naming one object in relation to another, e.g. Nicky at 1;7(29) says
‘apple’ while holding the refrigerator door; Matthew at 1;2(29) says
‘caca’ (cookie or cracker) while pointing in the direction of where
cookies are kept.
(b) Animate Being associated with Object or Location: when a child points
to an object and names its absent owner, e.g. Nicky at 1;6(19) says
‘Lauren’ while pointing at her empty bed; Matthew says ‘fishy’ at
1;3(19) while pointing at the empty fish tank.
The (b) example is like the Possessor-Possessed relation defined in the
comprehension task described in Table 6.13. The task itself does not
separate ‘association with’ from ‘possession’. The first cases of Animate
Being associated with Object or Location appear for Nicky at sessions IV
and V, but never become frequent in the subsequent sessions. For Nicky,
there are two instances at 1;6(27), but it only becomes frequent at 1;8(23).
The emergence of some form of a Possessor-Possessed relation, therefore,
is only weakly supported by the Greenfield & Smith data.
There is, however, some other support for the production use of
something like Possessor-Possessed in the data reported by Rescorla.
Recall that Rescorla defined one kind of overextension as a predicate
statement (see Table 6.8). Predicate statements look quite similar to the
function in (a) and (b) above. Rescorla notes the following (p. 330):
‘for the four children using some word combinations by the end of the
study, the peak in new predicate statement applications occurred in the
month preceding the onset of word combinations’. Such data can be
taken as indirect evidence of some parallels between the child’s recep-
tive and productive ability at the end of the period of single-word
utterances.
The data from studies like Halliday’s and Greenfield & Smith’s suggest
some general characteristics of early meaningful vocalizations. The first
ones may not look very much like adult words in the language, yet they
appear to express speech acts of wishing and noticing objects and events in
the environment. As the vocabulary expands to include more identifiable
adult words, the child’s words become more classifiable semantically. The
classification of these utterances into semantic categories appears to bear
some resemblance to the semantic categories that children comprehend
around the same time.
The theoretical implications of the specific proposals by the researchers
cited in Table 6.15 is critically discussed in Atkinson (1982). One concern of
Atkinson’s is that the taxonomies proposed are often not related to any
theory of language or, in most instances, theory of speech acts. Halliday is
an exception, in that he does have his own functional theory of adult
language. As we will see in Chapter 7, his functional categories must
undergo relatively extensive restructuring in Phases I1 and 111. The Green-
field & Smith semantic functions are not related to any theory of adult
language, and so they lack any explanation of how these lead to the
grammar of English. Another concern of Atkinson’s is at the other end of
development, that is, the relation of these taxonomies to some explanation
for their appearance. On this issue Greenfield & Smith fare better in that
they attempt to relate the semantic functions to earlier acquired concepts
and cognitive developments. It is less clear where Halliday’s functions
originate.
This discussion has concentrated on description, isolating what looks like
important observations. These observations will need to be part of any
attempt to explain the developments of this period. At the end of the
chapter I will address generally the Constructionist Assumption in regard to
the child’s ability to understand and produce language during the period of
single-word utterances. Before that, however, we need to consider the
possible effects of the child’s linguistic input during this period, as well as his
phonological development.
6.5 The onset of phonemic perception and production

The child has begun to develop a semantic system during the period, and
by its end shows some evidence of comprehending multiword utterances,
I
at least in terms of their major semantic categories. In this section we turn
now to the child’s emerging phonological system. The discussion will
begin with a look at perceptual development, followed by a treatment of
the phonological characteristics of the child’s first 50 words or so in
production. ,
6.5.1 Child speech perception

In Chapter 5 we began by defining infant speech perception as that ability of
the infant to perceive meaningless acoustic stimuli. The stimuli are
meaningless in that the infant does not yet see these sounds as parts of
words. Child speech perception, however, is operationalized here to refer
to the child’s ability to perceive speech sounds that are part of what the child
identifies as a word. Barton (1980: 87) discusses this same distinction with
the term phonemic perception. There are two components to phonemic
perception. One is the ability to discriminate speech sounds, that is, to hear
them as distinct. The other is the ability to classify the sounds discriminated
into phonological categories. The research with infants indicates that the
first of these abilities is quite developed in the first year of life; the second,
however, adds a great deal of complexity to the child’s task.
We have suggested earlier that the infant’s perceptual ability and
developing cognitive ability enable him at around 1year of age to recognize
words in the linguistic input. While this milestone, referred to as the
Principle of the Linguistic Sign in our discussion of Stern, is a unique genetic
feature of humans, it appears to be dependent in part on earlier develop-
ments. Its operation becomes particularly noticeable in comprehension
during this period, where the vocabulary grows to an impressive size before
much productive vocabulary appears. Given the size of this receptive
vocabulary, it seems reasonable to propose that phonemic perception is
occurring during this period. That is, the child is beginning to identify and
categorize the speech sounds in his words into linguistic categories.
We can briefly explore possible explanations of the child’s speech or
phonemic perception ability at this point in much the same way we did in
Chapter 5, section 5.3.1. First, we could propose a universal-maturationist
view. In its strongest form, this view states that the infant’s precocious
discrimination ability, combined with some linguistically programmed
milestone such as the Principle of the Linguistic Sign, triggers the innate
ability to develop phonological categories. We would expect then that the
child’s perceptual phonemic ability would be quite good, with little need for
development. Linguistic experience would be necessary to set the phonolo-
gical categories of the language being acquired. Assuming the child is
exposed to the range of speech sounds in the language, combined in words
with simple syllabic structure, then the child’s ability should be well-
developed after the acquisition of a relatively small vocabulary.
The attunement-refinement theory of Chapter 5 makes different claims.
It states that the child may begin with some initial structures as part of the
genetic program, but that these will be built upon through linguistic
experience. The child, in other words, needs to discover the phonology of
the language being acquired. As with infant speech perception, this theory
makes specific predictions about the emergence of phonemic perception:
(i) the development will go through stages, where the categories discovered
at one stage are built upon into greater differentiations at later stages;
(ii) there will be an effect of experience. Basically, the child who is exposed
to a phonological system of a certain kind will develop perceptual phonemic
abilities that will not be shown by children acquiring a different phonologi-
cal system.
Unfortunately, there has been little research on child speech (or phone-
mic) perception as compared to that on infant speech perception, and what
has been done is mostly on children in later periods than the one under dis-
cussion. The resolution of these predictions, therefore, will require future
cross-linguistic research. There is, however, one study that has pioneered
this area, that of Shvachkin (1948/73). In fact it remains today as one of the
truly classic studies in language acquisition. Because of its richness, in
method, description and explanation, we will look at it in some depth.
Shvachkin (1948/73) Shvachkin is just one of several Russian psycho-

linguists who have worked on language acquisition. The ‘Further reading’
section at the end of this chapter gives a guide to the relevant literature.
Method. Shvachkin’s (1948) paper would be of value even if he had
done no more than develop a method to test children’s speech perception.
The problems involved with testing comprehension with children around
1 year are formidable; they are increased when we attempt to focus on the
perception of individual segments. Shvachkin’s own remarks on this
problem are insightful. He states (p. 99; all page references for Shvachkin’s
work are from the translation in Ferguson & Slobin 1973): ‘. . .it was neces-
sary to work out a method which would correspond to the actual course of
development of phonemic perception in the child. This problem proved to
be quite difficult and required a great deal more time and effort than the
actual study of the facts themselves’ (my emphasis: DI).
When linguists seek the existence of phonemes in languages, one of the
primary methods is to find minimalpairs, i.e. words that differ in only one
segment, for example ‘pit’ vs. ‘bit’, or ‘tea’ vs. ‘toe’. When pairs like these
are found, it is concluded that the sounds that differ are contrastive, and the
pair itself is a contrast or opposition. For example, English has a contrast (or
opposition), of [p] vs. [b] or [i] vs. [o]. We can further conclude that the
sounds that contrast are phonemes, and can be shown in slants (e.g. /p/ vs.
/b/ and /i/vs. /o/).Also, it is proposed that the contrast is not just marked by
sounds, but by the distinctive feature in the opposition, e.g. [+voice] or
[hound]. Our goal in studying the phonemic perception of the child is to
see how the child recognizes oppositions in the language being acquired.
Table 6.20 Six steps followed by Shvachkin (1948173: 100-1) testing for
phonemic perception
1. Day 1. Spend time with the child teaching a nonsense word, e.g. ‘bak’, until the child shows
evidence of knowing the word.
2. Day 2. Teach the child a second word, e.g. ‘zub’, until the child shows evidence of learning
the word.
3. Test for non-minimal opposition:” Ask the child to respond to two nonsense words that have
non-minimal oppositions, e.g. ‘bak’ vs. ‘zub’. Do this until the child shows evidence of
hearing a difference.
4. (Next day? Time unclear). Teach the child a third nonsense word that contains a minimal
opposition to one of the earlier words, e.g. ‘mak’.
5 . Testsfor new non-minimal opposition:”Ask the child to respond to the newest word taught
with the second word taught, e.g. ‘mak’ vs. ‘zub’.
6. Test for minimal opposition: Place the objects for all three nonsense words in front of the
child. Test for perception, particularly in the cases where the child must perceive a minimal
difference, e.g. ‘mak’ vs. ‘bak’.
Examples in the text suggest that occasionally minimal oppositions would be tested in
step 3.
Shvachkin’s proposal is that we seek minimal pairs in the child’s per-

ception in order to be able to say with any confidence that a contrast exists.
For example, suppose we determine that the child hears the difference
between ‘pig’ and ‘cake’. Since the words differ in all three segments, we do
not know if the child processes all three segments, or focusses on only one.
We can conclude something, however, by showing the child perceives ‘tea’
vs. ‘toe’.
Shvachkin’s method was to teach the young child who is at the onset of
acquisition nonsense words that are minimal pairs. Each nonsense word was
assigned to an object that the child did not know. He used mostly solid
geometric figures such as wooden pyramids and cones. The steps that
Shvachkin followed are presented in Table 6.20. His method is strikingly
similar to Piaget’s clinical method discussed in section 5.4.4. He worked
with the child until he was confident the word was acquired, and adjusted
the teaching to the interests of the child.
A crucial part of this procedure is knowing that the child hears a minimal
contrast. Shvachkin was very aware that children may have contextually
based knowledge, that is, ability to do something in one context but not in
another. He used six methods, therefore, to test for the comprehension of a
test pair. The child would only be given credit for the acquisition of an
opposition if he succeeded in at least three of these methods. These
methods are given in Table 6.21. Unfortunately, Shvachkin does not
Table 6.21 Six methods used by Shvachkin (1948173: 101-2) to test
phonemic perception of a minimal contrast
1. Pointing to the object: The child would be asked to point out the object among a selection of
objects. The position of the object would be altered systematically.
2. Giving of the object: The child would be asked to hand the object to the experimenter,
under similar circumstances as in 1.
3. Placement of an object: The child would be asked to place several objects in different places
in the room.
4. Finding the object: The child would be asked to find one of the objects.
5. Operation of one object in relation to another object: The child would be asked to seek out
the object to put one object upon the other.
6. Substitution of objects: The child would be asked to get an object in a particular place, but
another object would be there instead. The child was observed to see if there was a reaction
of disappointment.
provide information about the number of times in a row that the child must
get a contrast correct before being assigned success on the method, a
characteristic of the clinical method.
There was apparently a long period of pilot testing until a word list was
decided upon to test the emergence of Russian contrasts. This word list
deservesa brief mention. It consisted of monosyllables that were CVCor VC.
The contrasts tested were: (i) between vowels in CVC syllables when the
consonants were the same; (ii) between initial consonants where the vowels
and final consonants were the same; (iii) CVC or VC where the vowel and
final consonants were the same. In other words, he worked with a very speci-
fic set of contexts, not looking at final consonants or multisyllabic words. This
point will be important when we look at the generality of his results.
Description. Shvachkin’s study had two phases to it. In the preliminary
phase, he worked carefully with individual children to establish what
appeared to be the earlier acquired contrasts. These results were then
checked in a more extensive study on other children, bringing the total
number of children studied to 19. All children were studied longitudinally for
approximately six months. The following is a profile of Shvachkin’s subjects:
Shvachkin’s subjects
Age at onset: range 0;10-1;6, mean: 1;3
Age at end: range 1;O-2;0, mean: 1;9
Sex: 14 girls, 5 boys
Unfortunately, Shvachkin gives us little specific information on the chil-
dren’s language abilities. If the children were normal, the mean age of 15
months at the onset suggests that they would only be producing single-word
Table 6.22 Shvachkin’s stages in the acquisition of Russian phonemic
contrasts, adapted from Shvachkin (1948173: 124, Table 2)
StagelDistinction Substages and contrasts
1. Between vowels 1. a vs. other vowels

2. i-u, e-o, i-0, e-u
3. i-e, u-o
2. Presence vs. absence of consonant e.g. bok-ok, vek-ek
3. Sonorants vs. articulated obstruents’ e.g. m-b, r-d, n-g, j-v
4. Palatalized vs. non-palatalized consonants e.g. n-n‘, m-m’, b-b’, v-v‘, z-z’, I-l’, r-r‘
5 . Distinction between sonorants 1. Nasals vs. liquids and /j-/
e.g. m-I, m-r, n-I, n-r, n-j, m-j
2. between nasals m-n
3. between liquids 1-r
6. Sonorants vs. non-labial fricatives e.g. m-z, I-x, n-i
7. Labials vs. non-labials e.g. b-d, b-g, v-z, f-x
8. Stops vs. spirants e.g. b-v, d-i, k-x, d-i
9. Velars vs. non-velars e.g. d-g, s-x, 5-x
10. Voiced vs. voiceless consonants e.g. p b , t-d, k-g, f-v, s-z, S-i
11. Palatal vs. alveolar fricatives e.g. 2-2, S-s
12. Distinction between liquids and j- e.g. r-j, 1-j
a These are those obstruents that can actually be produced by the child at the time of
testing. Their membership will thus vary from child to child. See discussion in text
under Explanation.
utterances at the beginning of the study. The 21-month final age in turn
suggests that they were probably beginning to produce multiword utteran-
ces by the end. The predominance of girls in the study is an interesting fact
that Shvachkin does not comment upon.
It is helpful to look at some results from the preliminary phase before
presenting the general results. The first discrimination Shvachkin found was
between vowels. The data for this were as follows:
Success Failure
Nina K. ‘kot’ vs. ‘kit’ ‘bak’ vs. ‘mak’
(1;3) ‘zuk’ vs. ‘zak’ ‘bak’ vs. ‘pak’
Zina P. ‘kot’ vs. ‘kit’ ‘bak’ vs. ‘pak’
(M) ‘zuk’ vs. ‘zak’
‘bak’ vs. ‘mak’
Nina could only successfully respond to contrastive pairs that differed only
in their vowels. She failed both ‘b’ vs. ‘p’ and ‘m’ vs. ‘b’. Zina, on the other
hand, could succeed in all the cases that Nina could, plus the ‘b’ vs. ‘m’
opposition.
Next, Shvachkin wanted to see when the ability to perceive CVC was
acquired in relation to the above sequence. To do this, he gave Nina K. the
words ‘os’,‘rnos’, ‘pos’, ‘bos’. She succeeded on ‘mos’ vs. ‘os’,‘bos’ vs. ‘os’
and ‘pos’ vs. ‘os’, but failed on ‘mos’ vs. ‘bod and ‘mos’ vs. ‘pos’. He con-
cluded then that the contrast of presence vs. absence of consonants occurs
after differentiation of vowels and before the differentiation of consonants.
In this way, he proceeded to develop a series of stages for the order in which
contrasts are perceived by Russian children. These stages were developed
in their final form after the data from all 19 children were analyzed.
Table 6.22 gives Shvachkin’s results on the stages his children followed.
To understand this table, it is useful to take a brief look at the phonemes of
Russian, based on the appendix in Shvachkin. The vowels and non-palata-
lized consonants are given below:
Vowels Consonants
i, e , a , 0 ,u P t k
b d g
f s s
v z i x
m n
1
r
W 1
Each of the consonants, except j , s, z, 5, and i has a palatalized counterpart
(for a description of palatalization, see Ladefoged 1975). It is a complex
system, and there is controversy concerning its exact nature. For example,
other descriptions of Russian include a ts and E , but Shvachkin makes no
reference to these in his data. Importantly, Shvachkin claims that children
show the acquisition of all the Russian contrasts by age 2 years at least in
simple CVC or VC syllables, and in initial position for consonants. The
findings show relatively rapid phonemic perception between 1 year and 2
years that parallels the rapid phonetic perceptual development between
birth and 1year of age. They do not, however, show that perceptual acqui-
sition is complete at 2 years; rather, they indicate that the basic perception
of Russian oppositions is available to the young child to use in later develop-
ment of more phonologically complex words. As we shall see in the discuss-
ion of production, most child words during the period of single-word utter-
ances are simple CV and CVC syllables.
We can get some ideas of age in relation to these stages by reworking
some of Shvachkin’s data. In his Table 1, he gives data showing that six of
the children only achieved stage 7 by the end of the study. These six children
ranged in age from 18to 21 months at the end of the study, with a mean of 19
months. Again, if we assume the children were normal, this would be about
the time that most children begin to produce multiword utterances. The
data suggest that children begin the phonological analysis of words percep-
tually during the period of single-word utterances, and that they develop
many of the basic contrasts (presumably in Russian children at least to
stage 7 ) .
The sequence of acquisition for Russian in Table 6.22 merits two specific
comments. First, observe that the distinction of palatalized vs. non-palata-
lized consonants occurs relatively early, in stage 4.This distinction does not
occur in English, and it might appear to the English speaker learning
Russian to be a difficult one. We need to be careful, however, not to be
Anglomorphic in our ideas about what will be difficult for a young child who
has not yet acquired language. For example, tones in Chinese are con-
sidered difficult to hear by many English speakers yet Chinese children
acquire them quite early (Tse 1982). Second, the voiced-voiceless oppo-
sition is very late (stage lo), yet we know from the infant perception studies
reviewed in the previous chapter that it is phonetically perceived early. We
will want to look for an explanation for this late appearance. As we will see,
Shvachkin has his own explanation to offer.
Explanation. Before looking at Shvachkin’s explanation of his results,
consider his use of ‘stage’. We need to extract his meaning for the term from
Table 6.22 since he offers no definition. If we compare the stages in 6.22 to
the kinds of stages presented in Chapter 3, several appear to be co-
occurrence stages, defined by a new behavior (Figure 3.6). Compare, for
example, stages 1 and 2. At stage 2, the vowel distinctions from stage 1
remain, and a new behavior is added, the ability to recognize the presence
or absence of an initial consonant. The new behavior is not directly the
result of an earlier one, and there is no logical relation between them. We
need, therefore, some principle to relate the two behaviors. This would
make the stages descriptive, in the sense of section 3.3. Shvachkin, in fact,
offers some principle to relate these behaviors, but to make them explana-
tory we will also need some independent evidence for the principles.
Shvachkin also makes a preliminary attempt to do that; we will consider his
proposals below.
Other stages of Table 6.22 are implicational stages in the sense of Figure
3.8, of Chapter 3 . Compare, for example, stages 2 and 3. The new behaviors
of stage 3 are not independent from stage 2. In fact, the occurrence of a
distinction between sonorants and stops and labial fricatives implies the
awareness of the presence of consonants (a stage 2 behavior). The sequenc-
ing of implicational stages does not require principles to relate the change
and is in a sense uninteresting. An implicational relation also exists between
stages 3 and 5 . Most of the relations between the stages, however, are
non-implicational and therefore require an explanation. For example, why
should the distinction between palatalized and non-palatalized consonants
precede that of voiced and voiceless consonants?
Shvachkin begins his explanation of phonemic perception by contrasting
two general periods or stages, the prephonemic period and phonemic
period. The prephonemic period is not what it might appear to be, that is, all
the development that occurs between birth and the first words; instead, it
refers to the period when the child first assigns meanings to sounds. The
child does not analyze the parts of the vocalization but maps meaning
directly onto the word. Shvachkin states (p. 94):
The unique phonemic features of the initial period in child speech
development correspond to the unique semantic features . . . it is not
the phoneme (a phonetic unit of speech) but the intonation, the
rhythm, and later a general sound picture of words which bear a
semantic load at this stage.
From the examples, this period occurs around 6 months to 1 year and
characterizes the child’s understanding of its earliest words. There are two
features to the stage: there is a correspondence between semantic differ-
ences and phonetic differences, and there is direct mapping of meaning onto
a general phonetic characteristic of the words. His use of ‘semantic
differences’ looks similar to Piaget’s discussion of ‘symbol’.
The phonemic period is the time when the child acquires the phonemic
contrasts of the language, as shown in Table 6.21. Shvachkin directly
addresses the possible cause of this change, which he argues to be the
acquisition of the semantic system of the language. He states (p. 96):
Thus, under the influence of semantic change the child moves toward
the phonemic perception of speech, which is connected with a radical
reconstruction of both articulation and speech perception. The
elements of this transformation are seen in the beginning of the second
year of life.
Linguistic organization, according to Shvachkin, begins very early, starting
with semantics.
There are a number of crucial aspects to Shvachkin’s proposal. One is
that he is proposing a radical reorganization around age 1 year. The
transition between prephonemic and phonemic development is one of
discontinuity, and there is no attempt to relate the developments of the
phonemic period to those of the prephonemic period. Ironically, his stages
for the phonemic period are developmental and imply a Constructionist
Assumption. This is not true for the onset of phonemic perception,
however, which would need some innate principles unrelated to previous
development, This is similar to Stern, who argues for a developmental
perspective and yet proposes a series of discontinuities.
Another aspect of this claim is that the proposal of semantics as the cause
of the onset of phonemic perception is not developed. It seems to me that
the claim is that the child’s receptive vocabulary grows and the semantic
system becomes organized in some way that triggers phonemic organi-
zation. We could reinterpret Shvachkin and argue that the Principle of the
Linguistic Sign precedes and triggers another linguistic principle. Let’s call
the latter the Principle offhonemic Organization. If so, we should realize
that the relation between the two is implicational; that is, a phoneme is a
unit that speakers use to differentiate words that differ in meaning. This
raises the question, then, as to whether or not there is any way to test this
claim.
I would like to propose the following. While the Principle of Phonemic
Organization implies the Principle of the Linguistic Sign, there is nothing in
this implication that requires the former immediately to follow the latter.
The child could function for at least some time with a notion of linguistic
signs without yet phonemically organizing words. Shvachkin’s claim,
however, is that the two occur closely together, with phonemic perception
developing gradually. If this claim is true, one should expect evidence of
early phonemic perception in both perception and production. Shvachkin is
very clear on this point. He believes that the child begins acquiring a
psychological representation of words that underlies his perception and
production.
The onset of the ability to organize phonetic segments into psychological
units or phonemes (the Principle of Phonemic Organization) only accounts
for the onset of the phonemic stage. Shvachkin still needs to account for the
sequence of stages in Table 6.21. Unfortunately, his account of this is not
always clear. A major reason for this is that his stages reflect a process of
differentiation: the child differentiates the most general classes of sounds,
and gradually discriminates between lower classes. To do this, however,
requires some theorizing about what are the natural classes in language and
their defining or distinctive features. Nowhere does Shvachkin discuss this
question. He could have proceeded in one of two ways: he could have taken
a proposed theory of phonological features and tested its predictions (a
study in Language Acquisition); or he could have analyzed his results into
classes and proposed a set of natural classes based on his results (a study of
Child Language). If anything, Shvachkin appears to be doing the latter.
Let’s assume that Shvachkin is proposing that there is a sequence of
acquisition that follows from some theory of phonological features. The
child will discriminate down the set of features, going from the most
general to the most specific. This is a viable proposal, and one that could
be seriously pursued today, given the range of theories we now have on
phonological features. Shvachkin’s own classes are in some cases unusual.
For example, in stage 3 he distinguishes between articulated and non-
articulated consonants. The examples in his table contain stops and labial
fricatives, which are not normally considered to be a natural class. For
reference, let’s call this proposal the Principle of Phonemic Differen-
tiation.
Shvachkin argues that differentiation is directly affected by two factors:
perceptual salience and articulation. Neither of these is discussed satisfac-
torily, but his remarks are suggestive. Perceptual salience means that some
distinctions are easier because they are in some sense easier to hear; he uses
this factor to explain why vowels are discriminated first. Apparently two
aspects of vowels lead to their salience: they are the loudest or most audible
segments and they are frequent (Le. in Russian they are about five times
more frequent than consonants). We are not given any suggestions,
however, on how these two factors interact. He uses these same factors to
explain the early differentiation of /a/ from other vowels.
Importantly, Shvachkin attempts to provide independent evidence in the
form of patterns found in adult language. For example, he cites languages
that only have a vocalic difference between /a/ and other vowels, and others
that only have /i/, /a/, /u/. The point appears to be that the stages the child
goes through should correspond to possible systems in the adult language.
This is very similar to White’s condition (1982) that the child’s language at
any stage should be a possible human language. Shvachkin does not,
however, provide evidence of this kind for his stages.
The second factor that aids differentiation is articulation. The point
appears to be that sounds that are easier to articulate tend to be discrimi-
nated earlier. He uses this factor to explain why children could at stage 3
discriminate between sonorants and so-called articulated consonants.
Examples of articulated consonants are /b/, /d/, /g/;of non-articulated /z/,
/a,/x/. (He is unclear in regard to /v/, treating it sometimes as articulated,
sometimes as non-articulated.) He says (p. 114):
The facts noted above reveal an interaction between hearing and

articulation in the process of phonemic development. Discussing
whether hearing determines the development of articulation or articu-
lation determines the development of hearing, the ontogenetic facts
speak for a two-way interaction of hearing and articulation.
Given the gap between receptive and expressive vocabulary, one would
expect this factor to become greater as development proceeds. It is not the
only factor, however, in that he acknowledges that there are cases where the
child will be able to discriminate but not produce a difference.
There is one last factor that Shvachkin mentions briefly (p. 115), the role
played by the language that the child is acquiring:
The Russian child discriminates palatalized and non-palatalized con-
sonants comparatively early; however, this cannot be explained
simply in terms of articulation and hearing. Here the linguistic
peculiarities of these phonemes in the Russian language manifest
themselves. First, in Russian (unlike English, French and German)
this distinction is of semantic significance. Second, the frequency of
alternation of palatalized and non-palatalized consonants plays a role
in Russian. Peshkovskiy’s data (1925) show that in most cases the child
encounters one palatalized consonant for every two non-palatalized
consonants.
If true, this shows a very early effect of the child’s linguistic environment.
We already discussed this possibility for infant speech perception. Here, we
see the suggestion of the development of an early phonological system that
is determined in part by the frequency of phonological oppositions in the
adult language. I will refer to contrasts that are prominent in a language as
having phonological salience. Such influence is, in fact, expected from a
theory of acquisition where the child’s linguistic system is the result of the
constant interaction of assimilation and accommodation. It is also in this
sense that Shvachkin’s study is Piagetian in nature.
6.5.2 Child speech production

Data from Shvachkin’s study suggest the onset of phonemic speech per-
ception during the period of single-word utterances. While this may be a
controversial claim, it is not an implausible one, given the size of the child’s
receptive vocabulary. The alternative would be that the child recognizes
and stores its 300 or so words by isolated acoustic cues. Apart from the
difficulty that arises when one tries to specify this disparate set of cues, there
is the problem of discontinuity; that is, when does phonemic categorization
begin, if not at this time? Given the current lack of counterevidence, I will
assume that Shvachkin’s position is the most viable one to pursue at this
time. It meets both the Constructionist Assumption and the Competence
Assumption. It only partially meets the Productivity Assumption. It does so
to the extent that the child’s ability was tested in a variety of contexts; but it
does not do so to the extent that contrasts were only tested between single
lexical items, e.g. ‘bak‘ vs. ‘mak’ and not multiple items, e.g. ‘mak‘vs. ‘mik’
vs. ‘bak’ vs. ‘bik’.
The next question is the nature of the child’s produced words during this
period. First, do they show phonological organization of the kind proposed
for perception? They should, if we follow Shvachkin’s suggestion that
phonemic organization underlies both perception and articulation as a more
abstract psychological level of organization. Second, do they follow an
orderly sequence of stages of the kind proposed by Shvachkin for per-
ception? Shvachkin’s proposals can be interpreted as predicting such a
sequence. They suggest some further predictions:
(a) they will not be isomorphic to the ones in perception, since the
articulatory factors are not always the same as perceptual ones;
(b) there will be some individual variation, just as there was in the
perceptual stages; and
(c) there should be cross-linguistic differences, since the phonological
salience of specific phonemes will differ from language to language.
This section will explore these proposals, first by examining the major
theory of phonological acquisition ever proposed, that of Jakobson
(1941168). As will be discussed later, this theory probably provided the
stimulus for Shvachkin’s research. Despite arguments by some to the
contrary, it remains the most useful theory ever proposed to account for
early phonological development. Its difficulty is that it was never developed
sufficiently, and also was never adequately tested. After a discussion of
Jakobson, I will outline a method that can be used to analyze the child’s
early words for phonological contrasts. Lastly I will present some of the
findings in recent years on the nature of the child’s early words. The current
data are preliminary, but provide some support for the proposals just cited.
Jakobson (1941/68) and Jakobson & Halle (1956) It is rare to find any
discussion of phonological acquisition that does not refer to the work of the
late Russian linguist Roman Jakobson. In 1941 he published in German a
short book (Kindersprache, Aphasie und allgemeine Lautgesetze) that
outlined a theory of phonological acquisition. This theory was initially
accepted in the 1940s and 50s without question, but recently it has gone
through a period of evaluation and criticism (see Menn 1980 for a discuss-
ion). In fact, in some places it is felt that the theory has been refuted
(Atkinson 1982: 1). Here, however, I will propose that the reports of the
death of Jakobson’s theory are greatly exaggerated; its demise will require
more elaboration of the theory to render it more testable, and better
methods to test it. We will begin with an outline of his theory, then turn to
methods of analysis in the next section.
Jakobson’s theory is most extensively described in Jakobson (1941/68;
page references are to 1968, the English translation). It is altered somewhat
in Jakobson & Halle (1956), and I discuss these changes later in this section.
Unlike Shvachkin, Jakobson (1941/68) begins with a theory of phonologi-
cal features. This theory, which is widely accepted in its most general form,
is that the phonemes of languages can be divided into a universal set of
distinctive features. (Changes in Jakobson’s theory can be frequently traced
to changes in his proposals concerning the set of universal features.) These
features can be seen as part of the child’s innate linguistic capabilities. The
child’s acquisition of a phonological system consists of his acquiring these
features in a consistent and predictable sequence.
In section 5.3.1 it was noted that Jakobson proposes a discontinuity
between babbling and language; we also noted that this discontinuity does
not appear to be supported. The counterevidence to that claim, however,
does not necessarily constitute counterevidence to his claims about devel-
opment during the period of single-word utterances. We can think of his
proposals about the child’s phonological system as linguistic constructs
imposed upon the child’s articulations.
Jakobson sees phonological acquisition as the result of the interaction of
the child’s internal structure (or innate knowledge) and the linguistic
environment. The child actively imposes structure on the linguistic input,
then alters this structure in response to the input. That the interaction is
very much like the processes of assimilation and accommodation proposed
by Piaget is shown by the following remarks, which also make clear that the
similarity is not accidental. The Genevan scholar referred to is de Saussure
(1922), who was writing at the same time as Piaget was doing his research on
infancy:
Accordingly, we recognize in the child’s acquisition of language the
same two mutually opposed but simultaneous driving forces that
control every linguistic event, which the great Genevan scholar
characterizes as the ‘particularist spirit’, on the one hand, and the
‘unifying force’, on the other. The effects of the separatist spirit and
the unifying force can vary in different proportions, but the two
factors are always present. (Jakobson 1941/68: 16)
The child’s system at any point in time, then, is the result of its internal
organization of the linguistic input and its adaptation to that input.
While he emphasizes this interaction, Jakobson nonetheless concludes
that the child constructs an invariant sequence of phonological contrasts.
This is an apparent contradiction in that one would expect some effects of
the nature of the linguistic input as a result of the interaction. He’does not
discuss when the child’s phonology of Language A will begin to look
Table 6.23 The first stage of phonological development, adapted from
Jakobson (1941148)
Substages
1. The acquisition of vowels and consonants develops from a basic CV syllable which
contains a forward articulated stop, and a wide vowel; it may appear singly, e.g. ‘pa’, or
reduplicated, e.g. ‘papa’.
2. The appearance of the first consonantal opposition, nasal vs. oral, e.g. ‘papa’, ‘mama’.
3. The appearance of the second consonantal opposition, labial vs. dental, e.g. ‘papa’ vs.
‘tata’, ‘mama’ vs. ‘nana’.
4. The appearance of the first vocalic opposition, narrow vs. wide vowel, e.g. ‘papa’ vs.
‘pipi’.
5 . The appearance of the second vocalic opposition, either:
(a) splitting of narrow vowel into front vs. back
e.g. ‘papa’ vs. ‘pipi’ vs. ‘pupu’
(b) splitting of narrow vowel into a more open vs. narrow opposition
e.g. ‘papa’ vs. ‘pipi’ vs. ‘pepe’
Minimal consonant system: m-n
P- t
Minimal vowel system: 1 (or) i
a e
a
different from another child’s phonology of Language B. As we will see,

however, his most specific predictions are only about the earliest words.
Another feature of Jakobson’s theory is that it is directed toward, and
based upon, speech production. Jakobson gives data from acquisition for 15
languages, based on published data. Most of the data is production data on
Czech, Bulgarian, Russian, Polish and Serbo-Croatian. While based on
spoken language, the theory makes the claim that the child’s words are
being restricted by an underlying linguistic system, not just by articulatory
constraints. For example, he cites data from Ament (1899) whose daughter
first varied in attempting words with /k/ between [k] and [t]. She eventually
settled on [t], not because she could not produce [k], but because the
linguistic system imposed it upon her.
The child’s linguistic system is under the influence of linguistic laws that
regulate the order of acquisition of oppositions. Table 6.23 presents a
summary of the proposed order of acquisition of the first phonological
oppositions, based on Jakobson (1941/68). These are part of what he calls
the first stage. The child develops the minimal consonant and vowel systems
upon which all further development is based. The CV syllable is seen as the
starting point in acquisition for both vowels and consonants. Two con-
sonantal features are first acquired: nasal vs. oral, and labial vs. dental. For
vowels, there are also two oppositions acquired. The second one allows
alternatives, one resulting in a triangular system, the other in a linear one.
In the first stages all children will sound the same. He states (p. 50): ‘the
child possesses in the beginning only those sounds which are common to all
the languages of the world, while those phonemes which distinguish the
mother tongue from the other languages of the world appear only later’.
To understand the developments for the next stage, we need to look
briefly at Jakobson’s methodology. First, Jakobson never states the data-
base for these substages of stage l. Presumably his reading of the diary
studies suggested that these forms were characteristic of early phonological
acquisition. His study, however, is not inductive in the sense of Child
Language, or even in the manner of Shvachkin’s study; that is, he is not
studying child language data and drawing conclusions from them. Remem-
ber that he is starting from a theory of phonological features. Further, as
part of his study of linguistics, Jakobson had studied the distribution of his
universal features across languages in the world. (The most impressive work
in this regard had been just completed by his close colleague Trubetzkoy,
translated into English in 1969.) This led him to propose what he called laws
of irreversible solidarity. Jakobson uses these laws to make predictions
about the child’s acquisition in the next stage, that is, the acquisition of
specific languages.
Laws of irreversible solidarity are claims about the distribution of
phonological features among the world’s languages. Presumably, these laws
are part of phonological theory. They state that, in the unmarked or normal
state, certain relations exist between the phonological features within a
language. These are implicational laws which state that the occurrence of
one feature (or class of sounds) implies the occurrence of another. For
example, Jakobson found in his survey that all languages had front
consonants but not all had back consonants: the law of irreversible
solidarity for this fact is that the presence of back consonants presupposes
the existence of front consonants.
Jakobson proposed that the laws of irreversible solidarity can be seen as
making predictions about the order of acquisition of phonological oppo-
sitions. The laws together indicate features or oppositions that are more
basic (or less marked). Children will acquire the basic or unmarked features
before the more marked ones: for example, children will acquire front
consonants before back ones. Table 6.24 gives some of the laws cited by
Jakobson and their predictions about acquisition.
As can be seen, these predictions are limited to a few classes of sounds
that occur in languages. There are a great number of further laws and
specific relations that are not given. For example, what about liquids?
Table 6.24 Some laws of irreversible solidarity and their predictions for
language acquisition (Jakobson 1941148)
Laws of irreversible solidarity Predictions for acquisition
CONSONANTS
1. The existence of fricatives implies the Stops are acquired before fricatives;
existence of stops (p. 51). fricatives are changed into stops, e.g.
lfl+lpl.
2. The existence of front consonants (labials Front consonants are acquired before back
and dentals) implies the existence of back consonants, back consonants are replaced
consonants (palatals and velars) (p. 53). by dentals, e.g. lkl+lt/.
3. If a language has only one fricative, it will The first fricative acquired is Is1 (a variable
be 1st. articulation between [SI and [J]); other
fricatives will be changed to 1st.
4. The existence of an affricate in opposition An affricate is acquired in opposition to a
to a stop implies a fricative within the same stop after the acquisition of the fricative in
series, e.g. Ipfl vs. /PI implies If/ vs. lpl. the same series; affricates are changed into
stops or fricatives, e.g. ltskltlorlsl.
VOWELS
5 . An opposition of two vowels with the same The opposition between vowels of similar
degree of aperture implies the existence of aperture is acquired late.
an opposition with a narrower aperture;
e.g. 191 vs. la1 implies la1 vs. le/.
6. An opposition between rounded vowels Oppositions between unrounded vowels
implies the same opposition between are acquired before those between rounded
unrounded vowels, e.g. /u/ vs. lo1 implies Iil vowels.
vs. le/.
Jakobson states that many languages will have a single liquid (usually Ill),
and that the acquisition of a second liquid is late. We are not told, however,
when the first liquid is acquired, nor its first opposition. One needs to
realize, therefore, that we are dealing with an incomplete theory, or
actually the outline of a theory. These laws will need to be elaborated,
altered and joined with other ones before a complete theory can be
proposed. Also, Jakobson is sometimes careless in the way he states his
predictions. For example, he states that front consonants precede back
consonants, as if he were dealing only with the acquisition of sounds. Given
that he is really concerned with distinctive features or oppositions, this is
quite misleading. What he actually means is that oppositions between front
consonants precede oppositions between back ones. For example, labial vs.
dental appears before palatal vs. velar, or even dental vs. velar.
Since Jakobson gives no analysis of child data, but only anecdotal
examples, we have no idea of when children acquiring different languages
will begin to sound different from each other. Since the child has to acquire
the oppositions of a specific language, however, one would expect some
differences before very long, but the processes that merge oppositions
should look similar across languages. The child’s oppositions, though, are
always acquired in relation to some adult language. This point is worth
emphasizing because it has sometimes been misunderstood in the literature.
Jakobson is not always clear on the level of representation of distinctive
features, but it appears that he recognizes a level of linguistic representation
that is distinct from the phonetic level. This level often appears to
correspond to the perceived surface structure of the adult word. There are
two kinds of examples which indicate a level of linguistic representation.
One is his discussion of the early merger of /k/+/t/. He states (p. 54):
‘Occasionally, an intermediate stage is introduced . . . in which, although
the velar series is not established, the two phonemes are already distin-
guished. In this case, a glottal stop corresponds to /k/ (or velars, in general)
of the mother tongue.’ In other words, a phonetic distinction between [t]
and [3] represents an underlying dental vs. velar contrast. The second
example concerns a French child who said ‘papa’, [dCdC] (for ‘tCter’) and
[de] (for ‘bC’) (p. 49). Jakobson proposes that the child has only one vowel
phoneme /a/with a rule that /a/+/e/ after dentals. The phonological
representation of the last two words ‘tCter’ and ‘be’ would be /dada/ and
/da/. This example differs from the first in that the child’s phonemic
representation is distinct from the perceived adult form. The second
example is similar to the first, however, in showing that there is a linguistic
representation that is distinct from, but related to, the perceived and
produced form of words.
In Part I11 of his book, Jakobson attempts to explain the laws of
irreversible solidarity by reference to the acoustic and articulatory prop-
erties of speech sounds. The style of explanation is similar to that discussed
by Shvachkin in that there is gradual differentiation. When discussing
Shvachkin I called this the Principle of Phonemic Differentiation. Jakob-
son’s term for the same general principle is the ‘principle of maximal
contrast’. In reference to his stages, he says (p. 68): ‘This sequence obeys
the principle of maximal contrast and proceeds from the simple and
undifferentiated to the stratified and differentiated.’ He discusses possible
explanations for the early oppositions acquired by the child and these are
summarized here in Table 6.25. They develop in this fashion, moving to
lesser degrees of contrast. This term is less than satisfactorily defined but
appears to refer to differences between features that are as acoustically
distinct as possible. This is similar to Shvachkin, except the latter also
discusses the role of frequency. In his later work, particularly Jakobson,
Fant & Halle (1963), Jakobson attempts to refine these acoustic differences
Table 6.25 The principle of maximal contrast applied to the first
oppositions acquired by children (Jakobson 1941168)
Opposition Principle of maximal contrast
The basic CV syllable The maximal contrast in the syllable is between closure and
opening. The maximal closure is a labial stop (which seals
the entire vocal tract); the maximal opening is the vowel /a/.
Nasal vs. oral consonants The maximal contrast within consonants is the obstructed
oral cavity in contrast to the open nasal cavity.
Labial vs. dental consonants Both of these involve pitch contrasts; dental consonants
and narrow vowels are greater in pitch than wide vowels
and labials.
Front vs. back consonants Back consonants contrast with front consonants by having
greater resonance (or loudness).
through spectrographic analysis. The important point here is that the

sequence of development is explained by marginally distinct acoustic
properties of the human repertoire of speech sounds.
Lastly, Jakobson attempts to provide independent evidence for his laws
of irreversible solidarity, citing predictions the laws make about language
acquisition and data from language disorders. He proposes that the laws
also predict the order in which language is lost in cases of language
disturbance, such as aphasia; in which case, however, we see the reverse
process - the last contrast acquired is the first lost. This hypothesis, although
controversial, has nonetheless stimulated a great deal of discussion (see
Caramazza & Zuriff 1978). The important point here is that Jakobson
attempts to make predictions in different domains to support his theory on
the nature of phonological universals.
Jakobson did not spend much effort in elaborating or altering the basic
theory presented in 1941. The most extensive development took place in the
refinement of his theory of distinctive features, as found in Jakobson, Fant &
Halle (1963). These new feature specifications were the primary topic in the
book Fundumentals of language by Jakobson & Halle (1956). In that work, he
also altered some of his claims about language acquisition, though these
alterations appear more in response to his changing view of features than his
response to language acquisition data, that is, his revised features led to
different predictions about the order of acquisition of phonological contrasts.
The major changes are as follows:
(i) a revision of stage 1 into what is referred to as the acquisition of the
primary triangle; the primary inventories that result from the primary
triangle are [p, t, k] for consonants, and [i, a, u] for vowels.
(ii) division of distinctive features into three general classes: eight sono-
rity features (including [& vocalic], [-t consonantal], [_+ nasal],
[f compact], [k voice]); protensity features [tense] vs. [lax], and three
tonalityfeatures. Importantly, the acquisition of features within a class
will follow a sequence, but there is independence between the classes.
For example, nasal vs. oral, an early sonority feature, may precede or
follow the acquisition of labial vs. dental, the first tonality feature (the
specific feature is grave vs. acute).
(iii) the universal syllable CV may contrast early with V or VC, i.e. /pa/
vs. /a/ or /pa/ vs. lap/.
The major innovation in these changes is the separation of the features

into three major classes. One result of this change is that the theory intro-
duces the possibility of variation between children at the earliest stages. For
example, some children may acquire nasal vs. oral first, while others may
acquire labial vs. dental first. By doing this, however, the theory runs the
risk of possibly becoming too powerful, i.e. being able to account for any
possible sequence of development. It is an essential condition of any theory
of phonological acquisition to be comprehensive enough to account for indi-
vidual variation without becoming a measurement sequence.
As pointed out by Atkinson (1982), the importance of Jakobson’s theory,
despite its incompleteness, resides in its attempt to provide a principled
theory of acquisition. Even though it has attracted widespread attention,
the theory has not been subject to rigorous development. It has either been
used without criticism or criticized without attempts to develop it. It is obvi-
ously the influence behind Shvachkin’s later study, even though the latter
never refers to it. The lack of such a reference, however, is not surprising
when one realizes that Jakobson was vilified in the Stalinist period after
World War 11. According to KuEera (1983: 878): ‘It was during the harshest
years of Stalinism that Jakobson was viciously attacked, for his linguistic
theories as well as his literary opinions, both in Moscow and Prague.’ He
was also denounced by the Moscow Linguistic Circle. Shvachkin’s attempt
to extend Jakobson’s work, under such circumstances, is truly remarkable.
6.5.3 Methods of phonological analysis

Theories like Jakobson’s need to be tested by the careful analysis of chil-
dren’s early word structure. In our discussion of Shvachkin we have seen a
method that can be used for phonemic perception. Most research, however,
has been limited to the analysis of spoken words. The problem then is how
to analyze the child’s early productive vocabulary for the development of
phonological categorization.
One of the most innovative ways would be to follow a child and actually
attempt to teach him specific phonological oppositions. This was, in fact,
attempted once by Braine (1974a), with his son Jonathan. Jonathan’s first
words were as follows:
(6.8) ‘that, there’ - - dae - de]
[da dA
‘see’ Pi1
‘no’ [dol
‘juice’ [dul
‘hi’ [?ai]
Braine hypothesized that Jonathan had only vocalic oppositions, and the ‘d’
was noncontrastive. That is, he was at Shvachkin’s stage 1(Table 6.22). To
test this, he attempted to teach Jonathan two words that would require
Shvachkin’s stage 2, that is, a contrast between the presence and absence of
a consonant. Braine made up two words: [i] meaning ‘cat’ or ‘food’, and
[dai] with a certain toy. If Jonathan had acquired ‘d’ contrastively, then the
pairs [di] vs. [i] and [?ai] vs. [dai] should be learnable. Jonathan readily
-
acquired both words, but changed them to [di] and [da d ~respectively.]
Braine concluded that Jonathan was not yet acquiring consonants contrast-
ively. The rule was that ‘d’ was placed automatically on simple vowels.
Ideally, the study of phonological development could be pursued in a
manner like Braine’s brief experiment. That is, we could continually
analyze the child’s system and teach words to test our analyses. In most
cases, suitable English words would be available. If not, nonsense forms
could be used. The viability of using nonsense words was recently shown by
Schwartz & Leonard (1982). In their study, Schwartz & Leonard sampled
12 children between 1;0(21) and 1;3(15) who were producing only single-
word utterances. They analyzed their data for each child’s more common
segments and syllables. They then composed two lists of eight nonsense
words, one with words like their own words (IN words), and one with words
not like their own words (OUT words). They taught the nonsense words to
the children over ten bi-weekly experimental sessions. They found that the
children showed a superior ability to acquire IN words, i.e. words near their
own phonological level.
Most analyses of child’s phonologies, however, are done on data col-
lected either through diaries or spontaneous language samples. Unfortu-
nately, it is not apparent that such data will ever be sufficient to test a theory
such as Jakobson’s and our only available data on most languages will
consist of diary material.
Given such restrictions, we need to develop ways to analyze the young
child’s phonological system through spontaneous data. Until recently, most
analysts have ‘done their own thing’ so to speak, thus making comparisons
Table 6.26 Phonological lexicon for T a t VI, taken from Ferguson &
Farwell (1973: 34)
Spontaneous Imitated
Lexical type phonetic type phonetic type
1. allgone 1. ?augho
2. ?auwo
3. ?okh6
4. ?okhu
2. baby 5. apepi
6. abi
7. bibi
3. ball 8. bau
4. blanket 9. bijae bjae
5. book 10. ag
11. bA?
6. bounce 12. bA
13. be
14. bwae
7. byebye 15. phaedi 16. bae'bae
8. cereal 17. "o$uiJ
9. cheese 18. $i
10. dog 19. "dae (2 tokens)
20. da
21. darhai
11. hi 22. ha1
12. ice 23. ai$ (2 tokens)
24. ?a
13. nightnight 25. naina
14. no 26. n6n6 27. nAh
28. n6
15. paper 29. 0etJa
30. baedu
16. pat 31. phaet (3 tokens)
32. phae
17. please 33. phe (2 tokens)
18. pretty 34. prhi
19. purse 35. phe
36. phe
20. rock 37. wakuak (2 tokens)
38. uakwak (2 tokens)
39. uakuak
40. wak
21. shoe 41. rJu
42. $'u
43. $u
44. thuh
22. tea 45. fhi
46. thi
23. thank you 47. thafhi
24. up 48. ?a?
25. yeah 49. ijAh
of different analysts difficult. In recent years, however, two distinct

approaches have appeared which attempt to analyze the child’s emerging
phonological system. Here we will take a brief look at each.
Phone classes and phone trees This method was originally proposed by
Ferguson & Farwell (1975). It has been used since by several other
researchers, in particular Shibamoto & Olmsted (1978), Leonard, Newhoff
& Mesalam (1980), and Stoel-Gammon & Cooper (1984). A reading of the
literature requires an understanding of the phone class/phone tree
methodology.
The study by Ferguson & Farwell limited itself to word-initial con-
sonants. The definition of phone class, however, can be extended to all
word positions (as is done in Shibamoto & Olmsted 1978) and this should be
kept in mind when we look at their definition of phone class. A p h o n e class
(Ferguson & Farwell 1975: 424) is determined as follows:
Then all words beginning with the same phone or set of variant phones
were put together. The set of initial-consonant variants of each of
these groups of words constitutes a ‘phone class’, and is represented
by the appropriate phonetic symbols in a box, or between vertical
-
lines. Thus a phone class [d th]consists of the initial consonants of all
those words whose initial-consonant sound varied between [d] and
[thl.
Phone classes, then, are primarily determined by the distribution of
segments across words.
In an earlier version of their paper, Ferguson and Farwell provide an
appendix giving the data which is the basis of their analysis. Table 6.26 gives
a phonological lexicon of their data for one of their subjects, T , at the sixth
sample session. The phone classes for this session were as follows (subscrip-
ted numbers refer to the number of tokens):
(6.9) Phone classes for T a t session VI
1. [b - - - - -
bw ph (I 01 for ‘baby’, ‘ball’, ‘blanket’, ‘book’,
‘bounce’, ‘bye-bye’, ‘paper’
2. IP”4 for ‘pat’, ‘please’, ‘pretty’, ‘purse’
3. [WII for ‘rock’
4. [dl for ‘dog’
5. [th12 for ‘tea’, ‘thankyou’
6. [n12 for ‘nightnight’, ‘no’
7. for ‘yeah’
8. [$ -‘q -
J th]3 for ‘cereal’, ‘cheese’, ‘show’
9. [hll for ‘hi’
-
10. [? 0 1 3 for ‘allgone’, ‘ice’, ‘up’
First of all, there is a problem concerning how words are assigned to
phone classes. Here it appears that words are grouped into phone classes if
some of the variants are the same. For example, ‘ball’ and ‘baby’ are
-
grouped together because the former’s [b] is a variant of the latter [0 b].
Shibamato & Olmsted (1978: 421-2), however, point out that for some data
this can result in some analyses that produce a very small number of phone
classes with very large membership. This happens in Ferguson & Farwell
(1975: 426), in fact, when T at VI1 has a phone class of the following:
- - - - - - - - -
[J tJ 3 j th s h q tq 91.
The next question, then, concerns the purpose of phone classes. Fer-
guson & Farwell (1975: 425) state it as follows:
The notion of ‘phone class’ here is similar to the notion of ‘phoneme’
of American structuralism, in that it refers to a class of phonetically
similar speech sounds believed to contrast with other classes, as shown
by lexical identification.
Phone classes, by this definition, are isolated in order to follow the
development of phonemic oppositions.
Ferguson & Farwell determine the phone classes of a longitudinal series
of language samples, and then connect the classes across time. These
longitudinal connections are phone trees. They state (p. 424):
If successive phone classes did not contain the same word but were
related to phone classes which did, dotted lines were drawn connect-
ing them. For example in T’s /m/ class:
/m/ (mama)
/m/ (milk)
I
/m/ (milk, mama)
Figure 6.2 gives T’s phone classes for nine longitudinal sessions, over 13
weeks, beginning at 11 months.
As stated earlier, this approach has become quite popular in recent years
for the phonological analysis of children during the period of single-word
utterances. In 6.5.4 we will present some results of these analyses. Here,
however, we would like to point out some concerns with this method. First,
there is a minor methodological point. These trees like Figure 6.2 are hard
to devise, since the procedures are not often clear, and they are difficult to
read. More elaborate phone trees look like the wiring diagram of a
television set. This would be worth the effort, of course, if they reflected the
child’s emerging phonological system. There are difficulties, however, in
arguing that they do.
The most serious problem with phone trees is that they are extremely
sensitive to the phonetic variability of a single lexical item. Suppose, for
example, that a child’s attempt at English /b-/ were as follows:
(6.10) ‘ball’ [ba] ‘button’ [bh] (6 tokens), [hth], [thth]
‘bike’ [bar]
‘book’ [bu]
- -
The phone class would be [b 0 t] suggesting extensive variability where,
in fact, only one out of four lexical types shows this variability. Further, the
variability of ‘button’ is not as great as it might appear, since six of the eight
phonetic tokens begin with [b]. The phone class analysis, however, misses
this point. This approach places tremendous emphasis on variability that
may result from performance factors, yet uses the results for conclusions
about the child’s phonological competence, i.e. his emerging phonological
oppositions. Still further, this variability is enhanced the more narrow the
phonetic transcription. The phone class given above for T at session VI1 is
an example of this. My general conclusion is that phone clasdtree analyses
are awkward to do, and they predispose the analyst to focus on the surface
variability of lexical items, variability that is subject to sampling conditions
and performance factors in general. It is useful as a method to draw
attention to such variation. As a measure of competence, however, it is a
measurement sequence in the sense of Chapter 3, that is, it can only lead to
conclusions that development is variable from lexical item to lexical item.
Phonetic inventories and phonological contrasts In a series of works

(Ingram 1981a, 1988; Ingram & Mitchell to appear), I have attempted to
develop a method of analysis that eliminates some of the variability in
production data due to performance factors. This method consists basically
of three steps: the determination of (i) a phonetic inventory, (ii) patterns of
substitution, and (iii) a set of phonological contrasts. Here I will briefly
outline the features of each of these steps. To reiterate, the process is
designed to reduce variability in order to determine evidence for underlying
linguistic competence.
First, we need to establish the child’s phonetic inventory; these are the
sounds that the child uses to construct his words. The analysis begins with an
abbreviated phonological lexicon. There are two initial steps to this process:
enter data in a broad phonetic transcription to eliminate transcriber
variability that occurs when fine phonetic transcriptions are done, then
select a typical phonetic type for each lexical type. The first step simply
eliminates diacritics and raised elements that were faintly heard or
questionable in the transcription. The second step requires some comment.
The idea is to eliminate some of the data by selecting a phonetic type that is
Table 6.27 An abbreviated phonological lexicon for T a t VI, with only the
most typical phonetic types indicated
Lexical type Phonetic type
1. allgone ?okho
2. baby bibi
3. ball bau
4. blanket bija: bja: (imitated)
5. book ag
6. bounce be
7. bye-bye phzdi
8. cereal SU’S
9. cheese FI
10. dog dz
11. hi hai
12. ice a1.r
13. nightnight naina (imitated)
14. no no
15. paper cpetsa
16. pat phzt
17. please Phe
18. pretty Pbhi
19. purse Phe
20. rock wakuak
21. shoe su
22. tea thi
23. thankyou thathi
24. up ?a?
25. yeah ijA
most typical of all the phonetic types for a lexical type. The following rules
seem to work most of the time:
(a) If a phonetic type occurs in a majority of the phonetic tokens, select it.
(b) If there are three or more phonetic types, select the one that shares the
most segments with the others.
(c) If there are two phonetic types, select the one that is not pronounced
correctly.
(d) If none of the above work, select the first phonetic type listed
We can see how these rules work by attempting to select the most typical
phonetic types in Table. 6.26. Rule (a) will apply in ‘ice’, which has the
phonetic types [arJ’] (two tokens) and [?a]. Since the former occurs in more
phonetic tokens than the latter, it is selected as the phonetic type. Rule (b)
will operate for ‘allgone’. The medial consonant that is most typical is ‘k‘
because three out of four phonetic types have a stop, and two out of three of
the stops are [k]. Three of the phonetic types end in [o], so it is the most
typical vowel. The only phonetic type with both of these is [?okho]. There
are no examples where rule (c) would apply in Table 6.25. An example
Table 6.28 The criterions of frequency for marginal, used and frequent
consonants at selected sample sizes
VOCABULARY SIZE CATEGORY

No. of lexical types marginal” used frequent
1-37 1 2,3 4 and up

38-67 0 2,3 4 and up
6%87 2 3.4 5 and up
88-112 2,3 4s 6 and up
(‘Normally, we want to avoid including a sound that only occurs once, even as marginal. An
exception is made for samples with 37 or fewer lexical types because the database is so
small.
would be if ‘dog’ were [dag] and [da]. The latter is selected by rule (c), for
we want to be conservative in our claims about the child’s ability. Lastly,
rule (d) will operate if the other rules do not apply. For ‘paper’, for example,
it would select [etJa]. Table 6.27 gives the abbreviated phonological lexicon
for T at session VI. (Imitations are included.)
The child’s phonetic inventory is determined separately for word-initial
and final consonants. Example (6.11) gives the initial and final consonants
that occur in Table 6.27 along with the phonetic forms in which they occur:
(6.11) Initial consonants Total phonetic forms
b- bibi, bau, bija bja, be 4
p- phadi, phat, phe (2 tokens), pghi 4
d- d a 1
t- thi, t h a thi 2
3- ?okho,?a? 2
fricative- ~ u I JGi,
, Ju 3
h- ha1 1
n- nama; no 2
0- 0etSa 1
w- wakuak 1
Final consonants
g- ag
-fricative SUIJ,arJ
-t phat
-k wakuak
-3 ?a?
The child’s phonetic inventory is determined by the frequency with which

consonants occur in distinct phonetic forms. In (6.11) 24 phonetic forms are
shown. According to their frequency, consonants are assigned to the phonetic
inventory as either marginal, used, or frequent. Marginal sounds are
placed in parentheses, and frequent sounds are marked by an asterisk, while
used sounds are shown without any marking. Table 6.28 gives the number of
times a sound must occur in samples of various sizes in order to occur in one of
these three categories. (These figures are arbitrary, based on the expectation
that used words should occur at least twice; see Ingram 1981afordetails.) Each
phonological sample, based on the number of phonetic forms, has a criterion
offrequency that is used to assign consonants to the phonetic inventory.
The phonetic inventory for T's session at VI is shown in (6.12).
(6.12) T's phonetic inventory at session VI
Initial Final
n- (-g)
*b- (d-) (-t) (4 (-3)
*p t- 3
-2 - -F
(0-1 s -F -s (h-)
(w-)
There are two frequent initial sounds, [*b] and *[PI, and three used ones [t,
3 , n] and the fricative group. The latter are placed together because they
would probably be given a single symbol in a broad phonetic transcription.
The second step is the determination of patterns of substitution. This is
done from Table 6.27 by looking at the single consonants only, in the adult
models. (Clusters are analyzed separately.) Example (6.13) gives the
substitution analysis for initial consonants, where correct productions of a
consonant are indicated with a 'C' (for correct), and substitutions are given:
(6.13) Lexical types Proportion correct
c c 0 c
b- baby, ball, book, bounce, g"ye-bye 31.5
0 c c
p- paper, pat, purse 213
C
d- dog 1/1
C
t- tea 1/1
C C
n- nightnight, no 212
s
s- cereal 011
F
tJ- cheese 011
C
J- shoe 111
C
h- hi 111
W
r- rock 0/1
th
0- thankyou 011
C
j- yeah 111
Once we have the substitution patterns, we need to decide when the child
has matched the adult model. I propose that we call a child’s attempts at an
adult target a march if they are correct over 50 percent of the time. Further,
we can specify a marginal march when there is only one attempt. In our
data, we have three matches, [b-, p-, n-1, and five marginal matches, [d-, t-,
J-, h-, and j-1.
These data are an example of the difficulties in working with spontaneous
samples. We have no data on adult words with /m-/, and we have five cases
of marginal matches. We would like to be able to visit this child at this point
and attempt to elicit more words in this area. These problems are somewhat
ameliorated when longitudinal data are involved.
The third step is to determine the child’sphonologicaf contrasts. Unfortu-
nately, it appears that we can never do this with much confidence with a
single sample of this kind: even so, we can suggest the following criteria:
A sound is considered part of the child’s phonological system (Le.
becomes a phoneme in the child’s phonology) when
(i) it is frequent; or
(ii) it is used, and it appears as a match, or substitute.
We thus need to compare the initial phonetic inventory in (6.12) with the
patterns of substitution in (6.13). By these criteria [n-, b-, p-, t-] and
fricative are part of the system. If we include these as marginal (in
parentheses), we get the following phonological system for T’s initial
consonants:
(6.14) n-
b- (d-)
p- t-
9
- - J (h-)
8 G
(w-)
It is proposed that data like these are more amenable to analysis to test
Jakobson’s theory than those of Figure 6.2 which are full of performance
factors.
The strongest evidence for the emergence of contrasts is obtained
through comparisons of analyses like the above at different time periods.
Table 6.29 gives such an analysis for T’s data in Ferguson & Farwell (1975)
across all nine sessions. For the longitudinal data, the analysis was altered as
follows. The phonological lexicon consisted of all the typical phonetic types
at a session, plus all previous typical types that were not altered at the
current session. For example, the phonological lexicon for session VI
(Table 6.27) contains one phonetic type beginning with a [d]: [dae] for ‘dog’.
Table 6.29 Phonetic inventories, matches, substitutions, and proposed phonemes (in circles) for T
Phonetic inventories Matches and substitutions
Sample
Session [m n b d g p t k f s 6 I tl h w] i m n h d g p t k f 0 s I t J d 3 w j r h/ size
I 4
II (?) 5
Vlll (m) (n) *h d p 't (k) ( f ) 6 5 (h) w
Only sounds are shown that occur in more than one session, except for IX.
Imitations are not included in the analysis.
Parentheses indicate a single occurrence of a sound
Criterion of frequency for sessions I-VI1 was 2(1); 2 for sessions VIII, IX.
A phonetic type beginning with [d] was also carried over from a previous
session, i.e. [dzdi] ‘daddy’ (V). Thus, [d] meets the criterion of frequency of
two for used sounds. We refer to such analyses as cumulative phonological
analyses, in that they carry data from previous samples over to the one
under analysis. A second feature of the analyses is that imitations were not
used. These were dropped because it was felt that claims about the child’s
emerging phonological system should only be made in reference to spon-
taneous language. The criterion of frequency for marginal and used sounds
was one and two respectively for all nine sessions, a departure from the
guidelines in Table 6.28 in that there should have been no marginal sounds
for sessions VI11 and IX, due to the sample size. This was changed in order
to present as much data as possible in Table 6.29.
Longitudinal data allow us to add one more important criterion for
defining an opposition:
A phoneme is acquired in the first session in which the criteria (i) and
(iii) are met, if they are also met in all further sessions.
We have circled those sounds in Table 6.29 that meet these criteria. As we
can see, /dl is acquired right from session I. /b/ is next at session V, followed
by /p/ at VI, and It/ at VII. It/ is not circled at session VI because it is a match
in only one lexical type. The last session shows evidence for a /k/,although it
would be nice to see data from subsequent sessions. The data on /s/ are not
as neat. It does not meet criterion (iii) in that there are no [s]s in sessions
VI1 and VIII. During these sessions, and IV for that matter, [s] and [F] are
phonetic alternants. We can circle Is/ at V, then, if we define this phoneme
as having two alternants for the child. The right side of the table shows that
these two are consistently used for adult words with Is/.
T’s data show the following developments:
(6.15) Sessions
I d dental
IV b d labial vs. dental
V b d dental stop vs. dental fricative
S
VI b d voiced vs. voiceless labial stop

S
P
VI1 b d voiced vs. voiceless stop
P t
S
VI11 b d
p t k velar vs. non-velar stop
S
It is data like these which can be used to test and elaborate Jakobson’s
theory, as well as others.
6.5.4 Some characteristics of early phonological development

We can now use the data from T, as analyzed by phone classics and phonetic
inventories, to discuss some of the general characteristics of the child’s
phonology during the period of single-word utterances. To do this, we will
focus on the study by Ferguson & Farwell (1975), using my reanalysis of the
data on T in Table 6.28 for parts of the discussion.
The use of phone classes in Ferguson & Farwell (1975) enables the
authors to identify some of the specific features of early phonological
development. The major conclusion of their analysis is what they refer to as
the lexical parameter. (The term ‘parameter’ was of no theoretical import-
ance here, and should not be confused with the parameter-setting model
discussed in Chapter 4.) The course of early phonological development is
heavily influenced by the properties of individual words. One aspect of this
is that the first contrasts may only be between individual words. This is
particularly clear in Table 6.29 for [m-] and [n-1: through most of the
sessions, [m] occurs in ‘mama’, and [n] occurs in ‘no’. I will refer to such
cases as lexical contrasfs.Other lexical contrasts occur for T with ‘he’ for [h]
and ‘where’ for [w]. None of these contrasts becomes generalized over the
nine sessions; thus, not only do some contrasts first appear as lexical, but
they may remain so over a long period of time.
Ferguson & Farwell also propose that the lexical parameter operates
when a contrast begins to generalize. They argue that it will spread
gradually word by word rather than suddenly to all relevant words. Velten
(1943) noted a lexical influence of a similar kind in his study of his daughter
Joan. In her case, a new contrast would be used for new words being
acquired, but older ones only changed gradually.
Table 6.29 shows some evidence for the gradual emergence of contrasts.
Look, for example, at T’s acquisition of /t/. For sessions 111, IV, and V it is
marginal, i.e. it is only a lexical opposition. At VI it becomes used, but it is
only at VI1 that it becomes frequent and meets the criteria for an
opposition. It is important to point out, however, that not all oppositions
begin lexically, as suggested by Ferguson & Farwell; /p/, for example,
suddenly emerges as a productive phoneme for T at VI. Ingram & Mitchell
(to appear) find several other examples of this in a study of three
English-learning and three French-learning children. There seem to be
three patterns, then, in the emergence of oppositions: (i) extended lexical
oppositions, (ii) gradual lexical spread, and (iii) sudden emergence. T’s
acquisition of /m/, /t/, and /p/ are examples of each pattern.
Another aspect of the lexical parameter pointed out by Ferguson &
Farwell is phonetic variability. While young children tend to show alter-
native pronunciations for lexical types, Ferguson & Farwell claim that some
words are more variable than others. They refer to those words which are
fixed in their pronunciation as stable forms. Unfortunately, they do not
provide a measure of stability. As examples, they provide T’s use of ‘baby’
as a variable form contrasted with ‘book’ as a stable form. As can be seen in
(6.16), however, both are only variable within certain limits. For example,
if we look at the shared features of the two words, it appears that their
underlying representations are something like /bebi/ and /bak/ respectively:
(6.16) ‘book‘ VI ag, ba?; VI1 bak; VI11 ba; IX bu
‘baby’ I11 de; VI bebr; V be@, ba, bebi; VI apepi, abi, bipi;
VI1 bebi; VI11 bebi; IX bebi
In fact, in some ways ‘baby’ is less variable than ‘book’ in that it shows a
repeated phonetic form [bebi], and it becomes stable at session VI1 whereas
‘book’ does not. Therefore, while there may be a lexical dimension to
variable pronunciation, we need to be cautious until stronger documen-
tation is provided. (See Leonard, Rowan, Morris & Fey, 1982 for an
attempt to pursue this further.)
Besides the lexical parameter, Ferguson & Farwell observe other char-
acteristics which seem the result of what we will call the phonological
parameter. This term refers to the restrictions imposed by a limited
phonological system. One of these characteristics is what is called a
phonological idiom, a term first used by Moskowitz (1971). This refers to
the child’s first pronunciations of some words which are superior to later
pronunciations. It is as if the first attempts were direct reflections of the
child’s perceptual and articulatory skills, without the interference of the
phonological system. Once they are incorporated into the system, they get a
simpler pronunciation. As an example, they cite Hildegard Leopold’s
acquisition of ‘pretty’ which was phonetically accurate at ten months
(although whispered), reduced to [p~ti]at 21 months, and eventually
became [brdi] at 22 months. They say (p. 432):
Progressive idioms suggest that a child’s perceptual and productive
abilities are more advanced than the phonological system seemingly
exemplified by most of his words.
They do not provide, however, an operational definition of how to
determine a phonological idiom, nor information on when they occur or
their frequency.
The proposal of something like a phonological idiom has been around for
some time. Jakobson (1941/68: 23) discussed this at some length:
First, one often secures the ‘parrot-like’ repetition of single sounds
and syllables from children, even though the very same sounds
continue to be absent when they talk spontaneously. Second, these
sounds are sometimes used correctly in the first acquisition of words,
and with the progressive acquisition of vocabulary, they disappear
from use without a trace.
He cites Abrahamsen (1938) and Ament (1899: 51ff.) as references on this.

Their occurrence is one reason why Jakobson argues for a phonological
level of representation in the child, distinct from the articulatory level.
Another characteristic of early phonological development that comes
under the phonological parameter is Ferguson & Farwell’s reference to
salience and avoidance. The claim is that children will tend to produce or
acquire words that contain sounds within their system, and avoid those that
do not. For one example, they cite T’s acquisition of [p]. Up to session IV,
she avoided saying any English word that began with [p]. At IV, however, it
became part of her system, and it was used frequently. The study cited in
6.5.3 by Schwartz & Leonard (1982) has experimentally supported the
observation that selection and avoidance occur. The fact that they do, of
course, raises our confidence level about the use of diary data, as long as
they are complete. We must remember, however, that these processes are
concerned with production. In Hoek, Ingram & Gibson (1986), for
example, we found that OUT words were acquired as easily in comprehension
as were I N words, even though I N words occurred earlier in production.
(Recall that IN words are nonsense words similar to the child’s own words,
and OUT words were ones that were relatively different in their phonological
form .)
The last aspect discussion by Ferguson & Farwell that is part of the
phonological parameter is the sequence of acquisition of contrasts in the
child’s system. As mentioned, Jakobson suggests that there is a universal
sequence, at least in his stage 1. While admitting some initial uniformity
among the children (p. 435), Ferguson & Farwell go on to emphasize how
different the children are from one another. They say, for example (p. 437),
‘. . .each of the three children is exhibiting a unique path of development
with individual strategies and preferences and own idiosyncratic lexicon’.
This, then, is about as opposite to Jakobson as one can get. Their study has
since been seen by many (e.g. Atkinson 1982) as the ultimate counter-
evidence to Jakobson’s theory. Further, it has been the basis for subsequent
studies which also focus on individual variation between children (e.g.
Stoel-Gammon & Cooper 1984). It is important, then, to take a close look
at their claims and the data on which they are based.
First, there are the problems cited in the previous section on the
measurement sequence that results from using phone class analyses as
evidence for claims about lexical influences and individual variation. I
would offer the qualification that their data show that children are quite
different from each other in their phonological development when perform-
ance factors are emphasized. The children will be acquiring different
vocabulary (to a degree at least), in varying contexts, with individual
preferences for certain objects and, probably, sounds. Further, the data will
be collected in varying circumstances, For example, in Ferguson & Farwell,
the data from T, a child sampled by the authors in a systematic longitudinal
study (in the sense of Chapter 2), are compared to Hildegard Leopold, a
bilingual English-German child who was observed in a diary study by her
father, Werner Leopold. Given this, it is not surprising that there are
individual differences.
The phonetic inventories/phonological contrasts method, designed to
minimize performance factors, can yield a rather different picture of
development. For example, Ferguson & Farwell state (p. 435): ‘T develops
her /m/ class just before her /w/, although the /w/ forms appear more stable.’
As Table 6.29 indicates, however, neither of these moves beyond lexical
contrasts. Further, T is contrasted with Hildegard because Hildegard ‘on
the other hand, seems to gain control over certain classes and then to prefer
to add new words to them’. As Table 6.28 suggests, T did this also at least
with /b/, /d/, /p/ and It/. It is the claim here, then, that claims about the
individual variation between children regarding their phonological com-
petence, as opposed to performance, will need to await more studies, using
a method more like the one of phonetic inventories/phonological contrasts.
Besides these methodological questions, there are theoretical issues to be
addressed. To understand Ferguson & Farwell, it is important to see their
initial orientation. They are very much part of Child Language, as discussed
in Chapter 4:
Our approach is to try to understand children’s phonological develop-

ment in itself so as to improve our phonological theory . . .
. . . some linguists at the present stage of the art might be well advised
to turn away from the fascination of writing rules of maximum
generality and conciseness for whole languages, and undertake
instead highly detailed analyses of the idiosyncratic paths which
particular children follow in learning to pronounce their languages.
(Ferguson & Farwell 1975: 457,438)
They are inductive, letting the data from children contribute to theory,
rather than vice versa. Interestingly, they also cite (p. 438) a form of the
Competence Assumption.
Given all of this, it is not surprising that they emphasize individual
variation and performance factors, both important aspects in Child Lan-
guage. In doing so, they need to explain how the child even reaches a
non-lexical ,systematic phonology. They are consistent in that they suggest
that phonological theory may need to be altered to be more lexical and
individualistic. They even hint at the possibility that all speakers may in a
sense have their own unique phonology.
Most important, however, is the need to explain what is meant by
individual strategy. This term is used throughout the Child Language
literature, but it is not often defined. Ferguson & Farwell never define the
term. They acknowledge (p. 435) that some differences between children
may be due to the linguistic environment, but propose that others are due to
individual strategies. Their definition is by example (p. 436):
Such individual strategies include preferences for certain sounds,
sound classes, or features (‘favorite sounds’); extensive use of redupli-
cation; special markers for certain classes of words, . . ; preferences
for lexical expansion or phonological differentiation at the expense of
the other; and persistent avoidance of particular ‘problem sounds’.
Some of these may possibly be instances of linguistic variation (in the sense
of section 4.5); most, however, seem to be cases of performance variation.
To summarize, Ferguson & Farwell have demonstrated through their
phone class analysis the role of the lexical parameter. At the same time,
though, they also provide evidence through phonological idioms and
salience and avoidance that there is a phonological parameter at work.
Their study, due to the limitations of their method, concentrates on
performance factors and individual variation. We are still waiting, however,
for extensive evidence for or against Jakobson’s theory as a theory of
phonological competence.
Despite numerous studies on languages other than English we know

relatively little about phonological development in other languages. Some
recent research, however, applying the phonetic inventories/phonological
contrasts method to other languages has indicated support for cross-linguis-
tic differences, and a linguistic level of representation (e.g. Ingram 1988,
1985b; Ingram & Mitchell, to appear). Here I conclude the section with a
brief look at one of these studies which has compared the acquisition of
English with Quiche, a language which is phonologically quite different.
Pye, Ingram & List (1987) Quiche is a Mayan language spoken by half a
million people in the western highland region of Guatemala. Its phonologi-
cal inventory is quite different from that of English. Table 6.30 provides the
Table 6.30 The word-initial consonants of adult Quiche'
Bilabial Alveolar Palatal Velar Uvular Glottal
Plosives p t, tS tS k q 2
Ejectives b' t', tS' tS' k' q'

Fricatives S s X
Nasals m n
Liquids r, 1
Glides w j
initial consonants that occur in QuichC. Stops and affricates are grouped
into a category labeled plosives. Quich6 initial consonants differ from
English in the following ways: (i) there is a series of electives (or glottalized
stops; see Ladefoged 1975); (ii) there are uvular and glottal stops /q/ /?/;
(iii) there is no /f/, but there is/x/, the voiceless velar fricative. This system is
sufficiently different from English to lead to some interesting tests of
predictions about phonological development. In section 5.3.1 we examined
Locke's (1983) maturational theory which predicted that there would be no
effects of the child's linguistic environment until some time after the first 50
words acquired (Locke 1983: 84). On the other hand, a constructionist
position such as Oller's (1981) suggests that some modifications due to the
linguistic environment should occur. In particular, Jakobson's theory
predicts linguistic effects after his stage 1 (the very earliest words) since the
child is proposed to have a phonological level of representation.
Phonological data from QuichC children were collected by Cliff Pye as
part of a longitudinal study on acquisition (Pye 1980). He visited QuichC
children in their homes over a nine-month period, approximately every two
weeks, for a one-hour play session. All sessions were tape-recorded and
transcribed by Pye with the help of two native QuichC speakers. In Pye,
Ingram & List (1987) the initial consonants from five of the children were
analyzed using the phonetic inventories method described above. The
sample sizes ranged from 23 to 115 lexical types with a mean of 68. The
children's vocabularies were larger than those of the children studied by
Ferguson & Farwell. According to Pye's morphological analysis (Pye
1983a), however, they were still primarily using single-word utterances.
Table 6.31 presents the phonetic inventories for the five children. Pye,
Ingram & List then determined a composite inventory for the five children,
using procedures for composite transcriptions. This composite phonetic
inventory was taken as representing a basic or core set of consonants that
Table 6.31 Phonetic inventories offive Quiche children, taken from Pye, lngram & List (1987)
Q U I C H ~CONSONANTS
Nasals Stops Affricates Glottalized stops Fricatives Liquids Glides

Child Age m n p t ts tf k q ' b' t' ts' tJ' k' q' s J x I r W J
A Tu:n 1 ;7 n (P) ts (k) ?** I* W*

A Li:n 2:o m n** P t tS* k* q ?** (b') x* I" w* (1)
AI Tiya:n 2:l n* P t* ts k* ?***
(b')
x* I*** W**
AI Cha:y 2;9 m n P* t ts* k ?* J* x I*** W*
A Carlos 3:O (m) n (P) t' (ts)tl* k ?* (b') k' s* (f) x* I W
Composite (m) n P t tl" k 7% (b') x* I* W*

QuichC children acquire early. It was then compared to a core set of initial
consonants for English proposed by Ingram (1981a), where 15 children
were analyzed by the same procedure. This comparison is shown in (6.17),
where marginal sounds are given within parentheses:
(6.17) The basic phonetic inventories of Quiche‘ and English
Quicht English
(m> n (m) n
(b’) b d (8)
p t t J k 3 P t k
X (f) (s) h
W W
1
While there are some similarities, particularly regarding nasals and
voiceless stops, there are more differences. Clearly QuichC and English
children are acquiring very different sets of initial consonants. The first
QuichC fricative acquired is [XI, whereas English children begin with [f] and
[SI. One could argue that English children do not show an [XI because it does
not occur in English. This is not true for [I] and [tJ], however, which occur in
English, but do not appear among children’s early initial consonants. These
two sounds, however, are two of the most used initial consonants by QuichC
children.
These data indicate that children will show the effects of linguistic input in
their phonetic inventories, at least toward the end of the period of
single-word utterances. In some ways this is not surprising, in that Shvach-
kin’s study has indicated that phonemic perception is well on its way by that
point. Importantly, the data show that articulatory constraints are not as
strong as one might expect, certainly not as strong as predicted by Locke’s
maturational theory.
Pye, Ingram & List give two possible explanations for the differences
shown in (6.17). First, they consider the articulatory learning theory
discussed in 5.3.1. This theory would predict that the child acquires the
sounds that are most frequently heard. The problem of correlating onset of
acquisition and frequency of occurrence in adult speech has been discussed
for English by Moskowitz (1970). Using the data from Wang & Crawford
(1960), Moskowitz points out that the most frequent English fricatives are,
in order of frequency, fsf, f6f, fzf, fvf. Only one of these, fsf, is early in
acquisition. Pye, Ingram & List conclude that absolute frequency does not
seem to be a primary factor.
They then explore the possibility of functional loud on phoneme fre-
quency. This has to do with the extent to which a phoneme is necessary to
the phonological system. The functional load of a phoneme is measured by
Table 6.32 The rank order frequencies for initial consonants common to
Quiche‘ and English
Sounds
Language /tJ w k p t I n s m r J j/
Quiche 1 2 3 4 5 6 7.5 7.5 9.5 9.5 11 12

English 10 7 1 2 6 11 8 3 9 12 13 4.5
the number of oppositions or minimal pairs it occurs in. English /g/, for
example, has a smaller functional load than /m/ because /g/ does not occur
initially. Wang & Crawford (1960) have found that English I6l is the second
most frequent fricative. This is because it occurs in a small class of frequent
words, like ‘the’, ‘this’, etc. Its functional load, however, is quite small in
that these words are a small class and subsequently enter into a small
number of minimal pairs. We could change all English /6/into [d]s and still
communicate.
To get an indirect measure of functional load, Pye, Ingram & List looked
at the frequency of English and Quiche initial consonants in terms of lexical
types. For English, we used the 500 most frequently used words of 5- and
6-year-old children in Birmingham, England, based on Burroughs (1957).
Since no adult data are available on QuichC, we had to resort to looking at
the frequency of Quiche initial consonants in the adult word types
attempted by the children. Table 6.32 gives the frequency rank order for
those sounds shared by the two languages. These frequencies seem to
account for the differences quite well. For example, /tJ/ occurs in many
more words in Quiche than in English, while the opposite is true for /SI;/I/ is
also nearly twice as frequent in lexical types in the words Quiche children
use as in English.
This preliminary study provides some suggestive data on the effects of the
input language on early phonological development. Similar analyses on
children acquiring other languages will need to be done before more
definitive conclusions can be drawn. If valid, the data indicate that children
begin during the period of single-word utterances to acquire the phonologi-
cal system of their language. While there may be some articulatory effects
that restrict the child’s production, these effects may be less than previously
expected. The primary factor behind the appearance of a sound in the
child’s early system appears to be its importance in the phonological system
which underlies the words that the child hears. It would not be surprising to
6. The linguistic environment 219
find that languages also tend to assign these highest functional loads to
sounds that are easier to say. We need to be cautious, therefore, in
concluding that the sounds that are acquired later in English are in some
sense harder to say. In some cases, such as /1/ and /tJ/, they may just be not
as important phonologically.
6.6 The linguistic environment

We saw in Chapter 5 that the language addressed to infants may be
influenced greatly by the culture in which the infant lives. Western culture
sees the infant as a potential conversational partner, and its burps and
smiles are accepted as turns in the conversation. Further, as found by both
Snow (1977b) and Kaye (1980a), this language to the child does not change
abruptly once the child begins to produce words. Adults continue to use a
simplified form of English which contains the features given in Table 5.11.
In this section I look first, during the period of single-word utterances, at
the child beginning to engage in what looks like conversation - while the
child still has relatively little to say, he begins to respond to adult speech
with language of his own. Second, I discuss the methodological means for
examining how the adult language directed to the child may influence his
language. Lastly, I look into the possibility that some individual differences
noted between children may be the result of their linguistic environment.
6.6.1 The onset of conversation

In section 5.5.3 an extract was given from Snow (1977b) of a conversation
between a mother and her infant. Later, two other examples were given
from Kaye (1980a). In each case, the mother was clearly the initiator and
controller of the conversation. During the period of single-word utterances,
the child begins to play a more important role in the process.
One of the first people to examine the development of conversation was
Lewis (1937) as part of his study of K. Lewis proposed that K went through
the following three stages of development (p. 67):
Lewis’s three stages of conversational development
1. Earliest months: the child responds to the adult’s acts by acts and
some utterance of sounds.
1. Towards the end of the first year: the child responds to the adult’s
speech by acts; and to his acts by speech.
3 . About 1;6: the child responds to the adult’s speech by speech.
Conversations for K, then, began toward the end of the period of single-
word utterances.
Table 6.33 Four examples of K s early conversation, taken from Lewis

(1937; Series 2: 73) (format amended)
Example no. K’s utterance Mother’s utterance
Age Orthography Phonetics
1. (Mother is dressing K in front of the gas fire; K points to fire and speaks)
1;5(10) ‘Aha’ aha
‘Fire’ fa
(Mother points to fire)
‘What’s that?’
‘Fire’ fa
2. (Around bath time, K climbs upstairs, with his mother following him)
1 ;W) ‘Where are you going?’
‘Bath’ ba
3. (The mother is standing by the bureau in which chocolate is kept)
1 ‘I’ve got something nice for you.’
‘Chocolate’ gjga
‘Chocolate’ gjga
4. (Mother is wheeling K s carriage toward a bed of tulips)
1;6(16) ‘What can you see?’
‘Flower’ fafa
Table 6.33 gives four examples of K’s first conversational exchanges. In

three of the examples, K responds to his mother’s speech. It is only in the
first example that K actually initiates the conversation. Lewis emphasizes
that the transition of his stage 2 to stage 3 is very gradual, although no
frequency counts are given. Further, he stresses the important part played
by K’s mother in eliciting K’s conversation. He states (p. 68): ‘The adult, by
gestures and intonation, continually incites the child to speak, and encour-
ages him with signs of approval when he has spoken.’ Again, we have to
remember that this is a feature of Western culture.
More recently, Lewis’s observations have been elaborated by Halliday
(1975) in his study of his son Nigel. Like Lewis, Halliday observed that his
son did not begin to engage in dialogue until around the end of Phase I (or
the period of single-word utterances) or 18 months. During Phase I, Nigel
engaged in what Halliday calls proto-dialogues. He could do the following
(p. 48):
(i) give three specific responses to calls, greetings and gifts;

(ii) answer questions of the type ‘Do you want. . . ?’; ‘Shall I?’, i.e. ones
where the answers required were instrumental, regulatory, or inter-
actional in function;
(iii) say ‘yes’ and ‘no’ in instrumental and regulatory ways (these were not
used to answer questions seeking information, e.g. ‘Did you see a
car?’).
Table 6.34 Nigel’sfive ways to engage in dialogue, taken from Halliday
(1975: 49)
(1) Respond to wh-question (provided the answer was already known to the questioner), e.g.
‘What are you eating?’ Nigel: ‘banana’.
(2) Respond to a command, e.g. ‘Take the toothpaste to Daddy and go and get your bib.’
Nigel does so, saying: ‘daddy. . . noddy . . . train’, [i.e. ‘Daddy (give) noddy (toothpaste to
him, and go and get your bib with the) train (on it)]
(3) Respond to a statement, e.g. ‘You went on a train yesterday.’ Nigel signals attention, by
repeating, and continues the conversation: ‘train . . . byebye’ [i.e. ‘Yes, I went on a train,
and then (when I got off) the train went away.’]
(4) Respond to a response, e.g. Nigel: ‘gravel’ Response: ‘Yes, you had some gravel in your
hand.’ Nigel: ‘ooh’ [Le. ‘It hurt me.’]
( 5 ) Initiate dialogue, e.g. Nigel: ‘what’s that?’ Response: ‘That’s butter.’ Nigel repeats:
‘butter’.
Proto-dialogue is restricted to the child’s communicative functions at the

time. Halliday does not see these events as dialogues, however, because
Nigel cannot initiate the dialogue and he cannot respond to information-
seeking questions.
Nigel’s first dialogues appeared around 18 months, over a two-week
period. At the end, he could participate in dialogue in five different ways.
These are provided with examples in Table 6.34. As with K, Nigel is mostly
limited to responses to the adult language. This is similar to Lewis’s stage 3.
The only means that Nigel has to initiate conversations is the question
‘what’s that?’. As we will see in the next chapter, the onset of dialogue for
Halliday is an important event which marks the end of Phase I and the start
of Phase 11.
Some further examples of the restricted dialogue of the child in this
period are given in Ingram (1981~).There I report on the early language of a
girl referred to as K. Like Nigel, K had a small number of proto-words that
assisted her in maintaining conversations with adults. One of these was
‘huh?’, which was not used to get the adult to clarify her remarks, but simply
to keep talking. Example (6.18) gives one such dialogue (p. 277; E =
Experimenter), at 1;2(15):
(6.18) E; What are those? (buttons in a jar)
K: huh?
E: What are those?
K: huh?
huh?
(pause)
huh?
tata please (three times)
E: Can’t you get them out?
K: huh?
huh?
E: Youdoit?
K: huh?
The session contained 61 uses of ‘huh’ out of 199 utterances. Two other such
words were ‘tata’, which was used when handing items back and forth, and
‘oooh!’, which indicated excitement.
Another aspect of K s language was that she and her mother (M) would
practice discourse routines. The most striking example occurred in a session
when K was 1;0(26) (p. 276):
(6.19) K: ‘Uncle Ben’ (akhaba]
(I don’t understand and turn to her mother)
M: ‘Uncle Ben’ (whispered to me)
M: ‘Uh huh. Ask her where he is’ (to me)
Me:‘Where’s Uncle Ben?’
K: ‘huh?’
‘Norway’ [i:wei]
M: ‘Norway’ (to me)
K: ‘Norway’ [ni:wei]
Me:‘What’s that?’ (to M)
M: ‘Norway’ (to me)
Me: ‘Norway?’
M: ‘Yeah, that’s where he is.’
K: ‘yeah’ [ja]
In this example, K is clearly too young to have any idea of where Norway is
or even what such an answer would mean to an adult. K’s mother felt that
Uncle Ben’s departure was an important event, and had discussed this at
length with K. Such discourse dialogues allow a mother and child to have
brief conversations, albeit one-sided ones.
6.6.2 Methods used to study the effects of linguistic input

In this period, we see that the child is beginning to understand language,
and can begin to engage in at least proto-dialogues in the sense of Halliday
(1975). From Chapter 5, we also know that at least English mothers are
speaking in a simplified fashion to their children. What effects, then, does
the mother’s input have on the child’s emerging linguistic system?
In Section 5.5.2, we briefly discussed predictions made by the behavior-
ist, constructionist and maturational views of acquisition. We pointed out
that the latter two views place the primary emphasis on the child adjusting
to a changing environment. Even so, it was felt that some environmental
effects may exist, due to the fact that some behaviorist principles will be part
of a theory of acquisition, even if they do not constitute the entire theory.
Also, such effects will need to be determined in order to separate those
aspects of the child’s linguistic behavior from others which are due to the
child’s internal language acquisition system.
The need to examine the role of the linguistic environment ,even without
a nativist orientation, is not always acknowledged by those with a matur-
ational orientation. A primary reason has been Chomsky’s claims
throughout his writings about the inadequacy of the linguistic environment
for the acquisition of language. For example, in Chomsky (1968:88) he
refers to the language that the child hears as ‘degenerate’. There are two
ways in which this term has been used: (i) it indicates that the language the
child hears is full of errors and false starts such that it could not possibly be
an adequate model for acquisition - research on English, at least, has
proven this aspect false, as indicated in section 5.5; (ii) it suggests that the
linguistic input does not provide sufficient information to allow the child to
induce the rich structural properties of human language. For example, a
well-formed, clearly spoken simple sentence is still nothing more than a
string of words, with no information in the signal about phrase boundaries,
the sites of moved NPs, the existence of empty categories, etc. The claim
that the child’s linguistic input is degenerate from this viewpoint is still valid.
In recent years, there has been a return to the issue of the child’s linguistic
environment within the maturational view. In Chapter 4, there was a
discussion of Chomsky’s parameter-setting model which claims that the
child establishes certain key features of the language being acquired which
enable a whole other series of assumptions to be made about the language.
These parameters will be triggered by specific linguistic events in the child’s
environment. The question then arises: what is a trigger experience for the
child? This issue has been raised recently by Lightfoot (1982) and Roeper
(1981), among others. To summarize, then, we now have three goals in
examining the linguistic input to the child: (i) to establish the effects that are
due to the behaviorist principles that will be part of the theory of
acquisition; (ii) the isolation of these effects in order to determine those
aspects of the child’s language that are due to the innate properties of the
child’s acquisition device; and (iii) to specify more clearly what constitutes
the relevant linguistic input for the child to formulate a linguistic rule (or,
within the parameter-setting model, to ‘set a parameter’).
There are several ways in which one might proceed to look for effects of
the linguistic environment. Most studies fall into one of the three following
general categories: (i) experimental studies; (ii) Time 2 studies, and
(iii) Time 1 vs. Time 2 studies. Here I will briefly discuss the general
features of these three types of studies, and then report in 6.6.3 on some
findings for this period. The major studies, however, have been done with
more advanced children; their results, therefore, will be presented in
Chapter 9.
The experimental studies use the traditional psychological procedure of
controlling the child’s linguistic environment, and then monitoring the
effects of that control. One can see the difficulties this presents with
children, since total control of their environment would be impossible
except through unethical procedures. There are, however, two ways in
which this may be done with a minimum of problems. One way is to present
the child with artificial data, which may be in the form of nonsense words, or
even an artificial language. Since these data are created by the experi-
menter, there is not the concern that the child may get additional experience
during the period of the study through normal linguistic interaction.
Another way is to select natural language forms, but only ones that are rare
in the child’s linguistic environment. For example, passive sentences such as
‘The cat was chased by the dog’ are not frequent in the speech of adults to
children. The presentation of such sentences could be controlled, say
through their inclusion in selected reading materials given to the child. Pre-
and post-tests of the child’s receptive or productive ability on the selected
structure could then be studied.
In experimental studies there are at least three aspects that can be
controlled. One is the frequency of presentation: it is useful to know how
often a child in any given stage needs to hear something before it is
acquired. The second aspect, which is related to frequency, is rate of
presentation: is acquisition enhanced more by frequent exposure over a
short period, or infrequent (but regular) exposure over a longer period?
Third, it is possible to look for generalizations from the presented stimuli to
other stimuli. For example, linguistic theory may claim that two structures,
SI and S2, are related by rule. We could present the child with only S2, and
then see if SI is also acquired; if so, the data could be used in support of the
linguistic theory which relates SI and SZ.
The second type of study is what I refer to as Time 2 studies. In these
studies, the mother’s and child’s language are compared at a selected time
of development on some feature or set of features. The studies attempt to
correlate the frequency of use of some aspect of the mother’s language with
its use in the child’s language. Such studies are also called correlational
studies. They typically examine a large number of mother-child pairs, and
select a controlled sample from each pair. For example, we could collect
samples from mothers of 3-year-olds and their children to see if the
frequency of use of passive sentences by mothers correlates with their use by
the children. There are three possible correlations: (i) positive, i.e. mothers
who use more passives have children who use more passives; (ii) negative,
i.e. mothers who use more passives have children who use fewer passives;
and (iii) no correlation. In this example, if we get a positive correlation, we
might conclude that the data suggest that the *other’s frequent use of
passives has caused the child to acquire passives. These studies are called
Time 2 studies since they imply an effect that presumably would have
occurred at an earlier time, say Time 1.
There are obvious difficulties with Time 2 studies. The major one is that
we can never conclude causal effects from correlations - it could be that the
mother is using more passives because her child has acquired them. Time 2
studies, therefore, can only be suggestive at best. Even a result of no
correlation does not mean that there may have been some effect at an
earlier time. Further, such studies imply that the mother has been the
primary linguistic influence on the child. Such studies therefore need to be
quite explicit about the general linguistic environment of the child.
The third type of study, Time I vs. Time 2 studies, has been developed to
overcome the problems of Time 2 studies. To explain them, consider the
following design:
Mothers MI M2
Children C, C?
Assumption: aU children are the same at Time 1 (C,)
Basic design of Time I vs. Time 2 studies
The general orientation of such studies is to observe the child’s language

before and after the acquisition of some feature of language. Their use of
the feature at Time 2 (where C2 = the child’s language at Time 2) is then
examined in relation to their mothers’ use of the feature at Time 1 (MI). If
there is a correlation between MI and C2, then it is concluded that a causal
relation exists.
The first major study of this kind was conducted by Newport, Gleitman &
Gleitman (1977). In their design, C1 consisted of three age groups of
children: 1year to 15 months, 18-21 months, and 2 years to 27 months. This
design was later criticized by Furrow, Nelson & Benedict (1979: 425) as
being based upon two unwarranted assumptions. First, it assumes that
possible effects of the linguistic input will be similar at different ages and
levels of development. Second, it assumes that changes in the use of
particular forms by children will be the same regardless of age or stage.
These are two very strong assumptions. It seems possible that effects, if any,
may differ dependent on the child’s level of development, and that they may
differ dependent on the structure being acquired. To get around these
assumptions, Furrow, Nelson & Benedict add the assumption that the
children must all be linguistically the same at Time 1; I have added this
assumption to the design above.
There is a further aspect of the Time 1vs. Time 2 study which needs to be
pointed out. Suppose that we get a MI, C2 correlation that is the same as the
M2, C2 correlation. According to Campbell & Stanley (1963), such a result
weakens the causal effects since the C2 data are equally interpretable as
being caused by Mz. If the M2,C2 correlation differs from the M1,C2 one,
however, then the only possible cause of the child’s language at Time 2 is
the mother’s language at Time 1.The arrow in the design above implies that
the MI ,C2 correlation differs from that of M2,C2.
Furrow, Nelson & Benedict point out that there m.ay be confounding
third variables that might lead to the correlations found in a Time 1 vs.
Time 2 study. The possibility of such confounding variables makes it
difficult to conclude a cause and effect relation. Even so, the design at least
provides one means to begin exploring the possible effects of the child’s
6.6.3 Some possible effects on lexical development

Most of the research looking for effects of the linguistic environment on the
child’s language has looked at grammatical development beyond the
current period of acquisition. A review of this research will be presented
in subsequent chapters. Here we will briefly examine two studies which
attempt to look at the possible effects of the most obvious milestone of this
period, lexical development.
An important aspect of the linguistic input that is inadequately under-
stood is the effect offrequency of exposure. That is, how often must a child
hear a particular aspect of language before it is acquired? There are several
points that need to be considered in studying this topic. The most important
point is that of readiness. Most likely the role of frequency is minimal until
the child is linguistically ready to develop particular linguistic features. At
the current time we have little understanding within either maturational or
constructionist orientations of readiness. In the area of language disorders
in children, the issue of readiness is a central one since the language
clinician needs to make assumptions about what the child is most ready to
acquire. In that field, such decisions are made by assuming that the normal
child’s order of acquisition reflects sequences of readiness (e.g. Miller &
Yoder 1974). We do not have, however, a well-defined theory of linguistic
readiness.
Given a child’s readiness, there are still the issues of number and rate
of exposures. Using the terminology of Schwartz & Terrell(1983), we can
separate massed presentations from distributed presentations. Massed
presentation refers to a situation where the item under acquisition occurs
very frequently over a short period of time. Moerk (1980), for example,
has proposed that children are very sensitive to short periods of increased
usage by their parents of selected linguistic features. Such an explanation
puts a great deal of responsibility on the parent to recognize the child’s
readiness and present the linguistic feature at the appropriate level of
frequency. Distributed presentation‘ refers to some minimal pattern of
exposure. For example, daily exposure over a particular period of time,
say a week, may be more important to acquisition than massed presen-
tation in a single day. Besides these two types, we can also propose single
event presentation, which means that the child will acquire an item after
just one, clear example. A strong biological theory should contain pre-
dictions about single-event presentations. It will need, however, to
account for how the child deals with input that contains grammatical
errors.
Schwartz & Terrell (1983) make a preliminary attempt to examine the
effects of frequency and rate of exposure on early lexical development,
using the experimental approach described in the previous section. Twelve
children (1;0(21) to 1;3(15) at the onset) were taught 16 nonsense words
(divided into objects and actions) during ten experimental sessions over a
four-month period. All of the children had five or fewer words in production
at the onset of the study. Each nonsense word was taught through exposure
to four exemplars. For example, one nonsense word was ‘objects at the end
of a string or chain’ which was taught with the following exemplars: 1. a
nose clip, 2. a yoyo, 3. a key chain, 4. a drain stopper on a chain. Within
each word, two exemplars were assigned to a frequently presented condition
(or FP condition), and two to an infrequently presented condition (or IP
condition). Exemplars in the FP condition were presented twice each
session, whereas those in the IP condition were presented once. Thus,
across the ten sessions IP exemplars were collectively presented 320 times
whereas FP exemplars were presented 640 times.
Table 6.35 presents the results for the three measures applied to the data
Table 6.35 Summary of results from Schwartz & Terrell (1983) on input
frequency
Condition
Measure FP IP
1. Percentage of exemplars named at least once 44% 29%

2. Mean number of presentations before naming:
(a) object exemplars 12.77 7.59
(b) action exemplars 12.19 7.33
3. Mean number of exemplars named during first ten presentations 4.33 9.42
by Schwartz & Terrell. The first measure shows that the child named more
of the frequently presented exemplars than the infrequent ones. There was
then a significant effect of frequency regarding naming. The next two
measures are particularly interesting. They show that the infrequently
presented exemplars were still acquired around the same rate as the
frequent ones, even though they were presented only half as often. This can
be seen by a reanalysis of measure 2. Measure 2 gives the mean number of
presentations that an exemplar needed to be first named. For the
infrequently presented exemplars, they need approximately seven-and-a-
half sessions (since one exemplar was given per session). The frequently
presented exemplars needed approximately six-and-a-half sessions (a
number obtained by dividing the mean of 12.77 and 12.19 by two, the
number of presentations per session). That is, even though the frequently
presented exemplars were presented twice as often, they were only first
named approximately one session before the infrequent exemplars. In these
data, then, rate of presentation appears to be nearly as important as
frequency.
The Schwartz & Terrell study is an insightful attempt to examine the
interaction of frequency and rate. Two reservations, however, need to be
made. First, they did not look at comprehension, only at production: the
next step would be to examine these factors in relation to the child’s growth
of receptive vocabulary. Second, the difference between frequent and
infrequent presentations within exemplars was not particularly great, and
the study, therefore, cannot be taken as definitive evidence against Moerk’s
proposals concerning the short-term effects of massed presentations.
Different results might have occurred if frequent presentations were
increased, say to four presentations per exemplar per session.
While Schwartz & Terrell studied the acquisition of nonsense words,
Della Corta, Benedict & Klein (1983) looked for environmental effects on
the acquisition of English vocabulary. In Section 5.5.3, there are examples
from Kaye (1980a) of two mothers, one of whom was very directive while the
other was more narrative. In this chapter, section 6.2.2, we have presented
Nelson’s (1973) analysis of children in this period falling into the groupings of
referential and expressive learners (see Table 6.4). Della Corta, Benedict &
Klein report on a study undertaken to see if the expressive vs. referential
distinction can be accounted for by environmental factors.
To study this issue they conducted a Time 2 study in the sense of the pre-
vious section. They collected 50-utterance samples from 16 mothers during
caretaking situations (diapering, dressing, bathing) with their children. The
children were between 1;3(15) and 1;7 and were divided into either
expressive or referential children. The samples from the mothers of the five
most referential and five most expressive children were then separated for
further analysis. Unfortunately, the authors provide no information about
the measures applied to the children to make this separation. Presumably
the children were only using single-word utterances at the time.
The mothers’ samples were subsequently analyzed for general structural
and pragmatic characteristics; an overview of the aspects measured is given
here:
General measures
(a) MLU; yesho questions; common nouns; 3rd person pronouns;
noun/pronoun ratio; noun type/token ratio
(b) number of utterances per caretaking event
Pragmatic measures
(a) communicative intent (13 measures): labeling; requests for infor-
mation; suggestions; prescriptives; proscriptives; phonological
corrections; referential corrections; description; displaced
speech; conventional social expressions; social play; fillers; other
(b) focus of attention: child- vs. mother-oriented speech; task- vs.
context-oriented speech; indeterminant
Evaluation: approval; disapproval; neutral
For communicative intent, the two categories of prescriptives and descrip-
tion are particularly important as possible sources for the referential and
expressive difference in the children. These two were defined as follows
(p. 38):
Prescriptives: Commands made in an attempt to direct the child’s
behavior or verbalizations, e.g. ‘Put the doll over here’, ‘Say Daddy’.
Description: Statements that describe a person’s behavior, actions,
feelings, appearance, etc. or an object or an event that is present in the
immediate situation, e.g. ‘Mommy’s getting your diaper’, ‘You look
sleepy’.
Table 6.36 Selected findings from Della Corta, Benedict & Klein (1983)
on the language of mothers of expressive and referential children
Mothers
Referential Expressive
Measures children children
General measures
MLU 4.30 4.03
no. of utterances per caretaking event *20.8 *6.2
other measures no significant effects
Pragmatic measures (mean no. of each category per 50 utterances)
(a) Communicative intent
description *19.6 * 13
prescriptives *6.8 *12.6
requests for information 8.0 8.8
fillers 6.6 6.6
(b) Focus ofattention
child-oriented speech 26.0 31.4
task-oriented speech 8.4 4.98
(c) Evaluation
approval 31.2 24.4
neutral 13.2 20.6
* Indicates significant differences between the two types of mothers.
Table 6.36 presents the major results of the study. Concerning general
measures, there were no significant differences between the speech of
mothers of referential and expressive children except for the number of
utterances per caretaking event. Even though both groups of mothers had
comparable MLUs (around 4.0) the referential mothers produced three
times as many utterances. Unfortunately, the study was not designed to see
if this difference would lead to a greater rate of acquisition for the
referential children. (We also do not know if the expressive mothers would
have spoken more to referential children.) Regarding the pragmatic
measures, there were no differences between the mothers on focus of
attention 'or evaluation. There were, however, significant differences on
two of the 13 measures of communicative intent, these being prescriptives
and description. The mothers of referential children used more description,
whereas the mothers of expressive children used more prescriptives. The
authors see the high use of description as a possible source of the high
proportion of general nominals in the speech of referential children. They
do not comment on how the high use of prescriptives would lead to the
relatively high use of personal-social words in expressive children (Table
6.4).
Further reading 231
The study by Della Corta, Benedict & Klein is an example of the
difficulties in doing a Time 2 study in order to determine causal effects. The
authors are aware of this problem, but say nonetheless: ‘Although a causal
inference cannot be made, the findings lend some support to the notion that
the mother’s speech has some influence on that of the child’ (p. 42). Of
course, we can also conclude that their mothers are simply responding to
differences in their children. A more definitive answer will require a Time 1
vs. Time 2 study. In such a case, Time 1would be at some point around the
onset of word comprehension. To date, such studies are quite small in
number; they will be looked at in Chapter 9.
Further reading
Lexical development
The early research on word acquisition (reviewed in McCarthy 1954)
concentrated on onset and rate for the purpose of determining norms of
acquisition. In recent years the research has been interested in developing a
theory of semantic acquisition, in large part inspired by Clark (1973a).
Much of the recent research has concentrated on early word over-
extensions.
McCarthy (1954: 523-34) provides a useful overview of the early studies
on word acquisition. Nelson’s (1973) study has provided the frequently
discussed distinction of referential vs. expressive children, and has inspired
a number of studies. A reading of the first three chapters (pp. 1-56) will
provide the major aspects of the study. Also read two of the studies that
follow Nelson’s study, Benedict (1979) and Rescorla (1980). Lastly, Piaget’s
observations on early lexical acquisition can be found in Piaget (1948) for
Stage V (pp. 52-4), Stage VI (pp. 62-72) and the emergence of signs
(pp. 215-24).
Pragmatic and grammatical development

General characteristics of single-word utterances were frequently discussed
by early investigators, with Jespersen (1922) and de Laguna (1927) most
frequently cited. The period was often referred to as the ‘holophrastic’ stage
to indicate that a whole idea was being expressed. A renewal of interest
occurred around 1970 (see Ingram 1971) and has led to an extensive
literature. This effort occurred after several years of concentrated study of
early word combinations (see Chapter 7), for the 1960s were directed
toward precursors to language in the periods of prelinguistic development
and single-word utterances.
A major debate on single-word utterances has been the extent to which
they can be said to contain structure. Some have argued that there is
semantic structure underlying single-word utterances which surfaces later in
two-word utterances (e.g. Ingram 1971; Greenfield & Smith 1976), while
others have denied this (e.g. Bloom 1973). Most of the studies have
concentrated exclusively on production data. The text has discussed some
of the small number of studies on comprehension. Shipley, Smith &
Gleitman (1969) only examined four children in this period, and is better
known for its results on older children (see Chapter 7). Huttenlocher (1974)
was the first study to concentrate exclusively on this period, although it is
only a preliminary report of a study that was never reported elsewhere.
Miller et al. (1980) is one of the few attempts to study comprehension in this
period for a relatively large number of children.
Huttenlocher (1974) provides a very insightful treatment of the problems
of testing comprehension in young children, as well as some important
preliminary results. Halliday’s approach can be seen in Halliday (1975:
8-36), and Greenfield & Smith (1976: 31-65). Reviews which argue for or
against structure at this period are Ingram (1979b) and Barrett (1982)
respectively.
Phonological development
Perception. While relatively little research has been done on phonemic
perception, even less has treated this period of acquisition. Elkonin (1971)
provides a summary in English of the rich literature on research in language
acquisition by Russian psycholinguists. Abstracts of 32 Russian studies are
provided by Slobin (1966) and comprise the major English summary of
Russian research. Most of the literature, however, is unavailable in English
at this time. Most of what is translated can be found in the collection of
articles in Ferguson & Slobin (1973). Thanks to the work of Dan Slobin, the
volume contains seven articles translated from Russian. One of these is an
overview by Elkonin of the famous Russian diary of Gvozdev (1949) of his
son, Zhenya. Another is the study (discussed in this chapter) by Shvachkin
on the development of phonemic perception. Shvachkin’s original study on
Russian children remains the major contribution. Unfortunately, little is
known about Shvachkin. The bibliography done by Slobin (1972) lists four
articles by Shvachkin between 1947 and 1954. According to the index at the
end, the first two articles are on the development of phonemic perception.
Since the latter of these was translated, we can assume that it was the more
important of the two. The last two papers, both published in 1954, are
indexed as studies on syntactic and semantic comprehension. Neither is
summarized in Slobin (1966). Shvachkin’s 1948 paper is valuable for its
creative methodology, descriptive findings, and theoretical proposals. In
the 1970s, a group at Stanford attempted to adopt Svachkin’s method and
Further reading 233
applied it to English children. A brief report of this effort is found in
Garnica (1973). The method was also tried with older children in Edwards
(1974). A different method for the study of phonemic perception was
developed by Eilers & Oller (1976). A general review of the topic for this
period and later ones is Barton (1980).
Production. There is a large and diverse literature on the child’s early
production of words - the text has been restricted to works focussing on the
acquisition of the first 50 words. Much of the literature in the field con-
centrates on children with larger vocabularies than this, usually in the 50 to
150 range.
It is necessary to read Jakobson directly to gain an appreciation of his
proposals for phonological development. Jakobson (1941/68) remains the
major place to begin. The following sections give the most important
discussions of child language: pp. 12-31’46-59’66-81,8491. The revisions
to this original theory are in Jakobson & Halle (1956), section 4 ‘Phonemic
patterning’, pp. 50-5. (A good summary of Jakobson’s theory can be found
in Ferguson & Garnica 1975: 162-9.) The lexical orientation of recent years
can be seen in Ferguson & Farwell (1975), and Stoel-Gammon & Cooper
(1984). An alternative point of view can be found in Goad & Ingram (1988).
Linguistic environment
The 1970s have seen the emergence of the study of the child’s input. In this
time there has arisen an extensive body of works on how adults speak to
their children. We will look at this literature in subsequent chapters. The
Bibliography here is restricted to studies which focus on speech to children
during the period of single-word utterances, and attempts to establish
effects. This limitation results in a much smaller list. The major collection of
articles on linguistic input is that by Snow & Ferguson (1977). An overview
of the topic can be found in Chapman (1981b).
7 The period of the first word combinations
7.1 Defining the period of the first word combinations

