Meaning, Form, and Use in Context: Linguistic Applications: Deborah Schiffrin

Meaning, Form, and Use
in Context:
Linguistic Applications
Deborah Schiffrin
editor
Meaning, Form, and Use
in Context:
Linguistic Applications
Deborah Schiffrin
editor
Georgetown University Press, Washington, D.C. 20057

BIBLIOGRAPHIC NOTICE
Since this series has been variously and confusingly cited as: George-
town University Monographic Series on Languages and Linguistics,
Monograph Series on Languages and Linguistics, Reports of the Annual
Round Table Meetings on Linguistics and Language Study, etc., beginning
with the 1973 volume, the title of the series was changed.
The new title of the series includes the year of a Round Table and
omits both the monograph number and the meeting number, thus:
Georgetown University Round Table on Languages and Linguistics 1984,
with the regular abbreviation GURT '84. Full bibliographic references
should show the form:
Kempson, Ruth M. 1984. Pragmatics, anaphora, and logical form. In:
Georgetown University Round Table on Languages and Linguistics
1984. Edited by Deborah Schiffrin. Washington, D.C.: Georgetown
University Press. 1-10.
Copyright (§) 1984 by Georgetown University Press

All rights reserved
Printed in the United States of America
Library of Congress Catalog Number: 58-31607
ISBN 0-87840-119-9
ISSN 0196-7207
CONTENTS
Welcoming Remarks
James E. Alatis
Dean, School of Languages and Linguistics vii
Introduction
Deborah Schiffrin
Chair, Georgetown University Round Table on
Languages and Linguistics 1984 ix
Meaning and Use
Ruth M. Kempson
Pragmatics, anaphora, and logical form 1
Laurence R. Horn
Toward a new taxonomy for pragmatic inference:
Q-based and R-based implicature 11
William Labov
Intensity 43
Michael L. Geis
On semantic and pragmatic competence 71
Form and Function
Sandra A. Thompson
'Subordination' in formal and informal discourse 85
Wallace Chafe
Speaking, writing, and prescriptivism 95
Gillian Sankoff
Substrate and universals in the Tok Pisin verb
phrase 104
in
iv / Contents
Talmy Givdn
The pragmatics of referentiality 120
Jerrold M. Sadock
Whither radical pragmatics? 139
Meaning and Form
Richard Hudson
A psychologically and socially plausible theory of
language structure 150
Richard Bauman
The making and breaking of context in West Texas
oral anecdotes 160
Thomas A. Sebeok
Enter textuality: Echoes from the extraterrestrial 175
Michael Silverstein
On the pragmatic 'poetry' of prose: Parallelism,
repetition, and cohesive structure in the time course
of dyadic conversation 181
Contexts: Institutional and Interpersonal
Thomas Kochman
The politics of politeness: Social warrants in
mainstream American public etiquette 200
Don H. Zimmerman
Talk and its occasion: The case of calling the police 210
Alan Davies
Idealization in sociolinguistics: The choice of the
standard dialect 229
Ellen F. Prince
Language and the law: Reference, stress, and context 240
Howard Giles and Mary Anne Fitzpatrick
Personal, group, and couple identities: Towards a
relational context for the study of language
attitudes and linguistic forms 253
John 3. Gumperz
Communicative competence revisited 278
v / Contents
Susan Gal
Phonological style in bilingualism: The interaction
of structure and use 290
The Acquisition of Meaning, Form, and Use
George A. Miller
Some comments on the subjective lexicon 303
Marilyn Shatz
A song without music and other stories: How
cognitive process constraints influence children's
oral and written narratives 313
Elinor Ochs
Clarification and culture 325
WELCOMING REMARKS
James E. Alatis
Dean, School of Languages and Linguistics
Georgetown University
Ladies and gentlemen, it gives me great pleasure to welcome you on
behalf of Georgetown University and the School of Languages and Lin-
guistics to the Georgetown University Round Table on Languages and
Linguistics (GURT '84). The chairman of this year's conference is Dr.
Deborah Schiffrin, who has chosen for this, the 35th annual Georgetown
University Round Table, the theme: 'Meaning, form, and use in con-
text: Linguistic applications.' The program she has prepared is impres-
sive, and the superb organization of all the conference details is her
work and that of her able assistant, Susan M. Hoyle.
Like many of you, I was fortunate enough to attend some of the pre-
conference sessions earlier today. One cannot help but remark that,
once again, the preconference sessions present almost as wide and as
interesting a range of topics as the conference itself. This is a tribute
to the energy and enthusiasm of Deborah Schiffrin and our good col-
leagues who dedicated so much time and effort to the success of the
presessions; to all these colleagues, I would like to offer our heartfelt
thanks. I especially want to extend a hearty welcome to the members of
the Inter-Agency Round Table of the U.S. Government. I assure them of
my intention to continue the tradition of acting as their host at all
future Georgetown University Round Tables on Languages and Lin-
guistics, for as long as they wish.
It is a pleasure to see that, as in previous years, some of the best
scholars in linguistics and related disciplines have assembled for this
year's Round Table, thus ensuring a most exciting and productive con-
ference. One of the few prerogatives reserved to the Dean of the School
of Languages and Linguistics is the honor of thanking our speakers for
having agreed to come, many from very long distances, to share the
results of their research with us. You do Georgetown University a great
honor, and render a great service to our colleagues at this Round Table.
Thank you for coming.
I do not wish to delay unduly what will surely be a very profitable and
Vll
viii / Welcoming Remarks
productive session. Before concluding, however, I am pleased to note

that other segments of society have come to appreciate both language
teaching and linguistics. Thanks to the dedicated research and skilled
writing of scholars in these fields, the world has begun to realize what
vital ingredients language teaching and linguistics are in the creation of
the mutual understanding necessary for people to function constructively
and beneficially in a multicultural, interdependent society. Surely then,
in a world that all too frequently edges toward the brink, I am very
pleased to welcome you to a conference that will contribute to the
educational goals which we all cherish and the sense of mission which we
all have.
INTRODUCTION
The theme of this year's Georgetown University Round Table on

Languages and Linguistics is 'Meaning, form, and use in context: Lin-
guistic applications.' Because this is so broad a theme, and one which
can be approached from so many different directions, I would like to say
a few words about how we will be approaching meaning, form and use of
language in context during the next few days at the Round Table.
A continuing task in any academic discipline is carving out a special
domain to be a focus of inquiry, and finding a special set of problems for
which to seek resolution. Sometimes, it seems that these tasks repeat
themselves over and over again, as we attempt to find some area of
scholarship (which necessarily gets smaller and smaller) within which to
achieve expertise.
During this progressive—and inevitable—narrowing of our interests,
our larger, more holistic frameworks may seem to disappear. Our initial
assumptions and definitions, concepts and ideas, theories and meta-
theories, may seem to be overshadowed by specific findings within a
very small area of research. But our larger frameworks never really
totally lose their salience. Rather, they continue to provide implicit
models—because it is within such frameworks that answers are sought
and problems are resolved. Even more basically, it is within such
frameworks that our questions and our problems actually take shape.
In a way, our disciplinary paradigms are like cultures: just as cultures
provide, organize, and sustain our perception of reality, our paradigms
provide, organize, and sustain our academic realities. They do so not
only by providing the questions that we ask and the problems that we try
to resolve, but by helping to form the methods that we bring to bear on
those problems. Our disciplinary paradigms also frame the interpreta-
tions that we provide for our findings. In addition, our paradigms help
define the everyday issues to which we seek to relate and apply our
findings. And just as we accept our cultural constructs without question,
so too, do we accept our academic constructs.
These comments are all relevant to the Georgetown University Round
Table on Languages and Linguistics because meaning, form, and use of
language, in context, have all been domains of inquiry for many very
different academic disciplines, each providing its own sets of paradigms,
IX
x / Introduction
models, and constructs: linguistics, anthropology, sociology, psychology,

philosophy, to name just a few. Furthermore, within those disciplines,
an inevitable division of labor often results in different subareas, each of
which focuses on one or another of those domains.
In linguistics, for example, we have the familiar three-part distinction
among semantics (for meaning), syntax (for form), and pragmatics (for
use). Focus on language in context actually undergoes further subdivi-
sion as different levels of contextual frames are pulled apart from one
another. For example, we differentiate sociolinguistics, which focuses
on social context, from psycholinguistics, which focuses on cognitive and
perceptual contexts; ethnolinguistics, which locates language in cultural
context, from discourse analysis, which (at least in one very narrow
interpretation) focuses on linguistic context.
The more we become partitioned by our particular academic worlds,
however, the more our overall academic world view becomes confined to
the paradigms, models, and constructs specific to our own relatively
narrow fields of interest. In a way, of course, such partitioning is a
strength, because it allows us to understand and describe, perhaps even
explain, something that has been perplexing. Thus, such partitioning
allows us to catch at least a glimmer of insight into some small part of
an often confusing world—and this is certainly a strength.
But our academic partitions can also be a weakness, because they can
prevent us from looking outside of our own paradigms, our own models,
and our own constructs.
It is my hope that the 1984 Georgetown University Round Table on
Languages and Linguistics can be one small step toward offsetting this
weakness. More specifically, it is my hope that the Round Table will not
only help us identify and appreciate the different frameworks through
which we understand meaning, form, and use of language in context, but
that it will also help us to use the diversity provided by frameworks that
are different from our own, to build a more holistic theory of language—
a theory which is based upon a wide empirical base, and which can
contribute to the resolution of a wide variety of real-world problems.
In sum, I hope that in listening to—and reacting to—the discussions of
the next few days, we will begin to view the diversity of frameworks for
analyzing meaning, form, and use of language in context as a strength,
and that we will be able to capitalize upon that strength when we even-
tually return to our own academic world views.
I would like now to thank several people who have contributed to the
materialization of this conference (although there are many more such
people than I can thank individually). First, I want to thank Dean James
E. Alatis for giving me the opportunity to organize this year's George-
town University Round Table on Languages and Linguistics. I also want
to thank my colleagues here at Georgetown for their moral support:
Deborah Tannen deserves special mention in this regard. The George-
town students, both graduate and undergraduate, have been extraordi-
nary in their willingness and eagerness to help in many different capac-
ities. And for this, they have my gratitude. In addition, the program
could not have taken shape without the cooperation and enthusiasm of
the organizers of special interest sessions, the participants in those
sessions, and, of course, the main session speakers. All of these people
have my heartfelt thanks.
Introduction / xi
Finally, I want to thank Susan M. Hoyle for her tremendously able

assistance, her never-ending patience, and her infectious cheerfulness in
wading through the countless tasks that are a necessary part of organiz-
ing a conference.
Deborah Schiffrin
PRAGMATICS, ANAPHORA, AND LOGICAL FORM
Ruth M. Kempson
University of London
This paper is simultaneously an outline of a unitary, pragmatic
account of anaphora and a sequence of arguments advocating the prag-
matic theory of Sperber and Wilson (forthcoming). First, I present
evidence that a unitary account of anaphora is the only correct one.
Second, I show that a natural account of anaphora emerges if we assume
the Sperber-Wilson approach to pragmatics. Finally, I draw out the
consequences of this conclusion for syntactic theory, and for the analysis
of discourse.
My argument concerns the parallelism not normally noticed between
bound-variable anaphora, as in (1), and discourse anaphora, as in (2).
(1) Every actor: worries that he:'s ugly.
(2) He^'sugly.
Despite the widespread assumption that these are discrete phenomena,
every pragmatic aspect of discourse anaphora is displayed also by bound-
variable anaphora. The parallels are as follows.
(A) The phenomenon of bridging cross-reference. This is where the
use of an anaphoric expression is licensed by a preceding noncoreferring
expression in virtue of some link between the objects those expressions
describe. Thus in (3) and (4) there is no relation of identity between a
car and its driver, and yet the introduction of the term car allows us to
use a definite NP in virtue of the link between cars and drivers.
(3) John lifted a car. The driver was underneath.
(4) John lifted a car with the driver.
Exactly parallel is (5), except that the expression a car can be inter-
preted within the scope of the quantifying expression everyone and the
term the driver is accordingly interpreted as a variable bound by the
expression a car, itself sensitive to the outermost quantifier.
2 / Ruth M. Kempson
(5) Everyone who was able to lift a car found the driver underneath.
A similar set of examples is given in (6) and (7).
(6) John's house is a mess. The roof needs mending.

(7) Every house needs the roof mended.
That this is not simply a matter of lexical specification of hidden argu-
ments is shown by (8): singer does not have accompanist as a lexically
specified argument.
(8) Every accompanist needs the singer to be quiet in the opening

bars.
(B) The use of anaphoric expressions can be licensed by extra in-
formation made available by the total implicit content of the preceding
sentence and not any one constituent. Thus (9) allows the term the
insult on the basis of the assumption that calling someone a Conserva-
tive is an insult. This has long been recognised as a pragmatic phenom-
enon. But there are exactly parallel examples with bound-variable
anaphora which is not recognised as a pragmatic phenomenon, as in (10).
(9) Jake called Jess a Conservative. The insult made him bristle.
(10) Everyone who called his neighbour a Conservative later
apologised for the insult.
(C) One may have to manipulate extra premises and principles of de-
duction in order to establish an antecedent. Thus in (11), we have to
know that Jaguars are cars and two negatives make a positive in order to
use the first disjunct in (11) to provide an antecedent for the definite NP
the car. And (12) is exactly parallel.
(11) Either my friend hasn't bought a Jaguar, or the car'U be in the

garage by now.
(12) Each of my millionaire friends who isn't so anti-British that
they haven't bought a Rolls-Royce will soon be fed up with
the car's gas consumption.
(D) The givenness associated with pronouns and definite NPs as op-
posed to the 'newness' of indefinite NPs. This is familiar, but it is not so
often pointed out that this concept of picking out something already
'given'/previously established is displayed equally by bound-variable
anaphora. Examples (13) and (14) are straightforward cases of definite
NP and pronominal anaphora. But (15) and (1) present the same phen-
omenon for bound-variable anaphora.
(13) John bought a house and discovered later that the house need-
ed damp-proofing.
A man came in. He sat down.
Pragmatics, anaphora, and logical form / 3
(15) Everyone who bought a house discovered later that the house
needed damp-proofing.
In these cases, the definite NP and pronoun are licensed and in some
sense given by the preceding quantified expression. The difference is, of
course, that with a quantified antecedent, the bound-variable anaphor is
in some sense given by each instantiation of the quantified expression.
(E) Similar to this is the property of uniqueness. This problem is

widely recognised to be pragmatic, relative to the circumstances in
some unanalysed way. Thus in (2), there is only one male assumed to be
under consideration, in (3) and (4) only one driver, and in (6) only one
roof. This phenomenon, too, is displayed equally by bound-variable
anaphora cases such as (1), (5), (7), (8), and (15), though in these cases
the value of the interpretation of the pronoun or definite NP is again
uniquely determined by each instantiation of the quantifying expression.
In (1), for example, for each instantiation of the subject quantifying
expression there is only one possible object picked out by hej for each
instantiation of car in (5) there is only one possible object picked out by
the driver; and so on. Of course, we analyse these in terms of identity
of variables; the pretheoretic phenomenon, however, that the anaphoric
expression is on each interpretation uniquely identified as something, is
exactly parallel in both referring and bound-variable uses of NPs and
pronouns. Thus for all the listable, apparently pragmatic properties of
referring, discourse-anaphoric uses of pronouns and definite NPs, there
is a direct analogue for bound-variable uses. Yet it is these pragmatic
properties which provide the motivation for analysing discourse anaphora
as a pragmatic phenomenon. But quantifier binding, which displays the
same properties, is universally assumed to be a phenomenon of grammar.
It seems that if we are to give a unitary account of problems of ana-
phora, we have to assume a pragmatic basis to bound-variable anaphora
as well as to discourse anaphora. This is what I am going to propose.
However, it does not follow from this conclusion that we are forced into
abandoning the formal characterisation of sentence-based grammars for
the more informal alternative of discourse grammars, or for the incor-
poration of pragmatics within the grammar. On the contrary, I argue
that an orthodox sentence-based grammar, in conjunction with the
principle of relevance, predicts exactly the right results. All that has to
be changed is the assumption that the articulation of semantics within
the grammar provides the full propositional content/ logical form/truth
conditions expressed by a sentence. In its place I propose that we should
look at the semantic content of any linguistic expression, simple or
complex, as a set of instructions on utterance interpretation. These
include a set of instructions on the construction of logical forms, but
may also include instructions on other aspects of interpretation. In all
cases, these instructions are fleshed out by pragmatic principles to give
the full content of the utterance on any occasion of use.
This conception of linguistic semantics, and the unitary analysis of
discourse and bound-variable anaphora I am going to propose, are both
dependent on the theory of relevance. There are five main claims of the
theory that I make direct use of.
4 / Ruth M. Kempson
1. Utterance interpretation is an inferential, largely deductive exer-

cise which involves the construction by the hearer of a context set of
premises which combine with the propositional form of what he hears to
yield indirect implications (the implicit content of the utterance—
roughly equivalent to implicatures). Thus, contextual information is not
in general antecedently given, nor accumulated throughout a discourse:
it is the construction of the context that is an essential part of
utterance interpretation.
2. The linguistic content of sentences of natural language is under-
determined with respect to propositional content; and the construction
of a context set of premises, the decision as to the propositional content
expressed by the sentence, and the consequent deduction of contextual
implications are driven by a single principle, that of relevance.
3. The speaker has done his best in the circumstances to say some-
thing of maximal relevance to the hearer. This principle of relevance is
the sole maxim of the theory. It needs some explication. Relevance
itself is defined as the nontrivial deduction of contextual implications
from a pair comprising a context set of premises, and a proposition
expressed. The hearer's task is to select what that pair should be. The
principle of relevance is the guarantee that the speaker believes that the
form which he uttered makes immediately accessible to the hearer a
context set and a proposition from which he can derive contextual impli-
cations. Implicit in the concept of maximal relevance is a trade-off
between maximum amount of information (contextual implications) for
minimal processing cost. (The motivation behind this is the stated aim
of Sperber and Wilson to provide a theory of pragmatics which is not yet
another module, but a theory of performance in which memory storage,
processing costs, inferential properties of the central cognitive
mechanism, information presented in a grammar, all come together.)
4. Linguistic specification of elements of language may involve a dual
specification: (1) their contribution to propositional content, (2) their
contribution to what information is made accessible for purposes of
context construction.
5. The only factor that constrains the construction of contexts in
addition to any relevant specification of elements of the language and
the principle of relevance, is the assumption that certain types of infor-
mation are immediately accessible, viz. the preceding utterance, the
scenario of the utterance itself, and information associated with con-
cepts expressed by the lexical items used (this last stored in a mental
lexicon of discrete concepts with both encyclopedic and analytic knowl-
edge entered with such concepts).
It is this concept of accessibility that is central to my analysis of
anaphora. What I propose is that the concept of definiteness associated
with both pronouns and definite NPs simply is that of guaranteed access-
ibility. If a speaker uses a pronoun or a definite NP, then he is indi-
cating to the hearer that a representation of an NP type is immediately
accessible to him in the sense specified—either from the scenario of the
utterance itself, or from the preceding utterance, or from preceding
parts of the same utterance, or from concepts expressed by what pre-
cedes the anaphor. Thus, in (2) the referent of he has to be accessible
from the immediate scenario. In (14) it is provided by the previous
utterance, in (13) by some previous representation in the same

utterance; and in the bridging cross-reference cases of (3) and (4) by a
premise, accessible from the conceptual address associated with car,
that cars normally have a driver. This analysis is, of course, merely a
restatement of earlier characterisations of the concept of definiteness
(cf. Jesperson 1942, Chafe 1976). But what is new in this account is the
way in which the bridging cross-reference cases and cases where extra
premises are required, i.e. (A)-(C), are all immediate consequences of
the analysis, given the Sperber and Wilson framework. Since the content
of all anaphoric expressions is a guarantee that an antecedent is imme-
diately available, we can predict that where no such antecedent is
provided—either by the explicit content of the discourse or indexically—
the anaphor will act as an instruction to the hearer to construct a con-
text premise which will provide that antecedent as the implicit content
of the discourse. Consider example (3): on this analysis, the use of an
anaphoric expression is by the principle of relevance a guarantee of
instant accessibility of some representation of an individual described by
the predicate 'driver' about which the speaker is making some asser-
tion. But the immediately accessible environment (in this case, just the
preceding utterance) does not provide such a representation. Yet the
speaker is using an anaphor as the guarantee of such a representation.
He must therefore be using the anaphor as an indication to the hearer
that he should construct a context premise such that the appropriate
representation is derived as the implicit content of what he hears. This
effect is directly predictable from the proposed analysis in terms of
accessibility and the principle of relevance, without any further postu-
lation. Cases such as (9) are simply the same. The only difference is
that the contextual premise required (triggered by use of the predicate
insult) cannot be accessed from any particular constituent of the
preceding utterance. It is the whole of Jake called Jess a Conservative
that combines with 'To call someone a Conservative is to give them an
insult' to provide by deduction the contextual implication 'Jake gave Jess
an insult'. And the only way to construe (9) so that the expression the
insult fulfils the guarantee of accessibility of its antecedent is to con-
struct such a context set.
I have so far characterised a definite NP as expressing, as its intrinsic
content, a guarantee of instant accessibility of its antecedent. But the
guarantee of immediate accessibility simply is an intrinsic part of the
principle of relevance. It is this that determines the context set of
propositions and the propositional content the hearer selects. Thus, all
we require of an analysis of anaphora is that an anaphor is some expres-
sion whose value is not given by rules of grammar. For all the rest will
fall out from the application of the principle of relevance. And this is
what my analysis will provide. An anaphor will be represented as a
metavariable whose value is not determined by any principle of gram-
mar. Given my assumption of the Sperber-Wilson framework, it follows
that it will have to be identified by a relevance-controlled principle of
antecedent identification.
What we need now is an account of quantification compatible with this
analysis. In order to get this account to extend to sequences relative to
the binding of a quantifier, all we need to do is to assume that we can
manipulate a name-like entity which can stand arbitrarily for any one of
6 / Ruth M. Kempson
the individuals over which the domain of the quantifier ranges. Arbi-
trary names, manipulated in natural deduction systems, do just this. And
this is what I shall do.
Given, then, that a quantifier introduces an arbitrary name, this name
will be accessible in just the same way as a referring expression, and can
enter into deductive and context-specifying processes. The only
difference is that its availability is restricted to the scope of the quanti-
fier that introduced it, so it is not invariably available like a name. And
this is just the distinction we want. In particular, we predict exactly the
phenomena listed in (A)-(C) initially:
(A) The phenomenon of bridging cross-reference