The onset of the period of the first word combinations is indicated by several
developments. First, there is a noticeable increase in the child’s spoken
vocabulary. In Smith’s (1926) classic study, for example, the vocabulary size
for her subjects expanded from a mean of 22 words at 18 months to 272
words by age 2. I give here a summary of this jump:
Age No. of subjects Mean no. of words
1;6 14 22
1;9 14 118
2;o 25 272
Further, recall that the child’s receptive vocabulary at 18 months is much
larger than the productive one. The estimate in Table 6.11 indicated that it
was as much as five times larger. There is active semantic development,
therefore, at this time.
A second important development is the gradual onset of multiword
utterances. Table 7.1 gives some general information on the emergence of
multiword utterances for five children. (Similar data are presented in
Table 6.18 for Matthew and Nicky from Greenfield & Smith 1976.) These
data show the gradual emergence of unique syntactic types over a four or
five-month period. Overall, the striking increases begin when the number of
these approaches 100, around 22 months, based on the grouped data. The
early multiword utterances go through a time of slow growth just as lexical
development did during the preceding period. We will refer to the marked
increase in syntactic types around the end of the second year as the syntactic
spurt.
Bloom, Lightbown & Hood (1975) provide MLU measures for two of the
children in Table 7.1, Eric and Peter. Eric’s MLU was 1.42 at 22 months
and 1.69 at 23 months. Peter’s MLU was 1.33 at 23 months and 1.75 at 2
years. That is, for these two children the syntactic spurt coincided with an
MLU of approximately 1.5. As indicated in Table 3.4, Roger Brown (1973)
234
7.1 Defining the period of the first word combinations 235
Table 7.1 The number of unique syntactic types for four children over the
first few months of multiword utterances. The children are Gregory
(Braine I963a), Jennika (author’s diary), Eric and Peter (Bloom,
Lightbown & Hood 1975)
Age Children Mean

Gregory Jennika Eric Peter
1 ;5 5 1
1 ;6 10 3
1 ;7 14 36 10 15
1;s 24 72 37 33
1 ;9 54 130 - 7 57
1;lO 89 256 108 70 131
1;11 350 - 401 81 272
2;0 1,400 - - 243 575
2;l 2,500+ - 902 458 1,029
refers to the development from MLU 1.0 to 1.5 as early Stage I. Here, we
will retain Brown’s term for this subperiod of the period of the first word
combinations, but alter its measurements. Somewhat arbitrarily, I will
specify this period as having the following boundaries. It begins when the
child has produced 25 recognizable syntactic types (i.e. multiword utteran-
ces). This measure is used because of the difficulty, to be discussed, of
identifying the child’s first multiword utterances. It ends with the syntactic
spurt, which is operationally defined as occurring when the child either has
an MLU of 1.5 or has used 100 syntactic types. At this point there is nothing
of theoretical importance in making these boundaries: their purpose is to
enable us to identify children at this stage in our review of relevant studies.
Most generally, I describe this first subperiod of the child’s first multiword
utterances as one during which the child very gradually develops the ability
to use words in novel combinations.
As with the spurt in vocabulary development, the syntactic spurt does not
necessarily indicate that the child’s linguistic system is now adult-like,
Bloom, Lightbown & Hood (1975) report on two patterns of syntactic-
semantic acquisition that their subjects followed after the syntactic spurt up
to approximately MLU 2.00 when their systems converged. Results like
these suggest that the syntactic spurt is followed by another formative
period in which fundamental aspects of the grammatical system are being
established. This subperiod roughly coincides with what Brown has referred
to as late Stage I when the child’s MLU is between 1.5 and 2.0. Here we will
define this second subperiod of the period of the first word combinations as
beginning at the end of early Stage 1 (as defined above), and ending when
the MLU reaches 2.0 or when there are at least 250 unique syntactic types.
236 T H E PERIOD O F T H E FIRST WORD COMBINATIONS
The latter number is based on the figures in Bloom, Lightbown & Hood
(1975) when the individual patterns seemed to merge. We speculate that the
child begins this second period with possibly an idiosyncratic grammar, and
ends it with a grammar that is much more adult-like.
7.2 Grammatical development: an overview
7.2.1 The major studies

This overview of early grammatical development begins with a brief look at
some of the major studies on the topic. As reported in Chapter 3, the first
extensive data were collected by Bloom, Brown, and Braine in the early
1970s. While other studies have followed, many have been reanalyses of
these original data. Given the importance of the claims one would want to
make about the child’s development during this time, it is crucial that we be
aware of the data used to make them. As will be argued, most of the claims
about early grammatical development are based upon a rather small
number of children. This, no doubt, is the result of how difficult and
time-consuming it is to collect extensive data from young children. It will be
necessary, however, to be quite cautious about claims of acquisition which
are based upon such a small sample size.
Of all the investigators in the field, Bloom and her colleagues have clearly
conducted the most extensive data collection and analysis in the area. These
studies cover subjects such as single-word utterances, early syntactic-
semantic relations, imitation, individual variation, discourse, and question
acquisition. They are based not only on Bloom’s original data, but also on
later data on several other children, including Bloom’s daughter Allison.
Together, they represent a formidable body of research that is necessary
reading for any serious student in the field.
Another important series of works have appeared as a result of analyses
conducted by Brown and his colleagues on the original data from Adam,
Eve, and Sarah. A related work is Bowerman (1973a) who was part of the
Brown project and subsequently recorded similar data from a young child
named Kendall. Brown has been quite generous in sharing his data with
other investigators, and Bowerman (1973a) presents the data from Kendall
in an appendix. A recent theoretical work of importance, Pinker (1984), is
in large part based on Brown’s original data.
Lastly, Braine has followed his original study of Andrew, Stephen, and
Gregory with the collection of additional samples from other children. In
1976 he published an influential monograph on early grammatical develop-
ment in which he presented new data from two subjects, David and
Table 7.2 Summary of the more frequently cited published data on
English children during the period of early word combinations
Child Age Source Form of data
Eric I1 1;8 Bloom (1970) Most of the syntactic types used are presented with
Gia I 1 ;7 discussion for Eric I1 and Gia I.
Gia I1 1;8 Examples of the 226 syntactic types produced in
Kathryn I 1;9 Gia 11, and of the 226 and the 767 syntactic types in
Kathryn I1 1;10 Kathryn I and I1 respectively.
Allison IV 1;10 Bloom (1973) Entire language sample of 271 utterances is given
in the Appendix.
Kendall I 1;10 Bowerman Lists in the Appendix are given for 102 syntactic
(1973a) types for Kendall I (MLU = 1.10) and 152
syntactic types for Kendall I1 MLU = 1.48).
Andrew 1;7-1;11 Braine (1963a) List of 102 first syntactic types for Andrew, of 33 of
Gregory 1;7-1;11 the first 89 syntactic types for Gregory, and of 82
Steven 1;ll-2;o first syntactic types for Steven.
David I 1;9 Braine (1976) List of 60 first syntactic types (MLU = 1.3) for
David I1 1;lO David I, and 149 syntactic types for David 11.
Jonathan I 1,11 List of 73 first syntactic types for Jonathan I, and
Jonathan I1 2;l 187 syntactic types for Jonathan 11.
Jonathan, and also reanalyses of data from several other published works. It
is both a useful source of data and an important attempt to present a lean
interpretation of children’s early utterances. Braine’s work has included a
series of experimental studies in addition to the observation of spontaneous
speech.
The subjects of these studies represent a large part of the data used to
make observations on the child’s acquisition of grammar during this period.
Table 7.2 gives a summary of the available published data on these
children. Two points are worth making. First, few attempts have been made
to divide development of this period into subperiods, as is done above.
When this is attempted, it is usually based solely on MLU or age. One has to
be very cautious, however, in comparing a child like Gia I1 (where the I1
refers to the sample time) with an MLU of 1.34 with Eric I1 with an MLU of
1.19, when both children were twenty months old. While they appear
similar on these superficial measures, Eric I1 used 37 unique syntactic types
while Gia I1 used 226. By our definition, Eric I1 is at the onset of early
Stage I while Gia I1 is at the end of late Stage I. Such comparisons are
vulnerable to type errors, as discussed in Chapter 3.
A second point concerns the small amount of data on children in early
Stage I as defined above. Many of the samples are for children who already
have 100 syntactic types, e.g. Kendall (see Table 7.2). Bloom (1973)
238 THE PERIOD OF THE FIRST WORD COMBINATIONS
exemplifies the problem in getting data on this subperiod. She presents
data for her daughter Allison at four ages, and the numbers of syntactic
types for these four samples are 11,4,19, and 94 respectively. We see that
the first three samples are before this subperiod, and the last one is at its
end. Allison’s data, therefore, tell us little about development up to the
syntactic spurt. Much of the data on this subperiod are from Braine’s sub-
jects. Unfortunately, they are presented without any information about
daily longitudinal changes. We can only conclude, therefore, that our data
on this period remain relatively meager, despite the incredible amounts of
effort expended on collection.
7.2.2 The transition to multiword utterances

The period of the first word combinations has been defined as beginning
when the child has shown evidence of 25 syntactic types in its productive
language. This has been suggested because the first multiword utterances
are not easy to identify, since they are difficult to distinguish from
sequences of single-word utterances. Through the period of single-word
utterances, the child will bring its single-word utterances closer and closer
together. This is true for the contextual aspects of the utterances as well as
their acoustic properties. We begin by describing the facts of this situation,
then turn to possible explanations.
Description Bloom (1973) observed this gradual approximation of single-

word utterances in her study of her daughter Allison, and separates them
into two types: chained, and holistic. Chained successive single-word utter-
ances are ones in which each utterance refers to its own event or activity.
An example adapted from Bloom (1973:47) is given in (7.1).
(7.1) Allison ut 16 months, 3 weeks
Allison: ‘cow’ (three times, reaching under a chair and picking up a
toy cow)
‘chair’ (twice, trying to put the cow on its hind legs on the
chair)
Mother: ‘What’s that?’
Allison: ‘Mama’ (giving cow to mother for help)
Each of Allison’s utterances has its own contextual reference. ‘Cow’ is used
referring to the act of picking up the toy cow, as if it meant ‘here’s the COW’.
‘Chair’ is spoken as Allison puts the cow on the chair with a meaning such
as ‘I’ll put it on the chair.’ Lastly, ‘Mama’ is a call to her mother to help
her.
Holistic successive single-word utterances differ from chained ones in
that the utterances are both referring to the same activity. An example from
Bloom (197351) is shown in (7.2):
(7.2) Allison at 20 months, 3 weeks
Allison: ‘up’ (twice, while wearing her coat and pointing to her neck)
Mother: ‘What?’
Allison: ‘neck’
‘up’
Mother: ‘Neck? What do you want? What?’
Allison: ‘neck‘
Mother: ‘What’s on your neck?’
Allison: ‘zip’ (three times, pointing to her zipper and lifting chin)
Here Allison is referring to a single activity, which is her desire to have her
mother zip up Allison’s coat. The words ‘up’, ‘neck’, and ‘zip’ are all used in
reference to this one act. The holistic sequences show an advance on the
child’s part, since the utterances are more closely bound to each other
contextually. Bloom points out that the holistic sequences only appear
toward the end of the period of single-word utterances, a few weeks before
the first multiword utterances. She sees these as important precursors to the
eventual development of grammatical structure which underlies multiword
utterances.
The gradual proximity of single-word utterances has also been demon-
strated acoustically in studies by F6nagy (1972) and Branigan (1979).
Fonagy is a well-known Hungarian phonetician who tape-recorded the early
development of his two children, Eve and Pierre, and acoustically analyzed
selected utterances. The first single-word utterances were clearly separate
utterances. There was a noticeable time between them and they each had
their own intonational contour and stress. The intonational contour con-
sisted of an initial rise in pitch to a peak, and then a terminal fall. He states
that between the use of single-word utterances of this kind and later
multiword utterances ‘there is a more or less long period of transition.
During this period single-word utterances show a tendency to group
together to form sequences, especially in pairs’ (p. 65; my translation).
These successive single-word utterances differ from multiword utterances in
having one or more of the following features: (i) a brief pause between
them, (ii) identical intonational contours on each word, (iii) primary stress
on both words. Fonagy measured his children’s successive single-word
utterances for these features and concluded that there were four stages to
the transition period. These are given in Table 7.3, with ‘stage’ here in the
sense of the co-occurrence of new behaviors with old ones. That is,
sequences of single-word utterances do not neatly follow through these four
stages as successive events; rather, new types will coexist with old ones.
240 THE PERIOD O F THE FIRST W O R D COMBINATIONS
Table 7.3 Four stages of successive single-word utterances proposed by
F6nagy (1972)
Stage I: Each word has the same intonational contour and a primary stress. It is only their
closeness in time which suggests any relation.
Stage 2: There is still a pause between the words, and each has a primary stress. The terminal
falling pitch of the first word, however, is not as great as that for the second one.
Stage 3: The stress on the first word is weaker than that of the second one, and the pause
between them is reduced (from examples, apparently to less than 500 msec).
Stage 4: There is no longer a pause between the two words. A succession of two words is still
indicated because of (i) the force of the two accents; (ii) frequently a glottal stop occurring
between the two words; and (iii) the terminal falling pitch of the first word.
Further, it appears that Fonagy presents these as a simplification of the facts

he observed. For example, his analysis of the types of utterances used by his
son at 20 months shows more kinds than suggested by the stages in
Table 7 . 3 . Nonetheless, his analysis provides important information about
the gradualness of the change from single- to multiword utterances.
Fonagy’s initial insights have received support in Branigan (1979).
Branigan acoustically analyzed the successive single-word utterances of
three children during the period leading up to the first multiword utteran-
ces. Using a range of information, Branigan found that adults perceive two
words as successive single-word utterances when the pause is between 400
and 1100 msec. He then measured the falling pitch level of words that were
in single-word utterances, successive single-word utterances, and multi-
word utterances. The results confirmed the observation made by F6nagy
about his stage 2, i.e. he found that the falling contour of the first words of
successive single-word utterances was less than that of the second word.
Importantly, he also found that these contours were similar to those in the
child’s multiword utterances. Branigan concludes from this that the succes-
sive single-word utterances are therefore closer in nature to multiword
utterances than to single-word utterances. While an interesting conclusion,
it needs to be viewed with some caution because of the measure he has used
to divide utterances into successive single-word utterances vs. multiword
utterances. By categorizing successive words with an interval of less than
400 msec as multiword utterances, he is placing most of the utterances of
Fonagy’s stages 3 and 4 into this category. We know from Fonagy’s study
that the first words in these stages do not yet have the terminal pitch of later
multiword sequences. It appears, therefore, that the results are confounded
by a measurement sequence of the kind discussed in Chapter 3.
Explanation The fact that successive single-word utterances come to refer

to one context as well as sharing acoustic properties has been used by some
investigators as evidence that these sequences share a common grammatical
structure (e.g. Scollon 1976; Branigan 1979; Ingram 1979b). The proposal
is something like this. During the period of single-word utterances, the child
begins to have some rudimentary understanding of the grammatical struc-
ture of the adult language. As discussed in the previous chapter, there is
little evidence of understanding syntactic structure, but there is some
indication that some semantic relations are understood. During at least the
latter part of this period, this knowledge underlies the child’s single-word
utterances. For example, the child who says ‘cow’in referring to picking up
a toy cow may be using that with the underlying structure of Action +
Object. Later, holistic sequences are further expressions of this same
semantic structure. The child at this point may say the sequence ‘up’ ‘cow’in
the same context, Later still, a multiword utterance such as ‘up cow’ will
also reflect this knowledge.
This viewpoint is what has come to be known as a rich interpretation of
single-word utterances. It has been argued for in Ingram (1971), Greenfield
& Smith (1976), and Scollon (1976), among others. Scollon (1976) has used
the terms ‘vertical construction’ and ‘horizontal construction’ to discuss this
issue. A horizontal construction is a construction which has its constituents
represented in a single sequence of words as in the English sentence ‘I want
a cookie’. Adult grammar is typically a series of horizontal constructions.
Scollon argues that children acquire structures as horizontal constructions
only after they first appear as vertical constructions. A vertical construction
is a construction which has its constituents appearing in a sequence of
utterances. Holistic sequences of single-word utterances would be called a
vertical construction by Scollon’s definition. An important feature of
Scollon’s claim is that it is not just restricted to the discussion of the
transition from single-word to two-word utterances. Using data from his
subject Brenda, he gives cases where three-word horizontal constructions
arise from earlier vertical constructions consisting of a sequence of a
single-word utterance and a two-word utterance. For example, the semantic
structure Agent+Action+Object of the utterance ‘baby eat cookie’ may
occur at an earlier time as ‘baby’ ‘eat cookie’.
The rich interpretation of single-word utterances has been a controversial
issue since its first inception. A major feature of this interpretation is that it
is consistent with a constructionist view of acquisition where development
of linguistic knowledge is seen as beginning in comprehension. In this case,
the primitive semantic relational knowledge emerging in comprehension
first appears as underlying single-word utterances, then successive single-
word utterances, and eventually multiword utterances. Further, it proposes
that new grammatical forms, such as successive single-word utterances, are
used with already acquired (or ‘old’) meanings. The idea that new forms
242 THE PERIOD OF THE FIRST W O R D COMBINATIONS
mark old meanings and that old forms mark new meanings is a basic
Piagetian notion that has been applied to child language by Slobin (1973).
The rich interpretation is in opposition to what can be called the lean
interpretation of successive single-word utterances. This position has been
expressed in Brown (1973), Bloom (1973), Dore (1975) and Barrett (1982),
among others. It has several sides to it which we will first introduce here and
then discuss. First, one version claims that the young child in this period is
mapping conceptual representations directly onto its single-word utterances
(Bloom 1973; Brown 1973). Allison’s expression of ‘cow’ in (7.2), for
example, is mapping the child’s knowledge of the toy cow in this context,
not a linguistic category like Object or Agent. Bloom (1973) relies heavily
on Piagetian theory in taking this position, emphasizing the limits of the
child’s cognitive ability at the end of the sensorimotor period. We have
discussed this aspect of development in the previous chapter, and have
pointed out its use to account for word overextensions. The argument is that
the sensorimotor child is unable to categorize and represent reality, abilities
necessary for the use of a grammatical system.
A second aspect of the lean interpretation is that it restricts the child‘s
meanings at this time to pragmatic functions such as expressions of desires
and demands, as discussed in the previous chapter. For example, Barrett
states (1982:62) that these utterances consist ‘of a lexical item serving a
particular communicative function, this function being communicated to
the listener by the intonation and gestures of the child’. Third, it argues that
predictions of the rich interpretation are not supported by the data. Two
predictions in particular are challenged: (i) that the child does not use
two-word utterances initially because of an inability to produce utterances
longer than a single word; and (ii) that successive single-word utterances
should show the word order of later multiword utterances.
In comparing these latter alternatives, it is important to point out at the
onset that they are not as opposed to each other as may initially seem. The
first rich interpretations appeared in the early 1970s as an exploration of the
hypothesis that features of two-word utterances may have their origins in
single-word utterances (e.g. Ingram 1971). These early analyses focussed on
single-word data independent of multiword utterances. In Ingram (1979b) ,
I looked at these data and found that most studies which are claimed to have
a rich interpretation divided the development of single-word utterances into
two periods. The first one lasts from around 1year to 16 months and has the
characteristics of the lean interpretation. That is, the child‘s single-word
utterances do not show structural properties, and their meaning appears to
be primarily functional in nature. The sequences of single-word utterances
that do occur are chained at this time. There is no substantive disagreement,
therefore, concerning this initial period.
The next period is one around 16-18 months when the single-word
utterances are analyzed as reflecting semantic categories like Action and
Object. As pointed out in Ingram (1979b), this period has the following
characteristics: (i) it occurs just a few weeks before the first multiword
utterances begin; (ii) it is characterized by holistic sequences of single-word
utterances; and (iii) it appears to coincide with other behaviors character-
istic of the onset of representation (i.e. the end of the sensorimotor period)
such as ‘what’s that?’ questions, symbolic play, references to past events,
and the onset of dialogue. It is in the description of this period where the
primary issue of a rich vs. a lean interpretation exists.
Since the critics of the rich interpretation tend to lump these two periods
together, it is not clear how they would directly support the lean interpreta-
tion for this second period. The general argument would probably be a
narrow interpretation of the Competence Assumption, as suggested in the
following statement from Bloom (1973:19-20):
Whether or not children do #havepre-syntax knowledge of grammar
may well be unknowable; that is, it is not clear that any amount of or
kind of evidence can demonstrate convincingly that children know
about sentences before they say sentences.
A similar note of caution appears in Barrett (1982). It is still the case,
however, as pointed out by Bloom (1973:19), that ‘The most persuasive
argument in favor of this position is that it offers a strong explanation for the
transition to syntax.’ While the lean interpretation may show a certain
methodological caution, it needs to answer some very difficult questions.
For example, how does it account for the discontinuity in development
between single-word and multiword utterances? Can the single-word utter-
ances and holistic sequences that occur after the onset of multiword
utterances be given a rich interpretation, and if not, why not?
The most serious arguments against the rich interpretation appear to be
the two stated above concerning the predictions it makes that do not appear
to be supported by the data. The first prediction is that the child has some
performance factor such as a physiological restriction of the length of
utterances which restricts h i d h e r to single-word utterances or sequences.
The data from F6nagy certainly suggest that this is the case, but Bloom
(1973) refers to data from her daughter Allison which seem to contradict
this position. During the period of single-word utterances, Allison pro-
duced a number of utterances which consisted of one of her words and a
phonetic form transcribed as lwidal. Bloom concluded that this form was an
empty form that could occur with any word in Allison’s language. ‘It has
been concluded that “wida” apparently referred to anything and every-
thing, and thus it “meant” nothing’ (Bloom 1973:34). These utterances with
/wida/ are taken as evidence that Allison could put two words together, but
didn’t because she had no grammar to allow her to do so.
Ingram (1979b:273) discusses some problems with these data; for
example, they contradict all the data such as F6nagy’s which indicate that
multiword utterances follow after a gradual approximation of single-word
utterances. Also, if this type of behavior is possible, it is surprising that none
of the many diary studies over the years have observed a similar case, i.e. it
is a unique characteristic of Allison’s language. Most important, there is a
much simpler explanation possible. With some very reasonable assump-
tions, it can be shown that Iwidal was probably Allison’s pronunciation of
her own name. To begin, /wida/ most often occurred as the variant form
/awida/. (7.4) shows how these two forms could result from the name Allison
by a series of common phonological processes of simplification which will be
discussed in the next chapter.
(7.3) Allison
alrsan
alIsan a + a vowel neutralization
alisan I + i vowel tensing
awisan 1 + w gliding
awisa n + 0 final consonant deletion
awita s + t stopping
awida t + d voicing
wida a + 0 syllable deletion (optional)
There is even support for this interpretation in Bloom’s own discussion of
this phenomenon. ‘. . . it occurred often in action events, as Allison either
did something or something happened; and it often occurred in situations
where Allison wanted to have or do something’ (Bloom 1973:34). If the
phonological explanation is correct, it removes this argument against the
rich interpretation, since these utterances with /wida/ would then be
multiword utterances.
A more serious challenge is the claim that the rich interpretation predicts
that sequences of single-word utterances will show the restrictions on word
order that are evident in the child’s first word combinations. As will be
discussed later, early word combinations appear to adhere to restrictions on
the order of their constituents. No such ordering, however, has yet to be
reported for sequences of single-word utterances. This is a problem, i
though, only if we can demonstrate that order is also a feature of the child’s
knowledge at the time when successive single-word utterances are occur-
ring. Evidence so far is lacking in this regard. It may be that children first
understand some basic semantic relations and only later work out their
ordering for the language. Also, there is some evidence (e.g. Braine 1976)
that not all early word combinations show such ordering. It may be that the
ordering is really the result of adults imposing adult interpretations upon
children’s utterances, as pointed out by Howe (1976). Even with this
response, it remains the case that the lack of ordering in these sequences is a
potential problem for the rich interpretation as currently presented.
Before concluding this section, there is one last point that needs to be
made about the rich interpretation. While we have used the term ‘rich’, the
view presented is by no means the richest interpretation that could be given.
We could propose, for example, that there is an underlying syntactic
structure, or even that the underlying structure is the grammatical structure
of the adult language being acquired. This is, however, far from the case.
The rich interpretation discussed above is actually quite conservative, and
conforms to the Competence Assumption of Chapter 3. That is, the
competence assigned to the child is based on behaviors such as the child’s
understanding and its use of sequences of single-word utterances. It only
claims that there is some evidence for primitive semantic structures under-
lying the child’s utterances during the weeks before the first multiword
utterances.
7.2.3 Some general characteristics

The discussion of early word combinations often starts from a list of unique
syntactic types that are used as data for the analysis being presented. While
this is a reasonable place to start an analysis, it can distort the picture of the
child’s data at this time. Children during this period do not speak in a
sequence of clear, unique syntactic types, and in fact the data used consist of
a short list of sentences that had been put together after several hours of data
collection. This section, then, presents a general discussion of what the
child’s actual language looks like during this period.
One aspect to keep in mind about this period is that it is named after the
new behavior which appears, that is, word combinations. Throughout it,
however, the most frequent production of the child is the single-word
utterance. This can be exemplified by looking at the data from one of
Bloom’s subjects, Eric. Based on data given in Bloom, Hood & Lightbown
(1974), Eric had an MLU of 1.42 at 22 months of age. This was calculated on
a sample of utterances collected during nearly seven hours. During this
time, 1,043 utterances were elicited. According to Bloom, Lightbown &
Hood (1975) only 165 of these were multiword utterances. Further, only 108
of these were unique syntactic types. Clearly, Eric’s predominant pro-
duction at this time near the end of early Stage I is the single-word
utterance. It is also striking that it took seven hours to collect 108 sentences
for a grammatical analysis. This fact alone is highly suggestive that Eric
must not yet have a very productive grammar, or else that severe perform-
ance factors are at work.
The limits of the child’s productive grammar can be seen by looking at a
sample of multiword utterances for a child during early Stage I. Table 7.4
gives all of the word combinations produced by Andrew over the first five
months of multiword usage, based on Braine (1963a). As mentioned in
Chapter 2, these data were collected by the mother through a parental
diary. The fact that it took five months to obtain the sample is another
example of the limited productivity of the grammar at this time. When we
examine this sample, we see that there is a small number of words that are
especially frequent, e.g. ‘all’, ‘no’, ‘more’, ‘there’. Also, they tend to occur
consistently in one sentence position: ‘no’, for example, occurs initially
while ‘there’ occurs finally. Braine emphasized these two facts in his
observations of this period and incorporated them into his explanation of
the child’s early grammar. This theory, which is referred to as ‘pivot
grammar’, will be discussed in 7.3.2.
Another feature of Andrew’s sample is its lexicon and how it enters into
word combinations. In Table 6.7 of the last chapter, we gave the semantic
categories proposed by Nelson for the child’s early words. When we look at
Andrew’s combinations, we see that they are predominantly nominals
combining with either action words, modifiers, or personal-social words.
This tendency in early grammars has been found in several studies, e.g.
Bloom (1970). While nominal+nominal combinations do occur, they do not
seem to be frequent in the earliest word combinations. A crucial issue in
explaining such early utterances is the interplay of the distributional
features of these words with their semantic properties. As we will see, some
accounts emphasize the distributional aspects of these words while others
concentrate on their semantic properties.
While most analyses usually work from lists like the one in Table 7.4, it is
important to realize that there are also contextual aspects of these utterances
that can aid in understanding their structure. For one thing, children do not
produce a sequence of unique lexical or syntactic types. The samples during
this period are full of repetitions and variations upon some idea being com-
municated. Although adults express a proposition within a single sentence,
the equivalent child thought may span several utterances. The various sen-
tences are connected for the child in the sense that they are combining to
express what appears to an adult speaker to be a single proposition. Such
sequences are similar to what Scollon (1976) has called vertical construc-
tions. Since this term implies the existence of a single structure, we will not
use it here. Instead, we will refer to such sequences as sentenceparugruphs.
This is a strictly descriptive term which refers to a sequence of child utteran-
ces that could be expressed as a single adult sentence.
Table 7.4 Andrew’s pivot grammar, as given in Braine ( 1 9 6 3 ~Table
: 2)
Pivotal constructions
all broke no bed more card other bib airplane byf

all buttoned no down“ more cereal other bread siren by
all clean no fix more cookie other milk
all done no home more fish other pants mail come
all dressed no mamab more high‘ other part mama come
all dry no more more hot other piece
all fix no pee more juice other pocket clock on there
all gone no plug more read other shirt up on there
all messy no water more sing other shoe hot in there
all shut no wetC more toast other side milk in there
all through more walk light up there
all wet see baby boot off fall down there
see pretty hi Calico light off kitty down there
I see see train hi mama pants off more down there
I shut hi papa shirt off sit down there
I sit shoe off cover down there
water off other cover down there
Other utterances
airplane all gone byebye back what’s that look at this

Calico all gone byebye Calico what’s this outside more
Calico all doneg byebye car mail man pants change
salt all shut byebyepapa mail car dry pants
all done milk Calico byebye our car off bib
all done now papa byebye our door down there
all gone juice papa away up on there some more
all gone outsideh
all gone pacifier
a ‘Don’t put me down’

‘I don’t want to go mama’
‘I’m not wet’
‘Drive me around some more’
e ‘There’s more up there’
f ‘A plane is flying past’
g Said after the death of Calico the cat
Said when the door is shut: ‘The outside is allgone’
Weir (1962) was one of the first to identify different types of sentence
paragraphs. She observed these in the presleep monologues of her son
Anthony between the ages of 26 and 28 months. The three types she found
are build-ups, breakdowns, and completions. In (7.4) these three types are
presented, with examples taken from Bloom’s (1973) corpus of Allison at 22
months of age. Allison had an MLU of 1.73 at this time, placing her in late
Stage I.
(7.4) a. Build-ups
‘baby eat’ ‘baby doll ride’
‘baby eat’ ‘baby doll ride’
‘cookie’ ‘truck’
‘baby eat cookie’ ‘baby doll ride truck’
b . Breakdowns
‘walking around’ ‘there baby’
‘around’ ‘there’
c. Completions
‘mommy’ ‘cow’
‘comb hair’ ‘stand up’
In Allison’s data, the build-ups where the child shows the parts of an
utterance before eventually getting them all together seem to be the most
common type. Breakdowns are similar to build-ups except that they are
going in the opposite direction. Both types indicate some awareness of the
parts of the longest sentence produced. There are two extreme explanations
that can be offered as to why the child is producing these. Using the terms in
Peters (1977), we could propose that the child is learning language either
analytically or synthetically. The child is being analytic if she is using rules of
combination to produce longer sentences. The other possibility is that she is
acquiring language synthetically, by learning an entire sentence and then
breaking it down into its different parts. Unfortunately, data like these do
not alone help us to determine which of these is happening. The com-
pletions show less evidence for a single underlying structure. Such an
assumption would be supported, however, if a single utterance such as
‘mommy comb hair’ and ‘cow stand up’ were found elsewhere in the sample.
Although we write the child’s sentences in adult orthography, it is
important to realize that the child is not actually pronouncing most words
‘correctly’. Later we will look at some of the kinds of phonological patterns
children use in this stage, but first we need to examine the implications of
these imprecise speech forms for the child’s speech and for our attempts to
analyze it.
One obvious effect of the child’s incomplete phonology is that there are a
number of child utterances which will be unintelligible. Bloom (1970: 106),
for example, reports that she was only able to elicit 490 intelligible
utterances from Eric in a six-hour visit when he was 20 months, 2 weeks old.
This can be contrasted with the 1,043 intelligible utterances she collected
from him during a comparable sampling six weeks later. His MLU was 1.19
and 1.42 for these two samples respectively, suggesting a noticeable
increase in intelligibility by the end of early Stage I. Also, during the time of
I
the first word combinations, some children show consistent uninterpretable

vocalizations alongside interpretable words. For example, Bloom (1970)
transcribed several sentences from Eric which contained an element given
as [a]. She was not able to identify a specific morpheme for this, although
possible candidates included ‘I’, ‘a’, and ‘the’. It was simply impossible on
the basis of its occurrence to decide the issue. Elsewhere, Dore, Franklin,
Miller & Ramer (1976) have observed a similar phenomenon and call these
phonetically consistent forms.They claim that these do not have any content
at this stage, but only reflect some awareness on the child’s part of some
phonetic elements occurring before the nouns and verbs it hears in adult
speech. We will return to this point in 7.2.4, when we discuss the child’s
comprehension ability during this period.
A last general aspect of the first word combinations is the question of
interpretation. Even when we can identify the words in the child’s sentence,
we need to attempt to determine what the child meant by it. In Table 7.4,
for example, Braine gives us in the footnotes several interpretations of
sentences. These interpretations are crucial in making judgements about
the underlying structure of the child’s sentences, but unfortunately there is
no totally satisfactory way of assigning meanings. There is always the
danger, as pointed out in Howe (1976), that we may impose our adult ideas
of meaning onto the child - the child could be intending something totally
different from what we think. One needs to be quite cautious, therefore, in
assigning meanings to children’s first sentences.
Even with this warning, it is possible to establish some guidelines for
interpreting children’s sentences. Bloom (1970) discusses three kinds of
evidence that can be used. First, there is the non-linguistic context of the
utterance. The child who says ‘mommy eat’ while watching the mother eat a
cookie is more likely to be referring to the mother as an Agent than as an
Object. Second, there are the preceding adult utterances. For example,
when Bloom (1973:246) says ‘What are you doing?’ to Allison, and she
responds ‘chewing’, it is likely she means ‘I’m chewing’ rather than ‘he’s
chewing’. Third, there are also the child’s subsequent utterances. These
may be build-ups or completions as shown in (7.4) which clarify the
intended meaning. Even with these guidelines, it is still not always clear
what the child intends. Bloom (197053) gives a useful taxonomy of the
kinds of problems in interpretation that may occur. These are given in
Table 7.5.
7.2.4 The onset of syntactic comprehension

In section 6.4.1 data were presented which suggested that during the period
of single-word utterances children are capable of understanding more than
Table 7.5 Four categories of interpretation for children’s sentences, taken
from Bloom (1970.53)
Ambiguous: a form having two (or more) possible interpretations that can be distinguished or
resolved - for example, ‘Mommy sock’ meaning ‘Mommy’s sock‘ or, alternatively, ‘Mommy
(verb) sock’;
Equivocd: a form having two (or more) possible interpretations that cannot be distinguished -
either one or the other interpretation being acceptable in the particular situation, for example,
‘Mommy iron’;
Indeterminare: a form for which an interpretation cannot be made, most often because of
insufficient evidence;
Anomalous: a form that appears to have no interpretation- the occurrence of an utterance in a
situation to which the linguistic expression bears no apparent relation- for example, Gia eating
peaches and saying ‘no more’.
one word in a sentence, and that they may even relate these semantically.
There was, however, no convincing evidence that they understand gram-
matical morphemes. In the period of the first word combinations, there is
some evidence which suggests that syntactic understanding is taking place.
As with the studies on the earlier period, these studies need to be
interpreted cautiously, owing to the inherent difficulties in testing com-
prehension in young children. Here we will review four studies which
indicate syntactic processing of grammatical morphemes in advance of their
appearance in spoken language. In one of these, we see another method for
determining a child’s grammatical knowledge, that of elicited imitation.
Shipley, Smith & Gleitman (1969) The major features of this study have
already been given in 6.4.1. To summarize: they asked parents of young
children to direct a series of commands to their children which varied in their
syntactic structures. There were three main commands: well-formed (VFN)
‘Throw me the ball!’; telegraphic (VN) ‘Throw ball!’; and holophrastic (N)
‘Ball!’ The children’s responses to these commands were monitored to see if
they would show a preference for one type over the others. Chapter 6 pre-
sented the results of this study with the four holophrastic subjects. We inter-
preted these results as suggestive of some preference for well-formed and
telegraphic commands over holophrastic commands, although this was only
a statistical tendency which did not prove significant.
Here we will present the more clear-cut results that were found for the
second group of subjects who were tested. This second group, referred to as
the telegraphic group, consisted of seven children between the ages of 19
and 32 months, with a mean age of 25 months. Their MLUs ranged from
1.40 to 1.85 with a mean of 1.57. This measure places these subjects in the
middle of the period being discussed in this chapter.
Table 7.6 Percentage of times each of seven children in the telegraphic
group in Shipley, Smith & Gleitman (1969) would touch an appropriate
toy in response to each of three constructions
Subject’s Structures
Name Age MLU N(%) VN(%) VFN(%)
Carl 23 1.85 33 33 58
Dottie 19 1.75 15 27 36
Eric 25 1.65 25 28 38
Fran 29 1.48 21 54 64
Gregory 28 1.43 37 25 57
Helen 21 1.41 33 38 62
Ira 32 1.40 50 33 54
Mean 31 34 53
Table 7.6 presents the percentages of times each of the telegraphic

children touched an appropriate toyjn response to each of the three types of
commands. These results can be compared with those for the holophrastic
group as presented in Table 6.12. While the holophrastic children showed a
preference for the simpler commands, the telegraphic ones showed the
opposite preference. A statistical analysis of the data revealed that ‘all
telegraphic speakers obey more often with well-formed commands than
with single word commands . . . The results for VN are similar but less
sharp’ (p. 329). Even though the children were producing predominantly
single-word and simple word combinations, they showed an awareness that
adult speech required more than what was occurring in their own output.
As pointed out in Gleitman, Shipley & Smith (1978), this result does not
mean that the telegraphic children understood the syntactic role of the
grammatical morphemes in the well-formed commands. It is, however, a
first step in demonstrating that they do. Also, even if they didn’t, the fact
that they recognized that ‘something’ needs to occur between lexical
morphemes provides some information on why they might use phonetically
consistent forms in their own speech, as mentioned at the end of the last
section. The next study, however, suggests that the children may process
these morphemes as more than meaningless noise.
Katz, Baker & Macnamara (1974) We know from Benedict’s study (1979)
discussed in 6.8.2 that specific and general nominals are a prominent part of
the receptive vocabulary of the holophrastic child. The paper by Katz,
Baker & Macnamara is a brief report on research into the question of
whether or not young children know that the difference between proper and
common nouns in English is normally marked by the absence or presence of
an article. In normal usage, proper nouns such as ‘John’, ‘Mary’, and
‘Rover’ occur without either an indefinite article ‘a’ or a definite article

‘the’. These are required, however, for common nouns, except in cases
where the plural is used for the class as a whole, e.g. ‘Dogs like to bark’.
Katz, Baker & Macnamara developed a clever experimental design to
test for this distinction. The children in the study were assigned to either the
common-noun or the proper-noun condition. The idea was to show the
children in both conditions a pair of dolls. The dolls were assigned labels
which were nonsense syllables, e.g. ‘zav’, ‘mef’ and ‘jop’, among others. In
the common-noun condition, one of the dolls was introduced to the child
as ‘a zav’ (or whatever nonsense name had been assigned to it. Here we
will use ‘zav’ for purposes of demonstration). In subsequent references, it
was called ‘the zav’. The other doll was never named or referred to. The
proper-noun condition was similar except that the doll was introduced as
‘Zav’, with no article ever used in referring to it.
During the learning period of the study, the experimenter attempted to
familiarize the child with the dolls. Each of the dolls was mentioned to the
child at least five times. The unnamed doll was referred to only as ‘the other
one’ or ‘this one’, whereas the named one was referred to according to
which of the two conditions was being tested. After this was done, testing
began. Both dolls would be placed within reach of the child, and he was
asked ‘to dress, undress, feed, hold, or bring the named doll to his mother’
(p. 470). The experiment was flexible in nature, using Piaget’s clinical
method discussed previously. The important thing was to determine which
doll the child selected for the requested activity. In the proper-name
condition, the child should select just the named doll if he knew that proper
nouns occur without articles. In the common-noun condition, the child
could select either doll, since the named doll was only introduced as ‘a zav’,
Le. one of the class of ‘zavs’ which could presumably include the other
unnamed doll. Every child was tested at least seven times, and many were
tested as many as 15 times.
This procedure was used in two experiments. The second experiment
differed from the first only in the nature of the dolls and in the subjects used.
In the second one, the dolls were made more distinct than in the first to see if
this would encourage a greater experimental effect. Information on the
subjects in the two experiments is given in Table 7.7. Unfortunately, we are
not given any information on the linguistic level of the subjects. The mean
ages, however, were typical of children in the period of the first word
combinations. Also, the youngest group in experiment 2 was quite young,
with a mean age of only 17 months, which would probably place them at the
beginning of early Stage I.
Table 7.7 gives the percentage of times that the subjects in the two
experiments selected the named doll when prompted to act upon the dolls.
Table 7.7 Mean percentage of times each experimental group selected the
named doll in Katz, Baker & Macnamara (1974)
Condition
Experiment Subjects Mean age Common-noun Proper-noun
1 30 girls 22 mos. 48 75
25 boys 24 mos. 47 51
2 10 girls 17 mos. 42 76
15 girls 22 mos. 53 72
Statistically significant differences were found between the two conditions

for the three groups of girls. The authors suggest that the boys did not yet
have the distinction, either because they were developmentally slower, or
because they were less experienced with dolls. Importantly, the results with
the girls, particularly with the 17-month-olds, indicates that they had some
awareness of the grammatical role of articles.
There is one further aspect of this study that needs to be mentioned.
Katz, Baker & Macnamara were not just interested in proving understand-
ing of one of the functions of articles. They also were interested in how this
acquisition takes place. There are two possible explanations that they
considered. One explanation, which we will call the cognitive explanation, is
that the child cognitively categorizes the world into classes that differ in the
importance of the individual members. Since people (and presumably dolls)
are named and individually stand out, children distinguish between specific
and general references, and thus labels. This distinction is then used to
determine the role of the articles. The alternative explanation, which we
will call the distributional explanation, is that the child’s syntactic processing
distinguishes the distributional properties of the articles. He then looks for
semantic properties for these grammatical morphemes.
To test these alternatives, they also tested the subjects above on the same
procedure, but with blocks instead of dolls. None of the groups showed any
differences between the two conditions with blocks. They argue that the
distributional explanation would predict that there should be a difference,
since the distributional analysis would precede the semantic one. The
cognitive explanation, however, predicts the results since the children
would not have had the same experience with naming blocks as they would
with dolls. Again, these conclusions should be viewed with some caution. A
proponent of the distributional explanation could argue that the child’s first
rule is semantically restricted at first, based on his linguistic experience with
naming practices. The importance of raising this issue here is that it is a
central one that will come up later. That is, we need to explain the
interaction between the child’s cognitive organization of the world, its
discovery of the semantic properties of individual languages, and the
distributional properties of grammatical categories of the language.
De Villiers & de Villiers (1973a) This study focusses on acquisition of

active and passive sentences by children at each of Brown’s five stages of
acquisition (see Table 3.4). They divided their subjects in Stage I into early
and late Stage I; thus, part of their results are relevant to the period we are
currently examining. Research into the comprehension of active and
passive sentences can provide us with information on children’s knowledge
of two aspects of English sentence structure: first, the ability to process
active sentences will indicate when children use word order to understand
sentences; second, the ability to process passive sentences tells us about
their knowledge of English verbal morphology.
In designing their study, the de Villiers had to face two problems in par-
ticular. One was how to get the child to respond to the test sentences. To do
this, they introduced each sentence with the command ‘Make’, for example
‘Make the dog bite the cat’. Despite using a sentence structure which was
relatively complex, the procedure appeared to work, except for the youngest
group. Another problem was how to get around the possibility that the chil-
dren might use semantic probabilities in responding to the sentences. For
example, the child could correctly act to the command ‘Make the cow eat the
flower’ because of its knowledge of ‘eat’, not because of word order - the
child knows that cows eat flowers but flowers do not eat cows. To avoid this
factor, they used what are called reversible sentences. These are sentences
where the subject and object are capable of being interchanged, for example
‘The boy hit the girl’ vs. ‘The girl hit the boy’. If children process such sen-
tences correctly, it is assumed it is because they have learned the English rule
that the unmarked word order in English is Agent, Action, Object.
The subjects included seven children in early Stage I (MLU 1.0-1.5),
with an age range of 19 to 24 months, and three in late Stage I (MLU
1.5-2.0), with an age range of 19.5 to 27 months. The latter small number is
unfortunate, in view of the importance of the results to be discussed. (7.5)
gives the basic set of sentences that was used:
(7.5) a. Make the dog bite the cat

b. Make the truck push the car
c. Make the mommy touch the daddy
d. Make the boat bump the train
e. Make the boy hit the girl
f. Make the horse kiss the cow
Table 7.8 Mean percentages of four kinds of responses for two test groups
on the understanding of active and passive sentences, taken from de
Villiers & de Villiers (1973~:Table 1).
early Stage I late Stage I

actives passives actives passives
Correct 31.0 28.6 75.5 28.9

Reversed 21.4 26.2 5.6 30.0
Child as Agent 33.3 30.9 12.2 28.9
Refusals 14.3 14.3 6.7 12.2
These are what we will refer to as the active sentences. A second set was
created from these, which we will call the reversed active sentences. These
are the same as those in (7.5) except that the subjects and objects are
reversed, e.g. ‘Make the cat bite the dog’. Two further sets were created,
the passives and reversed passives, which were the 12 active sentences made
into passives, e.g. ‘Make the dog be bitten by the cat’, and ‘Make the cat be
bitten by the dog’.
Each child was tested in two experimental sessions. In each session, the
child was shown 12 toys and asked to act upon them. Each session was
restricted to the testing of either active or passive sentences - pilot testing
indicated that mixing the sentences was difficult for the children. Half of the
children received the active sentences in the first session, and half received
the passives first. Six sentences were tested in each session, with the
presentation of reversed alternates being counterbalanced across children.
Table 7.8 presents the results in terms of the mean percentages of the four
different responses given by subjects in the two groups. ‘Correct’ indicates
that the child processed the sentence as an adult would; ‘Reversed’ indicates
that they reversed the subject and object, i.e. treated actives as passives and
vice versa. ‘Child as Agent’ is a response mode which was not anticipated.
Here the child would do the act itself on either the subject or object,
treating itself as the Agent of the sentence. ‘Refusals’, as expected, is when
the child did not respond to the test sentence.
The results indicate that the early Stage I children are understanding
neither the basic word order of English, nor the passive morphology. Given
the design of the study, we should restrict this to the claim that they could
not understand the constructions embedded in a sentence command begin-
ning with ‘Make’. It is still possible that they could process word order in a
simpler structure, such as ‘The cat is biting the dog’. The high number of
Child as Agent responses suggests that this structure might be having some
effect on their processing. Another possibility, however, and a very
reasonable one, is that this response is used when the child cannot process
the grammatical structure of the sentence. The 75.5 percent correct
responses of the three late Stage I children indicate that they have acquired
the English word order rule. This leads the authors to conclude that the
‘ability to use word order information in reversible active sentences first
appears in late Stage I’ (p. 338).
De Villiers & de Villiers do not comment on the responses of the subjects
in these groups on the passive sentences, except to say that there was little
evidence that they had acquired passives. It seems to us, however, that the
data in Table 7.8 indicate clearly that they are at least aware that there is
something about passives that is different from actives. The last three
measures, in fact, show evidence to this effect. Our interpretation is as
follows. If the children were treating the passives as actives, then the
percentage of reversals for passives should have been comparable to the
percentage of correct actives. This, however, was not the case, with only 30
percent of the passives being reversed. Instead, they just as often responded
to the passives with Child as Agent responses. This response occurred twice
as often for the passives as for the actives. We saw that this response type
occurred in the data for the early Stage I children as a strategy which was
used when they didn’t yet have the construction being tested. We take this
increased use of the Child as Agent response for passives as evidence that
they recognized these as a form they didn’t know. Lastly, there were twice
as many refusals for the passives as for the actives.
In sum, the data from de Villiers & de Villiers indicate that children are
receptively aware of the role of English word order by late Stage I. Further,
the data suggest that the same children distinguish passives from actives.
The children know that passives are a form that they have not yet acquired,
and respond differentially to it. They show three separate ways to respond
to such a dilemma: (i) treat passives as a form they have acquired, i.e.
actives; (ii) use a non-linguistic strategy such as Child as Agent, Le. when in
doubt, do the action yourself; or (iii) refuse to respond, i.e. ignore the
utterance.
The studies on comprehension which we have reviewed in the last two

chapters share a feature in common - they present the child with sentences
and then monitor the child’s response. The last study we will examine on
this topic uses a very different method, known as elicited imitation. It has
been observed in a range of studies that young children, when given a
sentence to imitate, often alter its form. In particular, they seem to change it
to conform to their grammatical system at the time. This has been referred
to as the ‘filter effect’, which claims that the child, when asked to imitate a
sentence beyond its current short-term memory ability, will encode it
through its own grammar. Like comprehension studies, this method has its
own problems - how do you get a child to imitate, and how can you be sure
that the sentence has been processed if it is correctly imitated? Because of
such problems, no doubt, there are relatively few studies which have used
this method. Those that have, however, have been highly suggestive. The
method has the virtue of providing us with information that is not available
from either comprehension studies or from the study of the child’s spon-
taneous utterances. This should be evident from our discussion of the next
study, which was one of the first to use this method.
Rodd & Braine (1970) This study examined three main areas of the child’s
grammar during the period of the first word combinations: (i) the gram-
matical morphemes which occur between verb+noun constructions; (ii) the
word order of subjects and predicates; and (iii) subject pronouns. We have
already looked at some information on the first two of these areas of
development, with Shipley, Smith & Gleitman’s research on the first, and
de Villiers & de Villiers’ research on the second. Here, we will address each
topic separately.
As stated, the method used was that of elicited imitation. The sessions
took place in the child’s home, where the experimenter and at least one
parent were present. The parents were instructed as to the nature of the
sentences to be tested, and asked to present them to the child, interspersed
with normal conversation. Rodd & Braine state (p. 432): ‘The model
offered for imitation was most often a question with a rising intonation,
intended to encourage relevant comment or assent from the child; such
comment often took the form of imitation.’ As might be expected, this
method takes time and a certain amount of patience. They found that the
children would imitate approximately one-third of the model sentences,
with variation between children in their interest to imitate. There were two
general kinds of data that were anticipated: they expected that the children
would selectively imitate, avoiding sentences that were not part of their
competence, and that they would show different kinds of imitations to
different constructions.
The first study was with a young boy, Owain, who was 23 months old.
Unfortunately, Rodd & Braine do not tell us much about his productive
language, except that he, as well as the other subjects, ‘had been producing
multi-morphemic utterances for at least a few weeks before the study began’
(p. 432). Owain was presented during two sessions with five kinds of
verb+noun constructions, which are shown in Table 7.9. Two of these,
verb+noun and verb+‘it’+noun, are ungrammatical phrases in English,
but occur in the speech of young children. They state, for example, that the
latter construction occasionally was used by Owain. Owain’s responses
Table 7.9 Percentages of occurrence of six kinds of intermediate elements
forfive constructions in the imitations of Owain, based upon Rodd &
Braine (1970: Table 1 )
Construction No. presented No. imitated Intermediate elementsD
0 [da] [a] [rn] [It] other
Verbnoun 74 21 38 38 19 - - 5
Verb ‘a’ noun 53 13 38 23 23 - - 16
Verb ‘the’ noun 64 26 - 65 11 - 8 16
Verb ‘it’ noun 45 14 7 57 14 - - 21
Verb ‘-ing’ noun 12 7 14 - 14 72 - -
a 0 indicates no intermediate element.
were placed into three categories: complete imitations, when the noun and
verb were both repeated with or without additions; partial imitations, with
either the noun or verb omitted; and no imitation. Only complete imitations
were used in the analysis.
Table 7.9 presents the results of this first study, given in the form of the
percentage of times each of six kinds of imitation took place. These six types
are defined by the element whch Owain inserted between the verb and
noun. The first finding was that Owain did not show selective imitation, as
expected. He imitated all models to about the same extent, which was
approximately 33 percent of the time. He did, however, show differential
patterns of imitation. Statistical analysis revealed the three patterns sum-
marized in (7.6):
(7.6) Pattern Dominant element inserted
1. verb ‘-ing’ noun [In1
2. verb ‘the’ noun [dal
verb ‘it’ noun
3. verbnoun @,[a],[dal
verb ‘a’ noun
That is, Owain treated the five constructions as if they were three, with
neither 0 and ‘a’, nor ‘the’ and ‘it’, distinguished from each other. This
supports Shipley, Smith & Gleitman’s finding that children in this period
have some grammatical awareness of the occurrence of morphemes
between verbs and nouns, as well as bearing out their caution that such
awareness does not necessarily mean adult grammatical knowledge. It also
suggests that ‘the’ rather than ‘a’ is the important article in the results of
Katz, Baker & Macnamara.
Table 7.10 Number of times that Owain and Carolyn maintained or
changed the word order of two kinds of constructions in their imitations,
based on Rodd & Braine (1970: Tables 3 and 4)
Subject and No. Response No. Order

construction presented type imitated same changed
Owain
NP + VP 185 substitution 10 10 0
complete 25 23 2
VP + NP 175 substitution 10 0 10
complete 25 2 19
Carolyn
NP + VP 153 substitution 11 11 0
complete 41 39 2
VP + NP 24 substitution 0 0 0
complete 14 3 11
The second study turned to the issue of the word order of intransitive
verbs and their subject noun phrases. Two types of constructions were used,
as exemplified in (7.7):
(7.7) 1. NPSVP e.g. (The) bird’s flying
Is the bird flying?
(The) bird flying?
2. VP+NP e.g. (He’s) flying, the bird
Is he flying, the bird?
Flying, the bird?
There were two children used, Owain and Carolyn. Owain was 24-25
months old for this study, and was tested over three sessions. Carolyn was
25 months old and tested in two sessions. The method was similar to that of
the previous study. The responses were categorized as in the first study,
except that substitutions were included in the analysis with complete
imitations. A substitution was when the child repeated the NP and VP but
replaced one or the other with a syntactically suitable substitute. The results
are given in Table 7.10, and are very straightforward. Both children
maintained the word order of NP+VP constructions, but changed it in the
VP+NP ones. The results support the findings of de Villiers & de Villiers
that children have acquired some knowledge of word order around 24
months of age.
The third study reports on the imitation of sentences with subject
pronouns. The test sentences were similar to the type 2 constructions in
(7.7), except that the following NP was missing, e.g. ‘He’s flying’. These
were elicited from Owain and Carolyn during the same sessions that were
used for the second study. They were also given to Owain again when he
was 28 months old. This later data is referred to as Owain2, and the earlier
data Owainl. There was also a third subject used in this study, Christine,
who was tested on the sentences in (7.7) and the subject pronoun sentences
during four sessions at 21-22 months of age.
The analysis of the data for the four sets of sessions (Carolyn, Christine,
Owainl, and Owain2) revealed three patterns of acquisition of the subject
pronouns. These are summarized below, with some relevant percentages
which were used to determine them.
Pattern I : no pronouns. Both pronoun+VP and VP test sentences
elicited most often VP responses, i.e. VPs without a subject (77% and
86% for Owainl, and 61% and 72% for both for Carolyn).
Pattern 2: optional subject prefixes. Pronouns occur with equal fre-
quency for both pronoun+VP and VP sentences (e.g. around 36%
and 27% respectively for Christine), with virtually no NP subjects for
either one.
Pattern 3: obligatory subject pronouns. Pronoun+ VP sentences elicit
a subject, either an NP or pronoun (33% and 38% respectively for
Owain2), and even VP sentences elicit either an NP or pronoun in
33% of the test sentences.
These patterns are presented in a way that suggests a developmental
sequence. The sessions for Owainl show little evidence of any knowledge of
subject pronouns (pattern 1). Christine seems to place the subject pronouns
on both VP and pronoun+VP constructions, suggesting to the authors the
followinp conclusion (p. 439): ‘A plausible interpretation of her total
response pattern is that Christine’s ‘pronouns’ are optional prefixes to
isolated VPs, without referential function.’ While quite an interesting
speculation, it needs to be taken cautiously for at least two reasons. First, if
this were true, one would expect uses of the pronoun in NP+VP test
sentences also, e.g. as ‘NP+pronoun+VP’, which apparently did not occur.
Second, there is no discussion of what would ever lead the child to this
grammar; the English child does not hear optional pronouns. This point is
one that we will return to later. As they stand, however, the data on
pattern 2 provide some evidence for English as being marked for the pro
drop parameter, a point discussed briefly in section 4.5. Pattern 3 is the
pattern expected for the English-learning child. Since we have little
information on Owain’s language at time 2, we don’t know when this change
took place. It seems that he is acquiring some knowledge of the status of
English subject pronouns around the end of this period.
7.3 The grammatical analysis 261
The discussion of pattern 2 by Rodd & Braine contains a particularly
useful exploration of its explanation. The authors propose pattern 2 by
assuming something very close to the Competence Assumption of
Chapter 4.They discuss the alternative that the child may have obligatory
subject pronouns at this time, but that there are performance factors which
inhibit their use. Two possible performance factors are considered. One is
that the child may be constrained to repeat after the adult verbatim, thereby
reflecting the nature of the stimulus sentences rather than its own grammar.
They point out that all the data suggest that the three subjects altered the
model sentences in systematic ways rather than repeating them as pre-
sented. A second factor is that the subjects may have been constrained by
the length of the model sentences. Upon analysis, however, they found no
effect of length of model sentences on the use of pronouns. After excluding
these possible performance factors, they support the analysis of data for the
four sets of sessions as given above.
7.3 The grammatical analysis of early word combinations

The literature on attempts to characterize the grammars of children during
this period is one of the largest, if not the largest, in the field of child
language. This is due probably both to a desire to understand the onset of
grammar and to the practical fact that the lists of syntactic types for this
period are somewhat less intimidating than later samples. Here we will
begin with some general issues in attempting to account for the child’s early
grammars, and then review the major studies.
7.3.1 General issues

The proposals that have been made about the young child’s early grammar
differ from one another on a number of dimensions. Most of these have
been mentioned at one point or another earlier in this volume, and this
section will outline several to be used in characterizing the various
approaches that have been attempted.
One aspect which sets off several approaches is the nature of the child’s
first grammatical categories. We have talked earlier about the young child’s
possible understanding of categories like Agent and Action. Such cate-
gories are part of most adult theories of grammar in some way, and are
referred to as semantic or thematic categories. We have also referred to
grammatical or syntactic categories like noun and verb. The question that
needs to be answered is when and how each of these categories emerges in
the child’s language. Some have argued that the child has syntactic
categories from the onset, and that the nature of these is highly constrained
by Universal Grammar. Others have argued that the child’s first categories
are semantic, and that these are used by the child to discover the syntactic
ones in the language. The interaction of syntactic and semantic categories,
then, is one aspect that needs to be part of any theory of grammatical
development.
Besides specifying the child’s categories as either semantic or syntactic,
child grammarians also differ in the degree of generality that is granted for
these categories. Categories can be claimed to be either specific or general.
To claim, for example, that a child has the category of Agent is to assign a
general semantic category to the child. An alternative is to say that the
child’s category is much more specific, for example, ‘people who play with
toys’. Or, that instead of Object, there are ‘things which can be eaten’. This
distinction can also be applied to syntactic categories. We can have, going
from most to least general, NP, noun, or nominals which occur only
sentence-initially or finally.
Another important distinction to be accounted for is that between lexical
and grammatical categories and their corresponding properties. Lexical
categories are those such as noun and verb, with their corresponding
semantic features. Grammatical categories are those such as number,
gender, tense, etc. which specify information about lexical categories.
There are two ways that the child could learn these two types of categories.
One way, which is commonly referred to as semantic bootstrapping, is for
the child to first determine the lexical categories, then use these to acquire
the grammatical ones. We saw an example of this when reviewing the paper
by Katz, Baker & Macnamara. Their proposal is that the child has the
cognitive distinction for the equivalent of common and proper nouns, and
then uses this to discover the way it is marked in English. The opposite view
is that of distributional learning, which is the claim that the child first notices
the regular occurrence of some morpheme, e.g. the presence or absence of
the article, and then seeks the corresponding semantic/syntactic properties.
Theories differ in the importance they assign to each of these properties
during early acquisition.
Child grammars also differ from one another concerning the amount of
productivity assigned to the child’s grammar. By productivity, I mean the
extent to which the grammar can generate sentences which have not been
attested for in the child’s spoken language. The grammar could have limited
productivity, in the sense that it mostly generates the sentences upon which
it was written. On the other hand, we could allow for extensive productivity
whereby the grammar generates many more sentences than have been
heard. This important point was emphasized in one of the first discussions of
grammatical development of the period of longitudinal studies, that of
Brown & Fraser (1964). At one point they wrote two grammars for 89
two-word utterances from Eve. One grammar only generated the 89
utterances recorded, while the other generated the original 89 utterances
plus 469 others. In their discussion of how to choose between such
grammars, they emphasize that the true test is the grammar’s ability to
make predictions. We will return later to the problem of testing predictions
of these kind.
The converse of productivity is a grammar’s observational adequacy. Here
I use this term to refer to a grammar’s ability to account for all of the attested
data. Some grammars may be lean in observational adequacy in the sense that
they only account for a subset of the data. Such grammars are proposed by
those who set strict criteria on what is a productive utterance in the child’s
language. These theories allow for nonproductive utterances for a variety of
reasons, e.g. because of imitation, memorization, emerging constructions
which are not yet productive, performance errors. Other grammars may be
rich in that they propose rules to account for all of the child’s utterances.
As with adult grammars, child grammars also differ in their complexity.
Some are written to be as simple as possible, with a small number of rules
and categories. Others tend to be complex, with a large number of rules and
categories. This aspect of grammars is not necessarily correlated with
productivity. Some simple grammars can be quite productive, while others
are not, depending on the kinds of categories used in the grammar.
In Chapter 4 we discussed at some length the issue of continuity vs.
discontinuity in development. One’s view on this affects one’s grammar for
the young child in two further ways. If one allows for a discontinuity in
development, this often leads to the postulation of non-adult categories in
the child’s language. This is in contrast to the view that the child’s categories
are constrained by Universal Grammar to be adult-like, or universal.
Within these distinctions, there is the further question of individual
variation. That is, to what extent can any one child’s grammar be different
from that of any other child? That is, we can propose individual grammars,
or non-individual or common grammars.
A last feature of the grammars to be discussed is their degree of formality.
There was a time in the early 1970s when it was common to see formal
grammars written for language samples. This was followed by a move away
from formalism toward taxonomic descriptions of the child data. Such
descriptions often take the form of lists of common constructions found in
the data. Recent signs indicate a move back toward more formal accounts.
7.3.2 Pivot grammar: Braine (1963a)

The first attempt to capture the young child’s grammar which gained a wide
appeal was that of Braine (1963a). We have already discussed Braine’s work
264 T H E PERIOD OF T H E FIRST WORD COMBINATIONS
in the overview in Chapter 2. To summarize, he asked mothers to keep
parental diaries of the first word combinations of Andrew, Gregory, and
Steven over a four- to five-month period. Braine’s paper gives most of the
data for the three children, which we summarize in Table 7.2. Table 7.4
gives Braine’s organization of the data from Andrew according to its
grammatical classification. By our definition, the children are in early
Stage I.
Pivot grammar is a theory that assigns a great deal of importance to
distributional learning. The child begins acquisition by selecting out a
small group of words for grammatical acquisition. Presumably from its
receptive knowledge, the child recognizes that certain words are quite
frequent in the input language. These words also are consistent in their
position, usually occurring before or after a wide range of other words.
Braine does not specify how the child selects these words, but the proso-
dic system of the language and the language of the parents would have to
play some role. For example, the child might hear ‘Hi’ stressed before a
range of words and notice its frequency and constant position. Or, the
parents might have some verbal games such as ‘see’+‘X, where X might
be a series of objects being pointed out. This small class of words
identified by the child is called the pivot class. All other words are defined
negatively in relation to the pivot class, and are thus simply the other
class, or what was called the open class. Since each child’s linguistic
experience will differ from that of other children, each child’s pivot class
will also differ.
The pivot class is then used to acquire other classes. When only two-word
combinations are being produced, there are only two kinds of possible
pivots, sentence-initial pivots or sentence-final pivots. A pivot class word is
always restricted by its position. If it is a sentence-initial pivot, it must
always occur in only that position. The child begins by learning which words
can occur with each pivot word. At the onset, the child might simply put
potentially any open class word with any pivot word. After a while,
however, he will recognize that certain open words only occur with certain
pivots. Gregory, for example, used distinct open words after initial, and
before final, pivots. Primarily English nouns and adjectives appeared after
initial pivots, while English verbs appeared before final pivots. Braine used
these observations to suggest that Gregory was beginning to acquire the
substantive-verb distinction in English. Since pivots are being used to
acquire word classes, they serve no purpose in occurring either alone or with
each other, so both of these distributions are disallowed by the theory.
Further, combinations of open+open class words are predicted to occur
only after the open class words have begun to differentiate into finer classes.
Table 7.11 The defining features of a pivot grammar
~
Feature Pivot class Open class
size small class large class

frequency frequent infrequent
position restricted not restricted
co-occurrence cannot co-occur may co-occur
isolation do not occur alone may occur alone
The possible combinations in a pivot grammar are summarized below.

Nonpermitted combinations are marked with an asterisk (*):
Possible word combinations of a pivot grammar
(a) P l + X (where P1 = initial pivot and X = open class words)
(b) X+P2 (where P2 = final pivot)
(c) X+X (after types a and b have occurred for some time)
(d) *P+P
(e) *p
To summarize, there are five criteria which define the pivot and open
classes; these are given in Table 7.11. Unfortunately, Braine does not
provide any procedures for determining them. For example, how frequent
must a word be before it is a pivot? Or, which of these is more important
than the others? To learn how to write a pivot grammar requires looking at
Braine’s grammars and learning from example. As will be seen, this lack of
precision has led to criticisms on the theory on the grounds of internal
inconsistencies, since some of Braine’s proposed pivots seem to violate one
or more of these criteria. Braine does, however, provide a distributional
criterion for defining a word:
Those segments are considered ‘words’ which are the longest seg-
ments that cannot be divided into two or more parts of which both are
English morphemes that occur in the corpus independently of the
others. Thus ‘ice cream’ and ‘all gone’ are each classified as one word
in Gregory’s speech, since neither ‘ice’ nor ‘cream’, nor ‘all’ nor
‘gone’, occur in other contexts or alone. However, for Andrew
‘allgone’ is classified as a combination of two words since ‘gone’ occurs
by itself, and ‘all’ occurs independently in ‘all wet’, ‘all dressed’, etc.
(Braine 1963a:3)
This criterion is important as one of the first attempts to define productivity.
We will discuss other ways to do this in section 7.5.
We can now look at the pivot grammar for Andrew, based on the data in
Table 7.4. Example (7.8) presents Andrew’s pivot grammar:
(7.8) S + Pl+X
S+ X+P2
P1+ ‘all’, ‘I’, ‘no’, ‘see’, ‘more’, ‘hi’, ‘other’
P2 + ‘off’, ‘by’, ‘come’, ‘there’
We can see that this is a very simple grammar, with a small number of rules
and only three classes of words. Andrew has a long way to go to get from
this grammar to that of adult English. There are also 29 of Andrew’s
utterances which are placed into the ‘other utterances’ category. Braine
does not address this problem, but it is one that needs explanation. It is
possible that these utterances are indicative of future advances, or the result
of performance factors. Any theory will need to account for the child’s
productions which are outside the predictions of the proposed grammar.
Braine’s pivot grammar is quite ‘lean’ in the sense used in the previous
section.
Braine’s analysis also gives some insights into the relative importance of
the defining characteristics of the pivot class. Two pivot words, ‘by’ and
‘come’, only occur twice, and three others, ‘I,, ‘see’ and ‘hi’, only occur three
times; frequency does not appear to be as important as positional consist-
ency. The data also show some internal inconsistencies in the analysis. One
sentence, ‘I see’, is a combination of two pivot words, which is a violation of
the theory. Also, the distribution of ‘byebye’ is peculiar. This is a frequent
word which is a likely candidate as a pivot word, yet it appears to occur in
free variation with open class words. Even ‘off’, a final pivot, can be found
in the list of ‘other utterances’ in initial position.
Assessment. At first glance, a pivot grammar has a certain appeal due to
its simplicity and apparent success at characterizing the child’s early
utterances. In the years after its first appearance, however, several investi-
gators began to challenge its validity. Substantial criticisms appeared in
Bloom (1971)’ Bowerman (1973a), and Brown (1973). By the mid 1970s
pivot grammar was generally considered to be discredited. Here we will
briefly review the arguments against it, and then provide a defence on its
behalf.
Brown (1973:97-111) presents three arguments against pivot grammar as
an adequate account of the child’s language at the onset of syntactic
acquisition. The first argument is that it doesn’t even account for Braine’s
original data. We have already commented on the internal inconsistencies
in the data for Andrew. The second argument is against its universality.
Both Bloom (1971) and Bowerman (1973a) found that it was too lean to
capture the grammars of the children they studied. Brown concludes that
the only data that show some support for it are those from children at the
very onset of acquisition. The third argument was on its grammatical 8
adequacy. As we will shortly see, Bloom (1970) pointed out semantic
regularities in early speech that are missed in the pivot account. In
particular, she emphasized the existence of ambiguous sentences like
‘mommysock’ which occurred in different contexts with different meanings
-one as ‘mommy is putting on your sock’, and the other as ‘mommy’ssock’.
Investigators have generally been convinced by these arguments but
some reservations are necessary. First, it is not clear that the theory itself
could not be revised to deal with some of the problems with the internal
consistencies in the original analyses as well as with its application to other
data. This would need to take the form of better criteria for determining the
defining features and their interaction. Also, it needs to be emphasized that
the theory was only designed to apply to the earliest sentences of young
children. Much of the data shown to be inadequately described by pivot
grammars are actually from children who are relatively advanced. The
‘mommy sock‘ example, for instance, comes from Kathryn I who, by our
definition, was at the end of late Stage I with 226 unique syntactic types in
her sample. The data that seem most like pivot grammars, e.g. those of
Eric 11, tend to be at much earlier points of development.
The strength of the pivot approach is that it offers an account of certain
distributional facts about the child’s early word combinations. It also offers
some idea, albeit imprecise, on how the child might achieve such a grammar
from the adult input. The suggestion that the child looks for positional
consistency in frequent words as a basis for building word classes and
presumably rules of word combination is a possible initial learning heuristic.
To make it work, however, requires additional specification about how the
classes and rules get established. In particular, it needs to include the role of
semantic information, and how the child gets from this primitive system to
the adult grammar. Braine (1976), in fact, attempts some revisions along
these lines, and we will look at these later in this chapter.
7.3.3 Transformational grammar: the Standard Theory

Around the time of Braine’s article on pivot grammar, other distributional
analyses were also being done (Brown & Fraser 1964; Miller & Ervin 1964).
These analyses were typical of child language research as presented in
Chapter 2. The focus was on the child data and there was little reference to
any theories of the adult language. In the 1960s, however, Chomsky’s work
in syntactic theory (Chomsky 1957, 1965) began to influence not only
grammatical theory, but also work in child language. This model was seen as
capable of providing a powerful means of describing and explaining
children’s early grammatical development. Grammatical research in the
period from 1965 to 1975 was dominated by the theory of syntax presented
in Chomsky (1965), which has since been referred to as the ‘Standard
Theory’.
In examining this research, it is important to realize that the study of
syntactic development was just beginning. The extensive child language
literature to that point had concentrated on data collection, and lexical,
morphological and phonological development. There were few data avail-
able on the child’s early word combinations, much less theories to account
for them. It was with some excitement that investigators brought the
descriptively powerful theory of transformational grammar to the study of
children’s language. It was applied in two distinct ways. On the one hand,
the theory was taken as the basis for an explanation of acquisition. On the
other, it was used as a descriptive device, a method to use in characterizing
the speech of young children.
McNeill (1970b) The first to use transformational grammar as an expla-

nation of language acquisition was McNeill (1966a). This paper was a
lengthy attempt to provide a nativist account of syntactic development,
using the Standard Theory. McNeill’s writings from this perspective include
McNeill (1966b), Miller & McNeill (1969), McNeill (1970a,b, 1971). As
expected, this view was striking in its contrast to the behaviorist orienta-
tion of Braine’s pivot grammar, and had quite an impact at the time. Most
of the discussion here, and all references are taken from McNeill (1970b).
According to McNeill, the child has genetically available at the onset of
language acquisition a highly constrained universal grammar. The foun-
I
dation of this grammar is the basic set of grammatical relations that
underlie all languages. The major ones are shown, with examples, in (7.9).
The first feature specifies its grammatical category, and the second its
subcategorization. For example, the relation ‘predicate’ is represented by a
VP, and it occurs after an NP which is its subject. The brackets in the
examples set off the domain of each relation, since each one enters into a
relation with one of the others. The constituent which represents each
relation is italicized in the examples. e
(7.9) Relation Feature Example

predicate [+VP, +NP-] [The dog ate the apple]
subject [+NP, +-VP] [The dog ate the apple] ,
main verb [+V, + N P ] The dog [ate the apple]
object [+NP, +V-] The dog [ate the apple]
modifier [+Det, +N][The dog] ate [the apple]
head [+N, +De+__] [The dog] ate [the apple]
The child’s task is to identify these relations in the language being
acquired. The child’s early grammar does not manifest all of them simul-
I
taneously; rather, they appear in a predictable order. McNeill discusses this
point in regard to Brown’s subject Adam. Adam’s sample included the fol-
lowing relations, with examples: predicate ‘change diaper’, direct object
‘see truck’, indirect object ‘write paper’, modifier ‘dirty paper’. McNeill
(p. 1090) makes the following comment:
It is impossible to say from Adam’s evidence whether or not these
relations had equal tenure in his grammar at 28 months. All four con-
ceivably existed at the holophrastic stage. But it is equally possible
that originally Adam’s utterances expressed only predication, to
which was first added modification (including possessives), then direct
objects of verbs, then subjects, and finally indirect objects of verbs -
this being the order of the frequency of these relations in Adam’s
speech at 28 months.
McNeill thus acknowledges that there is possibly an order of appearance of
these over the period of the first word combinations.
Importantly, McNeill accepts the superficial features of the early pivot
grammar. The child begins this period with a grammar which looks much
like the pivot grammars just discussed, but has, by its end, a basic phrase
structure grammar with the universal grammatical relations. The initial
pivot class, however, is not a random collection constrained by their fre-
quency and positional constancy. Rather, it is constrained by the hier-
archical nature of the universal grammatical relations. McNeill wants to
propose that the relations predicate, main verb, and modifier are grouped
together at a higher level of organization. In particular, they share the
feature of occurring with NP. The pivot class, therefore, is initially a class of
words that enter into relations with nominals.
The child begins grammatical acquisition by assigning relations to the lan-
guage it hears, presumably beginning in the holophrastic period. McNeill
states (p. 1096): ‘It is important to note that these features are automatic-
ally made available whenever a child obtains any meaning from adult
speech.’ To make these distinctions, it seems that the child would need to
use semantics. McNeill seems to imply this, since he refers to them as
‘logical relations’. Our discussion from the previous chapter indicates that
children have a range of semantic categories at this time. The universal
grammatical relations can be interpreted as constraints on how these sem-
antic categories can be related in speech. Table 6.8 gave the semantic cate-
gories of specific and general nominals, action words, modifiers, and
personal-social words. We can rephrase McNeill’s claim as something like
this: the most general grammatical relation is between nominals and the
other semantic classes, which are initially organized into a general class of
general predicates. Assign all members of the first class some feature, say
[+N], and all those of the latter class another feature, let’s say [+VI.
These initial semantically motivated distinctions immediately lead to
syntactic categories. The child assigns to each word it acquires the relevant
grammatical features from (7.9). Andrew’s pivot class, for example, has
two subclasses - one with the features [+V + N P ] , the other with the
features [+V, +N-]. Universal Grammar will require a set of
principles that specifies the phrasal structure of a sentence. It will require
that any constituent which has a lexical feature like [+N] or [+VI is also part
of a phrase which is an NP or VP. Such speculations are part of current
theory and are referred to as X-bar syntax (Jackendoff 1977).
Further development during the period leads to the emergence of the other
grammatical relations besides the initial one distinguishing general predi-
cates from nominals. For example, general predication will begin to divide
I
into other relations such as main verb by similar semantic guidelines. The
child’s acquisition device will know that objects affected by actions are [ +N]
categories which also bear the grammatical relation of object to the action. I
Words so identified are then assigned the features [+NP, +V-].

The assignment of grammatical features to the individual lexical entries
of words is an important part of McNeill’s theory. This assignment occurs in
a systematic fashion. At first, each word ‘has one and only one classification’
(p. 1096). A pivot class word, for example, cannot belong to both the P1
and P2 classes simultaneously. Another example comes from Brown’s data.
Two of his subjects used only animate nouns as subjects, and inanimate
nouns as objects. McNeill assigns these two classes the features [+NP,
+-VP, +animate] and [+NP, +V- , +inanimate] respectively:
‘Eventually all words are classified in several ways, thus enlarging the
distributional range of each word’ (p. 1096).
This picture of development leads the child to a grammar which at an
early age is very adult-like regarding its representation of grammatical
relations. It also claims that the grammars of young children should look
very similar from virtually the onset of acquisition. McNeill, in fact, makes
this very claim (p. 1099): ‘Insofar as the basic grammatical relations reflect
the innate abilities of children, the type of grammar just outlined will be
x
developed regardless of the language to which a child is exposed. It is a
universal child grammar. . .’
The grammar which McNeill wrote for Adam’s language sample at 28
months is given in (7.10). Adam has an MLU of just under 2.0 at this time so
is just at the end of the period which we are discussing. The ‘P’ here stands
for a modifier:
(7.10) rule 1 C (NP) (VP) ,

Table 7.12 Information of the two-word utterances expressing grammatical
relations for Adam at 28 months of age, taken from McNeill(l970b; Table
3)
Pattern Frequency Corresponding grammatical relations
P+N 23 modifier, head noun

N+N 115 modifier, head noun; subject, predicate
V+N 162 main verb, object
N+V 49 subject, predicate
This grammar has features in common with other grammars written for this
period of development. First, there are a number of optional elements,
since the child’s actual productions are much shorter than the longest
utterance which this grammar can generate. Adam, for example, produced
349 two-word utterances at this time, but only 49 three-word utterances.
Table 7.12 gives information on the grammatical relations which were
attested in Adam’s sample. The grammar is very powerful, therefore, in the
sense that it predicts a lot of utterances which were not attested. Second,
there are no grammatical morphemes, such as prepositions or articles,
represented; the theory implies, in fact, that these will be acquired after the
basic grammatical relations.
Assessment. McNeill’s theory is the first extensive theory on syntactic
acquisition to be proposed. It relies on the child’s innate ability to have and
determine the basic grammatical relations in language by the end of the
period of the first word combinations. It has a developmental component to
it in that it allows for the relations to emerge gradually over several months.
The theory also claims that all children will form the same universal
grammar at this point. It is less clear whether children will necessarily also
show the same order of emergence, but it implies that they will.
The theory has been attacked as assigning too much structure to the child.
If true, it would be in violation of the Competence Assumption. If we look
at the grammar proposed for the end of this period in (7.10) however, we
are hardly looking at the adult grammar of English. We would argue that
McNeill’s proposals are not nearly as rich in their assignment of structure to
the child as has been claimed by some critics. Also, this grammar is not the
one which he would write for early Stage I. He seems willing to accept the
surface patterns that Braine observed for the first multiword utterances.
The difference in Braine’s and McNeill’s grammars for the early data is not
in the number of categories and rules, but in their nature: McNeill wants to
constrain them in ways that make them closer to the adult ones which will
occur later.
We already have some evidence for syntactic processing from our review
of comprehension studies. The arguments for syntactic relations will need
to fall in large part on internal arguments of McNeill’s theory. That is, the
theory accounts for the emergence of syntax, and in part for the develop-
ment of an adult grammar in a way that pivot grammar does not. Later we
will look at a more recent theory by Pinker (1984) which is similar to
McNeill’s in many ways, but which addresses many of the issues left open.
The use of the Standard Theory as a descriptive tool was probably first
attempted by Menyuk, who wrote transformational grammars for a large
sample of children. The results of these analyses appeared in a series of
articles which are pulled together in Menyuk (1969), which provides a useful
introduction to the kinds of errors and omissions children show across
several years in acquiring the structure of English. The samples from
individual children, however, were relatively small, and no individual
grammars are presented. The first in-depth attempt to write individual
grammars was that of Bloom (1970).
Bloom (1970) The extensive data collected by Bloom have already been
discussed in a general way in Chapter 2 and again at the beginning of this
chapter. Bloom was interested in providing detailed analyses of the early
word combinations from the three children she studied. To do this, she used
the Standard Theory as a method to describe the child’s data. Her concern
was more with describing the children’s grammars, however, than testing
the predictions about language acquisition that follow from trans-
formational grammar. As a result, she changed the formal features of the
theory when necessary to enable it to account for the children’s patterns of
acquisition. The result was a set of grammars for the three subjects which
looked different from one another, and unlike that for adult English. 1
Superficially, Bloom’s grammars appear to fall somewhere between those
of Braine and McNeill. Those on data from early Stage I bear resemblances
to the pivot grammars of Braine, while those from late Stage I data look
more like McNeill’s grammar for Adam. Bloom’s early Stage I grammars,
however, assign more structure than that of a pivot grammar, while those
for late Stage I are less adult-like than McNeill’s, and require more
restructuring to reach the adult model. We will argue that Bloom’s
grammars are more similar to those of Braine and McNeill than is normally
credited.
Before looking at a sample of these grammars, however, we need to
examine Bloom’s procedure for selecting data for analysis. She was very
concerned about keeping close to the data and not assigning structures to
children on the basis of one or two examples. All sentences were assigned to
one of three categories.
Generally, a structure was unique if it occurred only once in the corpus
of a speech sample, marginal if it occurred fewer than five times with
different formatives in different situations, and productive if it occur-
red five times or more. (Bloom 1970: 17)
The grammars presented usually excluded unique and marginal structures,
although these were often referred to in the discussion.
We begin by looking at Bloom’s grammar for Gia I. This sample, collected
when Gia was 19 months, 2 weeks old, yielded 1,015 utterances, with an
MLU of 1.12. Bloom, Lightbown & Hood (1975) give 55 syntactic types for
Gia I, probably not counting the constructions with [a] to be discussed.
Gia I appears to be in early Stage I, at a point comparable to Braine’s
subjects. One would expect, therefore, that Gia’s grammar would be a pivot
grammar.
Bloom considers two phrase structure grammars for Gia I both of which
are given in (7.11). Examples of the constructions for the first grammar are
listed in Table 7.13 (Bloom uses VB for V, so I follow that practice here).
(7.11) Grammar I Grammar 2
-{;} } : {
P- ‘more’, “nother’,
Q--,‘more’, ‘another’ ‘hi’, [a]
A large number of recognizable words occurred with a preceding [a]. This

was a phonetically consistent form which had an indeterminate interpreta-
tion. Braine, for example, would have excluded these from his analysis.
Two other words, ‘more’ and “nother’, were used in the sense of recur-
rence. It was because of this consistent semantic usage that Bloom groups
these under the category Q. It is also productive by Bloom’s criterion since
it occurred at least five times. ‘Hi’ occurred in the context of Gia’s looking
out the window and waving. Bloom acknowledges the positional consist-
ency of these four forms and groups them as an initial pivot class in her
alternative grammar, shown above as grammar 2.
Bloom is more reluctant, however, to write a totally pivot grammar and
relegate the rest of the data to the category of ‘other utterances’. She feels
that the N+N and N+VB sentences suggest emerging semantic relations.
Since these are not frequent, she is cautious in concluding that they exist at
this time. The N+N constructions suggest a ‘locative’ relation, and also
either a ‘subject-object’ or ‘genitive’ one, as in the sentence ‘Gia eyes’, said
while reaching for her doll named ‘Blueyes’. The even less frequent N+VB
Table 7.13 Gia 1’s word combinations
[a1 + N Q + N o r VB N+N N+VB

29 combinations ‘more rabbit’ conjunction N subject
e.g. ‘[a] pen’ ‘more clown’ ‘wowwow car”’ ‘girl write’
‘[a] car’ ‘more write’“ ‘Mommy back‘d
‘more byebye’ subject-object ‘out Daddy’
[a] + VB “nother Ibzebal’ ‘Gia eyes’ ‘go Mommy’
6 combinations “nother bang’ ‘Gia Blueyes’
e.g. ‘[a] ride’ ‘lamb ear’ N object
‘[a] write’ ‘girl fish’ ‘slide go’
‘truck wheel’ ‘balloon throw’
‘Hi’ + N ‘hi’ + VB ‘girl ball’ ‘block ’way’
I O combinations none attested
e.g. ‘Hi Jocelyn’ locative
‘Hi Eric’ ‘block bag’
‘fly block’
‘fly blanket”
’Used in two contexts, one when she wanted to scribble again after her book was taken away,
the other when she was scribbling after she had scribbled before.
Used while looking at a picture of a dog running alongside a boy in a toy car.
Used twice, once when fly landed on her blanket, the other when it did so again.
Used twice, once waiting for mother to come back, the other as mother came through the
door.
sentences show emerging evidence for both ‘subject+verb’ and ‘verb+ob-

ject’ relations.
When we compare these data and their analysis to those for Andrew, we
see that there are noticeable similarities. Andrew’s sample also shows
evidence for emerging relations. For example, the pivots ‘more’ and ‘other’
could be grouped as a Q marking recurrence. The final pivots ‘off’, ‘by’ and
‘come’ could be classed as verbs in N+VB constructions. The issue becomes
not so much a difference in the data as in one’s emphasis on the role of
context in interpreting the child’s sentences. Both sets of data show the use
of a small class of words in positionally consistent ways, as well as some
evidence of consistent semantic patterns. f
The advances which are made over this period can be seen by looking at
Bloom’s grammar for Kathryn I. This sample yielded 1,225 utterances over
seven-and-a-half hours. Bloom states (p. 40) that there were 397 multi-
morphemic utterances, presumably tokens, and that the MLU was 1.32.
Bloom, Lightbown & Hood (1975) identify 226 syntactic types in five-and-a-
half hours of the sample. While the MLU suggests that she is in early
Stage I, the high number of syntactic types places her at the end of late
Stage I, and shows how MLU can be misleading as a measure of syntactic
development. The phrase structure rules for Kathryn I are presented in
(7.12):
(7.12)
1. SI-
2. s2+ Pivot+N
3. VP-
5 . Nom-
( :em
Ng -
Pivot -+ ‘hi’, ‘oh’, ‘OK, ‘thank you’
‘no’
The grammar for Kathryn I is actually relatively similar to McNeill’s

grammar for Adam, shown in (7.10). One difference is that there is a rule here
to generate pivot-like constructions. A look at the proposed pivots,
however, reveals that these are items that even occur in a pivot-like manner
in adult English. An example of this is ‘hi’, although Bloom mentions that it
was not used as a greeting, but more in the sense of ‘taking notice’ of
something (p. 41). It is possible that Adam produced constructions like
this, but that they were considered peripheral to the grammar. The
grammar also includes a phonetically consistent form [a] whose distribution
is more restricted than the one in Gia I. Here it only appears in NPs. There
were 36 constructions of the form [a]+N, and 13 of the form [a]+ADJ+N.
The [a] in this context looks like an emerging article which cannot be
specified as either ‘the’ or ‘a’ according to its pronunciation. The elicited
imitation technique used by Katz, Baker & Macnamara might have helped
to clarify the status of this element. Bloom also has an optional negative
morpheme, Ng for ‘no’, and a verb particle shown as Part, for sentences
such as ‘throw away’. Neither of these is mentioned in McNeill’s grammar.
There are two noticeable formal differences between this grammar and
the one for Adam. One is that, McNeill has restricted his categories to those
which appear in the adult grammar of English. This followed from his
theoretical position that the child’s universal grammar restricts the range of
possible categories from the onset of acquisition. Bloom, however, has
created the category Nom which includes N and Dem, but has done so in an
attempt to limit the generative power of the grammar. Nom is equivalent to
the subject NP of McNeill’s rule of C (NP)(VP). The problem with
having NP in this position is that it generates ADJ+N constructions as
subject NPs which are not attested in the data. As has been noted by others,
e.g. Ingram (1972), adjective phrases such as ‘more cereal’ first occur in
isolation, then inside of isolated VPs, e.g. ‘eat more cereal’, and only later
as subjects. Bloom’s Nom category avoids this problem, but creates the one
of accounting for (i) what allows the child to form such a category, and
(ii) how the child gets from it to the adult one. For purposes of discussion,
we will avoid this difficulty by revising Bloom’s first rule to read:
The second formal difference is the NP in Bloom’s first rule for Kathryn I.
This is needed to generate possessive constructions, e.g. ‘mommy sock’
meaning ‘mommy’s sock’, as will be shown below, and subject demonstra-
tives such as ‘this book’ meaning ‘this is a book’. McNeill captures these by
generating two Ns inside of the NP for possessives, and using P+N for
demonstratives, where P is the category symbol for modifiers. He points
out, however, that the latter construction may not be adequate if children
use demonstratives in two meanings, one as a subject of a predicate
nominative, the other as a modifier of a noun. Again, Bloom wishes to
avoid writing a grammar that is too powerful.
Despite being generatively weaker than McNeill’s grammar for Adam,
Bloom’s grammar for Kathryn generates basically the same grammatical
relations as does McNeill’s. For example, there were several cases of the
subject-predicate relation as in ‘mommy push’. Bloom discusses at length
(pp. 62-7) the inadequacy of a simple rule such as S -+ N+N to account
for the child’s early N+N constructions. She argues that there were five
distinct kinds of N+N sentences, based on the context of their use and their
consistent word order. They are each given a distinct underlying structure,
presented here in Figure 7.1. The 0 indicates underlying constituents which
do not appear in the surface structure of the sentence. Two of these
relations, conjunction and locative, are not actually generated by the rules
in (7.12) because they did not meet the productivity criterion of occurring at
least five times.
McNeill’s arguments for the importance of grammatical relations were
based primarily on logical grounds, that is, the innate nature of grammatical
relations provides the child with a starting point for acquisition of the adult
grammar. Bloom, however, was mostly concerned with proving the
existence of these relations on the basis of the child’s data. They arrived at
the same conclusions via deductive and inductive methods respectively, as
these terms were used in Chapter 4.
Bloom (pp. 50-62) describes five tests for proving the existence of these
relations in the child’s language. Each of these tests is summarized below; as
will be indicated, some proved to be better than others.
1. Sentence patterning. If a construction occurs in the same place as its
head without a change of meaning, then there is evidence of a constituent.
For example, if both ‘sock’ and ‘mommy sock’ occur as an object of the
same verb with the same meaning, then they are the same constituent, e.g.
conjunction (2) attribution (6) suhject - object (18)
A A A
N
I
umbrella
N
I
boot
I A
N
0 ADJ N
N
A NP
I I I I
party hat N
I O - I
I
Mommy sock
genitive (7) locative' (2) locative' ( I )
N NP
I
N
I I
N-
2
I N' l 0
Kathryn sock
0 v
I
PrepP
I
@v N NP
I
Wendy
N
I
e levator sweater
7
chair
Figure 7.1 The underlying structure of five kinds of noun+noun construction for
Kathryn I, along with their frequency (in parentheses), adapted from Bloom
(1970).
'give sock', 'give mommy sock'. This argument was also used by McNeill
(1970b:1079), but was not very helpful for Kathryn's data because most of
her multi-morphemic utterances were restricted to two words.
2. Linear order. If a construction shows the order of the adult construc-
tion, then it has the same grammatical relation. This proves to be a weak
test because some relations have the same order, e.g. subject-object and
genitive, and conjunctions in random order, yet these could be mis-
interpreted as another relation. Another problem with this test, not
mentioned by Bloom, is that adults tend to interpret child speech with the
meanings they would use as adults (Howe 1976).
3. Replacement sequences. If the child produces two or more sentences
in a row in the same context, and one is an expansion of the other, then the
shorter sentence has the same relation as the longer one. An example is the
sequence ‘mommy milk’, ‘mommy’s milk’, where the expansion indicates a
genitive relation. This test was first proposed by Braine (1973). These
constructions were not, however, very frequent in the data.
4. Replacement and deletion. This is a variation of the previous test
where expansion and simultaneous deletion indicates the relation. An
example is ‘Baby milk’, ‘touch milk’ where the second sentence suggests
that the first one is subject-object. This proved to be the most valuable
linguistic test.
5 . Non-linguistic context. The use of the non-linguistic context to infer
the child’s intended meaning. This includes the child’s overt behavior and
aspects of the environment at the time of the utterance.
While Bloom’s grammar is more restricted than McNeill’s, it still gener-
ates structures where constituents are proposed which do not occur in the
surface structure of the sentence. This is evident in the cases where 0 occurs
in Table 7.14. While there was evidence that three or more categories were
underlying a sentence, most showed only two at one time. Only four actual
three-term utterances occurred in the data: ‘me show Mommy’, “chine
make noise’, ‘man ride [a] bus’, and ‘I comb pigtail’. These sentences and
the patterns in two-word sentences led Bloom to conclude that Kathryn
had an underlying subject-verb-object order.
The problem arose, then, of how to account for this discrepancy between
the surface and underlying order. McNeill accounted for this by generating
optional subjects and predicates. Bloom does not do this because she wants
to represent the missing elements in the underlying structure of the
sentence. As the grammar in (7.12) currently operates, however, it will
generate surface strings with more categories than are appearing in Kath-
ryn’s language. Bloom adds an additional rule, called the reduction trans-
formation, which deletes one of the categories when three are generated by
the grammar. The selection of the category to be deleted is random, thereby
allowing for the various two-word combinations which appeared.
As pointed out in several places (e.g. Wall 1972; Brown 1973: 234-9), this
rule has several questionable characteristics; here we will briefly mention
two of them. One problem is that it is inconsistent with the formal
properties of transformations. Deletion transformations, for example,
apply in cases where information about the constituent deleted can be
‘recovered’ through other constituents in the surface structure. For
example, subject pronouns in languages like Spanish and Italian can be
deleted because information about the person of the sentence can be found
in the agreement affixes on the verb. No such recoverability, however,
operates in the case of the reduction transformation. Another problem is
that the rule violates the Constructionist Assumption. It is a rule that is
unique for this period, and which will be lost in later acquisition. Brown
(1973: 239) points out the peculiarity of having development proceed ‘from
the more complex to the less complex’. The rule’s real effect is not to map
from deep to surface structure, but rather from surface structure to speech.
That is, it is really a description of a performance factor which limits the
number of constituents that the child can say at one time. As a performance
factor, its eventual loss also becomes easier to understand.
Assessment. Bloom’s grammars are an important empirical attempt to
verify the existence of basic grammatical relations underlying children’s
early word combinations. They show that these relations are difficult to
establish in the earliest data, but that they exist at a more advanced level
such as that of Kathryn I. Importantly, this finding is based on evidence
found in the child’s language, not just of logical arguments following from
the theory of transformational grammar.
These grammars are also a serious attempt to restrict the generative
power of grammars for children in this period. While they go beyond the
restrictiveness of pivot grammars, they are less powerful than the grammars
of McNeill. As such, they are the first extensive attempt to implement the
Competence Assumption in writing formal grammars for children. Like-
wise, postulating the reduction transformation is a pioneering attempt to
provide some justification for the discrepancy between the child’$proposed
underlying grammar and its surface manifestation. The fact that the formal
properties of these grammars may be criticized should not take away from
the importance of their contribution.
7.3.4 Semantically oriented approaches

Establishing consistent relational meanings in the first word combinations
led to research exploring the relations in detail. In Child Language, at least,
the 1970sbecame absorbed with this question. A major impetus to this work
was the development of generative semantics in linguistics (see Newmeyer
1983). Generative semantics, in its most general form, was an attempt to
redefine the underlying relations in grammatical theory as semantic primes
such as Agent and Patient.
One of the first expressions of this approach was the case theory
expressed in Fillmore (1968). Fillmore noted that the notion ‘logical
subject’ covers a variety of semantic relations. Take, for example, the
sentences in (7.13) where the subject of each sentence is italicized.
(7.13) a. John opened the door

b. The door opened
c. The key opened the door
The semantic role of the subject of each of these sentences differs. In (a) it
is an Agent, in (b) an Object, Le. the object affected, and in ( c ) the
Instrument. Fillmore proposed a list of such basic semantic primes or ‘cases’
to underlie all sentences. The underlying structure of a sentence in this
theory is not an NP and VP which bear the syntactic relation of subject-
predicate, but rather a verb.and its semantic roles or cases.
These semantically oriented theories produced a great deal of debate
within the linguistic literature. The issue was whether these semantic roles
were primitive, as argued by Fillmore, or derived. They can be considered
derived in the sense that the semantic component of the grammar can
determine them from the underlying syntactic structure of the sentence and
the lexical features of its nouns and verbs. In such theories, they are
considered redundant. A response to case theory can be found in Chomsky
(1972). The general result of this debate has been a gradual incorporation of
such semantic relations into grammatical theory. One of the first attempts to
do this within the Standard Theory is Jackendoff (1977). Chomsky’s more
recent theory, known as Government and Binding Theory (Chomsky 1981),
includes a subtheory known as theta-theory to deal with them. Within
current literature, these roles are referred to as theta roles (see Radford
1981 for a useful summary). Recent theoretical work such as Marantz
(1984) and Culicover & Wilkins (1984) even have a separate linguistic level
where such roles are represented, distinct from the level where syntactic
ones occur.
Semantically oriented theories had an instant appeal to those working in
Child Language. It was noted that the range of semantic roles in children’s
early sentences was quite restricted. Sentences like (7.13b,c) did not look
like the sentences which occur in early language samples. The idea that a
grammar begins with semantic primes and later groups these into syntactic
relations looked like a viable model for explaining early language acqui-
sition. Further, it was suggested that such primes had as their roots the
child’s early concepts which appear at the end of the sensorimotor period
(e.g. Sinclair 1971). Child Language research in the 1970s explored these
ideas by examining the kinds of semantic relations which occur in early child
speech, and exploring the possibility that they are the result of the child’s
level of cognitive development.
Bowerman (1973a) The transition to more semantically oriented views of

early grammatical development began with comparisons of Standard
Theory grammars with case grammars. Two studies in particular stand out
in this regard. One of these is Bowerman (1973a), who examined the
acquisition of Finnish by two children, Seppo and Rina. Bowerman wrote
pivot, Standard Theory, and case grammars for her two Finnish subjects,
Table 7.14 Data used in Bowerman ( 1 9 7 3 ~for
) the comparison of pivot
grammar, the Standard Theory, and case grammar
No of
Age utterance
Child Sex (months) MLU types” Language Investigator
SePPo M 23 1.42 297 Finnish Bowerman (1973a)

SePPo M 26 1.81 437 Finnish Bowerman (1973a)
Rina F 25 1.83 338 Finnish Bowerman (1973a)
Kendall F 23 1.10 136 English Bowerman (1973a)
Adam M 27 2.06 505 English Brown (1973)
Eve F 18-19 1.68 309 English Brown (1973)
Sarah F 27-28 1.73 265 English Brown (1973)
5 subjects 49 Luo Blount (1969)
Sipili M 30 1.52 not given Samoan Kernan (1969)
a Total of unique types of syntactic and single-word utterances.
then compared these to comparable grammars written for selected other

subjects. One of these was an English child, Kendall, whose data were
collected by Bowerman. These were also compared with other English data
from Adam, Eve, and Sarah, and other non-English data from Luo (an
African language) and Samoan, collected by Blount and Kernan
respectively. The former investigator had a difficult time obtaining language
samples, discovering that Luo children are not usually addressed by adults,
much less foreign visitors. These data, therefore, needed to be grouped.
Table 7.14 gives a summary of relevant information about the samples.
Bowerman criticizes both the pivot grammars, for reasons already
discussed, and the Standard Theory grammars. While the Standard Theory
grammars improve upon pivot grammars in capturing more of the complex-
ities of the child’s grammar, they fail by assigning more structure to the child
than Bowerman feels is warranted by the data. Since these arguments have
proven to be the central ones against early syntactic grammars, we will look
at them more carefully (they are also discussed in Bowerman 1973b, 1975).
First I present the patterns observed in the data, then discuss their
interpretation.
As is evident from Table 7.14, the first sample from Seppo at 23 months
was the least developmentally advanced of the Finnish samples. The analysis
of it indicated that the construction N+V was by far the most frequent
construction, occurring in 30 syntactic types. The next most frequent one
was N+V+N which occurred in only six types. All other constructions
appeared in four types or fewer, and would have been considered marginal
by Bloom’s criterion of productivity. These two constructions also consist-
ently occurred with the word order indicated, although some of the less
frequent ones showed some variability. Although Finnish has a highly
variable word order, due to its rich morphological system, the unmarked or
most frequent one is N+V+N, as in English. An analysis of the mother’s
use of word order indicated that she most often used the unmarked order at
the time of the sample. Bowerman concludes that the mother’s word order
is probably the source of the consistent order used by Seppo. Another
feature is that subject and object NPs were always just an N. The only NPs
that consisted of a modifier and an N occurred in isolation. Seppo’s
Standard Theory grammar is given in example (7.14). ‘Proloc’ refers to a
class of grammatical morphemes called ‘prolocatives’, which are pronouns
with case endings which have meanings such as ‘here’ and ‘there’.
(7.14) Shl-
Sh2 - (M)(N) (Select at least one)
(N) vp
LOC --
Proloc
{ PNroloc
‘tuossa’ (‘there’)
The later grammar for Seppo at 26 months showed a number of advances
over the earlier one. Some of the more noteworthy are as follows: (1) while
the previous grammar was primarily restricted to two-word strings, three-
word strings became more frequent; (ii) modifier-noun strings could now
be embedded within subject and object NP; (iii) locatives could now
co-occur with object NPs; and (iv) new constructions that appeared
included (a) the use of adverbs with the previous constructions, and
(b) copular sentences. There was still no evidence of the development of
inflectional morphology. The flexible word order of adult Finnish also
became more evident. Bowerman includes optional transformations which
allow for each of the following reorderings of constituents from their
unmarked orders: reversal of prolocative and object NP; reversal of’verb
and direct object, locative, or adverb; reversal of subject and verb; and
movement of adverb or prolocative to front of sentence. Rina’s grammar at
approximately the same MLU showed several, but not all, of the same
features. Rina did show the same degree of variable word order. Impor-
tantly, the distribution of the varying orders of constituents for both
children correlated with that of their mothers.
Bowerman uses these data to argue against the existence of two gram-
matical aspects of the child’s grammar at this time. Specifically, she wants to
argue that there is no evidence for the existence of either the constituent
Table 7.15 A summary of Bowerman’s (1973~)arguments against the
existence of a VP in children’s grammars
Argument for Argument against
Children should respond to questions like No evidence exists that this is the case.
‘What’s it doing?’ with VPs.
V and V + N utterances in child speech should The data indicate that V shows the same
show the same privileges of occurrence. privileges of occurrence with V + N and
N+V.
The use of ‘do’ as a pro-verb for VP,e.g. as in Such sentences are not found in the language
‘daddy like cake. mommy does too.’ samples of children during this period.
V + N utterances are more frequent in the The data from Rina, Seppo, and Kendall
speech of children at this time than N+V show the opposite (Table 16 in Bowerman
strings. 1973a).
In replacement sequences with verbs, The data from Rina, Seppo, and Kendall
children will first expand to an object NP show both kinds of replacement sequence.
before adding a subject NP.
VP, or the relation ‘subject’. The hrguments are built upon a version of the
Competence Assumption. The first set of arguments are against the
postulation of a VP. These are summarized in Table 7.15. These arguments
claim that there is no valid test available to establish the existence of a VP.
Those tests based on frequency are inadequate, because they would lead to
the conclusion that subject and verb form a constituent for Seppo, Rina,
and Kendall, since subject+verb constructions for these children were more
frequent than verb-kobject ones.
A second set of arguments are made against the existence of the
grammatical relation ‘subject of’. Bowerman considers two reasons why the
notion of subject is needed in a grammar, and argues that neither of these
yet exist for the developmental data. One reason for subject is that it is
needed for certain grammatical transformations like Passive, where the
underlying subject becomes the surface object. This is not the case for this
period, because children at this time are not producing constructions like
the passive which require this category. The second reason is for the
operation of rules like Verb Agreement, Subject Pronominalization, and
Case Marking (in Finnish). Again, there is no evidence yet in the child’s
sentences for the operation of these rules. She comments that the assign-
ment of subject to the child’s early sentences is imposing our adult
knowledge of the language upon the child. She concludes that the data only
support the postulation of sequenced semantic categories.
Brown (1973) A second work which compared the various approaches to

the grammatical description of children in this period is Brown (1973).
Table 7.16 Definitions and examples of eleven semantic relations proposed

for Stage I grammars, taken from Brown (1973: 187-98)
Relation Definition and examples
1. Nomination The naming of a referent, without pointing, usually in response to the

question ‘What’s that?’ Often indicated with words such as ‘this’, ‘that’,
‘here’, ’there’. (Also see Demonstrative and Entity below.)
2. Recurrence The reappearance of a referent already seen, of a new instance of a
referent class already seen, or an additional quantity of some mass
already seen, e.g. ‘more’ or ‘another’ X.
3. Non-existence The disappearance of something which was in the visual field, e.g. ‘no
hat’, ‘allgone egg’.
Semantic Functions
4. Agent + The agent is ‘someone or something, usually but not necessarily animate,
Action which is perceived to have its own motivating force and to cause an action
of process’ (p. 193), e.g. ‘Adam go’, ‘car go’, ‘Susan off’.
5. Action + The object is ‘someone or something (usually something, or inanimate)
Object either suffering a change of state or simply receiving the force of an
action’ (p. 193).
6. Agent + A relation which uses the two definitions above. It can be considered a
Object direct relation without an intervening action.
7. Action + ‘the place or locus of an action’ (p. 194), as in ‘Tom sat in the chair.’ Often
Location marked by forms like ‘here’ and ‘there’.
8. Entity + The specification of the location of an entity, i.e. any being or thingwith a
Locative separate existence. These take a copula in adult English, e.g. ‘lady home’
meaning ‘the lady is home’.
9. Possessor + The specification of objects belonging to one person or another, e.g.
Possession ‘mommy chair’.
10. Entity + The specification of ‘some attribute of an entity which could not be known
Attribute from the class characteristics of the entity alone’ (p. 197), e.g. ‘yellow
block’, ‘little dog’.
11. Demonstrative The same as Nomination except that the child points and uses a demon-
and Entity strative.
Brown also provided an in-depth review and critique of pivot grammar, the
Standard Theory, and case theory. Like Bowerman, he took the position
that the primary development during this period is the acquisition of a basic
set of semantic relations, which are the building blocks of later develop-
ment. His work extended that of Bowerman, however, in attempting to
determine the set of semantic relations that children have.
Brown’s study of children’s semantic relations consisted of analyzing the
data from 24 language samples collected from 18 children. While the bulk of
the data are from English, there are also samples from Finnish, Swedish,
Samoan, and Mexican Spanish (see Brown 1973: 66, Table 9 for details).
He also used a second set of data from 17 children acquiring American
English, French, German, Hebrew, Japanese, Korean, Luo, and Russian.
These data are secondary in that there is less information on these samples
than the others. The two sets of data together constitute a broad base from
which some general statements can be made.
Brown’s analysis resulted in the claim that there are 11 kinds of semantic
relations that underlie the child’s early word combinations, divided into two
subtypes. First, there are the operations of reference. These are relations
where one term defines a property for a set of arguments. Brown specifies
three of these, which are defined here in Table 7.16. An example would be
Gia’s category of ‘recurrence’ discussed above for the grammar in (7.11).
The words marking recurrence, such as ‘more’, define a property of
nominals, in this case that they are repeated or occur in several instances.
Brown proposes that these are the relations that create the pivot look of
early grammars. The second subtype is a set of eight prevalent semantic
functions or relations. These relations consist of the categories that semanti-
cally underlie the grammatical relations such as those in (7.9). These are
called ‘prevalent’ because these are the ones that occurred in the samples
discussed above; Table 7.16 also gives definitions and examples of these.
In Brown’s analysis, all of the multi-morpheme types used by the first
group of subjects were categorized according to their semantic relations,
using the kind of evidence discussed above for Bloom. For purposes of
analysis, he collapsed Nomination with Demonstrative and Entity. The data
were scored according to the percentage of each child’s multi-morphemic
types which fell into each semantic relation. Brown found that the eight
‘semantic functions’ or ‘prevalent relations’ accounted for approximately 70
percent of children’s utterances (Brown 1973: 174, Table 22). Two of the
relations, Agent+ Object and Entity+Locative, were somewhat marginal,
in that neither constituted over 9 percent of any child’s sentences by this
measure.
Brown goes on to propose that this particular set of relations occurs
because they represent the knowledge acquired about the world during the
period of sensorimotor development. He states (p. 200): ‘Representation
starts with just those meanings that are most available to it, propositions
about action schemes involving agents and objects, assertions of nonex-
istence, recurrence, location, and so on.’ Since they are part of every child’s
early representation, they are also proposed as being universal. That is,
every child acquiring language will show the same relations.
Another important feature of Brown’s study is his comparison of
two-term and three-term multi-morpheme relations. He found that there
were two ways in which his subjects expanded their constructions. These are
what he calls concatenation-with-deletion and expansion-of-one-term.
Concatenation-with-deletion is the case where two appear to combine, with
the shared term being deleted in one of them. An example would be the
combination of Agent+Action with Action+Object to form a new
Agent+Action+Object construction. Expansion-of-one-term is the case
where categories that are NPs are expanded. For example, the Object in an
Action+Object construction will expand to also mark Possessor+Possess-
ion, as in ‘hit ball’ expanding to ‘hit Adam ball’. The latter type is
particularly of interest because such utterances can be taken as evidence of
an emerging NP structure. Brown’s analysis of these indicates that they are
relatively rare, only appearing toward the end of Stage I.
The above analysis led Brown to propose the law of cumulative com-
plexity :
It predicts that any child able to construct x+y will also be able to
construct either x or y alone. It does not predict . . . that a child able to
construct both x and y, severally, will be able to construct x+y. There
is evidently, and this is simply an empirical discovery, some additional
knowledge involved in putting the component items of knowledge
together to make a more complex construction. (Brown 1973: 186)
The fact that Brown does not want to make this law predict the existence of
x+y on the basis of the separate occurrences of x and y shows the
conservative stance which Brown takes about the child’s grammatical
knowledge. For example, he does not want to propose that the child has an
Agent+ Action+Object structure ( x + y ) on the basis of observing
Agent+Action ( x ) , and Action+Object 01).In this way he is more cautious
than Bloom, who does just that. A further feature of this law is that it is also
quite similar to what we have called the Constructionist Assumption. It
relates the structures at one point in acquisition to those which have
preceded them.
Brown’s study is representative of a wide range of research done in the

1970s. This research had the following characteristics. It was concerned
with accounting for the child’s early word combinations on the basis of
semantics rather than syntax. The only syntactic ability assigned to the child
was usually word order, although even this was not always found to be
consistently used. Further, these semantic relations were explained as
coming from the child’s sensorimotor knowledge. Language acquisition,
then, was seen as being closely tied to, and even the result of, cognitive
ability. Paradoxically, this continuity with early cognitive development was
juxtaposed with a discontinuity with later syntactic acquisition. That is, few
of these studies attempted to account for how the child ever got from a
semantic grammar to a syntactic one. One of the leaders in this area was
Schlesinger (1971), who developed his own theory on the acquisition of
semantic relations. A detailed account of his research on early grammatical
development appeared in Schlesinger (1982). While Brown and Schlesinger
tended to be conservative in the amount of semantic structure assigned to
the child’s sentences, others (e.g. Antinucci & Parisi 1973; Leonard 1976)
proposed quite rich semantic grammars. Leonard (1976) includes a com-
prehensive review of this literature (Appendix B, pp. 189-44).
While other investigators proposed similar sets of semantic relations,
they were by no means the same. Howe (1976) reviewed the ones presented
in Bloom (1973), Brown (1973), Schlesinger (1971) and Slobin (1970), and
pointed out a number of discrepancies between them. She also examined
what she saw as the two general methods or ‘strategies’ used to determine
them. These methods were needed since the meanings of early word
combinations are difficult to determine from their surface structure alone.
One method was to observe the referents of each word and then infer their
relation. The other was to rely heavily on the word order and expand the
child’s sentence into a grammatical adult one. Howe argued that these
strategies share one basic flaw - they assume that the adult and the child
share the same cognitive view of the world. Her conclusion about these
studies is therefore radically different:
Their most significant, though unintentional, contribution to the study
of child language development could be their demonstration of the
strategies parents use to interpret their children’s speech. Parents
assume that the two-word utterances of their children express one of
the meanings they would express. Parents use word order and their
perceptions of the relation between the referents of their children’s
utterances as clues to interpretation. (Howe 1976: 45)
Howe went on to suggest her own method for the study of the meaning of
children’s early sentences. This method has the following characteristics:
(i) ‘begin with the possibility that the two-word utterances of young
children refer to situations to which adults would never refer’ (p. 45);
(ii) ‘specify those situations which children can conceive and observe how
they refer to them’. She predicts that the categories that result from such an
analysis will be more general and fewer in number. She mentions three
likely ones - Actions, States, and Names of concrete objects.
Howe’s criticisms are useful in warning us against the pitfalls of imposing
aspects of the adult grammar onto children’s sentences. Her comparison of
the various semantic taxonomies of the time also points out the difficulty in
coming up with an acceptable set of semantic relations. Disagreements of
this kind have also appeared recently in the area of linguistic theory in
attempting to characterize thematic (or semantic) relations or roles. It is not
a problem, therefore, that is restricted to child language research. The
method she suggests is too general in its presentation to determine its
feasibility. It has, however, one useful feature to it. It begins with the
assumption that the child is not like the adult, and puts the onus on the
analysis to prove otherwise. This, we will argue, is an important methodolo-
288 T H E PERIOD OF T H E FIRST WORD COMBINATIONS
gical assumption which should underlie research on child language acqui-
sition. It does not logically exclude the possibility that we may ultimately
conclude that the child and the adult are the same, but it makes that point
the result of the analysis, not its premise.
Braine (1976) Another study which takes issue with the semantic rela-
tional analyses of that time is Braine (1976). Like Howe, Braine feels that
the number of semantic distinctions in child speech at the onset is much
smaller than those proposed by others such as Brown (1973). Unlike Howe,
he believes that these meanings are more specific, rather than more general.
In Braine’s view, the child sentences first express a small set of very specific
meanings that will vary from child to child. This position looks very much
like a revision of pivot grammar which takes into account the intended
meanings of the child’s sentences. Due to the extent of the data analyzed
and the methodological care taken, Braine’s study has had a major impact
on the analysis of early grammars. Here we will look at the results of his
study, as well as his methodology.
Braine conducted grammatical analyses on 16 corpora from 11 children
acquiring either English, Samoan, Finnish, Hebrew, or Swedish. Several of
these corpora have been discussed already. Collectively, they are children
who tended to be in early Stage I in our definition of this stage. The fact that
these children were at a relatively early point in grammatical acquisition
becomes important when we compare the results with the other studies we
have reviewed. Table 7.17 provides some information about the children
studied.
Of particular importance are the methodological procedures Braine used
in his analysis. First and foremost, Braine was very strict in setting criteria
for the existence of rule-based behavior. H e examined two properties in
making decisions of this kind - positional consistency and productivity.
Positional consistency refers to the requirement that a proposed pattern
reflect a strict ordering of its constituent parts. This property is never fully
justified but presumably follows from the assumption that children acquir-
ing languages with strict word order will use that in their first rules. In
specifying this, Braine also was responding to the earlier criticisms of pivot
grammar that no criterion was used to determine if a morpheme was
occurring in a fixed position or not. Positional consistency is operationally
defined as follows:
Criterion for positional consistency

No. of utterances with
(a) dominant order 6,7 8 9 etc.
(b) opposite order 0 1 2 etc.
First, the criteria require that at least six sentence types must occur before a
pattern can be considered. In addition, if there are only six or seven types,
then they all must show the dominant order, where dominant order means
the order of the rule producing them. If one sentence violates the appro-
priate order, then eight instances of correct order must appear (see Braine
1976:12 for details).
Productivity refers to the requirement that there be evidence of the
production of sentences that could not be accounted for by adult model-
ing. Andrew, for example, produced sentences like ‘no wet’ meaning
‘I’m not wet’, and ‘more hot’ meaning ‘another hot thing’. Such sen-
tences, Braine believed, would be evidence for an underlying rule. No
absolute criteria were set for this property, and it appeared to be deter-
mined by scanning the child’s sentences. In addition to positional consist-
ency and productivity, there is also the requirement of semantic consist-
ency which is never operationalized, but is presumed to be obvious from
the data.
The application of these properties to data from children resulted in the
following kinds of possible patterns:
positional productive patterns: patterns that meet the criteria of

positional consistency, productivity, and semantic consistency;
positional associative patterns: patterns which meet the criteria of
positional consistency and semantic consistency, but not the one of
productivity. These are so called because they combine positional
learning and the association of a word or words with a set of words
with which they have never been heard to occur;
groping patterns: patterns which only meet the criterion of semantic
consistency, and which are (i) small in number and (ii) produced with
some uncertainty. The child is attempting to express a new meaning
and is ‘groping’ for a way to do so;
free order patterns: patterns where productivity and semantic consist-
ency are met, but not positional consistency. These are also different
from groping patterns in that they do not meet both the criteria in (ci)
and (cii).
In addition to these patterns, there are approximately 25 percent of the data
which cannot be assigned. Braine proposes that such residual data are
probably imitations of adult models with no analysis yet done on them.
Braine used the data from Andrew, presented earlier in Table 7.4, to
illustrate the first three of these patterns. The positional productive patterns
include ‘no’+X, ‘more’+X, and ‘other’. Examples of its meeting the
productivity criterion are ‘no wet’, ‘more hot’ and ‘other milk’. The
constructions with ‘all’ are analyzed as a positional associative pattern.
While they meet the criterion of positional consistency, they do not meet
that of productivity. Braine states (p. 8):
It is unlikely, however, that the pattern is productive for the following
reason: because the set of English words that can co-occur with,‘all’is
rather small . . . it would follow that, if Andrew had some formula for
making combinations with ‘all’, one should expect that his use of the
formula would yield some . . . strange combinations he could not have
heard.
Since this is not the case, however, Braine places these into the positional
associative category. Lastly, Braine analyses the sentences with ‘all gone’
and ‘all done’ as examples of a groping pattern. They are small in number
and show variable word order. He also claims that they were uttered
‘hesitatingly’ (p. 11).
Braine did not write grammars for his data, but presented his results in
the form of lists of utterances grouped according to their semantic consist-
ency. Table 7.17 presents a summary of these semantic taxonomies. It gives
the meanings Braine proposed for each of the 11 children he studied along
with its pattern. This somewhat lengthy summary is given because his
analysis remains the most rigorous cross-linguistic grammatical study con-
ducted on this period. Also, the results of his analysis indicate the extent of
individual variation that appears to characterize early grammars once strict
criteria on productivity are applied.
When we compare the meanings used, we see that certain ones are more
frequent than others. Actor/Action, Possession, Location, and Identifica-
tion are the most frequent, followed by Recurrence and Requests. This is a
shorter list, however, than the one given by Brown. Importantly, Brown
was putting together a composite list, not trying to specify which patterns
would occur with individual children. Brown’s own data, in fact, show
variation of this kind, which Brown discusses. Each child appears to differ in
the meanings it selects for expression. Even though the labels suggest
comparable meanings, Braine also emphasizes that individual children may
restrict these in their own ways. These meanings are not universal €or
Braine in the way that they are for Brown. Braine states (p. 57): ‘Children
differ considerably in the kinds of contents expressed by their productive
patterns and in the order in which they acquire them.’ He points out, for
example, that there is virtually no overlap between Andrew’s patterns and
those in Kendall 11.
We have already discussed the issue of individual variation in regard to
phonological acquisition in 6.5.4. We pointed out then that the individual
differences claimed after analyses using the phone classes and phone trees
approach could be explained as a measurement sequence, i.e. that
Table 7.17 Summary of the subjects, meanings, and patterns found in
Braine (1976)
Children and
information’ Meanings Patterns
Andrew, Steven, and

Gregory2
(Braine 1963a)
Kendall I ActorlAction positional productive
(Bowerman 1973a) Possession positional productive
Location groping
Kendall I1 ActorlAction positional productive
Possession3 positional productive
Locatives with ‘here’, ‘there’ (~ncertain)~
Locatives (with other words) positional productive
Actor/Action/Locative (uncertain)
Identification’
1. ‘that’ + X positional productive
2.X+Y positional productive
??Verb/Object6 ??groping
SePPo ActorlAction positional productive
(23 months) Locatives with ‘tuossa’ (‘there’) groping
(Finnish) Locatives (with other words) positional productive
(Bowerman 1973a) ??Verb/Object6 ??groping
Sipili article-like Identification
(Samoan) 1. ‘le’ (the) positional productive
(Kernan 1969) 2. “0’(sign of nominative) positional productive
Possessives
+
1. X ‘ah’ (‘me’) positional productive
+
2. X “oe’ (‘you’) positional productive
3. ‘ma” + X positional productive
Tofi movement-to Locative positional productive
(26 months) (Verb + Location)
(MLU 1.6) ActiodActor (V + N)R positional productive
(Samoan) ActlPatient (V + N)9 positional productive
(Kernan 1969) Request
‘fia’ (‘want’) + X positional productive
Jonathan I Possessive positional productive
(see Table 7.2) size-Attribution
(Braine 1976) 1. ‘big’ 2. ‘little’ positional productive
Recurrence ‘more’ positional productive
+
color-word Object name’’ (uncertain)
Jonathan I1 Possessives
Property-indicating
1. ‘big’ 2. ‘little’ 3. ‘hot’ positional productive
4. ‘old’ 5. ‘hurt’ 6. color positional productive
7.‘(all) wet’ groping
Recurrence ‘more’, ‘other’ positional productive
Number ‘two’ positional productive
Disappearance ‘allgone’ positional productive
Locatives
1. objectfiocation positional productive
2. X with ‘there’ groping
?ActorlAction“ positional productive
Table 7.17 (cont.)

Odi Identifying patterns
(Hebrew) 1. ‘ze’ (that) + X positional productive
(23-26 months) +
2. ‘tirekiri’ (see) X positional productive
(MLU 1.4) +
3. ‘hine’ (here) X positional productive
(Braine 1976) Requests
1. ‘ten’ (give) + X positional productive
+
2. ‘efo’ (where) X positional productive
Locatives gropjng
Possessives groP’ng
David I Identifying form ‘here’ X + positional productive
(Table 7.2) Requesting expression
+
‘want’ X positional productive
David I1 Identifying forms
‘here’, ‘this’, ‘that’ positional productive
Requesting expression
‘want’, ‘Can-I’, positional productive
‘I-can’t’, ‘gimme’ positional productive
Recurrence ‘more’
Verb + ‘it’ positional productive
Embla 112 ActodAction
(Swedish) Locating or Identifying
(Lange & ,arsson 1973) ‘dar’ (there) X + positional productive
‘den’ (it) X+ positional productive
Embla 2 ActodAction positional productive
Locatives ‘ga’ (go) positional productive
Actor + ‘ga’ + Locative positional productive
Locative with ‘dar’
+
1. ‘dar’ X positional productive
+
2. X ‘dar’ positional productive
Conjunction ‘och’ (and) + X positional productive
Negation ‘inte’ (not) X + positional productive
’ These are only given for children not previously discussed in the text.
Data used to exemplify the kinds of patterns found. The meanings of individual patterns are
not usually given.
Braine actually labels these as Possessives for Kendall 11, but states in the text that he uses
‘Possessives’ and ‘Possession’ interchangeably (p. 16).
Braine speculates that ‘here’ was a positional (final) pattern at Kendall I, and that Kendall
has now learned that ‘here’ and ‘there’ can occur initially and finally.
Also called Class Membership.
There are several action phrases in the data without an agent, and with the object either
before or after the action word. Braine does not actually include these in his analysis because
the semantic basis for it is unclear.
’ Translated by Kernan as either ‘and’, ‘with’, or ‘for’. Braine is not sure if these are one or
several semantic patterns.
The basic word order in Samoan is Verb-Subject-Object although it is variable. Braine
+
combines Agent Action and Noncausative Movement Verbs (e.g. ‘pa’u:’ meaning ‘fall’) +
Object Moved in this pattern. He feels that the child makes a distinction between causative and
noncausative verbs, with the nominals having some kind of agent role.
+
This pattern includes Causative Movement Verbs Object Moved, and other transitive
Verb + Object sentences. The nominals are speculated as being Patients, i.e. objects affected.
lo These were semantically peculiar in that the color words were not used correctly.
Braine speculates that the nature of these sentences are such that they could be collapsed
+
with the Object Locative pattern on a semantic basis.
l2 Exact ages are not given, but the first samples were around 21-22 months, and the second
ones around 22-23 months.
approach gives a lot of importance to infrequent and idiosyncratic data. We
can now ask if Braine’s method does the same. In fact, by setting strict
criteria for what is considered a productive pattern, Braine has developed a
method which should minimize individual variation. Given this, his finding
of individual variation needs to be taken quite seriously.
Braine takes these grammars as an indication that children begin acqui-
sition by developing a small set of meanings for expression. These meanings
may be quite narrow, and have one member restricted to a single word, as in
David 1’s Identifying form ‘here’ or his Request expression ‘want’. Since
these early patterns appear to be independent, he doesn’t even want to call
them ‘rules’, since the latter implies the features of a generative grammar.
Instead, he calls them formulae of limited scope. They only become more
general as formulae with similar meanings are acquired.
Braine also accepts Bowerman’s criticisms against assigning a trans-
formational grammar to the child at this stage, but adds a third argument.
He gives the example of Kendall who acquired ActodAction as one of his
first productive patterns. According to the syntactic approach this should
lead to the rules:
rule 1 S- NP+VP
rule2 VP- (V)+(NP)
rule 2’ VP -
When the child acquires locatives, rule 2 would be then changed as shown:
Locative
Since the order of the subject in relation to the locative is done by rule 1,
then locatives should appear with appropriate ordering. The data from
Kendall and Seppo, however, indicate that Locatives were a groping
pattern at the time when ActodAction was acquired. These data, then, are
taken as evidence against the syntactic explanation of early word combin-
ations.
Assessment. Braine’s monograph is an important methodological and
descriptive contribution to our knowledge about early word combinations.
Its methodology is a major step towards extracting the productive utteran-
ces from a spontaneous language sample. The results of his analyses are an
impressive set of observations that require explanation. It appears that
children can be quite different in their selections from the set of meanings
which they could express. These data are quite a problem for any theory of
acquisition which states that all children begin in the same way.
We see two major limitations to Braine’s account of early word combin-
ations. One concerns the restriction of the data being considered to
spontaneous speech samples. While Braine is adhering strictly to his version
of the Competence Assumption, our formulation of it does not restrict data
in this way. Information about the child’s receptive knowledge as well as
other forms of expressive language can and should be used. We have
already seen the viability of the imitation technique in Rodd & Braine
(1970), and will explore still others in Section 7.5. The second limitation is
that it avoids accounting for how the child ever gets from this kind of
grammar to an adult one. Braine briefly discusses this fact toward the end of
his study (pp. 88-9). He correctly points out that there is no inherent
discontinuity in development in his proposals, should adult grammars be
different from the current syntactic models being considered. Only
additional research on later development will find if and when discontinuity
occurs.
Bloom, Lightbown & Hood (1975) The last study done during this period
of initial research into semantic relations is that of Bloom, Lightbown &
Hood (1975). Like the previous studies, it was concerned with examining
the language samples of young children for the kinds of semantic relations
which occurred. Its goals in doing this, however, extended far beyond
simply identifying a set of relations. They wanted to answer the last problem
just expressed concerning Braine’s study, i.e. how and when does the child
go from a semantic grammar to a syntactic one? They make this point very
clearly (p. 1) when discussing semantic relations:
It should not be surprising that these are the kinds of things that
children first learn to talk about. However, the linguistic means that
children learn for the representation of such notions, the sequence of
development in child grammar, and the relation of systems of child
language to the adult model remain to be determined.
In essence, this was an attempt to reconcile the apparent difference between
the proposals of McNeill (1970b) and Bloom (1970) with the results of the
studies by Bowerman (1973a), Brown (1973), and de facto Braine (1976).
That is, when does the child change from a semantic grammar to a syntactic
one? Or, stated differently, when does the child shift from semantic
categories (e.g. Agent) to syntactic ones (e.g. subject)?
The study was done by examining the spontaneous speech samples of four
children: Eric, Gia, Kathryn, and Peter. The first three children were the
ones studied in Bloom (1970), while Peter’s data were collected in a similar
fashion at a later time. Several samples were obtained from each child while
they were approximately 19-25 months of age (see Table 1 in Bloom,
Lightbown & Hood 1975: 7). Their MLUs across the samples ranged from
1.04 to 2.83. Based on the children’s MLUs, the authors felt that the
samples presented data from all four children on both early and late Stage I,
Table 7.18 Number of syntactic types for each sample period of four
children studied in Bloom, Lightbown & Hood (1975:7, Table 1); the
number of the sample period, and its MLU, are shown in parentheses
Ranges of
syntactic
types Peter Eric Gia Kathryn
0-50 7 (I, 1.04) 10 (I, 1.10)

5 (11, 1.09) 37 (11, 1.19)
70 (111, 1.37) 55 (I, 1.12)
80 (IV, 1.41)
81 (V,1.41)
101-150 108 (111, 1.42)
151-200
201-250 243 (VI, 1.75) 226 (11, 1.34) 226 (I, 1.32)
251-300 288 (111, 1.58)
301-350
351-400
401-450 401 (IV, 1.69)
451-500 458 (VII, 2.39) 457 (IV, 1.79) 767 (11, 1.89)
over 500 902 (V,2.63) 842 (V, 2.30) 1443 (111.2.83)
as defined by MLU values in Brown (1973), as well as Stage 11. This is so, as
can be seen by examining their MLUs, which are given in Table 7.18.
Recall, however, that we have challenged this definition at the beginning of
this chapter, and have defined the periods in Stage I by the number of
syntactic types as it interacts with MLU. Table 7.18 presents the four
subjects based on their number of syntactic types.
Table 7.18 reveals that these four children’s samples do not equally
represent both early and late Stage I, at least by our definition. There are
nine samples with 108 or fewer syntactic types, placing them in early
Stage I. Eight of these nine are from Peter and Eric. After a gap between
108 and 226 syntactic types, there are four samples between 226 and 288
which could be placed into late Stage I, assuming some flexibility in
interpreting our measure. Three of these four are from Gia and Kathryn,
and none are from Eric. The rest of the samples are beyond this range and
are in Stage I1 from our perspective. In other words, most of the early
Stage I data are from Eric and Peter, and most of the late Stage I data are
from Gia and Kathryn. This observation is important when we come to
analyze their data. We will only present their results regarding semantic
relations for those samples which fall within these two periods by our
definitions of them.
Bloom, Lightbown & Hood determined a semantic relation for each
syntactic type in the data. They did not use a predetermined list of relations,
but instead let the data reveal the meanings. Like Braine, they did not
provide a set of explicit criteria for this task, but imply as Braine does that
the meanings are in some way self-evident. They state (pp. 7-8):
Each multi-word utterance was examined and the semantic-syntactic

relations among words were identified by observing the relationship
between the utterance and aspects of the child’s behavior and the
situational context in which the utterance occurred. Obviously, one
cannot be confident that the semantic interpretation given to an
utterance by an adult does indeed equal the child’s semantic intent. A t
the least, it is necessary to establish that (1) any utterance can be
identified as a separate behavior (from the other linguistic behaviors
that occur) by observers, and (2) given the same information about
the utterance and nonlinguistic context and behavior, different
observers can assign the same interpretation to it.
The important advance here is that the validity of the relations is

determined by comparing the independent judgements of observers. In this
case, two interpreters were used and their judgements compared. This is an
advance over the other studies, which used a single individual for all
interpretations.
The analysis revealed over 20 semantic relations, although only a small
number of these proved to be frequent. To separate frequent from
infrequent relations, they used the same measure of productivity as in
Bloom (1970), i.e. a relation was considered to be productive if there were
at least five syntactic types for it. As mentioned in Bowerman (1975), this is
not a particularly good measure since it will be highly influenced by sample
size. Further, as pointed out in Ingram (1981d) and discussed here in
Section 7.6, it does not take into consideration crucial information about
the distributional properties of words. Five memorized phrases with similar
semantic properties could qualify as productive under this definition. Here
we will present their results with this measure with one proviso. A relation is
considered productive in the session when it meets this measure and
continues to do so on all further samples.
This adjusted measure reveals for the Stage I samples nine semantic
relations which were productive for at least two children. Here we summa-
rise Bloom, Lightbown & Hood’s definitions of each.
1. Action: sentences referring to movement where the purpose of the

movement was not to change the location of an object of person.
There were two subcategories - (a) transitive actions (our term for
this), where the semantic relation was agent-action-object, with any
two appearing in the sentence, e.g. ‘open drawer’, ‘I made’;
(b) intransitive actions, where only an agent and action were
involved, e.g. ‘Kathryn jumps’
2. Locative Action: ‘movement where the goal of movement was a
change in the location of a person or object’ (p. 11). As in the previous
relation, there were two kinds - (a) transitive locative actions (our
term for this), where the semantic relation was agent-action-object-
place, and at least two appeared in the sentence, e.g. ‘put in box’,
‘tape on there’; (b) intransitive locative actions, ‘where the agent and
affected object or person were the same’ (p. l l ) , e.g. ‘I get down’,
‘Mommy stand up (a) chair’. The semantic roles for this type were
mover-action-place. (Note that ‘mover’ is being distinguished from
‘agent’. This distinction will prove to be important later.)
3. Locative State: reference to a person or object and its location
without an indication of movement. The relation is object-state-place,
with at least two of these in the sentence, e.g. ‘I sitting’, ‘sitting on
chair’;
4. State: ‘reference to transitory states of affairs involving persons or
other animate beings’ (p. 112). There are three kinds of these -
(a) internal states with verbs like ‘like’, ‘need’, ‘want’, e.g. ‘I want
pretzel’, ‘Caroline sick’; (b) temporary ownership, e.g. ‘I have it’; and
(c) external states, e.g. ‘it’s dark outside’;
5. Existence: ‘pointing out or naming an object’ (p. 13);
6. Recurrence: ‘reference to “more” or another instance of an object or
event’ (p. 13);
7. Attribution: ‘counting, specifying or otherwise qualifying objects’
(p. 13);
8. Possession: (not defined);
9. Negation: ‘nonexistence, disappearance or rejection of objects or
events’ (p. 13).
These nine relations are remarkably similar to the frequent ones found by
Braine (1976). The primary differences appear to be in name. Braine’s
Requests and Identifications are classified here as State and Existence
respectively. Their terminology for locatives also differs. Braine refers to
Locative Actions as Movement-to-Locatives. Lastly, Negation occurs with
two of Bloom, Lightbown & Hood’s subjects, but it is not as frequent in
Braine’s data. Generally, though, there is a great deal of overlap.
Table 7.19 presents the acquisition of these relations by the four children
studied, divided into early and late Stage I. Like Braine’s study, the data
from Bloom, Lightbown & Hood reveal a certain degree of individual
variation. Even so, there are patterns which emerge. We can see this by
requiring that a relation exists when at least two children acquire it.
Table 7.19 The productive semantic relations for early and late stage I for
Eric, Peter, Gia, and Kathryn, adapted from Bloom, Lightbown & Hood
(1975: 15, Table 2). Relations are shown with the sample in which they
first appeared
Children
Stage Peter Eric Gia Kathryn
Early I: 111, IV I1 I
Action Action Existence
Recurrence Negation Recurrence
Attribution Existence
V I11
Locative Action Locative Action
State
Recurrence
Attribution
Late I: VI I1 I
Locative State Action Action
Locative Action Locative Action
Possession State
Attribution Existence
Negation
111 Recurrence
Locative State Possession
Attribution
Applying this measure to early Stage I, the following patterns appear in this
subperiod: Action, Recurrence, Existence, Locative Action, and Attri-
bution. The fact that two of the children show developmental changes even
allows us to impose an order to these. Action, Recurrence, and Existence
appear first, followed by Locative Action and Attribution. Given the
individual variation involved, however, we can only state this as a statistical
tendency. In late Stage 11, Locative State, State, Negation, and Possession
also meet our criterion. The data also indicate that there is an acquisition
order which is exceptionless in the data: Actions are acquired before
Locative Actions, which are in turn acquired before Locative States.
Bloom, Lightbown & Hood discuss these data in terms of their impli-
cations for the child’s grammar, particularly as evidence for the emergence
of the grammatical category of ‘subject’. First, several of these relations
involved initial nominal categories, i.e. Action, Locative Action, Locative
described syntactically by a rule of the form S -

State, and Attribution. Further, the initial nominals of these particular
relations occurred consistently in sentence-initial position. Each could be
noun or pronoun+verb
(or verb phrase). Recall that both Bowerman and Braine had argued that
there was no evidence to support such a claim, and proposed instead that
Table 7.20 The proportion of pronouns used to mark agents and affected
objects (Object) in Actions for Peter, Eric, Cia, and Kathryn, calculated
from Bloom, Lightbown & Hood (1975: Table 3)’
Ranges of
syntactic Peter Eric Gia Kathryn
types agent object agent object agent object agent object
0-50 - - - -
- - - 0.67
51-100 - - - -
0.86 0.74
0.75 0.89
101-150 0.88 0.78
151-200
201-250 1.00 0.64 0.00 0.05 0.17 0.08
25 1-300 0.01
301-350
351-400
401-450 0.75 0.49
451-500 0.76 0.45 0.06 0.12
over 500 0.77 0.39 0.77 0.30 0.08 0.34
0.57 0.38
Calculations are shown only for cases where Jive or more instances occurred.
children’s initial rules were solely semantic. Bloom, Lightbown & Hood
state (p. 28): ‘The kind of evidence that both Bowerman and Schlesinger
might accept in order to attribute such knowledge of grammatical relations
to the child would be the occurrence of superordinate categories whereby
words in the same syntactic position took on different semantic functions
relative to one another.’ They interpret the gradual and consistent spread of
initial nominals to different semantic roles as evidence for the syntactic
category of subject. In Kathryn I, for example, the subject of sentences
could be either an agent (in Actions), mover (in Locative Actions),
possessor (in Possessions), or object (in Locative States).
Besides examining the emergence of semantic relations, Bloom, Light-
bown & Hood looked into the acquisition of pronouns and nouns by their
subjects. They observed that the two boys, Eric and Peter, tended to mark
the semantic roles accompanying actions with pronouns. The two girls,
however, did so with nouns, and did not show pronouns until later in
development. This pattern was consistent for Possession, Action, and
Locative Actions. It was demonstrated by counting the number of times a
possessor, agent, mover, affected object, or place was indicated by either a
noun or pronoun in syntactic types. Here we will present their results
somewhat differently, using the format of Table 7.20, and a different
measure than absolute numbers. For each semantic role, I calculated the
proportion of pronouns, which is the proportion of times that the child in
any sample used a pronoun instead of a noun. These data are presented in
Table 7.20 for the agent and affected object roles of Actions. Both Eric and
Peter used pronouns predominantly until they were beyond Stage I. The
opposite was the case for Gia and Kathryn who used predominantly nouns.
All subjects eventually acquired the other category by the latest samples
where their proportions are relatively similar.
Bloom, Lightbown & Hood interpret these results as indicating that there
are two distinct strategies that children may use in acquiring an early
grammar. One is the categorization strategy which was used by Gia and
Kathryn. The girls used the semantic relations of early sentences to abstract
grammatical relations such as ‘subject’ and ‘predicate-object’. This strategy
has come to be known in more current literature as semantic bootstrapping.
We will examine this notion more carefully later, but essentially it claims
that the child’s language acquisition device looks for semantic roles which
pattern together for the purpose of forming more abstract categories. As
pointed out by Bloom and her colleagues (p. 32), these may not initially be
adult-like. Gia and Kathryn’s initial subject included possessors as well as
agents and movers. Even so, it presumably will restrict the possibilities
available in ways which are unknown at this time.
The second strategy is the pronominal strategy which was proposed for
Peter and Eric. This strategy is one which bears similarities to the early
pivot grammars and Braine’s more current view of formulae. Here we will
refer to it as an alternative to semantic bootstrapping which accounts for
early development by distributional learning. The children’s pronouns
operated like pivots with each marking a particular meaning. Peter and
Eric, for example, used ‘I’ to indicate agents, ‘my’ for possessors, ‘it’ and
‘this one’ for affected objects, and ‘here’ and ‘there’ as locations. If the
children were acquiring pronouns semantically for the purpose of estab-
lishing grammatical relations, we would expect a very different pattern.
Some pronoun such as ‘my’, ‘me’, or ‘I’, for example, would be expected to
represent several semantic roles as the subject, much the same as ‘Mommy’
did for Kathryn. We would then expect mistakes in their use until the adult
pronominal case system was acquired. This did not happen, however, and
instead the pronoun functioned as a formal marker of the individual
semantic role of the word with which it occurred.
If these pronouns are not leading to grammatical relations, what do they
help the child to acquire? As given in our discussion of pivot grammar, they
allow the child to acquire the grammatical classes of the language. For
example, the child might begin by identifying ‘it’ as a frequent morpheme
which occurs after a number of action words which it knows. ‘It’ then comes
to define that particular class as ‘words which occur with ‘‘it”’. Eventually,
as other properties are added, this becomes the class of English transitive
verbs. Another morpheme such as ‘the’ comes to identify the common
nouns. Only later will more detailed information be acquired about these
initially formal markers.
These two strategies lead to very different theories of early acquisition.
The categorization strategy (or semantic bootstrapping) leads to early
identification of grammatical relations, with subsequent distributional
learning. For example, the early identification of lexical items which
function both as agents and movers establishes an early category of subject.
Presumably Universal Grammar or subsequent distributional learning leads
to grammatical classes like noun or noun phrase. The pronominal strategy
(or more general distributional learning) predicts early specific learning of
positional word classes which develop eventually into grammatical cate-
gories, then grammatical relations. For example, words which occur with
‘it’ become verbs, and subsequently function as the predicate in the
predicate-object relation. In Section 7.5 we will look in more detail at a
recent proposal of this kind.
Assessment. The Bloom, Lightbown & Hood study was an important
first attempt to reconcile the discovery of semantic relational meanings in
children’s early word combinations with earlier proposals for the existence
of a syntactic grammar. They pointed out the need to account for the
transition to an adult grammar, and provided a major argument in support
of the existence of grammatical relations. This argument was empirically
based, i.e. it came from the analysis of children’s language. They also
noted that the development of these relations could occur in two very
different ways. One allowed for the early establishment of general semantic
categories, the other for early formal ones. The effort to decide between
these alternatives characterizes most of the subsequent theoretical research
in the field.
Its weaknesses stem from both its method and its gaps in theoretical
interpretation. The measure of productivity was too weak to establish with
confidence what the children’s grammars actually looked like. No formal
grammars were written, and claims about the child’s grammatical know-
ledge were usually in the form of taxonomies or lists of numbers. Even with
this reservation stated, the numbers in the larger studies, with our added
criterion, result in data that need to be explained by any serious candidate
for a theory of grammatical acquisition. The theoretical interpretations do
not deal with some important questions. One is how the child actually gets
from the semantic relations to grammatical ones, i.e. what is the child’s
learning mechanism? Later we will examine a proposal by Pinker (1984) for
one such mechanism. There is also little in the way of theoretical expla-
nation as to why children would use two apparently opposing ways to
acquire language. Bloom, Lightbown & Hood may be on the right track
when they suggest it may be due to alternative hypotheses on what type of
language English is. Most recent work attempts to select one or the other
procedure as the one children follow. The presentation of the data,
however, forces all present investigators who support one of these views to
account for the development of the two children who seem to have used the
other.
7.4 Current theoretical approaches

The research up to 1976 on early word combinations was characterized by
an effort to collect and analyze data from young children. Because of the
time:consuming nature of this effort, it resulted in data from a relatively
small number of children. It has been pointed out that the subject pool for
early Stage I, for example, is particularly small. Also, this research had only
begun to work out the methodological problems in analyzing data from such
young subjects. On the theoretical side, it laid out the basic alternative
explanations, and discussed ways these positions might be proven through
the analysis of child data. In the process, it concentrated on the child’s
creative role in acquisition. The research did not, however, focus on what is
the child’s learning procedure, given the primary linguistic data.
Much of the research since the mid 1970s has attempted to specify more
exactly the steps the child follows to get from the linguistic input to its
internal grammar. In doing so, it has concentrated on theoretical specu-
lation, with virtually no data collection, at least for the period under
discussion. As we will see, most of this research refers back to the data
collected in the studies just reviewed, and it is largely because of this fact
that these data were presented in relative detail. Much of the new data has
been on children beyond this period, often in the form of experimental
studies. This is not surprising, in that the concern to account for how the
child reaches an adult grammar led people to look at slightly older children.
It was influenced also no doubt by the apparent individual variation being
shown by the younger children.
While the recent work shares a theoretical emphasis, it does not neces-
sarily have a common theoretical orientation. Somewhat simplified, it can
be said that approaches differ in the amount of formal grammar assigned to
the child’s early grammar. Several accounts which assign very little gram-
matical knowledge to the child have developed from Braine (1976). One
such approach, which is referred to as functionalism, proposes that not only
does the child begin acquisition with a small set of restricted meanings, but
that these meanings may actually evolve into a more elaborate set which
underlies the adult grammar. It handles the problem of transition to an I
abstract syntax by denying the existence of the latter. Braine’s later work,
for example, attempts to explore the semantic relations of Fillmore’s case
grammar (Braine & Hardy 1982). Some people have tried to detail how the
early restricted formulae lead to more general grammatical categories. We
will refer to these as distributional theories. They vary from those which
reject abstract syntax (Kuczaj 1982a), to others which allow for the
possibility of its eventual emergence (Maratsos & Chalkley 1980). A major
concern has been to determine a set of procedures which the child applies in
acquisition. These procedures take the child through several steps before
adult-like rules are established.
Others have argued for the early appearance of abstract syntax, and
resemble McNeill’s pioneering approach in this regard. These researchers
assume the existence of an abstract grammar whose early existence can be
predicted on theory-internal grounds. This general orientation was dis-
cussed in Chapter 4 under the label of Language Acquisition. One
approach of this kind has been referred to as semantic bootstrapping. The
child begins the acquisition of formal syntax by applying a set of procedures
to the first semantic relations, resulting in essentially an immediate syntax.
This was proposed in one form by Macnamara (1972, 1982), and by
Grimshaw (1981) and Pinker (1984). Another rich interpretation is the
parameter-setting model of Chomsky (1981), in which the child begins
acquisition with a set of innate parameters about what is a possible
grammar. Some are unmarked and presumed to exist in the language unless
counterevidence is presented. The determination of a particular parameter
is an important event for the child because it predicts a range of other
characteristics about the language. This model has been applied to language
acquisition by Hyams (1983, 1984,1986), among others. Here we will look
at each of these four approaches in more detail.
7.4.1 Functionalism
This is a term which can be traced back to the distinction between ‘form’ and
‘function’in language, where form refers to grammatical properties such as
verb agreement, and function refers to the role these play in communi-
cation. Functional properties cover a range of things, for example the
semantic relations that underlie sentences, the pragmatic intentions of
speakers, the influence of cognition on language, and communicative
aspects such as informativeness. In its most broad use, functionalism can be
said to be the emphasis on the functional properties of language in general,
and in our case, in early acquisition. As such it can cover a diversity of
studies. We have already discussed a number of works that come under this
general use of the term - Halliday’s (1975) analysis of the first words and
Brown’s (1973) specification of universal semantic relations. Muma (1986)
surveys the range of research which comes under this general definition of
functionalism.
While a large number of studies can be considered functionalist, the term
itself did not become consistently used until around 1979, when it appeared
in the title of Karmiloff-Smith’s (1979) study of the acquisition of deter-
miners, and in a review of research by Bates & MacWhinney (1979). The
latter two in particular have developed the use of the term, as seen in their
I
subsequent review in Bates & MacWhinney (1982). In its broad or
descriptive use, functionalism refers to a range of research on pragmatic
aspects of language. The recent uses of it, however, are more as an
explanatory theory of language.
Bates 81 MacWhinney (1982) We can examine functionalism as an

explanatory theory by looking at the discussion in Bates & MacWhinney
(1982). They contrast two ways of doing psychological research into
language. One is to take linguistic theories or models and test their
a
psychological reality; such research, for example, would involve seeing if
more complex structures of the grammar do predict greater processing
time. Another way is to begin with the study of how people (adults and
children) process language, and use this processing as the explanation of the %
structure of language; i.e., to say that the structure of language is such

because of constraints imposed upon it by communicative needs. For
example, we have discussed the point that English has a grammatical
relation ‘subject’ which seems to conflate several semantic roles - see
(7.13). The formalist may be satisfied with noting this correspondence, and
then accounting for the properties of subjects within a formal system; the
functionalist, however, wants to explain why such a conflation occurs in the
first place. This use of functionalism emphasizes the processing of language,
and how this processing determines the form of language.
Bates & MacWhinney offer their own form of functionalism, called the
competition model. This preliminary model has been influenced by previous
research both in linguistics (e.g. Giv6n 1979), and in child language (e.g.
Slobin 1973, 1977; Karmiloff-Smith 1979). The term ‘competition’ comes
from the claim that the functions which underlie language must compete for
the small number of possible surface representations. For example, two
functions may ‘divide the spoils’ or share ‘peaceful coexistence’ (p. 192). If
they divide the spoils, then each function maps onto its own surface
representation. If they show peaceful coexistence, then they share the same
surface manifestation. The results on the use of pronouns and nouns in
Bloom, Lightbown & Hood (1975) exemplify these alternatives. Recall
from Table 7.21 that Eric and Peter used unique pronouns to mark each
semantic relation. This would be a case of choosing to divide the spoils.
Kathryn and Gia, on the other hand, show several semantic roles mapped
onto the same surface positions in the sentence. This, then, would be
peaceful coexistence. Presumably Bates & MacWhinney would account for
these two types of pattern as cases where the children differed initially in
the way to resolve the competition between their newly acquired functions.
The competition model presented in Bates & MacWhinney (1982), as
they admit, is still in a preliminary form. Also, since much of it is devoted
to outlining a model of human language processing, there is relatively little
detail about the form of either the child’s or adult’s grammar. They do
outline, however, how they see development proceeding from the onset of
word combinations to the adult language (pp. 183-7). I will present
these speculations in the form of a series of steps to the acquisition of
‘subject’, with our own labels for each stage:
Step 1: Establishment of basic functions. At least two functions seem to
be acquired before the onset of syntax: topic-comment, and agent. Topic-
comment is a fundamental function for Bates & MacWhinney which
underlies much of their acquisitional research. They define these terms as
follows (p. 199):
Topic is defined as what is being talked about, and comment is
defined as the point being made about the topic. In other words, the
topic-comment system involves a single communicative function of
point making, albeit applied in a variety of ways in the service of a
variety of motives.
These functions exist in the holophrastic period before the onset of syntax,
and thus word order.
Step 2: Competition of surface representations. The child seeks to repre-
sent these two functions by competing for surface forms. They eventually
will either map onto different surface forms (divide the spoils), or the same
one (peaceful coexistence). This step will reveal a certain amount of
groping and problem solving. The surface forms available in any language
are limited, due to acoustic-articulatory restrictions, to one or a combin-
ation of the following four choices (p. 190): 1. lexical items, 2. word-order
patterns, 3. morphological markings, and 4. intonational contours.
Research from Bates & MacWhinney (1979) indicates that the language
being acquired influences the patterns of this step. For example, Hun-
garian and Italian children first show a comment-topic sequence, whereas
English children show agent-action sequences. These two examples indi-
cate an important claim of functionalism which is summarized as follows:
initially, maps single forms onto single functions. This was argued for in
Karmiloff-Smith (1979) where the multifunctional nature of determiners
was only acquired after an initial period of assigning them to single

functions.
Step 3: Acquisition of the adult system of mappings. True competition
occurs as the child needs to map all functions onto the surface. through
either peaceful existence or dividing the spoils. Abstraction of the category
‘subject’ happens through a process to be determined.
A theoretical issue arises in step 3 as to whether or not this step also
involves the abstraction of purely formal categories such as ‘subject’. Bates
& MacWhinney mention the theories that have tried to do this, and we will
examine some of these later in this section. They are dubious, however,
about the postulation of abstract formal categories in the first place. They
discuss the discontinuity involved in shifting from semantic to syntactic
categories, which they call the developmental shift hypothesis. Their own
preference, however, is to form a fully functional adult grammar: ‘In the
final and most radical functionalist view, the categories of adult grammar
retain their full functionalist definitions’ (p. 187). This position is based
upon an assumed Constructionist Assumption.
Much of the research on functionalism in recent years has been on
children beyond the period of the first word combinations. Functionalist
accounts of early word combinations tend to be somewhat general, as in the
three steps just discussed. This account assigns little in the way of grammati-
cal knowledge to the child, and accounts for the child‘s first sequences of
words in terms of functional categories such as topic-comment. A similar
view, for example, which argues for early sequences on the basis of
‘informativeness’, can be found in Greenfield & Zukow (1978). The
functionalist view is also marked by a commitment to developing a theory of
grammar which is quite distinct from those characterized by the postulation
of abstract syntax.
As an account of early word combinations, there seems to be an
important aspect that needs further elaboration. The theory claims that
children begin acquisition with a set of basic functions. If so, one would
expect that the functions would be constant across children, and thus also
their early grammars. The example discussed above, however, indicates
that there are cross-linguistic differences in this regard. For this to be the
case, the child will need to have done some formal analyses of the input
language to determine the formal means in it for marking a particular
function. However, we are not told the procedures the child follows to
select any particular formal means over another, nor anything about the
potential individual variation that may occur, given the child’s ability to
grope for forms. The early appearance of cross-linguistic differences
actually seems to support claims for early distributional analysis of the
input.
Such apparently contradictory data also appear in Karmiloff-Smith
(1979). As part of a large sample cross-sectional study of French children,
Karmiloff-Smith examined children’s ability to use the gender distinction in
French articles. French articles distinguish two genders, masculine and
feminine, although its use with nouns is more arbitrary than semantic. For
example, tables, rain, and music are feminine while dogs, mountains, and
milk are masculine. The appearance of the two classes of articles, however,
is partly determined by phonological properties of the words, i.e. by formal
means. One would expect from the functionalist view that children would
first use the articles in a semantic way, and later learn their formal
properties. The results, however, indicated that the first uses of the articles
were formal rather than semantic. That is, the children showed awareness
of the phonological properties of nouns which occur with the articles before
an awareness of their semantic features. Results like these indicate that the
interaction of distributional learning and functional mapping within this
theory needs more clarification than exists to date.
7.4.2 Distributional learning

We need to be careful when discussing a model of acquisition as one which
can be characterized by distributional learning; all theories will need to
account for it in some way. The issue is one of when this learning begins and
how it proceeds. Semantically oriented approaches like functionalism and
semantic bootstrapping propose that semantic categories occur first and
provide the basis for the onset of distributional learning. We are reserving
the term distributional learning for those approaches which allow for
distributional learning at the onset of acquisition, and which even allow for
it to account for semantic development. Braine’s models are examples of an
approach of this kind.
Current approaches based on distributional learning have explored how
the child might move from restricted formulae of the kind discussed by
Braine to the categories of the adult language. These approaches character-
istically propose procedures on how the child builds more general cate-
gories out of more specific ones. As such, they propose models which
probably could be incorporated into the semantically oriented ones, at least
once the latter begin to allow for more formal acquisition. Here we will
consider two studies which have tried to determine procedures of this kind:
MacWhinney (1982), and Maratsos & Chalkley (1980).
MacWhinney (1982) In an earlier study, MacWhinney (1978) developed a

computational model for the acquisition of morphophonology. The present
paper was an attempt to extend those ideas to the acquisition of early word
308 T H E P E R I O D O F T H E FIRST W O R D COMBINATIONS
combinations. The earlier work even provided flow charts to demonstrate
the learning procedure presented. The major feature of both approaches is
the claim that the child follows an ordered set of strategies to acquire
language. As is typical of computer models, these strategies are discussed in
terms of the capabilities of the model, not necessarily of the child’s
psychological state. The success of such a model to account for the
acquisition data, however, can be used to infer its psychological validity.
MacWhinney’s model consists of three processing mechanisms: rote,
analogy, and combination. The latter mechanism can be broken down
further into four strategies, resulting in the six strategies which MacWhin-
ney suggests allow the child to develop from very restricted to productive
grammars.
The first strategy is rote. This refers to the claim which has appeared in
various places in the literature that many of the child’s first sentences are
simply internalized imitations of parts of the adult language (e.g. R. Clark
1977,1982; Peters 1983). These may range from obvious ones used in games
and ritual situations to less obvious ones that could mistakenly be identified
as productive sentences. It is the concern for possible rote sentences, in fact,
which underlies some of the criticism of Bloom’s (1970) productivity
measure. MacWhinney argues that rote can also explain some of the
well-known errors made by children, such as redundant pronouns, e.g.
‘have it egg’ (Braine 1976: 36) and redundant auxiliaries, e.g. ‘what’s that is’
(Hurford 1975: 300). Examples such as these are often presumed to be
rule-based and have led to claims about the nature of the child’sgrammar. It
may also be, however, that they are simply the combination of a memorized
unit such as ‘have it’ or ‘what’s that’ with another word. The methodological
procedures to be presented in section 7.5 attempt to isolate such rote forms
in early samples.
While rote may play a role in holding forms in memory or for idioms in
the adult language, it will yield to procedures for building up rules. The first
primitive rules are developed in MacWhinney’s model by the process of
analogy. These are very limited in that they consist of breaking up a rote
form into at least two parts and replacing one of them with another lexical
item. MacWhinney cites an example from R. Clark (1974) of a child who
said ‘wait for it to dry’, a purported analogy from ‘wait for it to cool’.
Another example given is the child who says ‘I are’ in response to the adult
question ‘Are you?’. These are distinguished from the combinations
produced by the next procedures by being limited to pairs of lexical items,
although MacWhinney acknowledges that the difference between analogies
and combinations is not a sharp one.
The next four procedures all fall under the general category of combin-
ation. Combination refers to ‘the set of principles that serves to transform
an unordered string of lexical items into an ordered string’ (p. 91). Four
such procedures are presented in increasing order of generality.
MacWhinney proposes that the first causes for the ordering of lexical
items in early word combinations are non-syntactic factors. Their order is
determined by several functional dispositions which order words in par-
ticular ways. The claim that these are the first influences on word order
reflects the functional influence on MacWhinney’s work. Some of the
dispositions that can affect word order in these early combinations are as
follows:
(a) informativeness - the first item in the combination will be the one
which is the newest or most informative;
(b) complexity - short, simple words precede long, complex ones;
(c) agency - nominals which are agents tend to precede other roles;
(d) perspective - the first item will tend to be an item with which the child
can identify.
These and other dispositions are given with limited information on how they
interact and affect any particular set of child data.
The next procedure creates bound or item-bused patterns. These are the
pivot-like combinations that were described as restricted formulae in Braine
(1976). Our discussion of the latter work should be sufficient for providing
an idea of what such combinations are like. Their occurrence in this set of
procedures represents the onset of distributional learning. The implications
are similar to those of Braine - they suggest that the child notes the
positional properties of specific words and uses these to determine the
grammatical patterns of the language being acquired.
Item-based patterns lead to more complex ones, which MacWhinney calls
free patterns. These result from rules which combine words based on
general semantic features such as Agent and Action. Unlike item-based
patterns, these ones are not lexically restricted to specific words. Any word
which has the semantic properties to be an agent can enter into such a
pattern. The proposal that these patterns are seen to develop after more
limited item-based ones is a major feature of distributional learning
theories. The distributional properties of the previous patterns presumably
aid in the eventual formation of the more general semantic ones.
MacWhinney is not clear in how he sees the interaction of these two kinds of
patterns in acquisition, but the discussion suggests that free patterns are the
later development.
The last patterns are those which MacWhinney refers to as class-bound
positional patterns. These are word combinations which result from rules
which refer to grammatical features. Such rules take the distributional
properties of a word into consideration, and presumably can refer to aspects
such as ‘subject’. Being a functionalist in orientation, MacWhinney is not
overly enthusiastic about proposing that the child actually ever acquires
rules of this sort. The point appears to be that such rules are more complex
and should be acquired last, if such rules exist in language in the first
place.
MacWhinney’s presentation of these six kinds of pattern provides a
useful taxonomy for the discussion of possible procedures for the acqui-
sition of grammatical structure. As stated, the attempt to determine pat-
terns which go from specific to general learning are characteristic of distri-
butional models. Also, these patterns appear such, in that distributional
patterns are suggested as preceding more general semantic ones. This
approach differs from functionalism and semantic bootstrapping which see
the acquisition of the general semantic categories as preceding any distri-
butional knowledge. MacWhinney, however, does not elaborate on how
the sequence of patterns emerges or interacts. A more detailed attempt to
do this is the work of Maratsos & Chalkley (1980).
Maratsos & Chalkley (1980) This long paper has become the most com-
monly cited approach attempting to outline how distributional learning
can take the child from specific knowledge about individual lexical items to
broad grammatical categories. A summary of some of its points can be
found in Maratsos (1982). Its presentation here combines the views of
both of these papers on the acquisition of rules of word combination and
i __ of grammatical classes.
Like other recent studies, Maratsos & Chalkley (1980) present a theo-
retical model based on their interpretations of data collected by the earlier
studies. Much of the work deals with acquisition in older children, and
particularly the acquisition of gender in German. There is, however, a
reasonably detailed later section on early development. Maratsos &
Chalkley do not work from a current model of grammar, but they do have
definite ideas about the nature of the adult language. They see language as
consisting of classes that are defined in large part by their distributional
properties: ‘It is probable that what are called syntactic categories, such as
verb, adjective, or gender class, are so called because they rest on differ-
entiation among themselves by partly distributional distinctions among the
sets of correlated semantic-distributional patterns in which their members
appear’. For example, a verb such as ‘spill’ is not just defined as an ‘action’
word, but also as a word which takes the progressive ‘-ing’ suffix, the
present tense suffix ‘-s’, and the past tense suffix ‘-ed’. The child’s goal,
then, is the discovery of these distributional properties.
As with our discussion of functionalism, we will present what we see as
the steps in this model for the acquisition of grammatical knowledge (the
authors do not actually present numbered steps as here, although they do
talk about sequences of development).
Step 1: Identification of individual morphemes. Use phonological,
semantic, and lexical properties of sentences to identify the separate
morphemes in utterances. Little is said as to how this takes place but it is an
important step in all theories of acquisition.
Step 2: Formation of the first semantic-distributional patterns. The
initial patterns have a form:
Word A
This states that the child has noted some property for an individual word
(word A), and thus assigns this word to a preliminary category (cate-
gory X). The specification of the category as X is important because the
child’s categories are not predetermined by Universal Grammar to be
highly restricted at the onset to noun, verb, or some other universal
category. Depending on the number of properties noted, the child may
begin with several restricted patterns. The first properties of these early
categories are likely to be positional ones such as ‘occurs with “it”, or some
other word. Each category which occurs is noted as a separate pattern. The
first categories are restricted to individual words, that is, they are not general.
Step 3A: The strengthening of patterns. Patterns become strengthened
in two ways. One is through the identification in the input of known words in
their previously established semantic-distributional pattern. The other way
is through the addition of new words to a pattern when those words are seen
to have the same categorical property. This generalization can be repre-
sented as:
(7.16)
Pattern 1 (. . . Category X .. . )
Word A WordB WordC ... etc.
The idea of strengthening is an adaptation of an original idea in Braine

(1971a).
Step 3B: The establishment of pathways between patterns. We label this
-
step 3B because Maratsos & Chalkley claim that it occurs simultaneously
with step 3A. At this time individual lexical items are being identified as
belonging to more than just one category. The diagram in (7.17) shows the
development of pathways in this fashion:
(7.17)
Pattern 1 (. .. Category W .. . )
Word A
:
Word B
Pattern 3 (. . . Category Y .. . )
connections occur, as in step 3A.

Word C
Pattern 2 (. . . Category X . ..)
Word D .. . etc.
Pattern 4 (. . . Category Z . ..)

These interconnected pathways become strengthened as more lexical
Step 4A: Pattern generalization and overgeneralization. Once patterns

are sufficiently strengthened or established, they ‘continually seek for
occasions for application, depending on their own proportional frequency
of use’ (p. 193). One way which this occurs is in overgeneralizations such as
saying ‘knowed’ instead of ‘know’. Suppose that the English past tense
morpheme defines category Win (7.17). A new word such as ‘know’ may be
first identified with categories X, Y, and Z , and thus enter into this network
of connections. Once so identified, ‘know’ can then be connected with
category W without previous experience with its having this property. An
important part of this step is determining how many connections are
necessary for a new word to become part of a network of connections.
Maratsos & Chalkley believe that this eventually will happen even after just
one connection, in large part due to the extensive redundancy that exists in
such a system. Three factors interact to determine how words are connected
to new patterns. These are (i) frequency - how often the established words
in a network, e.g. words A to D in (7.17), are used in all the categories;
(ii) variety -how many different pathways are shared by the categories; and
(iii) directness - the extent to which categories are connected through the
same lexical items.
Step 4B: The inhibition of overgeneralizations. While the model will
account for a large number of generalizations of words to new patterns, it
also predicts a large number of overgeneralizations which are not part of the
adult grammar. For example, noting that both ‘spill’ and ‘call up’ are actions
could lead the child to connect ‘call up’ to the past tense category and say
‘call upped’. Simultaneously with the generalization of patterns are restric-
tions or inhibitions. One factor which will restrict generalization is accuracy.
This is the extent to which a pattern which is established for a new word has
direct connections to another pattern. This is shown in (7.18):
(7.18)
WordA WordB WordC WordD WordE WordF New Word
Pattern 2 (. . . Category Y .. .)
Here all the words, including the new one, have been connected with
pattern 2. There are only three direct connections, however, between
patterns 2 and 1(words A , B, and C). The connection would not be accurate
enough (by measures to be determined) for the new word to overgeneralize
to pattern 1. The accuracy requirement predicts that certain overgenerali-
zations should never occur.
Step 5 : The correction of overgeneralizations. Accuracy is not suffi-
cient to prevent overgeneralizations from occurring occasionally: examples
are well-documented in the child language literature, e.g. ‘foots’, ‘knowed’,
etc . . . The child needs to correct those which occur. Maratsos & Chalkley
propose that this is done through indirect negative evidence. That is, the
child keeps track of the number of times he or she does or doesn’t hear a
word appearing with a particular grammatical property. For example, once
the child never hears ‘know’ occurring in the past with ‘-ed’, he or she
records as part of the lexical information for this morpheme that it is
restricted from doing so. They refer to these as positive inhibitory blocks.
Maratsos (1982) provides an example of how the child follows these steps
through his early word combinations. Using a slightly modified termino-
logy, Maratsos begins his example with the child acquiring the sentence
‘daddy walk’ with the meaning ‘that daddy moves himself along by
movement of the lower limbs’ (p. 256). This would be step 2, when the child
forms a restricted semantic-distributional pattern. The pattern consists of
two ‘sequential loci or slots’ (p. 256), with each slot defined by the
properties of the terms which can occur within the slot. (7.19) provides the
structure of this initial pattern:
(7.19) X + Y
I I
daddy walk
This is somewhat different from the discussion in Maratsos & Chalkley
where one of the morphemes would be the category (e.g. category X), and
the other would be its defining property. Here the defining property is also
assigned categorical status.
Step 3A occurs when the child next acquires ‘mommy walk’. The words
‘mommy’ and ‘daddy’ become members of the emerging category of things
that can walk. With the addition of other words, it can become ‘an
individual word formula’ (p. 257) with the structure of ‘walker+walk’. the
Y class can also expand in this manner, so that the eventual pattern may
have the meaning of ‘mover+movement’. Maratsos points out that the child
will be restricted in building a broader category from comparisons of ‘daddy
walk’ and ‘daddy chair’ with a possessive meaning. Maratsos states (p. 258):
‘The scheme thus builds grammatical categories on the basis of overall
sequential similarity, but also on the basis of semantic function and analysis
of parts of the sequence. It is hence not a purely distributional analysis.’ No
details are given, however, on the semantic properties that the child will use
to restrict such broader categories, and the example is not carried on
through any further developments.
Assessment. The work by Maratsos & Chalkley is a relatively detailed
attempt to follow acquisition after the early restricted formulae proposed by
Braine (1976) and to see how the child could develop adult categories.
Interestingly, it led them to a position that the adult grammar may consist of
complex networks of semantic-distributional patterns, with general terms
such as noun, verb, and subject being unnecessary: it is an ultimate example
of using child language data and something like the Constructionist
Assumption to develop a theory about the adult language.
Much of the demonstration of how such learning proceeds was done with
examples of the acquisition of grammatical morphemes to define grammati-
cal classes. For example, we discussed the proposed steps with examples of
the acquisition of categories such as words which take ‘-ed’, ‘-ing’, etc. (Le.
potential verbs). There is relatively little discussion on how general senten-
tial relations such as ‘subject-predicate’ (or the equivalent in their theory)
are acquired in relation to the acquisition of grammatical classes. For
example, do relations like agent-action precede, follow, or occur simul-
taneously and independently with emerging patterns like ‘words which take
“-ed”’? This is an important issue, as we will see in our discussion of
semantic bootstrapping, where semantic relations are claimed to precede
distributional learning. Maratsos (1982: 263) points out this problem,
among others, for his theory at this point in its development. He states
(p. 265): ‘These problems indicate that considerable empirical and theoreti-
cal analysis is required before we can be said to have any good idea of
plausible complete accounts of formal category formation.’
Pinker (1984) presents two arguments against the semantic-distributional
model of Maratsos & Chalkley. The first is what he refers to as the
learnability argument. This is the claim that the child in this model will
require some form of negative evidence in order to acquire the adult
language. We have already discussed the problems with negative evidence
in Chapter 2, and how Maratsos & Chalkley propose it in Step 5. The
problem of how children acquire language without negative evidence is a
serious one for a number of approaches in both grammatical theory and
acquisition (see Lightfoot 1982: Chapter 2). Pinker’s second argument is
the eficiency argument. Without better-defined measures of how patterns
correlate with each other, the number of potential patterns is enormous.
The theory as it now stands requires a tremendous amount of computation
and thus memory for it to take place. Pinker’s own solution to these
problems, as well as many others, is to restrict the range of possible
solutions in language acquisition to a limited number which are innately
endowed. The next section examines the views of Pinker, as well as of
others who assign a richer syntactic structure to the young child.
7.4.3 Semantic bootstrapping

At first glance, the term semantic bootstrapping looks like part of an
acquisition theory which evolved out of the research on semantically
oriented grammars. As we will see, however, this is far from the case. The
latter work has attempted to develop theories in which the child acquires a
set of semantic relations and functions, not a formal syntax. Semantic
bootstrapping, however, is an acquisition procedure needed for theories
which want to show the acquisition of abstract grammatical categories and
relations. The acquisition problem for any theory of abstract syntax is how
the child determines the language-specific features of the language being
acquired. For example, how do children come to know what are the subjects
and VPs in sentences? The answer to this dilemma for the preceding
approaches which believe in abstract syntax is semantic bootstrapping. The
formal grammatical properties are triggered by the appearance of certain
semantic properties. Using terminology from computer technology, the
claim is that semantics ‘boots’ the child’s syntax into operation.
Until the term semantic bootstrapping was created, this process was more
assumed than expressed. For example, McNeill’s model of acquisition
requires some way for the child to form categories like subject NP and VP
from the early meanings of the child’s utterances. The process of semantic
bootstrapping is also implied in the work by Bloom, Lightbown & Hood
(1975) where the category ‘subject’ emerges through the gradual broaden-
ing of the first subjects to more and more semantic categories. For this to
316 T H E P E R I O D OF T H E FIRST W O R D COMBINATIONS
occur, subject would have to be ‘triggered’ by the first one or two semantic
relations acquired.
The most explicit statement of semantic bootstrapping in the older
literature is in Macnamara (1972). Macnamara begins with the assumption
that the child first acquires the meanings of individual words, much like
step 1 in the discussion of Maratsos & Chalkley. This happens as the child
seeks word forms to represent semantic notions like specific and general
nominals. Then, the child develops combinations of meanings or semantic
relations of the kind which we have considered at various points in this
section. The child uses these meanings to acquire the syntax of the
language. It is at this point that Macnamara makes a somewhat vague
suggestion which could be interpreted as semantic bootstrapping. He states
(p. 395) that the child ‘. . .might, for example, suppose that a single
semantic relationship would always be expressed by means of a single
syntactic device or structure’. The crucial question here is whether Macna-
mara wants to claim that all children are innately endowed to map the same
single semantic relations into the same single or limited set of syntactic
devices. For example, do certain semantic relations like ‘Action’ always
map into a predetermined syntactic category such as VP? This point is never
clarified or elaborated upon.
It is only in the most recent literature that the claims of semantic
bootstrapping have been made more explicit, and even these attempts have
not been widespread. Macnamara (1982), among many things, expands his
original ideas on how semantics leads to syntax, focussing on the develop-
ment of categories like noun and verb from semantic notions like object and
action. Some preliminary ideas on how syntactic structures follow from
semantic ones can be found in Grimshaw (1981). Marantz (1982,1983) has
discussed two routes in which semantic bootstrapping could logically
proceed. Here, however, we will concentrate on what is clearly the most
detailed effort to develop a model along these lines, that of Pinker (1984).
Pinker (1984) This is a major work which examines the acquisition of a

wide range of grammatical phenomena including inflections, complemen-
tation, auxiliaries, and lexical entries. We will limit our treatment to his
explanation of the first grammatical acquisition of the child, the acquisition
of phrase structure rules for Stage I.
Pinker begins with laying out his assumptions about acquisition, the first
of which is the learnability condition. Pinker states (p. 5): ‘. . . we may view
the child’s abilities at any stage of development as a way-station in a process
that takes the child from a state at which he or she knows nothing about the
target language to a state at which he or she has acquired the language
completely’. The learnability condition has two parts. First, acquisition will
need to occur through the interaction of an acquisition mechanism which
begins with no knowledge about the target language and obtains this
knowledge from the linguistic input. This aspect of the condition takes the
position that acquisition must be accounted for by only positive evidence.
Second, the acquisition mechanism will need to be able to arrive at the adult
grammar through its exposure to the input. This condition expresses the
concern of recent approaches to take the child beyond the apparently
idiosyncratic features of the early word combinations to the adult grammar.
While the learnability condition requires continuity in development, it
still could allow for some restructuring during acquisition. The second
condition, however, the continuity condition,restricts this greatly. It states
that the most highly valued or explanatory theory will be the one which
posits the least number of mechanisms to get the child from the first stage of
acquisition to the last. ‘As Macnamara (1982) has put it, the null hypothesis
in developmental psychology is that the cognitive mechanisms of children
and adults are identical; hence it is a hypothesis that should not be rejected
until the data leave us no other choice’ (Pinker 1984: 7). Importantly, the
condition does not eliminate the possibility that the child may have a
grammar at some point which will be different from the adult one being
acquired. This would be the marked case, however, requiring the investi-
gator to propose a mechanism that manages to get the child back to the
adult grammar. In practice, however, Pinker does not explore this possi-
bility. Instead, he assumes that ‘the child’s grammatical rules should be
drawn from the same basic rule types, and be composed of primitive
symbols from the same class, as the grammatical rules attributed to adults in
standard linguistic investigations’ (p. 7).
These two conditions combine to define a theoretical approach which is
similar to what was called in Chapter 4 the strong inclusion hypothesis. The
child’s grammar from the onset looks very similar to, and is a subpart of, the
adult one. The conditions are also stronger than the Constructionist
Assumption, which resembles the learnability condition. We did not make
any assumptions at the onset, however, to restrict the extent of restructur-
ing to the degree that the continuity condition does.
As is characteristic of maturational theories like Pinker’s and others to be
discussed, Pinker selects a current grammatical theory to describe the
grammar that the child is acquiring. The theory selected is that of Lexical
Functional Grammar (or LFG as it is commonly called), as developed in
Bresnan (1978, 1982). One main feature of this theory is that it does not
map deep structures onto surface structures via transformations as does
Chomsky’s Standard Theory (Chomsky 1965). Instead, each sentence has
only one constituent structure which is generated by the phrase structure
rules. To use the terminology of the Standard Theory, the surface structure
is the deep structure. In such a theory, the notion of level of represen-
tation is meaningless. The phrase structure configurations are similar to
those of the Standard Theory, except that grammatical functions like
subject are indicated on categories like NP and VP. That is, the subject is
not just the NP under S as in the Standard Theory, but it is indicated as
NP,,bj. A second feature of the theory is that there is also a separate
functional structure which combines the functional information which is
indicated on the grammatical categories of the phrase structure with the
information in the individual lexical items in the sentence. The role of the
lexicon is quite substantial in this theory, with the lexicon accounting for
many of the grammatical aspects previously handled by transformations.
For example, there is no longer a Passive transformation, but instead
separate lexical entries for forms of verbs which can occur in passive
sentences.
One appeal of LFG (at least to some people) is its psychological
plausibility. Acquisition will consist primarily of establishing the phrase
structure of sentences and the lexical features of words. Since the only
structure to be acquired is that of the surface structure, the theory does
not present the problem of how the child reconstructs a deep structure
from a surface structure and acquires the complex set of trans-
formations that operate between the two. Also, it places most of acqui-
sition on acquiring information about words. Lexically oriented theories
are relatively better at dealing with individual variation and the specific
kinds of learning observed in Braine (1976). While children probably
hear relatively similar distributions of construction types, their experi-
ence with individual words probably varies more. While Pinker selects
LFG as his grammatical theory, he points out that the general features
of his acquisition model are applicable to other grammatical theories as
well.
In general terms, acquisition begins with semantic categories and rela-
tions. Semantic bootstrapping can be defined as follows (p. 40): ‘The
categorization of words can be inferred from their semantic properties, and
their grammatical relations can be inferred from the semantic relations in
the event witnessed.’ Pinker’s Table 2.1 gives a tentative list of some of the
syntactic categories that are booted in this way. A selection of these is given
in Table 7.21. These correspondences are innate in the child: that is, once
the child acquires the relevant semantic notion, then the syntactic property
appears. For example, the young child, possibly in the holophrastic stage,
who says ‘daddy’ as a specific nominal will automatically assign the feature
[+noun] to the word. This is true for all of the correspondences in
Table 7.21. The child’s semantic notion will also lead to phrase structure
information. The example under configurations claims that once the child
Table 7.21 Some syntactic properties and the semantic knowledge which
leads to their inference, selected from Pinker (1984: Table 2.1)
Syntactic properties Semantic basis categories
Categories
noun Name of person or thing
verb Action or change of state
adjective Attribute
preposition Spatial relation, path, or direction
sentence Main proposition
Functions
subject Agent of action; cause of causal event; subject of an attribution of
location, state, or circumstance
object Patient or Theme
oblique Source, Goal, Location, Instrument
Cases
nominative Agent of transitive action
accusative Patient of transitive action
dative Goal or Beneficiary
Features
tense Relative times of event, speech act, and reference point
aspect Durativity
Configurations
sister of X Argument of X
relates two semantic notions such that one is an argument of the other, the
child places them into a phrase structure configuration. For example,
Agents are arguments of Actions. The corresponding syntactic categories,
noun and verb, will therefore be sisters, i.e. attached to the node, in this
case S. If this looks familiar it should be; we saw a similar claim in McNeill
(1970b), exemplified in more detail in (7.9).
Up to this point, the approach looks very much like a combination of
the insights on semantic bootstrapping of Macnamara (1972) and the
claims about syntactic structure of McNeill (1970b), with supporting
evidence in the data in Bloom, Lightbown & Hood (1975). Pinker goes
on, however, to address the crucial issue of how the child acquires the
particular grammatical properties in languages. Different types of lan-
guages are discussed, but we will limit the examples to two problems in
acquiring English.
Two potential problems for the English-learning child are predicate
adjectives and oblique preposition phrases. (7.20) gives examples of each
with a simple phrase structure configuration of the adult structure for the
italicized part of each sentence:
(7.20) a. Predicate adjective b. Oblique prepositional phrase
‘The dog is big’ ‘I went to the store’
Thedog V AP went P NP
I I i n
is A to the store
I
big
The problem pointed out by Pinker is how to describe the grammar of the
child who says ‘dog big’ and ‘go store’ with the same intended meaning as
the adult sentences. Since the child does not have either the copula ‘is’ or
the preposition ‘to’, there is no evidence for the adult structures in (7.20).
Based on the child’s sentences, the child could just as well have the
structures in (7.21):
(7.21)
NP AP
b.
A
I I
The sentence ‘go store’, for example, would be similar to the adult phrase
‘go home’ which would have the structure of (7.21b). We also need to
account for how the child gets from categories like noun to phrases like
noun phrase (NP).
Pinker proposes a series of procedures that take the child from the initial
acquisition of semantics to syntactic structures. We will demonstrate these
in the same way as earlier, by defining a series of steps. (Pinker does not
provide labeled steps, but labels them as a series of procedures - P1 through
P6. We will give his procedures in the appropriate steps.) These will be
demonstrated by following through the acquisition of the structures in
(7.20) by a child who has only ‘dog big’ and ‘go store’ at the onset.
Step 1: Acquisition of basic semantic notions. This will be a subset of
those shown in Table 7.21. We assume in our example that the child has
acquired Proposition, Name of person or thing, Action, Attribute, Agent,
Theme (the future subject of the sentence ‘dog big’), Goal, and Argument
of x.
Step 2: Semantic bootstrapping. The child assigns syntactic properties
as shown in Table 7.21. The child has thus also acquired the syntactic
properties of sentence, noun, verb, adjective, subject, object, oblique, and
sister of X. These properties will allow the child to process a number of
simple sentences as determined by the following procedures.
Step 3: Application of existing phrase structure rules ( P I ) . Trees for
sentences are understood and produced by the syntactic properties avail-
able from step 2. Presumably these may include rules for word order. The
sentence ‘dog go’, for example, will have the structure in (7.22a):
(7.22) a. N V b. S C. S
I I I I
dog go NP VP VP
I I I
N V N V
dog go I I
dog go
Step 4: Apply X-bar theory (P2). X-bar theory refers to work since
Jackendoff (1977) to impose general conditions on grammatical categories.
Simply stated, it proposes that all categories share the same phrase structure
properties. For example, all Xs (where X=category) belong to a higher
category X-bar (usually shown as x). The X-bar for lexical categories like N
and V are phrases such as NP and VP. Within this theory, the higher
category of a VP is S. This step imposes this theory onto the child’s grammar
so that phrases are generated over the child’s lexical categories in the tree.
The results of this step for our example are shown in (7.22b). (See Radford
1981 for an introduction to X-bar theory.) While Pinker proposes this step,
he does allow for the possibility of restricting it to apply only if there is
evidence for the proposed phrases in the data (p. 108). If we apply this
adjusted P2, then our child would still only have the structure in (7.22a) for
NP since there is no evidence for the phrase, i.e. there are no adjective-
noun or determiner-noun constructions. There is evidence, however, for
the VP in that the child has the relation Object. Let’s assume that such is the
case and that she can produce sentences such as ‘eat apple’. Also, we have
evidence for S since the child has the notion of Proposition. This would then
give us the structure in (7.22~).
Step 5: The assignment of grammatical functions (P3). Functions like
subject and object are added to the syntactic categories in the tree, again
based on the correlations from Table 7.21. These functions are used in the
next step to connect categories to their appropriate nodes, for example
subjects to the S node, and objects to the VP node. The application of this
step to our example results in (7.23a):
(7.23) a. b.
S
- S
I /1'P
VP / \
I / I
Nwhl V N,",, V
I I I I
dog go dog go
Step 6: Connection of unattached branches (P4). Connect all unat-
tached branches to their appropriate categories based on their configura-
tions. Subjects, for example, will attach to S. This gives us (7.23b).
It is at this point that our child runs into a problem with its sentences 'dog
big' and 'go store'. After step 5 , they will have the structures in (7.24a) and
(7.24b) respectively (ignoring the broken lines for now). (Pinker represents
the critical nodes as AP and NP but the restricted form of P2 should leave
these as A and N in our example. We will use A and N, but the arguments
equally hold for AP and NP.) The problem is that the child does have more
than one option available for the attachment of the A and Nobliquesince
neither the copula nor the preposition are yet acquired to guide their
attachment. The A could attach to either S or VP, and the Nobliqueto either
VP or a newly created PP. Pinker refers to categories in this situation as
orphans. He does not wish to create lexically unfilled categories, for two
reasons: such categories would not be substantiated by the child's data (a
violation of the Competence Assumption), and, more seriously, they might
lead to the creation of categories which do not exist in the language.
Pinker's solution to this problem is to give orphans special marking (in our
example ?A? and ?N?obliquerespectively), and assign them temporarily to
an ancestor node. These connections are shown with the broken lines in
(7.24). We will return to the question of how orphans are assigned
eventually to the correct English nodes.
(7.24) a. b.1
S S
I
Nwhj
?A? VP
I I 1.
dog big
go store
Step 7: Creation and strengthening of phrase structure rules

(PS). Create phrase structure rules that will produce the tree which has
been built. If rules already exist, then strengthen them. This latter process is
similar to rule 3A in Maratsos & Chalkley. In our example, this step leads to
the following rules:
These rules will account for the four sentences that we have used so far: ‘dog
go’, ‘eat apple’, ‘dog big’ and ‘go store’. As can be seen, the orphans are
marked differently from the other categories.
Step 8: Lexical storage and strengthening ( L l ) . Add new words to the
lexicon with information about their grammatical properties. Verbs, for
example, will be entered with information about the complements which
they may take. Entries which already exist are strengthened.
Step 9:The collapsing of rules (P6). This procedure operates when a
new rule is added to the phrase structure which already has a rule which
expands the same category. For example, if the new rule is as in (7.25a) and
the old rule is (7.25b), then this procedure will collapse them to (7.26~).
Pinker provides a number of proposals on how collapsing takes place.
(7.26) a. VP- PP
b. VP- NP
We can now return to a discussion of how the child will adjust the status of
the orphans and reassign the orphans to their appropriate place in the tree.
Using the acquisition of prepositions as our example, it will happen as
follows. First, let’s assume that our young language learner acquires the
sentence ‘go to store’. In step 2, he or she identifies ‘to’ as a preposition
according to its semantic properties listed in Table 7.21. Step 3 (or P1)
allows the child to make the tree in (7.27a), since it now has the rules in
(7.25). Step 4 (or P2) then creates a PP (prepositional phrase node) above
the P, as in (7.27b). Such an expansion is allowed on the grounds that there
is a constituent present, i.e. ‘store’, which can function as the object of the
preposition. This structure creates a problem, however, when we come to
step 6 (or P4) and have to attach the PP to the VP. Since the ?N? is already
attached to VP, it needs to be detached and placed within the PP. Pinker
adds a condition to P4 (p. 104) that orphan nodes are replaced by a new
node (in our case PP) when the new node has the same structural relation to
the ancestor node (in our case VP) as the orphan. This then results in
(7.27~).
(7.27) a. b. C.
S S S
I I I
I '. \
VP
p. 1
VP
r\
v
I I
p
go to store
?~?,>l,Ilq,,<
I
go
v
I
pp
l
P
?N?,,,,,,,,,,,
l
store
YK
go p N<,l,,,<,,,,
I
to
I store
to
I
In my introduction to Pinker's assumption about acquisition, I reviewed
his continuity condition which assumes that child and adult grammars are
the same unless evidence shows otherwise. The orphans are such a
situation, since the child needs to restructure its grammar once the
prepositions are acquired. Such a restructuring would not be allowed under
our Constructionist Assumption, which requires that rules be retained once
they are formed. In this case we could claim that the rule in (7.25b) is
retained for oblique NPs, and used for phrases such as 'go home'. Pinker
considers, and rejects, the possibility that the child adds a new rule and
maintains the old one. Instead, he argues that the child is constrained by a
uniqueness principle from maintaining two rules to produce the same
function. This, admittedly vague, principle claims that the child, when
presented with alternative rules to represent the same function, will only
use the one for which positive evidence exists. In his account, rule (7.25b) is
deleted from the grammar, replaced with the one created from the structure
in (7.27~)by P5. The uniqueness principle, an innate property of the
language acquisition device, forces the child to do this. The new rules are
shown in (7.28). P6 will collapse the new rule with the old rule which
generates object NPs as in (7.28a).
b. PP- P NP
The initial semantic bootstrapping is followed by what Pinker calls
structure-dependent distributional learning. This is needed for acquisition of
the aspects of the language which are not semantically transparent. For
example, not all nouns are objects ('flight'), and not all subjects are agents
('John received a parcel'). Grammatical morphemes, like the articles and
verb inflections, can in turn be used by the child to identify abstract nouns
and verbs. This occurs through a procedure referred to as D1. Here we will
label this step 10:
Table 7.22 Pinker’s LFG grammar for late Stage I, from Pinker (1984:
11), with examples of possible utterances
Grammar Examples
‘mommy fix’
‘tree there’
-S NPsubj
‘pillow dirty’
?NP? ‘that man’
‘mommy soup’
‘pretty boat’
‘more tree’
‘put book here’
‘give food doggie’
‘put truck window’
‘take baby downstairs’
Step 10: Distributional learning. If a new word or phrase has the same
distributional properties as an acquired word or phrase, assign the new
word or phrase to the category of the acquired one. Pinker is purposely
vague on when this operation begins. He says at one point that it could
begin as soon as the second sentence heard (p. 53). More likely, however,
‘one expects the accretion of rules (and possibly the strengthening of rules)
to be slow enough that large amounts of structure-dependent distributional
learning do not follow the very first acquisitions’ (p. 56). It depends, in large
part, on the acquisition of grammatical morphemes, yet, as we will see,
Pinker does not see these as being part of the Stage I grammar. For this
reason we have placed this learning as the last step in Pinker’s sequence of
mechanisms.
These procedures, Pinker argues, will lead the child to the basic phrase
structure rules of the language. However, there are questions which arise
concerning how they operate. For example, it is not clear from the
presentation how the S is created for the sentence ‘dog big’ in (7.24), since
the S results from the operation of X-bar theory on V, not A. Also, it is not
clear how the child is going to distinguish sentences with verb particles from
those with prepositions such as ‘pick up ball’ vs. ‘sit on chair’. Other
questions arise as other examples are tried out (see Pye 1985). The
important point, however, is that Pinker has presented the most detailed
attempt yet to account for such facts, and some adjustments and clarifi-
cations may reasonably be expected.
Pinker uses this approach to write an LFG grammar for late Stage I or
early Stage 11, using the data cited by Brown (1973) in his discussion of
Stage I speech for Adam, Eve, and Sarah. Table 7.22 presents this
grammar along with some examples of the constructions it generates. The
grammar is a modified one which attempts not to assign structure to the

child unless there is evidence for it in children’s Stage I utterances. The
examples are ones that we have either taken from the text or made up, since
Pinker does not go through the grammar with examples.
We can compare this grammar to those proposed by McNeill, given in
(7.10) and by Bloom for Kathryn, in (7.12). Both McNeill’s and Pinker’s
grammars are attempts to characterize late Stage I in general, and both
were based in large part on data from Adam. While less detailed than
Pinker’s, McNeill’s grammar is the most powerful of the three, generating
sentences with several terms. Pinker’s and Bloom’s grammars share an
attempt to implement something like the Competence Assumption. Bloom
restricts her grammar by avoiding subject NPs, using ‘Nom’ instead. Pinker
restricts his by including orphan nodes and avoiding phrases unless there is
evidence for them. Bloom’s grammar generates a smaller range of struc-
tures than does Pinker’s, but this is not surprising since she is writing it for
just one subject. Overall, however, the grammars share some general
features.
Pinker’s discussion of his grammar consists of justification for five of its
features: (i) order and composition in phrase structure rules, (ii) identity of
categories across positions, (iii) use of syntactic categories, (iv) limitations
on the length of sentences, and (v) the use of universal symbols. Since these
are the major issues involved in writing grammars for young children, we
will look at a defense of each.
Order and composition refers to the issue of how the linear order and I
composition of constituents is acquired by the child. For example, one of
the rules in Pinker’s grammar is VP- V+NP. This rule combines two
aspects of the child grammar: that the verb precedes the NP (Le. linear
order), and that a verb phrase consists of a verb and a noun phrase (i.e.
composition). Such a rule assumes that children acquire these two aspects
together, but it is also possible that they are acquired separately; for
example, the child could identify the composition of phrases, but not the
linear order of their constituents.
Pinker points out that these two possibilities lead to different predictions
about the acquisition of languages with strict or with free word order.
Pinker’s position that linear order and composition are acquired simultane-
ously predicts that children should have little problem in acquiring word
order in languages (such as English) which have strict word order. It also
predicts that children acquiring free word order languages should show a
tendency to impose an order on them. The alternative theory predicts a very
different pattern in the acquisition of these two kinds of languages. If order
is not acquired until later, then children acquiring strict word order
languages should show a variety of orders, particularly ones that violate the
allowed sequences in the language. On the other hand, free word order
languages should be no problem.
The currently available data by and large support Pinker’s position.
Although there are sporadic reports of children using free word order in the
acquisition of English, most analyses show a high degree of accuracy.
Pinker cites Brown’s (1973) analysis in particular, which indicated very few
violations of appropriate usage. Regarding the acquisition of free word
order languages, Pinker cites Bowerman’s (1973b) analysis of Finnish, Luo,
and Samoan which revealed that children tended to use one order in
particular, despite having a range of options. Pinker concludes that amal-
gamating word order and constituent combinations in one rule reflects the
acquisition procedures used by children.
Another issue addressed by Pinker is the identity of categories across
positions. Recall in Bloom’s grammar for Kathryn that she used NP for
objects but Nom for subjects. This was because the only subjects which
Kathryn used were single nouns or demonstrative pronouns. Pinker,
however, has an NP for both positions. The major argument for this is a
logical one. If Universal Grammar specifies that there are a small number of
general categories, and that they are triggered semantically, then the
categories should exist once the trigger experience has taken place. Even
though Pinker allows for a restricted P4 to limit the use of phrasal categories
in the grammar before there is evidence for them, he feels that they need to
be represented on logical grounds once the evidence exists for the category
in any position.
The evidence for the claim that children do not have subject NPs is that
most of the subjects in early sentences are either nouns or pronouns, not
phrasal structures such as article+noun or adjective+noun. Also, the nouns
that occur in subject position tend to be inanimates. Pinker’s counter-
argument to these facts is that they are only tendencies. First, animates tend
to be subjects because most of the child’s early verbs are actions by people
upon objects. When data are observed more carefully, there are usually
examples of inanimate subjects, and nouns which can be found as both
subjects and objects. This is also true for the claim that expanded NPs never
occur in subject position. He cites examples from Adam such as ’Dale
Panda march’ and ‘Dale shoe right there’ which have subject NPs. This
uneven distribution of type frequencies, then, is accounted for by pragmatic
factors in speaking. A similar argument can be found in Limber (1976).
A third aspect of Pinker’s grammar is the use ofsyntactic categories and
rules instead of semantic and lexical ones. Pinker acknowledges that some
of the main kinds of evidence to prove the existence of syntactic categories
are not available yet at Stage I. Subject-verb agreement, for example,
would be one form of evidence. One could postulate syntactic categories if
the child showed that verbs agreed with the number of the subjects,
regardless of their semantic roles. Conversely, semantic theories should
predict that the first syntactic rules such as verb agreement should be
semantically restricted. Since such forms begin to be acquired during
Stage 11, their study could be useful in resolving this issue. Definitive results
on this, however, have not yet yet appeared. Pinker cites research in
progress which indicates that the first verb inflections occur as often for
action verbs as for non-action verbs. On the other hand, in a study not
mentioned by Pinker, Antinucci & Miller (1976) have claimed that the first
past tense affixes are only used on verbs which mark abrupt change.
One way that this issue can be addressed in Stage I data is to look at the
interaction of NP expansions with semantic categories. If the child’s
categories were syntactic, one would expect that expanded NPs would occur
across different semantic roles (Pinker refers to such a spread as
‘promiscuity’). His analysis of Adam’s speech showed such promiscuity. For
example, possessive+noun expansions occurred in at least four semantic
roles: patient of action ‘pull Dale bike’, object of experience ‘see Daddy
car’, predicate nominal ‘dat Dale baby’, and locative ‘spill Mommy face’.
In discussing this issue, Pinker responds to the argument given by Braine
against syntactic categories. I gave this argument during the presentation on
Braine (1976) in 7.3.4. To reiterate, Braine presented data from Kendall
and Seppo which showed an ordered rule for Agent+Action, but an
unordered rule or grouping pattern for Action+Locative. He used these
data to argue that the children therefore couldn’t have a rule of VP-NP or
PP (Locative) because such a rule would require agents to precede actions.
Pinker’s response is that his rules do not produce this effect. The locative in
this example would be an orphan, as in (7.24b), and produced by the rule
k ? N ? o b l i q u eThat
. is, the locative would not be generated in the VP at
that point in development.
Pinker also provides a general discussion of his view on the work of
Braine, MacWhinney, and Maratsos. He states (p. 146): ‘Like the logician
who saw a black sheep through a train window on his first trip to Scotland
and would conclude nothing more than that at least one sheep in Scotland
was black on at least one side, these theorists have offered a hypothesis that
is difficult to refute, given available data.’ He feels that his account
describes the available data as well as theirs, and has the advantage of
meeting the continuity condition presented at the onset. He also points out
that these theories are difficult to attack because they do not present a
detailed learning theory of the kind he has attempted (p. 151). Curiously,
he does not mention the one older study which tried to argue for syntactic
categories, that of Bloom, Lightbown & Hood (1975).
Pinker’s grammar for Stage I is relatively powerful, not only because it
generates a subject NP but also because it is shown as obligatory. The
grammar then produces a number of constructions that are longer than
those of the spontaneous sentences used in the analysis to construct the
grammar. Pinker therefore has to account for the limitations on the length of
the child’s sentences. Four hypotheses are considered to account for this.
The first is the solution used by Bloom (1970), that the child has deletion
rules which eliminate constituents. This is rejected for four reasons: (i) such
rules usually ‘do not account for the full set of apparent deletions in the data
(Braine 1976)’ (p. 156); (ii) such rules violate the formal conditions on
deletion rules in transformational grammar (Chomsky 1965); (iii) since
adult speech does not have such deletions, there is no positive evidence for
the child to use from the input to form such rules; and (iv) it requires that
the child unlearn rules to acquire the adult grammar. The second hypothesis
is that the child’s rules are incomplete. This is essentially the solution taken
by Brown (1973) in constructing his set of semantic relations (see
Table 7.16). Separate relations were proposed like Agent+ Action,
Agent+ Action, and Action+Object rather than one relation of Agent
+Action+Object. Pinker argues that to have separate abbreviated rules
misses the point that the children show these constructions in appropriate
orders, and will even produce sequences where the missing elements are
inserted. The third hypothesis is that the constituents in the rules are
optional. This was McNeill’s choice in his grammar. Pinker rejects this on
two grounds: (i) it has all the same problems as the deletion rules hypo-
thesis; (ii) if it were true, one would expect at least some sentences which
show all of the constituents generated, but such sentences usually don’t
appear.
Pinker’s solution is the last hypothesis that he considers, which is that
there are processing limitations. As is typical of such proposals, there is no
real psychological model of what the limitation is. Pinker speculates that
there may be limitations of mapping from functional structure to consti-
tuent structure. No independent evidence is presented, however, to support
that speculation.
The last issue that Pinker addresses is the use of universal symbols. We
have already treated Bowerman’s (1973a,b) arguments against the proposal
of a VP node for her subjects. Pinker accepts her arguments, and acknowl-
edges that ‘the sort of evidence necessary even for making tentative
hypotheses about a VP node for children, or other detailed aspects of their
tree geometry, does not currently exist’ (p. 162). He falls back on the logical
argument that since they are part of adult grammar, then the child must
have them innately (p. 165): ‘The use of universal adult symbols in
children’s first rules, then, is simply the theoretical commitment that the
child’s early rules are continuous with their later ones and with adult ones.’
Assessment. Pinker’s work shows a clear historical connection with
earlier work. The claim that the child begins with a syntactic grammar which
contains adult-like categories is very similar to the position taken by
McNeill (1966a, 1970b). It is also similar to McNeill in postulating that the
child’s lexicon plays a major role in the process. The attempt to tie the onset
of the syntactic development to semantics is an extension on the original
ideas in Macnamara (1972). Also, the argument that evidence can be found
in children’s speech for the existence of grammatical relations has its roots
in the work of Bloom, Lightbown & Hood (1975).
The originality in Pinker’s work is in the detailed speculation on the steps
that the child actually follows in constructing its grammar. It is also a
forceful demonstration of the kind of argumentation that needs to be done
to advance our understanding in this area of study. He provides a wide
range of arguments that can be the basis for future research.
While rich in theoretical ideas and argumentation, it has the same
weakness as in virtually all recent research. Unlike the work prior to 1976,
it does not provide detailed analyses of child data to support its claims.
Often the examples to support an argument are from different children,
and often from children at different points of acquisition. The arguments
also seem to be more appropriate for some data than for others. For
example, semantic bootstrapping appears to be a much more suitable
account of the acquisition of Kathryn and Gia in Bloom, Lightbown &
Hood (1975), than for Eric and Peter. (See Ramer 1976 and Ingram 1981d
for more data on these two patterns of acquisition.) The next step is the
collection and analysis of data through a new methodology that combines
the theoretical advances of recent years with the interest in data of the
earlier studies. An outline of what that methodology may look like is
presented in section 7.5.
7.4.4 Parameter setting

As pointed out in Chapter 4, a major feature of language acquisition studies
is the use of a theory of adult grammar for making predictions about child
data. Most work done of this kind, however, has been either on acquisition
in older children, e.g. papers in Tavakolian (1981a), or have not really used
child data at all. The latter studies are those on learnability which restrict
themselves to the logical problems of acquisition, e.g. Wexler & Culicover
(1980) and Culicover & Wilkins (1984). Pinker’s, then, is one of the few
studies to apply a current theory to early word combinations. Even so, it is
restricted to one particular theory, that of Lexical Functional Grammar.
Some other current theories, for example Relational Grammar (Perlmutter
1983) and Generalized Phrase Structure Grammar (Gazdar, Klein, Pullum
& Sag 1985), have produced little in the way of proposals for theories of
language acquisition. One current theory that is beginning to be applied
in this way is the Government and Binding (GB) Theory of Chomsky
(1981).
Until recent years, much of the work which could be characterized as
Language Acquisition had been working from the universal base hypo-
thesis. This is the claim that all languages share a single underlying ‘base’, or
phrase structure grammar. This was the assumption in McNeill’s work, as
well as others around that time. For example, Gruber (l967,1975a,b), who
developed his own theory of acquisition, says the following (Gruber 1975a:
71):
. . . thechild’s grammar does not develop in terms of increasing
complexity in the nature of his possible base structures, but rather in
terms of an increasing ability to analytically lexicalize those structures.
While the actual base structures produced may increase in complexity,
this is a performance, not a competence, factor. That is, we have
assumed that competence for the generation of base structures, being
innate, does not undergo development.
Gruber’s theory, in fact, is an excellent example of a strong innateness
theory. Pinker’s theory, by allowing more developmental change and the
possibility of discovering language-specificproperties, is a major movement
away from the universal base hypothesis. The exact extent to which this is
true, however, is speculative since Pinker does not provide analyses leading
to the possible Stage I grammars of young children.
Another alternative to the universal base hypothesis is expressed in
Chomsky (1981). Chomsky proposes a model of acquisition which he calls a
parurneter-setting model. We have already presented a brief discussion of
this model in Chapter 4, section 4.5. Most generally, the child is born with a
set of options or parameters about what is a possible language. These
options are such, however, that they are linked to each other in predictable
ways. The establishment of one parameter, then, informs the child that a
series of other characteristics should be expected. This simplifies the child’s
task to determining a presumably small set of parameters in acquiring a
language. Also, the child is born with an original or unmarked set of
expectations.
An example of the application of this theory to acquisition is Hyams
(1983,1986). Hyams looks at the acquisition of one parameter discussed in
Chomsky (1981) and numerous other works. This is the pro drop parameter
which distinguishes languages like Italian and Spanish from those like
English and German. Pro drop languages are characterized by several
features such as the optional use of subject pronouns, the use of rich verbal
332 THE PERIOD OF THE FIRST WORD COMBINATIONS I
agreement, and certain formal features of the GB model. Hyams proposes

that the child is born with the pro drop parameter as the unmarked
approach. English children, therefore, first treat English as a pro drop (*
language. Hyams argues that this accounts for several features in the
acquisition of English, including the early optional use of subject pronouns.
While Pinker treats this as a performance factor, Hyams proposes that it is
due to the child’s original analysis of a language which has optional subject
pronouns.
Since this theory is relatively recent, it is only now just being developed as
a possible theory of language acquisition: for example, it is still not clear
what the range of parameters is, nor how the parameters get set. The theory
is also undergoing change at a very rapid pace, which makes its adaptation
for child language somewhat difficult. I expect, however, that its use as an
explanatory theory of language acquisition will become more frequent.
7.5 The methodology of grammatical analyses of children

All of the approaches for writing grammars for children which we have
reviewed used spontaneous language samples for the analysis. This has
been the case despite the strong warning against this from Chomsky in 1964.
The lesson of the last twenty years is that such data are not sufficient to
determine the grammatical knowledge of the young language learner. If a
methodology exists which can, it will have to be of the form suggested by
Chomsky, that is, ‘devious kinds of observations of his performance, his
abilities, and his comprehension in many different kinds of circumstances’
(Chomsky 1964: 36). Also, we have not fully developed ways to use
spontaneous data and extract from them the most useful data for examining
grammatical competence. This section will outline some ways in which
these problems could be overcome. The discussion will be limited to one
goal - the collection of data from children in the period of first word
combinations for the purpose of determining the child’s grammar. It is
generalizable, however, to the study of other aspects of language acqui-
sition.
7.5.1 Spontaneous language samples

I do not suggest the elimination of spontaneous language samples; my
claim, however, is that such samples need to be the starting point for data
collection, not the end point. The typical use of such samples has been to
collect an extensive amount of language from the child, and then spend an
enormous time transcribing it. In principle, the transcription should occur
immediately after the sampling; in practice, it often occurs weeks, and even
months later, and this is not surprising, given the difficulty of the task. I
propose that the first sample should be relatively short, as little as 100
utterances, in order to get a general picture of the child’s current language
level. Next, if a longer sample has been collected, a selected transcription
can be done. We define a selective transcription as a transcription of a
subpart of a language sample of just those utterances needed for the analysis
being done. For example, the transcription could involve only the identifi-
able multiword utterances in the sample. We have already seen from
Bloom’s data how few of these can occur over several hours of sampling.
Importantly, the transcription should be completed within a day or two of
the sampling.
In making the above suggestions, it must be stressed that they are practical
in nature. It is certainly preferable to have an extensive transcription of two
or more hours of spontaneous language before beginning a grammatical
analysis; such transcriptions, however, are rarely available immediately
after a sampling. Here we place a higher importance on immediately obtain-
ing initial insights into the child’s language which can be used for further
observations. This is also practical in that a complete transcription does not
usually add that much more new information, relative to the time it takes to
do one, than does a selective transcription. I feel that the time spent on a
complete transcription can be better used planning experiments to pursue
initial hypotheses about the child’s grammar.
Of particular importance for grammatical analysis is the word index, the
sentence index, and the syntactic lexicon. As pointed out in our review of the
major analyses in the field, it is not enough to analyze a list of syntactic types.
As suggested by MacWhinney (1982), sentences can vary as to the extent to
which we can say that they are rule-based. For example, at one extreme, a
child utterance may simply be a memorized phrase, while at the other it
could be the result of a general combinatory rule. In other words, we need to
establish the productivity of individual sentences as well as proposed pat-
terns. Productivity is defined here as the extent to which a sentence or gram-
matical pattern can be said to be rule-based. The determination of such
information is needed before a grammar can be proposed for the child.
While the need to establish productivity has been expressed in recent
work, little has been done to develop a methodology. The simplest measure
criterion is a frequency criterion, such as the one used by Bloom (1970).
Recall that Bloom defined a pattern as productive if there were at least five
syntactic types which fit it. This measure, however, has a number of flaws. It
can (a) lead one to conclude that a pattern exists when the sentences which
meet the criterion are all memorized wholes, or (b) it can exclude
productive patterns which do not meet the criterion. Hypothetical examples
of each of these are given in (7.29):
(7.29) a. ride horsey b. eat apple
eat apple throw apple
throw ball throw ball
drink water eat cookie
tie shoe
Example (7.29a) contains five sentences which could qualify as an

Action+Object or verb+noun pattern by a frequency criterion of at least
five sentences. They are also possibly five reduced forms of sentences that
the child has frequently heard and has memorized. The sentences (7.29b)
would not meet the criterion. The fact that ‘eat’ and ‘throw’ occur with
different objects, however, is suggestive of a productive combinatory rule.
The first attempt to include co-occurrence in a definition of productivity
was Braine (1963a), where it was used to distinguish between Gregory’s and
Andrew’s use of ‘all gone’ (see above, section 7.3.1). Braine (1976) also
refers to productivity, but as in his earlier work, stops short of providing an
operational definition for it.
A more explicit attempt to define productivity is in Ingram (1981d). First,
any word which occurs in a child’s sentence is assigned to one of three
categories: not lexically free, lexically free, and grammatically free. In
principle, lexical freedom means that the child knows a word. Operation-
ally, a word is lexically free if either it occurs alone, that is, as a single-word
utterance (e.g. ‘cookie’), or in at least two sentences with different words
(e.g. ‘cookie here’, ‘want cookie’). If neither set of data is found, then the
word is ‘not lexically free’. Words which are lexically free may also be
grammatically free. Grammatical freedom means that the child knows the
word and has identified it as belonging to a grammatical class. A word is
operationally defined as being grammatically free if it is lexically free, by the
above definition, and occurs in at least two sentences with the same
relational meaning. In the previous example with ‘cookie’, ‘cookie’ is not
grammatically free because it has different semantic relations in the two
sentences. ‘Cookie’ becomes grammatically free if we add the example
‘cookie on table’, where ‘cookie’ can occur as a subject of locatives. (The
definition of possible relations is kept purposely vague, and can be altered
depending on one’s grammatical theory.) There are also two degrees of
grammatical freedom - weak and strong. Weak grammatical freedom
occurs when the word enters into only one relation; strong grammatical
freedom occurs when the word enters two or more relations.
The above definitions can be applied to every syntactic type in the data
under analysis. They lead to one of two kinds of further classification of the
data. First, they divide individual syntactic types into different degrees of
productivity, which will be shown first by presenting a set of hypothetical
data, and then using them to demonstrate the possible degrees of produc-
tivity. The adaptation of Ingram (1981d) found in Dale (1985) will also be
used. The hypothetical data in (7.20) are the collection of examples we have
used so far with some additions.
(7.30) a. ride horsey f. see apple j. cookie here
b. eat apple g. throw apple k. cookie on table
c. throw ball h. see ball 1. want cookie
d. drink water i. eat cookie m.table pretty
e. tie shoe j. table shoe n. see [a] cookie
The possible types of sentence are as follows:
nonproductive: sentences in which no word is lexically free. Examples
are ‘ride horsey’, ‘drink water’;
productive: sentences in which all words are grammatically free.
Examples are ‘eat apple’, ‘eat cookie’, ‘see apple’, ‘see ball’, ‘throw
ball’, ‘throw apple’;
productive / X sentences in which all words are grammatically free,
except for an unintelligible part. An example is ‘see [a] cookie’;
partially productive: sentences which fall between productive and
nonproductive. Dale (1985) distinguishes three kinds:
1. sentences with at least one word which is lexically free. An example
is ‘table pretty’ where ‘table’ is lexically free but ‘pretty’ is not;
2. sentences where all words are lexically free. An example is ‘table
shoe’ (with the meaning ‘table and shoe’) where both words are
lexically free;
3. sentences where at least one word is grammatically free. An
example is ‘want cookie’ where ‘cookie’ is grammatically free as
object of ‘see’ and ‘eat’, but ‘want’ is not lexically free;
Unclassified: all other sentences.
This classification can also be used to examine the productivity of specific
patterns proposed for the data, such as Agent+Action. Dale (1985)
proposes that a relation be considered productive if there is at least one
sentence in which both words are grammatically free for that relation.
These criteria provide an initial attempt to operationalize the notion of
productivity. They were applied in Ingram (1981c,d) to data from two
children to show that their grammars were quite similar despite the fact that
they were at different levels of cognitive development. The results were
used to argue that grammatical development appears to have its own form,
distinct from that cognitive development. Since the procedures are prelimi-
nary, they no doubt will need to be adjusted as they are developed. They
336 T H E P E R I O D O F T H E FIRST W O R D COMBINATIONS
provide, however, a useful way to begin an examination of the child’s
spontaneous sample.
The syntactic types obtained in the initial spontaneous sample from ths
child can be analyzed with some version of the criteria above to get a first
impression of the productive aspects of the child’s grammar. Again, this will
need to be done within a day or two of the initial sample. While these
analyses can be done by hand, a more efficient method would be the use of a
program for a microcomputer. Dale (1985) and Pye (1987) have recently
prepared programs that do analyses of just this kind. Given the avail-
ability of such programs, and the use of selective transcriptions, it is
reasonable to expect to be able to revisit a child within a day or two of the
first sampling to follow up the initial analysis of the child’s grammar.
7.5.2 Multiple testing procedures

It is suggested that all subsequent data from the child be the combination of
several procedures used to verify hypotheses about the child’s grammar.
These can be as numerous as practical and theoretical issues allow. Here I
will specify what I see as a possible program to follow the hypothetical
child whose data were given in (7.30).
First, we would want some information on the child’s receptive and
productive vocabulary. A parental diary technique of the kind used by
Benedict (1979) could be initiated either a week or more before the
spontaneous sample, or at that time. The parent’s diary could then be
checked by testing for the child’s response to words that are especially
relevant for the grammar suggested by the initial analysis. It would be
important to distinguish those words which are only in the receptive
vocabulary from those which are also produced.
Once some idea of the child’s vocabulary is established, we could begin
with more detailed study of the child’s receptive grammatical knowledge. A
useful test of receptive knowledge is that of Miller & Yoder (1984). Most
likely, we will want to develop specific tests for aspects noted in the initial
grammar. The techniques used by de Villiers & de Villiers (1973a) and
Katz, Baker & Macnamara (1974) are examples of possible procedures to
follow. For our subject, we would want to establish her receptive know-
ledge of word order across the semantic relations which appeared produc-
tive. We could also see if she shows some awareness of grammatical
morphemes like the articles and verb inflection which are not yet in the
productive language. The elicited imitation technique used by Rodd &
Braine (1970) is another way to get at this information.
Besides receptive knowledge, the subsequent visits to the child can be
used to test the predictive power of the grammar. A way to do this is to
Further reading 337
come with a list of sentences that might be predicted by the child’s patterns
but which are not attested. Our subject, for example, may have a productive
pattern for location, with ‘here’. The context of interaction with the child
could then be set up to be exactly as it was in the first session, except with the
replacement of an earlier item with a new one. For example, we could try to
replicate the context of the sentence ‘cookie here’, but provide an apple
instead, with the hope of eliciting ‘apple here’. This could be controlled by
also attempting to elicit previously produced sentences, and comparing our
success in the two contexts.
If well planned, the multiple testing procedures could be implemented
quickly enough that the child’s grammar would not have changed much
since the initial sampling. It is expected that such a procedure would provide
the kind of information needed to resolve the kinds of issues discussed by
Pinker concerning the nature of the child’s grammar. Syntactically rich
interpretations like Pinker’s would be supported by success on comprehen-
sion tests and the elicitation of gaps in the data. The leaner interpretations
could be supported by success at elicited attested, but not unattested,
sentences in the data. More detailed analyses of productivity should also
provide more information on the two kinds of learners found in Bloom,
Lightbown & Hood (1975).
Further reading
The major studies
Despite their age, the early studies by Bloom (1970, 1973), McNeill
(1970a), Braine (1963a, 1976), Brown (1973), and Bowerman (1973a) are
still necessary reading in the area of child language. Descriptions of Bloom’s
subjects and goals can be found in Bloom (1970: 1-23, 1973: 11-15), and
Bloom, Lightbown & Hood (1975: 1-8). Brown’s subjects are described in
Brown (1973: 51-8, 63-74), and I recommend the entire ‘Unbuttoned
introduction’. A brief introduction to Bowerman (1973a) can be gotten
from the first two chapters.
Syntactic comprehension
The three new studies reviewed in the text are all relatively short and should
be read: Katz, Baker & Macnamara (1974), de Villiers & de Villiers
(1973a), and Rodd & Braine (1970).
Early studies on grammatical development