(B) The use of anaphoric expressions can be licensed by extra
information made available by the total implicit content of
the preceding sentence and not any one constituent.
(C) One may have to manipulate extra premises and principles of
deduction in order to establish an antecedent.
Yet we also preclude examples such as (16).
(16) *A woman grinned at every man^. Hej smiled back.
This binding is precluded because the pronoun is not within the scope of
the quantifier, so the name associated with the quantifier is not access-
ible as an antecedent for that pronoun. Consider example (17), equival-
ent in all important respects to (S).
(17) Every singer worries that the accompanist is too loud.
This is a paradigm case where we have to manipulate information about

accompanists and singers in order to determine not merely what follows
from what the speaker has said, but also what it is that he has said. I
assume that the quantified expression every singer introduces an arbi-
trary name 'aj 1 , and definite NPs, like pronouns, are characterised by a
special variable ' /3', where /3 is some variable whose distinguishing
feature is that it is not characterised by the grammar. The expression
the accompanist is accordingly characterised as 'accompanist( 0)'. The
interpretation instructions associated with the lexical items and the
individual nodes of the tree provide a mapping up the tree in the manner
of Katz's (1972) 'projection' rules, projecting the sentence string onto:
singer(ai)—*aj worries that /3is too loud and accompanist (/3). Now aj
is not identical to /?, so there is no antecedent for/Sin this immediate
discourse. However, the (8-variable is used as a guarantee that there is
some such antecedent, even when the intended referent is not index-
ical. So, in order to identify the antecedent of /3 from this immediate
discourse, the hearer is forced to access what information he knows
about the concepts used in the sentence—in this case the information
associated with singer that 'Singers have accompanists'. This
information, by an orthodox process of logical deduction, enables him to
deduce an antecedent for /3 in the form of: 3y (accompanist(y) & aj has
y). With this information, he can therefore construct the total
propositional content asserted, to wit: Vx (singer(x) —» 3y (accompa-
nist^) & x has y & x worries that y is too loud)). The significance of this
example is twofold. On the one hand, the example demonstrates how the
interpretation of some element bound to a quantifier as a variable (in
this case, the definite NP the accompanist) may involve interaction
between information in the grammar and contingent world knowledge in
a way that is precluded by most current accounts of variable-binding.
On the other hand, it demonstrates how the process of working out the
implicit content of the utterance by the hearer takes place even for
subparts of a sentence-string. That is, the process involved in deducing
an antecedent for the accompanist is exactly that involved in deducing
implicatures of a standard sort; and, as in the case of implicature calcu-
lation, is a direct consequence of the principle of relevance.
I am not attempting here to give a formal specification of the interac-
tion between grammar and pragmatics, so this informal sketch is just a
promise of a properly formal account (cf. Kempson 1984 for the specifi-
cation of a formal recursive algorithm projecting linguistic strings onto
fully specified logical forms). However, even at this informal level,
there is more to say about this concept of accessibility. For I am claim-
ing that all anaphoric dependencies, quantifier-bound and discourse-
bound, can be explained in terms of this concept, that there is therefore
no configurational difference between the two types except as can be
predicted from the nature of the antecedent. How then can it explain
the asymmetry between (18) and (19)?
(18) His} mother likes John^.

(19) *Hisj mother likes everyone^.
These represent the so-called 'weak cross-over1 facts, a phenomenon
that constitutes a by now familiar argument for a distinction between
coreference and bound-variable pronominal anaphora and for a con-
figurational restriction associated with the latter which has to be con-
figurationally stipulated (cf. Chomsky 1977, Higginbotham 1980). The
analysis proposed here has a straightforward account, if we make one
plausible extension to it—that proper names should be analysed like
pronouns. So the proper name, John, like a pronoun, is a guarantee of
there being some immediately accessible individual named John. In (19)
then, given the system of projection rules associated with lexical items
and constituent structure, we predict that the quantifier every binds its
variable in object position, thereby guaranteeing that any arbitrary name
associated with it is not accessible for purposes of identification with
the subject, since the subject does not have information about the arbi-
trary name bound by the quantifier accessible to it. In (18), by contrast,
there is an independent guarantee (by the use of the name John) that
some individual named John is believed by the speaker to be immediately
accessible to the hearer. But if such an individual is immediately and
uniquely accessible, then that individual can be accessed directly by the
pronoun he_. To put it in possibly more familiar terms, he_ and John may
be coreferential but he_ is not referentially dependent on John; and the
account of anaphor-antecedent relations proposed here is exclusively in
8 / Ruth M. Kempson
terms of anaphoric dependency. So within the theory, the relation

between h<2 and John is coincidental yet predictable as a possiblity, given
the analysis of proper names.
But, you should reply, this assumes a simplistic left-to-right account
of both processing and anaphoric dependencies and we know this is not
right for two reasons: (1) not all pronouns display a dependency with an
antecedent to their left, as witness (20). (2) NPs are not necessarily
interpreted left-to-right, as witness the phenomenon of scope variation
exemplified by (21).
(20) By hisj bed each child i kept a teddy bear.

(21) Everyone in this room speaks two languages.
So why can't a hearer 'sit tight1 on the interpretation of the pronoun in
(19), so to speak, and identify it with the later arbitrary name associated
with the quantifier, this being given wide scope over the whole sen-
tence?
The answer is simple, and predictable. Hearers cannot randomly
reject a left-right processing mechanism, but only as the linguistic
structure of what they hear gives explicit directions to that effect.
There are two types of case where this is possible, one obligatory, the
other optional.
The obligatory case is represented by (20). Preposed prepositional
phrases are modifying expressions which are interpreted as modifications
either of VPs or of Ss. In semantic terms, they are functional expres-
sions which operate on VPs or Ss, to give another expression of that
same sort. In other words, they are expressions which have to be inter-
preted as modifications of the constituent they modify. Thus they are
only interpreted together with the interpretation of that constituent,
and the projection rule associated with them has to specify this. Hence
we get the reversed dependency directions in these constructions:
(22) Each child^ kept a teddy bear by his: bed.

(23) *He-l kept a teddy bear by each childj's bed.
(24) *By each childj's bed, hej kept a teddy bear.
(20) By his: bed, each child: kept a teddy bear.
The optional case is the scope vacillation displayed in (21). Orthodox
accounts of scope variation characterise it as a phenomenon to which all
NPs are subject, with the caveat that in the case of referential NPs it
makes no difference because their interpretation is independently
fixed. But suppose we say instead that scope variation between NPs is
available only to pairs (or triples) of nonreferential expressions. This
could either be a stipulation of the rules of interpretation associated
with quantifiers, or a direct consequence of the principle of relevance.
It is certainly not without pragmatic motivation—nonreferential NPs are
the only NPs whose interpretation is not fixed by the guarantee of
immediate accessibility. Either way, the predictions made are now
correct—(18) and (19) do not contain a pair of quantified expressions. In
particular, (19) contains only one. Hence it is not subject to scope
variation. Moreover, the pronoun is not contained within a constituent
which permits its interpretation to be subsequent to that of the quanti-

fied expression. Hence the interpretation is left-right and the pronoun
has to be identified independently of that quantified expression. The
account of anaphora in terms of accessibility is vindicated.
Despite the informality of this presentation, I hope that I have shown,
at least in principle, how with just three assumptions—relevance theory,
an uncontroversial mechanism for quantifiers, and a minimum linguistic
account of the definiteness intrinsic to pronouns and definite noun
phrases—we can predict the bridging cross-reference phenomenon, its
interaction with quantifier binding, and the weak cross-over effects, all
without stipulations particular to the structures in question. There are
two nontheoretical consequences of this analysis which should not be
missed. First, there should be no level of logical form within the gram-
mar which characterises the binding of definite anaphors since all
definite anaphors are subject to exactly the same pragmatic explanation.
Second, the linguistic content of a sentence is neither a full specifica-
tion of the truth conditions of the proposition it expresses, nor is it
exclusively directed towards such specification.
And here, finally, we have the basis for characterising discourse-
related phenomena without having to abandon the assumption that a
grammar is a sentence-generating mechanism. For the semantic specifi-
cation of linguistic content within this framework is not a set of rules
which predicts actual interpretations but rather is a set of instructions
on utterance interpretation, a set which may include instructions on
propositional content, indications of accessibility or more specific
indications of context-construction, or even indications as to the
speaker's attitude to what he says. Each of these instructions is then
fleshed out by the principle of relevance to give the full content of the
utterance. It is striking that those aspects of semantic specification
which were most embarrassing to a straight truth-conditional semantics
as part of sentence-based grammar fall out as a direct consequence of
the stand taken here, despite the fact that it, too, assumes that gram-
mars are sentence-generating devices. Thus, by way of a coda, I propose
that we abandon the Gazdar slogan 'PRAGMATICS = MEANING -
TRUTH CONDITIONS' (Gazdar 1979) in favour of a new slogan:
SEMANTICS = UTTERANCE INTERPRETATION - THE PRINCIPLE OF
RELEVANCE.
Note
1. There is no stipulation of uniqueness as an intrinsic property of
definiteness, either. Antecedent identification is made in virtue of the
guarantee of the principle of relevance that immediately accessible to
the hearer is a representation of an individual about whom he is to
understand the speaker as making an assertion. If there were any doubt
as to which individual that should be, the hearer would have to put
processing effort into deciding which individual it was. But the speaker's
utterance in that form is a guarantee that he believes no such processing
cost is necessary. Given the principle of relevance then, the speaker
must be intending to convey that there is only one such individual.
10 / Ruth M. Kempson
References
Chafe, Wallace L. 1976. Givenness, contrastiveness, definiteness,

subjects, topic, and point of view. In: Subject and topic. Edited by
Charles Li. New York: Academic Press. 25-56.
Chomsky, Noam. 1977. Conditions on rules of grammar. Essays in form
and interpretation. Amsterdam: North-Holland.
Gazdar, Gerald. 1979. Pragmatics. New York: Academic Press.
Higginbotham, James. 1980. Pronouns and bound variables. Linguistics
Inquiry 11.679-708.
Jesperson, Otto. 1942. A modern English grammar. London: Allen and
Unwin.
Katz, Jerrold. 1972. Semantic theory. New York: Harper.
Kempson, Ruth M. 1984. Anaphora, the compositionality requirement,
and the semantics-pragmatics distinction. North-Eastern Linguistics
Society Proceedings, vol. 14.
Sperber, Dan, and Deirdre Wilson, (forthcoming) Relevance: Founda-
tions of pragmatic theory.
TOWARD A NEW TAXONOMY FOR PRAGMATIC INFERENCE:
Q-BASED AND R-BASED IMPLICATURE
Laurence R. Horn
Yale University
1. The Principle of Least Effort (and the Principle of Sufficient
Effort). Thirty-five years ago, George Kingsley Zipf set out to explain
all of natural language (along with almost everything else in the human
universe, from dreams, art, and ritual to war, schizophrenia, and the
incest taboo) in terms of an overarching Principle of Least Effort. In
the linguistic realm, however, Zipf (1949:20ff.) acknowledged two basic
and competing forces. The Force of Unification, or Speaker's Economy,
is a direct least effort correlate, a drive toward simplification which,
operating unchecked, would result in the evolution of exactly one totally
unmarked infinitely ambiguous vocable (presumably uhhhh). The anti-
thetical Force of Diversification, or Auditor's Economy, is an anti-
ambiguity principle leading toward the establishment of as many differ-
ent expressions as there are messages to communicate. Given nn mean-
ings, the speaker's economy will tend toward 'a vocabulary of one word
which will refer to all the rn distinct meanings', while the hearer's econ-
omy will tend toward 'a vocabulary of nn different words with one dis-
tinct meaning for each word'. As Zipf (1949:21) (under)states, 'The two
opposing economies are in extreme conflict.'
It is in the crucible of this conflict, argue Zipf, Martinet, and allied
functionalists, that language change is forged. As Martinet notes
(1962:139):
In order to understand how and why a language changes, the
linguist must keep in mind two ever-present and antinomic fac-
tors: first, the requirements of communication, the need for the
speaker to convey his message, and second, the principle of least
effort, which makes him restrict his output of energy, both mental
and physical, to the minimum compatible with achieving his ends.
In this paper, I seek to demonstrate that these same two antinomic
forces—and the interaction between them—are largely responsible for
11
12 / Laurence R. Horn
generating Grice's conversational maxims and the schema for pragmatic

inference derived therefrom.
2. The maxims of conversation. In his ground-breaking work on

language use and the logic of conversation, Grice (1975) suggests a
procedure whereby participants in a conversational context may com-
pute what was meant (by a given speaker's contributing a given utter-
ance at a given point in the interaction) based on what was said (by that
speaker, in that utterance, at that point). The governing dictum is the
Cooperative Principle: 'Make your conversational contribution such as is
required, at the stage at which it occurs, by the accepted purpose or
direction of the talk exchange' (1975:45). Within this basic guideline,
Grice establishes four specific subprinciples, the general, almost trivial-
sounding, and presumably universal maxims of conversation which he
takes to govern all rational interchange (Grice 1975:45-46; cf. Levinson
1983:1 OOff. for discussion).
Quality: Try to make your contribution one that is true.

1. Do not say what you believe to be false.
2. Do not say that for which you lack evidence.
Quantity:
1. Make your contribution as informative as is required (for
the current purposes of the exchange).
2. Do not make your contribution more informative than is
required.
Relation:
Be relevant.
Manner:
1. Avoid obscurity of expression.
2. Avoid ambiguity.
3. Be brief. (Avoid unnecessary prolixity).
4. Be orderly.
The assumption that speaker and hearer are both observing the Coopera-
tive Principle and its component maxims permits the exploitation of
these maxims to generate conversational implicata, conveyed messages
which are meant without being said.
The partial reductionist program I envision would retain the Maxim of
Quality with its special character noted by Grice: unless Quality (or
what Lewis 1969 has called a Convention of Truthfulness) obtains, the
entire conversational and implicatural apparatus collapses. But the first
Quantity maxim (henceforth Quantity j , or simply Quantity) is essentially
Zipf's Auditor's Economy, the need for the speaker to convey his mes-
sage fully. Most if not all of the remaining Gricean rules respond to the
Speaker's Economy, either directly (as consequences of the least effort
principle) or indirectly (through the interaction of this principle with its
antithesis). Notice in particular that the second quantity principle, as
stated, is essentially akin to Relation (what would make a contribution
more informative than required, except the inclusion of material not
strictly relevant to and needed for the matter at hand?). Note also that
Grice in effect builds in Relation when defining Quantityj, and similarly
Toward a new taxonomy for pragmatic inference / 13
builds in Quantity j in defining Quantity2. In section 3, I have

tentatively boiled down the maxims (leaving aside Quality) to two
fundamental principles, echoing the two functional economies of Zipf
and Martinet.
3. Minding our Qs and Rs.
(la) The Q Principle (Hearer-based):

MAKE YOUR CONTRIBUTION SUFFICIENT (cf. Quantityi)
SAY AS MUCH AS YOU CAN (given R)
Lower-bounding principle, inducing upper-bounding implicata
(lb) The R Principle (Speaker-based):
MAKE YOUR CONTRIBUTION NECESSARY (cf. Relation,
Quantity2, Manner)
SAY NO MORE THAN YOU MUST (given Q)
Upper-bounding principle, inducing lower-bounding implicata
In the current coy style (cf. Chomsky 1982) which has given us D-struc-
tures and S-structures, LF and PF, R-expressions and A-binding (corre-
sponding, more or less, to Deep and Surface structures, Logical and
Phonetic form, Referring expressions, and Argument-binding, respec-
tively), I am using Q and R to evoke Quantity (i.e. Quantity^) and Rela-
tion while leaving open the extent to which my principles map onto these
two maxims.
The hearer-based Q Principle is essentially a sufficiency condition. A
lower-bounding law in terms of information structure, it may be (and
systematically is) exploited to generate upper-bounding conversational
implicata, as described by Grice (1975), Ducrot (1972), Horn (1972,
1973), and Gazdar (1979): a speaker, in saying '...p... 1 , implicates that
(for all she knows) '...at most p...'
The primary examples of generalized Q-based implicata arise from
scalar predications. If I tell you that some of my friends are Buddhists, I
license you to draw the inference that not all my friends are Buddhists.
(If I knew they all were, and this knowledge was relevant to your inter-
ests, it would have been incumbent on me to obey the Q Principle and
say so; the assumption that I am obeying Quantity allows you to infer
that I did not know for a fact that the stronger predication —All of my
friends are Buddhists—held.) Like all rules of pragmatic inference, Q-
based implicature is context-dependent; as a generalized implicatum,
the aforementioned inference goes through in unmarked contexts, but it
may be cancelled—explicitly (cf. Some, {if not all/and perhaps all,] of my
friends...) or implicitly (by establishing the appropriate context, in
which all that is relevant, or can be known, is the lower bound).
Examples of Q-based scalar implicature are legion. The following
sentences all assert (or entail) a lower bound ('at least ', their 'one-
sided' reading as Aristotle dubbed it), and characteristically implicate an
upper bound ('at most 0; the conjunction of the assertion and impli-
catum results in conveying the corresponding 'two-sided' understanding
('exactly ').
/ Laurence R. Horn
1-sided: 'at least 3'

(2a) He ate 3 carrots <: , e x aictly
2 $ i d e d : c t l y 3,
3'
1-sided: 'some if not all'
(2b) You ate some of the c o o k i e s < r -, ., , . • . . ...
1
——2-sided: 'some but not all
/ - v TJ . ... , ... . ____-—1-sided: 'possible if not certain 1
(2c) I t s possible
K she'll w i n < r i - > - J J I -ui u ^ ± ± •i
• .2-sided: 'possible but not certain'
/ o ,x , , . . . . .. . 1-sided = inclusive or
(2d) Maggie
&5 is Kpatriotic or Mquixotic B = ^ ^ o .. , . . —
•—2-sided = exclusive or_
1
/o \ Tl . _____—1 -sided: 'happy if not ecstatic
(2e) I'm happy VV3 <=cT -> J J 'happy
^--~-2-sided: .u u .. not.. ecstatic'
but ,. ^- ,
/->x\ TJ.t 1-sided: 'at least warm'
(2f) It's warms=ri! \ j j . .. .. u ..i
2-sided: 'warm ubut not hot1
(cf. Horn 1972, 1973, to appear b for additional details)
This analysis, foreshadowed a century ago by DeMorgan (1847:100) and

Mill (1867:501), allows us to preserve semantic parsimony by taking the
infinitely many weak scalar predicates to be logically unambiguous while
pragmatically ambivalent (cf. Hamilton 1860 and Kempson 1980 for
alternative accounts in which the sentences of (2) are treated as logi-
cally ambiguous; these accounts are reviewed and rebutted in Horn 1973,
to appear b).
If the Q-Principle corresponds to Quantityj, the countervailing R-
Principle collects not only Relation, but Quantity2 and possibly all the
manner maxims (although I shall not go through the arguments for that
claim here). The R-Principle, mirroring the effect of the Q-Principle
just discussed, is an upper-bounding principle which may be (and stan-
dardly is) exploited to generate lower-bounding implicata. A speaker
who says '...p... 1 may license the Q-inference that he meant '...at most
p...'; a speaker who says '..p...' may license the R-inference that he
meant '...more than p...' The locus classicus here is the indirect speech
act (Heringer 1972; Gordon and Lakoff 1975; Searle 1975): if I ask you
whether you can pass me the salt, in a context where your abilities to do
so are not in doubt, I license you to infer that I am doing something more
that asking you whether you can pass the salt—I am in fact asking you to
do it. (If I know for a fact that you can pass me the salt, the yes-no
question is pointless; the assumption that I am obeying the Relation
maxim allows you to infer that I mean something more than what I say.)
Grice notes that a speaker may 'quietly and unostentatiously violate a
maxim' as well as exploit it to generate an implicature. Clark and
Haviland (1977:2) have suggested that intentional covert maxim viola-
tions result in lies, while unintentional violations are simply misleading.
In fact, what is crucial is just which sort of maxim or pragmatic prin-
ciple is violated: intentional quality violations result in lies (another
reason for the special status of quality; cf. Coleman and Kay 1981 for
additional factors in defining jie), intentional violations of the Q-based
sufficiency principle result in a speaker's misleading the addressee, and
intentional violations of the R-based least effort principle are often
simply unhelpful or perverse. A courtroom witness must swear to tell
the whole truth and nothing but the truth, i.e. to obey quantity and
quality, while violations of relevance lead only to a possible lawyer's
objection or judge's scolding.
*. Q-based vs. R-based inference: Some early skirmishes. Like the

antinomic economies from which they derive, the Q-based and R-based
principles just outlined often directly collide. A speaker obeying only Q
would tend to say everything she knows on the off-chance that it might
prove informative, while a speaker obeying only R would probably, to be
on the safe side, not open her mouth. In fact, many of the maxim
clashes Grice and others have discussed do involve Quantity} vs. Rela-
tion. In delineating the operation of quantity to generate upper-bound-
ing scalar implicatures in my thesis (Horn 1972: Chapters 1 and 2), I was
bothered by the contrast between (3a), where Quantity is in force, and
(3b), where the principle of relevance is apparently responsible for the
implicatum (as Karttunen 1970 suggests).
(3a) It is possible that John solved the problem *

(For all S knows) John didn't solve the problem (Q)
(3b) John was able to solve the problem >
John solved the problem (R)
An application of quantity in the latter case would generate the opposite

implicatum, viz. that John didn't solve the problem. Similarly, in (4a)
quantity leads to an upper bound on the information communicated,
while in (4b) an R-based inference renders the indefinite more informa-
tive than its logical form suggests (both examples from Grice 1975:56;
cf. Harnish 1976).
(4a) X is meeting a woman this evening >

The woman in question is not X's wife, sister, or
close platonic friend (Q)
(4b) I broke a finger yesterday )
The finger is mine (R)
5. Politeness rules, pragmatic competence, and maxim clashes. A

parallel to these clashes emerges from work by Robin Lakoff on the
maxims of politeness outlined in (5), and by Brown and Levinson in their
reformulation of Lakoff's maxims in terms of face wants and needs:
(5) Rules of politeness (Lakoff 1973, Brown and Levinson 1979)
Lakoff: Brown and Levinson:
Rule 1. Deference: 'Don't Negative politeness, polite
impose', 'Keep aloof formality
Rule 2. Give options (special Respect 'negative face', i.e.
case of Rule 1, or freedom from imposition and
derived as theorem freedom of action.
from Rules 1 and 3?)
Rule 3. Be friendly Positive politeness, polite
(Camaraderie) friendliness
Respect 'positive face', i.e.
positive consistent self-
image and approval/ appre-
ciation by others.
Crucially for us, the Deference or Aloofness maxim, Rule 1, is an upper-

bounding R-based constraint on one's actions, while Friendliness, Rule 3,
is a lower-bounding Q-based imperative.
Lakoff has noted several instances in which Rule 1 Politeness seems to
clash with Rule 3, e.g. belching after a meal, which may signal appreci-
ation (via Rule 3) or may simply be taken as offensive (by Rule 1). More
systematically, when we look at languages like those in the Romance,
Germanic, Slavic, and Indie families of Indo-European, in which two
different forms for the second person singular coexist but are associated
with different presuppositions or conventional implicata (cf. Levinson
1983:128-29), we find that the use of a (familiar) T form rather than the
more formal V form for the addressee may convey polite friendliness by
Rule 3 or presumptuousness by Rule 1 (depending on the context and the
interlocutors' assumptions and beliefs). Similarly, the use of a V form
for the addressee may be polite by Rule 1, or aloof and unfriendly by
Rule 3.
A particularly devastating kind of maxim clash which has been ana-
lyzed in similar terms is the communication mixup example from work
by Tannen.
(6) Conversational breakdowns and marital breakups (Tannen
1975, 1979)
First exchange:
Wife: Bob's having a party. You wanna go?
Husband: OK.
Second exchange (later):
Wife: Are you sure you want to go?
Husband: OK. Let's not go. I'm tired anyway.
Post-mortem:
Wife: We didn't go to the party because you didn't want to.
Husband: ^ wanted to. You didn't want to.
On Tannen's gloss of this canonical interchange, one partner (the wife) is
operating on a direct strategy utilizing Rule 3 politeness: if one had
meant more, s/he would (and should) have said it. Her partner (or oppo-
nent), on the other hand, is employing an indirect, hint-seeking strategy
which emanates from Rule 1 politeness: avoid saying too much when you
can get it across by hints. As Tannen observes, each strategy may link
up with a different pragmatic competence, the difference involving not
the set of operative rules but the relative strength of opposing rules
within that set.
Notice that Tannen's account of the clash instantiated in (6) super-
imposes directly onto an alternative gloss utilizing Grice's Quantity and
Relation maxims (or Quantity j and Quantity 2 ), respectively. Crucially,
in either version we have to deal with Q-based vs. R-based inference
patterns.
A closely related example of maxim clash is cited by Keenan (1976).
As she reports and interprets the facts, the Malagasy-speaking culture of
Madagascar is a speech community in which the Cooperative Principle,
and in particular the maxim of Quantity, is not observed. Informative-
ness is apparently absent as a working principle in conversation, and the
participants in a talk-exchange tend not to draw what we have been

calling Q-based inferences in situations where such inferences would be
drawn in Western communities.
[Interlocutors] regularly provide less information than is required by

their conversational partner, even though they have access to the
necessary information. If A asks B, 'Where is your mother? 1 and B
responds, 'She is either in the house or at the market', B's utterance
is not usually taken to imply that B is unable to provide more
specific information needed by the hearer. The implicature is not
made, because the expectation that speakers will satisfy informa-
tional needs is not a basic norm. (Keenan 1976:70)
In fact, however, as Keenan makes clear in her discussion, Quantity

and Q-based inference do play a significant role in some conversational
contexts. Where they do not, it is because Quantity is overridden by a
countervailing R-based principle, in particular the imperative of avoid-
ing tsiny (the responsibility, guilt, or other unpleasant consequences
incurred by uttering claims which turn out to be false and/or offensive
to other members of the society living or dead). Sex of speaker, and
significance and accessibility of the information contributed, are other
variables influencing the relative weights of Q-based and R-based prin-
ciples and inference patterns in the Malagasy community. (Cf. Prince
1982 for a related critique of Keenan's conclusions.)
Both Tannen's and Keenan's cases involve situations in which different
speakers practice different utilizations of essentially the same tools;
pragmatic competence often differs across cultures—and across speakers
within the same culture—in accordance with the assignment of relative
weightings to different maxims or principles, and consequently with the
inference patterns associated with the exploitation of those maxims.
6. Quantity vs. informativeness. The most detailed and careful

discussion in the literature of Q vs. R clashes in English is due to Atlas
and Levinson (1981) (cf. Levinson 1983: section 3.2 for related discus-
sion). Atlas and Levinson begin by summarizing the evidence for the
inference from Quantity for the scalar cases. Examples (after Horn
1972 and Gazdar 1979) include the implicata in (7a); the general prin-
ciple for the Q-based cases is given in (7b).
(7) The inference from Quantity: What is communicated is more
informative, more definite than what is said.
(7a) some*—> not all

may<—>may not
3—> no more than 3
p or q—> not both p and q
a believes that p—»a does not know that p
(cf. discussion of example (3).)
(7b) 'Given that there is available an expression of roughly
equal length that is logically stronger and/or more informa-
tive, the failure to employ the stronger expression conveys
that the speaker is not in a position to employ it.' (Atlas and
Levinson 1981:38)
But, Atlas and Levinson point out, there is a substantial class of

(somewhat disparate) cases for which Quantity gives exactly the wrong
results. For these cases, including those in (8a), they invoke a Principle
of Informativeness, (8b).
(8) The inference from Informativeness: What is communicated

is more precise than (is a subcase of) what is said.
(8a) if p then q—>if -p then -q (Geis and Zwicky's 1971 'invited

inference' of conditional per-
fection)
p and q-—»p preceded q (cf. Grice 1975, Schmerling 1975 on
MJ 'asymmetric 1 conjunction)
p caused q
a and b VP'd—» a and b VP'd together
a ate the cake—»a ate the whole cake \ , , u . . 1Q7/ .s
(cf Harnish 1976)
a ate the a p p l e s - * a ate all the apples) '
Do you know the time?—> If you know the time, tell me what
it is. (Indirect speech acts, cf. Cole and Morgan 1975)
I don't think that p—> (I think that) not-p (Neg-raising, cf.
Horn 1978b)
I have a new car and the windows don't close—>...the windows
of my new car.... (Bridging inferences, cf. Clark and Havi-
land 1976)
(8b) The Principle of Informativeness: 'Read as much into an

utterance as is consistent with what you know about the
world' (Levinson 1983:146-47)
Atlas and Levinson (1981:42) formulate their informativeness-based

inference as an 'inference to the best interpretation'.
(8c) 'If a predicate Q is semantically nonspecific with respect to
predicates Pj, l < i < n , but for some j , l < j < n , P: is stereo-
typical of Qs, then in saying rQt1 a speaker will convey fP:t1.'
The key notion here is the restriction of a more general predicate to a

stereotypical instance. Two examples of this inference to a salient
subset or exemplar cited by Atlas and Levinson are those in (8d).
(8d) The secretary smiled. —>The female secretary smiled.

John had a drink.—* John had an alcoholic drink.
The class of indefinite descriptions provides a major source of clashes
between Quantity-based inference (cf. (9a)) and Informativeness-based
inference (cf. (9b)).
(9) Quantity vs. informativeness (Atlas and Levinson 1981:49ff.; cf.
Harnish 1976)
(9a) Quantity in force: (9b) Informativeness in force:
I slept on a boat yesterday—> I lost a book yesterday—>
The boat was not mine The book was mine.
I slept in a car yesterday—> I broke a finger yesterday—>
The car was not mine The finger was mine
Mort and David took a Mort and David bought a
shower—) piano >
They took separate They bought a piano
showers together
A number of factors are involved in determining which principle takes
precedence when the two are at odds; notice, for example, that the
speaker in (9a) could have chosen the more precise genitive form (I slept
on my boat, in my car) but did not do so, while the use of the genitive in
(9b) (my book, my finger) might suggest wrongly that I have but one book
or one finger. Atlas and Levinson cite two additional factors which are
relevant to the weighting of Quantity vs. Informativeness: if an entail-
ment-based 'Horn scale' (cf. Horn 1972) can be constructed on which the
predicates in question can be readily ranked, Quantity is more likely to
win out; if the application of Quantity tends to contradict our assumed
'Conventions of Noncontroversiality', Informativeness takes precedence.
Within the framework being explored here, Atlas and Levinson's in-
ference from Quantity corresponds directly to our Q Principle. There is
an equally strong parallel between Atlas and Levinson's inference from
Informativeness and our R Principle; my only objection to their formula-
tion concerns their terminology. First of all, Informativeness suggests
the Q Principle to me more strongly than, as intended, the R Principle
(note the language of Grice's Quantity! maxim, which Keenan 1976
indeed glosses 'Be informative'). Furthermore, Ducrot (1972), who
independently proposes a 'Loi d'exhaustivite'1 to do the work of Grice's
maxim of Quantity, also invokes a 'Loi d'informativite' of a rather dif-
ferent nature from Atlas and Levinson's Principle of Informativeness (cf.
Ducrot 1972:132-34). Thus, I feel justified in taking Atlas and Levin-
son's so-called Informativeness as an instance of the R Principle.
7. Negation as implicatum-canceller: Q-based vs. R-based infer-

ences. An opposition as fundamental as that between the Q Principle
and the R Principle would be expected to have major linguistic conse-
quences, and indeed this expectation is not disappointed. We begin by
considering the effect of negation on conversational implicata. What we
find is that Q-based implicata can be readily cancelled by a negation
which does not affect what is said (through what I have elsewhere
termed 'metalinguistic' negation: cf. Horn to appear a), but R-based
implicata cannot be cancelled by negation at all.
This is illustrated in what follows. First, some instances of negated
scalar implicata are given in (lOa-f). That negation is being used in a
marked way to deny the Q-implicated upper bound in these examples can
best be seen by considering the well-formed affirmative counterparts of
these sentences in (lOa'-f) (cf. (2); also Horn to appear b, and references
there).
(10a) He didn't eat 3 carrots—he ate k of them.

(10b) You didn't eat some of the cookies—you ate all of them.
(10c) It isn't possible she'll win—it's certain she will.
(lOd) She isn't patriotic or quixotic—she's both.
(lOe) I'm not happy—I'm ecstatic.
(lOf) It isn't warm—it's downright hot.
(10a1) He ate 3 carrots—in fact, he ate 4.
(10b1) You ate some of the cookies—indeed, you ate all of them.
(10c') Not only is it possible she'll win—it's certain she will.
(10d') She's patriotic or quixotic—in fact, she's both.
(10eO I'm happy—indeed, I'm ecstatic.
(10f) It's warm—in fact, it's downright hot.
But now consider R-based implicata. To say that someone was able to
solve the problem may R-implicate that she in fact solved it (Karttunen
1970). Similarly, to assert that someone was clever enough to do some-
thing generally implicates that he did it. As we have seen, my confiding
to you that I broke a finger would normally be taken to refer to a finger
of mine (given Atlas and Levinson's Conventions of Noncontroversiality)
unless I knew you knew that I was an enforcer for the mob. And, as
Wittgenstein and others have pointed out, an assertion of the form 'I
believe that S' would normally be taken as an indirect assertion of S
rather than merely a statement about my beliefs; this indirect speech
act, too, works by R-inference. Yet none of these implicata can be
cancelled by negation.
(lla) She wasn't able to solve the problem.

(4 She was able to solve it, but didn't)
(lib) He wasn't clever enough to figure out the answer.
(4 He was clever enough to do it, but he didn't)
(lie) I didn't break a finger yesterday.
(4 I broke a finger, but it wasn't one of mine)
(lid) I don't believe the Yankees will win the pennant.
(^ I believe they'll win the pennant, but I'm not fjveakljD
asserting that they will)
Why should this difference in cancellability exist? The answer lies in
the logic of Q-based and R-based inference. Let S represent a given
(stronger) proposition, and W the weaker proposition which it unilaterally
entails and from which the relevant implicata are to be drawn. In the
case of Q-based implicata, the assertion of 'W' Q-implicates -S. Where
W is a scalar predicate truth-conditionally defined by its lower bound,
the ordinary negation of W negates that lower bound, i.e. as 'less than
W', and is hence incompatible with S; the assertion that he did not eat
three carrots would be taken to amount to the assertion that he ate less
than three (and hence not four, five, or more). But 'not W1 uttered in a
context where S is affirmed (as in (10a-f)) self-destructs on the
unmarked 'less than W understanding and must therefore be sent back

through, in effect—whence the marked, metalinguistic quality of this
variety of negation.
In the case of R-based implicata, the assertion of 'W R-implicates not
-S but S: the proposition that she solved the problem unilaterally entails
the proposition that she was able to solve it (S entails W), but the asser-
tion that she was able to solve it may implicate that she in fact solved it
('W R-implicates S). Once again, 'not W signifies 'less than W and hence
licenses the inference of -S (via modus tollens from the original S tf-W
entailment). But crucially, there are no circumstances under which the
implicatum S is cancelled and 'not W cannot be interpreted consistently,
as an ordinary descriptive negation. The negation in (lla-d) thus never
gets sent back through to be interpreted metalinguistically, as an impli-
catum-canceller. Schematically, the situation we have is that in (12).
(12) Q-based implicata: R-based implicata:
S entails W S entails W
'W Q-implicates -S 'W R-implicates S
normally, 'not W = 'less normally, 'not W = 'less
than W, incompatible with S than W
'not W, asserted where S 'not W Ir-S (modus tollens)
is given, reinterpreted as 'not W never gets reinter-
metalinguistic negation preted, since it's always
compatible with -S (the
denial of W's implicatum)
Now, R-based implicata can be cancelled without negation, simply by
assigning the contradiction contour (cf. Liber man and Sag 1974) and
stressing the implicatum-inducing element.
(13) She was able to solve the problem (but she didn't solve it).
I believe the Yankees will win the pennant (but I'm not saying
they will).
I broke a finger today (but not one of mine).
Notice that we tend to get the opposite, Q-based implicatum in these
contexts; the contour which cancels the R-based inference sets up a
strong expectation for the kind of continuations exemplified above.
When we appear to get cancellation of an R-based implicatum by
negation, the implicatum in question has, in fact, become convention-
alized as part of literal meaning (cf. Grice 1975:58 and Morgan 1978 on
the gradual conventionalization of conversational implicata). Thus, for
example, predicate expressions which denote various personal relation-
ships may take on a narrowed symmetric sense (cf. X and Y are {mar-
ried/friends/lovers/in love |) but need not (cf. X and Y are spouses).
When the symmetric sense of these predicates is intended, negation may
leave the more general sense unaffected.
(1*0 They aren't jmarried/friends/in love}, (i.e. with or to each
other)
To confirm the claim that only conventionalized R-based implicata

can be cancelled by negation, we need only reconsider the pair of exam-
ples from Atlas and Levinson (1981) cited in (8d). Both speakers' intui-
tion and lexicographers' practice suggest that the implicatum associated
with drink ('alcoholic drink') has become fossilized into conventional
meaning, while the implicatum associated with secretary ('female secre-
tary') has not. Thus, in the terms of Horn (to appear b), drink represents
an autohyponymous lexical item while secretary does not. In this light,
negation behaves precisely as predicted.
(15a) My secretary didn't smile—I have a male secretary.
(15b) John didn't have a drink—that was a Shirley Temple.
A male secretary is still a secretary (although not one of the salient
variety), but a nonalcoholic beverage may or may not count as a drink.
8. The division of pragmatic labor. While the Q Principle and the R

Principle are diametrically opposed forces in inference strategies and
language change, it is perhaps in the resolution of the conflict between
them that they play their major role in both 'langue' and 'parole'. The
most general pattern for this resolution, the synthesis of the two anti-
theses, is summarized in (16) and derived more explicitly in (17a-f).
(16) The use of a marked (relatively complex and/or prolix) ex-

pression when a corresponding unmarked (simpler, less
'effortful') alternate expression is available tends to be
interpreted as conveying a marked message (one which
the unmarked alternative would not or could not have
conveyed).
(17a) The speaker used marked expression E' containing 'extra'

material (or otherwise less basic in form or distribution)
when a corresponding unmarked expression E, essentially
coextensive with it, was available.
(17b) Either (i) the 'extra' material was irrelevant and unneces-
sary, or (ii) it was necessary (i.e. E could not have been
appropriately used).
(17c) (17b(i)) is in conflict with the R Principle and is thus (ceteris
paribus) to be rejected.
(17d) Therefore, (17b(ii)), from (17b), (17c) by modus tollendo
ponens.
(17e) The unmarked alternative E tends to become associated (by
use or—through conventionalization—by meaning) with
unmarked situation s, representing stereotype or salient
member of extension of E/E1. (R-based inference; cf.
Atlas and Levinson, (8b), (8c).)
(17f) The marked alternative E1 tends to become associated with
the complement of s with respect to the original exten-
sion of E/E'. (Q-based inference; cf. (6.12))
The key steps in the argument are those sketched in (17a), (17e), and
(17f); they represent a characteristic shift which can be schematized as
in (18):
The result is an equilibrium which I shall call the division of pragmatic

labor. (The equilibrium is, in fact, somewhat unstable; in particular,
either the R-based inference represented in the second diagram or the
Q-based inference in the third can become conventionalized, as we shall
see later.) The remainder of this essay is devoted to rehearsing a num-
ber of instances of this pattern, varying in the details but essentially
parallel in terms of the overall dynamics.
9. Avoid Pronoun (and unavoidable pronouns). In considering the near-
complementary distribution between his abstract, phonetically unre-
alized PRO element (which usurps many of the functions of Equi-dele-
tion sites in the earlier Standard Theory) and overt pronominals, Chom-
sky (1982:65) arrives at a general principle which he calls Avoid Pronoun,
'interpreted as imposing a choice of PRO over an overt pronoun where
possible1. To illustrate, he cites the two cases of (19a-b).
(19a) John would much prefer Qiis going to the movie]
(19b) John would much prefer (his (own) booQ
PRO may appear in the frame of (19a) in place of the pronoun (cf. John
would much prefer going to the movie), and so his is taken here as non-
coreferential with John; in (19b), PRO cannot appear and the overt
pronominal his may be (and with own present must be) taken as corefer-
ential with John. As Chomsky notes, the Avoid Pronoun principle
might be regarded as a subcase of a conversational principle of not
saying more than is required, or might be related to a principle of
deletion-up-to-recoverability, but there is some reason to believe
that it functions as a principle of grammar. (Chomsky 1982:65)
Three important points emerge from this brief passage: the link
between Avoid Pronoun and analogous principles applying to deletion,
the grounding of the principle(s) in least effort (i.e. our R Principle), and
the fossilization or grammaticization of the functional principle into
conventional rules and constraints within particular grammars.
The transderivational nature of Avoid Pronoun is clear from
Chomsky's discussion of (19a-b) and an analogous contrast in French
(Chomsky 1982:146).
2* / Laurence R. Horn
(20a) Je veux qu'il vienne. 'I want him to come1
(20b) II: veut qu'il: vienne. 'He^ wants him: to come1 (where i t j)
(20c) *llj veut qu'il| vienne. 'He^ wants him(self)j to come1
Avoid Pronoun, in ruling out (20c), must refer to the availability of a
corresponding derivation, that of (21).
(21) II veut venir. 'He wants to come1
In (21) the subject of 'come' is realized as an abstract PRO argument of
the embedded infinitive rather than an overt pronominal subject of an
embedded subjunctive, as in (20a-c).
The applicability of Avoid Pronoun is also partially dependent on
ambiguity avoidance, a fact not discussed by Chomsky. Thus, in the
English versions of (20a-c) and similar sentences, coreference is ruled
out totally only in third person cases (cf. Morgan 1970:387).
(22a) *He-l wants him^ to win. (no coreference possible)
(22b) (?)l- want me to win. (coreference possible in contrastive
contexts)
(22c) (?)Youj want you^ to win.
(Technically, Avoid Pronoun should be activated in (22c), since one can
be speaking to different addressees within a single sentence, distinguish-
ing them by pointing or eye gaze. This situation, however, is too mar-
ginal to trigger the ambiguity avoidance condition on Avoid Pronoun, and
the second person case thus works like the first person case in (22b).)
The literature on reflexives offers further grist for a suitably up-
graded Avoid Pronoun mill. As is well known, it is usually (although not
always) the case that a nonreflexive pronoun can be interpreted as
coreferential with a given antecedent just in case a reflexive (bound by
that antecedent) could not have appeared in that position. Many in-
stances of this pattern respond to the Disjoint Reference principle
suggested in Postal (1974) and discussed further within the EST (cf.
Chomsky 1982). Thus, compare (23a) and (23b).
(23a) Hei likes j himself-/*him:J.
(23b) He- said that she tikes |*himself|/himj.
Here again, as with the 'Equi' cases, ambiguity avoidance is relevant in
the determination of when coreference is possible.
(24a) He's voting for him. (no coreference possible)
(24b) I'm voting for me. (OK contrastively)
In 'pro-drop' languages, PRO may appear in place of overt NPs freely
in a wide range of positions, including subject and object; generally, verb
agreement and/or pragmatic context permits recoverability of the
missing referent. As expected, the scope of the Avoid Pronoun principle
is commensurately widened in these languages. In her analysis of one
pro-drop language, Turkish, Enc (1982) shows that semantically
redundant pronouns (i.e. those whose referent would be recoverable) tend

to be interpreted as expressing contrast—including topic-change, which
she argues is a subtype of contrast. (Notice that in English, too, the
occurrence of an overt nonreflexive pronoun in a frame which permits a
referentially equivalent gap or reflexive, e.g. (22b) or (24b), is likewise
interpreted contrastively.) Enc argues that the interpretation of such
marked pronouns, whose appearance seems to violate parsimony, is based
on a conventionalization of a quantity-based inference—but she is re-
ferring here to Grice's Quantity2 which we have previously subsumed
under our R Principle (i.e. least effort).
In fact, while Avoid Pronoun, or whatever more general principle
ultimately includes it, is basically a least effort correlate (as Chomsky
observes), the division of labor we ultimately arrive at (in which abstract
pronouns, i.e. PROs, are interpreted one way and real pronouns another)
requires reference to both R and Q Principles, in the manner outlined in
section 8.
A related instantiation of this division of pragmatic labor makes this
clear. Chomsky proposes a 'general discourse principle' for R(eferring)-
expressions, i.e. names and descriptions (1982:227).
(25a) 'Avoid repetition 1of R(eferring)-expressions, except when
conditions warrant.
(25b) 'When conditions warrant, repeat.'
Note that this discourse principle comes in two parts, one (25a) R-based
and one (25b) Q-based; furthermore, as with Grice's two submaxims of
quantity, each of Chomsky's subprinciples is explicitly bounded in its
applicability by the other (as the subordinate clause material indicates).
The choice among referentially equivalent referring expressions for
picking out a given individual is also subject to R- and Q-based consider-
ations, as Prince shows in working out her Scale of Assumed Familiarity
(1981:245). The matter is rather complex, however, and I shall not
pursue it here.
One last example of the division of pragmatic labor emanating from
the domain of reference is presented by Levinson (1983:75):
The deictic words yesterday, today, and tomorrow pre-empt the
calendric or absolute ways of referring to the relevant days. Thus
the following, said on Thursday, can only be referring to next
Thursday (or perhaps some more remote Thursday), otherwise the
speaker should have said today:
I'll see you on Thursday.
The same holds if it is said on Wednesday, due to the pre-emptive
tomorrow.
10. The division of pragmatic labor and the lexicon. Aronoff has
shown (1976:43ff.) that the existence of a simple lexical item can block
the formation of an otherwise expected affixally derived form synony-
mous with it. In particular, the existence of a simple abstract nominal
underlying a given -ous adjective blocks or prevents the formation of an
-ity nominalization based on that adjective.
(26a) fury furious *furiosity