The first major publications in this area were the proceedings of conferen-
ces. Bellugi & Brown (1964) was the first, and contained the initial reports
by Roger Brown and Sue Ervin, as well as Chomsky’s early comments on
language acquisition. Two years later, two further collections appeared,
Lyons & Wales (1966) and Smith & Miller (1966). There were two papers
by McNeill in these works which provided an alternative to Braine’s pivot
grammar. The later collection had a particular impact and was required
reading at the time. Two other of this early series of conference papers were
Hayes (1970), and Slobin (1971) which contained Schlesinger’s early
attempts to develop semantically oriented grammars. The early 70s saw the
first appearance of single-authored, book-length research reports, and also
the first textbooks. Major works appeared by Bloom (1970,1973), Bower-
man (1973a), and Brown (1973). Other important collections of papers at
the same time were Ferguson & Slobin (1973), which contained the first
English translations of several classic works, and Moore (1973). A series of
monograph-length research reports by Bloom and her colleagues and
Braine’s (1976) monograph completed this initial period of data collection
and analysis. A useful set of review papers appeared in the same year by
Morehead & Morehead (1976), reinforced by a lengthy article by Bower-
man on early semantic development.
Suggested passages from the works reviewed in section 7.3 are as follows:
Braine (1963a), McNeill(1970b: 107&99), Bloom (1970: 34-79) ‘Kathryn’s
grammar at Time 1’; Bowerman (1973a: 217-28) ‘Summary and conclu-
sions’; Brown (1973: 91-111) discussion and critique of pivot grammar, and
pp. 169-98, ‘The major meanings at stage 1’; Braine (1976: 56-70) ‘The
structure common to all corpora’; Bloom, Lightbown & Hood (1975:
2540) ‘Discussion, and conclusion’; and Bowerman’s (1975) commentary
on Bloom, Lightbown & Hood (1975), and (1976a) commentary on Braine
(1976). I also recommend Howe’s (1976) critique, and Leonard’s (1976)
review of semantic approaches.
Current approaches
Two recent collections of articles, Kuczaj (1982b) and Wanner & Gleitman
(1982) are major sources of information about several of the recent
approaches. The first work contains papers by R. Clark, Kuczaj, and
MacWhinney which all tend toward a distributional learning view of
acquisition. The paper by MacWhinney (1982) is long, but provides a great
deal of information on the linguistic behavior of children around this period
of acquisition and beyond. Wanner & Gleitman (1982) contains the paper
on functionalism by Bates & MacWhinney and the summary by Maratsos of
his view of distributional learning. It also contains a summary by Wexler of
the 3asic features of learnability research. Since 1980, Keith Nelson has
edited six volumes of collected papers entitled Children’s language. The
most extensive book-length work by a single author is that of Pinker (1984).
Further reading 339
A feel for Pinker’s work can be gotten from the following sections:
(pp. 1-12) ‘Introduction’, (pp. 34-47) ‘The bootstrapping problem’,
(pp. 67-8) ‘Acquisitionmechanisms’,(pp. 101-28) ‘Threelearnability prob-
lems for the Stage I grammar’.
8 The period of simple sentences: phonological and
semantic acquisition
8.1 Introduction
Chapter 7 dealt exclusively with one topic, that of grammatical develop-
ment. The reasons for this were twofold. First, the literature in this area is
quite extensive and requires an unusually long discussion for comprehen-
sive coverage. Second, that aspect of acquisition is usually broken down
into a separate period coinciding with the period of the first word combin-
ations (i.e. approximately from ages 1;6 to 2;O). It is, however, possible to
combine our treatment of acquisition of phonology and semantics over the
previous period and the current one.
There have been few attempts to provide general stages of acquisition for
phonology. In one of the first attempts to do so (Ingram 1976a:ll) the
following are discussed:
1. Prelinguistic vocalization and perception (birth to 1;O)
2. Phonology ofthefirst 50 words (1;0-1;6)
3. Phonology of single morphemes (1;6-4;0)
‘Child begins to expand inventory of speech sounds. Phonological
processes that result in incorrect productions predominate until
around age 4 when most words of simple morphological structure are
correctly spoken. ’
These are very much descriptive stages in the sense of Chapter 3, i.e. they
do little more than document changes over time. Nonetheless, they provide
an overview of the major behavioral changes which occur.
I have already examined the first stage in Chapter 5, and the second one
in Chapter 6. In the present chapter I will look at the period of the
phonology of single morphemes, which is proposed to last usually from age
18 months to around 4 years. It is claimed to be a period during which the
child goes from a small set of spoken words (approximately 50) to a large
vocabulary of words that are relatively correctly spoken. These words are
claimed to be simple in that they mostly consist of a single morpheme,
except for inflectional endings which are also acquired during this time.
340
Later development, which I will only briefly discuss, includes the acquisition
of (i) a small set of difficult sounds (e.g. English dental fricatives), (ii) the
morphophonemic rules of the language, and (iii) the spelling system.
There have been even fewer attempts to divide semantic development
into stages. Often it is simply divided into two periods, i.e. before and after
the word spurt around 18 months. The interest in viewing semantic
development in relation to cognitive development led us to follow it (in
Chapters 5 and 6) in relation to Piagetian stages of cognitive development,
up to the end of the sensorimotor period of development. Chapter 6 also
included discussion of studies on word acquisition which covered data on
children up to age 2 years.
In this chapter I review semantic acquisition between the ages of 18
months and 4 years, as with phonological acquisition. This time period is the
first part of what Piaget has referred to as the period of concrete operations,
which in his earlier work he divides into three subperiods. The first two
subperiods are summarized below, based on Ingram (1976:11) where they
are compared to phonological stages of acquisition:
1. Preconcept subperiod (1;6-4;0)
‘The onset of symbolic representation. Child can now refer to past and
future, although most activity is in the here and now. Predominance of
symbolic play.’
2. Intuitional subperiod (4;O-7;O)
‘Child relies on immediate perception to solve various tasks. Begins to
develop the concept of reversibility. Child begins to be involved in
social games.’
An important feature of this period is that the child is primarily involved
with her immediate environment. This will be seen in the way she comes to
understand temporal and spatial terms. The early attempts to do this will
demonstrate the use of ‘immediate perception’, for example, in interpreting
two clauses as reflecting the natural order of the events. Besides examining
some of the semantic errors children make, we will also look at some of the
theories that have been proposed to account for them.
The topic of grammatical acquisition will be returned to in Chapter 9,
concentrating on the child’s acquisition of grammatical morphemes, but
also considering the child’s developing ability to comprehend and produce
more complex sentences.
8.2 The phonological acquisition of single morphemes

Based on our review of phonological development during the period of
single-word utterances, the child around 18 months can be seen to have
342 SIMPLE SENTENCES
made the following gains. First, receptive vocabulary is around 200 words,
and the productive vocabulary is around 50. During the present period,
which follows the word spurt, the child’s vocabulary undergoes a tre-
mendous increase. Second, our review of Shvachkin (1948173 and Jakobson
(1941/68) suggests that some preliminary phonological analysis has begun in
both perception and production, with the former leading the way. The
child’s perceptual ability appears such that she can identify differences
between words which are marked by differences between single segments,
although this ability appears to emerge only gradually. Preliminary data
also suggest that the child’s articulatory ability, although far from complete,
is capable of producing at least a basic set of phonemes of the language
being acquired.
Given the incomplete development that exists at the onset of this stage, it
is not surprising that the word spurt should put a tremendous strain upon
the child’s primitive phonological system. The result is a period of time
during which the child shows errors both in the perception and production
of words. An attempt will be made, first for perception, then for pro-
duction, to document the kinds of errors which occur and their extent. The
explanations that have been offered to account for these errors will then be
examined.
8.2.1 The growth of phonemic perception

Shvachkin claimed that his subjects showed acquisition of the phonological
oppositions of Russian by the time that they reached 2 years of age. Despite
proposing a development sequence which the children followed, this still is
potentially a very strong claim for early acquisition. Given the results of the
recent studies on infant speech perception (Chapter 5 ) , it is not surprising
that early perceptual development should be in advance of production.
Here we will explore the extent to which this is true.
To examine this issue, we will first look at the implications of Shvachkin’s
study. While the creative nature of Shvachkin’s method is acknowledged,
the report still lacks sufficient information to allow unequivocal acceptance
of the results. As pointed out in Barton (1980), Shvachkin does not provide
details on the number of trials he used, nor his exact criteria for acquisition.
Even if we accept the results, which we do in their general form, the claims
that follow need to be closely tied to the stimuli used, i.e. they show
perceptual acquisition of the Russian consonants in minimally paired CVC
nonsense forms that were conscientiously and systematically taught to the
children. The results still leave open the possibility that other oppositions or
more complex words require longer acquisition time, or even that the
sounds studied take more time to acquire under totally naturalistic con-
ditions. Even Shvachkin’sclaim about acquisition is somewhat cautious. He
does not say that the children know all the aspects of the phonological
system of Russian, such as its diverse phonotactic constraints; he only states
that the child has the basic oppositions for such acquisition to proceed.
If we examine Shvachkin’s explanation for the order of acquisition of
oppositions, we find three factors which interact - perceptual salience,
articulation, and phonological salience (6.5.1). These can also be viewed as
possible reasons for the later acquisition of an opposition. That is, an
opposition could be late because of the fact that its acoustic properties are
such that it is more difficult to process by the human auditory system (low
perceptual salience); or because it is difficult to articulate (should we accept
Shvachkin’s claim that correct articulation aids perception); or because it
does not have a central role in the phonological system; or some combin-
ation of these. As stated earlier, such factors are difficult to isolate without
some independent measure of them. Even so, they provide a basis for
discussion of the known results on phonological acquisition.
There has been a wide range of studies conducted on this topic. The initial
enthusiasm wavers, however, on closer inspection. First, most of the
research has been on English; the fact that there has been little in the way of
cross-linguisticresearch limits the range of our conclusions from the onset.
The most serious attempts to study the effects of phonological salience will
need to wait until such studies are done. Second, much of the English
research was done during the period of large sample studies and has all of its
associated flaws. Templin (1957), for example, conducted a speech sound
discrimination study on her subjects but does not give any results on specific
words or phonemes - results are only given in the form of scores for
particular age groups.
Even the more recent studies are questionable because of the many
factors that may affect a perception study. Both Barton (1980) and Strange
& Broen (1980) have discussed a number of such factors. Strange & Broen
(1980) distinguish between stimulus variables and task variables. Stimulus
variables ‘refer to the choice and structure of the set of to-be-perceived
items, or more precisely, contrasts between items’ (p. 122). We know from
research on acoustic phonetics that there are numerous acoustic features of
a speech sound which can be possible signals of its phonemic status in the
language. By using computer-generated synthetic speech, it is possible to
manipulate the test stimuli along several dimensions. We can use only
criteria1 aspects, i.e. those that are recognized by adults as separating
phonemes, or can include non-criteria1 ones, i.e. those such as pitch and
amplitude, which do not affect phonemic perception. While such stimuli
allow one to control what is heard, they can also be criticized as not being
natural sounds. Other stimulus variables pointed out by Strange & Broen
Table 8.1 General methodology of studies on phonemic perception,
presented in order of increasing ages of subjects
Study Subjects Stimulus items'

No. Ages Shape Words Contrasts Task
Garnica (1973) 15 1;s-1;lO cvc nonsense ? select one of

two objects
Barton (1976b) 10 1;8-2;o cvc real 2 select one of
four objects
Johnson, Hardee & 8 1 ;6-2;6 cv nonsense 11 select one of
Long (1981) two objects
Eilers & Oller (1976) 14 1;10-2;2 varied real & 8 select one of
nonsense two objects
Barton (1978) 20 2;3-2;11 cvc real 20 select one of
cvc two pictures
Garnica (1971) 8 19-3;s cvc nonsense ? select one of
two objects
Edwards (1974) 28 l;8-3;11 cvc nonsense 18 select one of
two objects
Strange & Broen 20 2;11-3;s cvc real 3 select one of
(1980) two pictures
Clumeck (1982) 31 3;1-5;10 varied real 19 select one of
three pictures
~~ ~~ ~
IThe capital letter of the syllable shape indicates the syllable position in which the contrasts
were tested.
are: (i) lexical status, Le. whether a word is a real word or a nonsense word,
(ii) frequency of usage, i.e. how well the child knows a word, and
(iii) grammatical variables.
Task variables are those which have to do with 'the configuration in which
the stimuli are presented and the method by which subjects respond to
them' (p. 124). One common procedure is to use a discrimination tusk in
which two items are presented and the child must decide whether they are
the same or different. Barton (1980), in particular, is critical of this task on
the grounds that the child could process the stimuli acoustically without
using his linguistic (or phonemic) system, much in the way the prelinguistic
infant does. The alternative is to use an identijication task in which the child
needs to match the stimulus to an internal model to identify something that
represents the meaning of the word, such as an object or picture. Other task
variables pointed out by Strange & Broen are the number of trials used, and
if the subjects are given feedback on their responses. Barton also mentions
the potential problems of experimenter effects when the stimuli are
presented live instead of being recorded.
With so many variables involved, it is not surprising that the results from
the literature in this field do not constitute a homogeneous body of findings.
It is also possible to find fault with most studies on phonemic perception, for
one reason or another, as has been done in Barton (1980). Table 8.1
presents a summary of the methodological features of a selection of these
studies. The list is restricted to those studies which used an identification
task, since these are felt to reflect linguistic processing more directly. Here I
highlight some of these findings, claiming that they are actually more in
agreement than has been thought by some - the controversy which exists
seems rather to be tied to their interpretation.
Most of these studies were done in the early 1970s when there was a flurry
of interest in this topic. The first were done at Stanford University by
Garnica, who adapted the technique used by Shvachkin. This adaptation,
which has been referred to by some as the Shvachkin-Garnica method,
changes the testing procedure into a more strict experimental format. In
particular, a criterion of acquisition was added that the child would need to
respond correctly in seven out of ten trials before it was concluded that the
contrast was acquired. This method was used in Garnica (1971, 1973), and
Edwards (1974). It was first criticized in Barton (1975) on the grounds that
the criterion of seven out of ten is not an adequate statistic. Barton also
comments: ‘approximately one in six of the results thought to be significant
will in fact not be SO’ (Barton 1980: 100). The technique was subsequently
adjusted by Clumeck (1982), who used the following criteria to judge a
child’s perceptual ability on a contrast:
Clumeck’s criteria for phonemic perception
1. If the first five responses are correct, then conclude correctness and go
on to next pair.
2. If not, give five more trials and conclude correctness if nine out of ten
are correct; if five or fewer are correct, conclude that the child cannot
discriminate the contrast.
3. If six to eight are correct, give ten more trials and conclude correctness
if 15 out of 20 are correct.
This was an adaptation of Barton (1976a, 1978) who used criterion 1 and
then if the first five were not all correct went on to 20 trials. In the latter
case, 15 out of 20 needed to be correct. Other criteria used are nine out of
ten in Strange & Broen (1980) and ten out of 12 in Johnson, Hardee & Long
(1981).
Barton has also criticized the Shvachkin-Garnica method on the grounds
that the use of nonsense words rather than real words makes the task more
difficult. Barton (1976a) found, for example, that his subjects did poorer on
words that they had to be taught than on ones they already knew before the
experiment. The effects of familiarity were also found by Clumeck (1982).
Such concerns have led some investigators to restrict their stimuli to real
and familiar words. Since this choice restricts one’s choice of possible test
items, others by necessity have still used nonsense forms (e.g. Johnson,
Hardee & Long 1981).
Before looking at some results of these studies, I would like to comment
on the two above criticisms of the earlier studies. Regarding the statistical
error of the early Shvachkin-Garnica studies, it should be pointed out that
this error is in the direction of overestimating the child’s perceptual abilities.
The results of these studies can be examined with this direction of error in
mind. They can be interpreted with more confidence, however, if their
results turn out to be comparable to those in studies with stricter measures
of acquisition. (It is somewhat ironic that Barton was the one to point out
this error, given that his own theoretical view is that children have complete
phonemic perception at the onset of acquisition.)
The second criticism about the use of nonsense forms raises a general
issue about our definition of rule acquisition. Recall that Chapter 6 began
with six definitions of what it means to say that a child has acquired a
word. Defining acquisition continually returns to the question of the
extent of a rule’s productivity. In various domains of language acqui-
sition, such as morphological development, productivity is typically
defined as the ability to apply a known rule to a new or novel form. The
same issue is at stake here. If a child has a general phonological oppo-
sition, then it should be applicable to new words being acquired. Indeed,
when we consider it, the thousands of words that the child acquires after
18 months of age are nonsense forms to the child when first heard. Little
in the phonemic perception literature has been concerned with the ques-
tion of the productivity of the system. Most work with real words tests
ones that are familiar to the child, often with oppositions tested in just
one pair of words. The nonsense word studies are actually the only data
we have that begin to address this issue. We suggest, therefore, that the
nonsense word data are actually the most robust data we have on phone-
mic perception.
It is necessary to raise a further methodological problem before exam-
ining the results to date on phonemic perception, a problem concerning
the selection of the contrasts used in the various studies. Even if we
restrict ourselves to simple CVC syllables, we still have an enormous
range of possible oppositions to test. Every English consonant and con-
sonant cluster can be paired with every other, both in initial and final
position (as well as medial position in longer words). Much of this
selection is arbitrary, by necessity, even when the domain is restricted, as
in Johnson, Hardee & Long (1981). As a result of the wide range of
possible choices, few studies have examined the exact same contrasts.
Here this problem will be attacked first by looking at one general study,
then by examining the acquisition of contrasts within general classes such
as fricatives and glides.
Table 8.2 The contrasts tested, stimulus words used, proportions of
response on all words and known words for 20 children between 2;3 and
2;11 tested in Barton (1976a), based on Figures 3.4 and 3.5
Contrasts Pairs used Proportion of subjects to meet criterion
All pairs Known pairs'

(no. of subjects) (no. of subjects)
STOPS
lp-,t-1 pie, tie 1.00 (15) 1.00 (13)
1-p,-tl cap, cat 0.90 (19) (1.00) (3)
lk-,g-l coat, goat 0.81 (16) 0.85 (13)
curl, girl 0.75 (16) 0.92 (12)
card, guard 0.93 (15) 1.00 (10)
l-k,-gl lock, log 0.92 (13) (1.00) (3)
back, bag 0.94 (16) 1.00 (15)
frock, frog 0.47 (15) - -
/b-,g-/ boat, goat 1.00 (17) 1.00 (15)
/b-,p-/ bear, pear 0.95 (19) 0.94 (17)
STOP-NASALS
/b-,m-I bat, mat 0.94 (16) 1.00 (11)
1-d,-nl head, hen 1.00 (15) 1.00 (17)
cloud. clown 0.50 (14) 0.83 (6)
FRICATIVES
Is-,f-1 seat, feet 1.00 (20) 1.00 (8)
I-s ,-e1 mouse, mouth 1.00 (19) 1.00 (17)
APPROXIMANTS
n-,r-/ lock, rock 0.79 (14) 1.00 (8)
/w-,r-l wing, ring 0.89 (19) 0.94 (16)
CLUSTERS
lgr-,gl-l grass, glass 0.83 (18) 0.83 (18)
lkr-,kl-l crown, clown 0.77 (17) 1.00 (7)
Itr-,tJr-/ train, chain 0.77 (17) (1.00) (4)
Parentheses are placed around those proportions where there were four or fewer subjects
who knew the pairs.
Barton (1976a, 1978) Barton's work is generally considered to be the

first major work to overcome some of the problems of the earlier studies.
He actually conducted two studies on this topic, which are reported in detail
in Barton (1976a) and much more briefly in Barton (1978) and (1976b).
They are summarized in Table 8.1. Here we will look at the first of these,
which is discussed in Chapter 3 of Barton (1976a) and in Barton
(1978).
This study examined the ability of 20 children between 2;3 and 2;11 to
perceive the difference between 20 pairs of English words. The contrasts
and words used are presented in Table 8.2. Each word was audio-recorded
on a card which could be played through a machine called a language
master. The child would put the card through the machine, hear the word,
and then identify a picture of the object named. The machine was used
because it could provide a constant model for all children to hear, and
because it created a task which interested the children.
The task consisted of two parts. First, there was an identification phase in
which the child put each card through the machine with the sound off and
was asked to identify its picture. The child’s responses for each word were
placed into one of three categories: (1) named, if the child appropriately
named the picture, (2) prompted, ‘if the experimenter named it only once
. . . and the child pointed it out and could later identify it’ (p. 90), or
(3) taught, if the experimenter had to name it more than once. Second,
there was the discrimination phase. The child would run each card through
the machine and be asked to identify one of the two pictures which
represented the test pair. The criteria mentioned on page 345 were used to
determine if the child received five or twenty trials.
Of the 20 subjects, 13 completed all 20 test pairs. The other seven only
completed ten pairs each, for a variety of practical reasons. The results are
given in Table 8.2 in the form of two measures. The first one, labeled ‘all
words’, is the proportion of subjects on each pair who met the criteria for
acquisition. The second measure, labeled ‘known words’, is the same except
it excludes data on all pairs where the child’s response during the identifica-
tion phase was categorized as ‘taught’. That is, the second measure is
presumed by Barton to be more reliable in that it only treats words already
known by the children.
At first glance, these results are quite impressive. Even on the 11
words measure, most sounds are correctly perceived by at least 90
percent of the subjects. Barton uses these results to conclude the gen-
eralization that phonemic perception is complete when words begin to be
produced. This is an important conclusion in that it allows us to assume
that the underlying representation of children’s words is at least the
adult surface form.
The results are less exciting, however, when we take a closer look at
them. Recall that the subjects are children between 2;3 and 2;11, with a
mean age of 2;7. Children at this age have acquired a fair amount of
language already. According to Smith (1926), as cited in Dale (1976: 174),
the average spoken vocabulary at 2;6 is 446 words. We have already
considered data suggesting that the receptive vocabulary is much larger
than the productive one. We would expect the child’s perception to be
reasonably good for at least a subset of English phonemes. The question
then becomes: what is this subset? Since Barton has only examined some of
the contrasts found in English, it is still possible that performance might not
be as good on other ones. He may have selected those contrasts which are
characteristic of the early vocabulary of young children.
This possibility is strengthened by looking at the actual contrasts that
Barton tested. In (8.1), we give the phonemes tested by Barton in the form
of a phonetic chart. Next to it, we repeat the phonetic inventory for the
English child (6.17). This is the phonetic inventory determined in Ingram
(1981a) to be that typical of the average 2-year-old English child, based on
data from 15 children.
(8.1) English phonemes tested Phonetic inventory at 2;O
m n
b,p d,t g,k

f S h
W
clusters:
gr, gl, kr, kl, tr, tf r
We can see that eleven of the phonemes tested by Barton are typically
acquired by English children by age 2 years (the shared sounds are indicated
within boxes). Interestingly, those few sounds tested that aren’t are the ones
which received the lowest scores in the experiment. For example, all of the
words with clusters appeared to create some difficulties. All of this suggests
that the only conclusion we can draw from Barton’s experiment is that
children between 2;3 and 2 3 1 can perceive differences between sounds that
they already have been able to produce since age 2. This limited conclusion
is not surprising in that the study is actually a measurement sequence in the
sense of Chapter 3. By selecting sounds to test that are among the first
acquired by children, and restricting the stimuli to known words that can be
named by the children already, the method predetermined its outcome.
The lesson to be learned from Barton’s study is that perception research
cannot be done without considering the child’s articulatory abilities, for
example, by using norms of acquisition as the basis for selecting stimuli.
As we shall see in the next section, several large sample studies exist
Table 8.3 Contrasts used and proportion of subjects to meet the criterion
of acquisition in perception and production for 28 children between 1;8
and 3;ll to test four hypotheses in Edwards (1974)
Hypotheses Proportion of subjects to meet criterion

& contrasts Perception (no. of subjects) Production (no. of subjects)
H1:
/s/-l~ 0.50 (316) 0.00 (016)
Itl-Id1 0.67 (416) 0.50 (316)
If/-/VI 0.57 (417) 0.71 (517)
lpl-lbl 1.00 (717) 0.86 (617)
/J/-lzl 0.80 (415) 0.20 (115)
/t$l-d3/ 0.80 (415) 0.80 (415)
101-161 0.56 (519) 0.11 (119)
It/-Id/ 0.89 (819) 0.78 (719)
Totals 0.72 (39154) 0.50 (27154)
H2:
/s/-ld 0.50 (418) 0.00 (018)
/s/-lJl 0.75 (618) 0.38 (318)
Id/-I61 1.00 (818) 0.25 (218)
If/-I01 0.75 (618) 0.38 (318)
Totals 0.75 (24132) 0.25 (8132)
H3:
/kl-/g/ 1.00 (818) 0.88 (718)
lfl-lvl 0.75 (618) 0.75 (618)
101-161 0.75 (618) 0.38 (3/8)
1x1-lyl 0.75 (618) 0.25 (218)
Totals 0.81 (26132) 0.56 (18132)
H4:
/I/-Id 0.90 (17119) 0.32 (6119)
Ill-IWI 0.90 (17119) 0.47 (9119)
l~l-lwl 0.79 (151191 0.21 (4119)
/Il-/j/ 0.74 (14119) 0.32 (6119)
lrl-ljl 0.68 (13119) 0.32 (6119)
lwl-ljl 0.84 (16119) 0.79 (15119)
Totals 0.81 (921114) 0.40 (461114)
which provide such norms. Sander (1961), for example, has found that the
last sounds acquired by English children between 4 and 6 years of age are
/e/, IS/, /i/,/j/,/v/, /e/, /a/, and /z/. Note that Barton only tested one of
these sounds, /e/, and only in opposition to an early one, /s/. Another
option is to study the child’s productive ability at the same time as
perception. We will now turn to a few studies which have, in fact,
attempted these options.
Edwards (1974) This study used the Shvachkin-Garnica technique to

study the acquisition of a range of contrasts in 28 children from 1;8 to 3;11.
The contrasts were organized so that four separate hypotheses about their
acquisition could be examined. The testing of each of these four hypotheses
constituted a separate study, and not all subjects participated in all the
experiments. The four hypotheses (H1-4) are follows:
H1: Voice distinction in stops precedes voice distinction in fricatives.
H2: In acquisition of an unmarked-marked opposition, (i) per-
ception precedes production and (ii) the unmarked member substi-
tutes for the marked member in production.
H3: Velar spirants are ‘acquired’ at the appropriate point in phonolo-
gical development even though they are absent in the model language.
H4: Phonemic perception precedes production in the acquisition of
every opposition among the liquids and glides ll,r,w,jl.
The contrasts used to test each of these hypotheses are given in Table 8.3.
This list also demonstrates the point about the range of possible contrasts
that can be tested. Only three of these are the same as contrasts used by
Barton: /kl-/g/, /l/-/r/, and /w/-/r/. Edwards’s list provides much more
information about both fricatives and glides.
Most generally, the results indicate that phonemic perception overall was
good. All the contrasts were acquired by at least one-half of the subjects
tested. The only one to show a relatively low score was /s/-/z/. A comparison
of the results of this study with those of Barton’s for the contrasts which
were the same is given (8.2):
(8.2) Contrast Barton Edwards
Jk/-Jgl 0.81 1.00
Ill-Irl 0.88 0.90
/w/-lr/ 0.63 0.79
We can see that the results are similar, with Edwards’s subjects showing
better performance than Barton’s. This could be due either to the fact that
Edwards’s subjects were older, or to the statistical problems already
discussed.
The performance in perception was consistently better than that of
production, indicating that the children were perceiving differences that
they were not making in their speech. Even so, Edwards emphasizes that
many children showed a lack of perception on some of the pairs tested. This
can be seen by looking at the individual scores on the six contrasts used to
test H4. Of the 19 subjects who participated, ten acquired all six, three
acquired five, one acquired four, four acquired three, and one acquired
one. Also, while a general pattern emerged, individual subjects would vary
regarding the set of approximants which they had acquired.
The data provided only partial support for the four hypotheses tested.
Since these hypotheses are not the primary reason for our review of
Edwards’s study, they will be only briefly discussed. H1 predicts that a voice
opposition will exist within the stops at any point of articulation before it
occurs between fricatives at the same point of articulation. For example,
It/-/dl will be acquired before Isl-lzl. The results given in Table 8.3
generally support this prediction. For the four pairs tested, three out of four
show better performance on the stops than on the fricatives in perception,
and all four show better performance on the stops in performance. Oddly
enough, Edwards does not interpret the data in production this way,
because of substitution patterns for individual subjects such as those in (8.3)
below, i.e. consonants are not always occurring in their correct phonetic
form. (‘Subject numbers’ here are simple arbitrary numbers assigned to
individual subjects.)
(8.3) Subject 119

/PI+-
-- [PI /t/--
Subject 114
[t]
Subject 111
--
/PI- [PI
-
/b/- [PI Id/ [al,[d3l,[zl /b/ [bl
If/ [f] /SI ------3 [SI If/ [f]
/VI [f] /z/ [SI /VI +- [w]
It seems to us, however, that these data do not contradict the predictions of
H1. Subject 199 has not acquired voicing in either pair, so it is irrelevant for
H1. Subject 114 supports H1 in that voicing is acquired for the stops, but not
the fricatives. Subject 111 is irrelevant because voicing has already been
acquired for both pairs.
The first part of H2 predicts perception before production and is
supported by the data in Table 8.3. The second part of H2 is a very strong
claim about the interaction of perception, production, and the child’s
substitutions. Edwards proposes a four-stage model of acquisition of a
contrast which she feels is predicted by H2(ii). This model is given here in
(8.4) with hypothetical examples for the acquisition of Is/ and Id. (The
unmarked members of each pair tested are shown on the left in Table 8.2.)
(8.4)
I.
Stage
Perception and production of
marked member as being
Is/
/z/+
-
Perception
[SI
[SI
Production
Is/ .--) [SI
/z/- [SI
-
unmarked
11. Substitution of unmarked /SI- [SI /SI* [SI
member for marked member in /z/ [z] /z/ [s]
production
111. Variable production of the Is/- -
[SI Is/ [SI
marked member
IV. Acquisition of both members in Is/
perception and production
[SI
/z/ -+ [z]
- -
/z/- [z] /z/-
Is/
[s],[z]
[SI
/z/ -+-[z]
Edwards claims that the data support the prediction strongly for the pairs
Is/-IS/ and /d/-l6/, and weakly for /f/-/O/ (since one subject replaced /O/ with
[SI instead of [f]). They did not support H2 for /s/-/z/ where four subjects
replaced Is/ with [z] instead of replacing /z/ with [SI. This kind of variability
in fricative substitutions has been more recently substantiated in Ingram,
Christensen, Veach & Webster (1980). Edwards also found that the
children were variable concerning which stages were occurring for which
pairs. Subject 110, for example, was in stage I1 for /s/-/J/, but stage I11 for
/d/-/O/.
Of particular interest in this model is the prediction of stage I, i.e. a stage
of incomplete perception. Unfortunately, few data are given to support its
existence: the only case cited is Subject 113 who was found to be in stage I
for the pair /f/-/O/. It is such data which will constitute true counterevidence
to Barton’s claim that children have correct perception of words before they
produce them. We will later report stronger evidence for stage I in the study
by Eilers & Oller (1976).
H3 is of interest because it tests for the perception of a distinction that
does not occur in English, that is /x/-/y/. The prediction is that the
acquisition of a voicing distinction in English fricatives should generalize to
this non-English pair. The data in Table 8.3 support H3, although they did
not fully support some other predictions about this set of contrasts, which
Edwards discusses. Lastly, H4 on the acquisition of glides was strongly
supported by the data, Le. the subjects showed better perception than
production of each of the pairs. Edwards also points out that the order of
the scores differed for perception and production, leading her to conclude
that the order of acquisition of contrasts may differ between perception and
production. Since the numbers are so small, however, this can only be taken
very tentatively.
Edwards draws seven conclusions from the results of her study
(pp. 218-19; her own words are used here).
(i) Phonemic perception develops in a gradual and patterned way;

(ii) Phonemic perception of a given sound difference generally precedes
correct production of the difference;
(iii) The order of acquisition tends to be uniform, but the details vary
greatly;
(iv) The order of development in phonemic perception may not be

identical with the order in production;
(v) Initial fricatives and glides (liquids and semivowels) are acquired as
whole segmental units rather than by successive additions of indi-
vidual features;
(vi) Substitutions are usually less marked than the model but details vary
greatly;
(vii) Factors not directly related to the content of features or feature
combinations may be important in the acquisition process.
(Edwards names frequency of occurrence as one such possible factor,
although this factor is not explored.)
Despite its statistical problems and occasionally questionable interpreta-

tions of the results, the Edwards study is probably the most insightful study
conducted on phonemic perception for the period under review. We will
consider one further study which provides support for some of its specu-
lations.
Eilers & Oller (1976) Eilers & Oller were interested in studying the
relationship between perception and production. In particular, they wanted
to see if the common errors in children’s productions were the result of
articulatory or perceptual factors. As such, the study investigates the
existence of stage I as proposed by Edwards. I have already pointed out
that Edwards’s data were not sufficient to conclude the existence of such a
stage. If it exists, then at least some of the child’s errors could be accounted
for as perceptual errors. If there is no such stage, which is Barton’s
conclusion, then substitution errors could only be the result of articulatory
or linguistic factors.
Eilers & Oller used three types of stimuli to study this question. Type 1
stimuli were pairs of words which differ in sounds which are often collapsed
in the speech of young children. One such pair, for example, is /r/-/w/ which
are often collapsed to [w] by children. Type 2 stimuli were pairs of words
which differ in sounds that children do not usually collapse, e.g. /p/-/t/.
Type 3 consisted of just one pair which differed in /lc-/p/. This was a control
item to ensure that the children understood the task. If the errors on type 1
pairs are due to perceptual factors, then the children should find these more
difficult than types 2 and 3. If there are no perceptual factors, then the
type 1 pairs should be perceived correctly. For purposes of the later
discussion, we shall refer to these predictions as the perceptual hypothesis
and articulutoryllinguistic hypothesis respectively.
The study is particularly interesting because of the stimuli it used. Instead
of using either pairs of nonsense words or real words, they used real words
Table 8.4 Information about stimulus items and the proportion of subjects
to meet the criterion of acquisition in perception and production for 14
children between 1 ; l O and 2;2 tested in Eilers & Oller (1976), adapted
from Tables 1 and 4
Stimulus items Proportion of subjects to meet criterion

Perception Production
Gloss Phonetic form (no. of subjects) (no. of subjects)
TYPE 1
car-gar [khar] [kar] 0.86 (12114) 0.21 (3114)
block-bock [ P W [PW 0.58 (7/12) 0.17 (2/12)
monkey-mucky [mKgki] [ m ~ k i ] 0.20 (2110) 0.00 (0110)
fish-thish [fJI[WI 0.00 (O/lO) 0.10 (UlO)’
rabbit-wabbit [raebrt] [waebtt] 0.60 (6110) 0.10 (1110)
TYPE 2
pig-tig [phigl [thigl 0.50 (7114) 0.86 (12114)
block-lock [ P W [19kl 0.83 (10112) 1.00 (12112)
TYPE 3
cow-bow [kawl [pawl 1.00 (14114) 1.00 (14114)
Production of subject was unstable.
paired with nonsense words. For example, the child would be shown a
familiar toy object, e.g. ‘car’, and one unfamiliar nonsense toy, e.g. ‘gar’.
They decided on this method because it avoids the problem with real words
of having a restricted number of possible minimal pairs, and the problem of
nonsense words of having to teach them. Table 8.4 presents the stimulus
pairs used in the experiment.
The subjects were 14 children between 1;lO and 2;2. The procedure was
to show the child two objects which were each placed on top of a container,
one with candy inside. The child was told that the candy was ‘under
the (name of one of the objects)’. Each subject was first tested on
nonminimal pairs such as ‘horse’ and ‘dog’ until four correct answers in a
row were obtained. Then the experimental sessions began. The testing of
each pair began with a warm-up session during which the child was
encouraged to play with the objects. Each pair of words was tested eight
times, with the real word correct four times and the nonsense word correct
four times. The orders of presentation and location were randomized to
avoid biases due to hand or object preferences. The children’s responses on
the pairs were scored as correct if they got seven out of eight. The children’s
production of the test words was also recorded. A child was scored as
correct in production if he imitated ‘both the nonsense word and real
object word in such a way as to maintain some clear phonetic contrast’
(p. 325).
The results are presented in Table 8.4. They do not support fully either
the perception hypothesis or the articulatory/linguistic hypothesis. The
type 1 pairs were neither all correctly perceived nor all the most difficult to
perceive. The statistical analysis of the perception data revealed three levels
of difficulty. For purposes of reference, we will refer to these as ‘difficult’,
‘moderate’, and ‘easy’. Example (8.5) shows how the pairs fell into these
three levels.
(8.5) Level Pair TYP
Difficult fish-thish 1
monke y-muck y 1
Moderate rabbit-wabbit 1
block-bock 1
block-lock 2
pig-tig 2
Easy pig-tig 2
car-gar 1
cow-bow 3
We can see that the type 1 pairs fall into all three levels of difficulty. They
constitute two of the most difficult pairs, as well as one of the easiest. The
score for ‘pig-tig’ fell between ‘moderate’ and ‘easy’, so could not be
assigned to a unique level.
Eilers & Oller also present results on how individual subjects performed
on each of the pairs. For the difficult pairs, most of the subjects showed
neither correct perception nor production (nine out of ten for ‘fish-thish’,
and eight out of ten for ‘monkey-muckey’). For the easy pair ‘car-gar’, on
the other hand, nine out of 14 subjects showed correct perception without
correct production. Subjects were more evenly divided between these two
patterns of response for the moderate pairs. These data thus constitute
much stronger evidence for Edwards’s proposed stage I than that presented
in Edwards (1974).
In the above analysis, one seemingly peculiar result emerged. On the two
type 2 distinctions, several children showed incorrect perception but
correct production (five subjects for ‘pig-tig’, and two for ‘block-lock’).
This result leads initially to the peculiar conclusion that production precedes
perception. I would like, however, to offer an alternative explanation.
Observe first that the criteria for correctness were more strict for perception
than production (see above). Pairs of words which are articulatorily easy for
children when compared to their perception will be likely to show the above
pattern when an imbalanced scoring system is applied which favors pro-
duction. We take these results to indicate that these two pairs are difficult to
perceive relative to their likelihood of being imitated successfully. This is
just one example of the general finding of this study that the relation of
perceptual and articulatory difficulty is a complex one.
The results support the first conclusion drawn by Edwards about phone-
mic perception: its acquisition is gradual, with certain distinctions appearing
earlier than others. They also support the speculation that the child may
have a perceptual representation of a word which is not a correct perception
of its adult surface form. They do not, however, support a view that all of
the child’s substitutions are the result of perceptual problems. At least one
pair tested was perceptually easy but articulatorily difficult for the children.
Since both Edwards and Eilers & Oller are cross-sectional studies,
neither can be taken as definitive evidence for the stages suggested by
Edwards. They are suggestive, however, of distinctions first appearing
perceptually, and then later in production. Nor do they provide an idea of
the extent to which we can say that a child’s perceptual development at any
time is incomplete. Studies on phonemic perception, apart from Barton
(1978) and Clumeck (1982) at least, tend to select stimuli that are likely to
pose problems for the child. We still have little in the way of direct evidence
on the extent to which incomplete perception characterizes the young
child’s total vocabulary.
So far, we have examined phonemic perception through reviewing experi-
mental studies which directly pair words for the child to process. We would
like to conclude our discussion of the topic by looking into a very different
approach. This is the possibility of making claims about the child’s percep-
tual ability on the basis of her productive language.
The general method is something like this. First we do phonological
analyses of the child’s matches and substitutions, presumably by methods
such as those discussed in 6.5.3. Different theories about the relation
between the child’s perception and production lead to different predictions
about the resulting analysis. I list here three possible theories (one with two
subtheories) and the kinds of analyses that each predicts. The subsequent
discussion will attempt to clarify their claims.
1. Globalperception theory. The child only identifies general features of
words without assigning them necessarily to individual segments,
Prediction: analyses should show poor correspondences between
adult and child sounds with variable pronunciations.
2. Partial perception theory. Segmental perception is good but incom-
plete for some more difficult sounds. Prediction: pairs of adult sounds
that are not distinguished will surface with similar substitutions.
3. Complete perception theory I , The child perceives all adult sounds.
Prediction: the child’s substitutions are use of an unmarked for a
marked sound, with variable articulation for the marked sound.
4.
Complete perception theory ZZ.The child perceives all adult sounds and
keeps them distinct in his own speech. Prediction: the child sounds
never correspond to more than one adult sound.
Waterson (1970, 1971) was one of the first to discuss the possibility of
drawing conclusions based on the child’s productive language. She dis-
cussed this in relation to the analysis she made of her son P around the ages
of 1;6 to 1;8. At this time P had about 155 words, with approximately 100 of
them monosyllables (Waterson 1970: 9). She found that several of his words
couldn’t be analyzed as simple correspondences between his sounds and
those of the adult language. She concluded that he was extracting only
certain features from the adult words which he used in his own. Since this
research is the strongest data in the literature in support of the global
perception theory, it will be examined more closely.
Example (8.6) gives a phonological lexicon of a subset of P’s vocabulary
upon which Waterson based her arguments.
(8.6) angel [ 4 p ] , [=ha], [aeha], [afibl fish [IS], [US1

another [JlaJlaI flower [va], [vaewa]
barrow [wawae], [ b ~ w u ] fly [wal, [bPaI, [Pael, 1VI.
biscuit [be:be:] honey [a$u:]
Bobby [babu :] kitty [t~ti]
brush bYS1 pudding [pupu]
bucket [babu:] Randall [pap]
dirty [&:ti] Rooney [ehe], [hehe]
dish [dIS1 vest [US]
fetch [IS1 window [pe:pe]
finger [gC:pC], [ p i : p ~ ]
In Table 8.5 we give two analyses of these data. The first is a substitution
analysis of P’s initial consonants of the kind presented in Chapter 6. The
second is a summary of Waterson’s analysis.
The segmental analysis in Analysis 1 supports the predictions for the
global perception theory. The correspondences look questionable in that
they are not what we would expect if the child were simply attempting an
adult sound. [JI] seems to occur for several sounds which it does not
particularly resemble. Also If/ shows extensive variable pronunciations.
The fact that [t] occurs for both /t/ and /k/suggests another possible
perceptual confusion. Waterson wants to suggest that such patterns are
indicative of the child’s incomplete perception. Instead of matching indi-
vidual segments, ‘the child perceives only certain of the features of the adult
utterance and reproduces only those that he is able to cope with’ (Waterson
1971: 181).
Table 8.5 Two analyses of the data for P in (8.6). Analysis 1: a segmental
analysis of P s initial consonants; Analysis 2: a summary of Waterson’s
(197I) analysis
Analysis 1
Matches and substitutions for initial consonants

Ip- t- k- b- d- f- h- v- w- r- n- I
(PI (t) 0) *b d 0 (0) (0) (51) (51) (51)
(w) V
Analysis 2: Five perceptual structures
I. Labial structure: wV(w(V) 11. Continuant structure: VhV

barrow [wzwz], [bnwu] angel [a@], [aeha], [afiir]
flower [vz], [vzwz] honey [a&:]
fly [W.I> [ b B 4 [ P 4 >[VaeI? Rooney [ C f i E ] , [hCfiE]
111. Sibilant structure: (C)V IV. Stop structure: C V
brush [by$] biscuit [be:be:]
dish [ W Bobby [bzbu:]
fetch [I$] bucket [baebu:]
fish [LrL [US1 dirty [&:ti]
vest IUS1 kitty [t I ti]
pudding [pupu]
V. Nasal structure: pVpV
another [papa]
finger [pe:p&:]. [pi:p~]
Randall [pap01
window [pe:pe]
As stated earlier, Waterson’s data make a strong case for the global
perception theory. There are, however, two reservations that stop us from
accepting this position. First, the data in (8.6) constitute only 21 of P’s 155
words. We are given little information about the nature of these other
words. It may be that the other words demonstrated correspondences more
consistent with segmental analyses. If so, then the data may be more
consistent with the partial perception theory in that they point out a small
number of the more difficult discriminations which P is still working out,
such as initial /f/-/fl/ and medial nasals. A second concern is that extensive
data of this kind are not widespread in other phonological diaries which
have been reported. This will be apparent when we review some of them in
the next section.
The opposite position to the global perception theory is the complete
perception theory II. This would be the strongest claim for early perception,
in that the child would never in its speech show mergers of adult categories.
Such a situation is logically possible on the grounds that the child’s language
acquisition device would dispose the child to attempt to maintain distinc-
tions of the adult language in his own language. We will, in fact, in our later
discussion of homonymy provide some evidence that such constraints are at
work. Current evidence indicates, however, that perceptual and/or articula-
tory factors still prevent this tendency from being totally successful.
This leaves us then with the partial perception and complete perception I
theories. It is our impression that the above data from perception studies
and data such as Waterson’s favor at this time the partial perception theory.
It is not possible to take a definitive position between these two, however,
because the crucial research to decide the issue has not been done. This
would be work which does phonological analyses on the child’s spontaneous
language, and then tests for perceptual awareness of those adult distinctions
which are merged by the child. In taking this position, I wish to clarify that I
do so with the feeling that the child’s incomplete perception will be more
noticeable in the child’s earliest words, and that it may never be apparent
for more than a few distinctions at any time.
The difficulty in determining which of these theories is at work can be
exemplified by looking at our analysis of T in Table 6.29. There are no cases
of two adult sounds being merged into one sound in T’s speech where both
cases involve non-lexical matches or substitutions. For example, /t/ is [t] and
/8/ is [t], but the latter only occurs in one word. Also, the cases of mergers
which occur tend to fit the predictions of the complete perception theory I.
For example, /t/ is [t] while /k/ appears both as [t] and [k], and /J/ appears as
[t] and [SI. These are the variable pronunciations of Edwards’s stage Ill
when perception is good.
Summary. We have explored the extent to which we can say that children
during the current period of acquisitions have acquired phonemic per-
ception. We have looked at both experimental studies and production data
in attempting to find an answer to this question. The evidence to date
suggests that children at this time have reasonably good perception of the
words they know, but not necessarily complete knowledge. There may be
particular distinctions which are inherently more difficult and which thus
take longer to acquire. We have also seen from Schwartz & Leonard (1982),
as discussed in Chapter 6, that children are selective in the words that they
use. We need to establish the extent to which such selectivity contributes to
this impression.
8.2.2 Phonological production

The discussion in Chapter 5 of early production for the first 50 words has
suggested that children begin acquisition with a small phonological inven-
tory which may show some influence of the native language. Despite this
early system, the child’s productive vocabulary is still quite small and variabi-
lity both within and across children is quite noticeable. I proposed a method
of analysis which reduced much of this variability to reveal the systematicity
which led Jakobson to his original proposals on language acquisition.
The present period begins with the word spurt and provides more
substantial data to make claims about phonological acquisition. While some
researchers deny phonological organization during the acquisition of the
first 50 words or so, most researchers will agree that the data of the period of
simple sentences shows such organization. Here we will begin with a
description of the kind of data available, and turn to some of the expla-
nations that have been put forward to account for it.
Large sample studies

A number of large sample studies have been conducted on phonological
acquisition in English, as well as in other languages (see the bibliography at
the end of the chapter). Many of these were conducted early on, e.g.
Wellman et al. (1931), Templin (1957), although there have been some
others since then, e.g. Olmsted (1971), Prather, Hedrick & Kern (1975).
Since the goal of most of them is to provide an overview of acquisition, they
usually study as many sounds of the native language as possible. Some
studies, however, have also concentrated on a single class of sounds, e.g.
Ingram et al. (1980) on fricatives and affricates in English.
Templin (1957) We can get an idea of the general nature of these studies
by examining the most frequently cited in Templin (1957). That this study is
still perceived by many as the single most important normative study on
English is supported by the fact that it is heavily cited for this purpose in the
most recent text on phonological acquisition by Stoel-Gammon & Dunn
(1985: 30-3). The general aspects of Templin’s study have already been
summarized in Table 2.3. To reiterate briefly, Templin tested 60 children at
each of eight age intervals between 3;O and 8;O on 176 sounds in English.
The younger children (3;0-5;0) were asked to identify a picture or asked to
imitate the experimenter. The older children either read the words or
repeated them.
Here we are particularly interested in how Templin constructed the test
words, recorded the responses, and scored for acquisition. Table 8.6 lists
the words used by Templin, distributed according to the consonants in
them. Importantly, each of the 176 test sounds was tested in only one word.
These test sounds are shown in Table 8.6 in capitals. For example, while
there were six words which began with a lpl, only the child’s response to
‘pie’ was scored for this sound. All responses were scored by Templin on
site without the use of a tape-recorder. Only the test sounds in a word were
Table 8.6 Test words used in Templin (1957)for children aged
3;O-5;0, arranged by the word position of their consonants. I =
word-initial, M = word-medial, and F = word-final, based on
Templin’s analysis of these words. Test sounds in each word are
capitalized
~ ~~~
STOPS
IPl I: paGE, peaCH, peas, Pie, piG, pIn, PROMPT, playTHing, PLeaSure M:
open, o’poSSUM, Upon F: CHip, SHip, soap, SOUP, SLeep, Sweep, graSP,
heLP, laMP, shaRP
It/ I: tIE, Toes, tuB, TRain, tweLFTH, TWiNKLE M: SKaTing F: baT, Feet,
Goat, Seat, WHite, biGGest, buRST, giFT, JumPED, neXT, PROMPT,
QUILT, SQUIRT, condUCT, elePHaNT
/k/ I:caGED, caR, caRD, Cards, cone, conduCT, clocks, cloud, CLown,
QUILT, CRaCKER, chaSM M: Vacuum-cleaner (X2), black-and-
WHite F: bAck, raKE, music, aSK, miLK, SHRiNK, WORK
/b/ I: bAck, bAll, baT, baTH, Beans, bEEts, beLL, bOILS, bOOKS, buLB,
buRST, baTHing, biGGer, biGGest, bo’lTLE, buBBLE, buCKLE, BLocks,
bloTTer, BRead, black-and-WHite M: rhuBaRB F: tuB, buLB, rhuBaRB
Id/ I: Dish, daNGLE, DRum, driving M: WaDing, black-and-WHite F:
BReaD, cloud, rEd, SPRead, caRD, saND
I: g i n , Goat, guM, gun, garaGE, giGGLE, GLass, gliMPSE, graSP,
GRass M: biGGest F: piG, icebeRG
FRICATIVES & AFFRICATES
I: Feet, fiSH, FLoWing, FRoZen M: elePHaNT F: kniFE, mySeLF, whaRF
I: THiNNer, THRee M: playTHing F: baTH, Mouth, eaRTH, heaLTH,
moNTH, tweLFTH
I: sAI1, saND, Seat, Soap, SOUP, SCRatch, SLeep, SMooTH, SNOW,Spoon,
SPRead, SQUirt, STairs, stoNE, stOVE, STRing, Sweep, swiNG, swiNGing,
SKaTing, SPLaSHing M: mySeLF, icebeRG F: GLass, GRass, HORse,
mouSE, bEEts, BLocks, bOOKS, clocks, gliMPSE
I: shaRP, SHip, ship-aHoy, SHRink M: SPLaSHing F:Dish, fiSH
I: Vacuum-cleaner M: driving F: stoVE
I: THose M: baTHing F: SMooTH
I: ZiPPER M: FRoZen, music F: peas, Toes, THose, Beans, bOILS,
Cards, STairs
M: PLeaSure F: rouGE*, garaGE
I: CHip, chuRCH, cheRRy M: reaCHing F:peaCH, SCRatch, chuRCH
I: JumPED M: aGing F: paGE, caGED, laRGE
maMa, miLK, moNTH, mouSE, Mouth, maSHER, moTHER, music,

mySeLF M: maMa F DRum, guM, laMB, aRM, chaSM, o’poSSUM
I: kniFE, Nail, neXT, nuMBER M: oNIon, conduCT F: CLown, cone,
gun, pIn, Spoon, stoNE, TRain, hoRN
M: swiNGing F: Ring, STRing, swiNG, aGing, baTHing, driving, FLoWing,
playTHing, reaCHing, swiNGing, WaDing, SKaTing
LIQUIDS & GLIDES
/I/ I: Lamb, IaMP, IaRGE M: elePHaNT, YeLLow F: bAll, beLL, Nail, sAII,
aPPLE, bo‘lTLE, buBBLE, buCKLE, daNGLE, giGGLE, ruFFLE,
TWiNKLE, whiSTLE
Irl I: raKE, rEd, Ring, reaCHing, rhuBaRB, ruBBER, ruFFLE M: cheRRy,
GaraGE F: caR, biGGER, aSTER, blo’lTER, CRaCKER, haMMER,
maSHER, moTHER, nuMBER, ODOR, OFFER, ruBBER, THiNNER,
Vacuum-cleaner, whiSKER, ZiPPER
lwl I: whaRF, WaDing, whiSKER, whiSTLE M: FLoWing
lhwl I: WHite M: black-and-WHite
lhl I: heaLTH, heLP, hoRN, HORse, House, HaMMer M: ship-aHoy
/j/ I: YeLLow M: oNIon
* used only if ‘garage’ was pronounced with an affricate
scored, and these were either (a) correct, (b) omitted, (c) substitutions
with other English sounds, or (d) substitutions with other sounds. A sound
was considered acquired when at least 75% of the subjects in an age group
produced it correctly. Templin provided a number of tables which allow
manipulation of her results, although she unfortunately did not present
information about the actual substitutions used by the subjects. Her results
will be discussed first for single consonants and then for clusters. The
discussion will concentrate on consonants because all twelve vowels which
were tested reached the criterion by 3;O.
By 3;0, all the nasals were acquired, and all the stops except /-k/, /-t-,/ and
final voiced ones /b,d,g/. All of these were subsequently acquired by 4;O
except /-t-/. If we look at the data more carefully, we can account for the
superficially individual behavior of /-kl and /-t-/. For /-k/, 73 percent of the
3-year-oldsproduced it correctly so that it just missed criterion. Also, at 4;0,it
scored 87 percent compared to only 75 percent for /-t/ which did meet the
criterion at 3;O. Further, l-W was tested in ‘rake’, which contains a front vowel
which fronts the tongue position of the /W and often influences children to
produce a more frontedsubstitute. Becauseof these factors, we will group/-k/
with the other voiceless stops and conclude that it is acquired by 3;0, lowering
the criterion for acquisition in the process to 70 percent. Medial position At-/
was tested in the word ‘skating’.It is a well-knownfact of English that both/-t-/
and I-d-/ undergo a rule of flapping where each is produced as [r]. Unfortu-
nately, the acquisition of this rule has yet to be studied. In particular, we
would want to know when children determine that the surface form [r] is the
surface neutralization of two phonemes. Such a study would include an exam-
ination of the use of flapping in the speech of parents to children. Templin’s
data indicate that the /-t-/vs./-d-/ contrast may be acquired late, but her scoring
system is not adequate to conclude this. A child’s flapping a I-&/ would be
scored as incorrect even though it is an appropriate English pronunciation.
The glides lwl, lhl, and Ijl are all acquired by 3;O. The first two are to be
expected from our study of the previous period of acquisition. The sound
Ijl is not particularly frequent in English and may be subject to the influ-
ences of functional load discussed in Chapter 6. The liquids I11 and Irl reach
criterion for acquisition by 4;0, with two exceptions: I-rl in ‘car’ is correct
80 percent of the time by 3;5, while 1-11 in ‘bell’ does not reach criterion
until 6;O. Both these sounds, however, need more careful testing than in
one word. As pointed out by Olmsted, /-rl is one phoneme, but is phoneti-
cally quite variable depending on the preceding vowel. He, in fact, uses
several phonetic symbols to transcribe this sound. Syllable-final 1-11 is also
quite different phonetically from syllable-initial /1/ by showing a high
degree of velarization (see Ladefoged 1975). The velarization of 1-11may be
so great as to reduce the sound to a vowel. Given these difficulties with 1 4
and /-11, we will separate them from the syllable-initial liquids and conclude
that the latter are acquired according to Templin’s data between 3;O and
4;O.
This leaves the fricative class. For a first look at this class, the percent-
ages of correct responses for the three word positions in (8.7) are collapsed
and are given here for each of five age groups. (The acquired sounds are in
bold type.)
(8.7) Percentages of correct responses in Templin (1957)
Age If1 I81 Is1 IS1 lv1 I61 lzl 131 Its1 la1
3;O 84 26 64 44 31 20 34 15 47 35
3;5 88 32 78 63 39 23 55 23 61 41
4;O 92 49 76 71 51 42 58 33 72 69
4;5 97 59 77 84 54 47 66 38 83 72
5;O 91 63 78 82 62 56 64 43 84 81
As we did with the acquisition of I-kl, we will modify the criterion of acqui-
sition from 75 percent to 70 percent. Also, for the years where half-year
data are available, we will average the two. For example, the percentage of
acquisition for 3;O will be the average of the figures €or 3;O and 3;5. This
loosening of the criterion is done on the grounds that 75 percent is a rela-
tively strict criterion. Recall that the measure for matches in Chapter 6 was
over 50 percent, a lower measure also used in Olmsted (1971). The revised
measure shows the following acquisition. First, If1 and Is1 are both acquired
at 3;O. This is consistent with the findings of English fricatives reported for
the period of the first 50 words. Next, the palatals IJI, Itll, and /@Iare
acquired at 4;O. None of the other sounds reach either Templin’s or our
adjusted criterion, indicating that these are the latest consonants acquired
in English.
While the positions for these sounds were merged, it can be seen that at
least some of the non-acquired fricatives were acquired differently in
different word positions. These differences are given in (8.8):
(8.8)
Age /v/ /z/ 131 ti/
I M F I M F M F I M F
3;O 12 60 22 30 47 25 18 12 23 17 20
3;5 30 65 23 55 75 35 25 20 32 27 12
4;O 40 75 38 62 72 42 45 22 57 45 23
4;5 47 78 38 72 75 50 48 28 60 52 30
5;O 55 80 50 65 75 53 57 28 62 67 40
The voiced fricatives /v/, /z/, and /3/ all show more of a tendency to occur
correctly in medial position than in either initial or final. The pattern is most
obvious for /v/, but is found for the other two also. The voiced dental /a/
shows a different pattern, in which the initial and medial positions are
noticeably better than final position. There is a slight tendency for /e/ to do
this also, but not enough to suggest grouping them into a class in this regard.
(8.9) presents a summary of our interpretation of Templin’s results on the
acquisition of English consonants.
(8.9) Summary of English consonant acquisition to 4;O

3;O nasals /m,n,g/; voiceless stops /p,t,k/; glides /w,j,h/; fricatives
/f,s/; voiced stops /b,d,g/ except finally
4;O voiced stops in final position, palatal fricatives and affricates
/J,tJ,@/; liquids /l,r/ at least in initial and medial position; medial
/V,d
This summary indicates that the majority of English phonemes are acquired
by 4;O. The sounds still to be acquired are the flapped intervocalic alveolar
stop and several fricatives, particularly the dentals /€),a/ and the voiced
palatal 131.
As can be seen from Table 8.6, Templin also examined a larger number of
consonant clusters, 71 with two consonants and 19 with three consonants by
her count. (We say ‘by her count’ because not everyone would classify
clusters in the way Templin does. This point will be returned to below.) A
brief summary of these data is as follows. (i) Active acquisition of fricatives
begins between 3;O and 4;O. At 3;0, only one cluster, /gk/, reached
Templin’s criterion; this increased to 15 clusters at 3;5 and to 35 or nearly
one-half of those tested by 4;O. (ii) The clusters acquired during this time
belong primarily to the following categories: (a) initial clusters with /s/ and a
following stop /p,t,k/ or nasal /m,n/; (b) initial clusters with stops
/p,t,k,b,d,g/ followed by a liquid A,r/ or glide /w/; and (c) final clusters with
a nasal /m,g/ followed by a voiceless stop /p,M. (iii) Those clusters which are
acquired later tend to fall into the following categories: (a) clusters with
three consonants - only three of the 19 tested reached criterion at 5;0,while
12 did at 6;O; (b) initial clusters which combine fricatives and liquids, e.g.
/sl-/, /fr-/; (c) final clusters which combine liquids /l,r/ with obstruents, e.g.
/-rb/, /If/; and (d) final clusters which combine /s/ with stops /p,t,k/.
Templin’s study provides a useful descriptive overview of English phono-
logical acquisition. It can be used for comparison with similar studies on
other languages, and as normative data for English acquisition. Here,
however, we will conclude with a caution about using large sample data such I
as these for anything more than the most general of purposes, setting out a
series of problems with Templin’s study in particular and large sample
studies in general. The limitations of such studies need to be emphasized
since their results may be inappropriately used both for theoretical and #
practical purposes, the latter including cases where a child might be

misidentified as being speech-delayed because of his performance of a
Templin-style articulation test.
A first point that can be raised about this study is that it is an articulation
study, not a linguistic one, that is, the results tell us something about the
degree of difficulty of certain sounds, but relatively little about the
phonological systems of children. The fact that children have more difficulty
with one sound over another does not of itself eliminate the possibility that
children may still have both sounds within their linguistic system. This is
particularly possible when a high criterion of acquisition is set and the two
sounds fall closely together on either side of it. The study is restricted from
being a linguistic study also because grouped data are being used for the
analysis: we are never talking about the language of a single child, so we can
never be sure whether the general pattern is reflected in individual children.
Such grouped results can only be interpreted with confidence when they are
supported by single subject analyses.
Other problems arise because of the nature of the words tested and the
way they are analyzed. For one thing, each sound is only tested in one word.
When we say that some sound ‘x’ was either acquired or not at some age,
what we are really saying is that this is so for this sound in the word tested.
For example, when we say that /v/ is not acquired until 6;0, what we are
saying is that 75 percent of children before 6 years of age cannot pronounce
[VI in ‘vacuum cleaner’. It may well be that children will do better with this
sound in another word such as ‘vase’. We will refer to such a possibility as
lexical variability. Despite Templin’s claims to the contrary, evidence exists
which suggests such lexical variation is due to the influence of syllabic
complexity and adjacent sounds (see Ingram 1976a: 83-5; and Ingram etal.
1980). This factor alone could influence greatly conclusions about the age of
acquisition of a sound when just one word is used for testing. It should be
pointed out that such problems are not inherent in studies of this kind. As
we can see in Table 8.6, there were often several words in the test which
would show each of the test sounds. If all of these data were recorded and
included in the analysis it would contribute greatly to eliminating the influ-
ence of lexical variability.
A factor which interacts with lexical variability is word familiarity. We
have already seen from the previous section that children are more likely to
perceive differences between sounds in words they know than in unfamiliar
ones. It is also possible that children’s pronunciation of words will be influ-
enced by this factor. A casual glance through Table 8.6 reveals several
words that do not appear to be typical of very young children, e.g. ‘aster’,
‘conduct’, ‘pleasure’ and ‘rhubarb’. Shriberg & Kwiatkowski (1980) have
criticized several such tests on the grounds that the words used are also
much more phonologically complex than those typical of children at the
ages being tested. They demonstrate this point with percentages of different
kinds of syllable structures’in words used in such tests as compared with the
distribution of such syllable structures in the natural speech of young chil-
dren. They, in fact, reach the conclusion that the only accurate way to assess
a child’s phonological system is through a naturalistic language sample.
A last problem with Templin’s study is the lack of linguistic sophistication
used in selecting the test sounds and words. I have already pointed out
the problem with the flapping rule for medial dentals, and the allophonic
variation between final and nonfinal /r/ and A/. A further set of such prob-
lems may arise because Templin does not distinguish syllable position from
word position. Sounds which are word-medial can occur in different syllable
positions; for example, medial /-f-/ and /-s-/ were tested in ‘elephant’ and
‘myself’respectively. The [f] in ‘elephant’ begins an unstressed syllable pre-
ceded by an unstressed vowel, and this is a different case from the [SI in
‘myself’ which begins a stressed syllable. Acquisition data suggest that
word-medial sounds beginning stressed syllables pattern similarly to word-
initial ones in stressed syllables, yet differently from those beginning
unstressed syllables. For example, the [SI in ‘myself’ is more likely to
pattern with the [SI in ‘seat’ than the [SI in ‘kissing’. We already observed in
Templin’s data a difference for several of the fricatives between sounds
beginning stressed and unstressed syllables. A further difficulty concerns
Templin’s treatment of syllabic consonants such as /r/ and /I/ in words like
‘masher’ and ‘whistle’. While their phonological status is controversial,
many linguists would agree that phonetically these sounds should be treated
as vocalic elements rather than members of clusters. If so, then words like
‘apple’ would not be analyzed as ending with a cluster ‘pl’ as Templin has
done, but rather as CVCV or CVCVC structures with the [p] treated as
intervocalic.
Limitations of these kinds are characteristic of large sample studies, but it
should be pointed out that they are not necessarily inherent in the studies,
but rather, they are characteristic of how such studies have been designed
and analyzed. A large sample study done in a different way could be a
valuable contribution to the study of phonological acquisition. For
example, suppose that Templin had tape-recorded all of her subjects and
done analyses of the kind discussed in Chapter 6 for all of the words listed in
Table 8.6 for each child. Also, suppose analyses for individual children
had been done within a week or two of the sampling so that follow-up visits I
could have pursued questionable aspects of the results, much in the way
suggested for syntactic studies at the end of Chapter 7. This approach would
have yielded tremendous information about the phonological acquisition of
English.
Diary studies As a consequence of limitations of the kind just outlined in

the analyses of data from large sample studies, the most insightful data on
the emergence of phonology come from diary studies. These studies
provide much more detail on the development of the phonological systems
of individual children. A small number of such studies has come to form the
basis of much which has been claimed about phonological acquisition. A
summary of the data from a few of the most famous ones on English
acquisition is presented here.
Velten (1943) Velten (1943) was the first major phonological diary to
appear in support of Jakobson’s (1941) theory of phonological acquisition,
newly published at that time in German. Velten reports on the acquisition
of his daughter Joan from her 11th to 36th month. Her words are presented
in a broad phonetic transcription, and are usually dated by the month in
which they appeared. No variable pronunciations are presented for indi-
vidual words and it appears that Velten has presented the most typical form
used by Joan for the period being discussed. Not all words are presented for
the entire period studied so that it is sometimes difficult to determine how
each word used was pronounced during each month of the study.
Before looking at Joan’s data, it is important to discuss her linguistic
background. The mother spoke American English, but other languages
were also used. Velten states (p. 285):
Since French and Norwegian are frequently spoken in the household
and used to address the child (though less often than English), she
acquires at an early age a large passive vocabulary, and is able to i
follow fairly complicated directions, in these languages . . . Normally

Joan responds in English when addressed in French or Norwegian.
Table 8.7 Words acquired by Joan Velten between the 11th and 22nd
months, taken from Velten (1943). Distinct phonetic types and lexical types
are numbered by their order of appearance*
~
Word English (month) Word English (month)

~
1. ap 1. ‘up’ (end of 11th) 18. ba: 15. ‘pie’ (15th)

2. ba 2.‘bottle’ (end of 11th) 19. dat 16. ‘duck’ (16th)
3. ba 3. ‘ban’ (end of 11th) 20. hap 17. ‘lamb’ (16th)
4. bas 4. ‘bus’ (12th) 21. am 18. ‘M’ (17th)
5. ba:’za 5. ‘put on’ (12th) 22. an 19. ‘N’(17th)
6. baza’ 5. ‘put on’ (later)’ 23. n 20. ‘in’ (17th)
7. za 6. ‘that’ (12th) 24. un 20. ‘in’ (later)’
8. da 7. ‘down’ (13th) 25. da: 21. ‘doll’ (18th-20th)
9. at 8. ‘out’ (13th) 26. as 22. ‘S’ (18th-20th)
10. ba:’ba 9. ‘away’ (13th) 27. u: 23. ‘0’ (18th-20th)
11. ba:’ba‘ 10. ‘outside’ (13th) 28. a 24. ‘ R (18th-20th)
12. bat 11. ‘pocket’ (13th) 29. nas 25. ‘nice’ (18th-20th)
13. af 12. ‘Fuff’2 (14th) 30. na:’na 26. ‘banana’ (18th-20th)
14. faf 12. ‘Fuff’ (later)’ 31. wa 27. ‘Y‘ (21st)
15. bada’ 5. ‘put on’ (14th) 32. ats 28. ‘ X (21st)
16. bus 13. ‘push’ (15th) 33. du‘du 29. ‘Doudou’3 (21st)
17. uf 14. ‘woof’ (15th) 34. hwut 30. ‘foot’ (21st)
*[:I indicates vowel length; [‘I occurs after a stressed syllable.

I A specific date is not given, but presumably occurs later the same month.
name of family cat
name of character in French children’s story
Unfortunately, little more information is provided about the adult models

for these languages. We speculate that Norwegian was used by Velten and
probably his Norwegian parent or parents, and that French was the
language of a nanny or housekeeper. The fact that French may have
influenced Joan’s language is an important point which we will return to
later.
Joan’s development is presented in essentially three parts. First, there is a
detailed discussion of the slow acquisition of the first 30 words between the
11th and 21st months. Then, there is the presentation of all the words
acquired during a five-week period beginning in the 22nd month when the
word spurt occurred, and her language expanded from 30 to 150 words.
Lastly, a series of subsequent developments are presented in detail, with
special consideration given to the acquisition of vowel length.
Table 8.7 presents the words acquired by Joan before the word spurt at 22
months. One general aspect of this list is that it consists predominantly of
monosyllables. The few multisyllables which occur fend to be redupli-
cations, the only exceptions being [ba:’za],[baza’], and [bada’]. The mono-
syllables not only dominate, but show only four syllabic shapes: VC, CV,
CVC, and V. Another feature is the way in which the voicing of consonants
occurs across syllable positions. Voiced consonants occur prevocalically,
and voiceless ones only at the end of the word. The only exceptions to this
are [f] and [hw]. An unusual aspect of the data is the highly restricted vocalic
inventory: [a] is the only vowel for several months, and only the two-vowel
system [a] and [u] occurs at the time of the word spurt.
Velten points out some other features of these early words, for example,
how frequent words tended to resist change. He gives the example of [ba]
for ‘bottle’ which did not change to a CVC form [baz] until a large number
of CVC words had been acquired. He states (pp. 284-5): ‘from almost
every stage of the development some of the most frequently used mor-
phemes are retained without change’. Another point he makes is how words
which contain a new sound seem at first like loan words, to the extent that it
may be a long time before the sound becomes established as part of the
child’s system.
Velten’s analysis of these data is that Joan has acquired nine phonemes at
this point (21 months). He doesn’t provide much in the way of a detailed
analysis, but presents them in the discussion of the next month when Joan
underwent a noticeable word spurt. Importantly, Velten emphasizes that
this rapid growth of vocabulary did not result in any expansion of the
phonological inventory. In fact, over a five-week period, there was only one
new phoneme added, this being [h] for adult words with /h/, e.g. ‘hole’ [hu],
‘head’ [hut]. Instead, the combinations of the acquired nine phonemes
increased dramatically. For example, Velten states (p. 286): ‘the utilization
of all theoretically possible combinations of these 9 phonemes rises, for
monosyllables, from 26% to about 86%’. At least these data, then, suggest
a relation of some kind between the word spurt and phonological produc-
tivity.
Table 8.8 presents Velten’s summary (p. 286) of these nine phonemes,
their allophones, and their correspondences to English sounds during this
subsequent period of development. (Following a practice used by Smith
1973, among others, we will use 1 1 around the proposed phonemes for the
child to distinguish these from the phonemes of the adult language.) The
system is relatively symmetrical, except for /f/ where there is no voiced
allophone occurring prevocalically. It also fits the phonetic inventory
discussed earlier for the period of the first 50 words, except for the early
occurrence of [z].
Besides information on allophonic variation, the analysis also provides
the adult models for the child’s sounds, in which we find both general
patterns and ones which are relatively unusual. The more common ones are
as set out below. (We will give these correspondences names that have been
used since Velten’s paper, although at this point these are only meant as
Table 8.8 Velten’s analysis of Joan at 22 months
Phoneme Allophones English target sounds
[b] before vowels /PI ,lbl iIv-1

[p] word-final
[d] before vowels ltl,ldl~l~l,l~l~l~-l,l~-l
[t] word-final
[fl If/;hw-l ,Isw-I$-VI;
1-01 and 1-61after labials
[z] before vowels lsl,l~l,lzl,l~l;ll-l,lj-l;
[SI word-final 181 and 161 initially and after dentals
[ml Iml
[nl lnl,lrJl
[WI lwl;lr-l;l-v-l
[a1 (to be discussed)
[.I (to be discussed)
descriptive labels; when necessary, some of the examples are from the data
in later samples.)
1. Prevocalic voicing; the voicing of obstruents before vowels, /p/ to [b],
e.g. ‘pocket’ [bat]; /t/ to [d], e.g. ‘toe’ [du:]; /k/ to [d], e.g. ‘cup’ [dap]
2. Final devoicing: the devoicing of final voiced obstruents, I-b/ to [p],
e.g. ‘knob’ [nap]; /d/ to [t], e.g. ‘bad’ [bat]; /-g/ to [t], e.g. ‘egg’ [ut]; /-v/
to [f], e.g. ‘stove’ [duf]
3. Stopping: the change of fricatives and affricates into stops, e.g. /v-/ to
[b], e.g. ‘vinegar’[bidu]; /iYto [d], e.g. ‘cherry’ [dawa]; [j] to [d], ‘jam’
[dab1
4. Fronting: the production of velar and palato-alveolar sounds as
alveolars, /k/to [t], e.g. ‘duck’ [dat]; /g/ to [d], e.g. ‘goat’ [dut]; /S/ to
[z], e.g. ‘shoes’ [zus]; /il to [z], e.g. ‘rouge’ [wu:~];/C/ to [ts], e.g.
‘match’ [mats]; /j/to [daz], e.g. ‘cabbage’ [ta:budz]
5. Gliding: the changing of a liquid into a glide, /r/ to [w], e.g. ‘rock‘
[wat].
Of the five patterns just named, the first two and stopping are the most
prevalent in the data. Stopping appears to be marginal and primarily affects
the initial affricates. This process is more widespread with other children,
but Joan clearly had an ability to produce fricatives from the onset of
acquisition. Gliding is restricted to initial /r/ and not /1/ according to
Table 8.8. Subsequent discussion in the text, however, shows that the
pattern is more widespread than this, as follows:
5. Gliding (revised): also affects intervocalic /-r-/ after the vowels /a/,
/ire/, and / c / ,e.g. ‘sorry’ [sawa], ‘Mary’ [mawa], ‘carry’ [dawa]
Some other common patterns occur in the data which are not included in
Velten’s summary on p. 286, but which are mentioned later in the text. One
such pattern affects certain liquids which surface as vowels, which we will
refer to as vocalization:
6. Vocalization: the production of syllabic or nonsyllabic liquids as
vocalic elements: syllabic /r/ and /1/ as [a] or [u] (these would usually
assimilate to the vowel of the preceding syllable), e.g. ‘hammer’
[hama], ‘table’ [dubu]; postvocalic /1/ after the vowels /A/, /ad,/ E / to
[w], e.g. ‘well’ [waw], ‘shall’ [saw], ‘belt’ [bawt]
There are also cases which involve deletion of either segments or
syllables, labeled as follows:
7. Final consonant deletion: the deletion of consonants at the end of
words to preserve open syllables, /-I/ after the vowels /d, /ai/, and /awl,
(which all surface as [a]) e.g. ‘doll’ [da], ‘while’ [fa], ‘owl’ [awl, and
after all the English vowels which surface as [u], e.g. ‘peel’ [pu]; /-r/,
e.g. ‘far’ [fa]
8. Cluster reduction: the simplification of consonant clusters into single
segments. The direction of the reduction is often predictable:
(a) stop and liquid /I/ or /r/ combinations are reduced to stops, e.g.
‘bread’ [but]; ‘glass’ [das];
(b) /s/ and stop or nasal initial clusters are reduced to stops, e.g. ‘star’
[da:], ‘snap’ [nap];
(c) final nasal and voiceless stop clusters are reduced to stops, e.g.
‘bent’ [bat];
(d) final nasal and voiced stop clusters are reduced to nasals, e.g.
‘hand’ [han].
9. Unstressed syllable deletion: delete unstressed syllables, particularly
if before a stressed syllable, e.g. ‘pocket’ [bat], ‘banana’ [na:’na].
Since Joan was reasonably good at producing syllables and final consonants,
neither final consonant deletion nor unstressed syllable deletion were
prominent in the data.
Besides the above patterns, Joan also showed some correspondences
which, based on Ingram (1981a), are less common in other children
acquiring English. One of these involved the deletion of liquids. This
process, intervocalic consonant deletion ,in Joan’s speech deleted intervoca-
lic /-r-/ and 1-14when they followed any English vowel which surfaced as [u],
e.g. ‘worry’ [wua], ‘pillow’ [pua]. Two others stand out in particular and
have been the basis for later analyses, particularly in Stampe (1973). They
are: (a) the change orfricution of /j-/, /l-/, /-I-/, /-r-/ and I-t-/ (presumably a
flapped consonant) into [z], (for the intervocalic liquids, this only applies to
cases where intervocalic consonant deletion does not apply); (b) the
denasalization of final nasals as in ‘lamb’ [bap]. Denasalization occurred in
words beginning with consonants or [hu], but not elsewhere.
(8.10) presents some examples of these particularly interesting changes.
(8.10) Frication Denasalization
Ij-1 ‘yellow’ [za:‘wa] ‘broom’ [bub]
‘yard’ [za:d] ‘spoon’ [bud]
11-1 ‘leaf’ [zuf] ‘jam’ [dab]
‘lady’ [zudu] ‘train’ [dud]
1-1-1 ‘color’ [daz] ‘swim’ [fub]
‘Napoleon’ [buz] ‘room’ [wub]
1-r-I ‘tomorrow’ [maz] ‘rain’ [wud]
(later [maza]) ‘home’ [hub]
‘Harry’ [haz] but:
(later [haza]) ‘arm’ [am]
1-t-I ‘bottle’ [baz] ‘moon’ [mun]
‘water’ [waz] ‘ham’ [ham]
A last unusual pattern is the reduction of all English vowels into just two
phonetic variants of [u] and [a]. We will present the English model vowels
for these two surface vowels through the vocalic chart given in (8.11), using
the phonetic symbols from Ladefoged.
(8.11) /i/ /ul
/I/ lul
lei/ /ow/
/E/ (before stops 131 before Ill
and spirants)
/a/ /31/
__________-_____________________________----------------
/E/ (before nasals /A/ Id
and liquids)
/E/ /ai/ /a/
Generally speaking, mid and high vowels surface as [u] and low vowels as
[a]. We will refer to these two patterns as vowel raising and vowel lowering
respectively. These can be exemplified while pointing out another aspect of
Joan’s phonology. She had a small set of phonetic forms which she used for
several adult words, resulting in a large degree of homonymy for them. The
three most used forms in this way were [bat], [but], and [bu]. (8.12) gives
adult words for these four (Velten 1943: 287-8) which can also be used as
examples for the vowel correspondences in (8.11).
(8.12) [bat] ‘black, bark, spot, pocket, bought, button, bent, bite’
[but] ‘brick, pig, bead, bed, break, board, boat, bird, put, boot’
[bu] ‘boy, pea, beer, bare, pear, ball, blow, bowl, blue’
Table 8.9 Summary of patterns affecting liquids (0 = deletion)
Pattern 11-1 1-1-1 1-11 1-11 lr-I I-r-I I-rl I-rl
Frication
Gliding
after la, E , E/
Vocalization
after /A, e, cl
Final consonant deletion
except after /A, ae, cl
Cluster reduction
Intervocalic consonant deletion
Frication
(if not deleted by above)
While many of Joan’s correspondences are relatively straightforward,

those for the liquids /I/ and /r/ are particularly varied. This is not necessarily
surprising in that we have already pointed out in relation to Templin’s study
that these sounds consist of several allophones. Table 8.9 is an attempt to
provide an overview of the various forms that these sounds surface in,
where 0 indicates deletion.
Velten states that all of the above correspondences remained unchanged
for approximately ten months, except for one important series of changes.
From the 23rd to 27th months, Joan acquired the distinction between voiced
and voiceless stops. This development coincided with a series of changes in
Joan’s use of vowel length. Velten divides these developments into five
stages of change, along with substages. Table 8.10 presents a summary of
this series of changes.
We have added a stage 0 to point out the situation with vowel length
before these stages begin. At 22 months, the only long vowels which occur
are in open syllables, with only four exceptions - words which were open
but had short vowels. All vowels in closed syllables were short. The first
stage brings the introduction of surface vowel length which later permeates
Joan’s system. The important stages are really stages 2 and 3. In stage 2,
Joan introduces a surface length distinction which marks the underlying
difference between voiced and voiceless consonants. This is not surprising
in that English tends to lengthen vowels before final voiced consonants. The
next stage shows some peculiar adjustments to this tendency. As final
voiced consonants emerge, one set of previously long vowels becomes
shortened, while another short set becomes lengthened. We will refer to
these two unique patterns which go against the normal direction of vowel
lengthening in English as vowel shortening and vowel lengthening
Table 8.10 Stages in the acquisition of the voicing distinction for Joan
Velten
Stage MonthDescription of change Examples
0 22nd Only long vowels are in words with ‘down’ [da] vs. ‘pie’ [pa:]
open syllables, with four exceptions. ‘shut’ [ba] vs. ‘toe’ [du:]
1. 23rd [a:] occurs for la1 in stressed closed ‘top’ [da:p] vs. ‘cup’ [dap]
syllables
2. 24th [a:] and [u:] occur for all vowels in ‘bad‘ [ba:t] vs. ‘back’ [bat]
stressed closed syllables before voiced ‘red’ [wu:t] vs. ‘wet’ [wut]
stops and spirants ‘bead’ [bu:t] vs. ‘beat’ [but]
‘nose’ [nu:s]
3c. 24th’ [u] for Ii, ei, a., ow, ul becomes [u:] in ‘goose’ [du:~],‘lake’ [du:t]
all stressed syllables ‘coat’ [du:t], ‘hurt’ [hu:t]
3b. 24th [a:] for lhl and /u:/ for I d become ‘red’ [wud]* (from [wu:t])
shortened ‘mud’ [madI2 (from [ma:t])
3a. 24th Final I-bl and l-dl are acquired ‘meat’ [mu:t]’ vs. ‘maid‘ [mu:d] vs.
‘moon’ [mu:nl
‘hat’ [hat] vs. ‘hug’ [hadI4 vs. ‘hen’
Ihan14
but ‘iib’ [wub] vs. ‘rim’ [wub]’
final [z] is acquired for I-z, -31 ‘nose’ [nu:z], ‘rouge’ [wu:~]
‘hose’ [hu:z]
25th6 [ I ] becomes an allophone of lul before ‘foot’ [frt] vs. ‘fit’ [fUt]7
dentals ‘whip’ [fup], ‘swim’ [fub]
4. 25th the oppositions lpl-lbl, It/-Id, and ‘cold’ [tu:d] vs. ‘Joan’ [du:d]
Is/-lzl appear in initial position ‘sign’ [sa:d] vs. ‘yard’ [za:d]
5. 27thR oppositions of stage 4 also occur for ‘needle’ [nu:du] vs. ‘naked’ [nu:tu],
medial position, except for I-t-1 ‘grocer’ [du:sa] vs. ‘daisy’ [du:zu] vs.
‘letter’ [zu:zu]
I These three substages occur simultaneously about three weeks after 2.

* See stage 3a for change of [t] to [d].
See stage 3c for lengthening of [u:].
See stage 3b for shortening of [a].
Due to denasalization, final nasal vs. stop is still neutralized in some cases.
This change occurs after those of stage 3 but does not affect the lengthening rules of stage
3.
This form is not attested in the data but it is hypothetically possible; we have placed it here
to show the surface contrast that now occurs.
This change takes over three months to spread throughout the lexicon.
respectively. The result is a four-way system of vowels which, Velten

claims, lasted for nine months. Velten summarizes this system as follows:
[a] for (a) /ae, 3, ai/ before voiceless consonants;
(b) /E/ before liquids, nasals, and /-t-/;
(c) /A/ in any position;
[a:] for (a) /ae, 3, ai/ before voiced consonants;
(b) /a/ in any stressed syllable;
[u] for (a) /E/ before stops and spirants;
(b) /I, u/ in any position;
[u:] for (a) /d/after initial labial;
(b) /i, ei, ow, a * , u, 31, IU/in any stressed syllable.
These, then, are the major patterns discussed by Velten for the acqui-
sition of his daughter’s phonology. The only other changes briefly discussed
are the loss of denasalization during the end of the third year, and the
eventual development of other vowels. Velten’s data show both the general
aspects of early phonological acquisition during this period and the pecu-
liarities that can be found in data from individual children. We have
identified nine common patterns in the data - common in the sense that they
can be found to some extent in the language of most other children
acquiring English. We have also pointed out several of Joan’s very unusual
patterns. Four of these affect vowels (vowel raising, lowering, shortening,
and lengthening), and two affect consonants (frication and denasalization) .
All of the patterns are striking in the detail in which they are found.
While Velten states that his data are in support of Jakobson’s theory, he
actually spends little time trying to justify this claim. Certainly Joan’s early
system does not match up particularly well with Jakobson’s first stage of
acquisition given in Table 6.23 on page 192. Nor does Velten attempt to
show how Joan’s complex patterns support Jakobson’s claims about an
invariant sequence of contrasts. What we have here, then, is a detailed set
of descriptive data which need to be explained: later I will examine a major
explanation for these data, proposed by Stampe (1969, 1973).
Leopold (1947) I have focussed on Velten’s data because of its detail and
uniqueness. Before turning to the issue of explanation, two further impor-
tant diary studies will be briefly considered; along with Velten, these
constitute the three most frequently cited sets of data.
The next significant diary study on an English-learning child was that of
Leopold on his daughter Hildegard. The report on Hildegard’s phonologi-
cal development appeared in 1947 in the second volume of Leopold’s four
volume work, subtitled ‘Sound learning in the first two years’. Like Joan,
Hildegard was not exposed just to English; she was acquiring both German
and English simultaneously, and Leopold reports data for both languages.
Her English data have been widely discussed, and her acquisition of her first
50 words has been analyzed in the study by Ferguson & Farwell (1975)
already discussed in Chapter 6.
Leopold reports on Hildegard’s development from the first words around
0;8 to the end of 1;11, with some selected remarks on later development. As
reported in Ferguson & Farwell (1975), the period of the phonology of the
first 50 words lasted up to 1;5 when a word spurt occurred. By the end of
1;11, Hildegard had acquired 400 words in her spoken language. Her
development of phonology was similar to Joan’s in the sense that she first
acquired a small inventory of sounds, then expanded the combinatory
potential of these during the word spurt. In other words, her development
from 1;6 to 1;11 was not marked by the addition of new sounds as much as
by increased use of the ones she had already acquired.
In Ingram (1985d), there was a brief report on Hildegard’s phonetic
inventory at 1;11, using the method presented in Chapter 6. This inventory
was as follows:
(8.13) Word-initial consonants Word-jinal consonants
m- n-
b- d- -t -k
W- h- -S
This is a strikingly small consonantal inventory for the size of her vocabu-
lary at this age. It is even small when compared with the English phonetic
inventory presented in Chapter 6 for the end of the phonology of the first
50 words. An important general aspect of Hildegard’s acquisition, then, is
a marked delay in phonetic acquisition compared with her lexical acqui-
sition.
One consequence of Hildegard’s slow phonetic development was the
extensive existence of homonymy, i.e. the use of a phonetic form for two
or more adult words. We pointed out that Joan also had extensive homo-
nymy, but it appeared to be tied to the use of a small number of phonetic
forms. Leopold’s list of Hildegard’s homonymous words (pp. 231-4) indi-
cates that it extended throughout her lexicon. In Ingram (1985d), Hilde-
gard’s use of homonymy was measured and compared with that of two
other children. The measure used was the proportion of homonymous
types in each child’s lexicon. A homonymous type is a word (or lexical
type) which has at least one phonetic type which is homonymous with some
other word in the lexicon. Some examples from Hildegard’s diary are [ba]
for ‘ball’, ‘box’, ‘block’, and ‘Paul’, [hat] for ‘hot’ and ‘hat’, and [titi] for
‘cookie’ and ‘sticky’. While homonymy decreased over time for the other
two children, it actually increased for Hildegard. For example, her propor-
tion of homonymous types was approximately 0.15 at 1;7 compared to over
0.40 at 1;ll. In other words, by 1;11 nearly every other word of hers was
homonymous with at least one other word.
Despite some peculiarities in individual lexical items, Hildegard’s substi-
tution patterns are by and large restricted to the set of common ones men-
tioned in the discussion on Joan’s language. (8.14) is a summary of these
with some examples for each pattern.
(8.14) Some phonological patterns of Hildegard Leopold
1. Prevocalic voicing: e.g. ‘Paul’ [ba], ‘toast’ [dok]
2. Final devoicing: e.g. ‘bug’ [bok]
3. Stopping: e.g. ‘Joey’ [do.i], ‘June’ [du]
4. Fronting: e.g. ‘Jack‘ [da]
5. Gliding: e.g. ‘lie7[jail, ‘rock’ [w3k]
6. Vocalization: ‘bottle’ [balu], ‘ball’ [bawl
7. Final consonant deletion: ‘poor’ [pu], ‘big’ [br], ‘comb’ [do],
‘June’ [du]
8. Cluster reduction: ‘stick’ [dik], ‘brush’ [baJ]
9. Unstressed syllable deletion: ‘pocketbook’ [babul
The peculiarities which occur in Hildegard’s data are restricted primarily
to the acquisition of voicing and the fricatives. The early data show the use
of both [p] and [b] word initially, with a later shift to only [b] with prevocalic
voicing of [p]. As Leopold points out, however, this was the consequence of
what he calls the ‘whispering stage’ (p. 208) which characterized Hilde-
gard’s first months of producing certain of her words. We have already
mentioned one example of this in 6.5.4, the word ‘pretty’. The [p] in these
words became [b] once the vowels also were pronounced. What initially
appears as a regression in development, then can actually be traced to
Hildegard’s tendency to whisper selected words.
Hildegard is noticeably different from Joan in her acquisition of frica-
tives. First, the only one in her phonetic repertoire is [J]which occurs finally
for /-f,/-s,/-z,/-~,/-~/.
This sound not only occurs as a final substitution, but
also as part of a preferred pattern for some words with a fricative elsewhere,
e.g. ‘stone’ [dorJ], ‘story’ [131J]. When we examine her substitution patterns
for initial fricatives, we find that there are very few words even attempted
with them. That is, as pointed out by Ferguson & Farwell (1975), Hildegard
appears actively to avoid attempting fricatives. The one exception is /f-/,
which becomes [w]. Interestingly, this is a substitution that was not found in
any of the 73 subjects studied in Ingram et al. (1980). It is a substitution
which suggests some influence from her simultaneous acquisition of
German.
Smith (1973) The last diary study to be reviewed here is Smith (1973) on
his son Amah1 (or A as he is referred to in the book). Like Joan and
Hildegard, A was raised in a multilingual environment. His father is British,
and his mother Indian. The mother spoke Hindi, Bengali, Marathi, and a
dialect of English referred to as Standard Indian English. Smith states that
A was exposed to Hindi and Marathi, but was still raised primarily as a
monolingual English speaker.
A started speaking late, around 20 months of age. Smith began his diary
observations somewhat later, when A was 2;2, and continued them until
3;11. The data are divided into 20 stages for purposes of analysis. At the
onset of the study, A had already acquired over 200 words. The study,
therefore, provides more data on later acquisition than do either Velten or
Leopold. All the data collected by Smith are conveniently given in
alphabetical order in an appendix at the end of the book. Given this, it is
somewhat surprising that more analyses of these data have not been done
since their first appearance.
Smith’s goal is to compare two kinds of analyses of A’s data. One analysis
assumes that the child is operating with the adult surface phonology of the
language, and that the child’s phonology is a set of realization rules which
map from the adult surface form to the child’s output. This position, then,
assumes the complete perception theory discussed on page 357, and that
the child’s competence is greater than what is seen in his performance. The
other analysis is one which treats the child’s system as independent from
that of the adult one, where the child moves ‘from a more idiosyncratic and
simple system to one which was more complex and more closely isomorphic
with the system of the adult language’ (p. 1). This alternative assumes
something like the global perception theory, and a universal sequence of
development which will determine the child’s acquisition order indepen-
dent of the adult language. The child’s performance is seen as virtually
isomorphic with his competence.
Smith considers these alternative theories by analysing the stage 1sample
separately for each one and following the changes that occur over time. The
rules for each analysis are presented in formal notation following the
general conventions of Chomsky & Halle (1968). Smith concludes that the
first analysis, which proposes realization rules, accounts for the data better
than the second. Here we will not go into detail over the arguments for or
against each analysis. It seems to us that the second analysis (or theory) in
its most general form is somewhat of a straw man. The only theories which
predict idiosyncratic and autonomous development (possibly such as that in
Ferguson & Farwell) are typically only for the phonology of the first 50
words, not for children as advanced as A. While Smith suggests Jakobson as
one such theory, we have already discussed in Chapter 6 how Jakobson sees
the influence of the adult language as always present and asserting a role
very early in acquisition. Smith’s observations are in fact consistent with
many proposals on phonological acquisition, including the one by Stampe
which we will examine in the next section.
We can get a feel for A’s data and Smith’s proposed realization rules by
looking at Smith’s analysis for the first sample (or stage 1). Table 8.11
presents a summary of this analysis along with some examples. We
Table 8.11 A’s realization rules at 2;2, taken from Smith (1973: 13-22)
1. A nasal consonant is deleted before any voiceless consonant, e.g. ‘stamp’ [dcp], ‘bump’
[bnp], ‘uncle’ [ngu].
2. A voiced consonant is deleted after a nasal consonant, e.g. ‘window’ [winu:], ‘mend‘
[men], ‘angry’ [cgi:], ‘finger’ [wina].
3. The alveolar consonants In/ and It,dI (but not continuants /s,z/) become velars, [rj] and [g],
before a syllabic [I], e.g. ‘handle’ [cr~u],‘pedal’ [bcgu], bur ‘whistle’ [wibu].
4. Syllabic [I] vocalizes to [u], e.g. ‘apple’ [cbu], ‘nipple’ [mibu].
5. A continuant consonant preceded by a nasal and a vowel, itself becomes a nasal
(optional), e.g. ‘noisy’ [ns:ni:], ‘smell’ [men], but ‘nice’ [nait], ‘mice’ [mait].
6. /I/ is deleted finally and preconsonantally, e.g. ‘ball’ [bs:], ‘bell’ [bc], ‘bolt’ [bs:t], ‘milk’
[mik].
7. /s/ is deleted preconsonantally, e.g. ‘spoon’ [bu:n], ‘swing’ [wig].
8. In ClwVC2 syllables, C2 assimilates to [p], [m], or [f], e.g. ‘squat’ [gsp], ‘queen’ [gi:m],
‘twice’ [daifl.
9. In sVC words, the /s/ is optionally deleted if the C is labial or alveolar, e.g. ‘soup’ [u:p],
‘seat’ [i:t], ‘sun’ [nn] or [dnn], but ‘sing’ [gig].
10. In J VC words, /$/ is optionally deleted if the C is labial or velar, e.g. ‘sharp’ [a:p], ‘sugar’
[ u p ] , but ‘shopping’ [wsbin], ‘shine’ [dain], ‘shirt’ [da:t].
11. IzI but not lslis deleted finally, e.g. ‘eyes’ [ai], ‘nose’ [nu:], ‘glasses’ [ga:gi:], but ‘kiss’ [gik],
‘mice’ [mait].
12. A nasal consonant following an unstressed vowel becomes alveolar, unless there is a
preceding velar consonant, e.g. ‘bottom’ [bsdin], ‘driving’ [waibin]; but ‘singing’ [gigin].
13. /h/ is deleted everywhere, e.g. ‘hair’ [E], ‘head’ [ct].
14. An initial or post-consonantal unstressed vowel is deleted, e.g. ‘away’ [we:], ‘banana’
[ba:na], ‘belong’ [bsn].
15. It/ and /d/ are optionally deleted before /r/, e.g. ‘driving’ [waibin], ‘troddler’ [Isla], ‘trolly’
[Ish:] (only cases found).
16. Post-consonantal sonorants /l,r,w,j/ are deleted, e.g. ‘play’ [be:], ‘brush’ [bnt], ‘new’
(/nju/) [nu:].
17. Non-nasal alveolar and palato-alveolar consonants harmonize to the point of articulation
of a preceding velar, e.g. ‘kiss’ [gik], ‘coach’ [go:k], but ‘skin’ [in].
18. /l,r,i/ (a) are neutralized as [I] _ - when they are the only consonants in the word, e.g. - ‘lorry’
[~s~i: I,. ‘yellow’ [MU: I;
(b) become [w] or are deleted when intervocalic, e.g. ‘telephone’ [dewi:bu:n], ‘follow’
[wswo:];
(c) elsewhere are neutralized as [d], e.g. ‘light’ [dait], ‘write’ [dait], ‘yes’ [dct].
19. Alveolar and palato-alveolar consonants harmonize to the point of articulation of a
following consonant, obligatorily if that consonant is velar, optionally if it is labial, e.g.
‘dark’ [ga:k], ‘snake’ [rje:k], ‘knife’ [maip], ‘nipple’ [mibu], but ‘stop’ [dsp], ‘drum’ [dnm].
20. /f, v/ become [w] prevocalically, e.g. ‘feet’ [wi:t], ‘fire’ [wae:].
21. Post-consonantal alveolar consonants are deleted, e.g. ‘empty’ [bi:].
22. Alveolar consonants are optionally deleted in final position, e.g. ‘moon’ [mu:], ‘open’
[ubu:] .
23. & 24. Alveolar and palato-alveolar fricatives, and affricates become alveolar [t] or [d], e.g.
‘bus’ [bnt], ‘zoo’ [du:], ‘brush’ [bnt], ‘church’ [da:t], ‘John’ [dm].
25. Prevocalic consonants are voiced and final ones are voiceless, e.g. ‘teddy’ [dcdi:], ‘kiss’
[iikl.
recognize many of the general patterns seen earlier in the data from Joan
and Hildegard. For example, rules 1,2,6,7,15,16,21, and 22 are all rules
of consonant cluster reduction. Smith’s analysis, however, makes strong
claims about the nature of such patterns. First, Smith argues that these rules
are formally separate from one another and strictly ordered. For example,
rule 11 which deletes final /-z/ must occur before rule 17 or else words like
‘keys’ would be [gi:k] instead of [gi:]. A complex set of such orderings is
proposed. The general patterns, therefore, are not psychologically real
rules themselves, but rather functional similarities that realization rules
share. The child’s actual rules will be a rather long and complex list of
realization rules which carry out a small number of functions. Smith
proposes that realization rules operate to perform one of the following four
functions (p. 162): (i) a tendency toward vowel and consonant harmony
(rules 3, 5 , 8, 12, 17, Ma, 19); (ii) cluster reduction (rules 1, 2, 6, 7, 15,
16, 21, 22); (iii) systematic simplification (replacing marked sounds with
unmarked ones such as replacing fricatives with stops - rules 13,14,18b,c,
20, 23, 24, 25); and (iv) grammatical simplification. The last function does
not operate in the first sample but is found later when rule 11 becomes
restricted to ungrammatical morphemes, e.g. ‘nose’ [nu:], ‘cheese’ [di:] vs.
‘pages’ [be:did], ‘peas’ [pi:d] (see Smith 1973: 67-70).
In attempting to describe A’s phonological development, Smith states at
the onset that he expected his task to be to account for an identifiable set of
regularities in the child’s spoken language. He says (p. 1): ‘It soon became
apparent that there were a wider range of phenomena . . . to be explained
. . . In addition to the mere fact of regularity, any theory of acquisition must
be able to account for at least the following classes of data observed in the
acquisitional process.’ He goes on to identify seven such phenomena, each
of which will be briefly examined here.
The first of these phenomena is the occurrence of systematic exceptions to
the general regularities. Exceptions can occur with individual lexical items,
but they can also apply to a group of words. We have already seen cases of
this, e.g. Joan’s lack of final vowel lengthening in the open syllables of
‘down’ and ‘shut’. The realization rules above show several kinds of
exceptions also. For example, rule 9 deletes initial /s/ in CVC words, yet it is
optional and does not apply if the .final consonant is a velar. Both the
optional forms like ‘sun’ [ d ~ nand
] velar cases like ‘sing’ [gig] are exceptions
to the generality of the rule.
There are probably several different reasons why exceptions arise. Smith
in his discussion of this point suggests one cause in particular, which he calls
restructuring (or relexicalization). He gives an example which arose when
the assimilation rule 19 dropped out of A’s system (p. 144). While words
like ‘duck’ changed from [gnk] to [ d ~ k ] the , word ‘take’ remained as
[kherk]. This also occurred in derived forms like ‘taking’ and ‘taken’ which
were [kerkin] and [kukan] respectively. Smith proposes that this is the result
of the child changing or ‘restructuring’ the internal lexical representation of
his word from that of its adult surface form to that of the child’s produced
form. We have seen a related set of facts earlier from Joan Velten, where
earlier acquired words resisted new sound changes. Whether we accept
Smith’s restructuring proposal or some other account, such exceptions need
to be accounted for in some way.
The second phenomenon raised by Smith is one which was also empha-
sized by Velten (1943: 288):
The child’s phoneme system and the phonetic structure of English are
obviously non-congruent. That is to say, not only does almost every
unit of the child’s system render several English phonemes, as might
be expected, but, conversely, the phonetic variants of a given English
unit may appear as different phonemes in the child’s language.
I will refer to this phenomenon as non-congruence, using Velten’s
terminology. Smith gives the examples of A’s [d] representing 13 different
English phonemes, e.g. ‘teeth’ [di:t], ‘zoo’ [du:], ‘chair’ [dr:],yet English Is/
surfacing in four different forms, e.g. ‘sun’ [m], ‘sock‘ [g3k], ‘mice’ [mait],
and ‘whistle’ [wibu].
The third phenomenon concerns the observation that changes in the
child’s phonological system usually occur across-the-board. That is, when
the child begins to change a particular correspondence between his sounds
and those of the adult language, the change usually maintains a systematic
correspondence that is narrower than the earlier one. For example, A at
one point produced /w/ and If/ as [w]. When A began to acquire the sound
[f], it was consistently used for If/ only, not for both /w/ and If/. Such facts
suggest that the child had perceived the adult distinctions, but had merged
them in his own speech. This kind of observation has been used to argue
against the global perception theory cited earlier. It is important to point
out that across-the-board change does not imply that the change is
instantaneous. It only refers to the correspondences that occur as a sound
emerges in the child’s system.
The next three phenomena are the occurrence of non-native sounds,
puzzles, and recidivism. Smith observed cases where A used sounds which
are not part of the phonetic inventory of English. Some examples are ‘slug’
[ h g ] with a lateral fricative, and ‘Smith’ [mis] with a voiceless nasal. The
term ‘puzzle’ is taken from the example Smith provides in which A
produced ‘puzzle’ as [ p ~ d a l ]but ‘puddle’ as [p~gal].That is, A could
articulate a medial [d] but not in ‘puddle’ where it is required. This is a
general observation pointed out long ago by Jakobson as evidence against
a strictly articulatory account of phonological acquisition. Recidivism is
Smith’s term to refer to ‘the loss of a contrast which has already been
established’ (p. 4). He gives the example of the words ‘side’ and ‘light’
changing over three time periods, reproduced here in (8.15).
(8.15) Time 1 Time 2 Time 3
‘side’ [dait] [dait] [lait]
‘light’ [dait] [lait] [lait]
Superficially at least, the child appears to have taken a step backward at
time 3 in the sense that a contrast between two words has actually been
lost.
The last phenomenon which Smith discusses is one he observed through
a clever experiment with his son. Smith tape-recorded A speaking and
played it back to him. When he heard himself, A was able to recognize his
reduced English words. For example, when A heard himself say ‘squeeze’
[gip], he responded by squeezing a piece of cloth. This ability was also
shown when Smith would say the words to A, e.g. Smith: ‘What’s a [sa:t]?’,
A points to his shirt, and Smith: ‘What’s a [su:]?’, A points to his shoe
(p. 136). Importantly, A not only recognized his reduced forms but was
aware that these were different from those of the adult language. For
example, when Smith pressured A to say ‘ship’, A responded ‘No. I can
only say [sip]’ (p. 137). Smith refers to this situation as the ‘child’s ability to
understand his own speech’ (p. 5). There were two limitations on this
ability. First, it only occurred when A was currently producing the word in
the reduced form he heard. Once his pronunciation improved, he would
not recognize his earlier form. Second, it would only occur if there wasn’t
an adult form that was homophonous with A’s reduced form. For example,
A pronounced ‘mouth’ as [maus] yet processed this as if it meant ‘mouse’
when Smith tested it in the above way.
Assessment. The development of two of the children in the phonological

diaries discussed above, Joan and A, was presented in some detail to
provide an idea of the degree of complex and individual patterning that is
found. They were not presented, however, to suggest in any way that they
are typical of how English children acquire their phonology. In fact, it is
likely that they are atypical in the sense that published diaries are usually
on children who are particularly interesting in their language acquisition;
for example, two of these children, Hildegard and A, both pronounced /f/
as [w], a pattern that was not found in any of the 73 children tested in
Ingram et al. (1980).
Accounting for phonological acquisition is a formidable task. We not
only need to explain the patterns or rules that children follow, but also the
kinds of phenomenon described by Smith. In the next section, we turn to
the issue of developing a theory of phonological acquisition.
8.2.3 The explanation of phonological acquisition