*cury curious curiosity
(26b) fallacy fallacious *fallacity
*tenacy tenacious tenacity
Aronoff's blocking phenomenon is the limiting case of a more general

pattern independently observed and discussed by McCawley (1978),
Kiparsky (1982), and indeed Breal (1900); a pattern which directly re-
flects the division of pragmatic labor sketched in section 8 and exempli-
fied in section 9: unmarked forms tend to be used for unmarked situa-
tions (via R) and marked forms for marked situations (via Q).
Kiparsky (1982) begins by noting that Aronoff's blocking paradigm is
both too strong and too weak. Contrary to Aronoff's predictions, pro-
ductive derivational processes are sometimes but not always blocked by
the existence of a less productive corresponding form: decency and
aberrancy block *decentness and *aberrantness, but gloriousness and
furiousness survive alongside glory and fury. Blocking may also extend
to inflectional processes as well, although again inconsistently: *mans/
men, *goed/went, but kneeled/knelt, dreamed/dreamt. Aronoff's formu-
lation of blocking, limited as it is to less than fully productive deriva-
tional processes, has nothing to say about these obviously related cases.
Kiparsky suggests a reformulation of Aronoff's blocking as a subcase
of the (Panini-Anderson-Kiparsky) Elsewhere Condition ('Special rules
block general rules in their shared domain1—specifically, in morphology,
irregular forms preclude regular forms). Kiparsky notes that 'blocking
can be partial in that the special Qess productive] affix occurs in some
restricted meaning and the general [more productive] affix picks up the
remaining meaning1. (These two components of the blocking process
correspond to the second and third diagrams in our representation of the
division of pragmatic labor, (18).) He cites as examples of partial block-
ing refrigerant/refrigerator, informant/informer, contestant/contester;
full blocking results when there is no meaning 'left over' for the more
productive form to pick up (*borer/bore^, *inhabiter/inhabitant).
To handle these and other cases, Kiparsky formulates a general condi-
tion which he calls Avoid Synonymy.
(27) Avoid Synonymy: 'The output of a lexical rule may not be
synonymous with an existing lexical item'.
This principle applies to both derivation and inflection, if we assume a

level-ordered morphology. Its transderivational nature allows blocking
between morphologically unrelated stems: thief blocks *stealer (but cf.
base-stealer, with a noncompositional meaning), cutter 4 knives, scis-
sors, etc.
While there is something right about this principle, it is still too
strong, as Kiparsky concedes. For one thing, we need to define a notion
of 'corresponding item' relativized to a given speech level or register;
words like fridge (underived), icebox (derived by compounding), and
refrigerator (derived by affixation) can coexist within a single idiolect
despite their referential equivalence. Indeed, the principle is almost
self-falsifying: the doublets synonymy and synonymity strike me as

perfect synonyms!
Working independently of the Aronoff-Kiparsky line, McCawley (1978)
collects a number of examples where the appropriate use of a given
expression formed by a relatively productive process (including syntactic
formations) is restricted by the existence of a more 'lexicalized' alterna-
tive to this expression (i.e. one formed by a relatively nonproductive
process). One case in point is originally due to Householder: the collo-
cation pale red is (in the language of Aronoff and Kiparsky) fully or
partially blocked by the lexical alternative pink. For some speakers pale
red is simply anomalous (or1 at least nonoccurring); for others, it picks up
whatever part of the 'pale domain of 'red' pink has not preempted. In
either case, pale red is limited in a way that pale blue and pale green are
not.
In the same way, McCawley observes, the distribution of productive
causatives (in English, Japanese, and other languages) is restricted by
the existence of a corresponding lexical causative. Lexical causatives
(e.g. (28a)) tend to be restricted in their distribution to the stereotypic
causative situation: direct, unmediated causation through physical
action.
(28a) Black Bart killed the sheriff.
(28b) Black Bart caused the sheriff to die.
This restriction can be viewed as a straightforward R-based conversa-
tional implicatum—an inference 'to the best interpretation', in the
language of Atlas and Levinson (1981). The use of the relatively marked,
morphologically more complex periphrastic causative (e.g. (28b)) will
then Q-implicate that the unmarked situation does not obtain. Thus,
(28b) suggests that (28a) could not have been used appropriately, possibly
because Bart caused the sheriff's gun to backfire by stuffing it with
cotton, or arranged for scorpions to be placed in the room of the sheriff
(who is known to have a weak heart), etc. Similarly, the use of the
unmarked lexical causative in (29a) R-implicates that the action was
brought about in an unmarked way (presumably by stepping on the brake
pedal), while the choice of the periphrastic (29b) correspondingly Q-
implicates that some unusual method was employed (pulling the emer-
gency brake, telekinesis, etc.).
(29a) Lee stopped the car.
(29b) Lee got the car to stop. (Lee made the car stop.)
McCawley's account of the division of labor between lexical and
periphrastic causatives, like Chomsky's Avoid Pronoun and Kiparsky's
Avoid Synonymy (or indeed, Grice's Avoid Prolixity maxim, which they
all seem to reflect), is explicitly transderivational, and thus runs into
some of the same problems of overgeneralization. Consider the data in
(30a-d), which McCawley borrows from Heringer (1976:207).
(30a) John made the plate move, (indirect)

(30b) John moved the plate, (direct)
(30c) John made the plate fall, (indirect or direct)
(30d) *John felled the plate.
Here, the periphrastic make versions, 'normally used only in a situation

involving indirect causation1, as in (30a) whose use is restricted (by the
existence of (30b)) to situations involving some sort of supernatural
power on John's part, may generalize (as in (30c)) to a wider domain of
causative situations 'just in those cases where no lexical causative exists
to express direct causation' (Heringer 1976:207) (cf. (30d)). But, as
Heringer and McCawley fail to notice, there is a simple lexical form
corresponding to (30c), (30d), viz.
(30e) John dropped the plate.
If drop does not count as a corresponding form for make...fall because it

involves an unrelated stem, then we cannot legitimately invoke (28a) to
predict the limited distribution of (28b).
Other problems for the analysis of McCawley (1978) are discussed in
Horn (1978a), where it is argued that the general 'least effort' principle
employed by McCawley is simply too powerful and, more specifically,
that we need to develop a notion of 'corresponding item' which will take
into account such variables as morphological relatedness, markedness,
speech register, and inherent complexity. For example, lexical causative
verbs may indeed be unmarked in English with respect to their peri-
phrastic make or cause counterparts, but this is not the case for lexical
causative adjectives. Thus, the distribution of the periphrastic forms in
(31a-c) is not constrained by the existence of the corresponding 'simple'
forms in (32a-c).
(31a) That sort of behavior really makes me angry.

(31b) I didn't know that teasing your dog would get you so upset.
(31c) Wild horses couldn't make me stay away.
(32a) That sort of behavior really angers me.
(32b) I didn't know that teasing your dog would upset you so
(much).
(32c) Wild horses couldn't keep me away.
Similarly, as a variety of linguistic and psycholinguistic evidence demon-
strates, incorporated negation remains relatively complex, and so the
distribution of (33a-b) is not constrained by the existence of the super-
ficially simpler examples of (34a-b).
(33a) I persuaded Bill not to date many girls.
(33b) It's not likely that your coin will land heads.
(34a) I dissuaded Bill from dating many girls.
(34b) It's unlikely that your coin wil land heads.
In particular, (33a) is neutral with respect to the (conventionalized) R-

implicatum associated with the use of (34a), viz. that Bill had previously
intended to date many girls, and (33b) is unspecified for the implicatum
associated with (34b), i.e. that it is likely your coin will not land heads
(the so-called neg-raised interpretation). No Q-implicata are triggered
in (31a-c) or (33a-b), and hence the working out of the division of prag-
matic labor is not complete. (Cf. Horn 1978a for further examples and
related discussion.)
While we must therefore refine and/or reevaluate the transderivation-
al mechanisms invoked by Aronoff, McCawley, and Kiparsky for describ-
ing the division of labor between the use (or existence) of a simple
lexical form and that of its more complex (lexical or phrasal) counter-
part—in particular, by developing the tools to predict just when a given
form counts as a counterpart to some other given form—the insight
behind their various accounts is real, and it is essentially a single in-
sight: the unmarked form is used for a stereotypical, unmarked situation
(via 1R-implicature) and the marked counterpart for the situations 'left
over (via Q-implicature). This is the division of pragmatic labor out-
lined in section 8.
11. Division of pragmatic labor: Additional cases. Two more in-
stances of the division of labor are worth mentioning here. First, con-
sider the realm of indirect speech acts (cf. Searle 1975, Gordon and
Lakoff 1975, and other articles in Cole and Morgan 1975). We find that
modal auxiliaries which can be associated with indirect speech acts
(iSAs) tend to become conventionally associated with the ISAs they
convey. If (following Searle) we derive the ISA by an exploitation of the
maxim of Relation (cf. section 3), we must nevertheless account for the
fact that (35a-b) are conventionally used to convey the request in (36),
while (35c-e)—which may (very indirectly) convey that request—are not
conventionally used to do so.
(35a) Can you (please) close the window?
(35b) Could you (please) close the window?
(35c) Are you able to (?please) close the window?
(35d) • Do you have the ability to (*please) close the window?
(35e) It's (**please) cold in here.
(36) (Please) close the window.
Searle observes (1975:76) that 'there can be conventions of usage that
are not meaning conventions' by which 'certain forms will tend to be-
come conventionally established as the standard idiomatic forms for
indirect speech acts', including the can you/could you forms of (35a-b)
for indirect requests. Morgan (1978) takes these usage1 conventions to
represent 'short-circuited conversational implicatures which, though
calculable, are no longer (after short-circuiting) actually calculated in
normal conversation. It is this short-circuiting of the R-based impli-
catum which licenses the preverbal please in (35a-b); cf. (35c-e) where
no short-circuiting has taken place! (A parallel account of the neg-
raising phenomenon and the triggering of negative polarity items is
offered in Horn and Bayer to appear.)
But if the modal auxiliaries, idiomatic and semantically versatile as

they are, become associated through a convention of usage or short-
circuited implicature with the ISAs they may be used to convey, their
periphrastic counterparts tend (predictably) to be interpreted literally.
Thus, compare the behavior of the modals in examples (37a), (38a), (39a),
and (40a) (where the implicatum to the ISA is short-circuited) with that
of their periphrastic equivalents in (37b), (38b), (39b), and (40b) (where
the indirect reading is not a conventional use of the expression and may
be totally unavailable).
(37a) Can you pass the salt? (request)
(37b) Are you able to pass the salt? (literal question)
(38a) Here, I can help you with that, (offer)
(38b) (?Here,) I am jable/allowedj to help you with that, (not an
offer)
(39a) Will you join us? (invitation)

(39b) Are you going to join us? (literal question)
(40a) I will marry you. (promise)
(40b) I am jgoing to/willing}to marry you. (only very indirect
promise)
Noting the tendency for the most colloquial expressions (typically, as
we have seen, the modal auxiliaries) to serve as the conventional forms
for conveying indirect speech acts, Searle invokes a new neo-Gricean
maxim, 'Speak idiomatically unless there is some reason not to1. He
observes (in language mutually intelligible with that spoken by Kiparsky
1982 and McCawley 1978),
In general, if one speaks unidiomatically, hearers assume that there

must be a special reason for it, and in consequence,...the normal
conversational assumptions on which the possibility of indirect
speech acts rests are in large part suspended. (Searle 1975:76-77)
The pattern discerned here by Searle responds not only to our division
of pragmatic labor, but also to the set of statistical correlations at the
core of Zipf's analysis of the Principle of Least Effort and its linguistic
reflexes (Zipf 1949). Zipf's Law of Abbreviation posits an inverse rela-
tion between the length of a word and the frequency of its tokens in an
arbitrary text. His Principle of Economic Versatility stipulates a direct
correlation between a word's frequency and its semantic versatility (i.e.
the number of discrete senses or meanings it allows). The Principle of
Economic Specialization states that the age of lexical item in the lan-
guage correlates inversely with its size and directly with its frequency.
By these measures, the relative simplicity of (35a), (35b) as against (35c)
is directly confirmed: can and could are historically older than their
periphrastic counterparts, phonologically simpler and shorter, more
frequent in text tokens, and certainly more versatile semantically, as
the contrasts in (41a-c) make clear.
(41a) I jcan/am able toj stand on my nose.
This knife |can/?is able toj cut the salami, (cf. ...is capable
of cutting it)
The salami jean be cut/*is able to be cut/is capable of being
cut| by this knife.
(41b) Can it really be raining out?/Is it possible that it's raining
out?
(41c) Can Billy come out and play?/Is Billy permitted to come out
and play?
It is significant that modals, the constituents par excellence of least
effort, figure so prominently among those forms conventionalized in
English and other languages as conveyors of indirect speech acts, while
their more expensive and complex counterparts do not.
Finally, we come to the case of logical double negation. What we find
here is that the two negatives of the form not- (not-p) do not cancel out
functionally even when they do semantically: they convey a positive
which is characteristically weaker than the corresponding simple affir-
mative. In Jespersen's words,
The two negatives Qn not common, not infrequent] do not exactly
cancel one another so that the result is identical with the simple
common, frequent; the longer expression is always weaker; 'this is
not unknown to me' or 'I am not ignorant of this' means 'I am to
some extent aware of it', etc. The psychological reason for this is
that the detour through the two mutually destructive negatives
weakens the mental energy of the listener and implies...a hesitation
which is absent from the blunt, outspoken common or known.
(Jespersen 1924:332; cf. Horn 1978c: section 3.1 for discussion)
Rather than appealing, with 3espersen, to the metaphysical (and some-
what Victorian) notion of double negation sapping the listener's mental
energy, we can more plausibly ascribe the weakening effect to the same
general tendency we have already observed: the use of a marked expres-
sion when there is a shorter and less 'effortful' alternative available
signals that the speaker felt s/he was not in a position to employ the
simpler version felicitously.
With double negation, as with indirect speech acts, we see an especial-
ly clear correlation between the stylistic naturalness of a given form, its
relative brevity and simplicity, and its use in stereotypic situations (via
R-implicature). The corresponding periphrastic forms, stylistically less
natural, longer and more complex, are restricted (via Q-implicature) to
those situations outside the stereotype, for which the unmarked expres-
sion would have been inappropriate.
12. Q-based and R-based processes in language change. Perhaps the
clearest lexical correlates of the clash and resolution of the R vs. Q
conflict are in the area of diachrony. The most obvious R-based effects
in language change are the reflexes in 'langue' of the well-known 'fast
speech phenomena' in 'parole': contraction, clipping or truncation,
assimilative shifts, and so on. Clipping (truncation) is a direct corollary
of the aforementioned Law of Abbreviation: the more frequently a word

or expression is used (within a given speech community), the shorter it
will tend to become.
Among the contemporary examples of clipping cited by Stern
(1931:258) in his standard work on lexical change are pram (from per-
ambulator), specs (< spectacles), rep Urepetition or reprobate); it is
striking that the same clipped forms persist even when their convention-
al value has shifted—specs is now likely to abbreviate specifications, and
rep would probably be taken to denote representative (or perhaps reputa-
tion, as in the somewhat dated bad rep). More clearly entered in our
current lexicon are such conventional truncations as TV (or Brit, telly),
phone, bus (ômnibus), and the conventionalized output of an extremely
productive process of acronym formation (e.g. US(A), EST (with at least
two senses for east-coast syntacticians) , UCLA, OSU--'Oregon State
University', 'Oklahoma State University1, or 'Ohio State University',
depending on the shared assumptions of the members of the speech
community).
In discussing the development of the truncated forms, Stern (1931:257-
5&) warns that 'the demands of the speech functions must set a limit to
the economic tendency1—i.e. the Q Principle constrains the power of the
R Principle. (He also disparages 'the use of pronouns, or of generic
words, to save mental effort' as 'especially characteristic of undeveloped
minds, unintelligent or immature 1 . Avoid Pronoun or else!)
A more complex area of language change, involving the interaction of
R-based and Q-based processes, is that of lexical shifts. Two traditional
categories of lexical change (discussed by Paul 1909 and Breal 1900,
inter alia) worth examining in this light are narrowing (or reduction) of
meaning and broadening (or expansion) of meaning.
Narrowing generally involves an R-based shift from a set denotation
to a subset (or member) of that set, representing the salient or stereo-
typical exemplar of the general category. Examples cited in the stan-
dard works (cf. Breal 1900, Stern 1931) which fit this definition include
Greek alogon (lit. 'speechless one') for 'horse', Latin fenum (orig. 'pro-
duce') for 'hay', and English poison (cognate with potion), liquor (cf.
liquid), undertaker (from 'one who undertakes' to 'mortician'), and corn
(used for whatever grain is the most important cereal crop of a parti-
cular region, e.g. wheat in England, oats in Scotland, maize in Australia
or the New World).
In these instances, the shift has become virtually complete (although
the original, broader extension may persist in marginal uses). Other
cases manifest the intermediate stage of 'autohyponymy' (Horn to
appear b), in which the basic, general sense survives in privative opposi-
tion with a specific sense derived from it. Autohyponyms which have
developed their specific meaning (indicated in parentheses) through R-
based narrowing include the following:
color (for 'hue', i.e. the range of colors excluding blacks,

whites, and grays): in color, color TV
temperature (for 'fever'): The baby has a temperature
number (for 'integer'): Pick a number from 1 to 10
drink (for 'drink alcohol1): I don't drink. Cf. the second exam-
ple of (8d)
smell (for 'stink'): Something smells here
Frau (Ger.), femme (Fr.), mujer (Sp.) (for 'wife' as well as
'woman')
Dismissing a contemporary account of this variety of lexical shift in
the evolution of homo 'man' from a source akin to humus 'earth', the first
century grammarian Quintilian asks rhetorically, 'Are we to believe that
homo comes from humus, because man is born of the earth, as if all
animals had not the same origin?' (cited by Breal 1900:114). Yet the
derivation is correct, as noted by Breal, who points out that alogon can
likewise designate 'horse' in modern Greek without implying that no
other animal lacks the faculty of speech. Like those logical fallacies
based on real pragmatic inference patterns (e.g. 'denial of the anteced-
ent', 'post hoc ergo propter hoc', and 'secundum quid1 (Horn 1973:212-13),
all responding to the workings of the R Principle as exemplified in (8a)),
the R-based inference from a set to a salient or stereotypical member is
as linguistically plausible as it is logically invalid.
In other cases, lexical narrowing is Q-based, typically resulting is
autohyponymy. Kempson (1980:15-16) offers a characterization of the
general process:
If for some general term, representing a lexical field, there is a gap
in the sub-parts of that field, with only one more narrowly specified
lexical item, then the gap may be filled by a more specific use of
the general term.
Thus, for example, the existence of bitch 'female dog', in the absence of
a sex-specific mate, leads the general term dog to develop a hyponymic
sense designating the male of the species.
The governing principle for such cases is given as follows (Kempson
1980:15):
If a lexical item Lj has as its extension a set S| which includes the
set S2 which a second lexical item \L^ has as its extension..., then
the lexical item Lj may be used to denote that subset of Sj which
excludes S 2 .
We thus have a partial reconstruction of the division of pragmatic labor,
in that the existence of a more informative, marked term (L2), together
with the choice by a (fully informed) speaker to employ a less informa-
tive, unmarked term (Lj) (in a context where the additional information
would have been relevant), licenses the Q-inference that the speaker was
not in a position to employ the more informative term. We can repre-
sent this state of affairs schematically as in ( )
3* / Laurence R. Horn
(Lj may be used to

denote Sj-S2, i«e« t n e
complement of Sj with
respect to S2)
Thus, we obtain a pragmatic resolution of the semantic asymmetry

inherent in the original state diagram. At the same time, Li may (and
usually does) retain its general application (so as to include £2) in con-
texts where there is no contrast at issue with L2: it is only when the sex
of the beast is relevant that bitches do not count as dogs.
Noncanine examples of autohyponymy (i.e. private polysemy) deriving
from Q-based narrowing include the following (cf. Kempson 1980, Horn
to appear b):
cow (excluding bulls)
rectangle (excluding squares)
finger (excluding thumbs)
gay (excluding lesbians)
player (excluding pitchers, in baseball)
I have argued elsewhere (Horn to appear b) that these cases are not
homogeneous in nature, differing in degree of conventionalization of the
relevant Q-based inference, and that (contra Kempson) they do not in
any case represent the sole source of autohyponymy, given the examples
of R-based narrowing cited earlier and those illustrating R-based broad-
ening to be discussed further on.
The results of R-based and Q-based narrowing may be synchronically
indistinguishable. As against the dog/bitch (and lion/lioness) variety, the
narrowing of man seems to have preceded the development of its coun-
terpart woman, and thus to represent an R-based rather than Q-based
shift, males presumably being reckoned as the salient members of the
species. More recent instances of what feminists have appropriately
dubbed the masculine usurpation of the generic include mankind, chair-
man, and poet. Here, it is clearly the prior (R-based) specialization of
the general term which created the perceived need for—and conscious
innovation of—the corresponding feminine form; it is not the existence
of sex-specific womankind, chairwoman, or poetess which led to a (Q-
based) restriction on the extension of the general terms.
The specific L2 term which triggers a Q-based restriction on the
meaning or use of the general Lj term must be sufficiently natural and
stylistically unmarked, or (as we have observed in our earlier survey of
Q-based restrictions) it will not count as a 'corresponding1 item. Thus,
Blackburn (1983:495) observes that animal may (or may not) be used so
as to exclude humans, and it may likewise be taken in the appropriate
context to exclude birds and/or fish, but there is no use of animal which
excludes mammals—despite the obvious fact that mammalia constitute
just as valid a subgrouping of animals as do birds and fish. Crucially,
however, mammal (unlike man, bird, fish) does not correspond to a basic
level category in the sense of Rosch (1977) and so cannot trigger the
division of pragmatic labor illustrated in (43).
We have seen that narrowing of a lexical item may be either R-based
(the spontaneous delimitation of a general term to a sense representing a
salient exemplar of the category denoted by that term) or Q-based (the
motivated specialization of a general term triggered by the prior exist-
ence of a hyponym of that term). The converse process—lexical broad-
ening or expansion—is always R-based: the generalization of a term for
a species to cover the encompassing genus, from genus to phylum, from
subset to superset. Thus, Latin pecunia, originally denoting 'property or
wealth in cattle 1 (cf. pecu 'livestock, cattle 1 ), generalizes to signify
'wealth' and eventually 'money', a shift paralleled in the English cognate
fee (<OE feoh 'cattle'-^'property'). As noted by Breal (1900) and Ullman
(1959), broadening is often accompanied by 'semantic impoverishment'
resulting from the attrition of a qualifying context, as in the expansion
of (assumed) Late Latin adripare, arripare, 'come to shore' into French
arriver 'arrive (tout court)', or the generalization of panarium 'bread
basket' into panier 'basket' (Ullman 1959:209).
Broadening tends to apply regularly with place and origin names, as
political entities grow and mutate; examples include the expansion of
Lat. romanus (or Eng. New Yorker) to designate someone or something
from the empire (state) at large, rather than specifically from its major
city. An even more productive source of lexical broadening involves
trade names which have lost their capital letters and become generics
(cf. Mason and Pimm 1982, Horn to appear b), including those in (45).
(45) xerox jello good humor

kleenex thermos toll-house cookies
scotch tape vaseline hoover (Br., 'vacuum cleaner')
As in many of the instances of narrowing discussed earlier, the net

result (or at least temporary equilibrium state) here is autohyponymy:
the broadened term retains its original specific meaning in at least some
contexts. We may even end up with multiple hyponymy, as in the case of
the enthogeographic label Yankee, with its three semantically nested
extensions cited by McCawley (1981:9-10) and standard lexicographers,
'native or inhabitant of New England; or, more generally, of northern
U.S.; or, more generally, of U.S.1
In these examples, broadening results when a specific term represent-
ing a salient exemplar (often the salient exemplar) of a wider class
generalizes to denote that wider class; lexical expansion thus constitutes
a perfect mirror image of R-based narrowing from a set to a salient
exemplar of it. Once the new value for the term is entered in the lexi-
con—alongside or in place of the original specific value—semantic shifts
(often culturally triggered) may ensue, obscuring the original set/salient
member relation. It was the Romans' use of livestock as the medium of
exchange which led to the broadening of pecunia, a derivative of the
term for 'livestock', into a term denoting 'property'—and it was the
subsequent abandonment of barter in favor of a monetary system that
led to the later loss of the etymological component altogether, to the
point where pecunia simply stood for 'wealth'. So too, in Mayan
languages studied by Lounsbury, a word referring to 'serpent' or 'jaguar'—

animals with a particular ritual significance—became generalized as the
standard term for 'animal', retaining its original specific meaning in
compounds. The manufacturers of the items in (45) should therefore find
a silver lining in the legal cloud of their copyright loss: their products
obviously represent the prototypes or epitomes of their respective kind,
even if this results in their brand name coming to stand for that kind.
Broadening and narrowing often operate in tandem within a given
language, or in complementary fashion across related languages. Thus,
German Tier has broadened from 'wild animal' to include domestic
livestock and pets as well as man (especially in compounds like Tierwelt
'animal kingdom'). At the same time, its English cognate has turned
autohyponymous (and eventually unambiguous once more) through nar-
rowing. OE deor, ME deer originally designated beasts in general, espe-
cially 'objects of chase' (Stern 1931:416), then became restricted (initial-
ly in the lexicon of hunters) to single out the object of chase par excel-
lence, fam. Cervidae. By the early modern period, the general use of
deer had been largely supplanted by the Romance loans beast, brute, and
animal, although it continued to retain a marginal application to the
class of quadrupedal mammals, undifferentiated for species (as in Shake-
speare's reference to rats and mice and such small deer). Eventually,
only the specialized hyponymic sense survived, spreading from hunters'
use into the general speech community.
The adjustment in the extension of deer vis-a-vis animal reflects a
general tendency in language insightfully described by Bre*al (1900:
27ff.). Bre'al's principle, in effect a diachronic precursor of Kiparsky's
Avoid Synonymy, is the Law of Differentiation, governing
the intentional C0> ordered process by which words, apparently

synonymous, and once synonyms, have nevertheless taken different
meanings, and can no longer be used indiscriminately...either they
are differentiated, or else one of the two terms ceases to exist.
(Breal 1900:27-28)
Typically, an older word for a given referent is retained, but limited to a

specialized (often humble, 'degraded', or 'trivial') domain. Once again,
the marked form is limited to a marked use.
Thus, Bre'al informs us, the Swiss patois word for 'room', palie", is
restricted to the meaning 'garret' after standard French chambre is
adopted as the unmarked term. The general use of Oscan popina 'kitch-
en', displaced by its Latin cognate coquina, eventually comes to denote
'tavern'. And now 'the Savoyard uses the names of pere and mere for his
parents, while he keeps for his cattle the old words pare" and metre1' (Breal
1900:29).
Among English examples of Breal's Law of Differentiation we might
reckon brethren (whose restricted use is motivated by the adoption of
the standard unmarked plural brothers), the deer/animal case already
discussed, and the related and notorious hound/dog affair.
Once again, we find broadening and narrowing operating hand-in-hand
(or paw-in-paw) until the eventual division of labor is arrived at. OE
dogca, referring to a particular breed of dog (it is not entirely clear
which one), represented a hyponym of the general term hound, then

denoting the entire kind 'dog' (as its German littermate Hund continues
to do). Sometime around the fourteenth century, when Chaucer's warn-
ing 'It is nought good a slepyng hound to wake1 was turning into Hey-
wood's 'It is evyll wakyng of a sleepyng dog', dog and hound were pre-
sumably both autohyponyms, with different specific understandings.
Eventually, hound was totally displaces by dog in its general application
but, in accordance with the Law of Differentiation, continued to retain
its specialized use (originally developed via R-based narrowing in the
vocabulary of hunters, for whom hounds were the salient representatives
of the species, dogs par excellence).
Thus, narrowing and broadening, separately and in conjunction, reflect
the centrality of R-based and Q-based shifts in the development of the
lexicon. In R-based and Q-based narrowing, in R-based broadening, and
in instantiations of Breal's Law of Differentiation, as in the dynamics of
use and meaning described by Chomsky (19&2), Aronoff (1976), Kiparsky
(1982), McCawley (1978), and Searle (1975), we descry the recurring
patterns of our two general pragmatic principles and of the division of
labor resulting from their interaction.
13. Varia and concluding remarks. Other apparent reflexes of the R

vs. Q dynamic cannot be discussed here in detail, but are worth mention-
ing as possible topics for further investigation. First, there is the class
of rhetorical figures of speech, including synecdoche, metonymy, and
litotes, representing the 'parole'-based equivalent of conventionalized
narrowing (part for whole) and broadening (whole for part).
Second, there is the privative relation between the meanings (and
uses) associated with nominative and ergative case marking for subjects
of intransitive verbs in 'fluid' or 'active' systems (cf. Dixon 1979 and
especially Holisky 1983). In Bats (a.k.a. Batsbiy and Tsova-Tush), a
Caucasian language investigated by Holisky, the facts appear to be as
follows:
With intransitive verbs which are typically agentive and
intentional (e.g. 'get dressed', 'wash', 'hide1, 'run', 'bump into'),
NOM(inative) case is semantically marked for nonagentivity
and ERG(ative) is unmarked.
With intransitives which are typically nonagentive (e.g. 'die',
'burn', 'become poor', 'forget', 'drown1, 'go crazy1), ERG is
marked as agentive and NOM is neutral.
(48) With intransitives which allow both agentive and nonagentive
interpretations equally freely (e.g. 'fall asleep', 'fall down',
'lose weight', 'be late'), ERG and NOM are equally natural and
no markedness relation obtains.
In the first two sets of examples, we are dealing with a privative opposi-
tion (Zwicky and Sadock 1975, Horn to appear b), in which 'the marked
member conveys its meaning truth-functionally, while the neutral mem-
ber does so by implicature'(Holisky 1983:5). The existence of the more
informative, marked form, together with the speaker's choice of the
unmarked, semantically neutral form, allows the addressee to construct
a Q-based implicature: the inference from the use of ERG with the
verbs of (46) that an agent was involved, and the corresponding inference
from the use of NOM with the verbs in (47) that the nonagentive inter-
pretation was intended. Once again, we arrive at a division of pragmatic
labor, in which the marked form is used for the marked situation (rela-
tivized to the semantics of the verb in question), and the unmarked form
for the unmarked situation.
Finally, one more possible locus of the dynamics and resolution of the
R/Q conflict is the range of 'switch-reference' constructions (cf. e.g.
Finer 1984). As I read the data, there seem to be some languages (in-
cluding Seri and Washo) in which the presence of a DS (different-subject)
marker indicates that an embedded clause has a different subject from
the main clause, while the lack of a marker is semantically unspecified
for same vs. different referent, but tends to be interpreted as indicating
that the subject is the same in contexts where the distinction is relevant
and no further disambiguating factors are available. The asymmetry
involved here is apparently analogous to that just touched on in Bats
case-marking (as well as other examples discussed earlier), and thus
similarly reflective of the use of Q-based implicature to complete the
division of labor, but further investigation is required to sharpen the
account of switch-reference and situate it more clearly within the
proposed framework.
We have surveyed (all too cursorily) a wide range of linguistic pheno-
mena, both synchronic and diachronic, both lexical and syntactic, both
'parole'-based and 'langue'-based, from conversation implicature and
politeness strategies to the interpretation of pronouns and gaps, from
blocking and distributional constraints on lexical items to indirect
speech acts, from lexical change to case marking. If I am right, these
apparently diverse and unrelated domains are all motivated and governed
by the same functional dynamic, the ongoing Zipfo-Gricean dialectic
between the Q-based Sufficiency Principle and the R-based Principle of
Least Effort.
Notes
1. Ducrot's model of pragmatic inference shares with Grice's the
crucial feature of indeterminacy. Given an utterance like (i),
(i) La situation n'est pas excellente. 'The situation isn't excel-

lent1
Ducrot notes (1972:132), an addressee may infer that the speaker in-
tended to convey (ii),
(ii) Elle est franchement mauvaise. 'It's pretty bad1
through the exploitation of the R-based 'Loi d'informativite'' (the as-

sumption that the hearer does not already know the information the
speaker is conveying); since he has not said (ii), however, the speaker can
always retreat to the literal meaning of what he has said, i.e. the weaker
(i). On the other hand, given the Q-based 'Loi d'exhaustivite1' (the prin-
ciple which demands that the speaker provide the strongest possible
information which he possesses and which he believes may interest the
hearer; cf. Ducrot 1972:134), someone who utters (0 may, in the appro-
priate context, implicate that the situation is pretty good. This inde-
terminacy is perhaps more apparent when the negation applies to a
semantically negative predication, as pointed out by Stern (1931:312):
Not bad, taken literally, leaves a large latitude, from indifferent to
excellent, and may mean GsicJ either, depending on the intonation
used and the circumstances.
2. Following Lyons (1977:9.4), A is called a 'hyponym' of B iff the
extension of A is properly included in that of B: Labrador retriever is a
hyponym of retriever, retriever of dog, dog of mammal, and so on. But
some words are hyponyms of themselves: dog and bitch are (sex-differ-
entiated) co-hyponyms of dog, lion and lioness of lion, etc. In these
cases, we can call the unmarked term (dog or lion) an 'autohyponym1.
Autohyponymy thus represents privative polysemy or ambiguity within a
single lexical item (cf. Zwicky and Sadock 1975 on privative opposition).
3. Breal (1900:108) suggests that in fact alogon [alo7o] came to stand
for the horse simply because 'the rider, speaking of his mount, was
accustomed to say "the animal"1; similarly, homines were so-called not
merely because of man's preeminent position among the creatures of
earth, but because of the intended opposition between the earthbound
human race as against "the inhabitants of the sky Dii or Superii' (Breal
1900:114).
4. As noted in Horn (to appear b), there is an even more narrowly
defined sense of Yankee, the sense in which the Kennedys are disquali-
fied from true Yankee status by their Irish Catholic heritage. On this
ultrarestrictive interpretation, a Yankee is someone from New England
who approximates to a sufficient degree the prototype WASP of the
Pepperidge Farm commercials. (We may need to invoke a Rosch (1977)-
style prototype theory in any case to explain why a Vermont farmer or a
Maine lobsterman is more of a Yankee than is a Greenwich stockbroker.)
5. The mirror image relation between R-based broadening (from
salient subset to superset) and R-based narrowing (from superset to
salient subset) is highlighted when we juxtapose the development of
pecunia with the opposite shift exemplified by ktimata in Greek, from
the general 'possessions' to the specific 'cattle' (Breal 1900:109).
References
Aronoff, M. 1976. Word formation in generative grammar. Cambridge,
Mass.: MIT Press.
Atlas, J., and S. Levinson. 1981. It-clefts, informativeness, and logical
form. In: Cole, ed. (1981:1-61).
Blackburn, W. 1973. Ambiguity and non-specificity: A reply to 3ay
David Atlas. Linguistics and Philosophy 6.479-98.
Breal, M. 1900. Semantics. Trans. Mrs. H. Cust. New York: Henry

Holt.
Brown, P., and S. Levinson. 1979. Universals in language use: Polite-
ness phenomena. In: Questions and politeness: Strategies in social
interaction. Edited by E. Goody. Cambridge: Cambridge University
Press. 56-311.
Chomsky, N. 1982. Lectures on government and binding, 2nd ed. Dor-
drecht: Foris.
Clark, H., and S. Haviland. 1977. Comprehension and the given-new
contract. In: Discourse production and comprehension. Edited by R.
Freedle. Hillside, N.J.: Erlbaum. 1-40.
Cole, P., ed. 1978. Syntax and semantics 9: Pragmatics. New York:
Academic Press.
Cole, P., ed. 1981. Radical pragmatics. New York: Academic Press.
Cole, P., and 3. Morgan, eds. 1975. Syntax and semantics 3: Speech
acts. New York: Academic Press.
Coleman, L., and P. Kay. 1981. Prototype semantics: The English word
lie. Lg. 57.26-44.
DeMorgan, A. 1847. Formal logic. [Reprinted London: Open Court
Co., 1926J
Dixon, R. M. W. 1979. Ergativity. Lg. 55.59-138.
Ducrot, O. 1972. Dire et ne pas dire. Paris: Hermann.
Enc, M. (to appear) Topic switching and pronominal subjects in Turk-
ish. In: Studies in Turkish linguistics. Edited by K. Zimmer.
Finer, D. 1984. The formal grammar of switch reference. University of
Massachusetts dissertation.
Gazdar, G. 1979. Pragmatics. New York: Academic Press.
Geis, M., and A. Zwicky. 1971. On invited inferences. Linguistic
Inquiry 2.561-65.
Gordon, D., and G. Lakoff. 1975. Conversational postulates. In: Cole
and Morgan, eds. (1975:41-58).
Hamilton, Sir W. 1860. Lectures on logic. Edinburgh: Blackwood and
Sons.
Harnish, R. M. 1976. Logical form and implicature. In: An integrated
theory of linguistic ability. Edited by T. Bever, 3. Katz, and D. T.
Langendoen. New York: Crowell. 464-79.
Heringer, 3. 1972. Some grammatical correlates of felicity conditions.
Ohio State University dissertation. Distributed by Indiana University
Linguistics Club.
Heringer, J. 1976. Idioms and lexicalization in English. In: Syntax and
semantics 6: The grammar of causative constructions. Edited by M.
Shibatani. New York: Academic Press. 205-16.
Holisky, D. 1983. The privative nature of agent-marking in Bats (and
English). Paper presented at annual meeting of the Linguistic Society
of America, Minneapolis.
Horn, L. 1972. On the semantic properties of logical operators in
English. University of California at Los Angeles dissertation. Distri-
buted by Indiana University Linguistics Club, 1976.
Horn, L. 1973. Greek Grice: A brief survey of proto-conversational
rules in the history of logic. In: Chicago Linguistic Society 9.205-14.
Toward a new taxonomy for pragmatic inference /
Horn, L. 1978a. Lexical incorporation, implicature, and the least effort

hypothesis. In: Papers from the Parasession on the Lexicon. Chi-
cago: Chicago Linguistic Society. 196-209.
Horn, L. 1978b. Remarks on neg-raising. In: Cole, ed. (1981:129-220).
Horn, L. 1978c. Some aspects of negation. In: Universals of human
language, vol. 4: Syntax. Stanford, Calif.: Stanford University
Press. 127-210.
Horn, L. (to appear a) Metalinguistic negation and pragmatic ambi-
guity. Lg.
Horn, L. (to appear b) Ambiguity, negation, and the London School of
Parsimony. In: Proceedings from New England Linguistic Society
14. Edited by C. Jones and P. Sells.
Horn, L., and S. Bayer, (to appear) Short-circuited implicature: A
negative contribution. Linguistics and Philosophy.
Jespersen, O. 1924. The philosophy of grammar. CReprinted New
York: Norton, 1976J
Karttunen, L. 1970. The logic of English predicate complement con-
structions. Unpublished MS, University of Texas.
Keenan, E. O. 1976. The universality of conversational postulates.
Language in Society 5.67-80.
Kempson, R. 1980. Ambiguity and word meaning. In: Studies in English
linguistics. Edited by S. Greenbaum, G. Leech, and J. Svartvik.
London: Longmans. 7-16.
Kiparsky, P. 1982. Word-formation and the lexicon. In: Proceedings of
the 1982 Mid-America Linguistic Conference, University of Kansas.
Edited by F. Ingemann.
Lakoff, R. 1973. The logic of politeness; or, minding your P's and Q's.
In: Chicago Linguistic Society 9.292-305.
Levinson, S. 1973. Pragmatics. Cambridge: Cambridge University
Press.
Lewis, D. 1969. Convention. Cambridge, Mass.: Harvard University
Press.
Liber man, M., and I. Sag. 1974. Prosodic form and discourse function.
Lyons, J. 1977. Semantics 1. Cambridge: Cambridge University Press.
McCawley, J. 1978. Conversational implicature and the lexicon. In:
Cole, ed. (1978:245-59).
McCawley, J. 1981. Everything that linguists have always wanted to
know about logic. Chicago: University of Chicago Press.
Martinet, A. 1962. A functional view of language. Oxford: Clarendon
Press.
Mason, J., and D. Pimm. 1982. Generic examples: Seeing the general in
the particular. Unpublished MS, The Open University, Milton Keynes,
England.
Mills, J. S. 1867. An examination of Sir William Hamilton's philosophy,
3rd ed. London: Longmans.
Morgan, J. 1970. On the criterion of identity for noun phrase deletion.
Morgan, 3. 1978. Two types of convention in indirect speech acts. In:
Cole, ed. (1973:261-80).
Wl / Laurence R. Horn
Paul, H. 1909. Prinzipien der Sprachgeschichte, 4th ed. Halle: M.
Niemeyer.
Postal, P. 1974. On raising. Cambridge, Mass.: MIT Press.
Prince, E. 1981. Toward a taxonomy of given-new information. In:
Cole, ed. (1981:223-55).
Prince, E. 1982. Grice and universality: A reappraisal. Unpublished
MS, University of Pennsylvania.
Rosch, E. 1977. Human categorization. In: Advances in cross-cultural
psychology. Edited by N. Warren. New York: Academic Press. 1-72.
Schmerling, S. 1975. Asymmetric conjuction and rules of conversa-
tion. In: Cole and Morgan, eds. (1975:211-31).
Searle, 3. 1975. Indirect speech acts. In: Cole and Morgan, eds.
(1975:59-82).
Stern, G. 1931. Meaning and change of meaning. Bloomington: Indiana
University Press.
Tannen, D. 1975. Communication mix and mixup, or how linguistics can
ruin a marriage. San 3ose State Occasional Papers in Linguistics.
Tannen, D. 1979. Ethnicity as conversational style. Sociolinguistics
Working Paper No. 55, Southwest Educational Development Labora-
tory, Austin, Texas.
Ullman, S. 1959. The principles of semantics, 2nd ed. Glasgow: Jack-
son, Son and Co.
Zipf, G. K. 1949. Human behavior and the principle of least effort.
Cambridge: Addison-Wesley.
Zwicky, A., and J. Sadock. 1975. Ambiguity tests and how to fail
them. In: Syntax and semantics 4. Edited by J. Kimball. New York:
Academic Press. 1-36.
INTENSITY
William Labov
University of Pennsylvania
At the heart oi social and emotional expression is the linguistic fea-
ture of intensity. It is a difficult feature to describe precisely. Inten-
sity by its very nature is not precise: first, because it is a gradient
feature, and second, because it is most often dependent on other lin-
guistic structures. Most discussions of intensity involve specialized
prosodic contours and a set of adverbs that code intensity directly:
really, so, and very in English. But intensity is most often expressed
through linguistic forms that are normatively devoted to logical rela-
tions and conceptual categories. The use of such forms to signal inten-
sity can lead to changes in the subsystems involved. If grammatical
descriptions don't take social and emotional expression into account, and
their effect on the underlying system, they will be incomplete and even
misleading for language learners.
1. Intensity in adverbs: Cognitive zeroes. Emotion is often ex-
pressed through peripheral, gradient systems: by prosody, vocal quali-
fier, and gesture. Information on emotional states can be conveyed by
the central grammatical apparatus, completely verbalized in proposi-
tional form, but with lower chances of success. Any imaginable emo-
tional state can be stated as a proposition: 'I am moderately angry with
you', or 'I'm entirely committed to this line of action'. But we are all
familiar with situations where listeners refuse to accept these words at
their face value. On the other hand, De Groot's principle (1949) tells us
that whenever the message conveyed by the intonational system is at
variance with the information contained by the words, the intonational
message will be understood as the one intended. It would follow that the
peripheral systems are the primary means of conveying social and emo-
tional information, and the grammatical mechanism is the primary
means for conveying referential and cognitive information.
Yet, some elements of the grammatical system are specifically de-
voted to emotional expression, and the most common of these are ad-
verbs that signal intensity. 'Intensity' is defined here as the emotional
to I William Labov
expression of social orientation toward the linguistic proposition: the