Although it is rare to see comparisons between syntactic and phonological
theories of acquisition, it is obvious that the same general issues exist for the
two. In our review of syntactic acquisition in Chapter 7 we touched upon
several of these issues, such as continuity vs. discontinuity and the relation
between the child’s competence and performance. The formulation of these
issues in the area of phonological acquisition will first be discussed, using
Jakobson’s theory as a basis of comparison. Afterwards, some specific
alternative proposals will be offered.
The issue of continuity matters when it is claimed that the child begins
adult-like organization of his phonological system. Jakobson’s theory
proposes that the child begins phonological acquisition with oppositions and
the use of distinctive features. At any point in acquisition his phonology will
contain a subset of the adult features. As such, it is a theory that is based on
continuity and in a very general way is comparable to Pinker’s view of
syntactic acquisition. This view is also characteristic of the theoretical
positions in Stampe (1969, 1973) and Smith (1973). An alternative view is
the claim that the child undergoes one or more discontinuities or reorgani-
zations during the course of acquisition. Probably the clearest statement of
such a position is Macken (1979) who claims that the child begins acquisition
with the word as the basic organizational unit, then switches to the
phoneme, and lastly to the feature. There are, then, at least two reorgani-
zations. Macken’s position appears to be constructionist in the sense of
Chapter 4,in that presumably the word and phoneme still play a role in the
child’s system when it goes through the final reorganization. (We say
‘presumably’ because Macken does not refer to a particular model of
phonology that the child is acquiring, so that the status of the earlier units
after reorganization is not clear.) Recall that the constructionist view does
not deny reorganization, but only restricts its extent. Lastly, we could have
total discontinuity in the sense that the child’s system up to some point bears
little relation to the adult system. We are unaware of anyone who takes this
view for phonological acquisition.
The issue of the relation between competence and performance translates
in phonological acquisition to the issue of levels of representation. There are
a number of positions on this point, recently summarized in Stoel-Gammon
(1986). Here I will simplify the possibilities to just two. One position is that
there are only two levels of representation: the child’s perceptual organi-
zation of the adult language (presumably an accurate phonemic represen-
tation of its surface structure) and the child’s phonetic representation or
output. There are then a set of realization rules (Smith 1973) or processes
(Stampe 1968, 1973) which map from the underlying representation to the
phonetic representation. This position argues that the child’s perception is
essentially adult-like at the onset of acquisition, and that there is a
considerable discrepancy between the child’s competence and her per-
formance.
The alternative to this view is that there is some intervening level (or
levels). Elsewhere (e.g. Ingram 1976a,b) this has been referred to as the
organizational level, where the child can be said to have an underlying
representation related to, but distinct from, the perceptual representation
of the adult word. The mapping from the organizational level to the
phonetic one is relatively close, so that in this view the child’s performance
is considered to be much closer to her competence.
The contrasting claims of these two viewpoints can be examplified by
looking at stage 2 in Table 8.10 for Joan Velten. In (8.16) I show how each
viewpoint would represent Joan’s words ‘bead’ and ‘beat’.
(8.16) Two Levels Three Levels
Adult form: /bit/ /bi:d/ /bit/ /bid:d/
(bit,dl (bi:t,d(
Child form: [bit] [bi:t] [bit] [bi:t]
In the two-level approach, the child’s underlying form would be the adult
phonemes, so in this case the child would have an underlying final voicing
distinction. There would then be a set of rules which map between the two.
The three-level approach, which is the one implied in Velten’s analysis, is
that the child may perceive the difference in final voicing (although this is
not necessary), but does not use it phonemically in his system (see Table
8.8.). For example, there is just one phoneme (t,d(with the [d] occurring
prevocalically and the [t] finally. Instead, vowel length is used to show a
phonemic distinction. The level indicated by 1 1 is the organization level
referred to above. Importantly, this level is not independent of influence
from the adult form. As pointed out by Jakobson, this level is constantly
under pressure of the ‘unifying force’ to conform to the adult form, which it
eventually does.
In the example just given, the influences on the child’s underlying
representation in the three-level approach come from the adult form. Some
versions of this general model also incorporate influences from the child
form (e.g. Macken 1979; Menn 1983). For example, we saw in the
discussion of Hildegard’s fricatives that the syllabic shape CV[J] was a
common form, e.g. ‘kiss’ [diJ], ‘knife’ [naiJ], ‘much’ [maJ], which also
extended itself to forms like ‘stone’ [do~J]and ‘story’ [brJ], distorting the
normal pattern of mapping from the adult form to that of the child’s.
Macken (1979) refers to these as ‘word patterns’, and gives an example from
a Spanish child referred to as Si who used the word pattern [p/b V t/d VI.
This preference to produce labial consonants before dental ones has been
reported elsewhere (Ingram 1974b). This pattern created some .radical
discrepancies between the adult form and Si’s when the order of consonants
in the adult words was different, e.g. ‘sopa’ [p’waet’a].This would be an
example of the ‘particularist spirit’ mentioned by Jakobson. The child’s
system at the organizational level in the three-level approach can be seen as
the result of the influence both of the adult form and child form.
We have interpreted Jakobson’s theory as one which recognizes three
levels, even though he never discussed it himself in those terms. This has
also been Velten’s and others’ interpretation, although it has sometimes
been misinterpreted to the extreme position that the organizational level
bears no relation to the adult model. (It is our impression that this
misinterpretation can be traced to Chao 1951, who claimed that his niece
Canta was not speaking Mandarin but her own form of it, which he referred
to as the Cantian idiolect.) For example, Stampe (1969: 446) states: ‘Most
modern students of child phonology have assumed that the child has a
phonemic system of his own, distinct from that of his standard language.’
This same view led Smith (1973) to his attempt to analyze A’s phonology as
an autonomous system without reference to the adult target sounds. We will
refer to this position as ‘Chao’s theory’ to distinguish it from Jakobson’s
which recognizes the importance of the adult language (see Ingram 1988 for
an elaboration of this point). The disagreement with Chao’s theory in the
late 1960s led to two well-known attempts to argue for a two-level model
(Stampe 1969; Smith 1973). Here we will look at the theory developed by
Stampe which has had a great impact upon research.
Stampe (1969, 1973) Stampe, like Jakobson, approaches the explanation

of phonological development from the orientation of Language Acqui-
sition. He has developed a theory of phonology which he refers to as
Natural Phonology, and uses data from acquisition and language change to
support it. Publications on Natural Phonology, however, have not been
abundant. Its first cited presentation is Stampe (1968), a paper which, to
my knowledge, has never been published. The most cited paper is Stampe
(1969), a brief work of only 12 pages. Stampe’s dissertation in 1973,
circulated at the time with the title ‘How I spent my summer vacation’, is ‘a
random selection of topics in phonology that occurred to me as I wrote’
(p. i). Stampe refers to a book Natural Phonology as forthcoming, but it
also has not appeared. There have been some papers since on specific
topics, e.g. Miller (1972), a thesis (Edwards 1970), a dissertation (Donegan
1986), and a more recent review (Donegan & Stampe 1979). The theory has
not been widely accepted among phonologists but it has had considerably
more success among researchers in language acquisition.
As stated earlier, Stampe sees perceptual development as far in advance
of production, with the child having an accurate phonological represen-
tation of the adult surface form of words. The child’s reduced productions
are the result of his applying an innate phonological system (Stampe 1969:
446) to the phonological representation. The innate phonological system
consists of a set of universal phonological processes which simplify the
phonological representation. These are like Smith’s realization rules except
that Stampe attempts to delimit the range of possible processes. It is
important to understand Stampe’s use of the term ‘phonological process’:
A phonological process merges a potential opposition into that
member of the opposition which least tries the restrictions of the
human speech capacity. (1969: 443)
A phonological process is a mental operation that applies in speech to
substitute, for a class of sounds or sound sequences presenting a
specific common difficulty to the speech capacity of the individual, an
alternative class identical but lacking the difficult property.
(1973: 1)
Although substitutions are mental in occurrence, their purpose is
clearly physical: to maximize the perceptual characteristics of speech
and minimize its articulatory difficulties. They are mental operations
performed on behalf of the physical systems involved in speech
perception and production. (1973: 9)
As such, Stampe’s theory is closest to what we referred to on page 357 as the
complete perception theory I.
Stampe does not provide a list of the universal processes, which he admits
is a formidable task, but does discuss several potential ones. One example
concerns the data presented above in (8.16). Stampe’s analysis for these
data (1969: 448) is given in (8.17).
(8.17) Adult phonetic form: [bit] [bi:d]
Child phonological representation: /bit/ /bid
length adjustment: bit bi:d
final devoicing: bit bi:t
Child form: [bit] [bi:t]
The child formulates a distinction between voiced and voiceless final
consonants in her phonological representation. Two processes then simplify
this representation: length adjustment, which lengthens vowels before
voiced consonants, and final devoicing of obstruents at the end of words.
Another example is Stampe’s analysis of Joan’s substitution of [z] for Ill.
Rather than proposing one process of frication, Stampe gives three pro-
cesses: [l] to [j] by delateralization; [j] to [g] by spirantization, and lastly [3]
to [z] by depalatalization.
(8.18) Adult phonetic form: [lif] [jard] [rug]
Child representation: /lif/ /jardl /rug/
delateralization: j
spirantization: 3 3
depalatalization: Z Z Z
other processes: U a: wu:
Child form: [zuf] [za:d] [wu:~]
Evidence for the separate processes for /1/ to [z] is that each one is
independently supported by data from other words. ‘Rouge’ provides
evidence for depalatalization and ‘yard’ for spirantization.
Besides showing three processes instead of one, the example in (8.18)
shows the processes applying in a particular order. At the onset of
acquisition, the order in which processes occur is ‘sequential application
without ordering’ (1973: 60), that is, they continue to apply in any order
until as many have applied as possible. The full range of processes, then, is
only seen at the very onset of acquisition. The child’s task in acquisition is to
overcome the application of the entire innate set: ‘The child’s task in
acquiring adult pronunciation is to revise all aspects of the system which
separate his pronunciation from the standard’ (1969: 444).
The child reduces the application of phonological processes in three
possible ways: (i) ordering, (ii) limitation, and (iii) suppression. These are
three mechanisms of change which operate upon the innate phonological
system.
Ordering refers to imposing a restriction on the natural order of applica-
tion of the processes just mentioned, such that the application of one
process eliminates one of the contexts in which another process normally
applies. (8.19) gives an example of this in the ordering of the processes of
depalatalization and despirantization ([dg] to [d]) and prevocalic voicing
([tJ] to [dg]) by Hildegard at 19 and 20 months (1969: 447).
(8.19) 19 months: ‘choo-choo’ ‘Joey’ 20 months: ‘choo-choo’ ‘Joey’
[tJl to id31 d3 d3 [d31 to [dl d
[dgl to [dl d d d [tJI to [dgl d3 d3
[dudu] [do:i] [dgudgu] [do:i]
The ordering of the prevocalic voicing process before depalatalization and
despirantization restricts the natural order of application, but brings the
child’s pronunciation closer to that required by the language.
A second mechanism of change is limitation, by which the child limits the
range of segments or contexts in which a process applies. A crucial aspect of
this mechanism is that each process contains ‘various subtle and strict
hierarchies, ranging from the greatest generality which is phonetically
motivated, to the complete suppression of the process’ (1969: 443). For
example, prevocalic voicing may be more favored by stops than fricatives,
or more likely intervocalically than initially. We saw an example of the first
situation from Joan: all stops were prevocalically voiced while the pattern
was more restricted for the fricatives. Stampe (1969: 447) gives two
examples of the latter case: (i) Hildegard’s development of ‘papa’ as [baba]
to [paba] and finally [papa], and (ii) Joan’s distinctions of ‘paper’ [pu:bu]
vs. ‘baby’ [bu:bu] at 25 months, and ‘puppy’ [papu] vs. [pabu] ‘probably’ at
27 months.
The third kind of change is suppression of a process, or sequence of
processes. In (8.18) the series of processes proposed for Joan’s substitution
of /1/ by [z] was shown. As pointed out by Stampe, it is not actually very
common to find children showing each of the potential processes in a
substitution. Another child, however, might begin acquisition with the
pronunciations of ‘leaf’ [jif], ‘yard’ [jad] and ‘rouge’ [wu]. We would then
say that the processes of spirantization and depalatalization have been
suppressed. The claim that individual children may suppress and limit
processes in different degrees is one of Stampe’s major ways of accounting
for individual variation in acquisition.
Stampe’s processes tend to be more specific than most of the patterns
identified here in Joan’s speech, but less so than some of Smith’s realization
rules. The reason for this is his concern to propose only those processes
which can be identified as natural. Stampe’s work, then, goes beyond that of
Velten and Smith by trying to explain the specific patterns found in the
speech of children.
This point can be demonstrated by looking at one of Smith’s rules for A.
As descriptive accounts of A’s behavior, several of A’s rules do not look
particularly natural. One such rule is rule 9, which deletes /s/ optionally for
IsNC words which end in labials and dentals, e.g. ‘soup’ [u:p] and ‘seat’
[i:t], but not velars, e.g. ‘sing’ [gig]. This does not have any apparent natural
motivation in terms of the restrictions of the human speech capacity.
Stampe would presumably identify at least the following interacting pro-
cesses: despirantization ([SI to [t]), prevocalic voicing ([t] to [d]), prevocalic
obstruent deletion (of [d]), and velar assimilation of coronals (or simply
velar assimilation, [d] to [g]). Stampe’s explanation of A’s data could take
one of at least two forms. One possible analysis would be that the process of
prevocalic obstruent deletion has been limited to just those contexts where
there are following labials and dentals. An alternative is that the process of
velar assimilation is now ordered before prevocalic obstruent deletion.
Importantly, these alternatives make different predictions about acqui-
sition in other children. If the limitation analysis is correct, then we should
find children who have ‘soup’ [u:p], ‘seat’ [i:t], and ‘sing’ [dig]. If the
ordering analysis is correct, then the preceding data should not be found,
and the proposed limitation would not be part of the hierarchical structure
of the prevocalic deletion process.
At first glance, Stampe’s phonological processes are similar to Jakobson’s
laws of irreversible solidarity. For example, the law that fricatives imply
stops (see Table 6.24 of Chapter 6) is another way to say that there is a
natural process whereby obstruents become stops. Stampe recognizes this
similarity and points out that Jakobson’s laws are a subset of phonological
processes. This subset consists of those processes which are context-free
and which operate on the phonemic inventories of languages. Stampe
argues, however, that there are other processes which also have to be
accounted for. These are the context-sensitive processes which are com-
monly in conflict with context-free ones. The three mechanisms of change
are needed to resolve such conflicts.
An example of such a conflict is that between the context-free process of
vowel denasalization and the context-sensitive process of vowel nasali-
zation. One of Jakobson’s implicational laws is that the existence of nasal
vowels in a language implies the existence of corresponding oral vowels.
The corresponding phonological process is vowel denasalization, which
states that nasal vowels are changed into non-nasal vowels. Evidence for
this process comes from acquisition studies on languages with nasal vowels
(such as French) which report that the child’s first attempts at nasal vowels
are oral ones (e.g. Jakobson 1941/68: 71-2). Such processes ‘respond to the
inner complexities of single segments’ (Stampe 1973: 20). At the same time,
there is the context-sensitive process of vowel nasalization which nasalizes
vowels next to nasal consonants, e.g. English ‘can’t’ [kgt] (see Ladefoged
1975: 81). Such a process responds ‘to the complexity of sequences of
segments’ (1969: 21). Importantly both processes often coexist in a lan-
guage, as in English where vowel nasalization exists at the same time as
vowel denasalization occurs in loan words. Our theory, therefore, needs to
account for the existence of both types of processes.
In discussing the existence of these two processes in English, Stampe
(1973) also points out a potential learnability problem in English. Because
of vowel nasalization (and subsequent nasal consonant deletion), children
will hear pairs such as ‘cat’ [kzt] vs. ‘can’t’ [kait] and ‘bees’ [bi:z] vs. ‘beans’
[bi:z], and words like ‘ant’ [St] which, unlike ‘bean’, do not have other
surface forms that provide information about the underlying nasal con-
sonant. These could be analyzed as having a phonemic distinction between
nasal and oral vowels. Stampe’s solution to this dilemma is to have
Table 8.12 A summary of Stampe’s (1973) analysis of selected forms from
Joan Velten
Adult Phonological processes (see below) Child

word A 1 2 3 4 5 form
‘room’
‘salmon’
‘home’
‘ham’
‘lion’
‘non’”
‘M’
‘bed’
Processes
A Child’s form after several phonological processes not related to the ones under analysis
1. Final obstruent devoicing (now ordered before 4).
2. Iterative regressive (vowel) nasalization: vowels, non-high nonsyllabics [ha], and
non-high sonorants [I] nasalize before nasal consonants. (The nasality of vocalics is
inversely related to their height, see Chen 1975.)
3. Nasalized [I] becomes [n].
4. Iterative progressive denasalization: of all nasal segments preceded by a non-nasal
segment .
5 . Denasalization of vocalics (context-free vowel denasalization).
~~~ ~ ~ ~ ~
a Joan produced a small set of French words which followed this pattern.
context-free processes innately (or naturally) ordered before context-

sensitive ones. Further, the English child innately assumes that the surface
nasal vowels have an underlying adjacent nasal consonant as their source,
i.e. that vowel nasality is derived. The English form ‘can’t’ /kaent/ is then
correctly produced by first applying vowel denasalization and then vowel
nasalization.
In looking at Stampe’s explanation of English vowel nasalization, we can
isolate three problems with his proposal. The first has to do with the
acquisition of languages with a phonemic difference between nasal and oral
vowels. It is not clear why the child would not be forced to assume that
nasality is also derived in such languages. It seems that Stampe needs to
treat all nasal vowels as underlying adjacent to nasal consonants, or else add
some mechanism for the child to distinguish derived from nonderived nasal
vowels. Based on his analysis of Joan (see below), it seems that he must take
the latter step.
A second dilemma concerns the so-called natural ordering of the two
processes. The first application of vowel denasalization contradicts the
earlier principle which stated that the natural order of processes is to allow
for their maximal application. This could be circumvented by stipulating the
principle that context-free processes always follow context-sensitive
ones, but then a different problem arises: such an ordering should be the
first ordering used by the child. Stampe (1973) presents a very ingenious
analysis of Joan’s more unusual forms, which is summarized here in
Table 8.12. In Stampe’s own analysis, vowel denasalization is ordered after
nasalization. Stampe’s only remark about this inconsistency is that ‘the
natural order is not yet manifest’ (p. 65).
The third dilemma is a more general problem for Natural Phonology.
This has to do with how the child creates phonemic representations for the
adult perceived forms. We have seen two examples of this problem: (i) how
does the child get from ‘can’t’ [kkt] to /kam/, and (ii) how does the child
get from ‘bead’ [bi:d] to /bid/? That is, the theory assumes a great deal of
phonological organization at the perceptual level which is never addressed.
In fact, given that the phonological processes in some ways are simply
performance factors in acquisition, it is these perceptual decisions which
are reflecting the child’s actual linguistic competence.
Stampe’s work has had a major impact on attempts to understand in
much more detail the range and limits to the kinds of natural correspon-
dences that exist between the child’s sounds and the adult target sounds. In
some ways it can be seen as an elaboration of Jakobson’s initial attempts to
determine the finite set of such correspondences. It has led others to
attempt to delimit the range of phonological processes, although most of
this work remains primarily descriptive and contradictory (e.g. Ingram
1974a, 1976a, 1979a, 1981a; Shriberg & Kwiatkowski 1980; Hodson &
Paden 1983). For example, one sees simultaneous natural processes of
fronting ([tJ] to [t]) and backing ([tJ] to [k]) being proposed. Such contra-
dictory processes have led others to attack the explanatory value of the
theory (e.g. Dinnsen 1980).
Macken & Ferguson (1983) Before concluding this section on expla-

nation, one other perspective on this topic requires brief discussion. In
recent years, the initial observations in Ferguson & Farwell (1975) have
been expanded in a series of articles to a point where they have come to be
seen as a new theory of phonological acquisition (e.g. Ferguson 1978;
Macken 1978, 1979; Macken & Ferguson 1983). Unfortunately, this
approach has acquired the name of a ‘cognitive theory’ of phonological
acquisition. I say ‘unfortunately’ because the use of the term is not with
the meaning used in this text and in child development in general; rather, it
is used in a very specific way to refer to the claim that the child actively
participates in the acquisition process through linguistic hypothesis testing
and revision. Since this general point is arguably part of both Jakobson’s
and Stampe’s theories (although not to the same degree), we will refer to
this approach as the Stanford theory, since most of its proponents have been
associated with the Stanford Child Phonology Project in one capacity or
another.
Chapter 6 examined some of the observations on phonological acqui-
sition made by Ferguson & Farwell (1975), and it was pointed out that their
approach, unlike that of Jakobson or Stampe, is much more characteristic
of Child Language. It arose from the detailed observations of individual
children, and does not make predictions about acquisition based on a
phonological theory of the adult language. To reiterate some of its aspects:
the Stanford theory emphasizes individual variation in acquisition across
children, and the variability of production within children. The former is
established through the phone classes and phone trees approach, and the
latter through the existence of lexical variability, phonetic variability,
phonological idioms, and selection and avoidance. It has also been argued
that much of the variability in the data they use results from their
methodology, which emphasizes performance factors.
The theoretical model includes some of the following claims. Regarding
perception, it proposes that some of the child’s errors in production may
stem from incomplete perception; as such, it is on the side of a global or
partial perception theory. Macken (1980), for example, gives an analysis of
Smith’s rule 3 in Table 8.11 which accounts for the pattern on the basis of
misperception. The theory also emphasizes the influence of the child’s
output or word patterns on the child’s underlying representation, as
mentioned above for Macken (1979). The theory, then, is like a three-level
model as discussed above. Regarding change, it proposes that there are
discontinuities in phonological acquisition, with the most explicit proposals
those of Macken (1979), i.e. that the child’s organization of phonology
changes from using the word to the phoneme, and finally to the feature.
In comparing the Stanford theory with others it is necessary to realize that
other theories have been aware of the kinds of observations it emphasizes.
For example, both Jakobson and Stampe were aware of the extent of
individual variation in phonological acquisition. Jakobson (1968: 31), for
example, states:
If all of the sound productions of the child are tossed into the same
heap, it is understandable that the laws of development cannot be
disclosed. By careful elimination, however, the regular succession of
acquired phonemic oppositions emerges.
Stampe (1973: 11) also recognizes the extent of this diversity:
Despite the opinion of early investigators, it is now generally recog-
nized that different children acquire sounds and sound-patterns in
quite different relative orders . . . children choose bewilderingly
diverse substitutions for the sounds and sound-patterns they have not
yet acquired. 1
While Jakobson concentrated on eliminating much of the variability by

methodological means, Stampe accepted it and incorporated its potential
within his theory.
It seems to us that the Stanford theory has isolated some important
observations on phonological acquisition, but that it needs to address some
unresolved issues before it can be considered as a serious challenger to
either Jakobson or Stampe. First, it needs to provide a theory for the
general aspects of acquisition: while emphasizing individual variation, it
misses the fact that general patterns do occur. More generally, the theory
needs to provide an explanation for the extent and limits of the individual
variation it observes. A step in this direction is Macken (1986), who
discusses a probabilistic component to phonological theory.
Second, the theory needs to account for the discontinuities that it
proposes. For example, what does it mean to say that the child changes from
words to phonemes to features? Elsewhere (Ingram 1986) I have pointed
out some concerns about these proposed discontinuities. For example, does
the child use the ‘word’ as an organizational unit for the 100 to 200 words she
has in her receptive vocabulary at the onset of word production? Further,
the ‘word’ as an organizational unit is never defined. If it were the unit of
organization, we would expect units like [ba], [di], and [mu] where
segments would be fused together, not the segmental patterns such as [ba],
[bu], [bi] which are more typical of early phonological acquisition.
Relatedly, the status of the former units during the next stage needs to be
more clearly spelled out. Lastly, the theory needs to address the cross-
linguistic patterns of acquisition observed in Pye, Ingram & List (1987). As
pointed out elsewhere (Ingram 1988), these patterns are consistent with
predictions of Jakobson’s theory.
8.3 The further development of word meaning

The cognitive precursors to the onset of word meaning and the onset of
early word acquisition were discussed in Chapters 7 and 8. This chapter will
follow the subsequent development of word meaning from the end of the
second year of life up to approximately 4 years of age, by presenting a
sampling of the observations made to date on this development, and by
discussing some of the theoretical proposals put forward to account for it. A
comment on the limited scope of this discussion on the acquisition of
meaning, however, is required here.
While tempted at times to do so, I have avoided labeling any of the
sections of the book with the title ‘semantic acquisition’. The reason for
this is that ‘semantic acquisition’ comprises a wide range of topics that I
could not begin to address. In fact, it is such a diverse set that no
book-length treatment of the entire range has yet been attempted. What
does exist, however, is a selection of books devoted to one or more of its
domains.
A comprehensive treatment of semantic acquisition would need to cover
at least the following topics. First, a broad distinction is made between
‘lexical semantics’ and ‘propositional semantics’, where lexical semantics
refers to the meanings of individual words, and propositional semantics
refers to the meaning that results from combining words. Within lexical
semantics, we can distinguish the study of the meaning of the different word
classes which occur in language. One class of such words, nominals or
referring expressions (e.g. ‘dog’, ‘flower’, etc.), has been a particularly
popular topic for acquisition studies. Other classes include verbs, adjec-
tives, and adverbs. Adjectives have also proven to be a rich area for study,
particularly those adjectives which refer to dimensions, such as ‘big’, ‘little’,
‘long’, and relations such as ‘more’, ‘less’, as well as particular domains such
as color. Lexical semantics also includes the meanings of a wide range of
grammatical morphemes such as pronouns, prepositions, and various
inflections.
Propositional semantics is loosely used here to capture a range of
phenomena which require reference to the constituent structure of sen-
tences. For example, there are at least (i) the semantic or logical relations
which underlie sentences, such as Agent and Action; (ii) the relations
between pronominals and their possible antecedents, as in sentences like
‘Behind him, John saw a snake’; (iii) thematic meaning (Leech 1974),
referring to the marking of parts of sentences which are topics or focussed
elements (for example, the thematic meaning of ‘John’ differs in the
sentences ‘It was John who I saw’ and ‘I saw John’); and (iv) scope
relations, as in ‘All of the arrows didn’t hit the target’. Still other kinds of
meanings include ‘stylistic’meaning, the differences in using words such as
‘steed’ (poetic), ‘horse’ (general), and ‘nag’ (slang); and ‘affective’
meaning, which distinguishes sentences like ‘Would you be so kind as to
lower your voice’ from ‘Shut up!’ (Leech 1974: 17-18).
The discussion in this section will be restricted to a selection of studies on
lexical semantics. For this reason, the term ‘word meaning’ is used instead
of ‘semantic development’. Semantic relations have already been examined
in Chapter 7, and will be returned to later in the next chapter. A discussion
of the many other aspects of meaning, unfortunately, is beyond the scope of
the present text.
8.3.1 Theoretical proposals: general issues

Chapter 6 discussed the child’s use of overextensions as one kind of
evidence that children do not always have the same meaning for a word as
adults have. The discussion was somewhat cautious, however, since it
appears that overextensions may result from several different causes (see
Section 6.3.3). Even with such caution, it was proposed that one cause is an
incomplete semantic system. Except for Piaget’s views on the transition
from symbols to signs, there was no examination there of how the child’s
semantic system develops. A review of the theories which have been
attempted is now presented, beginning with some general issues in this
section, and turning to some specific proposals in the next.
The general issues at stake in a theory of lexical semantic acquisition are
similar to ones we have considered in other domains. Again, we are faced
with the question of continuity. There is little in the way of proposals for
stages of lexical acquisition as there are for syntactic and phonological
acquisition: the single most important milestone appears to be the word
spurt around 1;6. Given this situation, it is not surprising that many of the
theories for lexical acquisition imply continuity of development. In such a
case, the child is claimed to be using the same units and mechanisms of
organization as the adult. Development, then, consists primarily in the child
determining the range of meanings used by the adult for the lexicon of the
language. Such theories may, at first glance, seem quite distinct from each
other because they propose different units and mechanisms. Their deeper
similarities appear, however, when the issue of continuity is considered.
As stated, much of the disagreement between theories can be traced to
differences in the units of organization proposed for the child. One general
difference is between theories which break down the meanings of words
into features and those which do not. The former will be referred to as
featural theories, and the latter as holistic ones, following Armstrong,
Gleitman & Gleitman (1983). For example, we can discuss the meaning of
the word ‘dog’ as consisting of a set of features which together construct the
total meaning, much as phonological features constitute the organizational
unit of phonology. The holistic alternative is that the meaning of ‘dog’ is the
concept DOG which cannot be broken down any further. Since holistic
theories require a separate concept for each word in languages, they expand
greatly the number of possible units for semantic structure (see Fodor
1975). Most accounts of early lexical acquisition can be classified as featural
theories.
Featural theories differ from one another on several dimensions, for
example, the kinds offeatures which are assigned to words by children, and
presumably adults. One widely discussed difference is between perceptual
features and functional ones. Perceptual features are features determined
for an object by the senses, e.g. shape, smell, and touch. ‘Ball’, for
example, can be categorized by a perceptual feature [+round]. Functional
features are those which concern the uses which objects have. A ball, for
example, is something which can be thrown and bounced. Two of the most
discussed theories of lexical acquisition, Clark (1973a) and Nelson (1974),
propose that word meaning consists of both kinds of features, which differ
in their emergence in the early months of acquisition. Clark suggests that
perceptual features take precedence at the onset, while Nelson argues that
functional ones do.
Another difference between featural theories is the way in which features
are organized into word meanings. The traditional approach is one which
has been referred to as the definitional view (Fodor, Garrett, Walker &
Parkes 1980). The definitional view is that the meaning of a word consists of
a set of invariant and necessary features. For example, the meaning of a
word like ‘ball’ will consist of features such as [+object, -animate,
+round, +throwable]. Any object identified as a ‘ball’ will need to have at
least these features. The definitional approach to word meaning was
challenged by Wittgenstein (1953), who pointed out the problems with
determining an invariant set of features for any particular word. The most
famous example of this is his discussion of the problem of defining a ‘game’.
We can identify a series of potential invariant features which all have
counterexamples (shown here in parentheses), e.g. use of a board (tennis),
play by at least two people (solitaire), play to win (ring around the roses).
This exercise can be replicated with virtually every referential word.
Wittgenstein’s solution is to propose that the meaning of words is a set of
cluster concepts. Words do not share a single set of features, but rather
participate in a network of features such that different words share different
features from the network.
Wittgenstein’s original ideas have been picked up in recent years by
Eleanor Rosch, who has developed a theoretical approach known as
prototype theory (e.g. Rosch 1973,1975,1978; Rosch & Mervis 1975; Rosch
et al. 1976). For any word meaning such as ‘ball’, all the members of the
class will share some of the set of defining features for a ball. Those items
which share the most features will be most typically balls, or the prototypes.
Those objects which share fewest of the family of features associated with
balls will be less likely to be so identified. A baseball, therefore, is more
typically a ball than a rolled up sheet of paper that is thrown around the
room. In the same way, a robin is more typically a bird than is a penguin or
an ostrich. Rosch’s research has consisted of asking adults to identify
instances of objects as being most typical of certain words, such as ‘fruit’ and
‘vehicle’. Adults appear to be able to grade objects in this way and agree in
their ratings. Prototype theory also accounts for the observation that the
boundary between categories can be somewhat ‘fuzzy’ when objects have
features from different categories. Prototype theory had an immediate
appeal to those working in language acquisition, although it has been
adapted in different ways (Anglin 1977; Bowerman 1978a; Barrett 1978,
1982; Greenberg & Kuczaj 1982).
Another issue to be addressed is how the units of organization appear
over time. A maturational theory of lexical development claims that the
child is born with the innate set of concepts which can be lexicalized in
human development. The child’s task then is basically to map each word
acquired onto the innate set. Fodor (1975) presents a theory of this kind.
The other featural theories that have been proposed have been more
constructionist in nature. In these, the child begins with some initial
hypotheses about the meaning of words, and then gradually narrows them
down. There are three different ways in which such a development might
proceed. First, the child may develop from general-to-specific categories.
For example, ‘ball’ might initially mean ‘round thing’, then ‘round thing
which can be thrown’ and so forth (see Clark 1973a, discussed below). A
second possibility is that the child may proceed from specific-to-general
categories. In this scenario, ‘ball’ might mean ‘this particular blue ball’, with
another red ball being labeled ‘red’. Only after later experience with a range
of balls would the term generalize (see Anglin 1970). Lastly, development
could take place from some intermediate level of organization. For example,
if children do indeed form prototypes, then the first experience with a ball
might lead to an initial prototype that is neither as general as ‘round thing’
nor as specific as ‘this particular blue ball’ (Anglin 1977).
8.3.2 Specific theoretical proposals

Somewhat simplistically, the history of the study of lexical acquisition has
seen the rise and fall of a series of theoretical proposals. This section
provides a brief historical overview of those which have been particularly
influential, using the issues just discussed as a frame of reference.
The semantic feature hypothesis The study of lexical acquisition is natur-

ally tied to the issue of cognitive development and categorization. The
meaning of a word for the child at any point in time is going to be influenced
by the child’s cognitive abilities. One claim about cognitive development
has been that the child develops by forming general concepts and then
narrowing them down, a view popular in the 1960s (Brown 1958b). Also in
the 1960s, a semantic theory developed within linguistics which proposed
that the meanings of words consist of a small set of universal semantic
features (Bierwisch 1970). Eve Clark (1973a) combined these two proposals
into a theory which she called the semantic feature hypothesis.
The semantic feature hypothesis is based upon a definitional view of word
meaning, that is, the meaning of a word consists of a set of necessary and
invariant semantic features. For purposes of discussion, we assume that the
features vary from general features to specific ones. The most general
features will be referred to as superordinate categories, and the most specific
ones as subordinate categories. A lexical example of this distinction is
‘animal’vs. ‘Collie’ with ‘dog’ at some intermediate level of generality. The
corresponding semantic features might be [+animate], [+long snout], and
[+canine] respectively.
The child acquires the meanings of words within this theory by first
acquiring the more general superordinate features. The first meaning of
‘dog’ for example, might just be [+animate], and later [+animate, +four-
legged]. Acquisition then goes from the more general to the more specific.
Also, the first features which are acquired are those which are perceptually
salient to the child. The most primitive categories involve movement,
shape, size, sound, taste, and texture.
An important aspect of the semantic feature hypothesis is that it makes
predictions about relational terms as well as referential ones. It predicts that
various antonyms like ‘same’ vs. ‘different’ and ‘more’ vs. ‘less’ will first be
acquired with the same meaning. In such cases, one of the terms is
considered to be more basic or ‘unmarked’, so that its meaning will be
acquired first. A series of predictions about the acquisition of dimension
terms are also made. For example, the pair ‘big’, ‘small’ is considered to be
more general than pairs like ‘tall’, ‘short’ and ‘high’, ‘low’ (H. Clark 1973),
and is thus acquired before them.
The semantic feature hypothesis had the virtue of developing current
views on semantics and acquisition into a coherent and testable theory of
development; because of this, it stimulated a wide range of studies, and was
still being critically reviewed in Carey (1982). It is very much a construc-
tionist theory in which the child’s lexical meanings are always a subset of
those of the adult. Development proceeds by the addition of features to
those already acquired. Its predictions about word acquisition are also
relatively straightforward: words will be overextended at the onset and then
eventually narrowed down until they are correct.
Functional core concept theory Katherine Nelson in Nelson (1974) puts

forward a theory referred to as the functional core concept theory. This
theory most generally differs from Eve Clark‘s in two ways. First, it proposes
that the child first forms a concept, and then attaches a word to it, whereas
the semantic feature hypothesis suggests that the child acquires a word
Table 8.13 Four steps in the acquisition of the functional core concept
‘ball’, adapted from Nelson (1974: 276-8)
1. ‘Identification of an individual whole’. The child recognizes the things it acts upon as
objects, not pictures or collections of features.
A ball is recognized as an object in space and time.
2. ‘Identification of important relationships of objects and assigning individuals on the basis of
their functional relations to a synthesized cognitive “chunk” or “concept”’. This process
can take place after experience with a single member of the class of objects.
A ball is known as an object which enters into several relations, such as the thing that can be
thrown, rolled, bounced, picked up, etc. For example, ‘mother throws ball in the living
room’. It is known both for its own actions and the uses it has. It does exist independent of
these relations.
3. ‘Identification of new concept instances by noting the salient stable (‘invariant’) character-
istics of members included in the concept on functional grounds and forming a hierarchy of
identification attributes therefore.’ From experience with other members of the class, the
child recognizes that certain attributes define the concept and others do not. ‘For some
concepts the child may retain relations that the adult regards as superfluous.’
The child sees a boy throw a ball in the park and identifies the relation of ‘throwing’ as a
defining feature, and location such as ‘in the living room’ vs. ‘in the park’ as not. A
superfluous feature such as ‘is thrown to child’ may be retained as a defining feature for a
while. The functional core concept of ‘ball’ is as follows:
BALL: Core functions = rolls, bounces
Location = Li i.e. any place for playing
Actor = Xj Le. any person
Action = Ak Le. other actions can be performed that can
be done to small objects
Direction = to child”
4. ‘Attach a name to the concept so formed.’
The child attaches the phonetic sequence [bd] to the functional core concept of ‘ball’.
” This feature will need to be eliminated from the concept before this process is complete, since
it is not a defining feature of ‘ball’.
and then attaches meaning to it. Second, Nelson argues that there is a level
of cognitive organization which exists intermediately between perception
and language. She interprets Clark as mapping directly from perception to
language.
Nelson outlines four processes, or steps, that the child follows in
acquiring a concept. These are summarized in Table 8.13, with the acqui-
sition of ‘ball’ as an example. The first process is the simple identification of
the ball as a real object. The second process involves the recognition of
functional features for the ball. At this point they are context-bound, and
the ball is not known outside of these relations. The third process separates
the core functions from the other relations which can be represented by
abstract markers such as Li ‘any place for playing’. The last process involves
the acquisition of a word for the concept.
We see then that the child acquires quite a bit of knowledge about a
concept before attaching a word to it. Also, as pointed out by Anglin
(1977), this suggests that the child has the meaning of words basically
correct when they are acquired. Should the child acquire a word earlier,
say during processes 2 or 3, then the error would be one of using a more
narrow meaning for the word than the adult does: that is, the direction
of error is in the direction of underextension. This prediction of the
functional core concept theory is markedly different from that of the
semantic feature hypothesis which predicts overextension of early word
meanings.
It is important to point out that the functional core concept theory and the
semantic feature hypothesis differ primarily in their claims about the onset
of acquisition. The former, for example, recognizes that perceptual features
will eventually be added to the child’s concept, just as the latter recognizes
the addition of functional features. They are both theories that approach
meaning from the definitional view of word meaning and see the acquisition
of word meaning as one which requires the establishment of an invariant set
of features. Macnamara (1982: 10) points out the overall similarities when
he states that the difference between the two theories is mainly one of
emphasis.
A last point about Nelson’s theory is that it is proposed for the acquisition
of referring expressions such as ‘ball’. It is not clear how Nelson would
account for the acquisition of relational terms such as ‘big’ and ‘little’. One
of the early appeals of Clark’s theory was its wider range of predictions.
Prototype theory The basic characteristics of prototype theory were

discussed in section 8.3.1. To reiterate: it is an approach developed by
Eleanor Rosch and her colleagues to account for the representation of
meaning by adults. The proposal is that the meaning of words is not a set of
invariant features, but rather a set of features which capture family
resemblances. Some objects will be most typical of the word’s meaning by
sharing more of the word’s features than others. Certain features, then, will
be more important in determining class membership than others, but none
are required by all members.
This theory appeared after the two just discussed and has been adapted
by several people for the acquisition of word meaning. In one version or
another it is probably the most widely held theory among those who
approach the topic from the perspective of Child Language. Here we will
examine two of the attempts to adapt it to language acquisition - Anglin
(1977) and Bowerman (1978a). Other attempts to do so can be found in
Greenberg & Kuczaj (1982) and Barrett (1978,1982).
Anglin (1977) One of the first to adapt prototype theory to child language
was Anglin (1977), which still represents one of the most extensive
databases on lexical acquisition. To follow Anglin’s proposals it is necessary
to define some of the terms he uses:
Basic definitions used by Anglin (1977:27)

Term ofreference: ‘a word . . .which denotes or refers to real objects’;
Extension: ‘all the objects which an individual is willing to denote’
with a term of reference;
Intension: ‘the set of properties which an individual believes to be
true of the instances of the category denoted’ by a term of reference;
Concept: ‘all of the knowledge possessed by an individual about the
category of objects denoted’ by a term of reference. ‘This knowledge
includes both knowledge of extension and knowledge of intension.’
Like Nelson (1974), Anglin was interested in the study of referring terms or
‘terms of reference’. He makes a distinction which has been made for years
in semantics between a word’s extension, i.e. the objects which a word refers
to in the real world, and its intension, i.e. the internal representation of the
word’s meaning. Importantly, the concept underlying the meaning of a
word contains knowledge about extension and intension.
Anglin (pp. 257-64) proposes that concepts or word meanings are
acquired in the following fashion. First, the child forms a perceptual schema
or representation of an object, based on his first experiences with it. This
schema is analogous to a prototype, in that it is a visual image of the object
that is not broken down into individual features. The child, therefore, does
not begin at some superordinate level (semantic feature hypothesis) or
subordinate level (functional core concept), but at an intermediate level.
For example, a ‘ball’ is neither ‘a round thing’ nor ‘the thing mommy throws
to me in the living room’, but the perceptual schema that represents the
balls which the child has seen. At first, the prototype will be limited to the
perceptual characteristics of the first instance so named, but will be
generalized as more instances are met.
It is at this point that the distinction between extension and intension
needs to be made. The child’s extensional knowledge of the object will
consist of the prototype which is constructed primarily from perceptual
categorization. Intensional knowledge, however, will involve not just
perceptual information, but also the functional characteristics discussed by
Nelson. Anglin states (p. 262):
. . . both the theory of concept formation based upon prototypes and
the theory based upon function . . . have relevance for a complete
description of the structure of the concepts underlying the child’s
terms of reference. The former is implicated in their extension while
the latter is implicated in their intension.
This account, then, incorporates the perceptual and functional aspects of
early lexical development.
A key aspect of Anglin’s theory is the way that extensional and intensional
knowledge develop over time. For the preschool child, these two aspects
of concepts are not well coordinated. That is, the child is not very efficient in
using its intensional knowledge when dealing with instances of the concept,
or in using its extensional knowledge when discussing its intensional
knowledge. This claim is an inductive one, based on Anglin’s research
which indicated that children’s overextensions (or identification of instances
of a concept) were primarily based on perception, while their definitions of
words (or intensions) were primarily functional, e.g. ‘apple’ is something
you eat. Interestingly, the claim is also consistent with Piagetian theory,
which sees coordination as an important mechanism in development.
Anglin also discusses other developments for the child. Not surprisingly,
the number of concepts increases over time, as well as the amount of
knowledge about each concept. More important, there is the development
in the school-age child of the ability to organize his knowledge into a
hierarchical system in which the child is aware of the relations between
superordinate and subordinate categories. For example, the preschool child
will know, at least consciously, that ‘trees’ are also ‘plants’. This ability
appears to develop quite late.
Anglin’s theory is similar to Clark’s and Nelson’s to the extent that it is
constructionist in nature. The child is using adult-like mechanisms to
organize his system. The changes in time are in adding information as well
as coordinating the knowledge already acquired. It differs in its predictions
about the nature of the child’s early word meanings: they are at an
intermediate level at the onset, and develop from there to more general and
more specific meanings. Also, Anglin’s theory differs in its view of the
relation of the child’s knowledge (or competence) to his performance. By
discussing two kinds of conceptual knowledge and their poor coordination,
he suggests that the child’s performance will only reflect part of the child’s
underlying knowledge; that is, he implies a larger gap between the two than
the other theories we have discussed.
Bowerman (1978a) Like Anglin, Bowerman developed her version of
prototype theory after empirical study of the speech of children. While
Anglin’s data were the result of experimental research on several children,
Bowerman’s data consisted of her careful diary notes on the language of her
own two children. Her data are the most detailed diary data presented on
the topic, and will be examined more carefully later.
Bowerman believes that both Clark and Nelson have proposed theories
which only account for part of the facts of lexical acquisition. In support of
this view, she presents two general claims in particular about lexical acqui-
sition. First, evidence suggests that children do not restrict themselves to
just perceptual or functional features, but use at least both. For example,
data indicate that the concepts behind early word meanings can be super-
ordinate for some words, prototypical for others, and even adult-like for
still others. From this she claims that children have ‘a variety of methods of
classification at their disposal’ (p. 283). Second, one type of categorization
does not necessarily replace the others over time; that is, the different kinds
of classification can be used simultaneously. Therefore, one type of catego-
rization is not more primitive than another. (We have already seen data in
Chapter 6 that suggest such a situation for overextensions in the study by
Rescorla 1980, as presented in Table 6.8 and the surrounding text.) From
these claims, Bowerman concludes that the representation of meaning as
features and prototypes needs to be incorporated into a single model.
The difference between Anglin and Bowerman represents that between
two views of prototype theory. Anglin’s view is that prototypes are like
mental images, where the prototype is an abstract image which looks like all
the members, yet is not necessarily any one in particular. As an image, it
cannot be broken down into featural components. Bowerman cites a
position which was expressed in Rosch & Mervis (1975) that prototypes can
be viewed as consisting of features. While none of the features are required
for class identification, prototypicality of any object is determined by the
number of features that it shares with other members of the class. This is the
version of prototype theory which was described above in 8.3.1. +
The development of a prototype, in Bowerman’s view, proceeds as
follows: (i) the child first hears a word frequently, and possibly exclusively,
in a particular context, e.g. ‘night-night’ while going to bed; (ii) the child
first acquires and uses the word in this context (from a few days to more than
a month) - presumably a preliminary prototype is formed at this time; (iii)
the child imposes a featural analysis upon the word’s prototypical meaning.
Further, some of these features can now be recognized in other contexts
without the other features with which they occurred in the previous stage.
This results in the associative complexes discussed in Chapter 6 in relation
to the study by Rescorla (1980). Bowerman sees the existence of such
complexes as compelling evidence for a prototype model. The features
selected for such overextensions can be quite variable from child to child, so
universal predictions for what will occur are not possible. They can be both
perceptual and functional, although the former appear to be more promi-
nent, at least in the early stages.
Bowerman’s study can only be taken as an outline of a prototype theory
of lexical acquisition. There are a number of gaps in it, such as details on
why some words overextend and others do not. Also, as pointed out by
Barrett (1978,1982), the theory does not account for why the child does not
end up with several words for the same object through overextension.
Except for the possibility of retrieval errors (see page 159), children do not
usually overextend a word to objects for which other words already exist.
Bowerman’s initial speculations, however, have provided a basis for further
refinement, as in Barrett (1982).
8.3.3 Some findings on the acquisition of nominals

Turning from theoretical speculations to look at some of the data which
have resulted from attempts to test theories such as those just discussed, it
can be seen from the above discussion that much of the research has
focussed on the acquisition of nominal or referring expressions. To get a feel
for some of the kinds of data on this topic, we will look at the most extensive
report on the topic, Anglin (1977). This will be followed by a discussion of
the diary data reported in Bowerman (1978a).
Anglin (1977) Anglin reports a series of studies conducted to answer three

general questions posed at the onset of the acquisition of terms of reference:
(i) what is their order of acquisition; (ii) what is their extension; and
(iii) what is their intension? We will look at a selection of the studies done
on each of these questions.
Anglin wanted to examine the issue of order of acquisition, i.e. whether the
child’s first terms of reference indicate superordinate, intermediate, or sub-
ordinate categories. The resultsof such research bear on the predictionsof the
theories discussed in the previous section. Previous research had already
provided some preliminary findingson this topic. In an important early paper,
Brown (1958b) had already pointed out that children do not begin with either
superordinate terms (e.g. ‘thing’) or subordinate terms (e.g. ‘dime’), but
intermediate ones (e.g. ‘money’). Studies of the early vocabulary of children,
such as Nelson (1973), also show this acrosschildren. Table 6.7 presented the
most common terms of reference used by the subjects in Nelson (1973). The
most frequent animal term, for example, was ‘dog’, a term intermediate
between the superordinate word ‘animal’ and the subordinate ones like
‘collie’or ‘boxer’. The problem with such studies, however, is that the child
might still use the terms in a more general or specific way. ‘Dog’, for example,
might mean ‘all animals’ to the child, but be used instead of ‘animal’because
of the naming practices of adults. This was, in fact, the claim in Brown
(1958b), who wanted to reconcile the use of intermediate vocabulary by chil-
dren with his position that cognitive development is general-to-specific.
Anglin conducted two studies to examine this issue in more detail. The
second of these (experiment 2.2) looked at the sets of terms given in (8.20).
(8.20) Level I I1 I11
superordinate animal plant food
intermediate dog flower fruit
subordinate collie rose apple
The stimuli consisted of 36 black-and-white pictures, 12 for each set of

terms. There were three pictures for each intermediate and subordinate
term. The pictures of superordinate terms were divided into two sets of
three, one for prototypical instances of the terms, another for peripheral
ones. The pictures were selected from 146 original pictures which were
shown to adults for rating of the extent to which they were typical of the test
words. For example, the three pictures showing collies were those judged
most like collies from 12 pictures presented. The three pictures of a
prototypical animal were of a King Charles spaniel, an African elephant,
and a cat. The pictures for ‘animal’which were judged peripheral were of a
bullfrog, a monarch butterfly, and a marsh hawk (see Table 2.4 in Anglin
1977 for a list of all the objects used and see p. 57 for the pictures used for
Set I). Each set of three pictures was put on a separate poster for
subsequent testing, for a total of twelve posters, four per set.
The subjects were 20 children between 2 and 5 years of age. They were
first told that they would be asked to name some pictures. After a
demonstration of the task with pictures testing the hierarchy ‘boy-child-
person’, they were randomly shown the 12 posters. The experimenter
pointed to each picture and asked ‘What’s this?’ After naming each
individual object on the poster, the child was then asked ‘What are all
these?’ The children were also encouraged to give another name if they did
not distinguish between objects, e.g. ‘dog’ for each of three dogs.
A simple summary of the major results is provided in Table 8.14, which
gives the findings for the answers to the second question ‘What are these?’
The adults categorized the posters as expected. The responses given in the
table are those which were most common for the adults and children. Two
results are shown for the children: ‘percent correct’ represents the percent-
age of times that the children used the same label for each poster as the
adults did, and this is followed by the most common response used by the
children.
For two of the three sets, the intermediate terms ‘dogs’ and ‘flowers’were
clearly the earliest acquired. Even the specific term ‘apple’ is somewhat
intermediate in the sense that there are varieties of apples which could be
labeled. Anglin concludes (p. 67): ‘There is neither a unidirectional
specific-to-general progression in vocabulary development nor a unidirec-
tional general-to-specific progression .’
In a later discussion, he proposes three factors which determine the words
Table 8.14 Results of Anglin’s experiment 2.2, reported in Anglin (1977)
Child
Adult Percent correct Most common response
animals 50 animals
dogs 100 dogs
co11i es 0 dogs
plants 15 flowers
flowers 70 flowers
roses 10 flowers
foods 40 foods
fruits 20 foods
apples 95 apples
that are first acquired. One of these is relevance, i.e. the extent to which the
object labeled is important to the child in its daily activities. This factor may
override the tendency towards intermediate terms, since specific individuals
like the mother and father are important to the child and will receive names.
‘Dog’ may be the usual intermediate term but the child will also learn the
name of the family dog. The second factor is function, which is used to
capture the fact that adults name objects at an intermediate level for a
particular purpose. They group objects together when they constitute a
group that the child should behave similarly toward. For example, the child
might not need to be careful with all animals, but dogs might constitute such
a group. Lastly, Anglin found that there was a correlation between order of
acquisition and frequency of occurrence of the words in the speech of
children. Anglin suggests that this correlation can be used as a general
heuristic for determining the categories used in children’s vocabularies. His
subsequent analysis provides a taxonomy very similar to the one in Nelson
(1973) which was reported in Chapter 6.
To answer his second research question, Anglin next examined the
occurrence of extension by studying children’s labeling in both comprehen-
sion and production. Chapter 6 already presented some studies on chil-
dren’s overextensions in comprehension and production. Those studies are
similar to Anglin’s, to the extent that they begin with spontaneous over-
extensions in the child’s spoken language, and use these as the basis of
further examination (e.g. Thomson & Chapman 1977). Anglin’s studies
differ in that they present a range of pictures to children and see how they
label them. The pictures, however, are not selected on the basis of the
child’s previous naming behavior. Anglin offers a general criticism of the
previous studies that they bias the study of word meaning in the direction of
overextensions.
To get around the above experimental bias, Anglin used a method which
would be sensitive to both over- and underextensions. His use of this
method in his experiment 4.1 is examined here. Three sets of test words
were used which differed in three degrees of generality: see (8.21) and
compare with (8.20):
(8.21) I I1 111
animal food plant
dog fruit flower
collie apple tulip
The subjects were two groups of 18 children each, with the ages 2;6 to 4;O
(younger group), and 4;6 to 6;O (older group). There were also 18 adults
tested. Each group was divided into three subgroups of six subjects each.
Within each subgroup, each subject would be shown 120 pictures distribu-
ted across three test words in the following way: 10 instances and 10
non-instances of a subordinate term, e.g. ‘collie’; 20 instances and 20
non-instances of an intermediate term from a different set, e.g. ‘fruit’; and
30 instances and 30 non-instances of the subordinate term from the
remaining set, e.g. ‘animal’. For each test picture, the child was asked ‘Is
this a (test word)?’
Detailed results were presented for each subject and selected pictures.
The general results can be summarized as follows. The children showed
both overextensions and underextensions, the latter being the more fre-
quent. For example, the younger children underextended 29% of the
possible underextensions, and overextended only 8 percent of the possible
overextensions. These figures were 16 percent and 6 percent respectively
for the older children. The most frequent underextension was the denial
that a picture of a woman was an ‘animal’. Other examples were rejecting
the pictures of a praying mantis and a caterpillar as ‘animals’.
There were three factors which influenced the data: (i) individual vari-
ation between children, (ii) the concept studied, and (iii) the nature of the
instances and non-instances being tested. Individual variation was evident
in that some terms were overextended by some children, underextended by
others, and correctly used by still others. Such data support the contention
of Chapter 6 that multiple factors may underlie a child’s errors in extension.
Despite the individual variation, there were some patterns for specific
concepts. ‘Flower’, for example, tended to be overextended to items such as
an elephant ear, a coconut, and a philodendron. ‘Plant’, on the other hand,
was usually underextended. In particular, trees were rejected as examples
of plants. Lastly, atypical or peripheral instances of a test word were more
likely to be incorrectly identified than others. This point was verified by
Anglin in subsequent studies.
This experiment was followed by a series of others to explore whether the
child’s extensional errors resulted more from perceptual or functional
factors. The findings supported Clark’s early findings that overextensions
tend to be primarily based on perceptual similarities.
Lastly, Anglin examined the intension of children’s early words. He
studied 14 children between 2;s and 6;7. Each child was interviewed and
asked to discuss the meanings of 12 words: ‘dog’, ‘food’, ‘flower’, ‘vehicle’,
‘animal’, ‘apple’, ‘rose’, ‘car’, ‘collie’, ‘fruit’, ‘plant’, and ‘Volkswagen’.
Anglin comments as follows (p. 189) on the methodology:
In general, we tried to ask five questions about each word in roughly
the following order: (1) ‘What is a -?’ (2) ‘Tell me everything
you can about (a) -.’ (3) ‘What kinds of -s are
there?’ (4) ‘What kind of thing is a -?’ ( 5 ) ‘Tell me a story
about a -?’ However, apart from the first one, these questions
were not always asked of each child for each word and quite often a
variety of other questions were asked.
The method, then, was very much in the spirit of Piaget’s clinical method.
This particular experiment, in fact, is similar to the one in Piaget (1929).
Anglin presented his results as a series of conversational samples with
individual children. Most generally, the children did not show knowledge of
the ability to give a set of defining properties for a word’s meaning. One
interview with a child Peter, age 2;8, particularly shows the limited kinds of
definitions available from young children. When discussing ‘dog’, Peter
states that ‘it goes woof, barks’. H e also recalls two dogs that he has seen
and what they looked like. That is, ‘the very young child’s expressible know-
ledge of a concept, as in this case, often appears to be “instance-oriented”
and quite possibly is based to a considerable degree on visual imagery’
(p. 191). Anglin emphasizes, however, in his analysis, that the definitions
are more often based upon the functional properties of objects rather than
strictly perceptual features.
Anglin cites the above findings on lexical acquisition as support for his
version of a prototype theory which we have already seen. The data on the
early acquisition of intermediate terms and the use of overextensions and
underextensions are cited as evidence for the child establishing early
prototypes which are perceptually based. The data from word definitions is
given in support of the claim that the child’s intensional knowledge is
functionally based. The fact that both kinds of knowledge appear to be
available to children leads to the proposal that a concept contains both
intensional and extensional knowledge.
Bowerman (1978a) In this paper, Bowerman presents data on the early

lexical acquisition of her two daughters, Eva and Christy, to support her
version of prototype theory given above. Bowerman states the following
Table 8.15 Summary of the examples of word meaning for Eva and
Christy given in four tables in Bowerman (1 978a)
ExampI es Tables and their contents

Eva Christy
‘moon’ ‘snow’ 1. Overextensions based on perceptual similarities, counter to

‘money’ known functional differences
‘there!’ ‘aha!’ 2. Words extended to novel situations on the basis of subjective
‘too tight’ ‘heavy’ experiences
‘heavy’
‘ball’ ‘on-off‘ 3. Words used noncomplexively for referents with shared
‘ice’ attributes
‘off
‘kick’ ‘night-night’ 4. Complexively used words with prototypical referents
‘close’ ‘open’
‘open’
‘giddiup’
‘moon
about her data (p. 264): ‘Fairlycomplete records are available on the way in
which almost every word was used from its first appearance in the child’s
speech to about 24 months. Data on word use continue beyond that point
but are more selective.’ Bowerman presents her data in four tables in the
form of selected examples, summarized here in Table 8.15. These examples
from Eva and Christy are frequently cited in discussions on the acquisition
of early word meaning.
The first table deals with the issue of whether the child’s early extensions
are on the basis of perceptual or functional similarity. First, Bowerman
observes that most of the girls’ overextensions appeared on the basis of
shared perceptual features, a finding consistent with Clark’s (1973a) diary
analyses and the findings from Anglin (1977) presented above. She points
out, however, that this finding would not be incompatible with Nelson’s
functional theory if the perceptual features were ‘used primarily to predict
the function of an object so that the object can be identified as a member of
a known function-based category’ (p. 265). The examples in her Table 1,
however, are given to show that such was not the case in most instances. For
example, Eva used ‘moon’ for the following situations, among others: ‘for a
ball of spinach she was about to eat, for hangnails she was pulling off, for a
magnetic capital letter D she was about to put on the refrigerator’ (pp. 265,
267). In each case, the perceptual feature of shape was similar while the
actions were quite distinct.
We have already pointed out that the functional core concept theory is
limited to predictions about the acquisition of nominals. Bowerman men-
tions this point, and turns to a discussion of the fact that Eva’s and Christy’s
non-nominal terms could not be explained by either functional or percep-
tual features. She argues that the overextension of several of these words
appeared to be based on the subjective experiences of the child. She offers
the following interpretations (8.22) of the words given in her Table 2:
(8.22) ‘there!’ the experience of completing a project;
‘aha!’ the experience of discovery and surprise;
‘too tight’ protest in situations involving physical restriction or
interference;
‘heavy’ physical exertion, often unsuccessful, with an object,
whether or not it is heavy;
Bowerman concludes from data such as that in her Tables 1 and 2 that
theories of word meaning need to expand to include the possibility that
overextensions can occur on the basis of several kinds of similarities, not
just one. In this way, her approach is similar to Anglin’s in trying to
incorporate at least both perceptual and functional features.
Bowerman’s next two tables turn to the issue of whether children’s first
overextensions are complexive or noncomplexive. We discussed this issue
in Chapter 6 in relation to Rescorla (1980), where complexive uses (or
associative complexes) refer to cases where the child links the use of a word
from one situation to another but does not relate all the uses to one shared
feature or set of features. Bowerman draws the following conclusions about
this issue: (i) both complexive and noncomplexive overextensions occur
from the onset of acquisition (a claim later supported by Rescorla’s study);
and (ii) complexive use was somewhat more common for action words than
for object words, suggesting that its use cannot be directly tied to the
attainment of object permanence (an issue discussed already in Chapter 6).
In presenting the examples in her Table 3 to show early noncomplexive
words, Bowerman emphasizes a striking feature of her daughters’ word
meanings, i.e. the way in which the girls could assign very different
meanings to the same words. This was particularly noteworthy since we can
assume that the environment was relatively similar for the two children. The
best example of this was Christy’s ‘on-off’ and Eva’s ‘off’ which are
summarized in (8.23):
(8.23) Christy ‘on-off’: ‘any act involving the separation or coming together
of two objects or parts of an object’ (p. 271), e.g. getting socks on or
off; getting off a spring-horse; taking pop-beads on and off; separat-
ing stacked cups; putting phone on hook; etc.
Eva ‘off’: ‘removal of clothes and other objects from the body’
(p. 271), e.g. for shoes, car safety harness, glasses, pacifier, bib,
diaper, etc.
Table 8.16 Prototypes proposed by Bowerman (1978a) for Eva and
Christy
Word Prototype Features
‘kick’ kicking a ball with the foot so that it is a. waving limb

propelled forward b. sudden sharp contact
c. object propelled
‘night-night’ person or doll lying down in bed or a. crib, bed
crib b. blanket
c. non-normal horizontal position
‘close’ closing drawers, doors, boxes, jars, a. bringing together two objects or
etc. parts
b. cause something to be concealed
‘open’ opening drawers, doors, boxes, jars, a. separation of parts which were in
etc. contact
b. cause something to be revealed
‘giddiup’ bouncing on a spring horse a. horse
b. bouncing motion
c. sitting on toy
‘moon’ the real moon a. circular shape
b. yellow color
c. shiny surface
d. viewing position
e. flatness
f. broad expanse
Such observations support the view of the child as an active organizer of her
linguistic experiences.
The last of Bowerman’s tables turns to examples of complexive use of
words, that is, cases where the features change from instance to instance.
Bowerman emphasizes that most of these cases indicated prototypical
structure, that is, the instances of overextension would share at least one
feature with some central or prototypical example. Such a situation is
distinct from the possible one in which the child links or ‘chains’ words from
situation to situation without any shared features across the instances.
Bowerman offers the analyses in Table 8.16 of the examples in her Table 4
(see Bowerman 1978a: 274-7 for details).
We can see how diverse these activities can be by looking at some of the
uses by Eva of the word ‘kick’. It was used in the following situations,
among others: seeing a picture of a kitten with a ball near its paw; watching
cartoon turtles on TV doing the can-can; pushing her stomach against a
mirror.
Bowerman’s analysis leads her to conclude that her daughters were
showing adult-like categories at a very early age of development. She states
(p. 278): ‘These findings suggest that there is less discontinuity between
child and adult methods of classification than has often been supposed.’ She
takes the examples of early complexive words as indicative that the children
were able to apply a featural analysis to the data, and that these features
were similar to those used by adults. She does not discuss, however, how the
child gets from its early prototypes to those of the adult language.
8.3.4 Some findings on the acquisition of relational words

While extensive attention has been paid to the acquisition of referential
words, there is also a large literature on the development of relational
words. These are words which ‘represent abstract concepts (relationships)
that cannot be directly perceived or referred in the environment’ (Blewitt
1982: 153). An example would be the temporal term ‘before’ which places
an event in relation to some other event. Here we will use the term to
include dimensional words which have some physical representation, for
example, spatial terms such as ‘big’ and ‘little’. These words share the
characteristic that their meaning is tied to relating one referential term to
another. For example, a ‘big’ mouse is actually smaller than a ‘small’
elephant, but the meaning of each word is relative to the norms set for the
sizes of the objects being discussed.
The study of relational word acquisition has differed in several respects
from the study of referential words. First, the theories which have been
proposed for referential words do not always work for relational words.
Neither functional core concept theory nor prototype theory make claims
on how such words are acquired. It is difficult, for example, to envision what
a prototype for a word like ‘before’ is supposed to be. It is not surprising,
therefore, that much of the research in this area has been concerned with
testing the predictions made by the semantic feature hypothesis.
A second difference in the study of relational words concerns the focus of
such studies. Referential word acquisition is often seen as a process which
involves assigning the relevant features to a word in order to have the adult
meaning. Importantly, these features are ones considered available to the
child, so that the process is one of selection from an available set of features
(as just discussed for Bowerman). The child, for example, is conceptually
aware of cats, dogs, cows, etc. and needs to attach the appropriate features
to the appropriate phonetic strings. Relational words, however, involve
more directly the issues of complexity and cognitive ability. Regarding
complexity, relational words can often be placed on a scale of least to most
complex. For example, ‘big’ and ‘little’ can be viewed as having a relation in
which ‘big’ is in some sense less complex than ‘small’since, in the unmarked
case, we ask ‘How big are you?’ not ‘How small are you?’ Such relations
lead to predictions that the less complex, or unmarked, terms are acquired
before more complex ones, i.e. that order of acquisition is determined by
featural complexity. Such relations are less identifiable with referential
words, although they do exist (e.g. ‘boy’ vs. ‘brother’).
Related to the issue of complexity is that of cognitive ability. There is the
possibility that more complex relational words are acquired late not because
they have more features to be acquired, but rather because the child has not
acquired the concept which underlies them. This can be exemplified by
looking at the acquisition of question words. It has been known for some
time that English children acquire wh-question words in a relatively
predictable order (e.g. M. Smith 1933; Tyack & Ingram 1977). The first
question words are typically ‘what’ and ‘where’, followed by ‘why’, then
‘how’ and ‘when’. This order involves increasing abstractness from the
relatively concrete terms ‘what’ and ‘where’ which refer to objects and
places to the more abstract temporal term ‘when’. The late acquisition of
‘when’, therefore, may tell more about the child’s cognitive development
than it does about the organization of the semantic system. If order of
acquisition of dimensional terms is strictly the result of cognitive develop-
ment, then we could hold a view that the child’s semantic classificatory
processes are adult-like, and that words are acquired appropriately as
concepts become available. For purposes of discussion, we will refer to this
hypothetical position as the cognition theory.
A further characteristic of relational word acquisition concerns the child’s
behavior when he has to process a relational word which has not been
acquired, be it due to an incomplete semantic system or incomplete
cognition. It has been observed by several investigators that in such cases
children will respond with some form of non-linguistic strategy. For
example, Eve Clark (1973b) has observed that children who have not
acquired the complete meanings of prepositions such as ‘in’ or ‘on’ will
simply do what looks most appropriate, given the situation. For example, if
handed a ball and an open container, the child will place the ball into the
container rather than alongside or underneath. If the object is a card and a
table, however, it is more likely that the card will be placed on the table. As
work proceeded on the acquisition of relational terms, it was found that
results of studies were frequently contradictory. This led to a review of the
methods used in such studies, and to the further observation that children
not only used non-linguistic strategies, but that these would vary according
to a number of possible factors in the test situation. Much of the work in this
area, then, has concentrated on determining the range of such non-linguis-
tic or cognitive strategies.
It can be seen from this discussion of the semantic feature hypothesis and
cognition theory that both approaches put the emphasis upon the child’s
cognitive rather than linguistic development. Further, a great deal of the
discussion has been by necessity on methodological issues. The result has
been that we know more about cognitive development, test methods, and
the general order of acquisition of some relational words. It is not clear,
however, that we know a lot more about the development of the semantic
system.
In an attempt to provide a flavor of the kind of research done in this area,
three kinds of relational words which have been studied in some detail will
be briefly considered here. These are the temporal terms ‘before’/’after’,
the dimensional terms ‘big’/‘tall’,and the locational terms ‘in’, ‘on’, ‘under’,
‘beside’, ‘front’ and ‘back’. As with other topics, specific pivotal studies will
be focussed upon before discussion of subsequent research.
Before/after One of Clark’s first examinations of the semantic feature

hypothesis was her study of children’s comprehension and production of the
temporal words ‘before’ and ‘after’ (Clark 1971). She selected these words
because she felt that they are a clear case where one term, ‘before’, could be
shown to be semantically simpler than the other, ‘after’. Her semantic
feature analysis of the words is presented in (8.24), where the features for
‘when’ are also given.
(8.24) Feature ‘before’ ‘after’ ‘when’
time + + +
simultaneous - - +
prior + -
‘When’ is the simplest of the three terms since it is not marked for the
feature [prior]. ‘Before’ is the next in complexity because it is positively
marked for [prior]. There is minimal justification for this decision, with
references to Leech (1970) and H. Clark (1970) for further argumentation.
The major claim is that ‘before’ is more basic because ‘it is used to describe
the visual perceptual field, whereas its counterpart “behind” . . . is negative
since it describes the area that is out of sight’ (Eve Clark 1973b: 267). The
predicted order of acquisition based on complexity then is ‘when’ > ‘before’
> ‘after’.
‘Before’ and ‘after’ were tested in four kinds of constructions, referred to
as ‘before]’, ‘before2’, ‘afterl’ and ‘afte1-2’. Each of the constructions
contained two clauses, so that the subscripts 1and 2 refer to the position in
the sentence of the clause that began with the test word. Examples of the
four types of construction are given in (8.25).
(8.25) Before2 ‘He jumped the gate before he patted the dog’ EIEz
Before1 ‘Before he patted the dog, he jumped the gate’ E2E1
After2 ‘He patted the dog after he jumped the gate’ E2E 1
After1 ‘After he jumped the gate, he patted the dog’ E1E2
A further feature of these constructions concerns the order of the events in

relation to their mention in the sentences. The ‘before2’ and ‘afterl’
sentences mention the events in the order in which they occurred. That is,
the first event, the jumping of the gate (El), is mentioned before the second
event, the patting of the dog (E2). Clark observes that children have been
found to acquire complex sentences like this with the main clause preceding
the subordinate clause. Notice that the ‘before2’ sentence with the ‘before’
in the second clause also mirrors the natural order of the events, i.e. E1E2.
She argues that these two factors make this sentence pattern easier to
acquire, thus another argument for the acquisition of ‘before’ ahead of
‘after’.
The subjects in the study were 40 children (24girls, 16 boys), ten in each
of four age groups: I 3;0-3;5,I1 3;6-3;11,I11 4;0-4;5,and IV 4;6-5;O.The
mean ages were 3;2,3;8,4;3, and 4;11respectively. Each child participated
in a production task, then a comprehension task. This order of presentation
is typical of studies which test both production and comprehension, since
the use of a comprehension task first might provide models to the child
which she could use in the production task. In the production task, the child
watched two events, and then was asked a ‘when’ question about each
event, e.g. ‘When did the boy pat the dog?’, ‘When did the boy jump the
gate?’ In the comprehension task, the child was presented with a variety of
toys and then asked to act out a sentence presented by the experimenter.
Each of the four sentence types was presented eight times for a total of 32
test sentences in each task.
On the comprehension task, an error was recorded when the child acted
out the two events in the wrong order. The measure used to score the results
was the percentage of errors on each of the four sentence types. The general
results on the comprehension task were as follows. First, each age group did
significantly better than the previous one. The percentages of errors for the
four groups were 49,38,28,and 17 respectively. Second, the four sentence
types showed three significant degrees of difficulty: ‘before2’ and ‘afterl’
were the easiest sentence types with only 4 percent and 11 percent of errors
overall respectively, followed by ‘before1’ with 48 percent, with ‘after2’ the
most difficult with 69 percent of errors. Since ‘before2’ and ‘after,’ both
involved the ELE2order of events, Clark interprets this as indicative that
some of the children were using an order of mention strategy. That is, they
had not yet acquired the [prior] feature for the words and were acting out
the events in the order in which they were mentioned. Third, the two
constructions with ‘before’ were significantly easier than the two with
‘after’, confirming Clark’s prediction that ‘before’ was easier than ‘after’.
A problem in interpreting the results was that ‘afterl’ was significantly
better than ‘after;?’,and that ‘before2’was significantly easier than ‘before,’,
Table 8.17 Stages, number of subjects, and percentage of errors on four
construction types for ‘before’ and ‘after’, adapted from Clark (1971)
Construction and order of events

No. of
Stage subjects before I (EzEl) before2 (EIE2) afterl (EI&) after2 ( E ~ E I )
A 21 80% (wrong) 4% (correct) 10% (correct) 83% (wrong)

B1 7 4% (correct) 4% (correct) 0% (correct) 89% (wrong)
I32 3 25% (correct) 8% (correct) 75% (wrong) 92% (wrong)
C 8 9% (correct) 9% (correct) 0% (correct) 6% (correct)
indicating that something else was going on in the data. Clark attempted to
capture this further factor by dividing the subjects into four stages of
acquisition. These stages are as follows:
Stage
A Use the order of mention strategy to interpret sentence
B1 Correct understanding of ‘before’; use order of mention strategy
on sentences with ‘after’
B2 Correct understanding of ‘before’; interpret ‘after’ as if it meant
‘before’
C Correct understanding of ‘before’ and ‘after’
The data on which these stages are based are presented in Table 8.17. The
stages above also lead to strong claims about the order of acquisition of
words like ‘after’ and ‘before’ which only differ in a single feature. Here we
restate the four stages above in more general terms:
Proposed stages in the acquisition of antonyms
Stage Interpretation and predictions
A Acquisition of superordinate feature: use a non-linguistic
strategy for both terms
B1 Acquisition of unmarked word: restrict use of non-linguistic
strategy to marked word
B2 Acquisition of unmarked word: treat marked word as if it has
the unmarked meaning
C Acquisition of marked word: correct understanding of both
words
Clark interprets the results as supporting the semantic feature hypothesis.
She states (p. 275): ‘children learn the meanings of words component by
component, and furthermore, where these components are hierarchically
related to each other, the feature hierarchy is learned in order beginning
with the superordinate component, e.g. +Time, followed by +/- Simul-
taneous, followed by +/- Prior’.
This study inspired a great deal of discussion and further study, both on
‘before’ and ‘after’ and on other works. In Ingram (1981b) some general
questions were raised about the interpretation of the data. As already men-
tioned, the difference between ‘beforel’ and ‘before2’ is not explained. It
appears to result from the drop in performance on ‘beforel’at stage B2. If we
look at B2, we see that this is the result of the use of an order of mention
strategy that affects both ‘after’ and ‘before’. Note that both pairs of these
words have 17percent more errorson the words with E2E1order, This creates
alogical problem for B1 and B2 since the child is supposed to have abandoned
the order of mention strategy for ‘before’ at B1 and for ‘after’ at B2. A further
problem is that there is no mechanism proposed which gets the child from B1
to B2. At B1, the child knows that ‘after’ is different from ‘before’ since it is
treated differently. What causes the child to abandon this knowledge and
treat them the same at the next stage? Lastly, there was no age effect between
B1 andB2, and thus no independent measure for them. They thus constitute a
‘stage error’ in the sense of Chapter 3. It is concluded in Ingram (1981b) that
there is little evidence to consider these as separate stages; they should only
possibly be considered as two different ways to acquire word meaning.
Further research on these words has led to other problems with Clark‘s
original study and further insights into the acquisition of word meaning.
These studies have been excellently reviewed in Blewitt (1982), on which
I will base my review here. The general findings have been as follows.
First, no study has replicated the finding that for some children ‘after’ is
interpreted as ‘before’. This is a striking result which has devastating effects
for the proposed stage B2. If true, it also is one that requires some
explanation, in that evidence exists that such a general stage may exist for
some other pairs, e.g. ‘more’ and ‘less’ (Donaldson & Wales 1970), ‘tell’
and ‘ask’ (C. Chomsky 1969), where children have been found to interpret
the second word in each pair as if it had the meaning of the first one. Second,
the order of acquisition of ‘before’ and ‘after’ has been controversial. Some
studies have replicated Clark’s result, while some others have found ‘after’
acquired first (Barrie-Blackley 1973; Amidon 1976; Feagans 1980), and
others have found no difference in their order of acquisition (Amidon &
Carey 1972; Friedman & Seeley 1976; French & Brown 1977).
Blewitt’s review of these studies leads her to propose several factors
which have affected children’s performance on a language comprehension
task. Three of these are summarized below:
(i) The structure ofthe sentence: Amidon & Carey (1972) used sentences
such as ‘Before you move the green car, move the red car’ where the
agent is 2nd instead of 3rd person and there is one action instead of
two. Children performed worse than in the Clark study.
(ii) The syntactic role of the test word: Coker (1978) tested with three
different tasks where ‘before’ and ‘after’ were prepositions in two
tasks and adverbs in the third, as in Clark’s study. ‘Before’ was harder
than ‘after’ on task 1, the same on task 2, and easier on task 3.
(iii) Semantic content of the sentence: French & Brown (1977) compared
sentences with logically constrained events, e.g. ‘After Raggedy Ann
fills the bottle, she feeds the baby’, with sentences that are not
logically constrained, i.e. as in the Clark study. Children performed
better on logically constrained sentences.
The first one, the structure ofthe sentence, is a factor which was also found in
Carol Chomsky’s study of ‘ask’ and ‘tell’. She found that children would
interpret ‘ask’ as if it meant ‘tell’, but that it would be restricted to only
certain sentence patterns, at least for older children. This factor suggests
that the child may have the correct meanings of the words, but that
processing more complex sentence structures affects the retrieval of more
recently acquired meanings. The second factor points out the crucial way in
which the syntactic role of the test word may influence the results of a study.
Coker replicated Clark’s results when testing ‘before’ and ‘after’ as subord-
inate adverbs, but found that children did better with ‘after’ when they were
tested as prepositions. This initially suggests the peculiar situation that the
relation of markedness between two words may vary according to their
grammatical role. Since the effect was influenced by the questions
addressed to the child, however, it appears more likely that it was an
experimentally induced bias. The third factor, the influence of the semantic
content ofthe sentence, is a performance factor like the first one. That is, it
indicates how earlier acquisition can be demonstrated if facilitative test
sentences are used.
Bidtall Possibly the most important claim of the semantic feature hypo-
thesis is that the order of acquisition of relational words is based on their
semantic complexity. While this claim did not hold up very well for ‘before’
and ‘after’, they are a single pair which are antonyms. It is still possible that
the claim may hold between related pairs of antonyms such as ‘big’l’little’
vs. ‘tall’l’short’. According to Bierwisch (1967), ‘big’/‘little’ are the least
complex size words because they are global and are not limited to a specific
dimension. ‘Tall’l’short’, however, are limited to the dimension of height.
Other such pairs in increasing complexity are ‘high’/’low’, ‘long’l’short’, and
‘wide’l’narrow’. Because dimensional terms can be graded in this fashion,
they have proved to be a rich area for the study of semantic acquisition in
general and of the semantic feature hypothesis in particular.
Several studies have supported the findings that such pairs have a
predictable order of acquisition based on their proposed semantic complex-
ity. Testing children between 2;6 and 3;6 on pairs of objects, Eilers, Oller &
Ellington (1974) found the acquisition order of ‘big’rlittle’ > ‘long’/
‘short’ > ‘wide’/(narrow’. Using the additional pair of ‘tall’/‘short’, Bartlett
(1976) replicated the results of Eilers, Oller & Ellington, with the additional
finding that the pair ‘tall’hhort’ was acquired simultaneously with ‘long’/
‘short’. Other studies with similar results include those by Clark (1972),
Klatzky, Clark & Macken (1973), Brewer & Stone (1975).
In the midst of studying this issue, one peculiar finding emerged. In a
series of studies on the meaning of ‘big’, Maratsos (1973,1974b) found that
3-year olds correctly used ‘big’ as predicted by the semantic feature
hypothesis, but that 4-and 5-year olds incorrectly processed ‘big’ as if it
meant ‘tall’. Further, ‘tall’ appeared to be correctly understood by all ages
tested. This finding is interesting for a number of reasons. First, it is a
problem for the semantic feature hypothesis since ‘tall’ seems to be easier to
acquire than ‘big’. Second, it poses a peculiar situation of children correctly
using a word, then changing to an incorrect meaning. Third, it has been
widely cited and has led to a number of subsequent studies. We will take a
closer look at one of the studies on which this finding is based, then turn to a
more recent study which appears to account for it.
In Experiment 1(of four) in Maratsos (1973), 30 children between 3 and 5
years of age were tested on their understanding of ‘big’. The children were
presented with pairs of cardboard rectangles and toy animals, and for each
pair, the child was asked ‘Which one is the big one?’ Maratsos states
(p. 748): ‘The comparative ‘bigger’ was not used because pilot work
observations indicated that preschool children do not have a clear under-
standing of the comparative.’ Since the stimuli for such experiments have
proved to be crucial for the results, we will present them in detail. Maratsos
used three test pairs and one pretest pair for each of the three types of
rectangles, and three pairs of toy animals. The stimuli used are summarized
below:
Pretest pairs: larger rectangle is higher and wider;
Unequal height and width pairs: one rectangle is higher and narrower
than the other (G & H in Figure 8.1);
Equal heightpairs: one rectangle is wider than the other but both are
equal height (C & D in Figure 8.1);
‘Animal‘pairs: giraffe vs. hippopotamus; donkey vs. bear; policeman
vs. leopard.
Facsimiles of the test pairs are shown in Figure 8.1 as pairs C, D, G, and H
in the inventory of stimuli later presented by R a m & Gelman (1984).
421
U N E Q U A L HEIGHT
and WIDTH
(1)
EXTREME (L)-
DIMENSION
nn n
(P) uu ((ir) Uo
Figure 8.1 Sets of rectangles constructed by Ravn & Gelman (1984) for testing
‘bigness’.
The results were scored in terms of percentages of correct responses. A
correct response was to pick the member of the pair with greater overall
size, e.g. the second rectangle in the pairs C, D, G , and Hof Figure 8.1. The
3-year-olds show the results of the studies mentioned above where young
children respond correctly to ‘big’. In this case 73 percent chose the correct
item. This correct response pattern, however, drops dramatically for the 4-
and 5-year olds to only 37 percent and 23 percent respectively. Maratsos
interprets this as indicating that the older children understood ‘big’ as if it
meant ‘tall’, citing the following anecdote (p. 749) as an example:
At one point, one boy, who consistently answered incorrectly, took
the shorter animal (the leopard) and set it on its rear legs, which made
it taller than the other figure. H e then exclaimed, ‘Look! Now it’s the
big one!‘
This general finding was substantiated in other experiments reported in
Maratsos (1973) and (1974).
This is a peculiar finding for several reasons. Besides those mentioned,
there is the further question of whether this result is unique to the
acquisition of ‘big’, or is a characteristic of semantic acquisition. We could
propose that it is a general feature of acquisition when the child is acquiring
words where one has a more specific meaning than another. This could be
done by adapting the stages on page 417 for the acquisition of pairs like ‘big’
and ‘tall’. If so, we would need a revised stage B, which is given below:
Stage B:
Loss of acquisition of unmarked word: upon acquisition of marked
term, reinterpret the unmarked word as having the meaning of the
marked one
As such, this proposal is very strong since it implies that the shift should
occur for a wide range of words which bear this relation. Since such shifts
are not widely attested in the literature, we will assume for the current
discussion that it is a specific aspect of the acquisition of size words.
A recent study by Ravn & Gelman (1984) has provided an explanation for
the results found by Maratsos. Ravn & Gelman picked up on an earlier
suggestion by Bausano & Jeffrey (1975) that older children may not be
treating ‘big’ as if it meant ‘tall’, but that another meaning might be used.
Bausano & Jeffrey point out that the same results would occur for the older
children if they were using a ‘salient-dimensional difference rule’, i.e.
picking out the dimension as ‘big’ which was most different between the
rectangles. Ravn & Gelman build upon this by proposing five possible rules
which children might use in judging ‘bigness’. These rules are summarized
below:
Table 8.18 Predicted patterns of response forfive rules on four types of
rectangle pairs from Figure 8.1, adapted from Ravn & Gelman
(1984:2145, Table 2). (1st = first rectangle selected; 2nd = second
rectangle selected)
Stimuli pairs
Rule K L M N
Correct rule 2nd 2nd 2nd 2nd

Height rule 2nd 2nd 1st 1st
Width rule 1st 1st 2nd 2nd
Salient-dimensional difference rule 1st 2nd 1st 2nd
Salient-dimension rule 1st 1st 1st 1st
Definition
Correct rule: Biggest area in two dimensions
Height rule: ‘tall’
Width rule: ‘wide’
Salient-dimensional difference rule: Biggest in whichever dimension
shows the greatest (most salient) difference
Salient-dimension rule: Biggest in the dimension of greatest extent
To test these rules, Ravn & Gelman constructed several sets of rectangles
which have been reproduced here as figure 8.1. They were constructed in a
way such that each of the five rules should lead to a different set of choices
from the different sets of rectangles. Table 8.18 gives the predicted patterns
of responses for each of the five rules for four of the pairs tested. These pairs
proved to be the most useful ones in distinguishing between the five rules
since the dimensional differences were most clear-cut.
As in the study by Maratsos, the subjects were 30 children between 3 and
5 years of age, ten at each age level. The method was also similar in that the
children were asked to point out which of the rectangles was the big one.
Each child’s responses were coded into five scores, one for each of the rules
tested. Here we will give just the results on the four pairs listed in
Table 8.18. The child was assumed to be using one of the five rules if all four
pairs were scored as correct for that rule. Table 8.19 gives the results for the
three age groups.
The results show clearly that the children at all ages have a preference for
the ‘height’ rule, the preference increasing with age. While the increased
use of the height rule is consistent with Maratsos’s results, the development
of the ‘correct’ rule is not. Recall that Maratsos found that 3-year olds
processed ‘big’ correctly, while here none of the 3-year olds have correct
Table 8.19 Percentage of children who used one ofjive rules adapted from
Ravn & Gelman (1984: 2147, Table 4)
Age in years
All
age
Rule 3 4 5 groups
Correct rule 0 10 20 10
Height rule 40 60 70 57
Width rule 0 0 0 0
Salient-dimensional difference rule 0 0 0 0
Salient-dimension rule 30 20 0 17
N o rule 30 10 10 17
understanding of ‘big’. Ravn & Gelman account for this discrepancy as

follows (p. 2147):
This seemingly paradoxical finding can be explained by the difference
in approach between our study and Maratsos’s. In each of his
experiments, Maratsos used a limited set of items and assumed that if
children were not using the rule being tested (height or width
depending on the experiment), they were using the correct rule. In
contrast, we selected a range of items that would allow us to pit several
rules against each other in one experiment.
To demonstrate the effect of the rectangle pairs, Ravn & Gelman scored the
percentages of correct responses on the pairs C , D, G, and H, the four pairs
used by Maratsos. They found the scores to be 65%, 52% and 40% for 3-,
4-, and 5-year olds respectively. These figures are somewhat higher than,
but consistent with, the results reported by Maratsos. The rectangles K
through N, then, proved to be a better test of the meaning of ‘big’ than C ,
D, G , and H.
Another analysis conducted by Ravn & Gelman looked at the consistency
of the children’s rule use. They assumed that the higher the score on a rule,
the more consistently it was being used. Statistical analysis revealed that the
older children were more consistent in their rule usage than the younger
children. Their conclusion about the acquisition of ‘big’, then, is somewhat
different than that drawn by Maratsos. Children do not start out with a
correct understanding of ‘big’ and then shift to a height rule for its meaning.
Rather, they begin with the height rule, but do not apply it as consistently at
the onset.
While their findings differ somewhat from Maratsos’s, they resemble his
in being a problem for the semantic feature hypothesis. If children begin
with a more narrow meaning for ‘big’, such as ‘tall’, then acquisition is not
proceeding by acquiring more general (or less complex) features to more
specific ones.
Given these difficulties with the semantic feature hypothesis, it is not

surprising that research has begun to explore other accounts of the
acquisition of relational terms. In her study of size terms, Bartlett observed
that the children’s responses seemed to be affected by the dimensions of
individual test items, a result just seen above for rectangles. Bartlett made
the following remark regarding areas for future research (1976: 214): ‘One
important task for future research, then, is to assess the relative effects of
cognitive complexity, word frequency and perceptual salience on orders of
acquisition for these dimensional features.’ Research exploring the last two
of these suggestions will be examined briefly here. The first suggestion will
be considered in the discussion on the acquisition of spatial terms.
Keil(l979) has attempted to develop an account of semantic acquisition
based on a version of prototype theory. Importantly, this work deals not
only with referential words, but with relational ones. Keil & Carroll (1980)
provide a useful summary of research from this perspective on the acqui-
sition of ‘tall’. A general premise is that the child first acquires a meaning for
a word that is restricted to the first instances in which it is learned. That is, it
is restricted to a small set of exemplars. Keil & Carroll state (1980: 21f):
The closest view to ours is Carey’s (1978) ‘missing features-plus-
haphazard-example-theory’. Carey suggests that the child may
acquire the meanings of spatial terms such as ‘tall’ by first learning
idiosyncratic, object-particular meanings and later learning a more
general concept that applies to all objects. Thus, the word ‘tall’ may
have different meanings for different objects.
That is, the meaning of the word for the child will be affected by the
perceptual salience of different dimensions for different objects.
Keil & Carroll pursue this theory by contrasting its predictions on
acquisition with those of the semantic feature hypothesis. If the child is
acquiring semantic features for word meanings, then she should apply them
consistently across objects. If the meaning is influenced by the aspects of
individual objects, then the meaning should vary.
To test these alternatives, Keil & Carroll report on two experiments on
the acquisition of ‘tall’. In the first one, 36 children between 3 and 6 years of
age were shown 18 pictures with three objects shown on each. There were
‘tall items’ where the objects varied in height but had the same width, and
‘wide items’ where the height was the same but the width varied. Keil &
Carroll comment (p. 23): ‘The objects ranged from familiar ones that are
easy to think of as tall to ones to which “tall” cannot be so naturally
applied.’ The children were told the following (p. 23): ‘Here are three X’s
(e.g. houses). Now is one of these the tallest, or are they all the same in
tallness?’ The results supported the predictions of Keil’s prototype model.
The children were correct for some items, but incorrect for others. They
were significantly better at objects which had vertical lines (e.g. pencils),
than they were for ones with curves (e.g. mountains).
The second study also supported this finding, but added some insight into
development over time. In the second experiment, the children were shown
the same pictures, but were told that they were different objects. For
example, the picture of mountains was now a picture of blankets. The
importance of this shift is that mountains are more naturally thought of as
being ‘tall’ than are ‘blankets’. The older children (4;6 to 7;6) were
significantly more likely to make an error on the less natural objects than
were the younger children. The fact that these changes in judgement
occurred with the same pictures indicates that the judgements were
semantic, not just perceptual, in nature. Also, the finding that older
children were consistent when interpreting the pictures differently indicates
that they were associating ‘tall’ with specific kinds of objects.
The finding of the second experiment shows that the child’s linguistic
experience with objects plays an important role in applying a word such as
‘tall’, overriding what looked like a perceptual effect in the first experiment.
Blewitt (1982) discusses input frequency as one such factor which may
influence how the child acquires such words. We have already discussed
briefly at the end of Chapter 6 the possible effects of frequency of exposure.
Blewitt reviews her own work as well as that of others which suggest that
orders of acquisition also correlate in many cases with frequency of use.
Regarding size words, she states the following (p. 169): ‘. . .the relative
frequency with which these words are used (e.g., KuCera & Francis 1967)
also matches their relative complexity. If the frequency of use in child-
directed speech is comparable to the general frequencies of use, relative
frequency also could account for the observed order of acquisition. . .’
While presenting this possibility, she also entertains another factor, that of
cognitive complexity. We turn to this aspect in the next section on the
acquisition of spatial words.
Spatial words Still another factor which could affect the order of acqui-
sition of a set of words is the child’s cognitive development. For example, we
could propose that a child has not acquired the meaning of a particular word
because he does not have the concept which underlies it. This possibility has
been demonstrated in particular with the acquisition of spatial terms. There
are two complementary kinds of evidence which can be brought forward to
prove this possibility: one would be the demonstration of similar develop-
ment in children across linguistic communities; the other would be the
demonstration that children acquire certain words only after they have
shown competence for their underlying concepts on a non-linguistic cogni-
tive task. Both kinds of evidence, in fact, have been presented in a series of
works by Johnston and Slobin (e.g. Johnston & Slobin 1979; Johnston
1979). An excellent summary of them can be found in Johnston (1986).
Johnston & Slobin (1979) examined the acquisition of the locative
expressions ‘in’, ‘on’, ‘under’, ‘beside’, ‘between’, ‘back’, and ‘front’ in four
languages: English, Italian, Serbo-Croatian, and Turkish. Based on
Piagetian research (Piaget & Inhelder 1967), they predicted that the order
of acquisition (below) should occur regardless of linguistic community if
cognitive development were the sole determinant of acquisition.
Predicted order of acquisition and justiJication (Johnston & Slobin
1979)
1. ‘in’, ‘on’, ‘under’: the first spatial concepts are of containment,
support, and occlusion.
2. ‘beside’: a purely spatial proximity relation, not dependent on the
speaker’s viewpoint.
3. ‘front;, ‘backi (Le. of objects which have inherent fronts and backs,
e.g. houses): proximity to an inherent feature.
4. ‘between’: coordination of two proximity relations.
5. ‘front’, ‘back’ (in relation to objects without inherent fronts and
backs, e.g. balls): coordination of the relative proximities of the
speaker, reference object, and located object.
Forty-eight children between the ages of 2 and 4 years of age were tested
in each of the four linguistic communities. Most generally, each child was
tested by an experimenter who placed a ‘reference object’ in front of the
child, and then a ‘located object’ in relation to it. For example, the child
might be shown a plate (reference object) with a stone (located object) next
to it, and then be asked ‘Where is the stone standing?’ Children were
credited with having acquired a particular locative word if they used it more
often in appropriate contexts than in inappropriate ones. The data from the
study were subjected to a variety of statistical analyses. The results from one
such analysis, involving Guttmann scaling procedures, are presented in
Table 8.20.
The analysis results in three groups of locative terms being acquired for
each language. Four of the words, ‘in’, ‘on’, ‘under’, and ‘beside’, were
always acquired before the others. This was followed by a second group of
‘between’, ‘back; and ‘front;. The last two words acquired were ‘back’ and
‘front’. This pattern is in general agreement with the cognitive predictions,
Table 8.20 Order of acquisition of locative expressions in four languages
and the percentage of subjects producing each, taken from Johnston &
Slobin (1979: Table 5) ~
Scale point English Italian Serbo-Croatian Turkish
in 90 in 91 on 88 in 90
on 83 on 88 in 84 on 80
under 81 under 84 beside 82 under 79
beside 74 beside 77 under 72 beside 79
between 49 between 57 back1 31 backt 71
frontr 30 backt 42 between 26 frontr 53
back1 21 frontt 41 frontt 19 between 50
back 14 back 23 back 16 back 7
front 3 front 18 front 12 front 4
with less specificity for ‘beside’ and ‘between’, which were each expected to
constitute a separate grouping. Also, there is variability within the group-
ings. For example, the English children acquired ‘between’ before ‘back;,
while the Turkish children showed the opposite order.
To account for the instances where cognitive complexity did not
account for the order of acquisition, Johnston & Slobin propose a second
factor, that of linguistic complexity. They discuss five ways in which
linguistic complexity might affect acquisition: (i) position of word - some
evidence suggests that postpositions (e.g. Turkish) are easier to acquire
than prepositions (English); (ii) lexical diversity - the number of terms
available in the language for a particular proximity (for example, Turkish
‘yannida’ captures the meanings of English ‘beside’, ‘by’, ‘next to’, ‘near’,
and ‘close to’); (iii) clear etymology - the extent to which a word’s
meaning is semantically transparent (for example, English ‘back’ names
an identifiable body part, while ‘between’ doe$ not); (iv) morphological
complexity - e.g. ‘in’ vs. ‘on top of’; and (v) homonymity - the extent to
which one term covers several meanings, e.g. ‘back‘ in English for both
featured and nonfeatured objects. Using an elaborate coding system,
Johnston & Slobin score the results for the influence of these five factors
and conclude (p. 541): ‘Wherever conceptual complexity fails to predict
actual order of acquisition, we find some pocket of relative linguistic
difficulty.’
The second kind of evidence in support of the effects of cognitive
complexity can be found in Johnston (1979). Johnston focussed on the
acquisition of the English terms ‘in front of’ and ‘in back of’. To acquire the
full range of meanings of these terms, the child needs to acquire several
concepts which are listed here in order of increasing complexity.
Concept Description
PROXIMITY The concept that two objects are in a
spatial relation to each other
OBJECT FEATURE The identification of an inherent spatial
orientation for an object such as having a
‘front’ and ‘back’
ORDER The concept that objects can be in a
particular linear arrangement with one
another
PROJECTIVE RELATIONS The conceptualization that the spatial
relation between objects is relative, such
as in relation to a viewer’s line of sight
Johnston hypothesized that children’s use of the terms ‘in front of’ and ‘in
back of’ would parallel the order of complexity given above. For example,
early uses of ‘in front of’ would be in the sense of ‘next to the front of some
object with an identifiable front’. This use (which we will call the featured
use) would only require the concepts PROXIMITY and OBJECT FEATURE. Later
uses with nonfeatured objects would have the meaning ‘first in line of sight’,
which requires the concepts ORDER and PROJECTIVE RELATIONS. (Following
Johnston, we will refer to this as the deictic use of the term.)
Johnston discusses three kinds of evidence that are necessary to demon-
strate that semantic development is determined by cognitive development.
First, there needs to be a correlation between performance on linguistic
tasks and the nonverbal cognitive tasks of the presumed cognitive precur-
sor. Second, the cognitive ability must precede the linguistic ability. Third,
there need to be similar orders of acquisition of concepts and their linguistic
forms. For example, children who use ‘in front of’ only to mean ‘in front of
some object with an identifiable front’ should (i) be able to pass nonverbal
tests of the concepts PROXIMITY and OBJECT FEATURE; (ii) show acquisition of
PROXIMITY and OBJECT FEATURE before the use of ‘in front of’; and (iii) show
an order of acquisition of the meanings of ‘in front of’ which is similar to the
order of acquisition of the related concepts.
Johnston tested the meanings of ‘in front of’ and ‘in back of’ in 33
children between 31 and 54 months of age on a variety of spatial tasks
designed to explore the range of their possible meanings. She also tested the
children on nonverbal tasks for the concepts listed above. She scored the
results in terms of the number of children to pass or fail the task on a
particular concept in relation to the number who passed or failed on the
related linguistic task. These results are summarized in Table 8.21.
The first two columns of results in Table 8.21 do not provide insight into
the relation between cognitive development and language use, since the
Table 8.21 Number of subjects tested on nonverbal (concept) tasks and
linguistic (‘in back of’and ‘infront of’) tasks who either (i) passed both
tasks, (ii) failed both tasks, (iii) passed only the linguistic task, or (iv)
passed only the nonverbal task, on three kinds of comparisons between
concept and use. Adapted from Johnston (1986: Table 1 )
Same on both tasks Pass only one task

Concept and use Pass Fail Linguistic Nonverbal
PROXIMITY, FEATURE
and featured use ‘back’ 9 3 1 20
‘front’ 6 3 1 23
ORDER and deictic use ‘back’ 6 20 1 6
‘front’ 0 21 0 12
PROJECTIVE RELATIONS
and deictic use ‘back’ 5 9 2 17
‘front’ 0 11 0 22
Totals 26 61 5 89
subjects here either passed or failed both tasks. The crucial data, however,
are in the third and fourth columns under ‘pass only one task‘. Here 89
responses were cases where the subjects passed the nonverbal cognitive
task, but failed the parallel linguistic one. The opposite result, however,
only occurred five times. These results show that children consistently
acquire the concepts which are related to these terms before the terms are
used in related ways.
These data provide strong evidence in support of the influence of
cognitive development on semantic development. In her recent review of
this relation, however, Johnston (1986: 983) stops short of concluding that
linguistic order of acquisition can always be determined by cognitive
development. Basing her review on some of the studies discussed above as
well as on Johnston & Slobin (1979), she states:
Taken together they [the studies reviewed] provide a solid platform
from which to argue that nonverbal conceptual development is an
important determinant of language growth . . . Further evidence from
child language studies, linguistic analysis, investigations of nonverbal
cognitive growth, and experimental tests of linguistic dependency did
not always confirm this initial estimate. Children’s learning of locative
prepositions, ‘more’/‘less’, and ‘big’/’little’ does seem to be affected by
the evolution of spatial and quantificational knowledge. But the
acquisition of dimensional adjectives other than ‘big’/‘little’ is not yet
so clearly related to parameters of conceptual growth. In that domain
factors such as frequency of input or semantic complexity may play a
more dominant role.
Cognitive factors, therefore, are an important influence, but not the only
one in explaining the child’s acquisition of word meaning.
8.3.5 Assessment
In this section we have reviewed a range of topics on the child’s acquisition
of word meaning, by outlining the most cited theories which have been
proposed to date, and then examining research into the areas of nominal
words and relational words. In doing so we have seen that a number of
predictions by the various theories have not been borne out. The semantic
feature hypothesis, for example, generated a great deal of research, and yet
support for it has been restricted at best. Carey (1982), for example, gives a
nice review of its rise and demise.
The one theory which has managed to survive so far, in several variants, is
prototype theory. This is true at least to the extent that it continues to be
mentioned in recent reviews as the theory most able to account for current
findings. Even prototype theory, however, is not without its critics. One
problem is that it does not lend itself to an account of relational words as
easily as it does to referring expressions. Another problem concerns even
the results on nominals. This form of criticism of prototype theory can be
found in Armstrong, Gleitman & Gleitman (1983). These authors attack
prototype theory on methodological grounds. They argue that the results of
numerous studies showing subjects judging exemplars as being more or less
typical of some prototype is the artifact of the way in which the studies are
conducted. Armstrong, Gleitman & Gleitman propose that human beings
are very able to provide relative judgements when asked to do so. This does
not, however, reflect their linguistic organization of the world.
Armstrong, Gleitman & Gleitman provide evidence for their claim with
results from a series of studies in which subjects were asked to judge which
exemplars were most typical of certain words. Some of the words tested
were similar to those used in other studies in support of prototype theory,
such as ‘fruit’ and ‘sport’. Other words tested, however, were what the
authors referred to as definitional concepts. These are terms which normally
cannot be broken down into features. One such term is the phrase ‘odd
number’, which is defined as ‘an integer not divisible by two without
remainder’. They found that subjects were just as able to give graded
responses to definitional terms such as ‘odd number’ as they were to other
terms such as ‘sport’. The number ‘3,, for example, was found to be the most
typical odd number. They conclude that graded responses in such experi-
ments do not reflect semantic structure, but rather an aspect of performance
which operates as a heuristic to assist people in identifying instances of
concepts in the real world. They go on to give a critique of featural theories
in general, and argue for a holistic theory of the kind mentioned briefly at
the beginning of 8.3.1.
All of this leaves the area of semantic acquisition somewhat in a
theoretical hiatus. The research that continues has turned to some of the
factors which have been discussed above, such as frequency of input and
cognitive determinants. Some recent studies, for example, have tried to see
how the linguistic presentation of words affects the child’s acquisition.
Carey & Bartlett (1978), for example, attempted to teach a child a new
meaning by contrasting it with one the child already knew. The new word
was ‘chromium’, which was given the meaning of the color olive-green.
Subjects were shown two trays which were exactly the same except that one
was red and the other was olive-green. They were then asked ‘Bring me the
chromium tray, not the red one, the chromium one.’ Other research of this
kind is reported in Dockrell(l981). The results so far have been mixed as to
the extent to which such linguistic framing aids in acquisition. Such studies
also often avoid discussion of broader theoretical issue.
It is difficult to predict at this time where research into semantic
acquisition will direct itself in the years immediately ahead. A comment
made by Armstrong, Gleitman & Gleitman about research in cognitive
psychology seems equally relevant to the current state of affairs in semantic
acquisition: ‘we ourselves are not optimistic that a general theory of
categorization, one that will answer to the serious problems (explication of
functions from words to the world, and of the units that figure in phrasal
meanings and in lexical entailments) is just around the corner.’ The lesson
so far has been caution about finding simple explanations, as well as
heightened awareness of the methodological problems involved in studying
this area.
Further reading
Phonemic perception
Much of the work in this area seems to have been done in a relatively short
period of time around the early 1970s. It is also our impression that some
have felt that Barton (1976a) was the definitive study on the topic (Clumeck
1982, for example, appears very much an extension of Barton’s work). It
should be clear from our discussion that we feel that this is far from true. An
insightful recent study that bears this out is Johnson, Hardee & Long
(1981). Edwards (1974) and Eilers & Oller (1976) remain classic studies in
Further reading 433
this difficult area. An insightful methodological review can be found in
Strange & Broen (1980). A more critical overview is that of Barton (1980).
Phonological production
There are currently no textbooks devoted exclusively to phonological
acquisition, although most of the general texts mentioned in Chapter 1
contain chapters on the topic. There are, however, several books on
phonological disorders in children which discuss normal acquisition in some
detail. The primary ones are Ingram (1976a), Edwards & Shriberg (1983)
and Stoel-Gammon & Dunn (1985).
The research on phonological production has gone through several
phases, based on the theoretical view popular at the time. Up until the late
1960s,most research was geared toward testing Jakobson’s theory. The early
1970s were marked by an interest in determining the set of phonological
processes (Stampe 1969) or realization rules (Smith 1973) that map from the
adult surface form to the child’s form. Much of the most recent work in this
area has been from the perspective of the Stanford theory, with its
corresponding emphasis on individual differences.
Overviews of phonological acquisition include Edwards & Shriberg
(1983) which provides a comprehensive discussion of phonological pro-
cesses, Chapter 2 in Ingram (1976a), Menn (1983) which presents the
Stanford theory, and Chapters 2 and 3 of Stoel-Gammon & Dunn (1985).
Their Chapter 3 provides a useful overview of current theories, updating
the discussion in Ferguson & Garnica (1975) which was mentioned in our
Chapter 6.
A feel for the large sample studies can be obtained from selected readings
from Templin (1957). Earlier in this volume Chapters I and I1 were
recommended; Chapter I11 on articulation of speech sounds should now be
added. Turning to the diary studies, Velten (1943) is standard reading.
Leopold (1947) is long and somewhat tedious to read, and can be scanned to
get an idea of its contents. The section entitled ‘General phonetic problems’
(pp. 207-56) contains a number of rich observations. A brief but effective
overview of Smith (1973) is found in his first chapter. Chapter 4 is the
central chapter in which he discusses the results and presents a defense .of
his theory. Stampe’s theory can be found in the short original form in
Stampe (1969), or in the longer, more recent, review in Donegan & Stampe
(1979). The Stanford theory is presented in a general form in Macken &
Ferguson (1983). Its use in analyzing data can be seen in Macken (1979).
Word meaning
As with phonological acquisition, there is little in the way of textbooks on
the topic. There are, however, some lengthy review articles, and some
edited books of articles. Most introductory textbooks such as those men-
tioned in Chapter 1 also have chapters on this topic. One of the first edited
texts is Moore (1973), which contains the first review articles presenting the
semantic feature hypothesis (Clark 1973a), and prototype theory (Rosch
1973). More recent collections are Kuczaj (1982b) and Kuczaj & Barrett
(1986), both of which reflect the more recent prominence of prototype
theory. Nelson’s own recent booklength treatment of her views in this area
is Nelson (1986). A good review of the topic is that of Blewitt (1982). Carey
(1982) reviews the impact of the semantic feature hypothesis, its failings,
and the current lack of an accepted theory. Rice & Kemper (1984) is a
highly readable discussion of the relation between cognition and word
acquisition.
If time allows, reading of the first statements of the three most influential
recent theories by Clark (1973a), Rosch (1973), and Nelson (1974) will
provide an insight into each. The latter attempt to develop these into a more
eclectic theory in Bowerman (1978a) is a widely cited article that is also
basic reading. The most extensive data-based treatment of the topic
remains Anglin (1977); the first chapter is an excellent review of the theories
up to that time and of Anglin’s viewpoint, and the last chapter is a good
overview of his results and the interpretations he gives them. Of the recent
reviews, I lean toward Blewitt (1982). Important recent criticisms of
attempts to explain word acquisition are Carey (1982) and Armstrong,
Gleitman & Gleitman (1983). A nice summary of the possible influences
between cognition and language is contained in the first three chapters of
Rice & Kemper (1984).
9 The period of simple sentences: the acquisition of
grammatical morphemes
9.1 Introduction
We have seen in Chapter 7 that the first sentences of children consist
primarily of words which belong in the adult language to lexical categories
such as noun, verb, and adjective. Because of this, most studies of the
period of the first word combinations have concentrated on explaining the
emergence of these early lexical forms. In particular, their semantic
characteristics have been emphasized. Brown (1973), for example, des-
cribes the period as one in which the first semantic relations appear. Even
while arguing for early syntactic structure, Pinker (1984) has nonetheless
also relied on the child’s first use of semantic information to enable him to
‘bootstrap’ into the syntax.
In this chapter, we turn to how the child acquires the range of morphemes
in language that do not have the transparent semantic meanings of the
major lexical words. These morphemes are often called grammatical
morphemes (or function words) in the sense that their meanings are either
partially or totally defined by the set of rules (or grammar) of a particular
language. Brown (1973: 253) refers to them as modulations of meaning in
that they:
. . . seem to ‘tune’ or ‘modulate’ the meanings associated with the
contentives in the sense that the modulation is inconceivable without
the more basic meanings. Thus ‘a’ and ‘the’ make the thing referred to
by a noun specific or nonspecific.
Gleitman in numerous places (e.g. Wanner & Gleitman 1982) refers to
them as a closed class (as opposed to the ‘open class’ of content words)
which can be defined by both syntactic (e.g. restricted membership and
positions) and phonological (e.g. unstressed, often nonsyllabic) properties.
These morphemes tend to be acquired later than lexical morphemes and
are usually associated with development beyond the period of the first word
combinations. Brown assigns their onset to Stage 11, although their devel-
opment takes several years. Their study has been mostly descriptive, i.e.
435
436 SIMPLE SENTENCES: ACQUISITION OF G R A M M A T I C A L M O R P H E M E S
Table 9.1 A sampling of some grammatical properties not found in
English, based on Comrie ( I 981, 1983)
1. Regular causative affixation of verbs, e.g. Songhai:

Ali nga-ndi tasu di
Ali eat-causative rice the
‘Ali got someone to eat the rice’
2. Ergativity, where the subjects of transitive verbs are marked differently than the subjects of
intransitives, e.g. Dyirbal (in which the normal word is OSV):
Transitive sentence
Balan dyugumbil bangul yarangu balgan
woman absolutive man ergative hit
‘the man hit the woman’
Intransitive sentence
Bayi yara baninyu
man absolutive came-here
‘the man came here’
3 . Switch reference, the marking in one clause to indicste whether its subject is coreferential
with the subject of another clause, e.g. Lenakel where SS = the morpheme which requires that
the subjects refer to the same person:
Different subjects
r -im-vin r-im-apul
he-PAST-go he-PAST- sleep
‘hei went and hej slept’
Same subjects
r-im-vin m-(im)-apul
he-PAST-go SS-PAST-Sleep
‘hei went and hei slept’
tracking when they occur and the kinds of error patterns in which they
appear, but more recently, some attempts have been made to explain how
they are acquired, both in terms of their syntactic properties and the
performance factors that affect them. This introduction begins with a
general discussion of the issues at stake in studying the acquisition of
grammatical morphemes.
A major issue is defining the class of grammatical morphemes. As noted,
several people have discussed them as a psychologically real category of
grammar, distinct from the lexical morphemes. No consensus has been
reached, however, on the extent to which this is true. We could, for
example, deny a unified class and propose a taxonomy of classes of grammar
that fall along some continuum in regard to one or more properties. The
problems involved can be exemplified by looking at the status of preposi-
tions. These are typically included in the class of grammatical morphemes,
yet Chomsky (1981) has included them within a feature system that includes
nouns, verbs, and adverbs. For example, they can function as heads of
constructions (prepositional phrases) and, like verbs, assign grammatical
case. The Government and Binding Theory of Chomsky, in fact, has yet to
come up with an adequate theory of the range of possible grammatical
categories. This is true, to some extent, of other theories as well; as a result,
the study of the acquisition of grammatical morphemes is often theory-
independent and based upon traditional category labels.
Much of the work in this area has focussed on English, but it is important
to recognize at the onset the range and variability of grammatical mor-
phemes cross-linguistically. Any account of their acquisition will need to
deal with the cross-linguistic cases also. Table 9.1 presents a very brief
sampling of the diversity which is involved. After reviewing the research in
English, we will discuss some of the cross-linguistic data studied.
Another important step in studying the acquisition of grammatical mor-
phemes is to establish the ways in which they differ from one another, and
how these differences interact. Brown (1973) was one of the first to do this.
One set of differences concern the effects of perceptual salience. This
includes the position of a morpheme in a sentence, and whether it is free or
affixed, syllabic or nonsyllabic, and stressed or unstressed. Another pos-
sible influence is frequency of occurrence in the parental language. As we
will see later, Newport, Gleitman & Gleitman (1977) present data which
suggest that children who acquire auxiliary inversion early have parents
who direct more questions to them with inverted auxiliaries than do other
parents. A third feature of morphemes is that they differ in semantic
complexity. We have already seen in the previous chapter, for example, that
certain spatial terms can be argued to be semantically simpler than others.
Lastly, there are a number of differences between morphemes which can be
placed under the category of grammatical complexity. These include
(i) redundancy, i.e. the extent to which a morpheme is predictable, e.g.
person on verbs is redundant when pronoun subjects are used; (ii) allomor-
phy, i.e. the number of allomorphs that occur, e.g. ‘-ing’ has one allomorph
in English but plural ‘-s’has three; (iii) paradigm regularity, e.g. regular
verb forms vs. irregular ones such as ‘to be’.
Still another issue in studying morphological acquisition is a methodolo-
gical one. That is, how do we study the child’s knowledge of these
grammatical forms? We have an immediate problem when we try to do this
exclusively from spontaneous language samples. As has been documented
in several places, some of the grammatical morphemes are not very
frequent even in adult language, e.g. the passive morphology on verbs.
Brown (1973) has gotten around this problem to some extent by his measure
of the percentage of obligatory occurrence which adjusts for frequency.
Even this measure, however, requires some minimum frequency which is
not always evident. Another problem with spontaneous samples is that it is
not always possible to determine from the context what morpheme a child
should have used. For example, if the child says ‘I eat apple?’, we can
438 SIMPLE SENTENCES: ACQUISITION OF GRAMMATICAL MORPHEMES
Table 9.2 A summary of the main grammatical morphemes found in
English
Inflectional morphemes
Nominal inflections:
plural {-s} ‘two boys’
possessive { -s) ‘the girl’s socks’
verbal inflections:
progressive {-ins) ‘the cat is running’
present tense {-s) ‘the dog sits on the rug’
past tense {-ed) ‘the teacher walked to school’
passive {-en} ‘the cake was eaten by the class’
perfect {-en} ‘I have eaten the apple’
adjectival inflections:
comparative {-er) ‘the cat is bigger than the dog’
superlative {-est} ‘this is the smallest class in the school’
Derivational morphemes’
adverbial {-ly} ‘Joe ran quickly to the store’
adjectival {-ful) ‘May is hopeful of getting the job’
agentive {-er} ‘she was the bearer of good news’
etc.
Free morphemes
verb negation {not} ‘do not go in there’
modal auxiliaries e.g. {will, can, may} ‘we will arrive by midnight’
infinitive {to} ‘Mary decided to leave’
personal pronouns e.g. {I, you, it) ‘they saw her near him’
possessive pronouns e.g. {his, her} ‘our books’
reflexive pronouns e.g. {himself] ‘they did it themselves’
demonstrative pronouns e.g. {this, that} ‘I bought this book in town’
locative pronouns e.g. {here, there} ‘put it down right there’
relative pronouns e.g. {which, that} ‘we saw the book which was on the table’
determiners {the, a} ‘the apple’
quantifiers e.g. {few, some, many) ‘we only saw a few people at the dance’
complementizers {that, for..to, -ing} ‘we knew that John was leaving’
gerund {-ing} ‘the running of the bulls was yesterday’
participle {-ing} ‘the falling leaves were lovely’
I Only a sampling of these are presented.
assume that an auxiliary is missing, but we can’t determine which one it is,
e.g. ‘can’, ‘may’, ‘should’, e t c . . .
This area requires as much methodological caution as any other. As I
will show, several innovative techniques have been developed to assist in
this study, including comprehension tasks, imitation tasks, and the use of
metalinguistic judgements. Since children are older at the time of morpho-
logical acquisition, it is possible to use more diverse tasks. Even so, we will
see that much of what is known is still based on spontaneous language
samples.
9.2 Morphological acquisition in English: a descriptive overview

I begin the presentation of findings on morphological acquisition with a
historical review of some early research on English. Much of this work was
done by Roger Brown and his colleagues, and they established many of the
methodological procedures used. Here, studies are considered in their
historical order to give a feel for how these findings became known and built
upon one another. Findings will be presented for children from the onset of
grammatical morphemes around age 2 years, up to children around 6 years
of age.
First it is necessary to give an overview of the range of the major
grammatical morphemes that occur in English and thus serve as potential
targets for research (Table 9.2). This is an extensive list and no study has yet
attempted to research them all in depth.
9.2.1 Three early studies

Three early studies between 1958 and 1963 provided a first insight into the
patternsof acquisition of selected English morphemes. It was a time when the
longitudinal study of Adam, Eve, and Sarah was underway and detailed
analysis of those data was not yet completed. The studies to be reported were
experimental ones which each developed a different method for the study of
morphological acquisition. In retrospect, the successfuldevelopment of these
methodologies will be as significant as the early findings which they provided.
Berko (1958) This study probably more than any other marked the onset
of the modern era of child language studies. It is not only one of the most
famous studies ever done, it is also the most replicated. Even with so many
replications, its general findings remain sound.
Berko selected a variety of grammatical morphemes for experimental
study. These are listed in (9.1).
(9.1) a. plural {-s} with three allophones:

[SI after [p,t,k,f,Ol
[az] after [s,z,S,i,E,j]
[z] after other voiced sounds
b. possessive {-s} with three allophones (same as plural)
c. present {-s} with three allophones (same as plural)
d. past {-ed} with three allophones:
[ad] after [t,d]
[t] after [p,k,E,f,O,s,S]
[d] after all voiced sounds except [d]
440 SIMPLE SENTENCES: ACQUISITION OF G R A M M A T I C A L MORPHEMES
e. progressive {-ins}
f. agentive { -er}
g. comparatives { -er} , {-est}
h. compounds
Here the discussion will be limited to study of the inflections selected, that is
items (9a)-(9e). The five inflectional morphemes chosen give an idea of the
order of acquisition of inflections. Also, the fact that the first three listed all
have the same allomorphs provides an important test case of the interaction
of phonological and semantic factors. For example, if inflectional order of
acquisition were determined solely by phonological factors, then these
three should be acquired at the same time. Lastly, the study provides the
first information on the order of acquisition of specific allomorphs.
Berko developed a technique to study these forms which has come to be
known as the ‘wug’ procedure - as famous as the results of the study. First,
Berko created a series of nonsense words which ended in sounds that would
elicit one of the possible allomorphs. For example, one of her nonsense
words was ‘wug’ which would require a [z] in the plural by the rules of
English. She then created nonsense drawings which were to represent
meanings for the nonsense words. A ‘wug’, for example, was a bird-like
animal. In the case of ‘wug’, there were two pictures, one with one wug and
one with two. The child was then tested on the following item:
(9.2) This is a wug. Now there’s another one.
There are two of them, there are two -
The child was expected to fill in the last word. Table 9.3 gives the words
tested for the five inflectional morphemes. As can be seen, there were
different words for different morphemes, and some real English words used
-these were added to see if the child would treat them differently from the
nonsense words.
There were two groups of children tested, a preschool group of children
who were 4 and 5 years old, and a first grade group of 5i to 7-year-olds.
There were 19 children in the preschool group and 61 in the first grade
group.
Table 9.3 gives the results in the form of the percentage of correct
responses for each of the test items. Since the number of test items is small,
Berko’s results must be treated with some caution but one striking result is
that the children showed acquisition of the single consonant allomorphs [SI,
[z], [t], and [d], but not the ones with schwa, Le. [az] and [ad]. That is, even
though children may appear to have acquired an inflection, accurate use of
all of its allomorphs can be quite late. A related finding concerns the
differences in the scores with nonsense words requiring [ad] and [az], and
Table 9.3 Percentage of correct productions of children studied in Berko
(1958) o n words requiring one of five English inflections
Nouns Verbs
Plural Past
glasses [az] 91 binged [dl 78
wugs [.I 91 glinged [dl 77
luns [zl 86 ricked [tl 73
tors [zl 85 melted [ad] 73
heafs’ [SI?[zl 82 spowed [dl 52
cras [zl 79 motted [ad] 33
tasses [az] 36 bodded [ad] 31
gutches [az] 36 rang 16
kazhes [az] 31
3rd singular
Possessive loodges [az] 56
nizzes [az] 28 nazzes [az] 48
bik’s [SI 87
wug’s [zl 84 Progressive
niz’s [az] 49 zibbing 90
~~ ~
Both ‘heaves’ and ‘heafs’ were considered correct responses.
the English words with the same requirements, that is, ‘glasses’ and
‘melted’. The scores on the latter were markedly better than the former.
This indicates that children may initially use words like these correctly, but
may not yet have productive use of the allomorph being used. It is an
important warning that spontaneous use is not sufficient evidence to
conclude rule-based behavior. At the same time, the strength of the rule
which was acquired for [d] was evident in the results for ‘rang’ where most of
the children regularized it.
Another finding concerned the order of acquisition of the morphemes.
The progressive with its single allomorph was the easiest one for the
children. When we compare the scores on the three morphemes which have
[az], we see that they are not the same. The children had more difficulties
using [az] with the plural than for the possessive and 3rd singular. That is,
the results suggest that there may be other factors involved than just
articulatory ones.
Brown & Fraser (1963) As with Berko (1958), this study was as important
methodologically as it was for its results. The study itself was very much a
preliminary one, with a small set of subjects and stimuli. It presented six
children between 2 and 3 years of age with a set of 13 sentences to imitate.
The goal was to see if the children would process the sentences through their
own grammatical system, or simply rote-imitate. The test sentences were as
shown in (9.3):
Table 9.4 Imitations of 13 sentences by the four youngest of six children
tested in Brown & Fraser (1963), as reported in Brown (1973:76)
Sentence
no. Eve Adam Helen Ian
(253 months) (284 months) (30 months) (314 months)
1. I show book (I show) book Correct I show you the book

2. (My) tall I (very) tall I very tall I’m very tall
3. Big box Big box In big box It goes in the box
4. Read book Read book - Read (a) book
5. Drawing dog I draw dog I drawing dog Dog
6. Read book I will read book I read the book I read the book
7. See cow I want see cow Correct cow
8. Do again I will that again I do that I again
9. I do apple I do a apple - I do not want apple
10. To read book? I read books? I read books? I read book?
11.’t car? Is it car? Car? That a car?
12. Where go? Go? Does it go? Where do it go?
13. Go? - - Correct
(9.3) 1. I showed you the book 8. I will not do that again

2. I am very tall 9. I do not want an apple
3. It goes in a big box 10. Do I like to read books?
4. Read the book 11. Is it a car?
5. I am drawing a dog 12. Where does it go?
6. I will read the book 13. Where shall I go?
7. I can see a cow
Brown & Fraser found that the children did imitate the sentences in a
systematic way. Specifically, they found that the children tended to retain
the lexical or contentive words but delete the grammatical ones. Second,
they found that the imitations tended to preserve the word order of the
original model. Lastly, the imitations struck Brown & Fraser as being very
similar to the utterances found in the spontaneous utterances of the same
children. They gave the label telegraphicspeech to this form of reduction. In
general, they found that the imitations provided information about the
children’s grammatical systems. Table 9.4 gives the imitations that were
elicited for the four youngest children.
We can see some patterns in the data by looking at the percentage of
correct imitations of specific grammatical morphemes. These are given in
Table 9.5 and can be compared with the Berko results in Table 9.3. Again,
the plural and the progressive appear to be relatively easier to imitate than
the other inflections. The past tense morpheme in particular was poorly
imitated, also somewhat in support of the Berko finding that it was more
difficult than the plural, progressive, and 3rd singular. The data suggest that
Table 9.5 Imitation of grammatical morphemes, as tested in Brown &
Fraser (1963)
Morpheme Word(s) % correctly imitated
Progressive drawing 67 (4/6)

Plural books 67 (416)
Articles the 53 (9/17)
a 37 (11/30)
Present tense goes 50 (3/6)
Modal auxiliaries will 50 (6/12)
can 50 (3/6)
shall 50 (316)
Infinitive to 50 (316)
Inverted copula is 50 (3/6)
Uninverted auxiliary am 33 (2/6)
Uninverted copula am 17 (1/6)
Past tense -ed 17 (1/6)
the modals may be easier than the copular and auxiliary forms of ‘to be’.
Lastly, there is an interesting pattern within the latter suggesting that the
uninverted forms may be more difficult than the inverted ones. These
preliminary results were to some degree substantiated in the in-depth
research which followed.
Fraser, Bellugi & Brown (1963) This frequently cited study set as its goal
the relation between imitation, comprehension, and production. Ten
grammatical contrasts were selected, and children were tested on their
ability to imitate, comprehend, and produce each one. Importantly, the
contrasts were selected in a way that the children would need to know
specific grammatical morphemes in order to process the sentences correctly.
For example, one contrast was between the sentences ‘The sheep is
jumping’ and ‘The sheep are jumping’. The authors state (p. 468): ‘No one
seems to have compared this kind of evidence for the passive control of
grammar in normal children with evidence of active or productive control in
the same children, and we have done an experiment to fill the gap.’ There is
no discussion of the rationale behind the selection of the ten contrasts used,
and also there are no predictions made about how the children would
perform with them. The ten contrasts tested with one of the sets of 40 test
sentences are presented in Table 9.6.
The subjects were twelve children between 37 and 43 months of age. Each
child underwent three tasks of imitation (I), comprehension (C), and
production (P). The test sentences consisted of three sets of 40 sentences
each. The three sets of test sentences were varied across the three tasks, and
Table 9.6 Ten grammatical contrasts with the 40 test sentences of Set A
taken from Fraser, Bellugi & Brown (1963: Table 2)
~ ~ ~~
1. Mass noun / Count noun Some mog / A dap

Some pim / A ked
2. Singular / plural, marked by The boy draws / The boys draw
inflection The kitten plays / The kittens play
3. Singular / plural, marked by The deer is running /The deer are running
‘is’ and ‘are’ The sheep is eating / T h e sheep are eating
4. Present progressive / past The paint is spilling /The paint spilled
tense The boy is jumping / The boy jumped
5 . Present progressive / future The girl is drinking / The girl will drink
tense The baby is climbing / T h e baby will climb
6. Affirmative / negative The girl is cooking / The girl is not cooking
The boy is sitting / The boy is not sitting
7. Singular / plural, of 3rd person His wagon /Their wagon
possessive pronouns Her dog /Their dog
8. Subject / object, in the active The train bumps the car / The car bumps the train
voice The mommy kisses the daddy I The daddy kisses the
mommy
9. Subject / object, in the passive The car is bumped by the train / The train is bumped
voice by the car
The daddy is kissed by the mommy / The mommy is
kissed by the daddy
10. Indirect object / direct object The girl shows the cat the dog / The girl shows the
dog the cat
The boy brings the fish the bird / The boy brings the
bird the fish
the order of presentation of the three tasks was varied across subjects. A
pair of pictures was designed for each pair of sentences tested. For example,
for ‘The sheep is jumping’ the picture showed a picture of a sheep jumping,
while for ‘The sheep are jumping’ the picture showed two sheep jumping.
On the imitation task, the child was asked to repeat the test sentences after
the experimenter. The pictures were only used in the comprehension and
production tasks. In comprehension, the child was shown the two pictures
and then heard one of the test pair of sentences. H e then had to point to the
picture named. Next he was given the second test sentence and again asked
to select the appropriate picture. The pictures were used again for pro-
duction, although they were a different set than those for either imitation or
comprehension. ‘The S is twice told the names of the two pictures but not
which name goes with which picture . . . After repeating the names of the
pictures E points to one picture at a time and asks S to name it’ (Fraser,
Bellugi & Brown 1963: 469-70). Correctness was determined only in
relation to the relevant morphemes in the test sentences. A correct response
on any pair required that both sentences be imitated, comprehended, or
produced correctly.
Table 9.7 Percentage of correct responses on the grammatical contrasts
across three tasks tested in Fraser, Bellugi & Brown (1963)
Contrast Imitation Comprehension Production Total
6. Affirmative / negative 75 71 50 65
7. Singular / plural, of 3rd person possessive
pronouns 96 63 33 64
8. Subject / object, in the active voice 79 67 46 64
4. Present progressive / future tense 83 67 24 58
3. Singular / plural, marked by ‘is’ and ‘are’ 83 50 29 54
5 . Present progressive / past tense 71 54 25 50
1 . Mass noun / count noun 50 54 4 36
2. Singular / plural, marked by inflection 58 29 4 31
9. Subject / object, in the passive voice 50 29 8 29
10. Indirect object / direct object 46 21 13 26
Table 9.7 presents the results in terms of percentages of correct responses

on each contrast across the three tasks. One general finding was that
imitation was easier than comprehension, and comprehension was easier
than production. This at first glance appears to contradict the result we just
saw in Brown & Fraser (1963) where the children appeared to process their
imitations through their grammatical systems. The authors offer the follow-
ing discussion of this contradiction:
It is very possible, however, that this . . . outcome would reverse with
still younger children. The longest sentences of the ICP Test were only
eight morphemes long which means they were easily within the
sentence-programming span of 3-year-old children. However, the
much shorter span of children at about 2;6 should compel such
younger children in the I task, to ‘reduce’ the model sentences. . .
(Fraser, Bellugi & Brown 1963: 484)
That is, it will be necessary to take the child’s short-term memory ability
into consideration when attempting to draw conclusions from imitation
tasks. As we will discuss, this was in fact done in Kuczaj & Maratsos (1975)
for the acquisition of auxiliaries.
The second finding, that comprehension was better than production,
supported the traditional view and the one expressed at various points in
this volume. It should be pointed out, however, that this result was aided by
the decision to score ‘no response’ in production as errors. Fernald (1972)
has reanalyzed these data with a different scoring system and found the
scores on the two tasks to be much closer.
There is relatively little qualitative analysis on the data discussed by
Fraser, Bellugi & Brown. One observation was that the children would
occasionally alter the sentences in production to fit their understanding.
One example was with the contrast with indirect objects (no. 10)where five
children inserted a preposition, saying for example, ‘The woman gives the
teddy to the bunny’ instead of ‘The woman gives the bunny the teddy’. This
response would be scored as an error in production, but this response
certainly indicates that it is understood correctly. Such examples are
important for at least two reasons: they show that production data can be
used to draw conclusions about comprehension, and they show how
production may lag behind comprehension, indicating that some additional
processing is involved.
A similar kind of observation was made in regard to the passive
sentences, which received a very low score in comprehension, indicating
that they were not yet understood. Five of the children responded to
sentences such as ‘The girl is pushed by the boy’ as if it meant ‘The girl
pushed the boy’, that is, they treated them as if they ignored the passive
morphology and understood it as an agent+action+object structure. This
observation has been supported by several subsequent studies which have
provided more details on how these sentences develop.
The authors do briefly address the possible reasons why certain mor-
phemes are acquired earlier than others:
. . . there does not seem to be any single dimension of the contrasts
that will account for the order. . . Probably the order obtained for the
problems is a complex resultant of many factors that increase diffi-
culty, including the perceptual obviousness of the contrast, the
amount of redundancy in the contrast, the length of the total sen-
tences, and the frequency with which the construction has been heard.
It should be possible to investigate the importance of these variables
with problems designed to that purpose. (P. 482)
The four possible variables they raise are similar to some of the ones we
discussed in the previous chapter for determining the order of acquisition
of words. Of some importance is the fact that the optimism of the last
sentence has not completely been borne out, although we will look later in
the chapter at some attempts to develop a theory that addresses these
issues.
9.2.2 The analysis of Adam, Eve, and Sarah

As the 1960s progressed, the language samples collected from Adam, Eve,
and Sarah were carefully analyzed for insights into English grammatical
development. Two aspects of these samples were analyzed in depth. One
was the development of selected grammatical morphemes, especially the
inflectional ones. This was first reported in Cazden (1968), and then com-
pleted in Brown (1973). The other area was the development of th.,e aux-
iliary, particularly in terms of its appearance in question words. We begin
by looking at the first of these two, and return to auxiliaries in a later
section.
Cazden (1968) This was the first detailed analysis of the development of
inflections by Adam, Eve, and Sarah, and it laid the groundwork for
Brown’s later report (1973). Cazden studied the five noun and verb inflec-
tions which occur in English, Le. the first five morphemes listed in (9.1).
To do this, Cazden made several important methodological decisions
which are needed in any in-depth study of grammatical development. We
will look at each in turn.
The first decision concerns how to measure whether the child has a
particular morpheme or not. We have already discussed the problem
with counting absolute frequencies as a measure of acquisition. The
problem is that we need both to see when a morpheme is used, and the
number of times that it is used appropriately. Cazden reasoned (p. 227)
that there are four possible ways in which an inflection could be scored
for a child:
S,: ‘supplied correctly’, Le. the child used an inflection in a context
for which it was appropriate, e.g. ‘two dogs’ in reference to two
dogs;
S,: ‘supplied in inappropriate contexts’, i.e. the child used an inflec-
tion in a place where it was inappropriate, e.g. ‘one dogs’ in refer-
ence to a single dog;
0: ‘required but omitted’, Le. the child did not use an inflection in a
context where it was required, based on the rules of English
grammar, e.g. ‘two dog’ to refer to two dogs;
OG: ‘overgeneralizations’, Le. use of the inflection in a context
where an alternative form was correct, e.g. ‘two foots’ used to
mean ‘two feet’.
Distinctions like this enabled Cazden to see more clearly the ways in which
inflections were used in the data.
A second decision concerns how one can determine when a context for
appropriate use has occurred in the data. In determining transliterations
for chi!dren’s utterances’, claims about missing morphemes have to be
made. Cazden set specific criteria for determining contexts when a mor-
pheme should be required; a feel for these can be seen by looking at her
criteria for determining when a plural morpheme would be required in a
sentence (p. 227):
448 S I M P L E SENTENCES: A C Q U I S I T I O N OF G R A M M A T I C A L M O R P H E M E S
Number - required after numbers except 1: ‘two minute’;
Linguistic - required on count nouns after such modifiers as ‘more’ or
‘some’: ‘more page’;
Interaction - required for discourse agreement: ‘shoe’ in response to
parent’s question, ‘what are those?’;
Normally plural - ‘stair’ (‘upstairs’);
Routines - either public, like nursery rhymes and the names of
cartoon characters - ‘Mr Ear’ (‘Mr Ears’) - or private, like Eve’s
telegraphic version of her mother’s oft-repeated explanation of
father’s work - ‘Make penny Ema’ Hall’ (‘He’s making pennies in
Emerson Hall’).
As Brown (1973) later pointed out, some grammatical morphemes are

easier for doing this than other. As Cazden’s stated criteria show, the
non-linguistic context of the utterance in very important in making such
decisions.
A brief comment is in order on the three measures Cazden used to score
the four kinds of uses (where S,, S,, 0, and OG refer to the number of
contexts scored as an instance of each of these categories):
Proportion of correct use = S, / S, + 0

Proportion of inappropriate use = S, / S, + S,
Proportion of overgeneralization = OG / S, + S,
One point about these measures is that all three need to be examined
(Cazden does this) before claims about adult-like acquisition can be made.
The child may have a high proportion of correct use, but still will not have
an adult system if the other two proportions are high. A second point
concerns the measure for the proportion of overgeneralization. This does
not appear to me to be a very useful measure since the S, contexts are not
ones where overgeneralizations can occur in the first place. I would
recommend an alternative, such as the following:
Suggested proportion of overgeneralization

+
OG / OG Oi + S,i
where Oi = irregular plurals where an overgeneralization could, but
did not, occur, e.g. ‘two foot’
where SCi= irregular plurals where the irregular form was correctly
used, e.g. ‘two feet’
In this formula, we could also distinguish between two types of OG, those
which occur on an unmarked irregular form, e.g. ‘two foots’, versus those
which occur on the marked form, e.g. ‘two feets’. Since Cazden does not
provide the needed information, I cannot replace her measure of over-
Table 9.8 Four periods in the acquisition of the English plural morpheme
for Adam, Eve, and Sarah, adapted from Cazden (1968: 228, table 1)
Period Proportion of correct use No. of inappropriate uses No. of overgeneralizations
Adam Eve Sarah Adam Eve Sarah Adam Eve Sarah
A - 0.00 - - 0 - - 0 -
B 0.36 0.15 0.13 0 0 0 0 0 0
C 0.68 0.86 0.86 25 10 4 2 8 0
D 0.94 0.98 0.98 54 1 28 40 I 23
generalization by mine. Instead, I simply replace her measure with a

simpler one: the number of overgeneralizations.
A similar criticism can be raised with the second measure, the proportion
of inappropriate use. In this case, it is difficult to see what kind of
proportion could be used, in that the number of possible inappropriate
contexts can be quite large. The only alternative I see is to use simply the
number of such instances, and I therefore use this measure in the review of
Cazden’s results.
Using the measures discussed, Cazden found four periods in the acqui-
sition of each inflection (while she does not say so, she appears to be using
the term ‘period’ in a descriptive sense to avoid the implications of the word
‘stage’ as we have discussed in Chapter 3). These periods are defined in my
words as follows:
Period definition
A The time before the appearance of any inflections
B The time of the first inflections, with no errors
C The time of widespread use of inflections, with concomitant
appearance of errors and over-generalizations
D The proportion of correct use reaches 90 percent, an arbitrary
measure of acquisition
Cazden states the following (p. 228):
These three divisions do not have the same status. The break between
C and D is clearly an arbitrary imposition. The meaning of the break
between A and B, signaling the onset of inflection, is unclear. But the
break between B and C represents a significant developmental
phenomenon, because systematic errors and overgeneralizations
provide convincing evidence that the child has a productive rule.
Table 9.8 presents the data for the four stages in the acquisition of the plural
inflection, where we give the original proportions of correct use along with
the two revised measures just discussed.
Cazden is certainly accurate in pointing to the significance of errors and
overgeneralizations as an indication that a productive rule is established.
We have already seen in Berko’s results that correct use of a morpheme, as
in ‘glasses’, cannot itself be taken as indicative of a rule. The fact that the
child may acquire a large number of inflected words before such evidence
occurs is quite striking in Cazden’s data. The numbers of S, tokens (i.e.
plural inflection supplied correctly) for Adam, Eve, and Sarah at Period C
are 142,136, and 124 respectively. While these are only token frequencies,
they certainly suggest a large number of words being marked correctly.
I would like to suggest that even the evidence at Period Cis rather lean in
regard to rule productivity. First, observe in Table 9.8 that very few
overgeneralizations appear at Cazden’s Period C. The average is only three
across the three subjects. We interpret these results as indicating that
overgeneralizations do not actually appear with any noticeable frequency
until Period D. The S , contexts for Adam, Eve, and Sarah at Period D are
927,217, and 722, adding even more evidence to the interpretation that the
child can acquire a very large vocabulary before a productive rule is
established. (Recall that in Chapter 4 we discussed this as the lexical
principle of acquisition .)
A possible problem for this interpretation is the fact that several
inappropriate uses of inflections do appear in Period C. Recall, however,
that inappropriate uses consist of using a correctly inflected word, e.g.
‘dogs’, where it should not be, e.g. ‘one dogs’ for one dog. (This, at least, is
my understanding of these errors. Cazden, unfortunately, says very little
about them.) Such errors are not the result of rule overgeneralizations, but
rather lexical errors. If the child has acquired two semantically related
words like ‘dog’ and ‘dogs’ as separate lexical items, it is not surprising to
see some confusion in the process. Some cases of an inflectional omission at
this period (e.g. ‘two dog’) in fact could be the result of the same confusion
or retrieval error. I take such data at Period C, therefore, as evidence for
such lexical problems rather than as indicative of a rule-based error.
Space does not allow examination of a number of other results from
Cazden’s study, for example, on order of acquisition of inflections, where
plural and the present progressive emerge as the first two inflections
acquired. Since Brown’s own analysis of these data will be considered later,
this topic is left until then. (Cazden also provides a useful discussion of the
kinds of errors found in the data.) Two other findings of general importance
may, however, be considered here.
One finding concerns the course of acquisition of the irregular forms of
verbs and nouns. Cazden found that use of these usually preceded the first
instances of overgeneralizations. For example, Eve used ‘came’correctly 11
times between 20 and 22 months before using ‘corned’ three times between
25 and 27 months. Also, Cazden states (p. 238): ‘Temporary coexistence of
the correct irregular form and the overgeneralization is common in our
records.’ In other words, the use of the overgeneralized form does not
appear to replace the previously acquired irregular form. This observation
leads us to hypothesize three periods in the acquisition of irregular forms:
1. Acquire parts of semantically related words as separate lexical entries,
i.e. the lexical principle, e.g. ‘come’, ‘came’, and ‘go’, ‘goes’.
2. Acquire a general rule, and determine for all previously learned pairs
which match the pattern and which do not; adjust lexical entries for
regular pairs, and retain the irregular ones, marking them as excep-
tions, e.g. ‘come’ [-regular past], ‘came’ [+past].
Overgeneralizations may occur during the period of lexical restructur-
ing, e.g. ‘corned’.
3. Attainment of adult lexical entries, which may take several months to
complete.
Another finding concerns the individual patterns of acquisition for
Adam, Eve, and Sarah. Eve acquired language rather quickly, and showed
relatively few errors and overgeneralizations. Adam was slower, but still
showed similar development to Eve’s when his proportions of correct use
were related to MLU. Sarah was the slowest learner of the three, at least as
measured by relating age and MLU, but she supplied grammatical mor-
phemes at a higher rate when related to MLU than either Eve or Adam.
Cazden used this difference to speculate that there are two ways to acquire
language: macrodevelopment and microdevelopment. She says (pp. 233-4):
‘Macrodevelopment refers to the elaborateness of the semantic plan for
speaking, while microdevelopment refers to the successful execution of
whatever plan has been executed.’ Macrodevelopment is seen as more
informative, in that the grammatical morphemes are often redundant and
can be supplied by the listener. This is an interesting speculation on two
possible routes or individual ways to acquire language, but Cazden says
little more about them, except a brief mention that they could be the result
of any of the three kinds of variation I have discussed at the end of
Chapter 4. The speculation has resurfaced in other places (e.g. Ramer
1976) where early accuracy in the use of semantic relations is associated with
more rapid learners. A cautious review of the distinction can be found in
Ingram (1981d).
Brown (1973) The most cited study of English morphological develop-

ment remains that of Brown (1973). We have already seen his discussion of
the semantic relations which appear in Stage I. The second half of his book
is concerned with the acquisition of 14 grammatical morphemes in par-
452 SIMPLE SENTENCES: ACQUISITION OF GRAMMATlCAL MORPHEMES
Table 9.9 The 14grammatical morphemes studied by Brown and their
mean order of acquisition across Adam, Eve and Sarah, taken from
Brown (1973: 274, Table 38)
Mean
rank Grammatical morphemes and general order of acquisition
2.33 1 . present progressive { -ing}

2.50 2.-3. the prepositions ‘in’ and ‘on’
3.00 4. plural {-s}
6.00 5. past irregular, e.g. ‘went’, ‘came’, etc.
6.33 6. possessive {-s}
6.50 7. *uncontractiblecopula {be}, e.g. ‘there it is’ (where the contracted form is
incorrect, e.g. ‘there it’s’)
7.00 8. articles ‘the’, ‘a’
9.00 9. past regular {-ed}
9.66 10. 3rd person regular {-s}
10.83 11. 3rd person irregular, e.g. ‘does’, ‘has’
11.66 12. uncontractible auxiliary {be},* e.g. ‘Are you going?’
12.66 13. contractible copula {be},* e.g. ‘I’m sick’
14.00 14. contractible auxiliary {be},* e.g. ‘She’s leaving’
* The contractible allomorphs discussed by Brown are {-s, -z, -m, -r}; the uncontractible ones
are {is, am, are}.
ticular (given here in Table 9.9), which begin development in Stage 11. The
methodology is very similar to that used by Cazden. Linguistic and
non-linguisticcontextual information was used to determine contexts where
the morphemes could be obligatory. Adam’s, Eve’s, and Sarah’s use of the
14 grammatical morphemes was then measured by the percentage of
obligatory morphemes supplied, which is more or less the same measure
used by Cazden which we called the proportion of correct use. A grammati-
cal morpheme was claimed to be acquired when this measure reached 90
percent. These 14 grammatical morphemes include the five studied by
Cazden .
Brown first presents his results in terms of the order of acquisition of the
14 morphemes for each of the three children. These famous results are
reproduced here in Table 9.10. Note about this table that nearly one-third
of the morphemes did not reach criterion at the last samples taken at
Stage V. The second point concerns the relative similarity across the
children of the first morphemes acquired.
Brown pursued the latter point by looking at the rank order of acquisition
of the 14 grammatical morphemes across the three children. For example,
the plural morpheme was acquired first by Sarah, and fourth by both Adam
and Eve for a mean rank of 3.00 (Le. 1+4+4 = 9 divided by 3 = 3.00).
Statistical analysis revealed a significant correlation between these rank
orders and MLU.
Table 9.10 The order of acquisition of Mgrammatical morphemes for
Adam, Eve, and Sarah, taken from Brown (1973: 271, Figure 14)
Stage Ages and morphemes at each stage
Adam Eve Sarah

I 2;3 1;6 2;3
(none) (none) plural {-s}
I1 2;6 1;9 2;lO
progressive {-ins} progressive {-ins) ‘in’, ‘on’
‘in’ ‘on’ progressive {-ing)
‘on’ ‘in’ past irregular
plural {-s} possessive {-s}
111 2;11 1;11 3;l
uncontractible copula {be} plural {-s} uncontractible copula {be}
past irregular possessive { -s} articles {the, a}
past regular {-ed}
IV 3;2 2;2 3;8
articles {the, a) (none) 3rd person regular { -s}
3rd person irregular
possessive { -s}
V’ 3;6 2;3 4;0
3rd person regular {-s} uncontractible copula {be} past regular {-ed}
past regular {-ed} past irregular uncontractible auxiliary {be}
uncontractible auxiliary {be}articles {the, a} contractible copula {be}
contractible copula {be} 3rd person regular {-s} 3rd person irregular
contractible auxiliary {be} 3rd person irregular contractible auxiliary {be}
uncontractible auxiliary {be}
contractible copula {be}
contractible auxiliary {be}
* These are listed by their order of decreasing percentages of obligatory occurrence.
Brown’s results represented a significant step in our descriptive data on

the acquisition of English morphemes. It expanded Cazden’s earlier study
by providing information on a wider range of morphemes, and also a detailed
discussion about why this particular order was found. The results were also
replicated in large part by a cross-sectional study conducted around the
same time by de Villiers & de Villiers (1973b). The latter study examined
the proportion of obligatory occurrence of the same 14 morphemes in
spontaneous language samples from 21 children between 1;4 and 3;4. Like
Brown, de Villiers & de Villiers found that the children showed an invariant
order of acquisition. There were , however, some specific differences
regarding the order of acquisition of specific morphemes. For example, if
we use a difference of at least three ranks between the two studies on the
individual morphemes, de Villiers & de Villiers found that the 3rd person
regular and contractible copula appeared earlier, and the present progress-
ive, possessive, and uncontractible copula appeared later. Perhaps more
interesting is the fact that relatively few of the 14 morphemes showed
acquisition before Stage V. Example (9.4) gives those morphemes that
were acquired, where acquisition is defined as the stage at which the
morpheme reaches a mean percentage of obligatory occurrence of 90
percent for that and all subsequent stages.
(9.4) Stage Morpheme acquired
I ‘on’, plural
I1 ‘-ing’, ‘in’, past irregular
111
IV possessive, contractible copula
This reveals that one half of these morphemes are not acquired until after an
MLU of 4.00, and shows the gradual acquisition that they undergo.
Both Brown and de Villiers & de Villiers explore three reasons for these
general findings: (i) frequency of the morphemes in parental speech,
(ii) syntactic complexity, and (iii) semantic complexity. Neither study
found any effects of (i), but both found that (ii) and (iii) had some predictive
power. This was a difficult issue to pursue, since both possible causes
require some theory of complexity for any effects to be determined. The
following extract captures the somewhat vague conclusions that were drawn
(de Villiers & de Villiers 1973b: 277):
In conclusion, both semantic and grammatical complexity to some
extent predict the order of acquisition of the morphemes, but with the
analyses of these variables currently available, it is impossible to
separate out the relative contributions of each type of complexity
since they make the same predictions. In fact, the order of acquisition
may best be predicted by some combination of grammatical and
semantic complexity, frequency, and perceptibility in speech. It is
possible that no one factor can be considered of primary importance in
determining the acquisition of the morphemes.
9.3 The acquisition of Aux in English questions

Section 9.2 provided a general overview on the acquisition in English of a
range of grammatical morphemes. It demonstrated the kinds of methods
which have been applied to the study of this topic, and gave some general
findings about their order of acquisition. Some general findings were also
presented on the gradual process of their acquisition, which indicated that
overgeneralizations appear only after a fair amount of development has
taken place. There was little discussion, however, of how the rules which
underlie these morphemes are acquired.
In this section we turn to a more detailed discussion of one class of
grammatical morphemes which has generated a wide range of studies, the
English auxiliary (or ‘Aux’ as it is commonly referred to in the syntactic
literature). Aux has been of general interest to linguists since the first
rules for its appearance were proposed in Chomsky (1957). Typically,
Aux is considered to contain all of the grammatical morphemes indicated
in (9.5):
(9.5) a. Tense (person, number, past)
b. Modals (e.g. ‘cad, ‘may’, ‘will’, etc.)
c. +
Progressive: ‘be’ ‘ing’
d. Perfect: ‘have’ + ‘-en’
Issues in grammatical theory regarding Aux include whether such a class
exists, and if so, where it occurs in the phrase structure of English (as well as
other languages).
Much of the acquisition literature assumes that Aux exists, and has
focussed on how children acquire the rules which invert Aux and subject
NPs to form questions in English. In yesho questions, Aux moves to the
front of the sentence as shown in (9.6b). I will refer to this movement as
Subject-Aux Inversion.
(9.6) a. John is going to the store
b. Is John going to the store?
c. John is doing what?
d. What is John doing?
e. John likes cookies
f. Does John like cookies?
This rule also occurs in what are called wh-questions as in ( 9 . 6 ~and d). It is
assumed that the wh- word, in this example ‘what’, is inserted into a place
where an NP can occur. The difference between the (c) and (d) sentences is
that the latter has had a second rule apply which moves the wh-word to the
front of the sentence. I will refer to this second rule as Wh-Movement. A
last aspect of English questions concerns ‘do’ and the appearance of tense.
In simple active declarative sentences, the tense occurs on the verb as in
(9.6e). Transformational grammar proposes a rule of Affix-Hopping which
in this case moves the tense from inside the Aux onto the verb. This happens
when there is no other element within Aux for the tense to attach to. Lastly,
in questions where only tense exists inside the inverted Aux, a ‘do’ is
inserted to carry tense. An example of this is given as (9.6f). We will refer to
this rule as Do-Support. In acquiring English questions, therefore, the child
needs to acquire the category Aux with its various constituents in (9.9, and
four rules -, Subject-Aux Inversion, Wh-Movement, Affix-Hopping, and
Do-Support .
456 SIMPLE SENTENCES: ACQUISITION OF GRAMMATICA L MORPHEMES
Table 9.11 Sentences used in Kuczaj & Maratsos. (1975) and A be’s pattern
of imitation
Sentences Imitation of Aux

with examples Correct Deleted Tense’ Reordered
PERIOD I
Declaratives
Grammatical
‘The nice monkey can kiss his little sister’ 38/48
Ungrammatical
Tensed verb: ‘The boy can pushed the elephant’ 0/31 13/31 14/31
Word order: ‘The boy push will the elephant’ O/26 1 1126 14/26
Wh questions
Grammatical
‘What will the boy eat?’ 117 517 1/7
Ungrammatical
Misplacement: ‘What the boy will eat?’ 3/11 8111
Yes/no questions
Grammatical
‘Can the boy push the elephant?’ 2/16 10116
Ungrammatical
Tensed verb: ‘Can the boy pushed the elephant?’0/21 19/21 2/21
PERIOD I1
Wh-questions
Grammatical
‘What will the boy eat?’ Oil0
Ungrammatical
Misplacement ‘What the boy will eat?’ 8/10
Yes/no questions
Grammatical
‘Can the boy push the elephant?’ 10110
Ungrammatical
Tensed verb: ‘Can the boy pushed the elephant?’l/lO 9/10
I This category records when Abe changes the tense of the verb, e.g. imitating ‘The boy can
pushed the elephant’ as ‘The boy can push the elephant.’
9.3.1 Patterns of acquisition

In this section, I review a few of the studies which have examined the
acquisition of sentences like those in (9.6). First, I consider the more
general patterns that have been found, and later I turn to some proposed
explanations.
Kuczaj & Maratsos (1975) This preliminary study on one child led to
several consequent studies (e.g. Kuczaj 1978; Maratsos & Kuczaj 1978;
Kuczaj & Brannick 1979). The study is useful in providing some more
detailed facts on acquisition by a single child, as well as demonstrating a
valuable technique. The subject was the first author’s son Abe. Example
(9.7) gives a summary of Abe’s use of Aux in his speech during the time he
was studied:
(9.7) period I 2,5(21)-2,7(15) with MLU 3.01 to 3.25:
The 3,058 sentences collected revealed only 4 with either
‘will’ or ‘can’.
period I1 2,9(16)-(19):
Aux was used in declaratives but not in questions.
As can be seen, Abe was not using Aux in his spontaneous speech during
period 1. The authors anticipated, however, that Abe might have already
begun the acquisition of Aux in his receptive language. To get at this, they
gave him several sentences to imitate which contained instances of ‘can’ and
‘will’. This is the technique we saw earlier in this chapter in the study by
Brown & Fraser (1963). It was hoped that Abe would show the ‘filter
effect’, i.e. that he would change the sentences to reflect his grammatical
knowledge. These sentences consisted of declarative sentences as well as
wh-questions and yesho questions. Each type of sentence also had sen-
tences which were either grammatical or ungrammatical based on the rules
of adult English speakers. The types of sentences used, with examples of
each, are given in Table 9.11.
Abe’s imitation showed a pattern indicating an underlying grammatical
system. An examination of his imitations of declaratives reveals an
awareness of the distribution of the English Aux. He correctly imitated it in
grammatical sentences, and changed ungrammatical ones to be grammati-
cal, either by deletion (of the Aux or tense on the verb) or by reordering. He
did not, however, show an awareness of how Aux occurs in questions. In
both grammatical and ungrammatical questions, he would delete the Aux.
These results lead to a first conclusion that Abe acquired Aux in declara-
tives before questions.
Table 9.11 also presents Abe’s pattern of response during period 11. At
this time, Abe repeats grammatical yesho questions correctly, and also
changes ungrammatical ones to be correct. This indicates that he had
acquired the rule of Subject-Aux Inversion, at least for questions. The
latter restriction is made because of his responses on wh-questions. On
these, he would correctly imitate the ungrammatical ones, and change the
grammatical ones into ungrammatical ones, i.e. he would change a question
such as ‘What will the boy eat?’ into ‘What the boy will eat?’ It should be
noted here that the study only examined the modals ‘will’ and ‘can’.
Bellugi (196711971) The results of Kuczaj & Maratsos (1975) supported

with a different methodology findings reported earlier by Bellugi. Bellugi
had been the member of Brown’s research team who had primary responsi-
bility for the collection of data from Adam. In her 1967 dissertation, she
analyzed the data from all three children and reported stages in the
acquisition of questions. Brief reports can be found in Klima & Bellugi-
Klima (1966) and Bellugi (1971). Ironically, the most complete published
summary of these stages is in Cazden (1970). A summary of these stages,
based on the latter report, is given in (9.8), which also gives Adam’s age
and stage at each period considered.
(9.8) period A: (I, 28 months) Children do not have Aux.
Yeslno questions marked by intonation, e.g. ‘sit chair?’,
‘ball go?’
Wh-questions restricted to ‘what’ and ‘where’in limited
forms, e.g. ‘what’s that?’, ‘where cookie go?’
period B: (11,111, 35 months)
period C: (IV, 38 months) Aux appears throughont system. Inver-
sion occurs in yeslno questions but not wh-questions.
Yes/no question, e.g. ‘does lions walk?’, ‘oh, did I
caught it?’, ‘will you help me?’
Wh-questions, e.g. ‘what he can ride in?’, ‘why kitty
can’t stand up?’, ‘what he can ride in?’
periods
D to F: (V and later) Gradual emergence of tag questions
These stages were widely accepted by investigators, as indicated by their
citation in texts at the time (e.g. Dale 1976). The most important stage
proposed is period C when children show inversion in yes/no questions but
not in wh-questions. This is also the pattern we just saw in Abe’s data.
Besides the uninverted wh-questions, there are two other kinds of errors
that need to be pointed out. One is the occurrence of what we will call
double tense marking, which refers to those sentences where tense is
marked twice, once in the Aux and once on the verb. An example from
(9.8) is ‘oh, did I caught it?’ Examples such as these pose an instant problem
for explanation, for they require one to account for tense being copied onto
the verb rather than being moved, which is the normal proposal for adult
English. A related error is what we will call double Aux marking. Hurford
(1975) cites examples of these from his daughter: ‘whose is that is?’ and
‘what did you did?’. These differ from double tense marking in that the
entire Aux is copied, not just the tense constituent. Examples of these other
two kinds of errors have been noted and discussed in works by Hurford
(1975), Prideaux (1976), Kuczaj (1976), Fay (1978), Maratsos & Kuczaj
Table 9.12 The proportion of auxiliaries and proportion of inversion
found in Ingram & Tyack (1979) for 21 subjects
~~~ ~ ~
Period No. of Prop. of auxiliaries Prop. of inversion

subjects
Yestno Wh Yeslno Wh
~~
A 7 0.22 0.44 0.55 0.77

B 3 0.68 0.70 0.81 0.91
C 5 0.89 0.88 0.91 0.96
D 4 1.00 0.93 0.98 0.95
E,F 2 0.99 1.00 0.97 0.98
(1978), Erreich, Valian & Winzemer (1980), and Klein (1982), among
others.
Subsequent research into these patterns of acquisition has qualified the facts
as they have been described. One qualification concerns the accuracy of the
periods as described by Bellugi. Most of the published accounts of these
periods are anecdotal in the sense that they are presented with examples
rather than precise measures. One exception is Bellugi (1971) where some
figures are given for Adam’s use of inverted and uninverted auxiliaries. At
3;6, Adam produced only eight inverted wh-questions out of 30, for a
proportion of 0.27. This changed to 0.87 at 3;11 (33 out of 38). Brown
(1968) also gives some absolute numbers for the occurrence of these for
Adam, Eve, and Sarah. He states that Adam produced 145 such questions
between Stages I11 through V, while Eve produced seven and Sarah 18.
Such data suggest that the pattern was clearly more characteristic of the
speech of Adam than of the other two children.
Ingram & Tyack (1979) attempted to test these periods further by using a
larger sample and more precise measures of inversion. The subjects were 21
children between 2;O and 3;11 who were placed into periods A through F
based on the MLUs given for these periods. Approximately 225 questions
were collected from each child. These were then examined for (i) the use of
Aux, and (ii) the occurrence of inversion. For the first point, the measure
used was the proportion of auxiliaries, that is, how often did the questions
have an Aux in them. The second point was measured by the proportion of
inversion. This measure was applied to just those questions which had an
Aux, and determined the proportion of those which showed the Aux
appearing before the subject NP. Table 9.12 gives the general results.
We see the gradual emergence of auxiliaries as characteristic of grammati-
460 S I M P L E SENTENCES: ACQUISITION OF G R A M M A T I C A L M O R P H E M E S
cal morphemes, as indicated at the onset of this chapter. Of some import-
ance is the observation that they occur even in periods A and B, at least as
these can be assigned on the basis of MLU. Of more importance is the lack
of a clearly marked period C, that is, at all times the children’s use of
inversions for wh-questions was as great as that for yesho questions.
The question arises, then, of how to account for the discrepancy between
these figures and the patterns reported for Abe and Adam. Ingram & Tyack
discuss this point, and mention that two of their subjects did indeed show
uninverted wh-questions of the sort reported in the literature. Their
appearance, however, did not occur to such a degree as to differ sub-
stantially from the yes/no questions. Also, they tended to occur more for
uncontracted auxiliaries than for contracted ones. These results, therefore,
suggested that the use of uninverted Aux is not as extensive as implied in
Bellugi’s stages and that they may be restricted to certain contexts.
Such restrictions have been confirmed by subsequent research by Labov
& Labov (1978), Kuczaj & Brannick (1979), and Erreich (1984). These first
two studies are methodologically complementary in that Labov & Labov
is an intensive single-subject study, while Kuczaj & Brannick is a cross-
sectional study of several children. Their general results show that the
inversion of Aux in wh-questions occurs gradually for specific wh-words.
For example, children will show inversion for ‘what’ questions while having
uninverted ‘why’ questions. Kuczaj & Brannick (1979: 43) state:
The results suggest that children learn to apply this rule [Subject-Aux
Inversion, DI] to questions beginning with one or two wh words, then
to questions beginning with another wh word, and so on, rather than
to all relevant question types simultaneously, indicating that the
acquisition of this syntactic rule is initially relatively specific.
Another qualification concerns the generality of double tense marking
and double Aux marking. As mentioned, examples of these are mostly
presented anecdotally with little measurement of their pervasiveness. An
exception is the report by Maratsos & Kuczaj (1978) where figures are given
on the occurrence of these errors in the speech of four subjects. For two
subjects, there were virtually no examples of such errors. For two others,
these were restricted to forms with ‘do’. These data are presented in
Table 9.13. They also cite data from 15 children discussed in Kuczaj (1976)
that show the same pattern. Further data in support of the restriction of
such errors in questions to ‘do’ appear in Davis (1987).
Both Hurford (1975) and Maratsos & Kuczaj (1978) report two further
aspects to these errors. One is that double tense marking is normally
restricted to irregular past forms (e.g. ‘broke’) instead of regular pasts (e.g.
‘missed’) or overgeneralized past forms (e.g. ‘breaked’). Also, Maratsos &
Table 9.13 Proportion of double tense marking for Jive possible Aux’s
reported for two subjects in Maratsos & Kuczaj (1978)
Subject Sample size Proportion of double tense marking for specific