commitment of the self to the proposition. The speaker relates future
estimates of his or her honesty, intelligence, and dependability to the
truth of the proposition. Intensity operates on a scale centered about
the zero, or unmarked expression, with both positive (aggravated or
intensified) and negative (mitigated or minimized) poles. A feature
notation with £+ intensity] is therefore not appropriate. Instead, I will
refer to position on an ordinal scale where features marked for intensity
raise an expression to a value greater than zero, and those marked for
deintensification lower expressions to values less than zero.
Really is one of the most frequent markers of intensity in colloquial
conversation, and must figure large in any first approach to intensity in
everyday communication. It makes little contribution to cognitive or
representational meaning, unless it is directly opposed to the unreal or
the insincere. In fact, really can be described as a 'cognitive zero1: it
would have zero representational content in context-free information
processing. Social and emotional meanings are difficult to demonstrate
directly, but the presence of cognitive zeroes or cognitive contradictions
can give indirect demonstrations of their presence.
To illustrate the use of really and other adverbs of intensity, I will
draw from one of the best records we have of intimate colloquial speech,
the family conversation of Dolly Ripley. She had just returned to New
York City from a visit to her family in North Carolina, helping to mind
her nieces and nephews, and she was giving the latest news of the family
to her first cousin over the telephone.
(la) I really worked while I was away!
(lb) He [grandfather] really came out of it good.
(lc) I got nothin1 to really stay here for.
(Id) They really had themselves a ball!
Here really acts to intensify the main predication. It can be considered
as a part of the mood system, pertaining to the existential status of the
predication. It establishes a 'surreal' mood, the converse of irrealis, in
designating a state of reality greater than normal.
The adverb sure also occurs in preverbal position; it is in general an
alternate of really, as shown in (2).
(2a) Sure it is!

(2b) 'Cause they sure work the hell out of me!
The adverb so is limited to adjectival modification, where it serves as
an alternant of sure and really.
(3a) Listen honey, they'll change clothes so fast down here!

(3b) I'm so glad she comes.
(3c) But they were so_ glad to see me!
Questions are intensified by the insertion of the hell after the WH-word.
(4a) Where the hell you been?
(4b) So what the hell you been doin1 for the summer?
Intensity / 45
The form used to intensify questions can be finely adjusted to the level
of mitigation or aggravation desired with a wide range of alternants.
The adverb just has the same privileges of occurrence as really, but
has a much wider range of meanings. Just can be a minimizer, as in_I
didn't yell and scream, I just went like this [gesture^. Here it can be
interpreted as 'no more than the following1. When just modifies a strong
action or expression or an unusual situation, it can be interpreted in the
opposite manner: an intensive meaning 'no less than this'. The examples
in (5) are drawn from a narrative of Bobby Andrucci, 23, of Ayr, Scot-
land.
(5a) I jus' you know whacked 'im.

(5b) He went straight through the glass doors an' I just put it into
him! Chit him hardO
(5c) But m' shirt an' all they was just covered with blood.
We encounter in spontaneous speech a wide variety of metaphors that
serve as intensifiers: bleeding like a pig, darker than pitch or pitch
dark. The metaphor in (6) was repeated many times in a description of a
mugging in a conversation among several middle-aged ladies in a park in
Carlisle in northern England:
(6) They said she was like a...oooh, they said she was like a
battered cucumber when they took her into the infirmary!
Closely related to these adverbial intensifiers is the use of the super-

lative form of attributive adjectives, as in the quotation from Dolly
Ripley in example (7).
(7) I told the biggest lie!
This review of adverbs of intensity is not intended as an inventory or

a taxonomy. The most common forms of intensification set the basis for
the recognition of the feature of intensity when it occurs in more covert
settings. Adverbs of intensity play a major role in the resolution of
semantic ambiguity in the study of quantifiers to follow.
The most thoroughgoing study of intensification is Thibault's (1977)
analysis of Montreal French. She draws on the Montreal corpus to
illustrate a wide range of intensifying adverbs, adjectives, interjections,
and verbal forms, including the highly developed Quebecois pattern of
swearing with forms of religious origin.
2. Intensity and aspect: Cognitive contradictions. The definition and

interpretation of aspect is one of the knottiest issues in linguistic de-
scription, and there is more disagreement here than in most areas of
linguistic inquiry. One reason is that aspect systems are not sets of
conjunctively defined concepts. At least some of the distinctions made
in the categories of tense are clear and distinct ideas, like past and
nonpast; but aspect, by definition, is a 'way of looking' at events. Ways
of looking at things are closer to associations than concepts, and there is
no clear route to agreement about how such ways of looking at things
are to be described.
46 / William Labov
A second reason for the difficulty is that certain aspect categories

tend to acquire the feature of intensity, and eventually the aspect
marker is used to signal intensity even when its other associations do not
apply. The result is a cognitive contradiction that defeats efforts to
find a conjunctive definition for the aspect category. This is particular-
ly true of the perfect aspect that signals the 'completed' state of ac-
tion. Though most English dialects do not have this aspect category,
the particle done serves as a perfect in many Southern dialects, in
Caribbean English, and in the Black English Vernacular (BEV). Intensity
plays an especially strong role in BEV aspect, which will serve to illus-
trate how association with intensity affects linguistic behavior and
linguistic description.
Done is quite frequent in BEV conversation, formal or informal, and
can be illustrated from any half-hour of recorded speech in vernacular
contexts. The characteristic, almost stereotypical use of done is shown
in example (8) from Baugh (1984).
(8a) I done forgot to turn off the stove.

(8b) Well, we usta get into trouble, and...if Pop'd catch us he
say, 'Boy—you done done it now!1
The usual interpretation of done as 'perfect' applies here, and can be

rendered by the translation with completely.
(8a1) I completely forgot to turn off the stove.

(8b1) Boy, you completely did it now!
But many speakers of BEV feel that the intensive meaning is lost here,
and that the force of the utterance is best captured as in (8").
(8a") I actually forgot to turn off the stove.

(8b") Boy, you've really done it now!
A further extension of done can be seen in the quotation in (9) from a
young woman from North Philadelphia.
(9) But next thing you know...she done stole some o' my
mother jewelry. Uerrie Chisholm, 17, PC193) 12
Stealing is an action that can be done partially, in the sense of an

incomplete attempt, or completely, in the sense of a clean sweep and a
getaway unobserved. But the intensive feature is also present, as the
intermediary for the sense of 'moral indignation1 that is evident here, as
in many other uses of done. These meanings—cognitive, emotional, and
social—are not inconsistent, and one could argue that the intensive and
indignant senses are derived pragmatically from the perfect meaning of
done. But the development of cognitive contradictions in the use of
done is seen in the quotation in (10), given by Baugh (1979:150), spoken
by a woman whose husband was cheating on her.
(10) So he went to where she was...and got the nerve to lie to
me...talking 'bout he done went to work.
Intensity / WJ
It is not at all appropriate to apply 'completely' to go to work, since

this is a discrete, socially defined action that is not done partially or
completely. Nor is it suitable to insert really, since there is no implica-
tion that anyone at that time had argued the reverse. The most likely
interpretation is that the done agrees with the intensive, morally indig-
nant use of got the nerve to, and is 'lowered' from that higher predicate
to the subordinated clause of indirect speech.
(10') ...and done got the nerve to lie to rne...talkin' 'bout he went
to the office.
Efforts to apply a conjunctive definition to done meet cognitive

contradictions in examples like (11), given by the speaker of (8).
(11) No, Miguel he done been to bed with Julia and been to bed
with Darlene...and sposta be a good friend o' Henry's.
Henry lives with Darlene, and Miguel's sleeping with Darlene is seen here
as a betrayal of Henry. The done cannot be interpreted here as the
normal 'completed' sense, since going to bed with someone is again a
socially defined act, which is not done partially or completely; or for
that matter, even intensively. The only coherent interpretation that
remains for this done is 'moral indignation".
As long as we can locate a plausible interpretation of aspect particles
in cognitive terms, oriented to the processing of information, it is not a
vital matter to recognize emotional meanings like 'intensive' or social
meanings like 'moral indignation'. They play no more important role
than any other redundant features that cluster about grammatical struc-
ture. It is quite otherwise when no cognitive or referential meaning
appears—a cognitive zero—or when the context is inconsistent with the
cognitive meanings usually recognized—a cognitive contradiction. We
then have no choice but to recognize social and emotional meanings as
an integral part of the central grammatical system.
The examples given from BEV indicate that the system of mood has
been truncated in most grammatical presentations. Mood is concerned
with the speaker's view of the existential status of the proposition: its
relation to reality. One pole of this dimension is the negative; the other
is usually considered the indicative. Yet the existence of surreal forms
like really and the intensive use of done, be, be done in BEV leads us to
the view that the unmarked indicative is the midpoint of a scale that
extends from the unreal to the surreal.
(12) You didn't do it.

You could have done it.
You must have done it.
You might have done it.
You seem to have done it.
You did it.
You really did it.
You done done it!
tZ I William Labov
Another way of looking at the matter is that a logical representation of

reality assigns probabilities of 0, 1, or some value between 0 and 1. But
in colloquial conversation, many utterances are assigned probabilities
well below 0 or considerably above 1.
3. The intensive use of quantifiers. There is no closed set of markers

of intensity. Intensity is signaled by a large and miscellaneous class of
devices, ranging from the most peripheral of prosodic variations to the
most central categories of the grammar. Labov and Waletzky (1967)
listed four kinds of intensifiers that serve as evaluating devices in narra-
tive: verbal and nonverbal gestures; expressive phonology, including
sudden changes in length, pitch, duration, and vowel quality; repetition;
and the use of quantifiers. The quantifier that drew the most attention
in these early analyses of intensification was all.
(13a) and then, when the man ran in the barber shop he was all
wounded
(13b) he had cuts all over
(13c) I knocked him all out in the street
In her study of the language used in meetings of a food cooperative, T.
Labov (1980:281-83) found that the quantifiers never, ever, always, and
all were used in a hyperbolic manner, to indicate degrees of activity that
were implausible, and assigned to them the same interpretation as the
intensifying adverbs.
For a wide range of colloquial speakers, about half of all the intensive
features in colloquial speech are accounted for by the universal quanti-
fiers: any, all, every, and ever. In the 20-minute conversation of Dolly
Ripley, there are 20 such examples. Example (14) presents character-
istic quotations.
(14a) She ain't had no kind o1 nobody to bring her up.

(14b) 3ust to say you been around and been some place, 'cause
you ain't never been no place.
(14c) I didn' bring none of my clothes back...I left 'em all down
there. That's right. I left all of 'em down there.
The use of jaU in (14) is distinct from the adverbial all of (13), which is
used to intensify participles and adverbs of place with the meaning of
'completely'. In (14c), Û is used as a universal quantifier of continuous
quantities with the mass noun clothes. It is the obverse of none: since
she left a\\^ of her clothes in North Carolina, Dolly Ripley has rio clothes
in New York.
The interpretation of adverbial aH is straightforward: we assign the
meaning 'thoroughly'. The translation of adverbial aU_ into a mark of
intensity then follows the same route as the perfect: there is a smooth
transition between saying that something is completely or thoroughly so,
and emphasizing the fact that it is indeed so.
On the other hand, the quantifier all in (14c) and the negative uni-
versal quantifiers in (14b) and (14c) involve cognitive contradictions. It
is possible in (14a) that a child had no one at all to care for him or her in
growing up, though it isn't likely. But it is not possible that the children
Intensity / *9
being addressed in (14b) had never been any place, nor is it possible,
looking at Dolly Ripley in New York City, to say that she had left all of
her clothes in North Carolina.
This investigation will approach these cognitive contradictions as the
result of the interaction of intensity with the semantics of the universal
quantifiers. But first it may be helpful to look at the pragmatic frame-
work in which this interaction is embedded.
3.1 Pragmatic approaches. One way of explaining (14b) is to uncover
implicit assumptions about what is relevant, or what counts as a member
of the set of places that fall within the scope of rio: 'You haven't been
any place that counts as far as places worth being to are concerned.' We
could abbreviate these assumptions by saying that all universal quanti-
fiers are bound in an implicit set of 'things that count' or 'things that are
worthy of mention': 'There is no X such that X is a member of places
that are worth mentioning and you have been to X.1
To deal with (14c) we would have to interpret 'things that count' as
'clothes that are needed for more than a few days'. The clothes that
Dolly Ripley is wearing, and the few that she needs for a short stay in
New York City, are perhaps not 'worth counting' in evaluating the
statement, I left all my clothes down there. The problem becomes more
difficult in dealing with the apparent contradiction of (15), spoken by a
working-class woman from eastern Tennessee, whose speech patterns are
examined in some detail in section 4.
(15) Seems like everybody's out for theirself. (Louise Adams,

42, Knoxville, Tenn.)
The normal interpretation of everybody does not include the speaker

and the addressee. No simple modification of the set of 'things worth
mentioning' will handle this situation. Rather than tailor pragmatic
rules to each new example, we may have to broaden the framework to
something as general as 'For all X, such that X is relevant in this con-
text... 1
This pragmatic approach accepts the meaning of universal quantifiers
that is conventional for sentence grammar, and tries to show how the
apparent illogic of usage is the result of interaction with a larger con-
text. A more direct approach would recognize the role of intensity by
stating that in the presence of a feature of intensity, the distinction
between all_ and almost all, between every and most, is neutralized. This
might be the result of a pragmatic rule that governs 'pardonable exag-
geration'. 16
(16) A universal quantifier in the presence of a marker of inten-

sity may be interpreted as less than universal...
There are more than a few difficulties facing this proposal. The three
dots indicate that there are many more conditions to be stated. We
would not want, for example, to include the sentence types of (17).
(17a) I swear to God I'll pay back every cent.
(17b) Every one of my children turned back Do another religion].
50 / William Labov
More generally, we have to recognize that most universal quantifiers can

be interpreted as carrying a feature of intensity without any additional
mark. It is well known that negative concord serves to intensify a
negation when it is optional (3espersen 1924:331, Labov 1972:177-78). In
BEV, where negative concord within the clause is semicategorical,
negatives can be intensified by multiplying the opportunities for nega-
tive concord through the introduction of more indeterminate quantifiers
any and ever. For all dialects, the introduction of these indeterminates
adds intensity, even without negation. Example (14a1) reflects an order
of intensity 1<2<3.
1. She hasn't had anybody to bring her up.

2. She hasn't had any kind of anybody to bring her up.
3. She hasn't had no kind of nobody to bring her up.
The same observation can be made about negative concord to the tem-
poral indeterminate ever in (14b), and the use of ever without negation.
(14b1)
1. 'Cause you haven't been any place.
2. 'Cause you haven't ever been any place.
3. 'Cause you haven't never been no place.
Similar interpretations can be made of the difference between My

children do what I say and All my children do what I say. But if all
universal quantifiers can be interpreted as carrying features of intensity,
then (16) amounts to saying that in colloquial speech, the distinction
between almost and all cannot be made. This certainly isn't so. Speak-
ers who use universal quantifiers very freely, like Louise Adams, occa-
sionally make it plain that exceptions count. At one point, she said:
(18) Now I think everybody's got a little fear in 'em.

She followed this immediately with a discussion of a possible exception:
her son Charles, who had said that he had such a good time in Vietnam
that he quit the service when they would not let him go back.
(18 I don't know...Charles said he wasn't scared over there but—

I don't know—I jus' think everybody's got a little bit o1 fear
in 'em...I got a feeling ol' Charlie boy was scared. You
know some people'll talk like they're not—all the time they
were.
We must conclude that no global rule accounts for the interpretation
of universal quantifiers. Some utterances, like (14b, 14c) are evidently
to be interpreted loosely, where all my clothes means 'most of my
clothes', and haven't been no place means 'have been to very few
places'. Even these glosses are too specific. From this point on, I will
use the term 'loose interpretation' to mean a sense of a universal quanti-
fier that focuses only on the whole and makes no division among the
Intensity / 51
members of the class, with no attention to possible exceptions. The
term 'strict interpretation 1 will apply to a sense of universal quantifiers
that conforms to logical practice and specifically rules out exceptions.
This is the sense needed to understand (IS).
The differences between the two interpretations are put into relief by
the differences in their entailments. In the strict interpretation of 'I
left all my clothes down there 1 it follows that 'There are not any clothes
that I haven't left down there'. But from the loose interpretation, no
such conclusion follows.
This distinction does not easily find its way into formal descriptions of
the English language. Among the dictionaries, the Oxford English Dic-
tionary and Websters New International Dictionary 2 agree that the
basic meaning of ^11 is 'the whole of, with a secondary meaning 'any
whatever; beyond all doubt'. It is possible to read into the various dis-
cussions a sense of loose and strict interpretation, but closer inspection
shows that the dictionaries are focusing on distinctions between the
quantifiers. It will appear further on in this paper that the universal
quantifiers do not differ in their susceptibility to loose or strict inter-
pretations.
The most detailed linguistic analysis of the semantics of universal
quantifiers is Sapir's essay on 'Totality' (1930). Sapir focuses on psycho-
logical processes that might support distinctions made in formal reflec-
tion, and specifically warns (1930:11):
It is not claimed for a moment that the ordinary English uses of 'the
whole1, 'all of, 'the whole of, and 'all' necessarily correspond to our
exacting distinctions, merely that they tend to do so. In actual
practice there is considerable confusion.
Sapir believes that all in generic statements like All men are mortal is
merely a class indicator, no different from Men are mortal, and not
'totalizers in the strictly logical sense'. But this suggestion is limited to
generics, and does not apply to the concrete statements about past and
particular events that are the main materials we have been examining.
To make any progress in locating the intersection of intensity and
quantification, we must first examine speech systematically, and locate
the conditions that favor the loose interpretation and those that favor
the strict interpretation. This reverses the procedure of the first half of
the paper. Instead of asking, 'How do people signal intensity?' I will ask,
'How do people interpret universal quantifiers?'
4. The interpretation of universal quantifiers. This section examines

100 universal quantifiers used by Louise Adams, of Knoxville, Tennes-
see. The focus is on local features of sentence grammar that may
influence the listener in the loose or strict interpretation of these quan-
tifiers. There are several special subsets that have acquired convention-
al meanings, irrespective of the local context. The main classification is
concerned with the observations that would have to be made to confirm
or disconfirm the statement within the scope of the universal
quantifier: primarily the denumerability of these observations, a feature
closely involved with other aspects of sentence grammar.
52 / William Labov
The classification is an ordered series of criteria. Any classification

that applies immediately determines the status of a sentence, without
regard to cross-classification by other criteria to follow. Following each
classification is an example from Louise Adams and an explication of
how the meaning of the quantifier would be determined by the ordered
series of criteria.
Type A. CONVENTIONAL. Sentences where strict or loose inter-

pretation is given by a fixed convention.
Several special uses of quantifiers have evolved over time which have
immediate interpretations as loose or strict, not relative to the imme-
diate context.
CO Unique. Modifiers of unique objects or events.
(19) Kay Francis was born 'fore I ever left 'em there.
Louise Adams left her in-laws' house to set up housekeeping only once.
Ever cannot be a quantifier, strict or loose, but is immediately taken as
intensive, like the examples of any prefixed to unique objects.
CO Approximate. The quantifier is the head of a noun phrase

introduced by a conjunction.
Such noun phrases are conventionally interpreted as invoking a set of

elements with a rough family resemblance to the prototype in the pre-
ceding phrase.
(20a) So I don't work or anything.