Aux’s No. of possible contexts (in parentheses)
~~~ ~
‘does’ ‘did’ ‘is’ ‘are’ modals

D.C. 24 hours 0.40 (20) 0.18 (40) 0.00 (56) 0.00 (15) 0.00 (?)
K.R. 50 hours 0.06 (17) 0.16 (32) 0.00 (38) 0.00 (19) 0.00 (?)
Kuczaj claim that double Aux marking is extremely rare, and that it
constitutes a different kind of error from double tense marking. They state
(p. 344): ‘Occasional errors such as ‘is this is the powder’ may not require
any explanation at all, given their extremely low frequency. . .’
In summary, several kinds of errors in the acquisition of Aux in English
have been reported. At first, these were proposed as general patterns in
acquisition, but have since been seen to occur (i) in some children more
often than in others, (ii) in more restricted contexts than first proposed, and
(iii) with differing degrees of frequency.
9.3.2 The explanation of Aux acquisition

The facts discussed above have generated several attempts at explanation.
These efforts can be placed into two general categories of competence
factors vs. performance factors. In the performance factors approach, the
child is seen as having the adult grammar, with performance factors
restricting the child from fully expressing his grammatical knowledge.
Those accounts which refer to competence factors assume that the errors
above reflect the child’s grammatical system, that is, the child has formed an
incorrect grammar of English and needs to reformulate it. This perspective
implies some form of Competence Assumption. We will briefly review some
of the proposals which have been made under each of these positions.
Performance factors The first attempts to account for Bellugi’s periods are
found in Bellugi (1967), Brown, Cazden & Bellugi (1969), and Brown
(1968). Brown divides the child’s acquisition of wh-questions into two
general stages. Before MLU Stage 111, the child’s questions are seen as
being either memorized routines or constructions which are nontransfor-
mational in nature. This is the time when Aux is rare, and when wh-
questions are restricted to forms such as ‘what that?’. By Stage 111,
however, evidence emerges for more productive rules. The general expla-
nation of these studies is that evidence exists that the child has both
Subject-Aux Inversion and Wh-Movement. The evidence for the former

rule is the occurrence of inverted yesho questions, and the evidence for the
latter rule is the occurrence of uninverted wh-questions. The general
proposal is that the child has both rules, but has a performance restriction
which allows him to apply only one transformation at a time. In yesho
questions only one rule is needed, Subject-Aux Inversion, so that the
resultant questions look adult-like. In wh-questions, however, two rules are
needed, Subject-Aux Inversion and Wh-Movement. The performance 4
restriction results in only Wh-Movement occurring, so that uninverted

wh-questions occur.
A potential problem for this explanation is why the rules are ordered for
wh-questions as they are. That is, why couldn’t the child apply Subject-Aux
Inversion in wh-questions instead of Wh-Movement and say questions of
the form ‘can John do what?’ Brown (1968) proposes that uninverted
wh-questions occur because the child hears the parent use what he calls
occasional questions. Brown found examples of the kinds of dialogue shown
in the samples from Adam, Eve, and Sarah.
(9.9) a. Child: I want milk
Adult: You want what?
b. Adult: What do you want?
Child: (no response)
c. Child: I want it
Occasional questions like this, particularly in dialogues such as (9.9b), show
the child that the wh-word can appear in two places, and that a movement
rule must be needed in the grammar of English.
While Brown’s observation about occasional questions provides the kind
of positive evidence needed to form Wh-Movement, it still does not explain
why the child chooses to select its operation in wh-questions instead of
Subject-Aux Inversion. It also implies a noticeable discontinuity in the
development of these rules. Before Stage 111,the child has neither rule, and
after, he has both. Another problem is that the explanation suffers
somewhat in light of the newer information about the restrictions on
uninverted wh-questions discussed above. Brown’s account, it seems,
would need to propose that the variation observed is in some way tied to the
variation between parents in their use of occasional questions in general and
with specific wh-questions. Brown, in fact, suggests such a possibility when
he observes that Adam’s mother used more occasional questions than either
Eve’s or Sarah’s mother. Lastly, Brown’s explanation still leaves the
occurrence of double Aux marking as a mystery.
Competence factors Subsequent attempts to account for these patterns of
errors have often assumed that they reflect the child’s grammatical system at
the time (e.g. Hurford 1975; Prideaux 1976; Fay 1978; Erreich, Valian &
Winzemer 1980; Klein 1982). The differences are usually in the kind of adult
grammar that is assumed to be acquired, rather than disagreement over the
data. Also, the focus is often on all of the error types discussed earlier.
Here we will briefly look at the proposals in Erreich, Valian & Winzemer
(1980) which are characteristic of this orientation. First, they assume that
the child’s language at any point in time should be a possible human
language, That is, the child’s grammar will always need to be constrained by
principles of Universal Grammar. To account for the child’s double
constructions, they propose that all movement rules (such as Wh-
Movement and Subject-Aux Inversion) involve two operations: copying
and deletion. Further, Universal Grammar instructs the child first to treat
all cases of movement as copying, and only later as subsequent deletion.
The shift from the incorrect rule to the correct one presumably occurs once
the child notices that deletion always follows copying for Subject-Aux
Inversion and Wh-Movement.
This explanation is very simple in its formulation and accounts super-
ficially for the child’s patterns of error. It has, however, certain difficulties
which are typical of such approaches. Like the pro-drop parameter dis-
cussed in Chapters 4 and 7, there are problems with learnability. First,
children are making these errors between Stages I11 and V, as found by
Bellugi. It seems hard to understand how the child after so much exposure
to English would not have noticed that English deletes the elements which
have been copied. Second, the account requires indirect negative evidence.
That is, the child will need to note that sentences with double AUX’Sand
wh-words never occur in the input language. There are also data problems
with these proposals. One, which is pointed out by Erreich, Valian &
Winzemer, is that it predicts that children should produce sentences like
‘what did you see what?’, yet such sentences are not reported for children
learning English. Further, it predicts that the errors should be made by all
children in a wide range of sentences. We have seen above, however, that
the data are much more restricted than this.
Maratsos and Kuczaj have published several studies on the acquisition of
English Aux, e.g. Maratsos & Kuczaj (1978), Kuczaj & Brannick (1979),
and Kuczaj & Maratsos (1983). These studies present an explanation of the
errors discussed above that combines both competence and performance
factors, and attempts to account for the restrictions on their occurrence.
The general orientation of their analysis is that children acquire these
rules very gradually, restricting them at first to very specific contexts, and
then generalizing them. Regarding uninverted wh-questions, Kuczaj &
Brannick (1979) conclude that the child first learns the rule of Subject-Aux
Inversion for one or two wh-words, and only gradually generalizes it to
other wh-words. They provide experimental results to support this con-
clusion.
If such is the case for wh-questions, then the question arises as to why the
rule appears to apply quite early and generally in yeslno questions. This
point is addressed in Kuczaj & Maratsos (1983), where it is argued that
Subject-Aux Inversion is not as general in yesho questions as has been
assumed. The authors claim that children initially acquire the use of 4
auxiliaries in yes/no questions separately from those used in declaratives,

and only later recognize that they are the same category Aux. Two kinds of
evidence are presented to support this conclusion, both argued with data
from Kuczaj’s two sons Abe and Ben. One is that children will show
knowledge of auxiliaries in declaratives but not in yeslno questions. Abe’s
acquisition of four modals is exemplified in (9.10). It indicates that he would
use ‘would’ and ‘could’ in declaratives for several months but would
substitute ‘will’ and ‘can’ respectively in questions.
(9.10) Abe’s acquisition of ‘will’, ‘would’, ‘can’, ‘could’

Age Yeslno questions Declaratives
2;9-3;4 ‘will’ ‘will’, ‘would’
‘will’ for ‘would’ (34 times)
2;ll-3;4 ‘can’ ‘can’, ‘could’
‘can’ for ‘could’ (201 times)
3;5 ‘could’ (15 times)
‘would’ (not given)
A second argument they give is that the children did not invert auxiliary
elements in questions for declarative sentences, as in ‘you better go’, or
similar sentences with ‘gonna’ and ‘wanna’. These should be predicted, they
claim, if the child has a general Aux category. Their account of the
acquisition of Subject-Aux Inversion, then, is that the child has not yet
formulated a general rule and is instead acquiring it in a very restricted way.
The explanation of double tense marking and double Aux marking is
taken up in Maratsos & Kuczaj (1978). They begin by acknowledging that
double tense marking is usually restricted to questions with forms of ‘do’
and an irregular verb. They then go on to assume that these are the result of
performance factors. They state (p. 343): ‘. . . w e assume that these are
errors of production, rather than representing underlying rule structure,
and that the child has by and large analysed the relevant sequences
accurately’. The performance factor which is at stake is one of lexical
retrieval, that is, in selecting the appropriate verb, the child picks the wrong
one. The reason this occurs with just ‘do’ is presumably because the only
function of ‘do’ is to indicate tense.
Maratsos & Kuczaj correctly point out the retrieval error solution will not
work for double Aux marking since this involves inserting an Aux where
one does not belong. These errors, however, are seen to be much less
frequent than those of double tense marking. The implication is that errors
of such minimal frequency are probably due to some other performance
factor, and that their rarity is sufficient to give them secondary importance.
Maratsos & Kuczaj account for uninverted wh-questions through com-
petence factors (specific learning) and double marking constructions
through performance factors. More than other proposals, they also take
into consideration the restrictions on the data on such errors. Despite their
concern with the data, their account is still not a completely satisfactory
explanation. A criticism is the same as the one directed to the discussion of
distributional learning in Chapter 7: such accounts do not work from a
theory of grammar and do not deal with how the child goes from specific
learning to general rules. In this particular case, there are also at least two
gaps in the explanation given. One concerns the claim that the child has
made retrieval errors; and it seems that some theory of when such errors do
and don’t occur is needed. For example, they do not explain why the child
wouldn’t show similar retrieval problems with other Aux elements besides
‘do’. Another problem is the claim that ‘wanna’ and ‘gonna’ are auxiliaries.
One could also claim that they are main verbs, and that the child has their
appropriate structure acquired (cf. Ingram 1985a).
9.4 Other aspects of English grammatical acquisition

This chapter has examined the acquisition of both bound and free gram-
matical morphemes in English. It began with an overview of their develop-
mental patterns, and then turned to a more detailed look at the acquisition
of Aux. In this last section on English, we will briefly look at a selection of
other aspects of English grammar which have been investigated in recent
years. In reviewing these topics, one goal is descriptive, that is, to provide
the reader with information about the grammatical study of English, but as
will be seen, these studies raise once more the general theoretical issue
about when children acquire adult-like knowledge of English.
9.4.1 Passives
In Chapter 7, section 7.2.4 we discussed comprehension of the passive, as
tested by de Villiers & de Villiers (1973a). I interpreted their results as
indicating that children at age 27 months have some awareness that passives
are different from actives, but that they don’t yet understand the semantics
of the structure. Here I will examine the development of this construction
over a longer period of time, first by looking at some production data and
then by turning to comprehension - a reversal of the approach in earlier
chapters. Unlike the literature for earlier periods of acquisition, that on
later grammatical development is more balanced in terms of the number of
studies in each area. Also, the constructions under discussion are infrequent
in spoken language and require careful comprehension studies.
Horgan (1978) This is the most comprehensive study of the use of passives
in spontaneous speech. Some basic information on the subjects who were
observed and the methods of data collection are given below:
group I: 54 children between 2;O and 4;2. Subjects described 44
pictures in each of two sessions, three months apart.
group 11: 180 children, 30 in each age group of 5, 6, 7, 9, 11, 13
years. Subjects told stories about pictures.
group 111: 262 university students. Subjects were asked to describe
pictures with passives within a structured task.
We see that Horgan not only studied children in the period which we are
reviewing, but also older children and adults. The adult data were obtained
so that normative comparisons could be established.
The first major point that Horgan makes concerns the difference between
full passives and truncated passives. These distinctions as well as others
which Horgan makes are given in (9.11):
(9.11) Types of passives
a. Truncated passives:
inanimate logical objects, e.g. ‘The lamp was broken’
animate logical objects, e.g. ‘The boy was chased’
statives with ‘get’, e.g. ‘It got broken’
b. Full passives:
reversible, e.g. ‘The boy was chased by the girl’
non-reversible, with agent as logical subject, e.g. ‘The lamp was
broken by the girl’
non-reversible, with instrument as logical subject, e.g. ‘The lamp
was broken by (or ‘with’) the ball’
Horgan excludes truncated passives from the study on the grounds that
they are a distinct structure from the full passive. That is, she proposes that
the two kinds of passives are not transformationally related, but instead are
generated independently of each other. She presents two arguments from
Table 9.14 The frequency of occurrence of three kinds of full passives for
three age groups of children, taken from Horgan (1978: Table 3)
Age Non-reversible Non-reversible

groups Reversible agentive instrumental Total
2-4 15 0 17 32
5-7 10 0 9 19
9-13 21 15 14 50
Total* 46 15 40 101
* Two passives were not classified because the subject was inaudible.
her data for this position. First, there is a noticeable difference in the
frequency of the two passives. Truncated passives are much more frequent
than full passives in the spontaneous speech of children. This observation is
not consistent with the predictions of the transformational analysis, since
the full passives are supposed to be more basic. The second argument is that
they have different characteristics. Most truncated passives in the child data
occurred with inanimate logical objects, while most full passives occurred
with animate logical objects. Further, truncated passives mostly had
familiar verbs like ‘break’ and were often used in a stative sense with ‘get’,
e.g. ‘it got broke’. Full passives, on the other hand, were used with a wider
variety of action verbs.
Despite the large samples, there were relatively few full passives pro-
duced by the children. Group I produced 32 full passives and group I1
produced 81. If we consider that the group I children were asked to produce
88 sentences across the two sessions, then the 32 passives occurred in a
sample of approximately 4,700 sentences. Table 9.14 gives a breakdown of
the use of the different kinds of full passives by the children. Not only are
passives infrequent, but no non-reversible agentive ones occurred until
age 9.
Horgan makes the following additional observations about these data.
First, ‘until age 11, no child produced both reversible and non-reversible
passives’ (p. 72). That is, individual children appeared to select one way or
the other to express passives. This created problems for children when they
needed to express something that required the other form. For example,
Horgan gives the following examples of sentences by children with only the
non-reversible construction, who wanted to express a reversible meaning:
(9.12) a. ‘the man was killed by the hand’
b. ‘the boy was kicked by the foot’
The second observation concerned the use of passives by the 2- and 4-year
olds. Most of their reversible passives had the wrong word order, so that
Table 9.15 Mean percentages of four kinds of responses for Stage II-V
groups on the understanding of active and passive sentences, taken from de
Villiers & de Villiers (1973a: Table I )
Stage I1 Stage I11 Early IV Late IV, V

(n = 7) (n = 7) (n = 4) (n = 5)
Active Passive Active Passive Active Passive Active Passive
Correct 70.0 36.3 81.0 42.3 85.5 13.3 83.3 39.4

Reversed 11.0 43.3 14.2 41.0 14.2 86.7 13.4 50.0
Child asagent 16.2 17.6 2.4 14.3 0.0 0.0 3.3 3.3
Refusals 2.8 2.8 2.4 2.4 0.0 0.0 0.0 7.3
‘the cat was chased by the girl’ was used to describe a picture of a cat chasing
a girl. This is consistent with our impression from de Villiers & de Villiers
that children have some awareness of the form of the passive, but not its
meaning.
Given production data of the kind cited above from Horgan, it is not
surprising that most studies on passives use more specific elicitation
procedures (e.g. Smith 1970; Whitehurst, Ironsmith & Goldfein 1974) or
comprehension tasks. The comprehension research will be reviewed by
returning to de Villiers & de Villiers (1973a) and then by looking at the
studies reported in Maratsos, Kuczaj, Fox & Chalkley (1979).
In section 7.2.3, we reviewed the study by de Villiers & de Villiers
(1973a) on the acquisition of English word order in the context of chil-
dren’s ability to process active and passive sentences. To review, they
tested 33 children across several ages on reversible active and passive
sentences. They placed the children’s responses into four categories:
correct, reversed, child as agent, and refusals. On the basis of their results
presented in Table 7.8, I concluded that the data indicated that late
Stage I children show awareness of the formal characteristics of passives
since they responded to them differently than they do actives (see 7.2.4
for details).
Table 9.15 presents their data for the children who were beyond Stage I.
As with the late Stage I children, the older children show an understanding
of the actives but not the passives. Do the data, however, provide evidence
that the children are responding differently to passives than actives, as it did
for late Stage I? Recall that a crucial indicator would be if the children’s
percentages of reversals for passives were the same as those for correct
responses to actives. We see in Table 9.15 that this is not true for Stages 11,
111, and late IV, V. We also get marked differences in the ‘child as agent’
responses for Stage 111. I interpret these differences as indicating that
several children must be aware of the fact that passives are formally
different from actives, and that other children can even interpret them
correctly.
There is in Table 9.15 a striking pattern of response for the early Stage IV
children: it appears that they are treating passives exactly like actives. This
is a pattern which has been observed by others, e.g. Bever (1970). There are
two ways in which we could interpret this. One is that this effect is due to the
small sample size and to a sampling error in which children had been
selected who had not yet acquired even an awareness of passives. The
alternative is that children around this point in acquisition adopt a strategy
for dealing with passives of treating them as corresponding actives. Note
that the existence of such a performance factor does not necessarily mean
that the children are unaware of the formal characteristics of passives, but
rather that they have reached a cognitive level where they have better
problem-solving abilities than earlier. Versions of this strategy have been
around since its first proposal in Bever (1970).
In viewing results such as those in Table 9.15, there is an alternative to
the claim that children either do or do not have the passive. This is that
the children may have some knowledge of the grammar of passives, but
incomplete knowledge. Such a possibility is suggested by the data from
Horgan’s study, in that truncated passives were acquired differently from
full ones, and even full ones came in differently from each other. This
alternative would also account for the data in Table 9.15 if we assume that
children are getting some forms of passive correct but not others. The
observation that children may acquire passives differently for different
verbs was first made in Sinclair & Ferreiro (1970). More recently it has
been argued for in Maratsos et al. (1979) and in Sudhalter & Braine
(1985).
Maratsos, Kuczaj, Fox & Chalkley (1979) This article is an extensive

discussion on the acquisition of several constructions of English. It assesses
the extent to which acquisition of syntax can be accounted for solely by
reference to the acquisition of formal structures independent of meaning.
Reiterating a point made in Maratsos (1978), they propose that the
semantic bootstrapping hypothesis combined with a purely formal rule for
passives such as that in (9.13) predicts across-the-board acquisition of
passives.
(9.13) NPI ‘be’ VERB ‘en’ ‘by’ NP2
By ‘across-the-board’ we mean application to sentences which show a range
of semantic roles for the syntactic subjects and objects, as exemplified in
(9.14):
Table 9.16 Proportions of correct responses on actional and nonactional
verbs, reported in Maratsos et al. (1979)
Age group Actionals Nonactionals
Actives Passives Actives Passives
4-year-olds 0.87 0.68 0.91 0.47

5-year-olds 0.94 0.66 0.91 0.35
(9.14) a. Tables are liked by John

INANIMATE OBJECT EXPERIENCER
b. John is pleased by tables
EXPERIENCER OBJECT
c. Mary was kissed by John
AGENT ANIMATE OBJECT
d. The door was opened by the key
INANIMATE OBJECT INSTRUMENT
Their purpose is to argue that passives are not acquired across-the-board,

but instead show gradual acquisition from one class of verbs to another.
Maratsos et al. report on two studies on the acquisition of passives; while
they differed in the methods used, the same general results were found. I
will present a brief review of the results from the first study.
Maratsos et al. tested 38 4- and 5-year olds on their ability to understand
passives for two main kinds of verbs: actional verbs such as ‘find’, ‘hold’,
‘wash’, ‘shake’, and nonactional verbs such as ‘remember’, ‘forget’, ‘know’,
‘like’, ‘miss’, ‘see’, ‘hear’, ‘watch’. The children were presented with finger
puppets and then told either an active or passive sentence about them - for
example, a typical test sentence might be ‘Donald was liked by Goofy’.
After each sentence, the children were then asked ‘Who did it?’ While this
question is appropriate for actional verbs, it is not for nonactional ones.
They discuss this possible effect on their results and state (p. 11) ‘. . .we
argue from the detailed pattern of the obtained results that the method did
not prove unduly discriminatory against adequate response to nonactional
passives. Another small study conducted afterwards supports further the
methodological adequacy of the main study.’
The proportions of correct responses for the two kinds of verbs are given
in Table 9.16. These figures are for the 31 of the subjects who were able to
complete the task.
There was a significant difference in the performance on passives for each
kind, indicating that passives were easier with actional verbs than non-
actional ones. This finding was confirmed in the second study, and also by a
later one by Sudhalter & Braine (1985). Further, there was no overlap
between the individual verbs within the classes, that is, all of the actional
verbs were easier to understand in the passive than any of the nonactional
ones. The authors take these results as confirmation of the claim that the
passive is acquired gradually over several years, and that it requires a lot of
experience with individual verbs. They also see it as counterevidence for a
strictly formal rule of Passive, at least up to the school years. They are not
explicit as to whether they think a restructuring to a more formal rule occurs
later, or that the adult rule requires semantic information. Their concern is
primarily to provide data showing that there are semantic restrictions on the
child’s understanding of passives at least up to age 5.
Explanation. Thus far we have presented data which indicate that
children may show some awareness of passives at an early age, but that it
takes several years before the complete range of its application is acquired.
We also mentioned in passing the implications that these data have for a
theory of acquisition. Horgan implies that truncated and full passives may
have different grammatical sources, at least while they are being acquired.
Maratsos et al. emphasize the semantic restrictions of the rule during the
same period. Neither, however, faces directly the implications these results
have for a theory of grammatical acquisition. That is, do they indicate that a
different kind of Passive rule needs to be proposed for the adult language,
or that children undergo a discontinuity in development? In addition to
these points, there is the issue of how the child identifies the passive
construction. That is, is it the result of semantic bootstrapping or some form
of distributional learning? This section will explore these issues more
directly.
The question of how the child determines the meaning and form of
passives is not one which has received much detailed discussion in the
literature. One exception to this is the discussion found in Pinker
(1984). We have already seen in Chapter 7 Pinker’s proposals about the
steps which the child follows in acquiring the basic phrase structure of
English, and the resulting grammar he gives for Stage I. That grammar
and the same steps will be used for the acquisition of passives, along
with some additional machinery. We will not go into the details about
how this occurs, or its concomitant problems, but will just outline the
approach.
Before passives are acquired, Pinker assumes that the child already has
acquired the rules which generate oblique prepositional phrases as in
(9.15a). When generating the phrase structure for a passive such as (9.15b),
the child has principles available which allow him to assume that the passive
has the same structural characteristics as the oblique phrase (see Pinker
1984: 76-83):
(9.15) a. The cat was sitting by the fence
b. The cat was bitten by the dog
This will eventually lead to a new phrase structure rule which generates
passives as well as oblique prepositional phrases. The difference between
the two rules will be in the grammatical roles of the NPs. A last step will be
the child’s acquiring a second lexical entry for the passive form of the verb.
For example, the child’s lexicon will have ‘bite’ for active sentences and
‘bitten’ for passive ones.
These steps indicate a continuous development in which the child
establishes the Passive rule as explained by the lexical functional approach,
without any extensive discontinuity. In Pinker’s discussion, in fact, there is
no reference to a need for orphan categories in acquiring the passive, as
there was with the oblique prepositional phrase. That is, once the structure
for the oblique PPs is acquired, the acquisition of passives should follow
smoothly, given appropriate input.
Given Pinker’s explanation, there is then the question of whether this
account fits the acquisition facts as described above. There are at least two
general findings that need to be explained: (i) the treatment of passive
sentences as if they were actives, and (ii) the incomplete acquisition of
passives, where the rule is applied to active verbs but not to nonactive ones.
A further, less well-documented observation is Horgan’s, that children may
express active meanings with passive morphology.
Pinker is aware of the phenomenon (i) and discusses it in his book. He
does not, however, directly offer an account for it. We will therefore
speculate about how he would deal with it. One possibility is to say that
the child has no contextual information in this situation. That is, the child
has no information which tells him that the subject is not an agent. If this
occurs, however, it seems that the procedures could lead the child t o a
grammar where actives can have either active or passive morphology.
This would account for Horgan’s observations that actives may be
expressed with passive formalism. Such a grammar could be restructured
if it contained some orphans and the child subsequently heard passives
with contextual information. Another possibility would be to consider
such examples as the result of a performance factor which has intervened
in the acquisition procedures. This performance mechanism would be one
that operates when the child can’t process all of a sentence, and yet needs
to determine some meaning. In this case, it would tell the child to ignore
the passive morphology and preposition, and impose some canonical
expected meaning. The consequences of such a mechanism is to say that
some sentences are processed for learning and others are not. This
appears to be Pinker’s proposal; in a general discussion of input, he states
(p. 28):
I also assume that not all sentences heard by the child, nor all parts of a
sentence, will be used as input to his or her acquisition mechanisms.
Presumably children encode most reliably the parts of sentences
whose words they understand individually, and the whole sentences
most of whose words they understand.
The second set of findings concerning the restricted application of the

Passive rule is directly addressed by Pinker. In his first explanation (Pinker
1982) he suggests that children initially only hear actives with semantic
restrictions, and thus formulate a restricted version with the appropriate
constraints. Such a rule is stated informally here:
(9.16) Initial Passive rule: Passive is restricted to sentences where the
surface grammatical subject has the semantic role of patient and the
surface grammatical object has the semantic role of agent.
He also proposes that this initial passive rule is part of Universal Grammar
and the one the child first applies to passive sentences. The child could then
extend the rule to more diverse semantic roles on the basis of positive
evidence.
Berwick & Weinberg (1984: 215-20) have criticized the above proposal
on theoretical grounds. They have pointed out that a universal rule such as
that in (9.16) predicts that there should be some languages which have such
a Passive rule. No such evidence, however, currently exists. They argue that
if this cross-linguistic observation is true, then Pinker’s proposal has a
problem in accounting for the acquisition data. It would force him to a
maturational model in which the child begins with such a rule, and then
reaches a point where it is programmed to drop out. Such an account,
however, would contradict Pinker’s own assumptions about restrictions on
changes in the child’s grammar.
Pinker (1984: 328-33) takes an alternative approach to accounting for the
restricted application of passive. He first attacks the findings themselves,
citing unpublished research of his own which found children passivizing
both actional and nonactional verbs. He then states (pp. 329-30): ‘neither
the child nor the adult entertains the hypothesis that passivization is
constrained by some blanket restriction like “agentive subjects only” or
“action verbs only”’, effectively rejecting his earlier proposal.
The problem left is to resolve the discrepancies between Pinker’s results
and those of Maratsos et al. (1979). Pinker does this by first discussing the
general constraints on passives in English. He refers to a proposal in
Jackendoff (1983) called the thematic hierarchy constraint, which states that
sentences in English can only passivize if the object has a semantic role that
is lower than that of the subject on the following hierarchy: Agent-
Location/Source/Goal-Theme (or Patient). This blocks passivization of a
sentence such as ‘The game weighs a dollar’ if we assume that the subject of
‘weigh’ is a theme and the object is a location, both roles thus sharing the
same place in the hierarchy. Presumably this is a constraint which the child
acquires from positive evidence. Also, some verbs do not have clear-cut
semantic roles, and require linguistic experience for the roles to be
determined. Pinker suggests that several of the verbs tested by Maratsos et
al. were such verbs. The verbs used by Pinker, however, had clearer
semantic structure and could therefore be passivized. Pinker also points out
that such a proposal about passives predicts that children will not over-
generalize passives to verbs like ‘weigh’, an observation supported by the
limited data available to date.
An alternative explanation for the acquisition of passives can be found in
Berwick & Weinberg (1984). We have already mentioned Berwick &
Weinberg’s criticism of Pinker’s first proposals about the acquisition of
passives. They follow those criticisms with their own suggestions on how
passives are acquired. Berwick & Weinberg (1984) approach language
acquisition from the Government-Binding theory of Chomsky (1981,1982,
1986). They assume that the child is born with a set of linguistic universals
which are available at the onset of acquisition. As such, their general
orientation is that the child’s grammar is essentially adult-like from the
onset, and that errors in acquisition will be restricted. One possible source
of errors will be when the language is marked for some aspect of grammar
and the child treats it as the unmarked case. We have already discussed this
form of the parameter-setting model in Chapter 7.
In order to understand Berwick & Weinberg’s account, it is necessary to
have some understanding of the way in which Government-Binding theory
(or GB) generates passive sentences. Here we will give an extremely brief
description, referring the reader to other sources for a more detailed
presentation (e.g. Radford 1981; Berwick & Weinberg 1984).
GB proposes that passive sentences like ‘John was liked by Fred’ have the
underlying structure of (9.17):
(9.17) NP, Past be like +en John by Fred
Unlike earlier accounts in transformational theory (e.g. Chomsky 1957),
passives have an underlying structure distinct from their corresponding
active sentences. NP, stands for the occurrence of a NP which is lexically
empty (hence ‘e’), that is, there is no word inside it at the underlying level.
The item ‘en’ represents the passive morpheme on the verb. A universal
principle requires that all nouns in a sentence receive what is called abstract
case. Nouns receive case from a restricted number of sources: subjects can
receive case from the Aux, objects from the main verb, and oblique nouns
from prepositions. Passives create a problem for case assignment, however,
because the passive suffix ‘en’ is seen as blocking it. In our example, ‘en’
blocks the main verb ‘like’ from assigning case to the object ‘John’. This
problem is resolved by having the object move into the subject position
where it can receive case from the Aux.
The child acquiring English is born with the knowledge about the
properties of case assignment. In acquiring English, she is faced with two
major tasks: (i) to identify the grammatical morphemes of English and
(ii) to realize that the English passive morpheme ‘en’ blocks case assign-
ment. The latter has to be acquired because the unmarked situation is for
affixes not to affect case assignment. For example, in the active sentence
‘Fred liked John’, the past tense morpheme ‘ed’ does not stop the main verb
‘like’ from assigning case to the object ‘John’.
Berwick & Weinberg argue that the problem in acquiring the passive
resides in the passive affix ‘en’. First, this morpheme has a common
allomorph ‘ed’ which is homophonous with the past tense morpheme ‘ed’.
This will create problems with identifying ‘en’ as a distinct morpheme which
is also marked for blocking case assignment. Because of this problem,
children will be very cautious when they begin to acquire this morpheme.
Berwick & Weinberg state (p. 219): ‘The result is a drive to limit the scope
of this rule’s affixation as much as possible. Hence the initial hypothesis that
the rule applies to the most minimal class of verbs consistent with the data
that is heard.’ Since the first verbs heard with passives are action verbs, only
they will have the passive morpheme associated with them. The account
differs from Pinker’s in placing the restriction on the association of the
passive morpheme to a particular class of verbs, rather than on the Passive
rule itself.
As with Pinker’s analysis, we can ask if the Berwick & Weinberg
explanation is consistent with the acquisition facts. Recall that our first
finding was that children initially treat passives as actives. Berwick &
Weinberg do not offer an account of why this occurs, and we can only
speculate that there is a performance factor at work of the kind discussed by
Pinker. This is a problem because their model assumes at the onset that the
child ‘knows about word segmentation and morphology, for example, that
‘s’ is a plural morpheme in “cats” ’ (p. 203). This difficulty reflects a general
one with their proposal which is that it is so incomplete that it immediately
becomes difficult to assess.
This difficulty continues with the second set of data, on the restricted use
of passives. There is no principle of acquisition given which tells the child to
restrict the passive morpheme to active verbs. Further, the account is highly
dependent on the child only hearing passives with active verbs, a point still
to be substantiated. The model suffers the same problem that all strong
nativist models have: if the child has the principles available, then slower
acquisition needs to be tied to lack of input. A last point concerns the lack of
discussion about the nature of the input that leads the child to assume that
the passive morpheme blocks case assignment.
9.4.2 Relative clauses

Up to now, we have examined aspects of English grammar which can be
described as part of the acquisition of simple sentences. In these last two
sections, we look at a few other aspects of grammatical acquisition which
concern complex sentences. Complex sentences are those sentences where
one sentence (the subordinate clause) occurs within another sentence (the
main clause). Example (9.18) gives two kinds of complex English sentences
(where the subordinate clauses are shown within square brackets):
(9.18) a. Relative clause: We saw the boy [who left]
b. Verb complement: We know [that Mary left]
We asked [Bill to leave]
We promised Bill [to leave]
(9.18a) is an example of a relative clause, which is a clause that modifies an
NP. Such clauses are introduced by relativepronouns such as ‘who’, ‘that’,
and ‘which’. The examples in (9.18b) demonstrate clauses which function as
complements to the verb. Such clauses are introduced by complementizers
such as ‘that’. Tenseless complement clauses have the infinitive ‘to’ which
marks the clause as such. Here we just look at the acquisition of relative
clauses.
A striking feature of the spontaneous speech of children between 2 and 5
years is that there are relatively few relative clauses. Instead of embedding
one clause into another, children tend to string clauses together with
numerous insertions of ‘and’. This can be exemplified by the following
passage from my daughter Jennika at 4;4 when she was asked to tell the
story of Little Red Riding Hood.
A little girl goes for a walk and her mom tells her not to talk to visitors
and she saw a big ole wolf. The big blue wolf said do you want to pick
some flowers for your grandmother? Where do you going to? Just
somewhere in the woods and let’s see who gets there first and then he
puts on the grandma’s clothes. . .
This lack of relative clauses in spontaneous speech was confirmed in
Ingram (1975). There I examined the grammatical structures which children
used to begin stories. The onset of children’s stories is a context where a
relative clause is normally used. Since such stories typically begin with an
introductory sentence such as ‘There once was an old lady who lived in a
shoe’, with the introduction of the main character, e.g. ‘There once was an
old lady’, and then a relative clause which provides the qualification, e.g.
‘who lived in a shoe’, it was expected that children would also use relative
clauses in their own stories if the construction was part of their grammar.
To study this, I looked at the first sentence of 360 stories told by 210
children between the ages of 2 and 5 years. The data were taken from a
collection of stories collected by Pitcher & Prelinger (1963). I found that
there were three ways with which children began their stories: these are
shown in (9.19) along with examples of each.
(9.19) a. Juxtaposition
‘A bus. He went up a hill.’ (Dale 2;8)
‘Boy fell out of car. He went in car again.’ (Cass 2;s)
‘Once there was a house. Peter lived in the house.’ (Porteus 2;lO)
b. Conjunction
‘Once there was a little boy and he went for a walk in the woods.’
(Kurt 3;1)
‘Once there was a little boy and he hit someone on the head.’
(Marla 3;4)
c. Relative clause
‘Once there was a kitty named Cindy.’ (Tina 5;O)
‘Once there was a big scarey man with lots of faces’ (Emmet 5;O)
‘Once upon a time there was a little girl who went for a walk to a
restaurant.’ (Chloe 4;2)
These three patterns show a gradual fusion of the introduction and quali-
fication of a story. Juxtaposition is a very loose connection between these
parts. Some children just named the main character, as in the first example,
and then placed the qualification as a separate sentence. The second example
has the two together, but as one main clause. The third example shows an
appropriate introduction, but the qualification as an independent clause.
Conjunction shows a closer relation by the inclusion of an ‘and’, but the two
parts still have equal status. It is only in the use of relative clause that the
appropriate subordination of the qualification occurs.
Table 9.17 presents the proportion of times each of these three begin-
nings were used by the children. For the youngest children, juxtaposition is
clearly the dominant pattern. This preference decreases over time, but is
still the most used for all groups except the 5-year olds. We also see that
relative clauses are very rare for the first two age groups, and are only
beginning to be frequent for the oldest children. Even for these children,
many of the examples are formulaic such as ‘kitty named Cindy’ and ‘man
with lots of faces’. Non-reduced relative clauses with relative pronouns and
verb phrases were never frequent in the data.
Table 9.17 The proportion of use of three ways in which children begin
stories, taken from Ingram (1975: 112, table 3)
Age groups
Relation 2;O-2;11 3;O-3;11 4;U;ll 5;0-5;11
Juxtaposition 0.75 0.57 0.40 0.25

Conjunction 0.22 0.40 0.37 0.40
Relative clause 0.03 0.03 0.23 0.35
Sheldon (1974) There have been several studies on children’s comprehen-

sion of relative clauses, e.g. Gaer (1969), H. Brown (1971), de Villiers,
Tager-Flusberg, Hakuta & Cohen (1979), Hamburger & Crain (1982). The
study by Sheldon (1974), however, remains the one which initiated exten-
sive discussion on the acquisition of relatives.
Sheldon studied the comprehension of relative clauses by 33 children
divided into three age groups of 11 children each: I (3;8-4;3), I1 (4;6-4;11),
and I11 (5;O-55). The children were tested on four types of sentence with
relative clauses, shown in (9.20):
(9.20) Subject relatives
SS The dog that jumps over the pig bumps into the lion
SO The lion that the horse bumps into jumps over the giraffe
Object relatives
OS The pig bumps into the horse that jumps over the giraffe
0 0 T h e dog stands on the horse that the giraffe jumps over
The sentences differed depending on (i) whether the subject or object of
the main clause contained the relative clause, and (ii) whether it was the
subject or object of the relative clause which was relativized. SO, for
example, indicates that the relative clause modified the subject of the
sentence (or S), and that the object of the relative clause (or 0) was
relativized. The children were also given coordinate sentences which
corresponded to each of the relative clause sentence types. A coordinate
sentence corresponding to OS, for example, was ‘The pig bumps into the
horse and the horse jumps over the giraffe’. These were given to see if the
children were treating the relative clauses sentences the same as coordi-
nations or not. Each child received three sentences of each of the eight types
for a total of 24 sentences.
Each child was presented with a collection of toy animals appropriate for
the sentences and asked to act out the test sentences. For a sentence to be
scored correct, the child needed to carry out both acts correctly, but could
Table 9.18 Mean number of correct answers by age groups on relative
sentences and coordinate sentences, taken from Sheldon (1974: Tables 3
and 4) (A perfect score would be 3, since there were three sentences per
type)
Relative sentences Coordinate sentences

Age group ss so os 00 ss so os 00
I (3;8-4;3)* 1.0 0.18 0.54 1.36 1.75 1.75 1.63 2.13

I1 (4;6-4;11) 1.45 0.73 0.91 1.64 1.82 2.0 1.82 2.0
111 ($0-5;5) 2.27 0.64 1.17 1.55 2.64 2.0 2.27 2.55
Total average 1.58 0.52 0.88 1.52 2.1 1.93 1.93 2.23
* Three children in group I did not complete the coordinate sentences.
do so in any order. While responses with only one action were scored as
incorrect, Sheldon comments that these never were more than 4-5 percent
of the total responses for any sentence type.
Table 9.18 gives the results on the eight sentence types tested. Statistical
tests revealed the following results:
(a) The coordinate sentences were easier than the relative ones, and
showed no differences across the age groups. Conclusion: the children
were aware that relative sentences are different from coordinate ones.
This finding reminds us of the one by de Villiers & de Villiers (1973a)
on actives and passives, where the children showed an awareness that
passives were different even though they hadn’t determined their
actual features.
(b) The SS and 00 relative sentences were significantly easier than
SO and OS relatives. Conclusion: sentences are easier to understand
when the NP which has a relative and the NP which is relativized have
the same grammatical function. Sheldon refers to this as the parallel
function hypothesis.
(c) SS relatives show significant improvement across age groups, but
the 00 relatives do not. Conclusion: there must be some additional
factor which causes this effect.
The first finding has been verified by other research and has not been
particularly controversial. The opposite has been the case, however,
concerning the second finding. A useful review of this issue can be found in
Bowerman (1979). This literature, in fact, has been so contradictory that
Bowerman has commented (p. 292) that it ‘presents a tangled web of
conflicting findings and alternative interpretations’. The criticisms can be
grouped into two categories. First, there are those who question whether
the data support the conclusion. Some studies have found results which
show different patterns of difficulty, e.g. de Villiers et ul. (1979). Others
(e.g. Hamburger & Crain 1982) have noted that the parallel function
hypothesis does not account for the fact that SS relatives are easier than 00
ones (although Sheldon has a proposal for this which will be discussed
shortly). Second, others have questioned the parallel function hypothesis
on theoretical grounds. It is not clear what role this function plays in the
acquisition of relatives. While it appears to be a performance factor, it could
operate under two possible conditions: (a) the child has acquired the
grammar of English relatives, but uses this as a processing constraint when
short-term memory fails, or (b) the child has not acquired the grammar of
English relatives, and uses this as a heuristic for responding while acquiring
them. As pointed out by Hamburger & Crain (1982), this is a claim about
the performance factors affecting relatives rather than one about how the
child’s competence emerges. Tavakolian (1981b) has rejected the parallel
function hypothesis in that it assigns too much competence to the child and
claims instead that the children are treating these sentences as if they were
simply two conjoined clauses.
Sheldon does offer a proposal to account for the third finding which
addresses the child’s linguistic competence. She argues that the pattern in
the third finding given above results from the child’s overuse of the English
rule of Extraposition. This is a rule in the adult language which moves a
relative clause from a subject NP to the end of the sentence. This rule, for
example, would operate on (9.21a) to produce (9.21b):
(9.21) a. The boy who left was sick
b. The boy was sick who left
c. The dog bumps into the horse that the giraffe jumps over
d. The dog that the giruffe jumps over bumps into the horse
Sheldon proposes that children have this rule, and thus treat 00 sentences
like (9.21~)as if they were an SO sentence as in (9.21d). The distribution of
the children’s responses, then, was as follows (figures are percentages):
ss so os 00
Correct 52 17 29 50
Parallel function 14 69 14 12
Extraposition 44 32
Other 33 13 12 5
An important set of evidence in support of the Extraposition argument is
that approximately one-third of the children only made this error on the
object relatives. Even more importantly, the same subjects virtually never
got any of the object relatives correct. The main criticisms of the proposal
are (i) that there is little other evidence to date that children have such a
rule at this point in acquisition, and (ii) doubt as to why the child would ever
form it this way since it is not a pattern found in the adult language. If it were
the result of overgeneralization of the rule, as Sheldon suggests, then the
children would require indirect negative evidence to overcome their
mistakes.
Sheldon’s study was important in drawing attention to children’s emerg-
ing ability to understand relatives, and it suggested some possible ways to
account for it. It also influenced the study on relatives reported in
Hamburger & Crain (1982).
Hamburger & Crain (1982) Hamburger & Crain approach the topic from
the perspective of Language Acquisition as discussed in Part I of this
volume. They are quick to point out that Sheldon’s study is more a study of
performance factors than competence factors. Their own orientation is
along the lines of what we have called a strong inclusion hypothesis. They
set out to show that children may have a better knowledge of relative
clauses than is indicated by research to date. Since they are in agreement
with Sheldon’s implicit claim that children have rules for relatives, they are
more concerned about the extent of errors found in the data. In particular,
they want to construct the experiments on the acquisition of relative clauses
in such a way as to minimize the conditions that encourage the use of
performance factors.
Hamburger & Crain studied the comprehension of relative clauses by
having children act out sentences, as Sheldon did. They adjusted other
aspects of Sheldon’s methodology in the following ways:
(a) Several exemplars are presented of the toy which is referred to by

the head of the relative clause. This is done because relative clauses
served to delimit a set of objects, and having a choice from this set
should make the task more natural or consistent with conversational
requirements, that is, we normally only use a relative clause if there is
a need to select one specific member of a set.
(b) The order in which the child performs the two actions in the
sentences is recorded and included in the analysis. Sheldon did not
take order into consideration, but it has implications for the child’s
interpretations (see below).
(c) Sentences of only one type were tested, the OS ones. They were
selected for two reasons: (i) they had resulted in the most errors in
previous studies (although note that the SO ones were the most difficult
category in Sheldon 1974); (ii) they show an order of mention that is

opposite to the conceptual order. For example, in the OS sentence
‘The pig bumps into the horse that jumps over the giraffe’ it is assumed
that the horse jumps over the giraffe’ before the pig bumps into it. No
independent psychological evidence, however, is presented to support
this claim.
The subjects were 18 children divided into groups of 3-, 4-, and 5-year
olds. Their results were scored in the following way. Each NP in the test
sentences was assigned a number as in (9.22):
(9.22) 1 2 3
The duck stands on the horse that jumps over the pig
The child’s actions were then coded according to their order and the NPs
used. For example, 23,12 means that the child had the horse jump over the
pig, and then the duck stand on the horse. Only two numbers were used if
only one action was performed.
Table 9.19 Percentages of use of three response patterns for OS sentences,

adapted from Hamburger & Crain (1982: 267, Table 7.2)
Patterns of response
Age groups 12,23 23,12 12 Other Sheldon’s scoring
(perfect score = 3)
3 years 42 27 0 31 2.07
4 years 18 43 13 26 1.83
5 years 5 35 55 5 1.2
Average 22 35 23 21 1.7
Table 9.19 presents the major results of the study. Hamburger & Crain
treat the three patterns given in Table 9.19 (excluding ‘other’) as all
correct, i.e. indicating correct processing of the sentence. They see the
patterns 12,23 and 23,12 as differing in that the latter reflects more
accurately the natural order of the events. They also score 12 as correct
because it indicates that the child is aware that the relative clause is
functioning to help delimit which toy to select, and that the main action is
having the main clause subject act upon the main clause object. By these
measures, then, the children can be interpreted as showing good ability at
OS structures, since the percentages of correct response are 60,74, and 95
for the three groups.
In Table 9.19, I have also entered the score which the children would
have received if Sheldon’s scoring methods had been used. By her measure,
both 12,23 and 23’12 would have been correct, but 12 would not. We then
multiplied by 3 the percentages obtained, to get an equivalent score. By
comparing these scores with those in Table 9.18, it is seen that the
Hamburger & Crain’s children did do better than those tested by Sheldon.
There appear to be two major differences in the two studies which led to
the better performance of the children: several exemplars of the head of the
relative clause were used, and only one sentence type with relatives was
tested. One can propose that these two changes made it much easier to do
the task, which indeed was the goal of the study. The general conclusion is
that at least some younger children can be shown to have more knowledge
of relative clauses than has been proposed (at least by some) when optimal
test conditions are involved.
While the above conclusion appears to be a reasonable one, I would like
to point out an alternative interpretation. If all three response types above
are correct ones, then one would expect a gradual improvement in all three
across time. This, in fact, is not the case. The 12,23 and 12 patterns are
actually going in opposite directions. A different interpretation results if we
assume that 12,23 results from incorrect processing and 12 results from
correct processing. This is also reasonable, in that 12,23 could also result
from treating the sentence as a form of coordination whereby the child
treats the last NP mentioned before the second verb as its subject. Under
this alternative, relative clause comprehension does not look as well
developed in children younger than 5 years.
A second study conducted by Hamburger & Crain involved the pro-
duction of relative clauses, which, as we observed earlier, are not common
in the speech of children at these ages. As in the above study, Hamburger &
Crain wanted to create an optimal situation for the elicitation of relatives.
To do this, they placed children in a situation where they had to describe
something to someone who was blindfolded. The child had to tell this
person to pick up one of a set of objects which was doing something. For
example, there could be a set of walruses where one was seen tickling a
zebra. An appropriate response would be ‘Pick up the walrus that is tickling
the zebra’. Ten of the 12 children responded and 72 percent of their
responses used some form of an OS sentence.
Summary. The results of the above studies demonstrate several points

about the acquisition of relative clauses. They do not appear to be abundant
in the speech of young children in this age range. There is evidence,
however, that they have some awareness of their existence in the language,
and that they can comprehend and even produce them under certain
conditions. We are reluctant, however, to conclude that children have
acquired relative clauses by the end of this period around age 4;O. It appears
that their acquisition has begun, but that the productivity in their use which
is characteristic of adult speech has still not occurred. The results, then, are
similar to those found in our review of passives.
484 S I M P L E SENTENCES: ACQUISITION OF G R A M M A T I C A L M O R P H E M E S
9.4.3 Pronominal reference

Pronominal reference concerns the conditions under which English (and
languages in general) will allow pronouns to refer to surrounding noun
phrases. The study of this topic is one that has received a great deal of
attention in recent years in both linguistic theory and language acquisition.
It has been found to be an area that is extremely complex in the range of
phenomena which it affects, and one which varies from language to
language. While linguistic theory has begun to study the range of such
conditions across languages, much of the work on language acquisition has
been restricted to English.
We will look at English pronominal reference by first distinguishing three
general conditions, adapted from Chomsky (1981). These will be referred
to as reflexive, pronominal, and nominal conditions on reference. They are
exemplified by the sentences given in (9.23).
(9.23) Conditions on coreference

a. Reflexive
1. John thinks that Bill likes himself (= Bill)
2. *John thinks that Bill likes himself (= John)
3. *Himself likes Bill
b. Pronominal
4. John thinks that Bill likes him (= John)
5. *John thinks that Bill likes him (= Bill)
c. Nominal
Blocked backwards pronominalization
6 . He (John) thinks that Bill left
7. *He (= Bill) thinks that Bill left
Backwards pronominalization
8. When he (= Bill) left, Bill was hungry
Blocked forwards pronominalization
9. In front of Bill, he (= John) saw a snake
10. *In front of Bill, he (= Bill) saw a snake
Reflexives are pronominal forms in English which add the morpheme

‘self’ to English pronouns. Their use in English has a requirement that the
NP (or antecedent) to which they refer must occur in the same clause.
Further, the NP antecedent must precede it in the sentence. These general
facts are exemplified in the sentences in (9.23). Sentences 1and 2 show that
the reflexive ‘himself’ can only refer to ‘Bill’ because only ‘Bill’ is in the
same clause. (‘John’ is in the higher clause which has the form ‘John thinks
X’.) Sentence 2 is therefore ungrammatical under the interpretation where
‘himself’ is meant to refer to ‘John’. Sentence 3 shows that the reflexive
cannot precede the antecedent.
Pronouns are referential forms in English that are normally divided into
subject, object, and possessive cases, e.g. ‘I’, ‘me’, and ‘my’. Their
conditions of reference are generally complementary to those of reflexives.
That is, pronouns cannot refer to an antecedent in the same clause. This is
shown in sentences 5 and 6 above, where ‘him’ can refer to ‘John’ but not
‘Bill’, where ‘John’ is outside the clause in which ‘him’ occurs but ‘Bill’ is
not.
The conditions on the possible order of pronouns and their antecedents
are more complex than the condition for reflexives. While reflexives cannot
precede their antecedents, pronouns can precede their antecedents under
certain conditions. There are also certain conditions under which pronouns
cannot follow their antecedents, but need to precede them. These sets of
conditions are exemplified in (9.23) as nominal conditions on reference.
They are grouped under the term ‘nominal’ because it appears that they can
best be described by referring to the position of the nominal antecedent in
relation to the pronoun.
The nominal condition on coreference can be seen by looking at the
relations between the pronouns and their potential antecedents in the three
types given in (9.23~).Blocked backwards pronominalization is a case
where the pronoun cannot precede its antecedent. This is similar to the
condition on reflexives and might initially lead one to propose that it is the
same for pronouns. The existence of backwards pronominalization,
however, indicates that the condition is not one of simple linear order. In
sentence 8, the pronoun precedes the antecedent ‘Bill’, and yet may refer to
it. Such sentences have led linguists to propose that the condition needs to
include reference to the relative position of the pronoun and the NP in the
structure in the sentence. Observe that the pronoun occurs in a higher
position than ‘Bill’ in sentence 7, but not in sentence 8: ‘He’ is the main
clause subject in 7, whereas ‘Bill’ is the main clause subject in 8. That is, it
appears that coreference is possible as long as the pronoun does not occur in
a higher place in the hierarchical structure of the sentence than does the
antecedent NP. This observation is supported further by the cases of
blocked forwards pronominalization. Here the NP cannot precede the
pronoun as shown in sentence 10. Note in 10 that the pronoun, as in 7,
occurs as the main clause subject. We can summarize the conditions
discussed as:
(9.24) (a) Reflexive: a reflexive pronoun must refer to an NP which

precedes it within the same clause;
486 SIMPLE SENTENCES : ACQUISITION O F GRAMMATICAL MORPHEMES
(b) Pronominal: a pronoun cannot refer to an NP which occurs
within the same clause;
(c) Nominal: a nominal cannot be the antecedent for a pronoun
which occurs in a higher position in the structure of the sentence.
The facts on pronominal reference have been overly simplified in order to

provide a general feel for the range of phenomena which are involved. The
exact conditions require more careful specification of the grammatical roles
of the pronominal and nominal forms as well as the grammatical structure of
the relevant constructions. Also, it needs to be pointed out that languages
differ from English in the way in which reference is handled. Japanese, for
example, allows what we can call ‘long distance reflexives’: it allows a
sentence like 2 in (9.23) where the reflexive refers to ‘John’. Further, while
it allows long distance reflexives, Japanese imposes a ‘subject condition’ on
reflexives, that is, the antecedent for the reflexive can only be the subject of
the sentence. Other languages will show still further differences, and the
facts on such variation are still being gathered (see Wexler & Manzini 1987).
Linguists differ in the ways they attempt to explain such facts. There are
two general issues that arise. One concerns whether or not linear order
needs to be stated as part of the conditions. In (9.24) I mention restrictions
on ordering in (a) and (b), but not (c). A more theoretical approach could
eliminate any reference to order whatsoever by restricting the account of
reference to structural conditions between the antecedent and the pronomi-
nal form.
The most commonly discussed structural relation of this kind is that of
c-command (Reinhart 1976,1981,1983). C-command, very simply stated, is
a relation between two constituents in a grammatical structure such that
they share a particular relation to some other grammatical category, let’s
say G. Under one definition, we can say that one category c-commands
another if the G that is directly above it is also above the other category.
There are two features of this relation that need to be made explicit. One
concerns the possible grammatical categories of G. These are usually
restricted to S (for sentence), and VP (for verb phrase). The other is the
notion of ‘directly above’. The constituent which is c-commanding another
needs to have a relation to the G category such that no other G category
intervenes.
We can demonstrate the use of c-command in accounting for the facts in
(9.24). In a simple English reflexive sentence like ‘John likes himself’, the
antecedent c-commands the reflexive since the first S over ‘John’ is also over
‘himself’. ‘Himself’, however, does not c-command ‘John’ because the first
G category over ‘himself’ is VP, and ‘John’ occurs outside the VP. These
relations can be seen by examining the grammatical structure in (9.25):
(9.25)
S
/ A
NP VP
I V
nNP
John
I I
likes himself
We can now eliminate the linear condition on reflexives in (9.24) by

referring to c-command. We can say that a reflexive cannot c-command its
antecedent. Further, we can make the nominal condition of reference more
explicit by using c-command. (9.24~)can be stated simply as ‘a pronoun
cannot c-command its antecedent’. This can be seen by comparing sen-
tences 7 and 8 in (9.23). In sentence 7, the pronoun c-commands the
antecedent and is therefore ungrammatical. In sentence 8, however, the
pronoun does not c-command the antecedent ‘Bill’, because the S immedi-
ately over the pronoun is the one for the clause ‘When he left’, and ‘Bill’ is
not inside of that clause. Hence, the coreference is allowed.
Many theoretical accounts of pronominal reference have attempted to
eliminate the need to refer to the linear order of constituents and use
c-command exclusively. The most famous of these is that of Chomsky
(1981) where the three kinds of conditions in (9.24) are stated in formal
terms using solely c-command. They are referred to as ‘binding conditions’
and labeled A-C respectively. (For example, a reference to ‘condition A’ in
the theoretical literature refers to the conditions on English reflexives.
Condition B refers to the condition on pronouns, and condition C is that on
nominals.) The term ‘binding’ simply refers to the situation where one
constituent is c-commanded by another. A reflexive, for example, is
‘bound’ by its antecedent, because it is c-commanded by it. While this is a
dominant trend, other attempts exist where linear order is retained in the
explanation of reference (e.g. O’Grady 1986).
A second issue in the theoretical explanation of pronominal reference is
the generality of the conditions on coreference. The appeal of a notion like
c-command is that it is a simple account of how reference is handled. We can
propose that Universal Grammar is such that all conditions on reference are
handled by c-command. The child then needs only to determine the
appropriate c-command conditions for its laEguage. (Presumably the range
of possible languages is also restricted, as we will discuss further below.)
Chomsky’s binding conditions are stated in such a way that a simple
statement about c-command relations applies to a wide range of construc-
tions. Take the three kinds of constructions given in (9.23) under nominal
conditions on reference (these are actually just a subset of the possible
constructions which are involved): all three are handled by condition C of
the binding conditions which states roughly that a pronoun cannot c-
command its antecedent. Carden (1986), however, has argued against a
general account of these constructions, using a range of linguistic and
psycholinguistic arguments. In particular, he has focussed on blocked
forwards pronominalization, where, under closer examination, the c-
command account has problems.
The topic of pronominal reference is of particular interest for language
acquisition because it provides a testing ground for a number of the major
issues in the field. This point was briefly discussed earlier in Chapter 4. One
general issue is the predictions that particular theories make about acqui-
sition data. It turns out that data from pronominal acquisition have been
interpreted as support for each of the general theories we have treated.
Related to this point is the extent to which data from pronominal acquisition
can be used to support one theory over another. Another issue is the
learnability of pronominal reference. There are important questions that
arise concerning what constitutes positive evidence for its acquisition, and
what the child needs as part of Universal Grammar for it to be acquired.
Each of these issues will be briefly reviewed here.
Data on the acquisition of pronominal reference Studies on pronominal

reference have been conducted primarily during the 1980s (see under
‘Further reading’ at the end of this chapter). There was, however, a much
earlier study by Carol Chomsky (1969) which provided some preliminary
data and interpretations. I will begin with that study and then turn to the
more recent work.
Carol Chomsky (1969) This study is a classic in the field for several
reasons. At the time of its appearance, there was a general impression
that language acquisition was complete by age 4 or 5 years. Chomsky
demonstrated that several of the more complex aspects of English syntax
may not be acquired completely until the primary school years. The
study was also important for isolating the areas which may be acquired
later. She conducted studies on children’s ability to process complex
sentences with the verbs ‘easy to see’, ‘promise’, ‘ask’, and ‘tell’. She also
proposed stages that children appear to go through in acquiring such
constructions.
The last study reported by Chomsky was one on pronominalization. It
looked at the ability of 40 children between 5 and 10 years of age to process
three specific constructions, exemplified here in (9.26). Chomsky simply
refers to these as ‘types’, and we have provided the labels for them, based
on terminology used in Ingram & Shaw (1981).
(9.26) type 1 Blocked backwards pronominalization
He found out that Mickey won the race
type 2: Backwards pronominalization
Before he went out, Pluto took a nap
type 3: Forwards pronominalization
Pluto thinks he knows everything
Each child took a 15-item comprehension test, with five sentences for each
of the three types. First the child was presented with two dolls, Mickey and
Pluto. Then the child was read the test sentence followed with a question
about its interpretation, e.g. ‘Pluto thinks that he knows everything. Pluto
thinks that who knows everything?’ The child then would answer with the
name of one of the dolls. Each response was then recorded as either
coreferential or non-coreferential.
Since the responses on types 2 and 3 could be either coreferential or non-
coreferential, Chomsky only considered the scores on type 1, i.e. blocked
backwards pronominalization. She considered this construction to be
acquired when all five test sentences were answered correctly, i.e. as being
non-coreferential. Of the 40 subjects, 31 (or 78 percent) showed acquisition
of type 1. Further, with just three exceptions, it was at age 5-6 years that
children began to meet her criterion for acquisition. She noted that 36 sub-
jects also showed at least some coreferential responses on type 2. From these
results she concluded (p. 109): ‘the principles of pronominalization appear
to be acquired by the majority of children at about the same age’. Impor-
tantly, her interpretation of the results is a maturational one. She proposes
that the principle (or principles) which children use for pronominal reference
matures around age 5 or 6 , and its emergence is independent of ‘linguistic
elaboration in the environment, intelligence, and rate of general cognitive
development’ (p. 110). This principle, using N. Chomsky’s more recent pro-
posals, would be condition C of the binding conditions.
Chomsky has little to say about the children’s performance before the
maturation of the appropriate grammatical principle. She does, however,
mention four subjects who got type 1 correct and never responded with
coreferential responses on Type 2 - backwards pronominalization. The
pattern of these subjects could result from a simple rule of linear order
rather than one of c-command. Chomsky, in fact, leans in this direction by
saying that these subjects ‘seem to be operating with the simple principle
that the basic function of a pronoun is to refer to what precedes, without
further refinements’ (p. 109). This opens the possibility that the structural
principle needed for backwards pronominalization is preceded in acqui-
sition by a simpler rule based on the order of appearance of the pronoun and
its antecedent.
Chomsky’s study was one which was clearly ahead of its time. The
development since then of linguistic theory on pronominal reference makes
it possible to restate her findings in more current terminology. Her
maturational explanation implies that c-command is a principle which
matures at a point around age 5 or 6. Before then, without that principle,
children rely on a rule which uses only the linear relation of the pronoun and
the antecedent. The result of this situation is a marked discontinuity in the
acquisition of pronominal reference.
More recent research has attempted to expand the range of constructions
studied (Ingram & Shaw 1981; Solan 1983,1987; Crain & McKee 1985), as
well as to examine other languages (O’Grady, Suzuki-Wei & Cho 1986).
This research has also included an extension of the topic to reflexive con-
structions (Wexler & Chen 1985).
One line of this research has been from a nativist perspective. The
primary support for the maturational explanation of C. Chomsky has come
from the research by Wexler & Chen (1985). They have found that the
proposed condition B, i.e. the pronominal condition in (9.24b), does not
appear to be consistently acquired until around age 6. They imply that the
principle matures around that time. Much of the other work, however, has
been in a direction away from a maturational explanation. Crain & McKee
(1985), for example, have adjusted the task conditions used and have been
able to get correct responses from children as young as 3 years of age. Their
research is in the same spirit as that reported earlier on relative clauses by
Hamburger & Crain (1982). They hope to prove, by showing correct
performance by young children, that the child’s grammar is essentially
adult-like.
A somewhat paradoxical study from the same perspective is that of Solan
(1983). Solan studied a range of constructions with pronominal reference in
four experiments on 36 children between 5;2 and 8;5. He found that the
children did well on cases of backwards pronominalization and concludes a
version of the strong inclusion hypothesis which is reproduced here:
The restrictive model
Children will have little difficulty using notions such as c-command,
d-command, precedence and clausematedness. Once they have
learned that backward anaphora is possible at all, they will quickly
hypothesize restrictions based on the correct definitions.
(Solan 1983: 92)
Solan contrasts this model with one which he calls the ‘developmental
model’. We find this paradoxical in that Solan’s results seem very much
interpretable within the constructionist perspective discussed in this text.
First, he acknowledges an early stage in the acquisition of pronominal
reference when children rely solely on linear order. Second, the use of
linear order (or precedence) does not drop out when structural properties
such as c-command appear, but remains as part of the rule. He thus adapts
the binding conditions of the adult grammar to include earlier properties
used by the child. This is very much in keeping with the Constructionist
Assumption of Chapter 4. Lastly, the subjects are relatively old, and at an
age when results in Carol Chomsky (1969) and Ingram & Shaw (1981) show
backwards pronominalization is acquired. Their use of a structural rule at
that point of acquisition is accepted in research from a developmental
perspective.
More explicit claims for a developmental or constructionist model of the
acquisition of pronominal reference can be found in Ingram & Shaw (1981)
and in work by O’Grady and his students (Taylor-Browne 1983; O’Grady
1986; O’Grady, Suzuki-Wei & Cho 1986). Ingram & Shaw (1981) examined
both backwards pronominalization and blocked forwards pronominali-
zation in 100 children between 3;O and 8;O. O’Grady, Suzuki-Wei & Cho
(1986) looked at pronominal reference in separate studies on Korean and
Japanese. These studies, as well as the review in Carden (1986), generally
propose four descriptive stages in the acquisition of pronominal reference.
They are summarized below, adapted from Ingram & Shaw (1981; see also
O’Grady 1986: 146-7):
Four stages in the acquisition of pronominal reference
stage 1: Use of coreference: a pronoun may refer to an NP in a clause
which may either precede or follow it;
stage2: Useoflinearorder:apronoun mayonlyreferto aprecedingNP;
stage 3: Use of dominance: a pronoun may refer to a following NP if
the appropriate structural conditions exist, i.e. children acquire
backwards pronominalization;
stage 4: Use of dominance: a pronoun cannot refer to a preceding NP
under certain structural conditions, Le. blocked forwards pronomi-
nalization is acquired.
As in Solan’s work, these studies show children moving from the exclusive
use of linear order to a system which takes into account the structural
properties of the sentence. Ingram & Shaw do not express a position on the
nature of the latter properties while Solan and O’Grady both do. The fact
that backwards pronominalization and blocked forwards pronominalization
are acquired at separate times has to be taken as evidence against the
generality and nature of condition C of Chomsky’s binding conditions.
Learnability Learnability has to do with how the child is able to determine

the conditions of pronominal reference from the sentences which it hears.
One issue is what constitutes positive evidence for pronominal reference.
Coreference is clearly an example of positive evidence, telling the child that
coreference is possible. What, however, about non-coreference? Since
many cases of coreference are optional, the child could assume that an
instance of non-coreference is just an optional application of a rule which
allows coreference. This can be circumvented by allowing indirect negative
evidence, or by arguing that non-coreference is itself a form of positive
evidence.
Another issue is how the child formulates the rules underlying pronomi-
nal reference. The potential problem here has to do with the possibility that
the child will overgeneralize a pattern and have no way to learn that the rule
is incorrect. For example, suppose that children do have a stage such as
Stage 1described above, where the rule is simply that a pronoun can refer to
any pronoun. How does the child narrow this rule down to the one in stage 2
where the pronoun can only refer to a preceding noun? Numerous other
such problems could potentially arise if children generalize patterns of
coreference. These problems are also compounded when we consider the
variability in pronominal reference across languages.
One approach to this problem is to argue that children are quite
conservative in the ways that they formulate rules for complex grammatical
processes like pronominal reference. This perspective is argued for in
O’Grady (1986), among others. The basic argument is that children acquire
the rule on a construction by construction basis, and do not overgeneralize
as they seem to do in morphology and semantics. It is somewhat ironic that
such piecemeal acquisition has normally been associated with Child Lan-
guage theories rather than those of Language Acquisition.
The problem with cross-linguistic acquisition has been addressed in
Wexler & Manzini (1987). Wexler & Manzini have discussed the possibility
that languages differ from each other on some grammatical construction in
such a way that the contexts of application in one language are always a
subset of those in another. For example, I have already mentioned that
Japanese differs from English by restricting reflexive antecedents to sub-
jects. The contexts for antecedents in Japanese are therefore a subset of
those for English. I also mentioned that Japanese allows long distance
reflexives whereas English does not. In this case, then, the English contexts
are a subset of the Japanese ones.
Wexler & Manzini, like Chomsky (1981), accept the parameter-setting
model of language acquisition. On any particular parameter of grammar,
Universal Grammar will allow a restricted range of possibilities. The child is
born with knowledge about these, and a principle of acquisition which
Wexler & Manzini refer to as the subset principle. This principle is
summarized as:
Subset principle: for any parameter of grammar, assume initially in
acquisition that the unmarked case is that which has the most
restricted range of application.
We can see how this principle operates by looking at the English and
Japanese differences on reflexives. Since the subject condition of Japanese
is more restricted than the one for English, it will be the unmarked case.
Conversely, the English clausemate restriction is more limited than the long
distance reflexives in Japanese and is thus the unmarked case. The child is
therefore born with the following two expectations about reflexives: (i) the
antecedent is restricted to subjects, and (ii) the antecedent must be in the
same clause as the reflexive (the clausemate condition). The sentences in
any language which have a wider range of application will constitute
positive evidence. For example, the English child will know that English
does not have a subject condition when it hears antecedents for reflexives
which are not subjects. So too, a Japanese child will realize that Japanese
does not have a clausemate restriction when it hears sentences with long
distance reflexives.
The subset principle is obviously a very powerful claim about how the
diverse patterns of language are acquired. An important point is that it is
also potentially testable as long as one does not take the maturational view
of acquisition, that is, we can test children in the relevant linguistic contexts
to see if the subset principle is at work.
9.5 Cross-linguistic morphological acquisition

Thus far our coverage of morphological acquisition has focussed on
English. As stated at the onset, however, languages differ greatly from one
another in their morphological structure. Further, since English has a
relatively limited morphology, it is not the best choice for the study of
morphological development - the concentration of studies on English is
more the result of demographic reasons than theoretical ones.
There are, in fact, a large number of studies on morphological acquisition
in languages other than English. A recent two-volume work edited by
Slobin (1986), for example, is devoted to the general issue of cross-linguistic
development. There is not, however, any in-depth review and assessment of
this wide range of data, although some selective comparisons have been
made (e.g. Slobin 1982). The reports are often primarily descriptive in
nature, and often not easily comparable to other studies as a result of
various methodological differences. Much of these data, therefore, have
gone largely unnoticed.
Here it is useful to take a brief look at just one aspect of these data, on the
acquisition of languages with greater morphological complexity than
English. As we have seen earlier, English children tend to acquire morpho-
logical endings after they have established some basic grammatical rela-
tions. We have also seen this pattern used to support the semantic
bootstrapping view of acquisition. The question then naturally arises
whether or not children acquiring more morphologically complex languages
will nonetheless show the same general sequence. This question may be
addressed by providing brief reports of studies on five diverse languages:
Dutch, Estonian, Greenlandic Eskimo, Hindi, and QuichC.
Dutch As a language closely related to English, Dutch is not as rich in

bound morphology as the other languages we will consider. (It does,
however, have a rich system of diminutive suffixes, whose acquisition is
discussed in Snow, Smith & Hoefnagel-Hohle 1980.) In fact, since it is close
to English, one would expect that the pattern of its acquisition should be
similar to the English data. A study by Arlman-Rupp, van Niekerk de Haan
& van de Sandt-Koenderman (1976) (henceforth ANS), however, has
concluded that such is not the case.
ANS examined four children longitudinally between 2;O and 2;3. The data
from each child consisted of approximately four-and-a-half hours of spon-
taneous language collected in five sessions at six-week intervals. The chil-
dren’s sentences were then classified into the ten semantic relations for Stage I
speech proposed by Brown (1973). Recall that this was done at a time when
Brown’s semantic taxonomy was having a significant impact on the field.
ANS found that the semantic classification accounted overall for less than
half of the children’s sentences. The actual percentages were 48,49,50, and
36. Many of the unclassified utterances contained what ANS referred to as
‘modal’ elements. They listed five of these, reproduced here in (9.27):
(9.27) a. adverbs of negation and affirmation, e.g. ‘niet’ (not), ‘nee’ (no),
‘ja’ (yes);
b. quasi-temporal adverbs, e.g. ‘nou’ (now);
c. other adverbs, e.g. ‘dan’ (then), ‘maar’ (just);
d. the interjection ‘hoor’ (hear, listen);
e. modal auxiliaries, e.g. ‘kan’ (can), ‘wil’ (will).
They also say (p. 273): ‘Other non-terms such as copulas, articles and
prepositions contribute yet more surplus to these utterances.’ In other
words, it was not apparent that the acquisition of a small basic set of
semantic relations was the primary motivation behind the acquisition of the
first sentences in Dutch.
Of importance in interpreting this result is the difference in basic sentence
structures between English and Dutch. Like German, Dutch uses the
surface word order of Subject-Aux-Object-Verb. One possibility is that the
position of the Aux between the subject and object and away from the main
verb may draw the child’s attention to it more than in English. ANS’s
own explanation is that the modal elements receive more frequent use in
Dutch and so are more prominent in the language. They state (p. 274): ‘The
strikingly early, frequent and appropriate use of modal elements by
children learning Dutch is very likely a reflection of the fact that modality is
much more frequently expressed in modal adverbs and other modal
elements by adult speakers of Dutch than in other languages studied.’
There are two features of these results which are of particular import-
ance: (i) they suggest some basic cross-linguistic differences in acquisition;
(ii) they indicate that type frequency may play an important role in syntactic
acquisition. (Note that we have already discussed the role of type frequency
in phonological acquisition.)
Estonian A Finno-Ugric language, Estonian is characterized by a rich

system of case suffixes on nouns, several verb inflections, and postpositions
instead of prepositions. It is also an agglutinative language, i.e. ‘words are
typically composed of a sequence of morphs with each morph representing
one morpheme’ (Lyons 1968: 188). Given the findings on morphological
development in English, one would expect that this complex system would
be difficult to acquire. Some preliminary data reported in Lipp (1977),
however, suggests that this may not be the case.
Lipp collected spontaneous language samples from three Estonian chil-
dren, Avo, Karen, and Erik, ranging from 2;6 to 3;O at the onset of the
study. They were observed four times over an eight-week period. For
comparative purposes, 250 utterances were selected from each sample for
subsequent analysis. The MLUs for each of the three children for the first
and last sessions were 1.87-2.91 (Avo), 2.75-2.68 (Karen), and 2.62-3.36
(Erik). Using Brown’s stages, Avo went from Stage I to 111, Karen was at
Stage I11 throughout the study, and Erik went from Stage 111to IV. A range
of grammatical morphemes was analyzed for each session, using frequency
and percentages of obligatory occurrence. A morpheme was considered as
acquired when it occurred at least four times in a session and was used
correctly in 90 percent of its obligatory contexts.
The first general finding was that 20 of the morphemes studied reached
the criterion of acquisition during the period of observations. This is a much
higher figure than that obtained by Brown for English. Karen, for example,
acquired ten morphemes while at MLU stage 111. Another finding con-
cerned the order of acquisition of the morphemes. Lipp divided them into
an early and a late group. She found that the early group shared several
similarities with the early morphemes acquired in English. One of the
exceptions was the Estonian copula in the 3rd person singular form. This
form was acquired much earlier than its English counterpart. In comparing
the late morphemes with the English data, however, Lipp found very little
overlap. The major characteristics of the late Estonian morphemes were:
(i) they consisted of combinations of several morphemes of agreement; and
(ii) ‘combined with fewer nouns and verb stems, and generally occurred less
frequently in the data’ (p. 318). Lipp also found several differences in the
order of acquisition for Estonian and Latvian, using the Latvian data in
Ruke-Draviga (1973).
1
Greenlandic Eskimo The North American languages constitute a rich and
diverse range of languages which have been studied in some detail. Due in
part to the fact that many of them are, or are on the verge of, becoming
extinct, their acquisition has not often been studied in detail. For example, a 1
recent attempt by Pye to study two native languages of British Columbia,

Thompson and Chilcotin, did not turn up a single subject acquiring either as
the first and only language (see Pye 1987). There are, however, a few of
these languages which are still being acquired as first languages and can be
studied if the investigator can overcome the geographic and political
problems which may be involved.
One such study is Fortescue (1984/85) on the acquisition of Greenlandic
Eskimo, or henceforth Eskimo. Eskimo constitutes a range of mutually
intelligible dialects that cross a vast geographical area in the north-most
regions of North America. (In recent years it is referred to as Inuit, the
preferred name of its native speakers). The language has been known for
years to be a polysynthetic language, that is, one which is very rich in its
morphological structure. Fortescue comments (pp. 101-2):
. . . the typical word form consists of a stem followed by from zero to at
least eight derivational affixes then an obligatory inflexional ending
(followed in turn by optional clitics), all bound together by complex
morphophonemic patterns of morpheme attachment and falling under
one potential intonational tone unit.
It is also agglutinative, in that there tends to be a single meaning attached to
each of the many morphemes. This is distinct from languages like Polish
where several grammatical meanings may be attached to a morph.
Fortescue examined a half-hour language sample from a 2-year-old boy,
Aqissiaq, acquiring Greenlandic Eskimo as a first language. He found that
Aqissiaq had already acquired a wide range of the numerous affixes of the
language. He states that the sample revealed 24 derivational affixes, 40
grammatical inflections (25 verbal and 15 nominal), and three enclitics.
Most importantly, he states that they were productive. By this he means,
‘All of these occur attached to more than one stem or on stems elsewhere
appearing with other productive affixes or zero following them’ (p. 103).
He is cautious in this interpretation and recognizes that some holistic learn-
ing may also be taking place. Also, he acknowledges the limitation of the
sample size used. Overall, however, his data are very suggestive that
grammatical morphemes are used from the earliest stages of the acquisition
of Eskimo. Furthermore, they are supported by other data reported on
synthetic languages, e.g. Solberg (1978) on Quechua, Pye (1983a) on
Quiche.
Hindi Varma (1979) reports on the spontaneous language of his daughter

Chopti, between the ages of 1;4 and 1;10. The child’s MLU was 1.05 at the
beginning of the observations and 3.43 at the end. This particular article
focusses on the child’s Stage I speech (in the sense of Brown’s stages). This
period lasted approximately 13 weeks, with the MLU in the last session at
1.88.
Varma states (p. 167): ‘One of the most interesting aspects is the rapid
development of verbal endings in Chopti’s speech.’ During Stage I, the
child showed evidence of using the imperative, present progressive, past,
and future suffixes. Also, the copula appeared during this period and ‘is well
established and used correctly by the end of Stage I’ (p. 168). One example
shows the copula being used along with the present progressive. With
nouns, Chopti uses some, but by no means all, of the possible postpositions
of Hindi. The earliest one acquired was the form ‘ka’ which marks
possessives. Of particular interest is the observation that Chopti did not use
noun combinations of the kinds reported for English children to express
semantic relations.
Varma’s general conclusion is that Chopti is acquiring Hindi by using
single inflected lexical categories rather than combinations of lexical
categories. That is, she is more concerned with acquiring function words
than semantic relations. The following summarizes his findings (p. 170): ‘it
is clear that Chopti’s MLU in Stage I increases generally as functors are
added, and not by concatenating relations as one would expect from
previous work on this stage’.
Quiche I have already discussed the research by Pye on QuichC in

Chapter 5 in relation to linguistic input, and in Chapter 6 in relation to
phonological acquisition. Here the patterns of its early grammatical acqui-
sition will be considered.
Pye (1983a) reports on the first morpheme combinations of four Quiche
children between 2;O and 3;O. Naturalistic samples were collected for an
hour every two weeks for approximately a nine-month period. Though a
498 SIMPLE SENTENCES: ACQUISlTlON OF GRAMMATICAL MORPHEMES
little older than comparable English children, they were at the onset of
multi-morphemic utterances at the beginning of the study. Recall from
our earlier discussion that QuichC mothers are not as verbal with their
infants as English ones, leading to a later onset of language (see Pye
1986).
Like Greenlandic Eskimo, QuichC is polysynthetic in that the verb takes
several prefixes and suffixes. The prefixes indicate person and aspect, while
the suffixes mark transitivity and termination. The termination affixes
constitute a complex set of facts that require some explanation. The choice 3
of particular termination morphemes is influenced by several factors includ-

ing the sentence’s aspect, root morpheme, and transitivity marker. There
are two phonological facts of importance about the termination markers.
One is that they do not always have their own syllable but result in
‘1
resyllabification of the verb. For example, the sentence ‘you see it’ has the
following morphemic structure: k-0-aw-il-oh (where ‘il’ is the verb ‘see’ and
‘oh’ is the termination suffix), but it has the following syllabic structure:
ka-wi-loh’. That is, the syllable which has the termination suffix begins with
the final consonant of the verbal root. A second feature is that the stress
tends to fall onto the final syllable. In the above example it is on the syllable
‘loh’.
Pye discusses QuichC as a good test case of semantic versus distributional
theories of acquisition, although he doesn’t use these particular terms. He
states (p. 591) ‘. . . there are two possible parts of the verb with which
Quiche children might begin: the semantically salient but perceptually weak
verb root, and the perceptually salient but semantically complex verb
suffixes’. His analysis indicates that the children consistently chose percep-
tual saliency over semantic complexity. From their first appearance, the
verbal suffixes were used correctly and apparently productively (we say
‘apparently’ because no strict productivity measure was applied). He makes
the following comment about the children’s use of the verbal suffixes
(p. 593):
There is also no indication that the children had any trouble distin-
guishing transitive from intransitive endings. Finally, the vowel in the
transitivizing suffix is lexically determined - it follows no regular
pattern. The children, however, seldom got the vowel wrong in the
utterances. All this indicates that the children had indeed learned to
use the status markers freely and appropriately at an early stage in
their linguistic development.
This pattern of acquisition is supported by similar results reported for
Dakota in Nokony (1978).
9.6 The explanation of morphological acquisition

Much of the theoretical discussion on morphological acquisition has been
done by identifying the factors which seem to be at work. A theory of such
acquisition, however, will require a more detailed model of how the child
identifies and processes grammatical morphemes. There have been two
major attempts to provide such a model: MacWhinney (1978) and Pinker
(1984).
MacWhinney (1978) This monograph attempts to develop a psychologi-

cally real model of morphological acquisition. We have already seen in
Chapter 7 MacWhinney’s six strategies of early word combinations (section
7.4.2). The 1978 monograph is an earlier attempt to focus on morphological
development, and refers to three of the six strategies: rote, analogy, and
combination.
MacWhinney refers to his approach as the dialectic model. It consists of
three major aspects: application, correction, and acquisition, which are
linked in a cyclic fashion, as in (9.28):
(9.28)
Application
(rote, analogy, combination)
Acquisition Correction
(mismatches)
These three aspects will operate in the child’s perception of language as well
as his production.
Application refers to the child’s initial step in the process of using and/or
acquiring grammatical morphemes. In reception, the goal is to understand
the semantic/syntactic meaning of morphemes, while in production it is to
retrieve and use them. In either modality, the child uses the three strategies
of rote, analogy, and combination in that order. For example, a word like
‘cats’ may initially be stored as a holistic unit. Next, it will be separated into
two parts, ‘cat’ and ‘s’,but the use of ‘s’will be restricted by analogy to a few
similar words, e.g. ‘bat’ and ‘mat’. Later still, combination will result
through the use of a general rule.
A crucial feature of following a child’s application is how he proceeds
from one strategy to another. These changes are apparently the result of the
child being able to segment morphemes and place them into one of three
categories: (1) words, (2) affixes, and (3) roots. At an early stage, for
example, a word like ‘cats’ may be placed in the word category with no
further segmentation. Later, ‘cat’ will be put in the root category and ‘s’ in
the affix category. During acquisition, specific kinds of potential errors
may be made, for example, having words in different categories: the child
could have ‘cats’ as a whole item as a word (category l), and also as a root
‘cat’ (category 2). This could result in the error of taking ‘cats’ from 1 and
+
adding the plural from 2, creating ‘cats s’. To avoid this, MacWhinney
proposes that the child has a process of affiw checking which restricts the
child from creating a meaning for a new category when that meaning
already exists in another category. Notice that this will restrict the child
from producing ‘feets’ if ‘feet’ already exists as a word.
As the child identifies roots and affixes, he will also need to deal with
the existence of words with different roots, e.g. ‘wife’, ‘wives’, and also the
allomorphs of grammatical morphemes, e.g. ‘cats’ [SI vs. ‘dogs’ [z]. The
major problem here is one which arises in linguistic theory in general -
what is the nature of the child’s morphological system? There are some
rules of combination which are phonetically motivated, such as the English
plural, and others which are not. An example of the latter would be
gender in languages such as French and German where the child needs to
select the correct form of some grammatical class to go with the gender of
the noun being used. MacWhinney does not provide a linguistic theory as
a framework, but falls back on the three strategies mentioned above - that
is, the child progressively works toward the general rules of the language
being acquired, whatever those rules may be.
The application phase of MacWhinney’s model only provides the child
with the ability to segment words into classes. This is not sufficient,
however, to acquire rules or avoid potential errors. For example, the child
still can select words like ‘foot’ and add ‘s’ to get ‘foots’, a well-documen-
ted kind of error. Application does not have the necessary mechanisms for
complete acquisition and correction. For that to occur, the next two
phases must operate.
The first step is referred to as correction. This step occurs when some
mismatch takes place in the perception or production of a morphological
combination. This state is somewhat like Piaget’s idea of disequilibrium, in
the sense that the child’s system may not conform to that of the input.
MacWhinney’s notion of mismatch, however, is broader because it also
refers to mismatches within the child’s system itself. Such mismatches
draw the child’s attention to the fact that there is a potential error in his
system. This can be seen by looking at the kinds of mismatches which can
occur.
MacWhinney discusses four types of mismatches (or what he calls dise-
quilibrated pairs), listed below:
Type I : Self corrections. The child overgeneralizes a rule when the
correct rule is available within the system, e.g. child says ‘wifes’, and
then ‘wives’;
Type 2: Semantic mismatches in production. The child expresses an
incomplete meaning, e.g. saying ‘dog’ to mean ‘dogs’;
Type 3: Semantic mismatches in reception. The child hears something
said which does not communicate the complete meaning intended,
e.g. hearing ‘dogs’ and only recognizing ‘dog’, yet seeing two dogs in
the context;
Type 4: Auto-instruction. A productive combination becomes a role
item, and elicits adults’ corrections, e.g. saying ‘foots’ and being cor-
rected to say ‘feet’.
MacWhinney is not completely clear how these kinds of mismatches
lead to correct acquisition, nor what mechanisms underlie each. Type 1
assumes some kind of self-monitoring of production in relation to the
current competence. This mechanism, however, could have negative
effects if the child’s competence is incorrect for it would wrongly mark an
incorrect production as correct. Type 4 brings direct positive evidence into
the process, although MacWhinney acknowledges that the child will resist
such corrections.
The real key to the child’s ability to correct itself lies in types 2 and 3.
These direct the child to seek out meanings for the speech it hears and
uses. Unfortunately, it is not clear how this takes place. The problem with
type 2 is delimiting how the child knows that what it wants to express exists
within the language being acquired. Type 3 is clearer to follow. It implies a
mechanism which leads the child to identify repeated stretches of speech
and to seek a meaning for them. No information is given, however, on
how this segmentation and identification takes place.
If the child uses a form and no mismatch results, then the form can
undergo acquisition. Acquisition here is used in a way roughly similar to
the notion of ‘strengthening’ that was discussed in Chapter 7 for syntax. As
acquisition of any form proceeds, the three strategies of rote, analogy, and
combination will compete for the way the form will be stored. In par-
ticular, rote will constantly attempt to override analogy and combination.
MacWhinney lays out five potential cycles that forms can follow in being
applied, corrected and acquired, summarized below:
Cycle 1: Amalgams. The child looks for forms to express the

meanings of its semantic system. When a form is heard which
appears to express that meaning, the child acquires it as an unanalyzed

whole.
Cycle 2: Allomorphs. The child identifies the meaning of an affix and
finds that there is another affix with the same meaning. They are
flagged as being allomorphs. As they are used, correct usages
strengthen the rule being acquired, and mismatches weaken it.
Cycle 3: Mod$cation productions. The child identifies that a mis-
match has occurred in his use of some allomorph. He compares his
form with the correct one and reformulates the rule used to create it.
Cycle 4: Selection productions. The child determines that a particular
mismatch resulted because the child generalized a minor pattern. The
rule is then flagged as a minor rule.
Cycle 5: Zmplicational table. When general or minor rules are not
adequate to capture the morphological patterns in the language, the
child resorts to lexical principles, that is, he memorizes that certain
morphemes are used with certain classes. This can be captured
through the use of an implicational table.
Cycle 1 occurs in response to type 2 and 3 mismatches. For example, the
child may wish to express ‘dogs’, but may only have ‘dog’. If he should hear
‘dogs’, it gets identified as a single form which expresses the plural of ‘dog’
and is stored as such. Cycle 2 involves the identification of allomorphs. The
next two cycles concern how the child forms a rule to use these allomorphs.
To understand these, we need to distinguish between general and minor
rules. A general rule is one which will apply to most forms which have the
required phonological shape. For example, the use of plural [SI after
certain voiceless consonants is a general rule. A minor rule is one which
only applies for a small class of words in the language. For example, the
change of ‘wife’ [waif] to ‘wive’ [waiv] in the plural is a minor rule in that
words with final ‘f’ will not all normally change in this fashion. Cycles 3 and
4 allow the child to determine through the examination of mismatches
whether a rule is general or minor.
The last cycle concerns what happens for those cases where the morpho-
logical selection is predominantly lexical. MacWhinney proposes that the
child sets up implicational tables of the relevant morphemes. He exempli-
fies this process with the articles in German. Example (9.29) gives an
adaptation of MacWhinney’s implicational table for German singular
definite articles, based on his Table 2 (p. 14).
(9.29) Roots Class Nominative Dative Accusative
Mann, etc. 1 ‘der’ ‘dem’ ‘den’
Frau, etc. 2 ‘die’ ‘der’ ‘die’
Pferd, etc. 3 ‘das’ ‘dem’ ‘das’
MacWhinney concludes his general discussion of his model with a section
on the major determinants of the child’s order of acquisition. His claims in
this regard are very straightforward: the sequence of acquisition of the first
amalgams ‘will be determined largely by the functional importance . . . of
the amalgams within which the affixes are embedded . . . as well as the
intonational packaging of these units’ (p. 16). Their subsequent productive
acquisition is determined primarily by their functional importance. H e
concludes (p. 17): ‘Nowhere in the model is there a role for either semantic
or grammatical complexity per se. Aather, the emphasis is on functionality
as a determinant of morpheme acquisition.’ The rest of his monograph is a
demonstration of the application of his model to data from Hungarian and
Finnish.
Despite a lack of specifics at several points, MacWhinney deals with the
major issues required for a model of morphological acquisition. There are
three particular aspects needing further development. First, he is not
working from any particular theory of the adult system: from our earlier
discussion of his competition model, it was clear that he is more disposed to
develop processing models than to test current linguistic theories. Another
aspect is what we will call the ‘segmentation’ problem: he points out that the
child needs to identify the morphemes, but says relatively little about how
that occurs. Lastly, there is the ‘identification’ problem: how exactly does
the child determine the appropriate grammatical/semantic meaning which
underlies the specific morpheme being acquired?
Pinker (1984) In Chapter 7 Pinker’s views on early syntactic development

were discussed. In its general orientation, his approach to morphological
development shares the same features. H e outlines a series of specific steps
that the child follows in reaching the adult system. These steps allow the
child to formulate some incorrect rules, but limit the range of incorrect rules
which are possible. This is done in detail, presenting arguments and
examples along the way.
Pinker deals with each of the three areas just mentioned for MacWhin-
ney’s model. As we noted earlier, he works from a particular theory of
grammar, Lexical Functional Grammar. In doing so, he also develops the
theory to make it more compatible with the facts of language acquisition.
He also addresses the segmentation problem, although he does not offer a
detailed account. He postpones this ‘for more precise notions to be taken
from a theory of phonology’ (p. 188). The identification problem, however,
occupies a central place in Pinker’s treatment of morphological acquisition,
and demonstrates the manner in which he approaches the topic.
Pinker outlines three specific proposals for handling the identification
problem, eventually rejecting the first two for the third. The process, in fact,
appears to be a historical account of Pinker’s own stages of thought on the
subject .
The first proposal is called exhaustive hypothesization and suggests that
the child first attaches a universal set of possible grammatical meanings to a
newly identified morpheme. The child then compares each possible
meaning with the context and rejects those meanings which are not
compatible. By this process of elimination, the child eventually narrows
down the meaning to the correct set. Pinker correctly points out three
serious problems for this proposal. First, it requires an enormous amount of
calculation: assuming the list of possible meanings can be quite long, the
child would be faced with an extensive range of possible meanings to be
checked out. Second, the proposal makes the false prediction that mor-
phemes with several grammatical meanings (syncretic ones) will be
acquired before ones with just one or two, since there would be fewer to be
eliminated. Facts from acquisition, however, indicate that the opposite is
the case. Third, the proposal makes false predictions about overgenerali-
zations, since it predicts that morphemes should be first undergeneralized.
In particular, the child should not overgeneralize a morpheme to a false
context if it is initially flagged with a range of appropriate possible
meanings. Such kinds of overgeneralizations, however, do occur.
The second proposal is one of hypothesis sampling. In this approach, the
child does not assign all possible meanings to a morpheme, but rather just
one at a time. Further, the list of possible meanings is weighted so that more
likely ones are tested first. For example, a language is more likely to have a
verbal affix which indicates aspect than one which indicates the definiteness
of the object, as is found in Hungarian. Pinker presents seven arguments
against the viability of the hypothesis-testing proposals, only a few of which
can be examined here. First, it doesn’t account for how zero morphemes are
acquired. It also ‘makes the implausible assumption that the child sooner or
later hypothesizes out of thin air every feature equation that languages in
the world can use’ (p. 173). Pinker feels that the model needs some
mechanism that actually draws the child’s attention to particularly unusual
meanings. Another difficulty is that the proposal does not account for
morphological imperialism (Slobin 1973), when the child uses one mor-
pheme based on one feature (e.g. nominative case) and then abandons it for
another morpheme based on a different feature (e.g. gender). Indeed, the
problems that Pinker points out for this proposal are also problems for the
earlier extensive hypothesization proposal and are important questions that
all accounts of morphological development need to address.
This brings us to Pinker’s own account of morphological acquisition. The
approach actually is rather similar to that already discussed by MacWhin-
ney. In particular, it builds upon the notion that the child constructs
implicational tables of grammatical morphemes. The latter are called ‘para-
digms’ by Pinker, and the idea is developed much more extensively. As in
Chapter 7 Pinker’s approach will be outlined by assigning a series of ‘steps’
to the acquisition process, even though he does not quite present it in this
fashion. The numbers for his procedures are indicated within parentheses.
While Pinker discusses several acquisition procedures, his approach will be
simplified here to two major steps.
Step 1: Establish word-speci)’ic paradigms (11, 12). The child begins
acquiring grammatical morphemes by acquiring the stem and the affix as a
complete unit. This is similar to MacWhinney’s first step of acquiring amal-
gams and assigning them to the class of words. The child at the onset has not
segmented the affix from the stem. The difference from MacWhinney is that
the child at the onset sets up paradigms for these. Example (9.30) gives one
such paradigm for the verb ‘walk’.
(9.30) Progressive Present Past
walking walks walked
This presumably also can occur for independent morphemes such as the
German determiners in (9.29). At first, the paradigms are given a single
dimension, but new dimensions can be added as new entries are added (12).
The advantage that this step has over MacWhinney is that it allows the child
to begin morphological identification before the affixes are actually deter-
mined. It creates a system, therefore, for them to be placed in once they are
segmented. A further point about this step concerns the criteria used for
establishing paradigms. Pinker allows for the possibility of using both sem-
antic and phonological features.
Step 2: Form general paradigms and word structure templates (1.5, 16). In
this step the child abstracts away from the individual word paradigms and
creates general paradigms in which individual grammatical morphemes are
entered. This requires the child to scan the words in the word-specific para-
digms and separate out the stems and affixes (roughly MacWhinney’s cycle
2). Further, as required by the grammatical model that Pinker is using, a
word structure template is set up (16). Example (9.31) shows this for ‘walk’.
(9.31) Class Progressive Present Past
‘walk’, etc. ing S ed
+
Word structure template: [”stem affix]
While these steps appear straightforward enough, there are numerous
specific problems which the child will face, including syncretic morphemes,
the acquisition of class differences, and paradigm splitting (the latter is the
situation where a morpheme will fill more than one place in a paradigm).
Pinker addresses all of these and offers procedures for accomplishing them.
506 SIMPLE SENTENCES : AC Q UISITION O F GRAMMATICAL M O R P H E M E S
Summary. The research on morphological development has resulted in a

number of observations on patterns of development. The models by Mac-
Whinney and Pinker are the most extensive attempts to date to explain in
more detail how these patterns emerge. Together they. have identified
much of what has to be accounted for. The model needs to be able to
segment morphemes, identify grammatical meanings and allomorphs, and
group morphemes into their appropriate patterns of arrangement. Both
approaches also share the feature of trying to develop a process model of
acquisition. A very interesting extension of some of MacWhinney’s origi-
nal ideas can be found in Rumelhart & McClelland (1986), where an
account of English verbal endings is given.
9.7 Linguistic input and grammatical acquisition

We last discussed the role of the linguistic environment in Chapter 6,
section 6.6, where we presented some information on early conversational
development, methodology, and effects on lexical development. Here we
turn to the issue of the role of the linguistic environment in the young
child’s grammatical development.
Throughout this text we have addressed the issue of the interplay of the
innate ability of the child and the effects of the environment. Gleitman,
Newport & Gleitman (1984) have summed up very nicely where the
controversy does and does not occur on this issue. They point out that
there is not much disagreement that ‘the way caretakers talk, and the
circumstances under which they talk, affect learning’ (p. 44). The con-
troversy resides in the extent to which the internal state of the child
intervenes and influences the issue. In this sense, the positions tend to fall
upon some point of a graded view of the influence of the linguistic
environment.
We can examine this issue in the context of the three general theories we
have considered throughout. The behaviorist wants to demonstrate a direct
relation at all points of acquisition, so that the child’s behavior can always
be traced to adult variables. The nativist positions we have discussed will
vary on this point. A maturationist will minimize the influence of the
environment: if a principle of grammar has not yet matured, then no
amount of linguistic input will lead to its acquisition; if it has matured, then
presumably some minimal exposure will be sufficient. The position of
strong inclusion proposes that the child’s readiness is set very early in
acquisition so that only a sufficient range of constructions need to be heard.
Lastly, a constructionist perspective falls between the two positions: readi-
ness will be a factor, and thus the child will not acquire a form until the
child’s system is ready for it, but at the same time, the development of a
structure will involve a set of interactions between the child’s internal
system and the linguistic environment.
A look at the literature on this topic in recent years suggests that most of
it has sought to demonstrate environmental effects of some kind. As such, it
has had a constructionist and even behaviorist flavor to it. The earlier
research was primarily interested in demonstrating that at least English
parents adjust or reduce their speech to children. Later, the obvious
question arose as to whether or not such adjustments contribute to
acquisition or not. It is this latter question which has dominated most recent
work.
Two distinct research programs into the effects of the linguistic environ-
ment on English grammatical acquisition will be considered here. The first
is the research of Lila Gleitman and her colleagues. Gleitman’s perspective
is primarily maturational in nature, as can be seen in Gleitman (1981). Her
position is that the course of acquisition is more determined by the child’s
internal state than by the language environment but that the latter does play
a role in the acquisition of language-specific grammatical morphemes. The
other approach is that of Keith Nelson and his colleagues. (References to
his work are indicated as K. E. Nelson to distinguish him from Katherine
Nelson.) K. E. Nelson takes what I interpret as a more constructionist
perspective. He sees acquisition as the interplay between facilitative adult
behavior and the child’s internal readiness.
Gleitman, Newport & Gleitman (1984) In the middle 1970s, researches

were very actively describing the attributes of the speech which parents direct
to their children. This literature often implied that these adjustments were
facilitative and even necessary for the child’s linguistic development. This
research was set back, however, by an influential study reported in
Newport, Gleitman & Gleitman (1977; henceforth NGG). NGG compared
the speech of children and their parents and found very few instances where
the parent’s behavior seemed to influence the child’s subsequent develop-
ment. They did suggest, however, that such influences occur with the
English auxiliary as it appears in yes/no questions. This led them to a
position that the effects of the linguistic environment are only found with
the more language-specific morphemes.
The NGG findings resulted in a new caution in drawing causal impli-
cations between adult linguistic behavior and language acquisition. Their
findings were also aided by the cross-linguistic research which was begin-
ning to appear and which indicated that some cultures did not have adult
speech adjustments. (The latter literature was briefly reviewed in Chapter 5
of this volume.) Since then, adult speech adjustment to children has been
seen as at most potentially facilitative, but by no means necessary and
sufficient. They also directed research toward more detailed work on what
might be facilitative and what might not.
The results in NGG were challenged by Furrow, Nelson & Benedict
(1979; henceforth FNB). FNB criticized a number of methodological
aspects of NGG, and reported their own results which indicated more adult
influences than found by NGG. As discussed in Chapter 6, they looked at
correlations of the mother’s speech at Time 1 (when the children were
around 1;6) and Time 2 (when the children were around 2,3), with the
children’s speech at Time 2 (around 2;3). Importantly, they found that
the correlations at Time 2 could not predict the correlations found at Time
1. They conclude from this that the possible correlations will be affected by
the child’s stage of development. They looked at correlations between
several measures of adult language and four measures of child behavior:
MLU, number of verbs per utterance, number of noun phrases per
utterance, and number of auxiliaries per verb phrase. Some of the sig-
nificant correlations they found between adult speech at Time 1 and the
children’s speech at Time 2 are summarized here:
Facilitative aspects
(a) Greater adult use of yeslno questions leads to:
i. more child use of auxiliaries,
ii. greater child MLU.
(b) Greater adult use of nouns leads to:
i. more child use of verbs,
ii. greater child MLU.
(c) Greater adult use of interjections leads to:
i. more child verbs per utterance.
Prohibitive aspects
(d) Greater adult use of pronouns leads to:
i. less child use of verbs,
ii. shorter child MLU.
(e) Greater adult use of different words leads to:
i. shorter child MLU,
ii. fewer verbs and noun phrases per utterance.
(f) Greater adult use of verbs and copulas leads to:
i. less child use of verbs,
ii. shorter child MLU.
FNB interpret their results as supporting the position that simpler adult
speech aids children in their acquisition. They manage this by offering the
following scenario. Simple speech will be characterized by salient aspects of
the grammar (such as fronted Auxs as in yesho questions), simple semantic
messages (such as having more nouns than verbs), and pauses which shorten
the sentence (interjections). These aspects all proved to lead to more
advanced child measures. Complex speech, on the other hand, will have
more pronouns, complex vocabulary, verbs, and copulas. Greater use of
these led to poorer child performance on the measures used.
The FNB results were a strong challenge to the earlier results in NGG.
The latter responded in great detail in Gleitman, Newport & Gleitman
(1984). First, they accepted the methodological criticisms of FNB and
provided an in-depth discussion of methodology. This was followed by a
reanalysis of their original data, as well as the data in FNB. Further, they
imposed a much stricter statistical measure of correlation to eliminate what
they refer to as ‘spurious correlations’, i.e. correlations which are found but
which seem to make little sense. Their results after this elaborate procedure
ended up being quite similar to those in NGG. All of the significant
correlations which they found are given below:
For younger children (18 to 21 months)
(a) Greater adult use of declaratives leads to more verbs per
utterance.
(b) Greater adult repetition leads to fewer child auxiliaries.
(c) Greater adult use of complex sentences leads to more child aux-
iliaries.
(d) Adult unintelligible speech leads to more child verbs (interpreted
as a spurious correlation).
For older children (24 to 27 months)
(a) Greater adult use of yes/no questions leads to more child verbs
and greater MLU.
Gleitman, Newport & Gleitman conclude from their study that there are
fewer effects than indicated in FNB once better statistical procedures are
used. Also, if such effects exist, they tend to be restricted to the younger
group of children - this partially supports the claim of FNB that such effects
may differ depending on the language level of the child. They emphasize,
however, that their reanalysis provides little support for most of the
correlations found in FNB, and thus for the FNB theory concerning the role
of the linguistic input. Lastly, they emphasize that the earlier finding about
the use of yes/no questions was replicated even with the revised statistical
procedure.
The finding about yes/no questions is important because it is at the core of
the Gleitman, Newport & Gleitman view of language acquisition. They
propose that the child acquires language most easily when the linguistic
input provides a range of grammatical constructions. They argue on logical
grounds against the FNB view about simple speech. They state (p. 69):
510 SIMPLE SENTENCES: ACQlJISITION OF GRAMMATICAL MORPHEMES
. . . it is relatively easy to show that the language is learnable if the
input includes complex sentences; it is awesomely harder to show
learnability if the input is restricted to the simplest sentences. This
position should not really come as a surprise.
-. .-
This point is supported by their findings that they did not get the corre-
lations in their data which are predicted by FNB. The fact that the yesho
questions in adult speech lead to more child auxiliaries is because yesho
questions place auxiliaries in a very prominent place in the sentence. It is
because of the prominence of the Aux that the effect occurs.
Gleitman, Newport & Gleitman (as well as NGG) go on from auxiliaries
to language in general. Auxiliaries are one member of a general class they
refer to as the closed class. They do not give a formal definition of these, but
say (p. 71) ‘Closed-class items, roughly, are the inflections and functors,
those items that can occur unstressed in the languages of the world.’ It is
because of their language-specific, and unstressed nature that these items
tend to be acquired later. They propose that adult speech can contribute to
the acquisition of these items if it places them in some more prominent
position for acquisition. This is the case for English yesho questions.
The studies by Gleitman er al. and by FNB serve to highlight two distinct
views on the role of adult speech in the child’s acquisition. While FNB want
to assign importance to simplified speech, Gleitman et al. want to play down
its importance, and emphasize instead the importance of hearing a range of
constructions. In the latter view, the adult speech can contribute if the
constructions used contain closed class items in prominent positions.
Nelson & Baker (1984) The studies just reviewed, as well as many of the
others on this topic, are based on the analysis of spontaneous language.
Such data are always difficult to assess, due to problems with sampling and
sample sizes. (Gleitman, Newport & Gleitman 1984, for example, criticize
FNB for using only 100 utterances per mother and child, while they used an
average of 513 per mother and up to 300 per child.) It is possible, however,
to examine this question from a more experimental perspective. Such a line
of research, in fact, has been pursued in research years in a series of studies
by K. E. Nelson and his colleagues.
The theoretical perspective which underlies Nelson’s experimental work
can best be seen in two review papers, K. E. Nelson (1981) and K. E.
Nelson, Denninger, Bonvillian, Kaplan & Baker (1984). (We refer to
Nelson here but assume his perspective has been developed with and is
shared by his several co-authors.) Nelson refers to this theory as a rare-event
comparison theory of acquisition (K. E. Nelson 1981). The theory accepts
the cross-linguistic research of recent years which indicates that many
features of parental speech adjustment are not necessary for children’s
grammatical development. He emphasizes, however, that such findings, as
well as those in NGG, only indicate that ‘high quantities’ of linguistic input
may not be needed. Nelson’s view on frequency is that there is not a directly
linear relation between adult frequency and acquisition rate (as is claimed in
FNB, for example). Rather, Nelson et al. state (1984:49): ‘in our view,
whether more frequently presented examples in input are more rapidly
acquired by the child than less frequent examples in input depends entirely
upon how those examples are embedded in the fabric of conversation’.
In Nelson’s approach, there are two additional factors besides frequency
which need to be considered. One is that the child has to be ready to acquire
the form under consideration. If the child is linguistically at a point where a
particular form can be acquired, the frequency of exposure to it can be quite
minimal. The other factor is the way in which the parent presents the form
to the child. Nelson has argued that there are two facilitative responses that
parents can make. One is what he calls a simple recast, a reply by the adult
which continues ‘reference to the central meanings in the child’s preceding
utterances’ but which departs ‘structurally by slight or moderate changes’
(p. 37). Such changes are limited to one major component of a child’s
utterance, e.g. the subject, verb, or object. For example, the child
utterance ‘broke’ can be recast as ‘the truck broke’. The other facilitative
feature is that the recast be a continuation of the discourse. A continuation
is an adult response which does not meet the definition of a recast but which
continues the topic under discussion. The claim is that constructions are
acquired more rapidly when the child hears them in the form of recasts and
continuations.
Nelson’s theory has been experimentally tested in a series of studies, e.g.
K. E. Nelson (1977), K. E. Nelson & Baker (1984), K. E. Nelson et al.
(1984). The general procedure is to pretest children on a construction (or
constructions) to be sure that it is not yet acquired. Next, there is an
experimental period from four to eight weeks when an experimenter meets
with the child on a regular basis and recasts the child’s speech, using the
construction(s) under study. At the same time, there is a control group
which receives no such intentional recasting. Finally there is a post-test to
see if the construction has been acquired. The results reported have in
general supported the theory.
We will present this experimental approach in more detail by focussing on
Study 2 in K. E. Nelson & Baker (1984) on the acquisition of passives,
relative clauses, and auxiliaries. The subjects were six children (four girls
and two boys) between 2;6 and 3;2 with MLUs from 3.07 to 4.17. In the
month before the study began, spontaneous language samples were col-
lected from conversations between the children and their mothers. The
Table 9.20 Number of spontaneous productions for six normal children of
passives, relative clauses, and ‘may’ and ‘could’, taken from Nelson &
Baker (1984: Tables 5 and 6)
Session Passives Relative clauses ‘may’ and ‘could’
1 0 0 1
2 0 0 2
3 1 0 -
4 2 2 2
I 3 0 1
9 1 4 1
11 2 1 7
12 7 3 5
Total 16 10 19
children were then placed into three pairs based on similarities on three
measures: MLU, noun and verb complexity, and use of complex forms.
Each was randomly assigned to two treatment groups with each member of
a pair assigned to a different group. One group (C/E) heard their sentences
recast while the other did not (E/E). The E/E group, however, did hear the
experimenter recast her own utterances.
The syntactic forms tested were passives, relative clauses, and two
auxiliaries, ‘may’ and ‘could’. Although data are not given regarding the
pretest samples, it is implied that none of these forms was used at that point.
The experimental phase of the study consisted of four weekly sessions for
three weeks, or 12 sessions. The first sessions consisted of 15 minutes of
recasting and five minutes of observations. The last session was for
post-testing. An additional language sample was subsequently collected in
each child’s home. Proportions of the use of each syntactic form were
calculated after each session to ensure that the number of recasts was
comparable for both groups.
Table 9.20 presents a summary of the spontaneous use of the syntactic
forms tested across a selection of tape-recorded sessions. The results for
both groups have been collapsed although there was a tendency for the C/E
group to produce more forms than the E/E group. The main effect is that
the spontaneous use of these forms increased over the test sessions. For
example, none of the children produced a passive sentence until the third
session, but 16 were produced by the end. Also, full passives (eight) were
nearly as frequent as truncated passives (nine). The results indicate that the
use of these recasts by the experimenter can lead to increased spontaneous
use in the child.
To support their conclusions, Nelson & Baker also compared their results
Further reading 513
to the data collected in Horgan (1978), whose study was summarized in
section 9.4.1. They calculated the proportion of different types of passives
used and compared figures with those in Horgan for the children aged
between 2 and 4 years. These results are summarized in (9.32):
(9.32) Type of passive Horgan Nelson & Baker

Reversible 47 % 63Yo
Agentive non-reversible 0% 25 %
Horgan found that agentive non-reversible passives were not used until age
9. Also, none of the children she studied produced both reversible and
non-reversible passives until age 11. The Nelson & Baker subjects had
higher percentages of both of these types, and one-half of their subjects
produced both reversible and non-reversible passives.
Further reading
English morphological acquisition
This area has generated a massive literature. In particular, the classic study
by Berko (1958) has been replicated, with various modifications, literally
hundreds of times. It remains a valuable article which is probably required
reading in just about everyone’s course on child language. Much of the cited
literature is the work of Brown and his colleagues. The study by Cazden
(1968) presents an important early analysis of the Brown data which was the
basis for the subsequent report in Brown (1973). Brown (1973) is an
important but very long discussion of the acquisition of 14 English mor-
phemes. A feel for the data and the results may be obtained from reading
pp. 249-73.
Aspects of English grammar

While many aspects of English have been studied from an acquisition
standpoint, the few areas reviewed in the text have received most attention.
Aux. The acquisition of auxiliaries in questions was first studied in detail
by Bellugi in her unpublished doctoral dissertation (1967). Her original
observations have led to a range of studies trying to replicate the pattern
and explain its causes. An excellent review and theoretical treatment of the
topic is Davis (1987).
Reading in this area should begin with looking at Bellugi’s original stages
of question acquisition. This can be gotten from Bellugi (1971), or Cazden
(1970) which is one of the clearest summaries ever written on the topic. The
study by Kuczaj & Maratsos (1975) presents an important methodology and
lays out the relevant issues at stake. Ingram & Tyack (1979) and Labov &
Labov (1978) provide grouped and single subject data respectively that
suggest that some caution is necessary in assuming the generality of the
stages proposed by Bellugi. A feel for the theoretical implications of aux-
iliary acquisition can be gotten from Brown (1968) and from Erreich,
Valian & Winzemer (1980).
Passive. The passive is another pattern which has been studied numerous
times, initiated in the late 1960sby studies by Turner & Rommetveit (1967),
Slobin (1968), Huttenlocher, Eisenberg & Strauss (1968). Other areas such
as relative clauses and pronominal reference have only been studied in more
recent years as syntactic research has expanded, despite the early research
by Carol Chomsky (1969). Pronominal reference especially has received
much recent discussion, e.g. Solan (1983), Lust (1986).
The first findings on passives were reported in Turner & Rommetveit
(1967) and Slobin (1968). Bever (1970: 303-12) is afrequently cited report of
when children begin to use strategies to cope with the passive construction.
Horgan (1978) gives a wealth of data on the use of passives in spontaneous
speech. The diverse theoretical interpretations on passives can be seen by
comparing the discussion in Maratsos et ai. (1979) with the nativist perspec-
tives in Berwick & Weinberg (1984: 212-22) and Borer & Wexler (1987).
Relative clauses. The first observations on relative clauses were made by
Limber (1973). The experimental work by Sheldon (1974) has inspired a
number of subsequent studies, such as Tavakolian (1981b). The develop-
ment viewpoint in Ingram (1975) can be contrasted with the nativist one in
Hamburger & Crain (1982).
Pronominal reference. Reading here should begin with Carol Chomsky’s
original study (1969: 102-11). Conveniently, Carden (1986) provides an
excellent review of much of the recent literature, as well as his own inter-
pretation of the results. A recent two-volume work on the topic is Lust
(1986). We also recommend looking at the cross-linguistic data presented
in O’Grady, Suzuki-Wei & Cho (1986).
Cross-linguistic morphological acquisition

Slobin’s bibliography (1972) gives separate indices for cross-linguistic
research up to that time. References are given for African, Algonquin,
Amerindian, Arabic, Bantu, Bulgarian, Burmese, Chinese, Comanche,
Czech, Danish, Dutch, English, Estonian, Finnish, French, Garo,
German, Gilyak, Greek, Hebrew, Hopi, Hungarian, Icelandic, Iroquois,
Italian, Japanese, Korean, Latvian, Malay, Marathi, Navaho, Norwegian,
Pidgin, Polish, Polynesian, Portuguese, Romanian, Russian, Ruthenian,
Samoan, Serbian, Slovenian, Spanish, Swedish, Ukrainian, Welsh, Wolof,
and Yiddish. His more recent edited work (1986) provides more up to date
information on a smaller selection of languages.
Further reading 515
Recent projects by Berman (e.g. Berman 1981a, b, 1982) and Weist (e.g.
Weist 1983; Weist et al. 1984; Weist & Konieczna 1985) provide insightful
new information on Hebrew and Polish respectively. Slobin (1982) gives a
nice review of his own research comparing English, Italian, Serbo-Croatian,
and Turkish. A comparison of English, Hungarian, and Italian can be found
in MacWhinney & Bates (1978).
It is insightful to read one or two descriptive papers on the acquisition of a
language which is very different from English to see how the pattern of
acquisition can vary. A reading of any of the articles reviewed in the text can
provide this, as well as numerous others. We feel that Pye (1983a) does this
very well, in addition to Slobin (1982) and MacWhinney & Bates (1978),
mentioned above. An excellent assessment of the implications of these
kinds of data is Bowerman (1986).
The explanation of morphological acquisition

We looked at just two attempts to develop a processing model for the
acquisition of morphology: MacWhinney (1978) and Pinker (1984). An
overview of MacWhinney’s approach is found in pp. 1-20 of his mono-
graph. Pinker’s discussion is in his Chapter 5, entitled ‘Inflection’. A more
recent work which has generated a great deal of interest is that of
Rumelhart & McClelland (1986).
Linguistic input
There have been a series of debates in the various attempts to demonstrate
causal effects between parental and child speech. The text discussed the one
between Furrow, Nelson & Benedict (1979) and Gleitman, Newport &
Gleitman (1984). Other reports have appeared which differ from these two
in various ways. Barnes et al. reports several correlations between adult and
child measures, but interpret the effects as reciprocal rather than unidirec-
tional. Scarborough & Wyckoff (1986) reanalyze their data and find
virtually no effects, that is, even fewer than Gleitman, Newport &
Gleitman.
A less prominent debate occurred after Moerk (1980). Moerk suggested
that the effects of parental speech may be more subtle than had been
suggested. In particular, he suggested that parents may increase their use of
specific forms for just a short period around the time that the child shows
evidence of acquiring them. This elicited a critique from Pinker (1981), as
well as a response from Moerk (1981).
A helpful review of this literature is that of Hoff-Ginsburg & Shatz
(1982). The article by Gleitman, Newport & Gleitman (1984) is not easy
reading, but it provides an in depth discussion of the relevant issues. Keith
Nelson’s views are probably most accessible in Nelson (1981).
10 Concluding remarks
I began this text with the goal of presenting an in-depth introduction to the
field of child language acquisition. This was to be done by providing a
balance between method, description, and explanation. We would like to
conclude by giving a brief assessment of where we now stand in regard to
these three aspects.
As should be apparent, the field has evolved tremendously in the area of
method, and should continue to do so in the years ahead. It has gone
beyond the simplistic view that a child’s linguistic knowledge can be
adequately assessed by collecting a spontaneous language sample. An
obvious way in which the approach has changed is the development of
diverse experimental methods to get at the many aspects of language that
are not immediately apparent from spontaneous speech. These have
included the use of imitation as in Kuczaj & Maratsos (1975), comprehen-
sion tests as in Thomson & Chapman (1977), and elicitation procedures as
in Berko (1958). Indeed, studies such as these across a range of linguistic
behaviors suggest that seemingly impossible aspects of language can be
studied as long as one brings a creative mind to the task.
While such experimental approaches have held great promise, their
details are still being developed. We have discussed cases where studies
have not been replicated, or where different methods of analysis have been
proposed. The area of environmental effects, for example, is still in the
midst of determining the appropriate methodological procedures needed to
determine causality. Another example is in the area of infant speech
perception where I presented the methodological debate between the
nativists and the developmentalists. So too, we are still attempting to
resolve the kinds of discrepancies that occur when an area is tested in both
comprehension and production: that is, which kinds of data are to be
considered most reliably an indication of a child’s linguistic ability? With
regard to the use of experimental techniques, therefore, particularly in
combination with each other, there are still a number of issues to resolve.
Methodological research has also sought to develop better ways to use
spontaneous speech. One effective means has been the improvements made
516
Concluding remarks 517
in the use of parental diaries to study the general aspects of early vocabulary
acquisition. Another concerns refinements in the linguistic methods used to
analyze language samples. We have addressed two areas in particular. In
the area of phonological acquisition, we discussed the need to set some
criteria for productivity as well as ways to separate individual from general
patterns. These same general issues returned when we discussed the
analysis of children’s early word combinations. Again the question of
productivity arose. In both domains we want to eliminate random and
memorized speech forms from the set of data being used to make claims
about a child’s rule system. I am somewhat concerned that the limitations of
spontaneous samples have led some to consider them to be of no value.
Instead, while recognizing such limitations, I believe that a better methodo-
logy can still lead to new insights into the developmental patterns of both
areas.
The descriptive aspect of the text deals with a review of the relevant
literature in the field. As stated in the Introduction, I have tried to limit this
aspect in the sense of restricting the range of studies discussed. Instead, I
have gone for a more intensive survey of a small set of relevant studies. The
tone of the review, therefore, is highly different from other attempts such as
McCarthy (1954).
A goal of the literature survey has been to present the reader with a feel
for the richness of the research which has already been done in the field. I
have presented a historical overview, and have opted to focus on what I
perceive to be classic studies, though, of course, including current work
which is perceived to be of importance. There is a concern among at least
some from the Child Language perspective that this literature is being lost.
This is true not only for older studies but even for more recent ones. I
pointed out, for example, the lack of any discussion of McNeill’s work in the
1970s in Pinker (1984), despite some very general similarities in the two
approaches.
The lack of interest in previous research among at least some current
researchers can be traced to two causes. One concerns the fact that current
linguistic theory has discovered new and interesting patterns of language
which have not been known before and there was no previous research,
therefore, which directly examined them. A related cause is the feeling that
the previous research was by and large descriptive. I acknowledge that there
may well be a range of linguistic phenomena which will require entirely new
research studies, particularly those phenomena which tend to be acquired
later in childhood. At the same time, however, I believe that a number of
previous studies can be reinterpreted in the light of more recent research. I
discussed, for example, the relevance of the data in Rodd & Braine (1970)
to current discussions about the acquisition of subject pronouns. While it is
518 CONCLUDING REMARKS
true that some of the earlier research was descriptive, much still contains
insightful and speculative theoretical suggestions. It just may not be
characterized by the current linguistic jargon.
Certainly the attention to explanation has expanded and now is central to
most research currently underway. I have tried to capture the directions of
this work by comparing these different general approaches to the question.
The more traditional developmental view that children go through pro-
gressive stages of acquisition has been contrasted with two nativist views.
One of these, which I have called ‘maturational’, allows maturational
milestones and is quite comfortable with discontinuities in acquisition. The
other, which I labeled a ‘strong inclusion’ view, sees the child as essentially
adult-like from the onset of acquisition.
These differences of opinion are currently being assessed by all sides as to
their predictions about language acquisition. I have some concern about the
polarization which may result from these differences, and have contrasted
Child Language and Language Acquisition as a way to characterize the
respective camps. I have some worries about the future in this regard, but
see no inherent differences in the general goal of all sides to better
understand the child’s language acquisition process. Indeed, I believe that
the disagreements may come down to whether one assigns a child’s
linguistic behavior as the result of its performance or competence. If such is
the case, we may all be satisfied at some point with the data, and only debate
its interpretation.
The more serious point perhaps, at least as concerns the future of our
field, is the interest in data at all. If language acquisition shows little about
the child’s emerging linguistic competence, then linguists will presumably
lose interest in child language data altogether. This does not mean, of
course, that there will be no more issues of interest to study. Certainly an
account of the performance mechanisms at work will be a primary issue. So
too, the logical problem of how specific constructions can be acquired from
positive evidence will keep us busy for many years to come. Even so, much
of the research in the field has been driven by the assumption that children’s
patterns of speech reveal something of interest about their linguistic rules. It
is certainly this interest which has led me to highlight the constructionist
perspective in this text. The resolution of this point, to my mind, will
constitute the single most important event in the history of the field.
Bibliography
Abrahamsen, A. 1977. Child language: an interdisciplinary guide to theory and

research. Baltimore, Md.: University Park Press.
Abrahamsen, D. 1938. The function of language and its development in early
childhood. Acta Psychiatrica et Neurologica 13: 649-58.
Ament, W. 1899. Die Entwicklung von Sprechen und Denken beim Kinde. Leipzig:
Ernst Wunderlich.
Amidon, A. 1976. Children’s understanding of sentences with contingent relations:
why are temporal and conditional connectives so difficult? Journal of Experi-
mental Child Psychology 22: 423-37.
Amidon, A. & Carey, S. 1972. Why five-year-olds cannot understand before and
after. Journal of Verbal Learning and Verbal Behavior 11: 417-23.
Anastasi, A. & D’Angelo, R. 1952. A comparison of Negro and white preschool
children in language development and Goodenough Draw-a-Man I.Q. Journal
of Genetic Psychology 81: 147-65.
Anglin, J. 1970. The growth of word meaning. Cambridge, Mass: MIT Press.
1977. Word, object, and conceptual development. New York: Norton.
Anisfeld, M. 1984, Language development from birth to three. Hillsdale, NJ:
Lawrence Erlbaum.
Antinucci, F. & Miller, R. 1976. How children talk about what happened. Journal
of Child Language 3: 167-89.
Antinucci, F. & Parisi, D. 1973. Early language acquisition: a model and some data.
In Ferguson & Slobin (1973: 607-19).
Arlman-Rupp, A , , van Niekerk de Haan, D. & van de Sandt-Koenderman, M.
1976. Brown’s early stages: some evidence from Dutch. Journal of Child
Language 3: 267-74.
Armstrong, S., Gleitman, L. & Gleitman, H. 1983. What some concepts might not
be. Cognition 13: 263-308.
A s h , R. & Pisoni, D. 1980. Some developmental processes in speech perception.
In Yeni-Komshian, Kavanaugh & Ferguson (1980b: 75-102).
Atkinson, M. 1982. Explanation in the study of language acquisition. Cambridge:
Cambridge University Press.
Bach, E. & Harms, R. (eds.) 1968. Universals in linguistic theory. New York: Holt,
Rinehart & Winston.
Baker, C. L. 1979. Syntactic theory and the projection problem. Linguistic Inquiry
10: 533-81.
Baker, C . L. & McCarthy, J. (eds.) 1981. The logical problem of language
acquisition. Cambridge, Mass.: MIT Press.
519
520 BIBLIOGRAPHY
Baldie, B. 1976. The acquisition of the passive voice. Journal of Child Language 3:
33148.
Bar-Adon, A. & Leopold, W. (eds.) 1971. Child language. A book of readings.
Englewood Cliffs, NJ: Prentice-Hall.
Barnes, S., Gutfreund, M., Satterly, D. & Wells, G. 1983. Characteristics of adult
speech which predict children’s language development. Journal of Child Lan-
guage IO: 65-84.
Barrett, M. 1978. Lexical development and overextension in child language. Journal
1982. The holophrastic hypothesis: conceptual and empirical issues. Cognition 11:
47-76.
Barrie-Blackley, S. 1973. Six-year-old children’s understanding of sentences
adjoined with time adverbs. Journal of Psycholinguistic Research 2: 153-65.
Bartlett, E. 1976. Sizing things up: the acquisition of the meaning of dimensional
adjectives. Journal of Child Language 3: 205-20.
Barton, D. 1975. Statistical significance in phonemic perception experiments.
Journal of Child Language 2: 297-8.
1976a. The role of perception in the acquisition of phonology. Doctoral disser-
tation, University of London.
1976b. Phonemic discrimination and the knowledge of words in children under
three years of age. Papers and Reports on Child Language Development 11: 61-8.
1978. The discrimination of minimally-different pairs of real words by children
aged 2;3 to 2;11. In Waterson & Snow (1978: 255-61).
1980. Phonemic perception in children. In Yeni-Komshian, Kavanaugh & Fer-
guson (1980b: 97-116).
Bateman, W. 1916. The language status of three children at the same ages.
Pedagogical Seminary 23 ; 2 11-40.
Bates, E. 1976. Language and context: the acquisition of pragmatics. New York:
Academic Press.
1979. The emergence of symbols: cognition and communication in infancy. New
York: Academic Press.
Bates, E., Camaioni, L. & Volterra, V. 1975. The acquisition of performatives prior
to speech. Merrill-Palmer Quarterly 21: 205-36.
Bates, E. & MacWhinney, B. 1979. The functionalist approach to the acquisition of
grammar. In Ochs & Schiefflin (1979: 167-211).
1982. The development of grammar. In Wanner and Gleitman (1982: 173-218).
Bates, E. & Rankin, J. 1979. Morphological development in Italian: connotation
and denotation. Journal of Child Language 6: 29-52.
Bateson, M. C. 1975. Mother-infant exchanges: the epigenesis of conversational
interaction. In D. Aaronson & R. Rieber (eds.) Developmental psycholinguis-
tics and communication disorders. Annals of the New York Academy of Science
263: 21-9.
Bausano, M. K. & Jeffrey, W. E. 1975. Dimensional salience and judgements of
bigness by three-year-old children. Child Development 46: 988-91.
Bavin, E. & Shopin, T. 1985. Children’s acquisition of Walpiri: comprehension of
transitive sentences. Journal of Child Language 12: 597-610.
Bellugi, U. 1967. The acquisition of the system of negation in children’s speech.
Doctoral dissertation, Harvard University.
Bibliography 521
1971. Simplification in children’s language. In Huxley & Ingram (1971: 95-119).
Bellugi, U. & Brown, R. (eds.) 1964. The acquisition of language. Monographs of
the Society for Research in Child Development 29.
Benedict, H. 1979. Early lexical development: comprehension and production.
Berko, J. 1958. The child’s learning of English morphology. Word 14: 150-77.
Berko-Gleason, J. 1975. Fathers and other strangers. In Dato (1975: 289-97).
Berman, R. 1981a. Regularity vs. anomaly: the acquisition of Hebrew inflectional
morphophonology. Journal of Child Language 8: 265-82.
1981b. Language development and language knowledge: evidence from the
acquisition of Hebrew morphophonology. Journal of Child Language 8: 609-26.
1982. Verb-pattern alternation: the interface of morphology, syntax, and seman-
tics in Hebrew child language. Journal of Child Language 9: 169-91.
Berwick, R. & Weinberg, A. 1984. The grammatical basis of linguistic performance.
Cambridge, Mass.: MIT Press.
Bever, T. 1970. The cognitive basis for linguistic structures. In Hayes (1970:
274-353).
Bierwisch, M. 1967. Some semantic universals of German adjectivals. Foundations
of Language 3: 1-36.
1970. Semantics. In J. Lyons (ed.) New horizons in linguistics. London: Penguin,
pp. 16684.
Blewitt, P. 1982. Word meaning acquisition in young children: a review of theory
and research. In H. Reese & L. Lipsitt (eds.) Advances in child development and
behavior, vol. 17. New York: Academic Press, pp. 140-95.
Bloom, L. 1970. Language development: form and function in emerging grammars.
1971. Why not pivot grammar? Journal of Speech and Hearing Disorders 36: 4@50.
1973. One word at a time. The Hague: Mouton.
(ed.) 1978. Readings in language development. New York: John Wiley.
Bloom, L. & Lahey, M. (eds.) 1978. Language development and language disorders.
New York: John Wiley.
Bloom, L., Hood, L. & Lightbown, P. 1974. Imitation in language acquisition: if,
when, and why? Cognitive Psychology 6: 38W20. Reprinted in L. Bloom (1978:
452-88).
Bloom, L., Lifter, K. & Hafitz, J. 1980. Semantics of verbs and the development of
verb inflections in child language. Language 56: 38W12.
Bloom, L., Lightbown, P. & Hood, L. 1975. Structure and variation in child
language. Monographs of the Society for Research in Child Development 40 (2,
serial no. 160).
Bloom, L., Miller, P. & Hood, L. 1975. Variation and reduction as aspects of
competence in language development. In A. Pick (ed.) Minnesota Symposia on
Child Psychology 9: 3-55. Reprinted in Bloom (1978: 169-216).
Bloom, L., Rocissano, L. & Hood, L. 1976. Adult-child discourse: developmental
interaction between information processing and linguistic knowledge. Cognitive
Psychology 8: 521-52.
Bloom, L., Lahey, M., Hood, L. Lifter, K. & Fiess, K. 1980. Complex sentences:
acquisition of syntactic connectives and the semantic relations they encode.
522 BIBLIOGRAPHY
Bloomfield, L. 1933. Language. New York: Henry Holt.
Blount, B. 1969. Acquisition of language by Luo children. Doctoral dissertation,
University of California, Berkeley.
1972. The prelinguistic system of Luo children. Anthropological Linguistics 12:
326-42.
Bohn, W. 1914. First steps in verbal expression. Pedagogical Seminary 21: 578-95.
Borer, H. & Wexler, K . 1987. The maturation of syntax. In Roeper & Williams
(1987: 123-72).
Bowerman, M. 1973a. Early syntactic development: a cross-linguistic study with
special reference to Finnish. Cambridge: Cambridge University Press.
1973b. Structural relationships in children’s utterances: syntactic or semantic? In
Moore (1973: 197-213).
1974. Learning the structure of causative verbs: a study in the relationship of
cognitive, semantic and syntactic development. Papers and Reports on Child
Language Development 8: 142-78.
1975. Commentary, In Bloom, Lightbown & Hood (1975: 80-90).
1976a. Commentary. In Braine (1976: 98-104).
1976b. Semantic factors in the acquisition of rules for word use and sentence
construction. In Morehead & Morehead (1976: 99-179).
1977, The acquisition of rules governing ‘possible lexical items’: evidence from
spontaneous speech errors. Papers and Reports on Child Language Development
13: 148-56.
1978a. The acquisition of word meaning: an investigation into some current
conflicts. In Waterson & Snow (1978: 263-87).
1978b. Semantic and syntactic development: a review of what, when, and how in
language acquisition. In Schiefelbusch (1978: 97-189).
1978c. Systematizing semantic knowledge: changes over time in the child’s
organization of meaning. Child Development 49: 977-87.
1979. The acquisition of complex sentences. In Fletcher & Garman (1979:
285-305).
1981. Beyond communicative competency: from piecemeal knowledge to an
integrated system in the child’s acquisition of language. Papers and Reports on
Child Language Development 20: 1-24.
1986. What shapes children’s grammars? In Slobin (1986: 1257-319).
Boyd, W. 1913. The beginnings of syntactical speech: a study in child linguistics.
Child Study 6: 21-4 and 47-51.
Braine, M. D. S. 1963a. The ontogeny of English phrase structure: the first phase.
Language 39: 1-13.
1963b. On learning the grammatical order of words. Psychological Review 70:
323-48.
1971a. On two types of models of the internalization of grammars. In Slobin
(1971: 153-86).
1971b. The acquisition of language in infant and child. In C. E. Reed (ed.) The
learning of language. New York: Appleton-Century-Crofts, pp. 7-95.
1973. Three suggestions regarding grammatical analyses of children’s language.
In Ferguson & Slobin 1973: 421-9.
1974a. On what might constitute a learnable phonology. Language 50: 270-99.
1974b. Length constraints, reduction rules, and holophrastic processes in chil-
Bibliography 523
dren’s word combinations. Journal of Verbal Learning and Verbal Behavior 13:
448-56.
1976. Children’s first word combinations. Monographs of the Society for Research
in Child Development 41 (1, serial no. 164).
Braine, M. D. S. & Hardy, J. 1981. Case categories. In Deutsch (1981: 201-22).
1982. On what case categories there are, why they are, and how they develop: an
amalgam of a priori considerations, speculation, and evidence from children. In
Wanner & Gleitman (1982: 219-39).
Braine, M. D. S. & Wells, R. 1978. Case-like categories in children: the Actor and
more related categories. Cognitive Psychology 10: 100-22.
Brainerd, C. 1978. The stage question in cognitive-developmental theory. The
Behavioral and Brain Sciences 2: 173-213.
Brandenburg, G. 1915. The language of a three-year-old child. Pedagogical
Seminary 22: 89-120.
Branigan, G. 1979. Some reasons why successive single word utterances are not.
Braunwald, S. 1978. Context, word and meaning: towards a communicational
analysis of lexical acquisition. In Lock (1978: 485-527).
Bresnan, J. 1978. A realistic transformational grammar. In Halle, Bresnan & Miller
(1978: 1-59).
(ed.) 1982. The mental representation of grammatical relations. Cambridge,
Mass.: MIT Press.
Bretherton, I., McNew, S., Snyder, L. & Bates, E. 1983. Individual differences at
20 months: analytic and holistic strategies in language acquisition. Journal of
Child Language 10: 293-320.
Brewer, W. & Stone, J. 1975. Acquisition of spatial antonym pairs. Journal of
Experimental Child Psychology 19: 299-307.
Brown, H. D. 1971. Children’s comprehension of relativized English sentences.
Child Development 42: 1923-6.
Brown, R. 1958a. Words and things. Glencoe, Ill.: Free Press.
1958b. How shall a thing be called? Psychological Review 65: 14-21.
1968. The development of wh questions in child speech. Journal of Verbal
Learning and Verbal Behavior 7: 277-90.
1970a. Psycholinguistics: selected papers of Roger Brown. New York: Free
Press.
1970b. The first sentences of child and chimpanzee. In Brown (1970a: 208-31).
1973. A first language: the early stages. Cambridge, Mass.: Harvard University
Press.
1977. Introduction. In Snow & Ferguson (1977: 1-27).
Brown, R. & Bellugi, U. 1964. Three processes in the child’s acquisition of syntax.
Harvard Educational Review 34: 133-5. Also in Lenneberg (1964).
Brown, R. & Fraser, C . 1963. The acquisition of syntax. In C. Cofer & B. Musgrave
(eds.) Verbal behavior and learning: problems and processes. New York:
McGraw-Hill. pp. 158-201.
1964. The acquisition of syntax. In Bellugi & Brown (1964: 43-79).
Brown, R., Cazden, C. & Bellugi, U. 1969. The child‘s grammar from I to 111. In
J. P. Hill (ed.). Minnesota Symposium on Child Psychology, vol. 2. Minneapo-
lis, Minn.: University of Minnesota Press, pp. 28-73.
524 BIBLIOGRAPHY
Brown, R. & Hanlon, C. 1970. Derivational complexity and order of acquisition. In
Hayes (1970: 11-53).
Brown, R., Fraser, C. & Bellugi, U. 1964. Explorations in grammar evaluation, In I
Bellugi & Brown (1964: 79-92).
Bruner, J. 1975. The ontogenesis of speech acts. Journal of Child Language 2: 1-21.
Bruner, J. & Sherwood, V. 1976. Peekaboo and the learning of rule structures. In
J. S. Bruner, A. Jolly & K. Sylva (eds.) Play - its role in development and
evolution. Harmondsworth, Middx: Penguin, pp. 277-85.
Buhler, C. 1931. Kindheit und Jugend. Leipzig: Hirzel.
Burroughs, G. 1957. A study of the vocabulary ofyoung children. Edinburgh: Oliver
and Boyd.
Butterfield, E. & Cairns, G. 1974. Discussion summary - infant reception research.
In Schiefelbusch and Lloyd (1974: 75-102).
Bybee, J. 1979. Child morphology and morphophonemic change. Linguistics 17:
21-50.
Campbell, D. & Stanley, J. 1963. Experimental and quasi-experimental design for
research in teaching. In N. Gage (ed.) Handbook of research on teaching.
Chicago, Ill.: Rand McNally.
Caramazza, A. & Zuriff, E. (eds.) 1978. Language acquisition and language
breakdown. Baltimore, Md.: The Johns Hopkins University Press.
Carden, G. 1986. Blocked forwards anaphora: theoretical implications of the
acquisition data. In Lust (1986: 31S57).
Carey, S. 1978. The child as word learner. In Halle, Bresnan & Miller (1978: 26493).
1982. Semantic development: the state of the art. In Wanner & Gleitman (1982:
347-89).
Carey, S. & Bartlett, E. 1978. Acquiring a single word. Papers and Reports on Child
Carter, A. L. 1975a. The transformation of sensorimotor morphemes into words: a
case study of the development of ‘here’ and ‘there’. Papers and Reports on Child
1975b. The transformation of sensorimotor morphemes into words: a case study
of the development of ‘more’ and ‘mine’. Journal of Child Language 2: 233-50.
1978a. The development of systematic vocalizations prior to words: a case study.
In Waterson & Snow (1978: 127-38).
1978b. From sensorimotor vocalizations to words: a case study of the evolution of
attention-directing communication in the second year. In Lock (1978: 309-49).
1979. Prespeech meaning relations: an outline of one infant’s sensorimotor
morpheme development. In Fletcher & Garman (1979: 71-92).
Caudill, W. & Weinstein, H. 1969. Maternal care and infant behavior in Japan and
America. Psychiatry 32: 12-43.
Cazden, C. 1968. The acquisition of noun and verb inflections. Child Development
39: 433-48.
1970. Children’s questions: their form, functions, and roles in education. Young
Children, March, pp. 202-20.
1972. Child language and education. New York: Holt, Rinehart & Winston.
Chamberlain, A. & Chamberlain, I. 1904. Studies of a child. Pedagogical Seminary
11: 26691,452-83.
,
Bibliography 525
1905. Studies of a child. Pedagogical Seminary 12: 427-53.
1909. ‘Meanings’ and ‘definitions’ in the 47th and 48th months. Pedagogical
Seminary 16: 64-103.
Chao, Y.-R. 1951. The Cantian idiolect. University of California Publications in
Semitic Philology 2: 27-44. Reprinted in Bar-Adon & Leopold (1971: 116-30).
Chapman, R. 1981a. Cognitive development and language comprehension in
10-21-month-olds.In Stark (1981: 359-91).
1981b. Mother-child interaction in the second year of life: its role in language
development. In Schiefelbusch & Bricker (1981: 201-50).
Chapman, R. & Thomson, J. 1980. What is the source of overextension errors in
comprehension testing of two-year-olds? A response to Fremgen & Fay. Journal
Chen, M. 1975. An areal study of nasalization in Chinese. In C. Ferguson, L.
Hyman & J. Ohala (eds.) Nasalfest: Papers from a symposium on nasals and
nasalization. Palo Alto, Calif. : Stanford University Press.
Chomsky, C. 1969. The acquisition of syntax in children from 5 to 10. Cambridge,
Mass.: MIT Press.
Chomsky, N. 1957. Syntactic structures. The Hague: Mouton.
1959. Review of Verbal behavior, by B. F. Skinner. Language 35: 26-58.
1964. Formal discussion. In Bellugi & Brown (1964: 35-9).
1965. Aspects of the theory of syntax. Cambridge, Mass.: MIT Press.
1968. Language and mind. New York: Harcourt Brace Jovanovich.
1972. Studies on semantics in generative grammar. The Hague: Mouton.
1975. Reflections on language. New York: Pantheon.
1981. Lectures on government and binding: the Pisa lectures. Dordrecht: Foris.
1982. Some concepts and consequences of the theory of government and binding.
1986. Barriers. Cambridge, Mass. : MIT Press.
Chomsky, N. & Halle, M. 1968. Thesoundpattern of English. New York: Harper &
Row.
Clark, E. 1971. On the acquisition of the meaning of ‘before’ and ‘after’. Journal of
Verbal Learning and Verbal Behavior 10: 266-75.
1972. On the child’s acquisition of antonyms in two semantic fields. Journal of
Verbal Learning and Verbal Behavior 11: 75@8.
1973a. What’s in a word? On the child’s acquisition of semantics in his first
language. In Moore (1973: 65-110).
1973b. Nonlinguistic strategies and the acquisition of word meanings. Cognition
2: 161-82.
Clark, H. 1970. How young children describe events in time. In G. Flores d’Arcais
& W. Levelt (eds.) Advances in psycholinguistics. Amsterdam: North-Holland.
1973. Space, time, semantics, and the child. In Moore (1973: 27-63).
Clark, H. & Clark, E. 1977. Psychology and language: an introduction. New York:
Harcourt Brace Jovanovich.
Clark, R. 1974. Performing without competence. Journal of Child Language 1: 1-10.
1977. What’s the use of imitation? Journal of Child Language 4: 341-58.
1982. Theory and method in child-language research: are we assuming too much?
In Kuczaj (1982b: 1-36).
Clumeck, H. 1982. The effect of word familiarity on phonemic recognition in
526 BIBLIOGRAPHY
preschool children aged 3 to 5 years. In C. E. Johnson & C. L. Thew (eds.)
Proceedings of the Second International Congress for the Study of Child
Language. Vol. 1. Washington, DC: University Press of America, pp. 58-77.
Coker, P. 1978. Syntactic and semantic factors in the acquisition of before and after.
Papers and Reports on Child Language Development 12: 81-8.
Comrie, B. 1981. Language universals and linguistic typology. Chicago, Ill.:
University of Chicago Press.
1983. Switch-reference in Huichol: a typological study. In J. Haiman & P. Munro
(eds.) Switch-reference and universal grammar. Amsterdam: John Benjamins,
pp. 17-37.
Crain, S. & McKee, C. 1985. Acquisition of structural relations on anaphora. NELS
16: 94-110.
Cruttenden, A. 1970. A phonetic study of late babbling. British Journal of Disorders
of Communication 5: 110-18.
Crystal, D., Fletcher, P. & Garman, M. 1976. The grammatical analysis of language
disability. London: Edward Arnold, pp. 200-7.
Culicover, P. & Wilkins, W. 1984. Locality in linguistic theory. Orlando, Fla.:
Academic Press.
Dale, P. 1976. Language development: structure and function. New York: Holt,
Rinehart & Winston.
1985, Productivity analysis of early child language (PANEL): user’s guide. IBM
versions. Unpublished paper.
Dale, P. & Ingram, D., (eds.) 1980. Child language: an international perspective.
Baltimore, Md.: University Park Press.
Darwin, C. 1877. A bibliographical sketch of an infant. Mind 2: 285-94.
Dato, D. (ed.) 1975. Developmental psycholinguistics: theory and application.
Georgetown University Round Table on Languages and Linguistics 1975. Wash-
ington: Georgetown University.
Davis, E. 1937. The development of linguisticskills in twins, singletons with siblings,
and only children from age five to ten years. University of Minnesota Institute of
Child Welfare, Monograph Series 14.
Davis, H. 1983. Restructuring in child language. Paper presented to the Canadian
Linguistic Association annual meeting, Vancouver, BC.
1987. The acquisition of the English auxiliary system and its relation to linguistic
theory. Doctoral dissertation, University of British Columbia.
Day, E. 1932. The development of language in twins: 1. A comparison of twins and
single children. 2. The development of twins: their resemblances and differ-
ences. Child Development 3: 179-99.
de Boysson-Bardies, B., Sagart, L. & Durand, C. 1984. Discernible differences in
the babbling of infants according to target language. Journal of Child Language
11: 1-15.
de Laguna, G. 1927. Speech: its function and development. New Haven, Conn.: Yale
University Press.
Della Corta, M., Benedict, R. & Klein, D. 1983. The relationship of pragmatic I
dimensions of mothers’ speech to the referential-expressive distinction. Journal

de Saussure, F. 1922. Cows de linguistique gtntrale. Paris: Pavot, 5th edn 1955.
(1st edn 1916). English translation by Wade Baskin, 1955. Course in general I
linguistics. New York: Philosophical Library, 1959.

Bibliography 527
Deutsch, W. (ed.) 1981. The child’s construction of language. New York: Academic
Press.
Devereux, G. 1949. Mohave voice and speech mannerisms. Word 5: 268-72.
de Villiers, J. (1980). The process of rule learning in child speech: a new look. In
K. E. Nelson (1980:1-44).
de Villiers, J. & de Villiers, P. 1973a. P. Development of the use of word order in
comprehension. Journal of Psycholinguistic Research 2: 331-41.
1973b. A cross-sectional study of the acquisition of grammatical morphemes in
child speech. Journal of Psycholinguistic Research 2: 267-78.
1978. Language acquisition. Cambridge, Mass. : Harvard University Press.
de Villiers, J., Tager-Flusberg, H., Hakuta, K. & Cohen, M. 1979. Children’s
comprehension of relative clauses. Journal of Psycholinguistic Research 8:
499-518.
Dinnsen, D. A. 1980. Phonological rules and phonetic explanation. Journal of
Linguistics 16: 171-91.
Dockrell, J. E. 1981. The child’s acquisition of unfamiliar words: an experimental
study. Unpublished doctoral dissertation, Stirling University, Scotland.
Donaldson, M. & Wales, R. 1970. On the acquisition of some relational terms. In
Hayes (1970: 235-68).
Donegan, P. J. 1986. On the naturalphonology of vowels. Outstanding Dissertations
in Linguistics series. New York: Garland.
Donegan, P. J. & Stampe, D. 1979. The study of natural phonology. In D. A.
Dinnsen (ed.) Current approaches to phonological theory. Bloomington, Ind. :
Indiana University Press, pp. 126-73.
Dore, J. 1975. Holophrases, speech acts and language universals. Journal of Child
Language 3: 22-39.
1979. What’s so conceptual about the acquisition of linguistics structures? Journal
Dore, J., Franklin, M., Miller, R. T. & Ramer, A. 1976. Transitional phenomena in
early acquisition. Journal of Child Language 3: 13-28.
Dresher, E. 1981. On the learnability of abstract phonology. In Baker & McCarthy
(1981: 188-210).
Drummond, M. 1916. Notes on speech development, 1-2. Child Study 9: 83-6,95-9.
Edwards, M. L. 1970. The acquisition of liquids. MA dissertation, Ohio State
University.
1974. Perception and production in child phonology: the testing of four hypo-
theses. Journal of Child Language 1: 205-19.
Edwards, M. L. & Shriberg, L. D. 1983. Phonology: applications in communication
disorders. San Diego, Calif. College-Hill Press.
Eilers, R. 1980. Infant speech perception: history and mystery. In Yeni-Komshian,
Kavanaugh & Ferguson (1980b: 23-39).
Eilers, R. & Gavin, W. 1981. The evaluation of infant speech perception skills:
statistical techniques and theory development. In Stark (1981: 185-213).
Eilers, R. & Oller, D. K. 1976. The role of speech discriminations in developmental
sound substitutions. Journal of Child Language 3: 319-29.
Eilers, R., Gavin, W. & Oller, K. 1982. Cross-linguistic perception in infancy: early
effects of linguistic experience. Journal of Child Language 9: 289-3b2.
Eilers, R., Gavin, W. & Wilson, W. 1979. Linguistic experience and phonemic
perception in infancy: a cross-linguistic study. Child Development 50: 14-18.
528 BIBLIOGRAPHY
Eilers, R., Oller, D. K. & Ellington, J. 1974. The acquisition of word meaning for
dimensional adjectives: the long and the short of it. Journal of Child Language 1:
195-204.
Eilers, R., Wilson, W. & Moore, M. 1976. Discrimination of synthetic prevoiced
labial stops by infants and adults. Journal of the Acoustical Society of America
60, Supplement 1, S91 (abstract).
1977. Developmental changes in speech discrimination in infants. Journal of
Speech and Hearing Research 20: 766-80.
Eilers, R. Oller, D. K., Bull, D. & Gavin, W. 1984. Linguistic experience and infant
perception: a reply to Jusczyk, Shea & Aslin (1984). Journal of Child Language
11: 467-75.
Eimas, P. 1974. Linguistic processing of speech by young infants. In Schiefelbusch &
Lloyd (1974: 55-73).
Eimas, P., Miller, J. & Jusczyk, P. In press. On infant speech perception and the
acquisition of language. In S. Harnad (ed.) Categorical perception. Cambridge:
Cambridge University Press. I
Eimas, P., Siqueland, E., Jusczyk, P. & Vigorito, J. 1971. Speech perception in
infants. Science 171: 303-18. Reprinted in Bloom (1978: 87-93).
Elbers, L. 1982. Operating principles in repetitive babbling: a cognitive continuity
approach. Cognition 12: 45-63.
Elkonin, D. B. 1971. Development of speech. In A. Va. Zaporozhets & D. B.
Elkonin (eds.) The psychology of preschool children. Cambridge, Mass. : MIT
Press, pp. 111-85.
Emonds, J. 1976. A transformational approach to English syntax: root, structure-
preserving and local transformations. New York: Academic Press.
Erreich, A. 1984. Learning how to ask: patterns of inversion in yes-no and wh
questions. Journal of Child Language 11: 579-92.
Erreich, A., Valian, V. & Winzemer, J. 1980. Aspects of a theory of language
acquisition. Journal of Child Language 2: 157-79.
Fay, D. 1978. Reply to Kuczaj (1976). Journal of Child Language 5: 143-9.
Feagans, L. 1980. How to make sense of temporalkpatial ‘before’ and ‘after’.
Ferguson, C. A. 1964.Babytalkinsix languages. AmericanAnthropologist66: 103-14.
1978. Learning to pronounce: the earliest stages of phonological development in
the child. In Minifie & Lloyd (1978: 237-97).
Ferguson, C. A. & Farwell, C. B. 1975. Words and sounds in early language
acquisition. Language 51: 49-39. An earlier version appeared in 1973 in Papers
and Reports on Child Language Development 6: 1-60.
Ferguson, C . A. & Garnica, 0. K. 1975. Theories of phonological development. In
Lenneberg & Lenneberg (1975: 153-80).
Ferguson, C . A. & Slobin, D. (eds.) 1973. Studies of child language development.
New York: Holt, Rinehart & Winston.
Fernald, C. 1972. Control of grammar in imitation, comprehension, and pro-
duction: problems of replication. Journal of Verbal Learning and Verbal f
Behavior 11: 606-13.

Fillmore, C. 1968. The case for case. In E. Bach & R. Harms (eds.) Universals in
linguistic theory. New York: Holt, Rinehart & Winston, pp. 1-88.
Fisher, M. 1934. Language patterns of preschool children. Child Development J
Monographs 15. New York: Teachers’ College, Columbia University.
Bibliography 529
Flavell, J. 1963. The developmental psychology of Jean Piaget. New York: Van
Nostrand.
Fletcher, P. & Garman, M. (eds.) 1979. 2nd edn 1986. Language acquisition.
Cambridge: Cambridge University Press.
Fodor, J. 1975. The language of thought. New York: Crowell.
Fodor, J., Garrett, M., Walker, E. & Parkes, C . 1980. Against definitions.
Cognition 8: 26347.
Fonagy, I. 1972. A propos de la genese de la phrase enfantine. Lingua 30:
31-71.
Fortescue, M. 1984/5. Learning to speak Greenlandic. First Language 5: 101-13.
Francis, H. 1969. Structure in the speech of a 24-year-old. British Journal of
Educational Psychology 39: 291-302.
Fraser, C., Bellugi, U. & Brown, R. 1963. Control of grammar in imitation,
comprehension, and production. Journal of Verbal Learning and Verbal Behav-
ior 2: 121-35.
Freedle, R. & Lewis, M. 1977. Prelinguistic conversation. In M. Lewis & P.
Rosenbaum (eds.) Interaction, conversation and the development of language.
New York: John Wiley, pp. 157-85.
Fremgen, A , & Fay, D. 1980. Overextensions in production and comprehension: a
methodological clarification. Journal of Child Language 7: 205-1 1.
French, L. & Brown, A. 1977. Comprehension of ‘before’ and ‘after’ in logical and
arbitrary sequences. Journal of Child Language 4: 247-56.
Friedman, W. & Seeley, P. 1976. The child’s acquisition of spatial and temporal
word meanings. Child Development 47: 1103-8.
Furrow, D., Nelson, K. & Benedict, H. 1979. Mothers’ speech to children and
syntactic development: some simple relationships. Journal of Child Language 6:
42342.
Gaer, E. P. 1969. Children’s understanding and production of sentences. Journal of
Verbal Learning and Verbal Behavior 8: 289-94.
Garnica, 0. K. 1971. The development of the perception of phonemic differences in
initial consonants by English-speaking children: a pilot study. Papers and
Reports on Child Language Development 3: 1-29.
1973.The development of phonemic speech perception. In Moore (1973: 214-22).
1977. Some prosodic and paralinguistic features of speech to young
children. In Snow & Ferguson (1977: 63-88).
Gazdar, G., Klein, E., Pullum, G . & Sag, I. 1985. Generalized phrase structure
grammar. Cambridge, Mass. : Harvard University Press.
Gilbert, J. 1982. Babbling and the deaf child: a commentary on Lenneberg et a/.
(1965) and Lenneberg (1967). Journal of Child Language 9: 511-15.
Ginsberg, H. & Opper, S. 1969. Piaget’s theory of intellectual development: an
introduction. Englewood Cliffs, NJ: Prentice-Hall.
Givon, T. 1979. On understanding grammar. New York: Academic Press.
Gleitman, L. 1981. Maturational determinants of language growth. Cognition 10:
103-14.
Gleitman, L., Newport, M. & Gleitman, H. 1984. The current status of the
motherese hypothesis. Journal of Child Language 11: 43-79.
Gleitman, L., Shipley, E. & Smith, C. 1978. Old and new ways not to study
comprehension: comments on Petretic & Tweney’s (1977) experimental review
of Shipley, Smith & Gleitman (1969). Journal of Child Language 5: 501-19.
530 BIBLIOGRAPHY
Goad, H. & Ingram, D. 1988. Individual variation and its relevance to a theory of
phonological acquisition. Journal of Child Language 14: 419-32.
Goertz, C. 1973. The impact of the concept of culture on the concept of man. In C.
Goertz, Interpretation of cultures. New York: Basic Books, pp. 33-54.
Goldstein, U. 1979. Modeling children’s vocal tracts. Paper presented at the 97th
Meeting of the Acoustical Society of America, Cambridge, Mass., June 12-16.
Gottlieb, G. 1976. Conceptions of prenatal development: behavioral embryology.
Psychological Review 83: 215-34.
Graham, L. W. & House, A. S. 1971. Phonological oppositions in children: a
perceptual study. Journal of the Acoustical Society of America 49: 559-66.
Greenberg, J. & Kuczaj, S. Towards a theory of substantive word meaning
acquisition. In Kuczaj (1982: 275-311).
Greenfield, P. & Smith, J. 1976. The structure of communication in early language
development. New York: Academic Press.
Greenfield, P. & Zukow, P. 1978. Why do children say what they say when they say
I
it? An experimental approach to the psychogenesis of presupposition. In K. E.
Nelson (1978).
GrCgoire, A. 1937, 1947. L’apprentissage du langage. Vol. 1 (1937); Vol. 2 (1947).
Libge, Paris: Droz.
Grimshaw, J. 1981. Form, function, and the language acquisition device. In Baker &
McCarthy (1981: 165-82).
Gruber, J. S. 1967. Topicalization in child language. Foundations of Language 3:
37-65. Reprinted in Bar-Adon & Leopold (1971: 364-81).
1975a. ‘Topicalization’ revisited. Foundations of Language 13: 57-72.
1975b. Performative-constative transition in child language. Foundations of
Language 12: 513-27.
Grunwell, P. 1985. Phonological assessment of child speech (PACS). San Diego,
Calif.: College-Hill Press.
Guillaume, P. 1927/28. First stages of sentence formation in children’s speech. In
Bloom (1978: 13148).
Gvozdev, A. 1949. Formirovanie u rebenka grammaticheskogo stroya russkogo
yazyka (Formation in the child grammatical structure of the Russian language),
parts 1 & 2. Moscow: Akad. Pedag. Nauk RSFSR.
Halle, M., Bresnan, J. &Miller, G. (eds.) 1978. Linguistic theory and psychological
reality. Cambridge, Mass.: MIT Press.
Halliday, M. A. K. 1975. Learning how to mean: explorations in the development of
language development. London: Edward Arnold.
Hamburger, H. & Crain, S . 1982. Relative acquisition. In Kuczaj (1982b: 245-74).
Harding, C. & Golinkoff, R. 1979. The origins of intentional vocalizations in
prelinguistic infants. Child Development 50: 33-40.
Hayes, J. R. (ed.) 1970. Cognition and the development of language. New York:
John Wiley.
Hayhurst, H. 1967. Some errors of young children in producing passive sentences.
Journal of Verbal Learning and Verbal Behavior 6: 634-9.
Hill, J. 1984. Combining two-term relations: evidence in support of flat structure.
Hills, E. 1914. The speech of a child two years of age. Dialect Notes 4: 84-100.
Hodson, B. & Paden, E. 1983. Targeting intelligible speech: a phonological
approach to remediation. San Diego, Calif.: College-Hill Press.
Bibliography 531
Hoek, D., Ingram, D. & Gibson, D. 1986. An examination of the possible causes of
children's early word extensions. Journal of Child Language 13: 477-94.
Hoff-Ginsburg, E. & Shatz, M. 1982. Linguistic input and the child's acquisition of
language. Psychological Bulletin 92: 2-26.
Hogan, L. 1898. A study of a child, illustrated with over 500 original drawings by the
child. New York: Harper.
Holmes, U. 1927. The phonology of an English-speaking child. American Speech 2:
219-25.
Hood, L. & Bloom, L. 1979. What, when and how about why: a longitudinal study
of expressions of causality in the language development of two-year-old
children. Society for Research in Child Development, serial no. 181, vol. 44,
no. 6.
Horgan, D. 1978. The development of the full passive. Journal of Child Language 5 :
56-80.
Howard, R. 1946. Language development in a group of triplets. Journalof Genetic
Howe, C. 1976. The meanings of two-word utterances in the speech of young
children. Journal of Child Language 3: 29-47.
Hoyer, A. & Hoyer, G. 1924. Uber die Lallsprache eines Kindes. Zeitschrift fur
angewandte Psychologie 24: 363-84.
Humphreys, M. 1880. A contribution to infantile linguistics. Transactions of the
American Philological Association 11: 5-17.
Hurford, J. 1975. A child and the English question formation rule. Journal of Child
Huttenlocher, J. 1974. The origins of language comprehension. In R. Solso (ed.)
Theories in cognitive psychology. New York: Erlbaum, pp. 331-68.
Huttenlocher, J., Eisenberg, K. & Strauss, S. 1968. Comprehension: relation
between perceived actor and logical subject. Journal of Verbal Learning and
Verbal Behaviour 7: 527-30.
Huxley, R. & Ingram, E. (eds.) 1971. Language acquisition: models and methods.
New York: Academic Press.
Hyams, N. 1983. The acquisition of parameterized grammars. Doctoral disser-
tation, City University of New York.
1984. Semantically based child grammars: some empirical inadequacies. Papers
and Reports on Child Language Development 23: 58-65.
1986. Language acquisition and the theory of parameters. Dordrecht: D. Reidel.
Ingram, D. 1971. Transitivity in child language. Language 47: 889-910.
1972. The development of phrase structure rules. Language Learning 22: 65-77.
1974a. Phonological rules in young children. Journal of Child Languagel: 49-64.
1974b. Fronting in child phonology. Journal of Child Language 1: 233-41.
1975. If and when transformations are acquired by children. Monograph Series on
Language and Linguistics. Georgetown University 27: 99-127.
1976a. Phonological disability in children. London: Edward Arnold.
1976b. Phonological analysis of a child. Glossa 10: 3-27.
1978. Sensorimotor intelligence and language development. In Lock (1978:
261-90).
1979a. Phonological patterns in the speech of young children. In Fletcher &
Garman (1979: 133-49).
532 BIBLIOGRAPHY
1979b. Stages in the acquisition of one-word utterances. In P. French (ed.) The
development of meaning. Hiroshima: Bunka Hyoron, pp. 256-81.
1981a. Procedures for the phonological analysis of children’s language. Baltimore,
Md.: University Park Press.
1981b. The notion of ‘stage’ in language acquisition studies. Proceedings from the
Second Wisconsin Symposium on Research in Child Language Disorders 2: 1-16.
Revised version to appear in H. Wittje (ed.) Festschrift Els Oksaar. Tubingen:
Gunter Narr.
1981c. The transition from early symbols to syntax. In Schiefelbusch & Bricker
(1981: 259-86).
1981d. Early patterns of grammatical development. In Stark (1981: 327-52).
1981e. On variant patterns of language acquisition. Paper presented to the
Biennial Meeting of the Society for Research in Child Development, Boston,
Mass.
1981/2. The emerging phonological system of an Italian-English bilingual child.
Journal of Italian Linguistics 2: 95-113.
1985a. The psychological reality of children’s grammars and its relation to
grammatical theory. Lingua 66: 79-103.
1985b. [VI: the acquisition. Paper presented to the Western Conference of
Linguistics, University of Victoria, October. Revised version to appear in
Language and Speech.
1985c. The acquisition of Dative Movement from positive evidence. Unpublished
paper.
1985d. On children’s homonyms. Journal of Child Language 12: 671-80.
1986. In defense of the segment in phonological acquisition. Paper presented to
the Linguistic Society of America, New York.
1988. Jakobson revisited: some evidence from the acquisition of Polish phono-
logy. Lingua 75: 55-82.
Ingram, D. & Mitchell, G. to appear. A longitudinal comparison of French and
English phonological development.
Ingram, D. & Shaw, C. 1981. The comprehension of pronominal reference in
children. Unpublished paper, University of British Columbia. Revised version
to appear in The Canadian Journal of Linguistics.
Ingram, D. & Tyack, D. 1979. The inversion of subject NP and Aux in children’s
questions. Journal of Psycholinguistic Research 4: 333-41.
Ingram, D., Christensen, L., Veach, S. & Webster, B. 1980. The acquisition of
word-initial fricatives and affricates in English by children between 2 and 6 years.
In Yeni-Komshian, Kavanaugh & Ferguson (1980a: 169-92).
Inhelder, B. 1963. Some aspects of Piaget’s genetic approach to cognition. In Kissan
& Kohlman (eds.) Thought and the young child, pp. 17-32.
Inhelder, B. & Piaget, J. 1964. The growth of logic in the child: classifcation and
seriation. London: Routledge & Kegan Paul.
Irwin, 0. C . 1941-49. See references in McCarthy (1954).
Jackendoff, R. 1977. Xsyntax: a study of phrase structure. Cambridge, Mass.: MIT
Press.
1983. Semantics and cognition. Cambridge, Mass.: MIT Press.
Jakobson, R. 1939. The sound laws of child language and their place in general
phonology. Communication prepared at Charlottenlund, Denmark, in the
Bibliography 533
summer of 1939 for the 5th International Congress of Linguists at Brussels,
September, 1939. Reprinted in Bar-Adon & Leopold (1971).
1941168. Child language, aphasia, and phonological universals. The Hague:
Mouton. Translation by R. Keiler of original German version of 1941.
Jakobson, R., Fant, G. & Halle, H. 1963. Preliminaries to speech analysis: the
distinctive features and their correlates. Technical Report 13, MIT Acoustics
Laboratory, 1952. Cambridge, Mass.: MIT Press.
Jakobson, R. & Halle, M. 1956. Fundamentals of language. The Hague: Mouton.
Jegi, J. 1901.The vocabulary of a two-year-old child. Child Study Monthly 6: 241-61.
Jenkins, J. & Palermo, D. 1964. Mediation processes and the acquisition of
linguistic structure. In Bellugi & Brown (1964: 141-69).
Jespersen, 0. 1922. Language: its nature, development, and origin. London: Allen
& Unwin.
Johnson, C. 1983. The development of children’s interrogatives: from formulas to
rules. Papers and Reports on Child Language Development 22: 108-15.
Johnson, C. J., Hardee, W. P. & Long, S. H. 1981. Perceptual basis for phonologi-
cal simplification. Paper presented to the American Speech, Language, and
Hearing Association, Los Angeles, California.
Johnston, J. 1979. A study of spatial thought and expression: ‘in back’ and ‘in front’.
Doctoral dissertation, University of California, Berkeley.
1986. Cognitive prerequisites: the evidence from children learning English. In
Slobin (1986: 961-1004).
Johnston, J. & Slobin, D. 1979. The development of locative expressions in English,
Italian, Serbo-Croatian, and Turkish. Journal of Child Language 6: 529-45.
Jusczyk, P. 1981. Infant speech perception: a critical appraisal. In P. Eimas & J.
Miller (eds.) Perspectives on the study of speech. Hillsdale, NJ: Erlbaum.
Jusczyk, P., Shea, S. & A s h , R. 1984. Linguistic experience and infant speech
perception: a re-examination of Eilers, Gavin and Oller (1982). Journal of Child
Kaczmarek, L. 195311. A general view of the formation of child language. In
Bar-Adon & Leopold (1971: 13M).Translation of original Polish.
Kahn, L. 1986. Basics of phonological analysis: a programmed learning text. San
Diego, Calif.: College-Hill Press.
Kamil, M. L. & Rudegeair, R. E. 1972. Methodological improvements in the
assessment of phonological discrimination in children. Child Development 43:
1087-91.
Kaplan, E. & Kaplan, G. 1971. The prelinguistic child. In J. Eliot (ed.) Human
development and cognitive processes. New York: Holt, Rinehart & Winston,
pp. 358-81.
Karmiloff-Smith, A. 1979. A functional approach to child language: a study of
determiners and reference. Cambridge: Cambridge University Press.
Katz, N., Baker, E. & Macnamara, J. 1974. What’s in a name? A study of how
children learn common and proper names. Child Development 45: 469-73.
Kaye, K. 1980a. Why don’t we talk ‘baby talk‘ to babies. Journal of Child Language
7: 489-507.
1980b. The infant as a projective stimulus. American Journal of Orthopsychiatry.
Kaye, K. & Charney, R. 1981. Conversational asymmetry between mothers and
children. Journal of Child Language 8: 35-49.
534 BIBLIOGRAPHY
Keil, F. C. 1979. Semantic and conceptual development: an ontological perspective.
Cambridge, Mass. : Harvard University Press.
Keil, F. C. & Carroll, J. J. 1980. The child’s acquisition of ‘tall’: implications for an
alternative view of semantic development. Papers and Reports on Child Lan-
guage Development 19: 21-8.
Kent, R. 1981. Articulatory-acoustic perspectives on speech development. In Stark
(1981: 105-26).
Kernan, K. 1969. The acquisition of language by Samoan children. Doctoral
dissertation, University of California, Berkeley.
Kiparsky, P. & Menn, L. 1977. On the acquisition of phonology. In Macnamara
(1977: 47-78).
Klatzky, R., Clark, E. & Macken, M. A. 1973. Asymmetries in the acquisition of
polar adjectives: linguistic or conceptual? Journal of Experimental Child
Klee, T. & Fitzgerald, M. D. 1985. The relation between grammatical development
and mean length of utterance in morphemes. Journal of Child Language 12:
251-69.
Klein, S. 1982. Syntactic theory and the developing grammar: reestablishing the
relationship between linguistic theory and data from language acquisition.
Doctoral dissertation, UCLA.
Klima, E. & Bellugi-Klima, U. 1966. Syntactic regularities in the speech of children.
In Lyons & Wales (1966: 183-208). Reprinted in Bar-Adon & Leopold (1971:
412-24).
KuEera, H. 1983. Roman Jakobson. Language 59: 871-83.
KuEera, H. & Francis, W. 1967. Computational analysis of present-day American
English. Providence, Rhode Island. Brown University Press.
Kuczaj, S. A. 1976. Arguments against Hurford’s ‘Aux copying rule’. Journal of
1978. Why do children fail to overgeneralize the progressive inflection? Journal of
1982a. On the nature of syntactic development. In Kuczaj (1982b: 37-71).
(ed.) 1982b. Language development. Vol. 1. Syntax and semantics. Hillsdale, NJ:
Lawrence Erlbaum.
Kuczaj, S. A. & Barrett, M. (eds.) 1986. The development of word meaning:
progress in cognitive development research. New York and Berlin: Springer-
Verlag.
Kuczaj, S. A. & Brannick, N. 1979. Children’s use of the wh question modal
auxiliary placement rule. Journal of Experimental Child Psychology 28: 43-67.
Kuczaj, S. A. & Maratsos, M. 1975. What children ‘can’ say before they ‘will’.
Merrill-Palmer Quarterly of Behavior and Development 21: 89-111.
1983. Initial verbs of yes-no questions: a different kind of general grammatical
category. Developmental Psychology 19: 440-4.
Kuhl, P. 1980. Perceptual constancy for speech-sound categories in early infancy. In
Yeni-Komshian, Kavanaugh & Ferguson (1980b: 41-66).
1981. Auditory category formation and development of speech perception. In
Stark (1981: 165-83).
Kuhl, P. & Miller, J. D. 1975. Speech perception by the chinchilla: voiced-voiceless
distinction in alveolar-plosive consonants. Science 1%:69-72.
Bibliography 535
Labov, W. & Labov, T. 1978. Learning the syntaxof questions. In R. Campbell & P.
Smith (eds.) Recent advances in the psychology of language: language develop-
ment and mother-child interaction, vol. 4b. New York & London: Plenum,
pp. 1-44.
Ladefoged, P. 1975. A course in phonetics. New York: Harcourt Brace
Jovanovich.
Lange, S. & Larsson, K. 1973. Syntactic development of a Swedish girl, Embla,
between 20 and 42 months of age, I: age 20-25 months. Report No. I, Project
Child Language Syntax. Stockholm: Institutionem for nordiska sprak,
Stockholms Universitet.
Lasky, R. E., Syrdal-Lasky, A. & Klein, R. E. 1975. VOT discrimination by four to
six-and-a-half month old infants from Spanish environments. Journal of Experi-
mental Child Psychology 20: 215-25.
Leech, G. 1970. Towards a semantic description of English. Bloomington, Ind.:
Indiana University Press.
1974. Semantics. Harmondsworth, Middx: Penguin.
Lenneberg, E. H. (ed.) 1964. New directions in the study of language. Cambridge,
Mass.: MIT Press.
1967. Biological foundations of language. New York: Wiley.
Lenneberg, E. H. & Lenneberg, E. (eds.) 1975. Foundations of language develop-
ment, New York: Academic Press.
Lenneberg, E., Rebelsky, G. & Nichols, I. 1965. The vocalizations of infants born
to deaf and to hearing parents. Human Development 8: 23-37.
Leonard, L. 1976. Meaning in child language: issues in the study of early semantic
development. New York: Grune & Stratton.
Leonard, L., Newhoff, M. & Mesalam, L. 1980. Individual differences in early
phonology. Journal of Applied Psycholinguistics 1: 7-30.
Leonard, L., Rowan, L., Morris, L. & Fey, M. 1982. Intra-word phonological
variability in young children. Journal of Child Language 9: 55-69.
Leopold, W. 1939-49. Speech development of a bilingual child: a linguist’s record.
Vol. 1: Vocabulary growth in the first two years (1939). Vol. 2: Sound learning in
the first two years (1947). Vol. 3: Grammar and general problems in the first two
years (1949). Vol. 4: Diary from age two. (1949). Evanston, Ill.: Northwestern
University Press.
1952. Bibliography of child language. Evanston, Ill. : Northwestern University
Press.
Lewis, M. M. 1936151. Infant speech: a study of the beginnings of language. 2nd edn.
New York: Harcourt Brace. (First edition published in 1936.)
1937. The beginning of reference to past and future in a child’s speech. British
Journal of Educational Psychology 7: 39-56.
Lieberman, P. 1975. On the origins of language. New York: MacMillan.
1980. On the development of vowel production in young children. In Yeni-
Komshian, Kavanaugh & Ferguson (1980a: 11342).
Lightfoot, D. 1979. Principles of diachronic syntax. Cambridge: Cambridge Univer-
sity Press.
1982. The language lottery. Towards a biology of grammars. Cambridge, Mass.:
MIT Press.
Limber, J. 1973. The genesis of complex sentences. In Moore (1973: 169-85).
536 BIBLIOGRAPHY
1976. Unravelling competence, performance and pragmatics in the speech of
young children. Journal of Child Language 3: 309-18.
Lipp, E. 1977. The acquisition of Estonian inflections. Journal of Child Language 4:
313-19.
Lisker, L. & Abramson, A. S. 1967. The voicing dimension: some experiments in
comparative phonetics. Proceedings of the Sixth International Congress of
Phonetic Sciences, Prague.
Loban, W. 1976. Language development: kindergarten through twelve. Urbana, Ill.:
National Council of Teachers of English.
Lock, A. (ed.) 1978. Action, gesture and symbol: the emergence of language.
London: Academic Press.
Locke, J. L. 1971. Phonemic perception in 2- and 3-year-old children. Perceptual
and Motor Skills 32: 215-17.
1983. Phonological acquisition and change. New York: Academic Press.
Lovell, K. & Dixon, E. 1967. The growth of the control of grammar in imitation,
comprehension and production. Journal of Child Psychology and Psychiatry 8:
31-9.
Lukens, H . 1894. Preliminary report on the learning of language. Pedagogical
Seminary 3: 424-60.
Lust, B. 1986. Studies in first language acquisition of anaphora: defining the
constraints. Boston, Mass.: D. Reidel.
Lynip, A. 1951. The use of magnetic devices in the collection and analysis of the
preverbal utterances of an infant. Genetic Psychology Monographs 44: 221-62.
Lyons, J. 1968. Theoretical linguistics. Cambridge: Cambridge University Press.
Lyons, J. & Wales, R. (eds.) 1966. Psycholinguisticpapers. Edinburgh: University
of Edinburgh Press.
MacKain, K. S. & Stern, D. N. 1985. The concept of experience in speech
development. In K. E. Nelson (1985: 1-33).
Macken, M. A. 1978. Permitted complexity in phonological development: one
child‘s acquisition of Spanish consonants. Lingua 44: 219-53.
1979. Developmental reorganization of phonology: a hierarchy of basic units of
acquisition. Lingua 49: 11-49.
1980. The child’s lexical representation: the ‘puzzle-puddle-pickle’ evidence.
Journal of Linguistics 16: 1-17.
1986. Representation, rules and overgeneralization in phonology. Unpublished
paper, Stanford University.
Macken, M. A. & Ferguson, C. A. 1983. Cognitive aspects of phonological
development: model, evidence, and issues. In K. E. Nelson (1983: 256-82).
Macnamara, J. 1972. Cognitive basis of language learning in infants. Psychological
Review 79: 1-13. Reprinted in Bloom (1978: 390-406).
(ed.) 1977. Language learning and thought. New York: Academic Press.
1982. Names for things: a study of human learning. Cambridge, Mass.: MIT Press.
MacWhinney, B. 1975. Rules, rote, and analogy in morphological formations by
Hungarian children. Journal of Child Language 2: 65-77.
1976. Hungarian research on the acquisition of morphology and syntax. Journal of
1978. Processing a first language: the acquisition of morphophonology. Mono-
graphs of the Society for Research in Child Development 43, no. 174.
Bibliography 537
1982. Basic syntactic processes. In Kuczaj (1982b: 73-136).
MacWhinney, B. & Bates, E. 1978. Sentential devices for conveying givenness and
newness: a cross-cultural developmental study. Journal of Verbal Learning and
Verbal Behavior 17: 539-58.
Malrieu, P. 1973. Aspects psychologiques de la construction de la phrase chez
I’enfant. Journal de Psychologie 70: 157-74.
Marantz, A. 1982. On the acquisition of grammatical relations. Linguistische
Berichte 8 W 2 : 32-69.
1983. The connection between grammatical relations and lexical categories in
language acquisition and linguistic theory. In Otsu et al. (1973: 165-7).
1984. On the nature of grammatical relations. Cambridge, Mass.: MIT Press.
Maratsos, M. 1973. Decrease in the understanding of the word ‘big’ in preschool
children. Child Development 44: 747-52.
1974a. Children who get worse at understanding the passive: a replication of
%ever. Journal of Psycholinguistic Research 3: 65-74.
1974b. When is a high thing a big one? Developmental Psychology 10: 367-75.
1978, New models in linguistics and language acquisition. In M. Halle, J. Bresnan
& G. Miller (eds.) Linguistic theory and psychological reality. Cambridge,
Mass.: MIT Press, pp. 247-63.
1982. The child’s construction of grammatical categories. In Wanner & Gleitman
(1982: 240-66).
1984. The acquisition of syntax. In J. Flavell & E. Markman (eds.) Carmichael’s
handbook of child psychology, vol. 3,4th edn. New York: Wiley.
Maratsos, M. & Abramovitch, R. 1975. How children understand full, truncated,
and anomalous passives. Journal of Verbal Learning and Verbal Behavior 14:
145-57.
Maratsos, M. & Chalkley, M. A. 1980. The internal language of children’s syntax:
the ontogenesis and representation of syntactic categories. In K. E. Nelson
(1980: 127-214).
Maratsos, M. & Kuczaj, S. A. 1978. Against the transformational account: a simpler
analysis of auxiliary overmarkings. Journal of Child Language 5 : 337-45.
Maratsos, M., Kuczaj, S . A., Fox, D. & Chalkley, M. A. 1979. Some empirical
studies in the acquisition of transformational relations: passives, negatives, and
the past tense. In W. Collins (ed.) Children’s language and communication.
Minnesota Symposia on Child Psychology. Hillsdale, NJ: Erlbaum, pp. 1-45.
Mazurkewich, I. 1982. Second language acquisition of the dative alternation and
markedness: the best theory. Doctoral dissertation, Universitt de Montreal.
1984. The acquisition of the dative alternation by second language learners and
linguistic theory. Language Learning 34: 91-109.
Mazurkewich, I. & White, L. 1984. The acquisition of the dative alternation:
unlearning overgeneralizations. Cognition 16: 261-83.
McCarthy, D. 1930. The language development of the preschool child. Institute of
Child Welfare Monograph Series 4. Mineapolis: University of Minnesota Press.
1954. Language development in children. In Carmichael (ed.) Manual of child
psychology. New York: John Wiley, pp. 492-630.
McNeill, D. 1966a. Developmental psycholinguistics. In Smith & Miller (1966:
15-84).
1966b. The creation of language by children. In Lyons & Wales (1966: 99-115).
538 BIBLIOGRAPHY
1970a. The acquisition of language. New York: Harper & Row.
1970b. The development of language. In P. Mussen (ed.) Carmichael’s manual of
childpsychology, vol. 1, pp. 1061-1161.
1971. The capacity for grammatical development in children. In Slobin (1971:
17-40).
Mead, M. 1963. Sex and temperament in three primitive societies. New York:
Morrow.
Menn, L. 1971. Phonotactic rules in beginning speech. Lingua 26: 225-51.
1978. Pattern, control and contrast in beginning speech, a case study in the
development of word form and word function. Bloomington, Ind.: Indiana
University Linguistics Club.
1980. Phonological theory and child phonology. In Yeni-Komshian, Kavanaugh
& Ferguson (1980: 2341).
1983. Development of articulatory, phonetic, and phonological capabilities. In B.
Butterworth (ed.) Language production: development, writing and other lan-
guageprocesses. Vol. 2. New York: Academic Press, pp. 3-50.
Menyuk, P. 1969. Sentences children use. Cambridge, Mass: MIT Press.
1971. The acquisition and development of language. Englewood Cliffs, NJ:
Prentice-Hall.
Menyuk, P. & Anderson, S. 1969. Children’s identification and reproduction of lwl,
Id, and Ill, Journal of Speech and Hearing Research 12: 39-52.
Menyuk, P. & Menn, L. 1979. Early strategies for the perception and production of
words and sounds. In Fletcher & Garman (1979: 49-70).
Mervis, C. & Canada, K. 1983. On the existence of competence errors in early
comprehension: a reply to Fremgen & Fay and Chapman & Thomson. Journal of
Miller, G. & McNeill, D. 1969. Psycholinguistics. In G. Lindzey & E. Aaronson
(eds.) Handbook of social psychology. Reading, Mass: Addison-Wesley,
pp. 666-794.
Miller, J. 1981. Assessing languageproduction in children. Baltimore, Md.: Univer-
sity Park Press.
Miller, J., Chapman, R., Bronston, M. & Reichle, J. 1980. Language comprehen-
sion in sensorimotor stages V and VI. Journal of Speech and Hearing Research
23: 284-311.
Miller, J. & Yoder, D. 1974. An ontogeneticteaching strategy for retarded children.
In Schiefelbusch & Lloyd (1974: 505-28).
1984. Language comprehension text (clinical edition). Baltimore, Md. : University
Park Press.
Miller, P. D. 1972. Vowel neutralization and vowel reduction. Papers from the
Eighth Regional Meeting, Chicago Linguistic Society, pp. 482-9.
Miller, W. & Ervin, S. 1964. The development of grammar in child language. In
Bellugi & Brown (1964: 9-34).
Minifie, F. & Lloyd, L. (eds.) 1978. Communicative and cognitive abilities in early
language assessment. Baltimore, Md. : University Park Press.
Moerk, E. 1980. Relationships between input frequencies and children’s language
acquisition: a reanalysis of Brown’s data. Journal of Child Language 7: 105-18.
1981. To attend or not to attend to unwelcome reanalyses? A reply to Pinker.
Bibliography 539
Moffitt, A. 1971. Consonant cue perception by 22 twenty-four-week-old infants.
Child Development 42: 717-32.
Moore, T. (ed.) 1973. Cognitive development and the acquisition of meaning. New
York: Academic Press.
Morehead, D. M. & Ingram, D. 1973. The development of base syntax in normal and
linguisticallydeviant children. Journalof Speech and Hearing Research 16:330-52.
Morehead, D. M. & Morehead, A. E. (eds.) 1976. Normal and dejicient child
language. Baltimore, Md.: University Park Press.
Morse, P. 1974. Infant speech perception: a preliminary model and review of the
literature. In Schiefelbusch & Lloyd (1974: 19-53).
Moskowitz, A. 1970. The two-year-old stage in the acquisition of English phono-
logy. Language 46: 426-41.
1971. Acquisition of phonology. Doctoral dissertation, University of California,
Berkeley.
Mowrer, 0. 1960. Learning theory and symbolicprocesses. New York: John Wiley.
Muma, J. 1986. Language acquisition: a functional perspective. Austin, Texas:
Pro-Ed.
Munson, J. & Ingram, D. 1985. Morphology before syntax: a case study from
language acquisition. Journal of Child Language 12: 681-4.
Nelson, K. 1973. Structure and strategy in learning to talk. Monographs of the
Society of Research in Child Development 38, no. 149.
1974. Concept, word and sentence: interrelations in acquisition and development.
Psychological Review 81: 267-85.
1975. The nominal shift in semantic-syntactic development. Cognitive Psychology
7: 461-79.
1986. Event knowledge: structure and function in development. Hillsdale, NJ:
Erlbaum.
Nelson, K. E. 1977. Facilitating children’s syntax acquisition. Developmental
(ed.) 1978. Children’s language. Vol. 1 . New York: Gardner Press.
(ed.) 1980a. Children’s language. Vol. 2 . New York: Gardner Press.
1981. Toward a rare event cognitive comparison theory of syntax acquisition:
insights from work with recasts. In Dale & Ingram (1981: 22940).
(ed.) 1984. Children’s language. Vol. 4 . Hillsdale, NJ: Erlbaum.
(ed.) 1985. Children’s language. Vol. 5 . Hillsdale, NJ: Erlbaum.
1981. Toward a rare-event comparison theory of syntax acquisition. In Dale &
Ingram (1981).
Nelson, K. E. & Baker, N. 1984. Recasting. First Language 5: 3-21.
Nelson, K. E. & Bonvillian, J. (1973). Concepts and words in the 18-month-old:
acquiring concept names under controlled conditions. Cognition 2: 435-50.
Nelson, K. E., Carskaddoq, G., & Bonvillian, J. 1973. Syntax acquisition: impact of
experimental variation in adult verbal interaction with the child. Child Develop-
ment 44: 497-504.
Nelson, K. E., Denninger, M., Bonvillian, J., Kaplan, B., & Baker, N. 1984.
Maternal input adjustments and non-adjustments as related to children’s linguis-
tic advances and to language acquisition theories. In A. D. Pelligrini & T. D.
Yawkey (eds.) The development of oral and written languages: readings in
developmental applied linguistics. New York: Ablex, pp. 31-56.
540 BIBLIOGRAPHY
Newmeyer, F. 1983. Grammatical theory: its limits and its possibilities. Chicago, Ill. :
University of Chicago Press.
Newport, E., Gleitman, H. & Gleitman, L. 1977. Mother, I’d rather do it myself
some effects and non-effects of maternal speech style. In Snow & Ferguson
(1977: 101-49).
Nice, M. 1917. The speech development of a child from eighteen months to six
years. Pedagogical Seminary 24: 204-43.
1920. Concerning all-day conversations. Pedagogical Seminary 27: 166-77.
1925. Length of sentences as a criterion of child’s progress in speech. Journal of
Educational Psychology 16: 370-9.
Ninio, A. & Bruner, J. 1978. The achievement and antecedents of labelling. Journal
Nokony, A. 1978. Word and gesture by an Indian child. In Lock (1978: 291-307).
Ochs, E. 1979. Transcription as theory. In Ochs & Schiefflin (1979: 43-72).
1982a.Ergativity and word order in Samoan child language. Language 58: 646-71.
1982b. Talking to children in Western Samoa. Language in Society 11: 77-104.
Ochs, E. & Schiefflin, B. (eds.) 1979. Developmental pragmatics. New York:
Academic Press.
O’Donnell, R. C., Griffin, W. J. & Norris, R. C. 1967. Syntax of kindergarten and
elementary school children. Champaign, Ill.: National Council of Teachers.
O’Grady, W. 1986. Principles of grammar and learning. Chicago, Ill.: University of
Chicago Press.
O’Grady, W., Suzuki-Wei, Y. & Cho, S. W. 1986. Directionality preferences in the
interpretation of anaphora: data from Korean and Japanese. Journal of Child
Oksaar, 0. 1983. Language acquisition in the early years. New York: Saint Martin’s
Press. Translated from the German original of 1977.
Oller, D. K. 1980. The emergence of the sounds of speech in infancy. In
Yeni-Komshian, Kavanaugh & Ferguson (1980: 93-112).
1981. Infant vocalizations: exploration and reflexivity. In Stark (1981: 85-103).
Oller, D. K. & Eilers, R. 1982. Similarity of babbling in Spanish- and English-
learning babies. Journal of Child Language 9: 565-77.
Oller, D. K., Eilers, R., Bull, D. & Carney, A. 1985. Prespeech vocalizations of a
deaf infant: a comparison with normal metaphonological development. Journal
of Speech and Hearing Research 28: 47-63.
Oller, D. K., Wieman, L., Doyle, W. & Ross, C. 1976. Infant babbling and speech.
Olmsted, D. 1971. Out of the mouth of babes. The Hague: Mouton.
Olsen-Fulero, L. 1982. Style and stability in mother conversational behaviour: a
study of individual differences. Journal of Child Language 9: 543-64.
O’Shea, M. 1907. Linguistic development and education. London: Macmillan.
Otsu, Y., van Riemsdijk, H., Inoue, K., Kamio, A. & Kawasaki, N. (eds.) 1983.
Studies in generative grammar and language acquisition. Tokyo: Editorial
Committee. (Inside jacket states that copies can be ordered from N. Kawasaki,
c/o Division of Languages, International Christian University, Tokyo.)
Park, T.-Z. 1978. Plurals in child speech. Journal of Child Language 5: 237-50.
Pelsrna, J. 1910. A child vocabulary and its development. Pedagogical Seminary 17:
328-69.
Bibliography 541
Penman, R., Cross, T., Milgrom-Friedman, J. & Meares, R. (1983). Mothers’
speech to prelinguistic infants: a pragmatic analysis. Journal of Child Language
10: 17-34.
Perlmutter, D. 1983. Studies in relational grammar 1. Chicago, Ill.: University of
Chicago Press.
Perroni Semoes, M. & Stoel-Gammon, C. 1979. The acquisition of inflections in
Portuguese: a study of the development of person markers on verbs. Journal of
Peshkovskiy, A. M. 1925. Desyat’ tysyach Zvukov. Sb statey.
Peters, A. 1977. Language learning strategies. Language 53: 56-73.
1983. The units of language acquisition. Cambridge: Cambridge University Press.
Petretic, P. & Tweney, R. 1977. Does comprehension precede production? The
development of children’s responses to telegraphic sentences of varying gram-
matical adequacy. Journal of Child Language 4: 201-9.
Piaget, J. 1929. The child’s conception of the world. New York: Harcourt Brace.
1948. Play, dreams and imitation in childhood. New York: Norton.
1952. The origins of intelligence in children. New York: International Universities
Press.
1954. The construction of reality in the child. New York: Basic Books.
1955. The language and thought of the child. Cleveland, 111.: World Publishing Co.
1971. Biology and knowledge. Chicago, Ill.: University of Chicago Press.
Piaget, J. & Inhelder, B. 1967. The child’s conception of space. New York: Norton.
1969. The psychology of the child. New York: Basic Books.
Piattelli-Palmarini, M. (ed.) 1980. Language and learning: the debate between Jean
Piaget and Noam Chomsky. Cambridge, Mass.: Harvard University Press.
Pinker, S. 1981. On the acquisition of grammatical morphemes. Journal of Child
Language 8: 477-84.
1982. A theory of the acquisition of lexical interpretive grammars. In J. Bresnan
(ed.) The mental representation of grammatical relations. Cambridge, Mass. :
MIT Press.
1984. Language learnability and language development. Cambridge, Mass. :
Harvard University Press.
Pitcher, E. G. & Prelinger, E. 1963. Children tellstories. New York: International
Universities Press.
Pollock, F. 1878. An infant’s progress in language. Mind 3: 392-401.
Prather, E. M., Hedrick, D. L. & Kern, C. A. 1975. Articulation development in
children aged two to four years. Journal of Speech and Hearing Disorders 40:
179-91.
Preyer, W. 1889. The mind of the child. New York: Appleton. Translation of
original German edn of 1882.
Prideaux, G. 1976. A functional analysis of English question acquisition: a response
to Hurford. Journal of Child Language 3: 417-22.
Pye, C. 1980. The acquisition of grammatical morphemes in Quicht Mayan.
Doctoral dissertation, University of Pittsburgh.
1983a. Mayan telegraphese: intonational determinants of inflectional develop-
ment in Quicht Mayan. Language 59: 583-604.
1983b. Mayan motherese: an ethnography of Quicht Mayan speech to young
children. Unpublished paper.
542 BIBLIOGRAPHY
1985. Review of Language learnability and language development by Pinker.
Language 61: 903-7.
1986. QuichC Mayan speech to children. Journal of Child Language 13: 85-100.
1987. Language loss in the Chilcotin. Unpublished paper, University of Kansas.
Pye, C., Ingram, D. & List, H. 1987. A comparison of initial consonant acquisition
in English and QuichC. In K. E. Nelson & A. Van Kleeck (eds.) Children’s
language, vol. 6: 175-90 Hillsdale, NJ: Erlbaum.
Radford, A. 1981. Transformational syntax: a student’s guide to Chomsky’s
Extended Standard Theory. Cambridge: Cambridge University Press.
Ramer, A. 1976. Syntactic styles in emerging speech. Journal of Child Language 3:
49-62.
Ratner, N. & Bruner, J. 1978. Games, social exchange, and the acquisition of
language. Journal of Child Language 5: 391-401.
Ravn, K. E. & Gelman, S . A. 1984. Rule usage in children’s understanding of ‘big’
and ‘little’. Child Development 55: 2141-50.
Reinhart, T. 1976. The syntactic domain of anaphora. Doctoral dissertation, MIT.
1981. Definite NP anaphora and c-command domains. Linguistic Inquiry 12:
605-35.
1983. Anaphora and semantic interpretation. Chicago, Ill.: University of Chicago
Press.
Rescorla, L. 1980. Overextensions in early language development. Journal of Child
Language 7: 321-35.
Rice, M. L. & Kemper, S . 1984. Child language and cognition. Baltimore:
University Park Press.
Rodd, L. & Braine, M. D. S. 1970. Children’s imitations of syntactic constructions
as a measure of linguistic competence. Journal of Verbal Learning and Verbal
Behavior 10: 430-43.
Rodgon, M. 1976. Single-word usage, cognitive development and the beginnings of
combinatorial speech. Cambridge: Cambridge University Press.
Rodgon, M., Jankowski, W. & Alenskas, L. 1977. A multi-functional approach to
single-word usage. Journal of Child Language 4: 23-44.
Roeper, T. 1981. In pursuit of a deductive model of language acquisition. In Baker
& McCarthy (1981: 121-50).
Roeper, T. & Williams, E. (eds.) 1982. Parameter setting. Dordrecht: D. Reidel.
Rosch, E. 1973. On the internal structure of perceptual and semantic categories. In
Moore (1973: 111-44).
1975. Cognitive representation of semantic categories. Journal of Experimental
Psychology: General 104: 192-233.
1978. Principles of categorization. In E. Rosch & B. Lloyd (eds.) Cognition and
categorization. Hillsdale, NJ: Erlbaum.
Rosch, E. & Mervis, C. 1975. Family resemblances: studies in the internal structure
of categories. Cognitive Psychology 7: 573-605.
Rosch, E., Mervis, C., Gray, W., Johnson, D. & Boyes-Braem, P. 1976. Basic
objects in natural categories. Cognitive Psychology 8: 382-439.
Roussey, C. 1899-1900. Notes sur I’apprentissage de la parole chez un enfant. La
Parole 1: 790-880,2: 23-40, 86-97.
RCke-Dravipa, V. 1973. On the emergence of inflection in child language: a
contribution based on Latvian data. In Ferguson & Slobin (1973: 252-67).
Bibliography 543
Rumelhart, D. & McClelland, J. 1986. On learning the past tenses of English verbs.
In J. McClelland & D. Rumelhart (eds.) Parallel distributed processing: explor-
ations in the microstructure of cognition. Vol. 2. Psychological and biological
models. Cambridge, Mass.: MIT Press, pp. 216-71.
Ryan, J. (1973). Interpretation and imitation in early language acquisition. In R.
Hinde & J. Hinde (eds.) Constraints on learning: limitations and predispositions.
London: Academic Press, pp. 427-44.
Sachs, J. 1977. The adaptive significance of linguistic input to prelinguistic infants.
In Snow & Ferguson (1977: 51-61).
Sachs, J. & Truswell, L. 1978. Comprehension of two-word instructions by children
in the one-word stage. Journal of Child Language 5: 17-24.
Samarin, W. 1967. Field linguistics: a guide to linguisticfield work. New York: Holt,
Rinehart & Winston.
Sander, E. K. 1961. When are speech sounds learned? Journal of Speech and
Hearing Disorders 37: 55-63.
Scarborough, H. & Wyckoff, J. 1986. Mother, I’d still rather do it myself: some
further non-effects of ‘motherese’. Journal of Child Language 13: 431-7.
Schaerlaekens, A. 1973. The two-word sentence in child language development. The
Hague: Mouton.
Schiefelbusch, R. (ed.) 1978. Bases of language intervention. Baltimore, Md. :
University Park Press.
Schiefelbusch, ,R. & Bricker, D. (eds.) 1981. Early language: acquisition and
intervention. Baltimore, Md. : University Park Press.
Schiefelbusch, R. & Lloyd, L. (eds.) 1974. Language perspectives: acquisition,
retardation, and intervention. Baltimore, Md.: University Park Press.
Schiefflin, B. 1979. Getting it together: an ethnographic approach to the study of the
development of communicativecompetence. In Ochs & Schiefflin (1979: 73-108).
Schlesinger, I. M. 1971. The production of utterances and language acquisition. In
Slobin (1971: 63-101).
1974. Relational concepts underlying language. In Schiefelbusch & Lloyd (1974:
129-51).
1977. Production and comprehension of utterances. Hillsdale, NJ: Erlbaum.
1982. Steps toward language: toward a theory of native language acquisition.
Hillsdale, NJ: Erlbaum.
Schwartz, R. & Camarata, S. 1985. Examining relationships between input and
language development: some statistical issues. Journal of Child Language 12:
199-207.
Schwartz, R. & Leonard, L. 1982. Do children pick and choose? An examination of
phonological selection and avoidance in early acquisition. Journal of Child
Language 9: 319-36.
Schwartz, R. & Terrell, B. 1983. The role of input frequency in lexical acquisition.
Scollon, R. 1976. Conversations with a one year old. Honolulu: University Press of
Hawaii.
Sheldon, A. 1974. The parallel function in the acquisition of relative clauses in
English. Journal of Verbal Learning and Verbal Behavior 13: 272-81.
Sherrod, K., Friedman, S., Crawley, S., Drake, D. & Devieux, J. 1977. Maternal
languageto prelinguisticinfants: syntactic aspects. Child Development48: 1662-5.
544 BIBLIOGRAPHY
Shibamoto, J. S. & Olmsted, D. 1978. Lexical and syllabic patterns in phonological
acquisition. Journal of Child Language 5 : 417-57.
Shipley, E., Smith, C. & Gleitman, L. 1969. A study in the acquisition of language:
free responses to commands. Language 45: 32242.
Shirley, M. M. 1933. The first two years: a study of twenty-five babies. Vol. 2.
Institute of Child Welfare Monograph Series. Minneapolis, Minn. : University of
Minnesota Press.
Shriberg, L. & Kwiatkowski, J. 1980. Natural process analysis (NPA). New York:
John Wiley.
Shvachkin, N. 1948/73. The development of phonemic speech perception in
children. In Ferguson & Slobin (1973: 91-127).
Sinclair, H. 1971. Sensorimotor action patterns as a condition for the acquisition of
syntax. In Huxley & Ingram (1971: 121-30).
Sinclair, H. & Ferreiro, E. 1970. Etude gCnCtique de la comprehension, production
et rePCtition des phrases a mode passif. Archives de Psychologie 40: 1 4 2 .
Skinner, 0. F. 1957. Verbal behavior. New York: Appleton-Century-Crofts.
Slobin, D. 1966. Abstracts of Soviet studies of child language. In Smith & Miller
(1966: 361-86).
(ed.) 1967. Afield manual for cross-cultural study of the acquisition of communica-
tive competence. Berkeley, Calif. : University of California.
1968. Recall of full and truncated passive sentences in connected
discourse. Journal of Verbal Learning and Verbal Behavior 7: 876-81. ,
1970. Universals of grammatical development. In G. Flores D’Arcais & W. Levelt

(eds.) Advances in psycholinguistics. Amsterdam: North-Holland, pp. 174-86.
(ed.) 1971. The ontogenesis of grammar. New York: Academic Press.
1972. Leopold’s bibliography of child language. Bloomington, Ind.: Indiana
University Press.
1973. Cognitive prerequisites for the development of grammar. In Ferguson &
Slobin (1973: 175-208).
1977. Language change in childhood and history. In Macnamara (1977: 185-214).
1982. Universal and particular in the acquisition of language. In Wanner &
Gleitman (1982: 128-70).
(ed.) 1986. Cross-linguistic study of language acquisition. Hillsdale, NJ: Erlbaum.
Smith, C. 1970. An experimental approach to children’s linguistic competence. In
Hayes (1970: 109-35).
Smith, F. & Miller, G. A. 1966. The genesis of language. Cambridge, Mass.: MIT
Press.
Smith, M. 1926. An investigation of the development of the sentence and the extent
of vocabulary in young children. University of Iowa Studies in Child Welfare
vol. 3, no. 5.
1933. Grammatical errors in the speech of preschool children. Child Development
4: 183-90.
1935. A study of some factors influencing the development of the sentence in
preschool children. Journal of Genetic Psychology 46: 182-212.
Smith, N. V. 1973. The acquisition of phonology: a case study. Cambridge:
Cambridge University Press.
1982. Review of Language acquisition, by P. Fletcher & M. Garman (eds.).
Language 58: 470-4.
Bibliography 545
Snow, C. E. 1977a. Mothers’ speech research: from input to interaction. In Snow &
Ferguson (1977: 31-49).
1977b. The development of conversation between mothers and babies. Journal of
Snow, C. E. & Ferguson, C. A. 1977. Talking to children: language input and
acquisition. Cambridge: Cambridge University Press.
Snow, C. E. & Goldfeld, B. 1983. Turn the page please: situation-specific language
acquisition. Journal of Child Language 10: 551-70.
Snow, C. E., Smith, N. S. & Hoefnagel-Hohle, M. 1980. The acquisition of some
Dutch morphological rules. Journal of Child Language 7: 539-53.
Snyder, L., Bates, E. & Betherington, I. 1981. Content and context in early lexical
development. Journal of Child Language 8: 565-82.
Solan, L. 1983. Pronominal reference: child language and the theory of grammar.
Dordrecht: D. Reidel.
1987. Parameter setting and the development of pronouns and reflexives. In
Roeper & Williams (1987: 189-210).
Solberg, M. E. 1978. Constant and structure contingent characteristics of the
language array. In R. Campbell & P. Smith (eds.) Recent advances in the
psychology of language: language development and mother-child interaction.
New York: Plenum.
Stampe, D. 1968. Yes, Virginia. . . Unpublished paper presented to the 4th Annual
Regional Meeting of the Chicago Linguistic Society.
1969. The acquisition of phonemic representation. Proceedings of the Fifth
Regional Meeting of the Chicago Linguistic Society, pp. 433-44.
1973. A dissertation on natural phonology. Doctoral dissertation, University of
Chicago.
Stark, R. 1980. Stages of development in the first year of life. In Yeni-Komshian,
Kavanaugh & Ferguson (1980: 73-92).
(ed.) 1981. Language behavior in infancy and early childhood. New York:
Elsevier North-Holland.
Stark, R., Rose, S. & McLagan, M. 1975. Features of infant sounds: the first eight
weeks of life. Journal of Child Language 2: 205-21.
Starr, S. 1975. The relationship of single words to two-word sentences. Child
Development 46: 701-8.
Stern, C. & Stern, W. 1907. Die Kindersprache. Leipzig: Barth.
Stern, D., Spierker, S., Barnett, R. & MacKain, K. 1983. The prosody of maternal
speech: infant age and context related changes. Journal of Child Language 10:
17-34.
Stern, W. 1924. Psychology of early childhood up to the sixth year of age. New York:
Holt.
Stockman, I., Woods, D. & Tishman, A. 1981. Listener agreement on phonetic
segments in early childhood vocalizations. Journal of Psycholinguistic Research
10: 593-617.
Stoel-Gammon, C. 1986. Underlying representations in child phonology: current
models, sources of evidence, and future research. Unpublished paper, Univer-
sity of Washington.
Stoel-Gammon, C. & Cooper, J. 1984. Patterns of early lexical and phonological
development. Journal of Child Language 11: 247-71.
546 BIBLIOGRAPHY
Stoel-Gammon, C. & Dum, C. 1985. Normal and disordered phonology in children.
Baltimore, Md.: University Park Press.
Strange, W. & Broen, P. A. 1980. Perception and production of approximant
consonants by 3-year-olds: a first study. In Yeni-Komshian, Kavanaugh &
Ferguson (1980b: 117-54).
Streeter, L. A. 1976. Language perception of two-month-old infants shows effects of
both innate mechanisms and experience. Nature 259: 39-41.
Strohner, J. & Nelson, K. E. 1974. The young child’s development of sentence
comprehensions: influence of event probability, nonverbal context, syntactic
form, and strategies. Child Development 45: 564-76.
Sudhalter, V. & Braine, M. D. S. 1985. How does comprehension of passives
develop? A comparison of actional and experiential verbs. Journal of Child
Taine, H. 1877. On the acquisition of language by children. Mind 2: 252-9.
Tavakolian, S. (ed.) 1981a. Linguistic theory and language acquisition. Cambridge:
Mass.: MIT Press.
1981b. The conjoined-clause analysis of relative clauses. In Tavakolian (1981a:
167-87).
Taylor-Browne, K. 1983. Acquiring restrictions on forward anaphora: a pilot study.
Calgary Working Papers in Linguistics 9: 75-100.
Templin, M. 1957. Certain language skills in children. University of Minnesota
Institute of Child Welfare Monograph Series 26. Minneapolis: University of
Minnesota Press.
Thomson, J. R. & Chapman, R. 1977. Who is ‘Daddy’ revisited: the status of
two-year-olds’ over-extended words in use and comprehension. Journal of Child
Language 4: 359-75.
Thorndike, E. & Lorge, I. 1944. The teacher’s word book of 30,000 w o r k . New
York: Teachers College, Columbia University.
Trehub, S., Bull, D. & Schneider,B. 1981. Infant speech and nonspeech perception:
a review and reevaluation. In R. Schiefelbusch & D. Bricker (1981: !9-50).
Trubetzkoy, N. 1969. Principles of phonology. Berkeley and Los Angeles: Univer-
sity of California Press. Translation of original German version.
Tse, S.-M. 1982. The acquisition of Cantonese phonology. Doctoral dissertation,
The University of British Columbia.
Turner, E. & Rommetveit, R. 1967. The ’acquisition of sentence voice and
reversibility. Child Development 38: 649-60.
Tyack, D. & Gottsleben, R. 1974. Language sampling, analysis, and training. Palo
Alto, Calif.: Consulting Psychologists Press.
Tyack, D. & Ingram, D. 1977. Children’s production and comprehension of
questions. Journal of Child Language 4: 211-24.
Uzgiris, I. & Hunt, J. McV. 1975. Assessment in infancy: ordinal scales of
psychological development. Urbana, Ill. : University of Illinois Press.
Varma, T. 1979. Stage I speech of a Hindi-speaking child. Journal of Child
Language 6: 167-73.
Velten, H. 1943. The growth of phonemic and lexical patterns in infant speech.
Vihman, M. 1981. Phonology and the development of the lexicon: evidence from
children’s errors. Journal of Child Language 8: 239-64.
Bibliography 547
Vinson, J. 1915. Observations sur le dheloppement du langage chez I’enfant. Revue
Linguistique 49: 1-39.
Vygotsky, L. 1962. Thought and language. Cambridge, Mass.: MIT Press. Trans-
lation of original Russian version of 1934.
Wall, R. 1972. Review of Language development: form and function in emerging
grammars, by L. Bloom. Language Sciences 19: 21-7.
Wang, W. & Crawford, J. 1960. Frequency studies of English consonants. Language
and Speech 3: 131-9.
Wanner, E. & Gleitman, L. (eds.) 1982. Language acquisition: the state of the art.
Wasow, T. 1983. Some remarks on developmental psycholinguistics. In Otsu et al.
(1983: 191-6).
Waterson, N. 1970. Some speech forms of an English child: a phonological study.
Transactions of the Philological Society 1: 1-24.
1971. Child phonology: a prosodic view. Journal of Linguistics 7: 179-211.
Waterson, N. & Snow C. (eds.) 1978. The development of communication. New
York: John Wiley.
Watt, W. 1970. On two hypotheses concerning psycholinguistics. In Hayes (1970:
137-220).
Weiner, F. 1979. Phonological process analysis. Baltimore, Md. : University Park
Press.
Weir, R. 1962. Language in the crib. The Hague: Mouton.
Weist, R. 1983. Prefix versus suffix information processing in the comprehension of
tense and aspect. Journal of Child Language 10: 85-96.
Weist, R., Wysocka, A. Witkowska-Stadnik, K., Buczowska, E. & Konieczna, E.
1984. The defective tense hypothesis: on the emergence of tense and aspect in
child Polish. Journal of Child Language 11: 347-74.
Weist, R. & Konieczna, E. 1985. Affix processing strategies and linguistic systems.
Wellman, B., Case, I., Mengert, I. & Bradbury, D. 1931. Speech sounds of young
children. University of Iowa Studies in Child Welfare 5 .
Werker, J. & Tees, R. 1984. Cross-language speech perception: evidence for
perceptual reorganization during the first year of life. Infant Behavior and
Development 7: 49-63.
Werner, H. & Kaplan, B. 1963. Symbol formation. New York: John Wiley.
Wexler, K. 1982. A principle theory for language acquisition. In Wanner &
Gleitman (1982: 288-315).
Wexler, K. & Chen, Y.-C. 1985. The development of lexical anaphors and
pronouns. Papers and Reports on Child Language Development 24: 138-49.
Wexler, K. & Culicover, P. 1980. Formal principles of language acquisition.
Wexler, K. & Manzini, M. R. 1987. Parameters and learnability in binding theory.
In Roeper & E. Williams (1987: 77-89).
White, L. 1982. Grammatical theory and language acquisition. Hingham, Mass.:
Kluwer.
Whitehurst, G., Ironsmith, M. & Goldfein, M. 1974. Selective imitation of the
passive construction through modelling. Journal of Experimental Child Psychol-
ogy 17: 288-302.
548 BIBLIOGRAPHY
Wilkinson, A. 1971. The foundations of language: talking and reading in young

children. Oxford University Press.
Wilson, L. 1898. Bibliography of child study. Pedagogical Seminary 5: 541-89.
Winitz, H., 1969. Articulation, acquisition and behavior. New York: Appleton-
Century-Crofts.
Wittgenstein, L. 1953. Philosophical investigations. New York: Macmillan.
Yeni-Komshian, G., Kavanaugh, J. & Ferguson, C. (eds.) (1980a). Child phono-
logy. Vol. 1. Production. New York: Academic Press.
(eds.) 1980b. Child phonology: Vol. 2. Perception. New York: Academic Press.
Young, F. 1941. An analysis of certain variables in a developmental study of
language. Genetic Psychology Monographs 23: 3-141.
Zarcbina, M. 1965. Ksztaltownaie sie systemu jeszykowego dziecka. Krakow.
Oddzial w. Krakowie: Polska Akademia Nauk.
Zukow, P. 1982. Transcription systems for video-taped interactions: some advan-
tages and limitations of manual and computer rendering techniques. Journal of
Applied Psycholinguistics 3: 61-79.
Zwicky, A. 1970. A double regularity in the acquisition of English verb morphology.
Papers in Linguistics 3: 411-18.
Author index
Abrahamsen, A., 1977: 30 Barrie-Blackley, S., 1973: 418

Abrahamsen, D., 1938: 212 Bartlett, E., 1976: 420, 425
Abramson, A. S.: see Lisker, L. See Carey, S.
Ament, W., 1899: 192,212 Barton, D., 1975: 345
Amidon, A., 1976: 418 1976a: 345,347,432
Amidon, A. & Carey, S., 1972: 418 1976b: 344,347
Anglin, J., 1970: 398 1978: 3445,347,357
1977: 148, 159,398,400,402-3, 1980: 179,233,342-5,433
405-10,434 Bateman, W., 1916: 9
Anisfeld, M., 1984: 4, 58, 116, 138 Bates, E.: see Snyder, L.;
Antinucci, F. & Miller, R., 1976: 328 MacWhinney, B.
Antinucci, F. & Parisi, D., 1973: 286 Bates, E., Camaioni, L. &
Arlman-Rupp, A., van Niekerk de Volterra, V., 1975: 168-9
Haan, D., & van de Bates, E. & MacWhinney, B., 1979:
Sandt-Koenderman, M., 1976: 494 304-5
Armstrong, S., Gleitman, L. & 1982: 304-7
Gleitman, H., 1983: 396, 431,434 Bausano, M. K. &Jeffrey, W. E.,
Adin, R.: see Jusczyk, P. 1975: 422
A s h , R. & Pisoni, D., 1980: 84-6, 138 Bellugi(-Klima), U., 1967: 457, 461,
Atkinson, M., 1982: 80, 178, 190, 197, 513
212 1971: 457-61,513
See Brown, R.; Klima, E.
Bach, E. & Harms, R. (eds.), 1968: see Bellugi, U. & Brown, R. (eds.), 1964:
Fillmore, C. 23,337
Baker, C. L., 1979: 29,62-3,67 See Brown, R. & Fraser, C.;
Baker, C. L. & McCarthy, J. (eds.), Jenkins, J. & Palermo, D.;
1981: see Dresher, E.; Miller, W. & Ervin, S.
Grimshaw, J., Roeper, T. Benedict, R., 1979: 1414,146, 148,
Baker, E.: see Macnamara, J. 150, 159-62,231,251,336
Baker, N.: see Nelson, K. E. See Della Corta, M.
Bar-Adon, A. & Leopold, W. (eds.), Berko, J., 1958: 43941,513,516
1971: 30-1,57 Berman, R., 1981a: 515
See Kaczmarek, L.; Klima, E. & 1981b: 515
Bellugi-Klima, U.; Jakobson, R. 1982: 515
Barrett, M., 1978: 398,400,405 Berwick, R. & Weinberg, A., 1984:
1982: 232,242-3,398,400,405 473-4,514
See Kuczaj, S. A. Bever, T., 1970: 469,514
549
550 AUTHOR INDEX
Bierwisch, M., 1967: 419 Brannick, N.: see Kuczaj, S. A.
1970: 399 Branston, M.: see Miller, J.
Blewitt, P., 1982: 413,418,426,434 Braunwald, S., 1978: 149-50
Bloom, L., 1970: 22,237,246,248-50, Bresnan, J., 1978: 317
267,272-9,294,296, 308, 329, (ed.) 1982: see Pinker, S .
333,337-8 See Halle, M.
1971: 266 Bretherton, I.: see Snyder, L.
1973: 159, 169,232,237-9,2424, Brewer, W. & Stone, J., 1975: 420
247-9,287,337-8 Brown, H. D., 1971: 478
(ed.), 1978: 30,59 Brown, R., 1958: 148,398,405
See Eimas, P., Siqueland, E., 1968: 459,461-2,514
Jusczyk, P. & Vigorito, J. 1970: 61
Bloom, L., Hood, L. & Lightbown, P., 1973: 22,24,49-53,57,234-5,242,
1974: 245 266,278-9,281,283-8,294-5,
Bloom, L. & Lahey, M. (eds.), 1978: 4 304,325,327,329,331-8,435,
Bloom, L., Lightbown, P. & Hood, L., 437,442,447-8,4514,494
1975: 234-6,245,274,294-302, See Bellugi, U.
304,315,319,328,330,337-8 Brown, R., Cazden, C ., &
Bloomfield, L., 1933: 18-20,26,31 Bellugi, U., 1969: 461
Blount, B., 1969: 281 Brown, R. & Fraser, C., 1963: 441-3,
1972: 138 445,457
Bonvillian, J.: see Nelson, K. E. 1964: 2634,267
Borer, H. & Wexler, K., 1987: 514 Bruner, J., 1975: 130,138
Bowerman, M., 1973a: 236-7,266, See Ninio, A.; Ratner, N.
280-3, 291, 294,329, 337-8 Bruner, J. & Sherwood, V., 1976: 129
1973b: 281,324,329 Buczowska, E.: see Weist, R.
1975: 281,296,338 Buhler, C., 1931: 140
1976: 338 Bull, D.: see Eilers, R.; Oller, D. K.;
1978: 398,403-5,409-13,434 Trehub, S .
1979: 479 Burroughs, G., 1957: 218
1986: 515 Butterfield, E. & Cairns, G . , 1974: 93,
Boyes-Braem, P.: see Rosche, E. 137
Bradbury, D.: see Wellman, B
Braine, M. D. S., 1963a: 22,235-7, Cairns, G . :see Butterfield, E.
245-7,263-7,291,334,337-8 Camaioni, L.: see Bates, E.
1963b: 263 Campbell, D. & Stanley, J., 1963: 226
1971: 28-9,311 Canada, K.: see Mervis, C.
1973: 278 Caramazza, A. & Zuriff, E. (eds.),
1974: 198 1978: 196
1976: 236-7,2454,267,288-94, Carden, G . , 1986: 488,491,514
297,302,308-9,314,318,328-9, Carey, S., 1978: 425
334,331-8 1982: 431,434
See Rodd, L.; Sudhalter, V. See Amidon, A.
Braine, M. D. S. & Hardy, J., 1981: Carey, S . & Bartlett, E., 1978: 432
303 Carney, A.: see Oller, D. K.
Brainerd, C., 1978: 54-8, 127 Carter, A. L., 1975a: 169
Brandenburg, G . , 1915: 9,21 1975b: 169
Branigan, G . , 1979: 239-41 1978a: 169
Author index 551
1978b: 169 Comrie, B., 1981: 436
1979: 174 1983: 436
Case, I.: see Wellman, B. Cooper, J.: see Stoel-Gammon, C.
Caudill, W. & Weinstein, H., 1969: Crain, S. & McKee, C., 1985: 490
129 Crawford, J.: see Wang, W.
Cazden, C., 1968: 446-51,513 Cruttenden, A., 1970: 106,138
1970: 458,513 Culicover, P. & Wilkins, W., 1984:
1972: 4 280,330
See Brown, R.
Chalkley, M. A.: see Maratsos, M. Dale, P., 1976: 4,14,348,458
Chamberlain, A. & Chamberlain, I., 1985: 335-6
1904: 9,30 Dale, P. & Ingram, D. (eds.), 1980:
1905: 9,30 see Nelson, K. E.
1909: 9,30 Darwin, C., 1877: 30
Chamberlain, I.: see Chamberlain, A. Dato, D. (ed.), 1975: see Ingram, D.
Chao, Y.-R., 1951: 386 Davis, E., 1937: 14
Chapman, R., 1981a: 166-8 Davis, H., 1983: 75
1981b: 233 1987: 460, 513
See Miller, J. Day, E., 1932: 14
Charney, R.: see Kaye, K. de Boysson-Bardies, B., Sagart, L. &
Chen, M., 1975: 391 Durand, C., 1984: 113
Chomsky, C., 1969: 71,418,488-9, 514 de Laguna, G., 1927: 231
Chomsky, N., 1957: 23,267,455,474 Della Corta, M., Benedict, R. &
1964: 245,332 Klein, D., 1983: 229-31
1965: 24-5,61,63,267-8,317,329 Denninger, M.: see Nelson, K. E.
1968: 514 de Saussure, F., 1922: 191
1972: 280 Devereux, G., 1949: 129
1975: 64,66, 80,125-6 de Villiers, J. & de Villiers, P., 1973a:
1981: 25,61,63-4,74-5,79,280, 2547,3367,465,468,479
303,331,436,474,484,492 1973b: 453-4
1982: 474 1978: 4
1986: 474 de Villiers, J., Tager-Flusberg, H.,
Chomsky, N. & Halle, M., 1968: 379 Hakuta, K. & Cohen, M., 1979:
Christensen, L.: see Ingram, D. 478,480
Clark, E., 1971: 415-17 Dinnsen, D. A . , 1980: 392
1972: 420 Dockrell, J. E., 1981: 432
1973a: 150, 156,231,397-9,410,434 Donaldson, M. & Wales, R . , 1970: 418
1973b: 41415 Donegan, P. J., 1986: 387
See Clark, H.; Klatzky, R. Donegan, P. J. & Stampe, D., 1979:
Clark, H., 1970: 415 387,433
1973: 339 Dore, J., 1975: 242
Clark, H. & Clark, E., 1977: 4 Dore, J., Franklin, M., Miller, R. T. &
Clark, R., 1974: 308 Ramer, A., 1976: 169, 174,249
1977: 308 Doyle, W.: see Oller, D. K.
1982: 308 Dresher, E., 1981: 65
Clumeck, H., 1982: 3 6 5 , 3 5 7 , 4 3 2 Dunn, C.: see Stoel-Gammon, C.
Cohyen, M.: see de Villiers, J. Durand, C.: see de
Coker, P., 1978: 419 Boysson-Bardies, B.
552 AUTHOR INDEX
Edwards, M. L., 1970: 386 Fernald, C., 1972: 445

1974: 233,3445,3504,356,432 Fey, M.: see Leonard, L.
Edwards, M. L. & Shriberg, L. D., Fillmore, C., 1968: 61, 279-80
1983: 433 Fisher, M., 1934: 14
Eilers, R., 1980: 87 Flavell, J. 1963: 138
See Oller, D. K. Fletcher, P. & Garman, M. (eds.),
Eilers, R. & Gavin, W., 1981: 95 1979: 4
Eilers, R., Gavin, W. & Wilson, W., See Bowerman, M.
1979: 93-5 Fodor, J., 1975: 396,398
Eilers, R. & Oller, D. K . , 1976: 233, Fodor, J., Garrett, M., Walker, E. &
344,35340,432 Parkes, C., 1980: 397
Eilers, R., Oller, D. K., Bull, D. & Fonagy, I., 1972: 23940
Gavin, W., 1984: 95 Fortescue, M., 198415: 496
Eilers, R., Oller, D. K. & Fox, D.: see Maratsos, M.
Ellington, J., 1974: 420 Francis, H., 1969: 69
Eilers, R., Wilson, W. & Moore, M., Franklin, M. see Dore, J.
1976: 88 Fraser, C.: see Brown, R.
1977: 88 Fraser, C . , Bellugi, U. & Brown, R.,
Eimas, P., 1974: 137 1963: 443-6
Eimas, P., Siqueland, E., Jusczyk, P. French, L. & Brown, A., 1977:
& Vigorito, J., 1971: 87, 90-3 418-19
Eisenberg, K.: see Huttenlocher, J. Friedman, W. & Seeley, P., 1976: 418
Elbers, L., 1982: 138 Furrow, D., Nelson, K. &
Elkonin, D. B., 1971: 232 Benedict, H., 1979: 132,226, 508,
Ellington, J.: see Eilers, R. 515
Emonds, J., 1976: 73
Erreich, A . , 1984: 460 Gaer, E. P . , 1969: 478
Erreich, A., Valian, V. & Garman, M.: see Fletcher, P.
Winzemer, J., 1980: 459, 463, 514 Garnica, 0. K., 1971: 344-5
Ervin, S.: see Miller, W. 1973: 233, 344-5
1977: 132
Fant, G.: see Jakobson, R. See Ferguson, C. A.
Fay, D., 1978: 458,463 Garrett, M.: see Fodor, J.
Feagans, L., 1980: 418 Gavin, W.: see Eilers, R.
Ferguson, C. A., 1964: 132 Gazdar, G., Klein, E., Pullum, G. &
1978: 392 Sag, I., 1985: 331
See Macken, M. A.; Snow, C.; Gelman, S. A.: see Ravn, K. E.
Yeni-Komshian, G. Gibson, D.: see Hoek, D.
Ferguson, C. A. & Farwell, C. B., Gilbert, J., 1982: 112
1973: 199-200 Ginsberg, H. & Opper, S., 1969: 138
1975: 2004,207,210--14,233,376, Givbn, T., 1979: 304
378,392-3 Gleitman, H.: see Armstrong, S.;
Ferguson, C. A. & Garnica, O., 1975: Gleitman, L.; Newport, M.
233,433 Gleitman, L., 1981: 507
Ferguson, C. A., & Slobin, D. (eds.), See Armstrong, S.; Newport, M.;
1973: 30,180,232,338 Shipley, E.; Wanner, E.
See Antinucci, F. & Parisi, D.; Gleitman, L., Newport, M. &
Braine, M. D. S. Gleitman, H., 1984: 50610,515
Author index 553
Gleitman, L., Shipley, E. & Smith, C., Hoyer, A. & Hoyer, G., 1924: 103
1978: 162,251 Hunt, J. McV.: see Uzgiris, I.
Goad, H. & Ingram, D., 1988: 233 Hurford, J., 1975: 308,458,460, 463
Goertz, C., 1973: 129 Huttenlocher, J., 1974: 153, 156,
Goldfein, M.: see Whitehurst, G. 164-6,232
Goldstein, U., 1979: 98 Huttenlocher, J., Eisenberg, K. &
Gottlieb, G., 1976: 84 Strauss, S., 1968: 514
Gray, W.: see Rosch, E. Huxley, R. & Ingram, E. (eds.), 1971:
Greenberg, J. & Kuczaj, S., 1982: 398, see Bellugi, U.
400 Hyams, N., 1983: 303,331
Greenfield, P. & Smith, J., 1976: 169, 1984: 303
1748,232,234, 241 1986: 303, 331
Greenfield, P. & Zukow, P., 1978: 306
Gregoire, A., 1937: 10 Ingram, D., 1971: 231-2,241-2
Grimshaw, J., 1981: 303,316 1972: 75, 275
Gruber, J. S., 1967: 331 1974a: 392
1975a: 331 1974b: 386
1975b: 331 1975: 476,478, 514
Guillaume, P., 192718: 58 1976a: 14, 58,340, 366,385,392,433
Gvozdev, A . , 1949: 10,232 1976b: 385
1978: 11, 121, 157
Hakuta, K.: see de Villiers, J. 1979a: 392
Halle, M., Bresnan, J. & Miller G. 1979b: 175,232,241-4
(eds.), 1978: see Bresnan, J.; 1981a: 203, 217,349,372, 292
Carey, S.; Maratsos, M. 1981b: 57-8,418
See Chomsky, N.; Jakobson, R. 1981~:221-2,335
Halliday, M. A. K., 1975: 57, 169-74, 1981d: 76,296,330,3345,451
178,220-2, 232, 303 1981e: 77-8
Hamburger, H. & Crain, S., 1982: 478, 1985a: 25,65,67,75, 80,465
480-3, 490, 514 1985b: 214
Harding, C. & Golinkoff, R., 1979: 1985~:63
168 1985d: 377
Hardy, J.: see Braine, M. D. S. 1986: 394
Harms, R.: see Bach, E. 1988: 203,214,386,394
Hayes, J. R. (ed.), 1970: 338 See Goad, H.; Hoek, D.;
See Smith, C. Munson, J.; Tyack, D.
Hedrick, D. L.: see Prather, E. M. Ingram, D., Christensen, L., Veach, S.
Hodson, B. & Paden, E., 1983: 392 & Webster, B., 1980: 17, 353, 361,
Hoefnagel-Hohle, M.: see Snow, C. E. 366,378,383
Hoek, D., Ingram, D. & Gibson, D., Ingram, D. & Mitchell, G., to appear:
1986: 147, 153, 159-60, 212 17,203
Hoff-Ginsburg, E. & Shatz, M., 1982: Ingram, D. & Shaw, C . , 1981: 488,
515 490-1
Hogan, L., 1898: 9 Ingram, D. & Tyack, D., 1979: 459,
Hood, L.: see Bloom, L. 513
Horgan, D., 1978: 466-9,513-14 Inhelder, B.: see Piaget, J.
Howe, C., 1976: 245,249,277,287, Inhelder, B. & Piaget, J., 1964: 125
338 Inoue, K.: see Otsu, Y.
554 AUTHOR INDEX
Ironsmith, M.: see Whitehurst, G. Klein, S., 1982: 459,463
Irwin, O., 1941-9: 104-5, 138 Klima, E. & Bellugi-Klima, U., 1966:
52,458
Jackendoff, R., 1977: 280,321 Konieczna, E.: see Weist, R.
1983: 26, 126,473 KuEera, H., 1983: 197
Jakobson, R., 1939: 99 KuEera, H. & Francis, W., 1967: 426
1941/68: 98-9,106, 190-7,211-12, Kuczaj, S. A., 1976: 458,460
233,342,368,390,393 1978: 456
Jakobson, R., Fant, G. & Halle, H., 1982a: 303,434
1963: 19% (ed.), 1982b: 153,338; see Clark, R.;
Jakobson, R. & Halle, M., 1956: Greenberg, J. & Kuczaj, S.;
190-1,196-7,233 Hamburger, H. & Crain, S.;
Jeffrey, W. E.: see Bausano, M. K. Kuczaj, S. A.; MacWhinney, B.
Jenkins, J. & Palermo, D., 1964: 21 See Maratsos, M.
Jespersen, O., 1922: 41,44,231 Kuczaj, S. A. & Barrett, M. D. (eds.),
Johnson, C. J., Hardee, W. P. & 1986: 434
Long, S. H.,. 1981: 344-6,432 Kuczaj, S. A. & Brannick, N., 1979:
Johnson, D.: see Rosch, E. 456,460,463
Johnston, J.: 1979: 427-8 Kuczaj, S. A. & Maratsos, M., 1975:
1986: 421,430 445,456-7,513,516
Johnston, J. & Slobin, D., 1979: 1983: 463-4
427-8,430 Kuhl, P., 1980: 138
Jusczyk, P.: see Eimas, P. 1981: 88
Jusczyk, P., Shea, S. & A s h , R., Kuhl, P. & Miller, J. D., 1975: 92, 95
1984: 95
Labov, W. & Labov, T., 1978: 460,
Kaczmarek, L., 1953/71: 57 513-14
Kamio, A.: see Otsu, Y. Ladefoged, P., 1975: 89,215,364,390
Kaplan, B.: see Nelson, K. E. Lange, S. & Larsson, K., 1973: 292
Kaplan, E. & Kaplan, G., 1971: 139 Larsson, K.: see Lange, S.
Karmiloff-Smith, A., 1979: 304-5, 307 Lasky, R. E., Syrdal-Lasky, A. &
Katz, N., Baker, E. & Macnamara, J., Klein, R. E., 1975: 93
1974: 2514,262,336-7 Leech, G., 1970: 415
Kavanaugh, J.: see Yeni-Komshian, G. 1974: 395
Kawasaki, N.: see Otsu, Y. Lenneberg, E., Rebelsky, G. &
Kaye, K., 1980a: 131-2,135-7,219, Nichols, I., 1965: 112
229 Lenneberg, E. H. & Lenneberg, E.
1980b: 137 (eds.), 1975: see Ferguson, C. A.
Kaye, K. & Charney, R., 1981: 135 & Garnica, 0.
Keil, F. C., 1979: 425 Leonard, L., 1976: 286-7,338
Keil, F. C. & Carroll, J. J., 1980: 425 See Schwartz, R.
Kent, R., 1981: 98,109-10 Leonard, L., Newhoff, M. &
Kern, C. A.: see Prather, E. M. Mesalam, L., 1980: 200
Kernan, K., 1969: 281,291 Leonard, L., Rowan, L., Morris, L. &
Klatzky, R., Clark, E. & Macken, Fey, M., 1982: 211
M. A., 1973: 420 Leopold, W., 1939-49: 10,376,378,433
Klein, D.: see Della Corta, M. 1952: 30,168
Klein, R. E.: see Lasky, R. E. See Bar-Adon, A.
A u t h o r index 555
Lewis, M. M., 1936/51: 10,96,103, Marantz, A., 1982: 316
105-6,138 1983: 316
1937: 219-20 1984: 280
Lieberman, P., 1975: 97 Maratsos, M., 1973: 420,422
1980: 101 1974: 420, 422
Lightbown, P.: see Bloom, L. 1978: 75
Lightfoot, D., 1979: 62, 73 1982: 310, 313-14
1982: 126,223,315 See Kuczaj, S. A.
Limber, J., 1973: 514 Maratsos, M. & Chalkley, M. A.,
1976: 327 1980: 303,307,310-15
Lipp, E., 1977: 495 Maratsos, M. & Kuczaj, S. A., 1978:
Lisker, L. & Abramson, A. S., 1967: 456,458-61,463-4
90 Maratsos, M., Kuczaj, S. A., Fox, D.
Lloyd, L.: see Minified, F.; & Chalkley, M. A., 1979: 468-76,
Schieffelbusch, R. 514
Lock, A. (ed.), 1978: see Mazurkewich, I., 1982: 63
Braunwald, S.; Ingram, D. 1984: 63
Locke, J. L., 1983: 100-1, 111, 113, McCarthy, D., 1930: 14
138,215 1954: 12,14,16,104-5,139,171,
Lukens, H., 1894: 9 231,516; see Irwin, 0.
Lust, B., 1986: 574 McCarthy, J.: see Baker, C. L.
See Carden, G. McLagan, M.: see Stark,R.
Lynip, A., 1951: 101 McNeill, D., 1966a: 268, 330
Lyons, J., 1968: 495 1966b: 268
Lyons, J. & Wales, R. (eds.), 1966: 1970a: 4,268,337
338 1970b: 268-72,294,319,330,338
See McNeill, D., 196613 1971: 268
Mead, M., 1963: 129
Macken, M. A., 1978: 392 Mengert, I.: see Wellman, B.
1979: 384-6,392-3,433 Menn, L., 1980: 190
1980: 393 1983: 385,433
1986: 394 Menyuk, P., 1969: 272
See Klatzky, R. 1971: 4
Macken, M. A. & Ferguson, C. A., Mervis, C.: see Rosch, E.
1983: 3924,433 Mervis, C. & Canada, K., 1983: 153
Macnamara, J., 1972: 303,316, 319, Mesalam, L.: see Leonard, L.
330 Miller, G.: see Halle, M.; Smith, F.
(ed.), 1977: see Slobin,D. Miller, G. & McNeill, D., 1969: 268
1982: 303,316-17 Miller, J., Chapman, R., Branston, M.
See Katz, N. & Reichle, J., 1980: 166-8,232
MacWhinney, B., 1978: 304,499-503, Miller, J. & Yoder, D., 1974: 227
515 1984: 336
1982: 76,307-10,333,338 Miller, J. D.: see Kuhl, P.
See Bates, E. Miller, P. D. 1972: 386
MacWhinney, B. & Bates, E., 1978: Miller, R.: see Antinucci, F.
515 Miller, R. T.: see Dore, J.
Malrieu, P., 1973: 58 Miller, W. & Ervin, S., 1964: 22,24,
Manzini, M. R.: see Wexler, K. 267
556 AUTHOR INDEX
Minifie, F. & Lloyd, L. (eds.), 1978: Ochs, E., 1982a: 78
see Ferguson, C. A. 1982b: 129
Moerk, E., 1980: 227,515 Ochs, E. & Schiefflin, B. (eds.), 1979:
1981: 515 see Bates, E. & MacWhinney, B.
Moffitt, A., 1971: 87 O’Grady, W., W., 1986: 487,491-2
Moore, M.: see Eilers, R. O’Grady, W., Suzuku-Wei, Y. & Cho,
Moore, T. (ed.), 1973: 338,434 S. W., 1986: 490-1,514
See Bowerman, M., 1973b; Oksaar, O., 1983: 4, 38
Clark, E., 1973a; Clark, H.; Oller, D. K., 1980: 100-1, 106-11
Garnica, 0. K.; Rosch, E. 1981: 100-1, 111, 138,215
Morehead, D. M. & Morehead, A. E. See Eilers, R.
(eds.), 1976: 338 Oller, D. K. & Eilers, R.,1982: 113
Morris, L.: see Leonard, L. Oller, D . K., Eilers, R.,Bull, D. &
Morse, P., 1974: 137 Carney, A., 1985: 112
Moskowitz, A., 1970: 217 Oller, D. K., Wieman, L., Doyle, W.
1971: 211 & Ross, C., 1976: 138
Mowrer, O., 1960: 20,99 Olmsted, D., 1971: 16, 361, 364
Muma, J., 1986: 304 See Shibamoto, J. S.
Munson, J. & Ingram, D., 1985: 56 Otsu, Y., van Riemsdijk, H.,
Inoue, K., Kamio, A. &
Nelson, K., 1973: 57, 144-6, 148-50, Kawasaki, N., (eds.), 1983: see
161,229,231,405,407 Wasow, T.
1974: 397,399400,402,434
1986: 434 Parisi, D.: see Antinucci, F.
Nelson, K. E., 1977: 511 Parkes, C.: see Fodor, J.
(ed.), 1978: see Greenfield, P. & Pelsma, J., 1910: 9
Zukow, P. Perlmutter, D., 1983: 331
(ed.), 1980: see Maratsos, M. & Peshkovskiy, A. M., 1925: 189
Chalkley, M. A. Peters, A., 1977: 248
1981: 72,510,515 1983: 308
(ed.), 1983: see Macken, M. A. & Piaget, J., 1929: 1256,406
Ferguson, C. A. 1948: 11619, 122, 138, 1574,231
Nelson, K. E. & Baker, N., 1984: 1952: 115, 118-19, 1224, 138, 157
510-13 1954: 115, 119, 138
Nelson, K. E., Denniger, M., 1955: 126
Bonvillian, J., Kaplan, B., & 1971: 2 6 7 , 117, 126, 128, 130
Baker, N., 1984: 510-11 Piaget, J., & Inhelder, B., 1967: 427
Newhoff, M.: see Leonard, L. 1969: 138
Newmeyer, F., 1983: 279 Piatelli-Palmarini, M. (ed.), 1980:
Newport, E., Gleitman, H. & 73
Gleitman, L., 1977: 226,437,507 Pinker, S., 1981: 515
Newport, M.: see Gleitman, L. 1982: 473
Nice, M., 1917: 9 1984: 236,272,301,303,315-30,
1920: 9 338,435,471,473,499,503-6,
1925: 1 4 , 4 6 9 , 5 7 515-16
Ninio, A. & Bruner, J., 1978: Pisoni, D.: see A s h , R.
129-30 Pitcher, E. G. & Prelinger, E., 1963:
Nokony, A., 1978: 498 477
Author index 557
Prather, E. M., Hedrick, D. L. & Sagart, L.: see de Boysson-Bardies, B.
Kern, C. A . , 1975: 361 Sander, E. K., 1961: 350
Prelinger, E.: see Pitcher, E. G . Scarborough, H. & Wyckoff, J., 1986:
Preyer, W., 1888-9: 7,8, 11, 103 515 r
Pye, C., 1980: 215 Schieffelbusch, R. & Bricker, D.

1983a: 131,215,497,515 (eds.), 1981: see Chapman, R.,
1983b: 128-9 1981b; Ingram, 1981c
1985: 325 Schieffelbusch, R. & Lloyd, L. (eds.),
1986: 129,498 1974: see Butterfield, E. &
1987: 336,496 Cairns, G . ;Eimas, P.; Miller, J.
Pye, C., Ingram, D. & List, H., 1987: & Yoder, D.; Morse P.
21418.394 Schiefflin, B., 1979: 129, 138
Schlesinger, I. M., 1971: 286-7
Radford, A . , 1981: 280,474 1982: 286
Ramer, A., 1976: 330,451 Schneider, B.: see Trehub, S.
See Dore, J. Schwartz, R. & Leonard, L., 1982:
Ratner, N. & Bruner, J., 1978: 129 160,198,212,360
Ravn, K. E. & Gelman, S . A . , 1984: Schwartz, R. & Terrell, B., 1983:
420-4 227-9
Reichle, J.: see Miller, J. Scollon, R., 1976: 241-6
Reinhart, T., 1976: 486 Shea, S.: see Jusczyk, P.
1981: 486 Sheldon, A . , 1974: 478-82,514
1983: 486 Shibamoto, J. S. & Olmsted, D., 1978:
Rescorla, L., 1980: 150-3, 159-60, 177, 200-1
231,404,411 Shipley, E.: see Gleitman, L.
Rodd, L. & Braine, M. D. S., 1970: Shipley, E., Smith, C. & Gleitman, L.,
25661,294,3367,516 1969: 162-6,232,25&1,257-8
Roeper, T., 1981: 223 Shriberg, L. & Kwiatkowski, J., 1980:
Roeper, T. & Williams, E. (eds.), 367,392
1987: see Borer, H. & Wexler, K.; Shvachkin, N., 1948173: 180-9,232-3,
Wexler, K. & Manzini, M. R. 342
Roussey, C., 1899-1900: 7 Sinclair, H., 1971: 280
Rosch, E., 1973: 397,434 Sinclair, H. & Ferreiro, E., 1970: 469
1975: 397 Siqueland, E.: see Eimas, P.
1978: 397 Skinner, B. F., 1957: 20, 99
Rosch, E. & Mervis, C., 1975: 397 Slobin, D., 1966: 232
Rosch, E., Mervis, C., Gray, W., 1968: 514
Johnson, D. & Boyes-Braem, P., 1970: 287
1976: 397 (ed.), 1971: 338; see Braine,
Rose, S.: see Stark, R. M. D. S.; McNeill, D.;
Ross, C.: see Oller, D. K. Schlesinger, I. M.
Rowan, L.: see Leonard, L. 1972: 30,232, 514
Riike-Draviga, V., 1973: 496 1973: 68,242, 309, 504
Rumelhart, D. & McClelland, J., 1986: 1977: 304
506,515 1982: 493, 515
(ed.), 1986: see Bowerman, M.;
Sachs, J., 1977: 132 Johnston, J.
Sachs, J. & Truswell, L., 1978: 165-6 See Ferguson, C. A.
558 AUTHOR INDEX
Smith, C., 1970: 468 Strange, W. & Broen, P. A., 1980:
See Gleitman, L.; Shipley, E.
Smith, F. & Miller, G. (eds.), 1966:
338 Skeeter, L. A., 1976: 93
See McNeill, D. 19661; Slobin, D Sudhalter, V. & Braine, M. D. S.,
Smith, J.: see Greenfield, P. 1985: 469,471
Smith, M., 1926: 13, 141,234,348 Syrdal-Lasky, A.: see Lasky, R. E.
1933: 30,414
Smith, N. S.: see Snow, C. E. Tager-Flusberg, H.: see de Villiers, J.
Smith, N. V., 1973: 10,370,378-86, Taine, H., 1877: 8,9,30,31, 170
433 Tavakolian, S. (ed.), 1981a: 330
1982: 59 1981b: 480,514
Snow, C. E., 1977a: 132 Taylor-Browne, K., 1983: 491
1977b: 132-8,219 Tees, R.: see Werker, J.
See Waterson, N. Templin, M., 1957: 14-16,31,343,
Snow, C. E. & Ferguson, C. A. (eds.), 361-8,433
1977: 233 Terrell, B.: see Schwartz, R.
See Garnica, 0. K.; Newport, E., Thomson, J. R. & Chapman, R., 1977:
Gleitman, H. & Gleitman, L.; 153-5, 159, 161,407,516
Sachs, J.; Snow, C. E., 1977a Tishman, A.: see Stockman, I.
Snow, C. E. & Goldfeld, B., 1983: 129 Trehub, S., Bull, D. & Schneider, B.,
Snow, C. E., Smith, N. S. & 1981: 138
Hoefnagel-Hohle, M., 1980: 494 Trubetzkoy, N., 1969: 193
Snyder, L., Bates, E. & Tse, S.-M., 1982: 185
Betherington, I., 1981: 148 Turner, E. & Rommetveit, R., 1967:
Solan, L., 1983: 490,514 514
1987: 490 Tyack, D. & Ingram, D., 1977: 414
Solberg, M. E., 1978: 497
Stampe, D., 1968: 385-6 Uzgiris, I. & Hunt, J. McV., 1975: 36,
1969: 376,384,38692,433 127
1973: 372,376,384-93
Stark, R., 1980: 101, 103 van der Sandt-Koenderman, M.: see
(ed.), 1981: 138 Arlman-Rupp, A.
See Chapman, R., 1981a; Eilers, R. van Niekerk de Haan, D.: see
& Gavin, W.; Kuhl, P., Oller, Arlman-Rupp, A.
D. K. van Riemsdijk, H.: see Otsu, Y.
Stark, R., Rose, S. & McLagan, M., Varma, T., 1979: 497
1975: 101-4, 118 Veach, S.: see Ingram, D.
Stern, C. & Stern, W., 1907: 8, 14, 38, Velten, H. 1943: 17,210,368-76,382,
103 433
Stern, W., 1924: 9,38-46,57 Vigorito, J.: see Eimas, P.
Stockman, I., Woods, D. & Vinson, J., 1915: 7
Tishman, A., 1981: 106 Volterra, V.: see Bates, E.
Stoel-Gammon, C., 1986: 384 Vygotsky, L., 1962: 150
Stoel-Gammon, C. & Cooper, J.,
1984: 200,212,233 Walker, E.: see Fodor, J.
Stoel-Gammon, C. & Dunn, C., 1985: Wall, R., 1972: 278
361,433 Wang, W. & Crawford, J., 1960: 217-18
Author index 559
Wanner, E. & Gleitman, L. (eds.), White, L., 1982: 80, 188
1982: 338,435 See Mazurkewich, I.
See Bates, E. & MacWhinney, B.; Whitehurst, G., Ironsmith, M. &
Braine, M. D. S. & Hardy, J.; Goldfein, M., 1974: 468
Carey, S.; Maratsos, M.; Wieman, L.: see Oller, D. K.
Slobin, D.; Wexler, K. Williams, E.: see Roeper, T.
Wasow, T., 1983: 27 Wilson, L., 1898: 9, 30
Waterson, N., 1970: 358 Wilson, W.: see Eilers, R.
1971: 358-9 Winitz, H., 1969: 99, 104, 138
Waterson, N. & Snow, C. (eds.), 1978: Witkowska-Stadnik, K.: see Weist, R.
see Bowerman, M. Wittgenstein, L., 1953: 397
Watt, W., 1970: 70 Woods, D.: see Stockman, I.
Webster, B.: see Ingram, D. Wysocka, A.: see Weist, R.
Weinberg, A,: see Berwick, R.
Weir, R., 1962: 10, 247 Yenj-Komshian, G., Kavanaugh, J. &
Weist, R. & Konieczna, E., 1985: 515 Ferguson, C. A. (eds.) 1980a: 138
Weist, R., Wysocka, A., See Aslin, R. & Pisoni, D.;
Witkowska-Stadnik, K., Ingram, D., Christensen, L.,
Buczowska, E. & E., 1984: 515 Veach, S. & Webster, B.;
Wellman, B., Case, I., Mengert, I., & Lieberman, P.; Menn, L.; Oller,
Bradbury, D., 1931: 14,361 D. K.; Stark, R.;
Wells, G.: see Barnes, S. (eds.), 1980b: 137-8
Werker, J. & Tees, R., 1984: 95 See Eilers, R.; Kuhl, P.; Strange, W.
Wexler, K., 1982: 338 & Broen, P. A.
See Borer, H. Yoder, D.: see Miller, J.
Wexler, K. & Chen, Y.-C., 1985: 490 Young, F., 1941: 14, 17
Wexler, K. & Culicover, P., 1980: 66,
80,330 Zarebina, M., 1965: 10, 58
Wexler, K. & Manzini, M. R., 1987: Zukow, P.: see Greenfield, P.
486,492 Zwicky, A., 1970: 29
General index
A/Amahl (Smith) (child Ann (Snow) (child subjects) 133-5

subjects) 10, 378-84, 386, 389 Anthony (Weir) (child
Abe (Kuczaj) (child subjects) 10, 247
subjects) 4574,460 aphasia 196
acceleration requirementhtage 3 4 5 , Aqissiaq (Fortescue) (child
54 subjects) 496
accommodation (cognition) 117-18, Arabic (languages) 133,514
191 articles 251-3,2574,304,307,324
action stage 42 articulation 13-15, 188, 366, 383
action words 144-7, 162 articulatory learning theory 97, 99,
Adam (Brown) (child 217
subjects) 22,49,51,236,269-71, articulatory/linguistic hypothesis 354
281,325-6, 328, 439,442,446-53, aspiration (stop) 89,90
462 assimilation (cognition) 117, 191
Adjective Preposing 71-2 assimilation: see phonological processes
adult-child interaction 134-5, 137, associations 18-231; see also
219-22 behaviorism
affix: see inflection; morphological associative complex 150, 153
development attunement theory 85-6,93-5,
affix checking 500 113-15, 179-80
‘after’: see ‘before’/‘after’ auditory judgement-spectral feature
Algonquin (languages) 514 system 101-2
Allison (Bloom) (child Aux/auxiliary (verb) 17, 26, 52, 66,
subjects) 169,236-9,243-4,247-8 308,454-65, 507-14: Bellugi’s
Amahl: see A/Amahl stages 458-60; explanation of Aux
amalgams 501-2 acquisition 461-5
Ament, daughter of (Ament) (child Avo (Lipp) (child subjects) 495
subjects) 192 Axel (Preyer) (child subjects) 8,
analogical overextensions 151-2 11,40, 103, 140
analogy 20,44,308
analytic learning strategy; see babbling 39,84,97-9, 108; see also
individual variation infant speech production
anaphora: see pronominal reference baby ratio 48
ancestor node 322 baby talk 131-2; see also linguistic
Andrew (Braine) (child environment
subjects) 22, 2367,2467,264-6, backing 392
274, 291
560
General index 561
Bantu (languages) 514 investigator): see Abe (Kuczaj);
‘before’l‘after’ 415-19 Adam (Brown); Allison (Bloom);
behaviorism 3,12,18-21,25,70,130, AlAmahl (Smith);
222-3; see also articulatory learning Andrew (Braine); Ann (Snow);
theory; explanation; perceptual Antony (Weir);
learning theory Asquith (Fortescue); Avo (Lipp);
Ben (Kuczaj & Maratsos) (child Axel (Preyer); Ben (Kuczaj &
subjects) 464 Maratsos); Canta (Chao);
Bengali (languages) 378 Carl (Miller & Ervin);
‘big’/’tall’ 419-26 Carolyn (Rodd & Braine);
biological factors: see innate abilities Chopti (Varma); Christine (Rodd
breakdowns 247-8 & Braine); Christy (Bowerman);
build-ups 247-8 Christy (Miller & Ervin);
Bulgarian (languages) 192, 514 Claire (Hoek, Ingram & Gibson);
Burmese (languages) 514 Craig (Huttenlocher) ;
daughter (Ament);
Canta (Chao) (child subjects) 386 daughter (Taine);
Cantonese (languages) 113 David (Braine); David (Carter);
Carl (Miller & Ervin) (child Embla (Lang & Larsson);
subjects) 22 Eric (Bloom); Erik (Lipp);
Carolyn (Rodd & Braine) 259-60 Eva (Bowerman); Eve (Brown);
case grammar 61,279-81,284; see Eve (Fonagy); Gia (Bloom);
also semantically oriented grammars; Gregory (Braine); Gunter (Stern
semantic relations and Stern); Harlan (Miller &
categorical overextensions 151-2 Ervin); Hilde (Stern & Stern);
c-command 486-8 Hildegard (Leopold);
chained successive single-word Jacqueline (Piaget)’
utterances 238 Jennika (Ingram); Joan (Velten);
Chao’s theory 386 Jonathan (Braine); K (Lewis);
child-directed speech: see baby talk; Karen (Lipp); Kathryn (Bloom);
linguistic environment Kendall (Bowerman); L (Oller);
Child Language vs. Language Laurent (Piaget); Lisa (Miller &
Acquisition 27-9,32, 3 8 , 6 W , Ervin); Lucienne (Piaget);
69-70,79, 187, 193,213-14,280, Mary (Snow);
331,401,517 Matthew (Greenfield & Smith);
child speech perception 15, Nicky (Greenfield & Smith);
178-89,232-3: Shvachkin’s Nigel (Halliday);
explanation of 185-8; see also child Nina (Shvachkin); Odi (Braine);
speech perception; infant speech Owain (Rodd & Braine);
perception Peter (Bloom, Lightbown &
child speech production 189-219,233 Hood); Pierre (Fonagy);
Jakobson’s theory of 190-7,215; Rachel (Rescorla);
see abo infant speech production; Rina (Bowerman);
phonological development; Sarah (Brown);
phonological production Seppo (Bowerman); Si (Macken);
Child Study Institute, Clark Sipili (Kernan); Steven (Braine);
University 9 Susan (Miller & Ervin);
child subjects (with name of principal T (Ferguson & Farwell); Thomas
562 GENERAL INDEX
(Huttenlocher); Tofi (Kernan); completions 247-8

Wendy (Huttenlocher); complex sentences: see relative clauses
Zina (Shvachkin) compounds 46
chinchillas 92 comprehension (vs. production) 3,
Chinese (languages) 185, 514; see 11,20, 39,42-3, 139-43, 146-7,
also Cantonese; Mandarin 153-5, 160-8, 189,249-61, 336,
Chopti (Varma) (child 350-3,443-6
subjects) 497 compre hension
Christy (Bowerman) (child testing (methodology) 1, 153-5,
subjects) 409-12 161-2
Christy (Miller & Ervin) (child concrete operations 341
subjects) 22 conditioning 18-19; see also
Claire (Hoek, Ingram & behaviorism
Gibson) (child subjects) 159-60 consonants and vowels, development
class-bound positional of: see infant speech perception;
patterns 309-10 infant speech production
class differences 14 consonant clusters 365-6
clinical method 125, 252 Constructionist Assumption 73, 77,
closed class 435, 510 11415, 178, 187, 189, 279,306, 314,
cluster reduction 372, 378 317,324
cognition (in relation to language constructionist theory 3 , 2 6 9 , 69-76,
acquisition) 11, 40, 58, 116, 120-4, 109, 114, 116, 126, 130-1,222-3,
126, 156-9,285-6, 335, 341, 414, 384,403,491,507; see also
42631,469; see also attunement theory; explanation;
accommodation; assimilation; object Oller’s theorylstages of babbling;
permanence; organization; Piaget’s Piaget’s theory; refinement theory
theory; sensorimotor activity/level; continuity requirementktage 33, 35,
symbolic play 49
cognition theory 414 convergence 40
cognitive development 115-27, 138; co-occurrence
see also cognition requirementhtage 36-7,42, 51, 109
cognitive explanation 253 cooing 84, 103; see also infant speech
cognitive theory (Macken & production
Ferguson) 392-4 coreference: see pronominal
Comanche (languages) 514 coreference
common nouns: see nominals correction by parents: see negative
competence: see linguistic competence evidence
vs. linguistic performance correlational stage 37-8
Competence Assumption 76-7, 143, Craig (Huttenlocher) (child
156, 189,213,243,261,271, 279, subjects) 164-5
283,294,322,326 cross-sectional
competence factors: see linguistic studies (methodology) 13
competence vs. linguistic culture 128-9,219; see also linguistic
performance environment
competition model 304-7 Czech (languages) 192,514
complementizer 745,476
complete perception theory 357-9, Dakota (languages) 498
387 Danish (languages) 514
+
General index 563
dative 28,29,62-3,445-6 ergativity 436
David (Braine) (child subjects) Eric (Bloom) (child
2367,292 subjects) 2345,237,245,249,267,
David (Carter) (child subjects) 169 294-5,298-300,304,330
Dedham, Mass. 23 Erik (Lipp) (child subjects) 495
degenerate input 26,223; see also Estonian (languages) 494-6, 514
linguistic environment Eva (Bowerman) (child
delateralization 388 subjects) 409-12
denasalization 373, 375-6 Eve (Brown) (child subjects) 22,
depalatalization 388-9 49,51,236,281,325,439,442,
description 2,9, 12,23,49, 125-6, 446-53,462
272,517 Eve (Fonagy) (child subjects) 239
descriptive stage 3, 32, 54-5, 185; see exhaustive hypothesization 504
also explanatory stage; stage of explanation 2-3; 9-10, 12, 18-21,
development 23-9, Ch. 4 (59-80), 84-7, 130-1,
despirantization 389 155-60,178,178-80,185-9,240-5,
determiner: see article 253,343,357-60,379,384-94,431,
developmental shift hypothesis 306 461-5,471-6; see also attunement
dialectic model 499 theory; behaviorism; cognitive
diary studies: see parental diary; period explanation; constructionist theory;
of diary studies distributional explanation;
directive 137 grammatical development; innate
discontinuity in development (vs. abilities/principles; maturational
continuity) 186-7, 191,243, 263, theory; refinement theory; stage of
267,286,294,384,394,396,412-13 development; theory of language
disorders: see linguistic deficit acquisition
dispositions 309 explanatory stage 3,32,55-7, 185-6;
distinctive features 180 see also descriptive stage; stages of
distributional explanation 253 development
distributional learning 262, 264, expressive vs. referential speech: see
300-1, 324; see also distributional individual variation
theories extension 402-3,407-9
distributional theories 303, 338,498; extraposition 480-1
MacWhinney 307-10; Maratsos &
Chalkley 310-15 final consonant deletion 372,374,378
dominance relation: see c-command Finnish (languages) 280-4, 291, 327,
double Aux marking 458-61 503,514
double tense marking 458-61 first words 18-19,38, 139-61:
Dutch (languages) 494-5, 514 cognitive factors 156-9; general
Dyirbal 436 semantic categories 143-7; onset
and rate 140-3; see also single-word
efficiency argument 315 utterances; word acquisition
elicited imitation (methodology): see first word combinations 2,21,22,23,
imitation 33,36,56, Ch. 7,234-339: general
Embla (Lange & Larsson) (child characteristics 245-9; general
subjects) 292 issues 261-3; major studies
environment: see behaviorism; culture; of 2368,337; see also
linguistic environment distributional theories;
564 G E N E R A L INDEX
functionalism; lexical functional definition 435-8; see also period of

grammar; parameter setting; pivot simple sentences: grammatical
grammar; semantically oriented morphemes
grammar; transformational grammar grammatical relations 268-72, 276-8;
flapping 363-4 tests for 276-8
formulae of limited scope 293 Greek (languages) 514
free patterns 309 Greenlandic
French (languages) l o , % , 113,284, Eskimo (languages) 494,496-8
368,390-2, 514 Gregory (Braine) (child
frequency of exposure 2264,426, subjects) 22,235- 7,2645,291,
437 334
frication 372-4 Gunter (Stern & Stern) (child
fricatives 87,3645, 378 subjects) 9,40, 44,45, 103
fronting 371, 378, 392
functional core concept Hall, G. Stanley 7, 8, 12
theory 399-401,410 Harlan (Miller & Ervin) (child
functional load 217-19 subjects) 22
functionalism 302-7, 310, 338 heart-rate paradigm 87-8
Hebrew (languages) 284,292,
Garo (languages) 514 514-15
gender 307,310 high-amplitude sucking paradigm 87,
general nominals: see nominals 9&2
generalization principle 65, 68 Hilde (Stern & Stern) 9, 44, 45
Generalized Phrase Structure Hildegard (Leopold) (child
Grammar 331 subjects) 10, 16, 168,213, 376-8,
generative semantics 279; see also 383,385,388-9
semantic grammars Hindi (languages) 378,497
German (languages) 140, 190, 284, history 3, 7-31 (Ch. 2); see also
310, 331, 368, 495,502, 514 period of large sample studies;
Gia (Bloom) (child subjects) 237, period of longitudinal sampling;
273-5,2945,298-300, 305, 330 period of diary studies
gifted children 14 holistic successive single-word
Gilyak (languages) 514 utterances 238-9
gliding 371, 374, 378 holophrase: see first words
global perception theory 357,379, homonymy 377
393 Hopi (languages) 514
Government and Binding horizontal construction 241
theory 331-2,436 Hungarian (languages) 503, 514-15
grammatical development 2 , 3 ; see hypothesis sampling 504
also first word combinations;
passives; pivot grammar; Icelandic (languages) 514
productivity; single-word utterances; imitation (methodology): direct 43,
subject pronouns; syntactic 120; elicited 1,256-7, 294, 336,
analysis (of children's language); 441-6; indirect 43;
syntactic comprehension; syntactic spontaneous 19-21,39; see also
spurt expansion; modeling; self-
grammatical freedom 334-5 repetition
grammatical morphemes, index 124
General index 565
individual variation 77-8, 212-14, intelligibility 248-9; see also
235, 318, 393-4,408; analytic vs. interpretation
synthetic learning 248; categorization intension 402-3, 409
vs. pronominal strategy 299-301; interaction: see adult-child interaction
expressive vs. referential speech 145, interpretation by adults 249-50, 287:
229-31 lean 242-4; rich 52,241-4
induction 2, 27 intervocalic consonant deletion 372,
infant-adult interaction 84 374
infant speech perception 83-96, invisible displacement 157
137-8, 140; see afso attunement Iroquois (languages) 514
theory; child speech perception; Italian (languages) 79, 331, 427-8,
heart-rate paradigm; high-amplitude 514-15
sucking paradigm; maturational item-based patterns 309
theory; perceptual learning theory;
phonemic perception; universal Jacqueline (Piaget) (child
theory; visually reinforced infant subjects) 115, 124
speech discrimination paradigm Jakobson’s theory of babbling 98-9
infant speech production 84,96115, Japanese 129,284,492-3,514
138; see also auditory Javanese 129
judgement-spectral feature system; Jennika (Ingram) (child
child speech production; Jakobson’s subjects) 235, 476
theory of babbling; Lockes’s theory Joan (Velten) (child subjects) 17,
of babbling; maturational theory; 368-76,381,385,388-9,391-2
Oller’s theoryktages of babbling; Jonathan (Braine) 198, 237, 291
phonological production; phonetic journals/monographs: see Monograph
transcription; recognition Series of the University of Minnesota
techniques; refinement theory; Institute of Child Welfare
spectrographic analysis; universal (journalshnonographs); Pedagogical
theory Seminary; The Journal of Child
infant vocal behavior; see prespeech Language; University of Iowa Studies
inference in Child Welfare
inflection 43-4,52,282,43&54;
Cazden’s stages 449-51; see also K (Lewis) (child subjects) 10, 96,
morphological development 103,219-20
innate abilities/principles 9, 12, 41, K (Ingram) (child subjects) 221-2
43-6,65-8, 117; see also Kaluli (languages) 129
maturational theory; generalization Karen (Lipp) (child subjects) 495
principle; lexical principle; principle Kathryn (Bloom) (child
of Derivational Morphology; subjects) 237,267, 2 7 6 9 , 2 9 4 5 ,
principle of linear sequence; 298-300,305,326-7,330,338
Principle of the Linguistic Sign; Kendall (Bowerman) (child
Principle of Phonemic Organization; subjects) 236-7,281, 292, 328
Principle of Subordination; Principle Korean (languages) 284, 514
of Syntactic Structure; uniqueness
principle L (Oller) (child subjects) 106, 108
input: see baby talk, child-directed Language Acquisition: see Child
speech; linguistic environment; Language vs. Language Acquisition
mother’s speech to children
566 G E N E R A L INDEX
language famine 42 lexical variability 366-7

language level: see stage of lexicon: see word acquisition
development linguistic competence vs. linguistic
languages cited (other than Modern performance 3,245,52-3,60,
English): see Algonquin; Arabic; 6 4 9 , 7 4 7 , 774,203, 213,243-4,
Bantu; Bengali; Bulgarian; Burmese; 261,329,3845,4614,472-3,
Cantonese; Comanche; Czech; 480-1, 518; see also explanation;
Dakota; Danish; Dutch; Dyirbal; operating principles
Estonian; Finnish; French, Garo; linguistic deficit 1; see also aphasia
German; Gilyak; Greek; linguistic environmenthnput 2, 77-9,
Greenlandic Eskimo; Hebrew; 95, 127-38, 217-31,233,506-15; see
Hindi; Hopi; Hungarian; Icelandic; also adult-child interaction;
Iroquois; Italian; Japanese; behaviorism; culture; degenerate
Javanese; Kaluli; Korean; Latvian; input
Lenakel; Luo; Malay; Marathi; linguistic performance: see linguistic
Mohave; Mundugumor; Navaho; competence vs. linguistic
Norwegian; Polish; Portuguese; performance; explanation
Romanian; Russian; Ruthenian; linguistic sign, development of 1 2 W ,
Samoan; Serbo-Croatian; Slovenian; 157-8
Songhai; Spanish; Swedish; Turkish; linguistic universals: see grammar;
Ukrainian; Welsh; Wolof, Yiddish universal grammar
large sample studies: see period of Lisa (Miller & Ervin) (child
large sample studies subjects) 22
Laurent (Piaget) (child Locke’s theory of babbling 100,
subjects) 115, 124 111-12,215
Latvian (languages) 496, 514 longitudinal language
law of cumulative complexity 286 sampling (methodology) 13
laws of irreversible solidarity 193-4, Lucienne (Piaget) (child
390 subjects) 115
lean interpretation: see interpretation Luo (languages) 281,284,327
learnability 2&9,62-3,315,330,
491-3; see also parameters macrodevelopment 451
learnability argument 315 Malay (languages) 514
learnability condition 316-17 Mandarin (languages) 386; see also
Lenakal 436 Chinese
lengthkomplexity indices: average Marathi (languages) 378, 514
length of sentence (ALS) 47-9; markedness 351-4
mean length of utterance Mary (Snow) (child
(MLU) 50-1; sentence length, subjects) 133-5
general 13-15 match 207
lexical contrasts 210 Matthew (Greenfield & Smith) 169,
lexical freedom 334-5 1747,234
lexical parameter 210, 214 maturational theory 3, 26-9, 70-6,
lexical principle 65-6, 68 78, 130-1,222-3.489-91, 506, 518:
Lexical Functional Grammar 317-18, of child speech perception 179; of
325,330,503; Pinker’s model of infant speech perception 85-6; of
acquisition 316-30 infant speech production 97, 100,
lexical semantics 395 111-15; see also explanation; innate
General index 567
maturational theory (cont.) Mohave (languages) 129
abilities/principles; restructuring Monograph Series of the University of
hypothesis; strong inclusion Minnesota Institute of Child
hypothesis; universal theory Welfare (journals/
mean length of utterance: see monographs) 14-15
length/complexity indices morphological analysis (of children’s
meaning: see semantics language) 3
measurement 1,12-13, 17-18, 47-9 morphological development 65-6;
measurement sequence: 153,203,213, 435-515: Adam, Eve and Sarah,
240, 349: logical 56-7; of 446-54; Brown’s stages 451-4;
quantitative 56-7 cross-linguistic 493-9, 514-15; early
measurement typology: logical 57; studies in English 439-46;
quantitative 57 explanation of 454,499-506,515;
mechanisms of change 388-90 linguistic environment 506-13, 515;
metalinguistic see also Aux; inflection; passives;
judgements (methodology) 1 plurals
metaphonological parameters 109 morphological imperialism 504
methodology/procedure 1-2,3,7-8, motherese 131
11-18,21-5,63,87-8, 101-2, 180-2, mother’s speech to children: see baby
197-210,2224,332-7,343-6,437-8, talk; linguistic environment
447-8; see also auditory multiword utterances: onset 42-3, 48,
judgement-spectral feature system; 49, 51-2; see also two-word
maturational theory; phonetic utterances
transcription; comprehension testing; Mundugumor (languages) 129
crosslinguistic studies; cross-sectional
studies; elicited imitation; nativism 25,27-9, 85-7; see also
experimental studies, identification innate abilities/principles
task; imitation; longitudinal language Natural Phonology 386-92
sampling; metalinguistic judgements; naturalistic studies (methodology) 1
modeling; Navaho (languages) 514
morphologylmorphological negative evidence: direct,
analysis (of children’s language); correction 20, 28-9; indirect 29,
naturalistic studies; parental diary; 315
phonemic perception; phonetic Nicky (Greenfield & Smith) (child
transcription; subjects) 169, 1 7 4 7 , 2 3 4
phonologylphonological Nigel (Halliday) (child
analysis (of children’s language); subjects) 169-73,220-1
recognition techniques; Nina (Shvachkin) (child
semantics/semantic analysis (of subjects) 183-4
children’s language); spectrographic nominals 246,251,269-70,405-13,
analysis; spontaneous language 431: common vs. proper
samples; syntadsyntactic analysis (of nouns 251-4; general vs. specific
children’s language); visually nominals 144-7
reinforced infant speech non-congruence 382
discrimination paradigm non-linguistic strategy 414
microdevelopment 451 Norwegian (languages) 368-9,514
minimal pairs 180-1
modifiers 144-5 object permanence (cognition) 120
568 G E N E R A L I N D E X
oblique prepositional phrases 319-24, period of concrete operations 341
471-2 period of diary studies 7-11; see also
observational adequacy 263 history; period of large sample
Odi (Braine) (child subjects) 292 studies; period of longitudinal
Oller’s theorylstages of language sampling
babbling 100-1, 10613,215 period of first word combinations 3,
one-word utterances: see single-word 234339: definition of 2 3 4 6 ; see
utterances also periods of language
onomatopoeia 40 development; first word combinations
operating principles 68-9; see also period of large sample studies 7,
linguistic competence vs. linguistic 11-21; see also history; period of
performance diary studies; period of longitudinal
operations of reference 285 language sampling
order of mention strategy 416-17 period of longitudinal language
organization (cognition) 117 sampling 7, 21-9; see also history;
overextensions 149-55,404, 410-12 period of diary studies; period of
overgeneralizations 447-51 large sample studies
Owain (Rodd & Braine) (child period of prelinguistic
subjects) 257-60 development 2; see also periods of
language development
P (Waterson) (child period of simple sentences 3:
subjects) 358-9 acquisition of grammatical
parallel function hypothesis 479-80 morphemes 435-515; phonological
parameterslparameter setting 64, and semantic acquisition 340-434;
330-2 see also periods of language
parental diary 7-11, 12, 21 development
partial perception theory 357,360,393 period of single-word utterances 2, 3,
passives 2245,2544,465-76, 53,139-233; see also periods of
511-14: explanation of language development; single-word
acquisition 471-6 utterances
past tense affix 310,312,314 periods of language development 2,
Pedagogical 3; see also period of first word
Seminary (journalslmonographs) 9 combinations; period of prelinguistic
percentage of obligatory development; period of single-word
occurrence 52 utterances
perception: see child speech personal-social words 145-7
perception; infant speech perception; Peter (Bloom, Lightbown &
phonemic perception Hood) (child subjects) 234-5,
perceptual hypothesis 354 294-5,298-300,304,330
perpetual learning theory 84-6 phatic utterance 136
perceptual salience 188,437 phone classes and phone trees 200-3,
performance (vs. competence): see 393
linguistic competence; linguistic phoneme 16,201
performance phonemic perception 179, 342-60,
performance factors: see linguistic 432-3: major studies 347-60;
competence vs. linguistic methodology 343-6; see also child
performance speech perception; infant speech
performatives 168-9, 176 perception
General index 569
phonetic inventories and phonological speech production; phonemic
contrasts 203-10 production
phonetic transcription 10, 16,203: of phonological salience 189,212, 343
infant speech 101,332-3 phonological stages 340-1
phonetic variability 203, 211 phrase structure 32C5; see also
phonetically consistent forms 174, 249 grammatical development
phonological analysis of children’s physiological development 95, 97-8,
language 1, 197-210; see also 109-11
phone classes and phone trees; Piaget’s theory 4, 115-27
phonetic inventories and Pierre (Fonagy) (child
phonological contrasts subjects) 239
phonological contrasts 207-9 pivot grammar 246,263-7,269,273,
phonological development 2, 3, 127, 284,288: arguments against 266-7
178-219,232-3,340-94: avoidance plateau requirementktage 33-5, 49
strategies 160,212, 378; levels of plural, English, regular and
representation 195,384-6; lexical irregular 65-6,310,447-50
factors 210; of single Polish (languages) 10,57-8,514-15
morphemes 340-94; see afso child Portuguese (languages) 514
speech perception; child speech positional consistency 266, 288-90
production; infant speech positive evidence 28-9; see also
perception; infant speech negative evidence
production; phonemic production; positive inhibitory blocks 313
phonological production practice 19
phonological idiom 211, 214, 393 predicate adjectives 319-24
phonological parameter 211, 214 predicate statements 151-2, 177
phonological perception: see child prelinguistic development: see period
speech perception of prelinguistic development
phonological processes 371-3, prevision 124
387-92; see also cluster reduction; prevocalic obstruent deletion 389-90
delateralization; denasalization; primary circular reaction 118-19
depalatalization; despirantization; Principle of Derivational
final consonant deletion; frication; Morphology 46
fronting; gliding; intervocalic principle of linear sequence 67-8
consonant deletion; prevocalic Principle of the Linguistic Sign 41,
obstruent deletion; spirantization; 42,54,179, 187
stopping; unstressed syllable principle of maximal contrast 195-6
deletion; velar assimilation; Principle of Phonemic
vocalization; voicing; vowel Organization 187-8, 195
lengthening; vowel lowering; vowel principle of structure dependent
nasalization; vowel raising; vowel rules 66-8
shortening Principle of Subordination 45
phonological production 360-94,433: Principle of Syntactic
diary studies: Velten 368-76; Structure 43-5
Leopold 376-8; Smith 378-84; principle requirementktage 36-8
explanation 38494; Macken & principles: see innate abilities
Ferguson 3924; Stampe 386-92; PRO 74
large sample study of 361-8; see also pro drop languages 79
child speech production; infant pro drop parameter 260, 331-2,463
570 GENERAL INDEX
production vs. comprehension: see relevance 407
comprehension vs. production restructuring (of grammar) 70-3
productivity 4,767,2454,262-3, restructuring (of lexicon) 381
265,273,281,289-90,296,333-7: restructuring hypothesis 70-3
Braine’s criteria for 289-90; criteria reversible sentences 254-5; see also
for 3 3 4 5 passives
Productivity Assumption 77, 189 rich interpretation: see interpretation
progressive affix 258, 310 Rina (Bowerman) (child
promiscuity (Pinker’s term) 328 subjects) 280-1
pronominal reference 70-1,484-93, Romanian (languages) 514
514; stages of 491 rote: memorization 4,26,43,76, 308;
pronominalization: see pronominal see also productivity
coreference Russian (languages) 10,182-5,189,
proper nouns: see nominals 284,342-3, 514
propositional semantics 395 Ruthenian (languages) 514
proto-dialogues 220-1
prototype theory 397,401-5,425-6, Samoan (languages) 78, 129,281,
431,434 284,291,327,514
proto-words 170-1 Sarah (Brown) (child subjects) 22,
puzzles (Smith’s term) 382 49,51,236,281,325,439,446-53,
462
QRNs 106-8 secondary circular reaction 118-19
questions: see Subject-Aux Inversion self-repetition 69
Quiche Mayan (languages) 128-9, semantic analysis (of children’s
131-2,21419,497-8 language) 10
semantic bootstrapping 262, 300, 303,
Rachel (Rescorla) (child 310, 315-16, 318-21,330; see also
subjects) 159 cognitive explanation; distributional
rare-event comparison theory 510-11 learning; Lexical-Functional
rate of acquisition 141-3,234-5 Grammar
realization rules 379, 385 semantic development 2,3, 18-19,
recidivism (Smith’s term) 382-3 120-4, 13940,394432; see also
recognition techniques 102, 113 word acquisition; lexical semantics;
recurrence 273,285 propositional semantics
reduction transformation 278-9 semantic feature hypothesis 398-9,
referential speech: see individual 41425,434
variation semantic relations, in early syntax 51,
refinement theory 87, 100-1, 113-15, 162,165-8,241-3,284302,494:
179-80; see also Oller’s theorylstages criticism of 287-8
of babbling semantic theory 155
reflexives 4 8 4 6 semantically oriented
relation and attribute stage 42 grammar 279-302: Bloom,
Relational Grammar 330-1 Lightbown & Hood 294-302; t
relational words 413-31 Bowerman 280-3; Braine 288-94;

relative clauses 28, 61, 72-3, 476-83, Brown 283-8
511-12,514; see also Adjective sensorimotor
Preposing activityllevel (cognition) 36,
relative pronouns 44
General index 571
116-24, 156-9, 168,2854,341 Standard Theory 268,272, 280-1,
sensorimotor communicative 284,317-18
schemas 174 Stanford theory 3 9 W , 433
sentence length: see lengthkomplexity Steven (Braine) (child
indices subjects) 22,236-7, 264,291
sentence paragraph 246-8 stopping 371, 378
sequences of single-word strategy: see individual variation
utterances 238-45: explanation of strengthening 311, 323
240-5; Fonagy's stages of 239-40 strong inclusion hypothesis 70-3, 317,
Seppo (Bowerman) (child 506,518
subjects) 280-2,291,328 structure dependent distributional
Serbo-Croatian (languages) 192, learning 324
427-8,514-15 structure preservation 73
serial verbs 75 style of acquisition: see individual
shifting reference (reference): see variation
deixis subject 283,298-9,304,310,314-16,
Si (Macken) (child subjects) 386 327
signals 122-4, 156-7 subject pronouns 260-1, 332; see also
single-word utterances 33, 3 4 , 4 6 1 , pro drop parameter
48,56, 168-78,231-2; see also period Subject-Aux Inversion 66, 455-64
of single-word utterances; sequences Subject-Verb Agreement 67
of single word utterances subset principle 492-3
Sipili (Kernan) (child substance stage 42
subjects) 291 succession requirementlstage 36-7
Slovenian (languages) 514 Susan (Miller & Ervin) (child
Songhai (languages) 436 subjects) 22
Spanish (languages) 89,93-4, 113, Swedish (languages) 284,514
284,331,386, 514 switch reference 436
spatial words 42631 syllables 68, 367
specific nominals: see nominals symbolic play 38, 55, 158, 341; see
spectographic analysis, of infant also cognition
speech 101 symbolic stage 38
spirantization 388-9 symbols 122, 157-8; see also linguistic
spontaneous language sign
samples (methodology) 1 , 3 , syntactic analysis (of children's
21-2,332-6 language) 1 , 3 , 332-7
stable forms 211 syntactic comprehension 249-61, 337
stage error 57 syntactic development: see grammatical
stages of development, general 3, development
32-58,64, Ch. 3, 340-1; syntactic spurt 234-5
Brown's 49-53; Nice's 47-9,53; syntactic vs. semantic categories 52,
Stern's 38-47, 53; see also Aux; 261-2,327-8
cognition; descriptive stage; synthetic learning strategy: see
explanatory stage; infant speech individual variation
production; inflection; child speech
production; periods of language T (Ferguson & Farwell) (child
development; pronominal reference; subjects) 199-213, 360
sequences of single word utterances Taine, daughter (Taine) (child
572 GENERAL INDEX
,
vocabulary \ 13-15
subjects) 8
i
Taine, daughter (Taine) (ch Id
subjects) 8 k
vocalizatidb’ 372,374, 378
voicing 4 0 : final 371, 378;
‘tall’: see ‘big’ytall’ prevocalic 371,378,388-9
telegraphic speech 442 voice onset time,
term of reference 402 VOT, (phonetics) 89-90:
tertiary circular reaction 157 development 90-4
The Journal of Child Language vowel lengthening 374-6
journals/monographs) 1, 95 vowel lowering 373
thematic hierarchy constraint 473-4 vowel nasalization 390-2
theory of language acquisition 3, vowel raising 373
63-9; see also explanation vowel shortening 374-6
Thomas (Huttenlocher) (child
subjects) 164 Watson, Thomas
Time 1 vs. Time 2 studies 2246, 508 B. (behaviorism) 12
Tofi (Kernan) (child subjects) 291 Welsh (languages) 514
tokens (vs. types) 105 Wendy (Huttenlocher) (child
topic 305 subjects) 164
transformational grammar 23,25, Wh Movement: see Subject-Aux
267-8: as a model of acquisition: Inversion
Bloom 272-9; McNeill 268-72; Wolof (languages) 514
see also grammatical development word, definition of 13940
transition requiremendstage 34-5, 40 word acquisition 18-19,83, 139-61,
Turkish (languages) 427-8, 515 231, 394432,434: environmental
twins 14 effects 226-31; explanation
typeerror 57 of 396405,431; see also first
types (vs. tokens) 105 words; functional core concept
theory; nominals; prototype theory;
Ukrainian (languages) 514 relational words; semantic feature
underextensions 148-9, 152 hypothesis; word spurt
uniqueness principle 66, 68, 324 word familiarity 367
universal base hypothesis 331 word order 5 1 , 2 4 6 5 , 2 5 4 7 , 2 5 9 ,
Universal Grammar (UG) 25-6, 282, 309-10, 326-7; see also
63-7,79,262,327 passives
universal symbols 329 word spurt 34-5,41,42,54,141-3,
universal theory 85-6,93-9, 113-15, 159,234,342, 369,370,377; see
179 lexicon; morphology; vocabulary
University of Iowa Studies in Child wug procedure 440
Welfare (journals/monographs) 14
unstressed syllable deletion 372, 378 X-bar syntax (x) 270,321,324
velar assimilation 389-90
vertical construction 241, 246 Yiddish (languages) 514
visually reinforced infant speech
discrimination paradigm 88, 9 3 4 , Zina (Shvachkin) (child
167 subjects) 183-4

(David Ingram) First Language Acquisition Method

Uploaded by

Copyright:

Available Formats

(David Ingram) First Language Acquisition Method

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(David Ingram) First Language Acquisition Method

Uploaded by

Copyright:

Available Formats

First Language Acquisition

Method, Description, and Explanation

llK Unrvcrw has

Cambridge University Pres

0Cambridge University Press 1989

Printed at The Bath Press, Avon

British Library cataloguing in publication data

Library of Congress cataloguing in publication data

ISBN 0 521 34109 4 hard covers

Author index 549

This enterprise officially began in April, 1984, in the Dallas-Fort Worth

The study of children’s language acquisition is a field that comprises a large

1, techniques of data collection - from spontaneous language sampling

of ‘productivity’. This is the concern with separating the child‘s linguistic

2.1 The period of diary studies (1876-1926)

eventually published, sometimes in their raw form with little commentary

2.1.2 Major diary studies

2.1.3 The value of diary studies

2.2 The period of large sample studies (19261957)

psychology that has come to be known as ‘behaviorism’. Just as the earlier

Study Sample characteristics Topic

Subjects: 430 total

directed several studies that appeared in the Monograph Series of the

2.2.2 Strengths and weaknesses

contribution with their emphasis on measurement. A continuing weakness

2.2.3 Theoretical orientation

This account of acquiring early word meaning attempts to combine the

2.3 The period of longitudinal language sampling (1957 to present)

2.3.1 Methodological orientation

Table 2.5 General information on four major studies using longitudinal

Brown (1973) Adam

The longitudinal language samples incorporate some of the experimental

1964) remain an important comment on the field of language acquisition. In

2.3.2 Theoretical orientation

language and isolates predictions those theories make about language

3.1 Some possible meanings of ‘stage’

3.1.1 Single behaviors

continuous stage plateau stage

transition stage acceleration stage

3.1.2 Multiple behaviors

succession stage co-occurrence stage

co-occurrence stage principle or implicational stage

3.2 Some proposals on stages of acquisition

3.2.1 Stern (1924)

While there is no language during the first year, Stern sees it as an

association of a sound or set of sounds with certain actions and contexts.

Table 3.2 A summary of the major characteristics of Nice’s (1925) stages

I. Single word stage The child uses single-word utterances.

This change is apparently a separate co-occurring behavior in the area of

(3.5) a. fire box [a fireproof casserole over a stove]

Average Length of Sentence (words) 1.00 3.5-3.5 5.0

Proportion of complex sentences very low more adult-like

3.2.2 Nice (1925)

Further, there is the emphasis on the development of measures of super-

3.2.3 Roger Brown (1973)

Table 3.4 A summary of Brown’s (1973)five stages of grammatical

Stage Range of MLU Upper Midpoint Stage name and description

1.oo The period of single-word utterances

analysis. It was with this selective process in mind that he originally