(20b) She puts all of it on him, all the cleanes', 'n' everything.
The interpretation is automatically loose.
GO Adverbial. AU functioning as adverb rather than deter-
miner, adjective or noun.
This use is recognized by Websters New International Dictionary 2 with
the note: 'jall^ is frequently placed before a word or phrase to add the
idea of completely, intensity or thoroughness.' As discussed in section 3,
these uses of aU are all inevitably the loose interpretation, since they
are not focused on discrete elements that can form exceptions. The
second ajU in (20b) is an example, as are the instances in (21).
(21a) And he said, 'When you been like 'at and been all over, you
see things', V he said, 'Marriage just ain't for it.'
(21b) Jimmy's been aU through it all.
(21c) When he w's a kid, I had to go all the way around the state
here gittin' that young 'un.
We must also include the adverbial at all, which is immediately inter-

preted as C+ intensive).
Intensity / 53
(21d) It's not like it used to be at all.
DO Limiter. The quantifier is in a sentential context that

denies or limits universality.
Such sentences do not make any sense at all if we do not assign them
strict interpretation. Speaking of wild game that the men in her family
brought home from hunting, Louise Adams said:
(22) Now me, I don't eat everything come in the house.
The loose interpretation would amount to saying 'It is not true that I eat
almost everything that comes into the house', but that might very well
be true in this case. This sentence demands a strict interpretation for
everything, just as all but one does for all.
D3 Nonliteral. If the strict interpretation is assumed, the

statement is obviously false.
This is the most common of all the conventional uses of quantifiers,
exemplified by (14b, 14c) from the speech of Dolly Ripley. It is also
quite common in the speech of Louise Adams.
(23a) She always been like 'at, all through her life.
(23b) He eats all the time and don't even get fat.
(23c) 3immy has give those kids—you ask her—they've got every-
thing!
(23d) All he wants t'do is wrassle and that's it.
(23e) There ain't anything he wouldn' do for me, for her or for
anybody.
(23f) She can. She can doctor anything.
(23g) CThey shoo-Q squirrels, possum...anything they can see!
(23h) Everything's changed.
(23i) Seems like everybody's out for theirself! (=(15))
Interpretation is necessarily loose. Section 3 considered some of the

possibilities for pragmatic rules that would reinterpret the strict inter-
pretation. This seems especially promising when the listener can con-
struct an unstated context that allows the strict interpretation. For
(23d), for example, we can imagine a context 'Whenever you go to play
with him...' But in the great majority of nonliteral types, for Louise
Adams and others, these reconstructed contexts are hard to find and
arbitrary at best.
Type B. UNOBSERVABLE. Sentences where no objective observa-

tions can be made to confirm or disconfirm the statement.
Whereas the preceding set of sentences have automatic interpreta-

tions, there are a great many sentences where the listener has no means
of knowing from the immediate context whether strict or loose interpre-
tation is to be made. From this point on, the questions pertain directly
to observations that would confirm or disconfirm the statement within
the scope of the quantifier.
5* / William Labov
[jQ Self-defined. The occurrence of the event is determined by

the observation, so that any possible observation is a
confirmation.
(24a) I could always get in a good argument.

(24b) That would git 'im, oh God, that would git 'im every time.
(24c) In fact, I always said that I had two too many children.
Consider (24c): unless we take the impossible interpretation that Louise

Adams repeated the statement endlessly, it can only mean that when
there was an occasion for Louise Adams to say that, she said it. Only
the speaker would know when the occasion to say it had arisen.
C7D Opinion. The statement is a matter of subjective opinion so

that only the person directly involved can decide if it is
true or not.
In such cases, the listener can hardly assign an interpretation without

the help of further information from the speaker. Usually, we are
dealing with the opinion of the speaker, but sometimes, as in (25c), it is
someone else.
(25a) That didn1 do him no good, and still ain' doin1 nobody no
good.
(25b) That youngest 'un and that girl o1 mine, they don't care
anything about wrestlin1.
(25c) I run from anything like that.
CThe opinion involved in (25c) is the question of similarity:
what is like that? Only the speaker can decided
(25d) If I stand and look at anything like 'at, everything jus' turns
dark to me, and I pass out.
(25e) Her daddy got a goat one time, I never will forget this.
(25f) Jus' plain ol1 cows milk, that's all I like.
(25g) You ain1 gon get me on no boat. You ain' gon' get me on no
airplane.
(25h) I said [pf politicians], 'Well, when one's done one thing dirty,
you jus' go t'look at 'em back, they're all_ dirty. They'll
come out dirty. They go in clean, they come out dirty.1
(25i) Now I think everybody's got a little fear in 'em.
In section 3, (25i) was interpreted Qis (1&D by Louise Adams herself in

the strict sense as she proceeded to consider counterexamples. There is
no doubt that the interpretation of universal quantifiers is in the hands
of both speaker and listener, and the meaning is often negotiated
through social interaction. As outside observers, we must be willing to
wait and see what people do make of utterances. But as speakers of the
language, we are entitled to search for whatever clues the grammar
gives us on the interpretation of such critical matters, before we decide
to raise a question or objection. Sentence (25h) expresses a common
theme in American political life; a politician addressing Louise Adams
would have to decide whether she intended the loose or strict
Intensity / 55
interpretation before knowing what to say next. Sentences of type B
offer less opportunity for us to estimate the probability of a strict
interpretation than those to follow.
The great majority of Unobservables are in the opinion class, and in
the analysis to follow, self-defined are merged with opinions. An over-
view of all speakers to be considered shows that the self-defined class
does properly fall within Type B; in all respects these sentences prove to
be prototypical Unobservables, even though the actual statements refer
to events that can be observed.
Type C: OBSERVABLES. Sentences where objective observations
might be made by listeners or others.
DO Single. A single observation would confirm the statement.
In spite of the fact that we are dealing with large, inclusive, and often
badly defined sets, the deictic configuration sometimes allows the
universal to be confirmed in a single observation. This happens when the
field is suddenly cleared by an exit (Everybody ran; and I was standing
there) or when the field is blank to begin with (compare Nobody came to
Nobody left).
(26) They ain't nobody to take him up on it Cher husband
Charlie's offer to go fishing].
Since Charlie is not going fishing, it is clear that no one has appeared.
Sentences like this have the highest probability of a strict interpreta-
tion, since they are the most easily confirmed. Because they are rela-
tively rare in the data, they will be merged with the following class of
known quantities.
GO Known. The set to be observed is a known quantity.
The most common examples of such known quantities are the speaker's
children, members of a team or a jury. Families as a whole are usually
defined more loosely.
(27a) He gets in on time for us all to go to wrestlin' on Friday
night, and we go.
(27b) Yeh, they all Qhe children] went to f ishin'.
(27c) Now all my kids go t1 water but not me.
(27d) An' every one o' my kids turned back [changed their denom-
inations].
The likelihood of a strict interpretation here is almost as high as in the
preceding case.
QcO Denumerable. The quantity of observations to be made is
not known, but it could be enumerated.
56 / William Labov
(28a) I go every Friday night.

(28b) He tries to see Mama goes somewhere every week.
(28c) He'il bite, but he won't marry, marry no girl.
(28d) He had his head sewed up about 4, 5 times; go swimmin',
he'll go swimmin1, an1 he gets his head cut ever1 time he
goes.
The high frequency of these nondenumerable observables is partly the
result of a rhetorical process that introduces universal quantifiers in
situations that logically might call for only simple declarative state-
ments. Louise Adams shows this tendency more than most speakers.
When I asked her what diapers she used, she responded with (28f), and
(28g-281) followed in quick succession.
(28f) That's all we ever had back 'en.

(28g) You never heard o' Pampers 'til after.
(28h) That's all I ever used.
(28i) That's all I ever did for my kids.
(28j) They never w's bothered with it Giiaper rasrj].
(28k) I always changed 'em.
(281) I don' remember non o' my kids bein1—what you'd say real
sick.
Though (28a-28l) vary in likelihood, they may in general be considered

less likely to receive a strict interpretation than those utterances that
could be more easily confirmed.
This classification presumes that listeners and speakers are possessed
of a sense of fact. Up to this point, there has been no evidence to
support this idea. The following has been established for Louise Adams:
1. The variety of interpretations required for universal quanti-
fiers is considerable.
2. There are many sentences where the loose interpretation is
required.
3. There are many sentences where the choice between loose
and strict interpretation is not easily made.
4. Though sentences vary in the probability of strict interpre-
tation, the only sentences where the strict interpretation is
required are those where it is denied.
Section 5 advances indirect evidence that speakers and listeners are

alert to the categories that suggest an empirical basis for the choice of
strict and loose interpretations.
5. The distribution of intensity. The categorization of quantifiers by

the local context has established that in about one-third of our cases,
strict quantification would not make sense; in another third, the dif-
ference between strict and loose interpretation would be impossible to
decide by objective means; and in most of the remaining third, disproof
by counterexamples is possible, but positive confirmation is not. It now
remains to explore the relevance of these categories to intensity.
Intensity / 57
Some of the conventional uses of quantifiers are invariant in respect
to intensity. The approximants and everything, and all, and anything like
that are unstressed and never signal intensity in themselves. Limiting
expressions that deny universality also act as deintensifiers: I don't eat
everything comes in the house. Certain minimizing expressions use
universal quantifiers, as in That's all, normally deintensifying.
On the other hand, adverbial quantifiers like all ruined are uniformly
intensive, along with a number of fixed intensifying expressions, e.g.
Every time I turn around, all of a sudden. Intensity seems to be
heavily concentrated in certain conventional expressions that do not
permit strict interpretation, particularly the nonliteral uses. It is true
that intensity can be conveyed by universals that are marked for strict
interpretation, such as every single one of, and absolutely all of. Yet
the use of quantifiers as markers of intensity seems more likely when
the loose intepretation is more likely.
Given the loose interpretation, one can ask: why was every used in
place of most, all in place of almost all? This choice may be motivated
by the fact that every and al[ can also signal intensity. Given a great
likelihood of loose interpretation, a universal quantifier may be assigned
the intensive feature even when no other markers of intensity are
present.
Since marks of intensity usually cluster, it should be possible to con-
firm this notion by plotting the frequency of other marks of intensity in
combination with universal quantifiers:
prosodic contrast: stress on the quantifier beyond that predictable
from phrase and sentence structure; also laughter,
negative concord: the incorporation of negative particles in inde-
terminates any, ever, either following the first negative of the
sentence,
adverbs of intensity: really and other adverbs of intensity with
scope including the universal quantifier,
repetition: usually in the form of the repetition of whole clauses in
rapid succession,
inversion: shift of placement of the quantifier from its normal
position after the first tensed member of the auxiliary: I never
will forget this.
Figure 1 shows the distribution of the other marks of intensity that
accompany the 100 universal quantifiers of Louise Adams. Slightly more
than a third are Unobservables: these show the highest concentration of
intensifiers. A little more than a quarter are Observables. The known
class is too small to be shown by itself, and it is combined with the
denumerable class, so that Observables are divided into only two types.
Of these the largest part are nondenumerable, with a low frequency of
intensifiers, and even lower are the smaller denumerable group. The
conventional uses, with fixed interpretations, show an intermediate
range of intensifiers.
The numerical values are given in Table 1, along with the numbers for
all other diagrams to follow.
58 / William Labov
Non-denumerable
OBSERVABLE
UNOBSERYABLE
Denumerable
• _
Approximate
Adverbial
Non-literal
CONVENTIONAL
Fig. 1. Per cent markers of intensity accompanying
universal quantifiers: Louise Adams. Knoxville, TE
IN = 100]
91- 51-90 i|i|!i 26-50 11-25 1-10 00
Table 1. Distribution of quantifier classes and accompanying

marks of intensity for five English speakers.
L. Adams 3. Lynch S. Romano A. P. M. Beachy
Conventional:
Limiter 0/2/1 0/1/3 0/22 0/18/5 0/1
Unique 1/1 0/0 0/0 0/0 0/0
Approximate 1/1 0/20 0/1 0/1 0/0
Adverbial 4/5 5/23 0/16 0/4/1 0/11/5
Nonliteral 8/8/1 7/7 14/19 0/6 0/0
Unobservable:
Self-defined 3/0 2/0 3/1 0/2/1 0/1
Opinion 25/8/3 11/8 13/30/2 2/8/15 4/3/6
Observable:
Single 1/0 2/4 0/1 0/4/1 0/1
Known 0/4 0/0 2/5 0/14/1 1/0
Denumerable 1/5/1 0/5 0/5 0/6/2 0/2/1
Nondenumerable 3/13 7/18 11/14 1/3/11 1/2/5
Totals 47/47/6 34/86/3 43/114/2 3/66/37 3/25/ll
N: 100 123 159 106 44
% intensified 47 28 27 03 14
% deintensified 06 02 01 35 39
x/yA/z) = intensified/not intensified l/deintensified).
Intensity / 53
There is evidently a firm relation between Louise Adams' use of

intensifiers and the probability of loose or strict interpretation for the
listener. The more likely the strict interpretation of the universal
quantifiers, the less likely it is that she will use other intensifiers. It
follows that the more likely the strict interpretation, the less likely that
the listener will interpret the quantifier as an intensifier itself.
The pattern of Figure 1 is not idiosyncratic to Louise Adams: it is
characteristic of almost all of the speakers I have examined so far.
Figure 2 shows the corresponding distributions for two older white
Philadelphia working-class men. Jim Lynch is a retired stevedore and
factory worker from Kensington in North Philadelphiat Stanley Romano
is a barber from the central area of South Philadelphia.
Lynch shows a much heavier use of adverbial all than Adams (all over,
all the way up, all around); a greater use of approximates like anything
else, or anything; and a more limited use of quantifiers in opinions. The
Unobservable area is considerably reduced: Lynch does not express a
great many opinions with universal quantifiers. Romano, on the other
hand, uses roughly the same proportion of Unobservables and nonliteral
quantifiers as Louise Adams, but shows a sizeable number of limiting
forms: these are mostly the expression that's all.
These differences in the distribution of the quantifiers are relatively
superficial differences in personal style and lexical choice. The constant
features of the quantifier distributions are: (1) Observables account for
only about one-quarter of the data; (2) of these, the greatest part are
nondenumerables; and (3) nonliteral uses of quantifiers are quite fre-
quent.
The distribution of intensifiers for Lynch and Romano is basically the
same as for Adams. The heaviest use of intensifiers is with Unobserv-
ables; next among the Observables are the nondenumerables; and least,
the denumerable types. This is the ordering that relates intensity to
observability. There is some variation in the amount of intensity associ-
ated with the nonliteral forms: with Lynch it is as high as the Unobserv-
ables, and there is some fluctuation with the approximates. The only
variation in the observability hierarchy is that for Romano: there is no
greater use of intensity with Unobservables than with nondenumer-
ables. Within the gross categories of intensity level established here, it
appears that:
nonliteral
= < Unobservable
= > nondenumerable
= > denumerable
These speakers appear to have a sense of fact. Their use of universal
quantifiers to signal intensity is proportionate to the possibility of the
statement being contradicted. Yet the dominant semantic of the uni-
versal quantifiers is the loose interpretation, as shown by the free use of
nonliteral expressions that would be obviously false if the strict inter-
pretation held; and by the heavy cooccurrence of universal quantifiers
with other intensifiers in the majority of uses outlined by Figures 1 and
2.
60 / William Labov
JIM LYNCH, 72 r No. Philadelphia [N = 123]
UNOB5ERVABLE Non-denumerable
OBSERVABLE
Non-literal
Denumerable
Approximate
Adverbial
CONVENTIONAL
STANLEY ROMANO, 66, So. Philadelphia [N = 159]
Non-denumerable
UNOBSERVABLE OBSERVABLE
Denumerable
Limiting
Non-literal Adverbial
CONVENTIONAL
Fig. 2. Per cent markers of intensity accompanying
universal quantifiers for two Philadelphia speakers
51-90 | | | 2 6 - 5 0 :Wl 11-25 1-10 00
Intensity / 61
6. Exceptional speakers. This exploratory analysis of intensity is not
intended as a sociolinguistic study of intensification and quantifiers
throughout society. But it is evident that the pattern of quantifier use
seen here is a part of the colloquial language. It is not confined to
vernacular settings, but is used in spontaneous speech in relatively
formal situations. The speakers studied here have been recorded in an
interview format that sometimes approaches vernacular use (Labov
1984); but much of the speech produced would be classified as 'careful
speech1 in a thoroughgoing stylistic analysis.
A different pattern prevails in formal writing, where the balance of
semantic interpretation tilts strongly toward the strict interpretation,
and most of the conventional uses of the quantifiers are missing. There
is therefore a strong stylistic constraint on the use of universal quanti-
fiers to signal intensity, and on the loose interpretation. Whether or not
there is social stratification remains to be seen. It may be that the
pattern of Figures 1 and 2 is typical of all English vernaculars, no matter
what the social level of the speaker. But it is not binding on all sponta-
neous speech. To examine the maximum variation, I turned to an older
Philadelphian who is a dominant figure in the upper social stratum of the
city. A.P. was 65 years old, head of a family that has influenced
Philadelphia social and economic life for many generations. He is a
prominent lawyer, has taught law in universities, and has made policy in
many of the social clubs that regulate the upper class social life of
Philadelphia. His speech is archetypical of the pattern described by
Kroch (1980): a conservative but relaxed Philadelphia phonology; a
precise syntax with careful attention to logical distinctions but none of
the intense concern with grammatical niceties that marks less secure
speakers; and a strong tendency to intensify attributions with adverbs
and a regressive accent: 'vastly higher', 'a very very small distance1, 'a
rather small child', 'perfectly satisfactory', 'completely local'.
In an hour of speech, A.P. used 106 universal quantifiers, much less
than the rate of Lynch and Romano. The distribution by observability
was quite different from the other speakers we have examined. There
was about the same proportion of Unobservables (a third), but a greatly
reduced proportion of conventional uses (a sixth) and a greatly expanded
set of Observables. This expansion is entirely due to an increase in the
number of known sets (18%). A.P. was quite conscious of his family
connections, and referred to them quite often with universal quantifiers
where others might have used simpler expressions: all four of my grand-
parents, None of the immediate family, my own children have, all three
of them... Beyond the family relations, we find such constructions as
three people all of whom already had their Ph.Ds.
Among the Conventional uses of the quantifiers, A.P. uses a few
approximates and a small number of nonliteral uses (7%). In spite of his
precise habits of speech, he occasionally uses nonliteral expressions that
can only be heard as the loose interpretation of the quantifier, as shown
in (29).
(29a) ODf an early relative] The girl who identified all the
witches.
62 / William Labov
(29b) They were interrelated as all the people in that county

were.
(29c) Cpf the Betsy Ross Bridge] Nobody uses it.
A more profound difference between A.P.'s speech and the others is
found in the use of intensifiers. There are only two clear cases in his
speech, given in (30).
(30a) None of the immediate family have any connections up
there at all.
(30b) COf the use of any word other than nigger in the rhyme
'Eeny meeny miny mo1] Never, never gave it any—never
would have occurred to any of us.
The absence of negative concord in A.P.'s speech accounts for a part of
this effect. In Louise Adams' speech negative concord accounted for 15
of the 50 marks of intensity that accompanied universal quantifiers. If
we subtract these, we are left with 35 intensifiers on her part as against
2 on his. We might also consider the occurrence of multiple indeter-
minates in A.P.'s speech as marks of intensity, since there is no contrast
with negative concord. Multiple indeterminates are found in (30a, 30b)
along with repetition, inversion, and the intensifier at all, and in two
other cases, given in (31).
(31a) I never had any trouble with that group.
(31b) I don't think it ever did me any harm.
All four of these intensifiers are found in the Unobservable class. This
confirms the conclusion drawn from A.P.'s use of nonliterals. He has
available to him the same mechanism that other speakers use: intensifi-
cation supporting the loose interpretation. He uses it much less often.
This similarity does not disguise a bigger difference between A.P. and
the speakers of Figures 1 and 2: a great development of limiters. The
other speakers occasionally weaken or limit the universal quantifiers by
several different devices. The main assertion may be mitigated by ]_
guess, I imagine, I think, etc. Louise Adams said:
(32) I imagine ever1 one o1 them had t1 take their turn there Con
the front lines].
In example (33), Lynch modifies the claim that all of his friends were
married with pretty near, which I take to mean that most of them were
married, not that all were about to be married.
(33) We were all pretty near married.
As noted, an adverb of intensity can be inserted within the scope of the
quantifier. In example (34), Lynch uses an intensifier both without and
within, strengthening and weakening the statement.
I never really got a real beating.
Intensity / 63
There are not many more examples of limiters in the speech of the first
three persons studied. But there are 23 in the interview with A.P., in
addition to many limiters that do not accompany universal quantifiers.
There are many examples of mitigated assertions, as in (31b). We find a
variety of other mitigating devices—adjectives, adverbs, and adverbial
phrases—as indicated in example (35).
(35a) All the people from that peninsula have a little bit of that
in varying degrees.
(35b) I never knew her entirely well.
(35c) Chestnut Hill Academy's never been a major factor for
Harvard.
(35d) Over any period of time, it's pretty substantial.
(35e) We've never been able to reduce it quite to that.
It is not uncommon for the universal quantifier to be directly denied or
restricted in a way that demands strict interpretation, as example (36)
shows.
(36a) It's not an absolute or anywhere near it.
(36b) Every night or nearly every night.
A.P. often uses an accumulation of limiting devices:
(37a) until I figure nearly virtually until his death.
(37b) I don't think the instruction is necessarily any better.
Examples (31b) and (37b) show both mitigation and intensification. How
these patterns combine is a nice problem for semantic calculation: the
difficulty is probably no less for the native listener.
The study of intensity must then consider its obverse, the process of
deintensification exemplified by these limiting devices. Figure 3 maps
the distribution of deintensifying marks used by A.P. against the same
classes of universal quantifiers that were used in Figures 1 and 2. The
known class, which was combined with denumerables in the other dia-
grams, is here separated since the numbers are relatively large. The
highest concentration of deintensifiers is found in the unobservable and
nondenumerable classes of Observables. Fewer deintensifiers are found
in the denumerable class, and even fewer in the known class. A.P.'s use
of deintensifiers is parallel to the other speakers' use of intensifies.
This similarity in pattern should not conceal the semantic opposition.
The more likely the loose interpretation, the more likely is A.P. to
deintensify it. This suggests that his basic definition of the universal
quantifiers is the strict interpretation, and he takes care to avoid a loose
interpretation by his listeners. In contrast, the other speakers use
intensification to reinforce the probability of a loose interpretation as it
becomes more probable. They take as target what A.P. takes as pitfall.
/ William Labov
Non-denumerable
UNOB5ERVABLE
OBSERVABLE
Denumerable
Known
CONVENTIONAL
Fig. 3- Per cent markers of de-intensification with
universal quantifiers: A.P., 72, Philadelphia [N = 1061
91- 51-90 J26-50 | : ; : : i ! J l l - 2 5 |- ' : | l - 1 0 | | 00
Non-denumerable
UNOB5ERYABLE
OBSERVABLE
Denumerable
Adverbial
CONVENTIONAL
Fig. 4 Per cent markers of intensity and de-intensity
with universal quantifiers: Morris Beachy, Iowa
IN = 441
91- 51-90 26-50 11-25 1-10 n 00
Intensity / 65
The advanced education and high social status of A.P. might be
thought to account for this reversal of semantic process. But this pre-
cision of speech and the preference for the strict interpretation is not
limited to upper class speakers. For a second exceptional speaker, I
turned to an interview with Morris Beachy, 65, from a small town in
south central Iowa. Beachy had only a ninth grade education, and he
came from a family of small farmers who were members of the Mennon-
ite community which emigrated from Pennsylvania to Indiana and Illi-
nois, and then to Iowa several generations ago. Beachy had followed his
father in assuming the duties of bishop of the local Conservative Men-
nonite church. While he was more open and easygoing than A.P. in his
surface manner, he shared with A.P. a long personal history of directing
the affairs of others, and setting policies for organizations. He had an
even stronger tendency to a judicious and guarded statement of his ideas
and policies.
The distribution of Beachy's universal quantifiers was not so very
different from the norm established by the first three speakers. He had
the usual one-third of Unobservable sentence types, and one-third Ob-
servable, with the usual distribution of two-third nondenumerable and
one-third denumerable. There was no expansion of the known category
as with A.P. The most striking differences emerged in the Conventional
area. Beachy did not use a single nonliteral form, and no approximates.
All of the conventional forms are adverbial all, with a strong concentra-
tion on at all. The other major difference is quantitative: in more than
an hour of free conversation, where Beachy spoke almost continuously,
he used only M universal quantifiers.
The small number of intensifiers showed the usual distribution: mostly
in the Unobservables, with fewer in the nondenumerable category, and
even fewer in the denumerable category. Figure k shows this distribu-
tion in the central circle, and in the outer circle, the distribution of
deintensifying devices. Beachy is particularly rich in this area. He used
a great many nonuniversal quantifiers that are limited by definition, and
qualified them even further, as shown in (38).
(38a) We pretty well would agree on basic doctrines, I think.
(38b) I think probably largely so, probably.
(38c) Most of our growth, and we've had some growth...
(38d) We emphasize that quite a little.
(38e) Not nearly as many as one time here.
As example (39) shows, Beachy qualifies most statements carefuly, but
he also intensifies them, often combining both intensifying and deinten-
sifying devices.
(39a) I think it would be almost very very necessary.

(39b) ...most of our people, and I can speak only for our people,
that I know real well...
(39c) I think that people may have that sometimes, I really do.
(39d) I don't suppose that I know a lot of people that really
strongly believe that.
(39e) COn whether he knows everyone's church membership!] Not
quite. But there're not too many in the community that I'm
not aware where they go to church at all.
66 / William Labov
The same combination of intensifiers and deintensifiers is found in

statements with universal quantifiers, often with logical double nega-
tion.
(40a) I guess we've never really regretted...
(40b) There's some of that in all the Mennonite groups, really.
(40c) I can't say that it doesn't affect doctrine at all.
Some of these sentences include propositions reinforced by intensifiers,
but the strengthened proposition is qualified even as it is strengthened.
A paradigmatic example is:
(41) I'm not sure that's not true with aU branches of Mennonites
to some degree too.
These deintensifying practices of qualification and restriction suggest a
strong orientation toward the strict interpretation of universal quanti-
fiers. This comes to the fore in the exchange in example (42). Beachy
made an intensified and unmitigated statement about the church's policy
on television, which differentiated it from the most liberal sect.
(42) Our three churches, we have no television at all. It's been a
church agreement not to have television.
This prompted me to ask if there were any exceptions, assuming the
strict interpretation:
W.L.: No one has a television set in their homes?
M.B.: If they do, it's not known to me. I wouldn't say they are
not—absolutely are not. There are not expected to be, but
then...
Like A.P., Beachy behaves as if the strict interpretation were the un-
marked meaning of the universal quantifiers.
7. Grammatical heterogeneity. The first three speakers examined
here showed considerable similarity in their use of quantifiers and inten-
sifiers. The fact that two other speakers showed radical differences
should not suggest that the English speech community is evenly divided
in this respect. The two exceptional speakers were picked because I
expected them to be exceptional. They use a formal English that has
many features in common with the written language. It is certainly
speech, and there are undoubtedly many other people who talk this
way. But these differences in the pattern of intensification and quanti-
fication may indicate an imperfect penetration of the vernacular lan-
guage by patterns that are developed primarily in the written language.
No mention has been made of the universal quantifier each, which has
played a role in many recent linguistic analyses of quantifiers. Each
illustrates the heterogeneous character of the grammar in a dramatic
way. It is first of all an exception to the pattern of loose and strict
interpretation, since it seems to have only the strict interpretation.
There is also reason to think that the role of each in the spoken language
Intensity / 67
is quite limited. There are two occurrrences of each among the 500-odd
quantifiers examined here: one is used by A.P., and the other by Morris
Beachy.
I won't try to resolve here the problem of writing a realistic grammar
of universal quantifiers. But if these exploratory studies of intensity
have any significance, that problem must be high on our agenda. It will
be hard to solve because the description must deal with the social con-
flict that is symbolized, and sometimes implemented, by the differential
treatment of intensity and quantifiers. It remains to be seen whether we
can describe language as it is, and not the way it is expected to be.
Notes
1. This paper is the latest in a series aimed at a description of lan-

guage that recognizes social and emotional elements in the central
components of linguistic structure. These included papers given at the
Cognitive Science Society in 1981 ('On restricting the meaning of mean-
ing'); the LSA in 1981 ('A sociolinguistic approach to problems of mean-
ing'); at the University of Colorado ('Defining the meaning of meaning');
at Harvard in 1983 ('How things go together'). This is the first to be
published, since I think it is the first that succeeds in building the empir-
ical base that promises cumulative results.
I am indebted to many along the way for constructive and destructive
criticism that led me to take down a number of formal models that are
all too easy to put up: in particular for brief exchanges with George
Lakoff, Charles Fillmore, Guy Carden, David Premack, and Michael
Geis. I am grateful to Michael Silverstein for bringing to my attention
Sapir's important work on quantifiers. The current version owes much to
extended discussions with my colleagues Gillian Sankoff and Tony
Kroch. Kroch's contributions are too many to enumerate, including
many continued reservations that have made it plain to me that I have
not yet got to wherever I am trying to get. Convergence with the work
of T. Labov is evident throughout the paper: I am indebted to her for
many observations which anticipated my own, and for corrections of
many lapses of logic and detail.
2. There is a curious asymmetry of persons involved here. I.am angry
needs appropriate peripheral signals to be effective. But He is angry is
not conveyed better by adding the prosodic signals of intensity, which
are not easily transferred to third person objects but continue to refer to
the state of the speaker. The same asymmetry applies to social infor-
mation. A number of linguistic variables operate as claims to specific
social identities of the speaker or the addressee, or as signals of accom-
modation to the addressee. But the social characteristics of the addres-
see can only be inferred indirectly and with much less certainty from the
forms used by the speaker.
3. We will encounter some expressions with both markers of intensifi-
cation and deintensification. Their position on the scale of intensity is
indeterminate, since there is no way of assessing the quantitative value
of such marks at present.
4. A pragmatic translation of really is not hard to find. We can say
that really informs us that there are others who do not support the
proposition that it modifies (that the proposition is a D-event in the
68 / William Labov
Labov and Fanshel 1977 terminology). The pragmatic periphrasis cannot

be entirely satisfactory because it loses the emotional content expressed
by intensity, shared by very, actually, basically, and really.
5. The Ripley interview was part of the Lower East Side study of New
York City (Labov 1966). Dolly Ripley is a black woman, 46 years old in
1963, whose family was divided between a small town in North Carolina
and New York City. In the midst of the interview, the telephone rang. I
went into the next room and talked with her nephew, while she talked
for 20 minutes with her cousin and her cousin's husband. She was wear-
ing a lavaliere microphone and the Nagra tape recorder was turning in
full view, but she unconsciously assumed that the interview situation had
been transferred to the conversation in the next room.
6. Note that in (lc) the adverb is lowered to the infinitival comple-
ment, but the semantic interpretation normally assumed continues to
apply to the matrix clause.
7. The formulation of a surreal grammatical category is difficult to
justify on the basis of a semigrammaticalized form like really. In the
Black English Vernacular, there are several elements of the grammatical
system usually considered aspect markers which appear to have shifted
into the mood system with the intensive feature as a major semantic
element: be, be done, and (see further on) done.
8. Among the many variants of this intensifier, there are polite
substitutions like the heck, and mitigated forms like in heaven's name, jn
the world. An alternant of the hell is the devil, with euphemistic vari-
ants like the deuce and the dickens. A common aggravated form is what
the fuck...
9. Recorded by Teresa Labov in 3une 1971.
10. The term 'perfective' is now generally applied to aspect cate-
gories like the English present perfect, which are attached to actions
that precede and condition following events of interest. The 'perfect' is
used here in the conventional sense of a 'completed' action, viewed as a
whole. It may be the prior condition for a present situation of interest,
or the result of another event.
11. It is often possible to produce translations with the present per-
fect of other English dialects, since there is considerable overlap in the
appropriateness of the meanings of 'current relevance' and 'completed1.
But such translations are only accidental, and not possible in the exam-
ples of (9)-(ll) to follow.
12. Interviewed by Wendell Harris as part of the NSF project on 'The
influence of urban minorities on linguistic change'.
13. As for example, the associations of words tapped by the semantic
differential procedure of Osgood, Suci, and Tannenbaum (1957). Though
such associations may be reliable and replicable, no one would suggest
that dictionaries add for each entry a rating on the dimensions of power,
activity, and value.
\k. T. Labov found many nonliteral uses of quantifiers in her data,
e.g.: 'In other words, the second time you default you're thrown out as a
working member, Daughter]] and you will never work for the Co-op
again...Qaughtec]1 (1980:281). As she observes, the laughter indicates the
listeners' recognition that never is an implausible use if taken literally.
In this, as in the case of intensifying adverbs, 'it is often difficult to
discover any literal contribution to the meaning1.
Intensity / 69
15. Louise Adams is a 42-year-old white woman, born and raised in
Knoxville, Tennessee, of a middle working-class family. Her East Ten-
nessee dialect is quite representative of the Appalachian English de-
scribed by Wolfram and Christian (1975). She is an extremely voluble
speaker, and uses more universal quantifiers than most. The 100 uni-
versal quantifiers analyzed in the next section were used in 20 minutes.
16. I owe this term to Guy Carden, together with other insights on the
interpretation of quantifiers.
17. There are a large number of other constructions with all_ that
show idiosyncratic properties. When alj[ is the subject head of a relative
clause, it can be a mitigating, deintensifying form, as in All(s) I know
is..., or intensive in All he wants to do is... Exactly where All I can is...
fits is hard to say.
18. It was noted earlier that intensity can be signaled by the implan-
tation of the indeterminers ever and any in the sentence structure
without negative concord. However, this is a much weaker signal than
the negative concord itself, and it is therefore not counted for those
speakers who do use negative concord. I am not sure how its effect
should be estimated for speakers who do not.
19. When the adverb of intensity falls inside the scope of the uni-
versal quantifier, it has the opposite effect: of limiting the statement
made by the quantifier. Compare There really isn't any honor today with
There isn't any real honor today. The second intensifier falls within the
scope of the quantifier and it can be heard as a modification or limita-
tion of the first. See further on for the effects of such limitation.
20. Both speakers were interviewed by Anne Bower, as part of re-
search on the social origins of sound change in Philadelphia, under a
grant from NSF. Both use universal quantifiers quite freely, though not
at the level of Louise Adams. While Adams produced the 100 quantifiers
in 20 minutes, the 123 items summarized for Lynch, and the 159 for
Romano, were used over the course of an hour.
21. Interviewed by Kroch in his study of the upper class of Phila-
delphia. Data from this series are incorporated in the study of the social
origins of sound change (Labov et al. 1982, Labov 1984).
References
Baugh, 3ohn. 1979. Linguistic style-shifting in Black English. Unpub-
lished dissertation, University of Pennsylvania.
Baugh, John. 1983. Black street speech: Its history, structure and
survival. Austin: University of Texas Press.
De Groot, A. 1949. Structural linguistics and syntactic laws. Word 5.1-
12.
Jespersen, Otto. 1924. The philosophy of grammar. London: Allen and
Unwin.
Kroch, Anthony. 1980. Dialect and style in the speech of upper class
Philadelphia. Paper given at Conference on Social Psychology and
Language, Bristol, England.
Labov, Teresa. 1980. The communication of morality: Cooperation and
commitment in a food cooperative. Unpublished dissertation, Colum-
bia University.
70 / William Labov
Labov, William. 1966. The social stratification of English in New York

City. Washington, D.C.: Center for Applied Linguistics.
Labov, William. 1972. Language in the inner city. Philadelphia: Uni-
versity of Pennsylvania Press.
Labov, William. 1984. Field methods of the project on linguistic change
and variation. In: Language in use. Edited by J. Baugh and 3.
Sherzer. Englewood Cliffs, N.J.: Prentice-Hall.
Labov, W., A. Bower, E. Dayton, D. Hindle, T. Kroch, M. Lennig, and D.
Schiffrin. 1980. The social determinants of sound change in Phila-
delphia. Philadelphia: U.S. Regional Survey.
Labov, William, and David Fanshel. 1977. Therapeutic discourse. New
York: Academic Press.
Labov, William, and Joshua Waletzky. 1967. Narrative analysis. In:
Essays on the verbal and visual arts. Edited by J. Helm. Seattle:
University of Washington Press. 12-44.
Osgood, C. E., G. J. Suci, and P. H. Tannenbaum. 1957. The measure-
ment of meaning. Urbana: University of Illinois Press.
Sapir, Edward. 1930. Totality. Language Monographs 6.
Thibault, 3ohanne. 1977. Les marqueurs d'intensite' en fran?ais Mont-
re*alais. Master's essay, University of Montreal.
Wolfram, W., and D. Christian. 1976. Appalachian speech. Washington,
D.C.: Center for Applied Linguistics.
ON SEMANTIC AND PRAGMATIC COMPETENCE
Michael L. Geis
The Ohio State University
0. Introduction. The manufacturer of an antiseptic product once ran
a television advertisement saying that it
(1) ...gently penetrates skin injuries to kill germs that can cause
infection, helps speed healing (CBS, 3uly 18, 1978, 10:55
a.m.) 1
a.m.)
2
This is a rather common kind of claim in contemporary advertising. It
is generic in form and seems to make a rather general claim about the
efficacy of this product—call it 'P1—but it contains two so-called 'weasel
words' (Stevens 1978), can and help, that are said to weaken claims
substantially. Stevens (p. 75) claims, for instance, that 'by adding that
one little 1word, help, in front, we can use the strongest language possible
afterward , and that claims employing can are (p. 79) 'indicative of an
ideal situation.' Even the phrase penetrates skin injuries to kill germs
would be considered weak from a literalist perspective, for it would be
said to be true even if P only sometimes penetrates skin injuries and only
sometimes kills germs. In short, (1) would be considered as true even if
P almost never works.
Interestingly, language scholars have taken a rather different view of
generic sentences like (1), according to which they make relatively
strong claims. Jackendoff (1972:309), for instance, has said that 'seman-
tically, generic sentences resemble sentences with a universal quan-
tifier,1 claiming that (2) 'can be paraphrased by' (3).
(2) A rhinoceros eats small snakes.
(3) Every rhinoceros eats small snakes.
In his introductory logic text, Thomason (1970:158) notes that generic
sentences are 'very tricky and treacherous,' and then says (p. 170) that a
generic sentence like (k) means that 'most carrots are good to eat, or
71
72 / Michael L. Geis
that almost all carrots are good to eat 1 (his emphasis), or that 'as a rule,
any carrot is good to eat 1 (his emphasis).
(4) Carrots are good to eat.

Thus, while Stevens would say that (1) could be true if P almost never
kills infection-causing germs, 3ackendoff and Thomason would say that
(1) cannot be true unless P almost always succeeds in killing infection-
causing germs. More recently, Carlson (1979) has provided an explicit,
model-theoretic semantic analysis of generics which accords well with
the views of Thomason and Jackendoff.
A language must be a peculiar thing indeed to succor such radically
different theories of generics. In fact, of course, the problem lies not
with English but with differences in theories of sentence meaning. If
one takes the position that the truth conditions of sentences must be
stated without reference to context (except for indexical expressions, of
course), then the literalist view of the meaning of (1) must be correct.
In this view, no generic sentence is stronger than the weakest generic
sentence, for all must have the same truth conditions. However, this
literalist theory of sentence meaning cannot be correct. As Carlson
(1979) notes, the truth conditions of pairs like (1) and (5) must be rather
different.
(5) John Jones executes criminals for the state of California.
We might say that (5) would be true even if John Jones has done his job
only once or twice, but we would not say that (1) is true if P has killed
only one or two infection-causing germs. Thus, the assumption that the
truth-conditions of generic sentences can be defined independently of
context is clearly wrong.
The narrow truth-conditional approach to a sentence like (1) is wrong
in a still deeper way. Sentence (1) is uttered by way of making an offer,
an offer to sell a product which is intended by the advertiser to be used
by the consumer to satisfy two consumer needs: chemical protection
from infection and as speedy a recovery as is possible. Surely, this also
is important to an understanding of a sentence like (1).
Most linguists would say, I think, that the truth-conditional meaning of
(1), and the fact that it is being used to make an offer, both have a
bearing on how (1) is interpreted. The Standard Theory (see Gazdar
1979) is, in fact, that the full meaning of an utterance is a function of
the truth conditions of the sentence uttered and such pragmatic consid-
erations as speech act felicity conditions (Austin 1965, Searle 1969),
rules of cooperative conversation (Grice 1975), politeness conventions
(R. Lakoff 1977), widely accepted factual premises (Boe*r and Lycan
1976) and context, etc. On this view, the fact that (1) is uttered by
way of making an offer is crucial to its interpretation.
Although this Standard Theory of sentence meaning is very widely
accepted today, at least in principle, by both pragmaticists and seman-
ticists (see Barwise and Perry 1983), I am reliably informed that it has
very little currency in legal circles. Lawyers tend to take the same
literalist view of meaning as did Stevens. They, in fact, are surely the
source of it. If this Standard Theory of sentence meaning is the correct
On semantic and pragmatic competence / 73
one, then we should get the story out, for if we do not, the literalist
perspective of sentence meaning will continue to dominate in legal
contexts, especially in regard to the issue of what advertising claims
actually do claim.
In this paper, I propose to do this by arguing that the literalist theory
of sentence meaning fails, not just because it makes incorrect predic-
tions about how people will interpret sentences like (1), which it does,
but, more fundamentally, because it assumes something false about
human linguistic competence, namely, that logically untutored people
have some sort of substantive logical capacity. So I have a practical
goal, but I have theoretical and methodological aims as well. I argue
that truth-conditional semantics is to some degree based on this same
false assumption about human linguistic competence, and I then discuss
the implications of this fact for an empirically substantive theory of
sentence understanding by logically naive speakers.
1. Validity judgments. Logicians, philosophers of language, and

linguists would agree with Davidson (1967), I think, that an adequate
description of the meaning of sentence (6) should account for the fact
that utterance of (6) commits the speaker to the truth of (7a-c).
(6) John buttered the toast with a knife in the bathroom at

midnight.
(7a) John did something at midnight.
(7b) John used a knife to butter his toast.
(7c) John was in the bathroom at midnight.
During the heyday of generative semantics, many linguists and some

philosophers of language took the line, made most explicit by Harman
(1972), that the deep structure or semantic representation of a sentence
should be its logical form, that it should be such that it accounts for
what inferences are and are not entailed by the sentence. On this view,
a correct description of the meaning of (6) should account for the fact
that it commits the speaker to the truth of (7a-c). I would heartily
agree with this particular claim. However, I would like to question
whether or not, in general, speakers of natural languages have the se-
mantic competence presumed by this truth-conditional approach to
sentence meaning. I believe and argue that the answer to this is 'no1, and
that we must to some degree rethink the intellectual foundations of
semantics and pragmatics as well as certain of our methodological
assumptions in these areas. I also draw out some practical consequences
of my views for a theory of truth in advertising and for teaching reason-
ing in the schools.
Doing truth-conditional semantics presumes the capacity somehow to
discriminate valid from invalid inferences. However, making validity
judgments is sometimes anything but a trivial intellectual feat. As
Partee (1982) has noted in connection with belief-sentences, it is some-
times very difficult for even the most sophisticated speakers to make
sound synonymy judgments. And, while most would readily assent to the
view that (6) entails (7a-c), do we as readily assent that (8a) and (8b) are
synonymous (i.e. entail each other), or that (9a) and (9b) are synonymous,
or that (10a) entails (10b)?
7* / Michael L. Geis
(8a) If you open the refrigerator, it won't explode.

(8b) If you open the refrigerator, then it won't explode.
(9a) I will leave if you don't leave.
(9b) I will leave unless you leave.
(10a) If Boris had gone to the party, Olga would have gone.
(10b) If Olga had not gone to the party, Boris would not have gone.
I hope that we would not assent to these judgments, for they are quite
wrong.
It should be clear that judgments of validity are not always quickly
and easily made. They often require substantial imagination (to con-
struct an illuminating possible world or situation to test the sentences
against) as well as reasoning (to sort out whether the pertinent sentences
or propositions are true or are false of these possible worlds or situa-
tions). Now, if the validity judgments of semanticians are to serve as
the data for descriptions of the semantic competence of ordinary speak-
ers of language, then we must provide some explanation of how ordinary
speakers might acquire this semantic competence. The question is: can
native speakers of the many languages of the world be assumed to have a
substantive, validity-based reasoning capacity (i.eâ mental logic) which
they employ in learning and using their languages?
In 1982, Johnson-Laird published a very important paper called 'Think-
ing is a skill' (Johnson-Laird 1982b). In it he considered the question
whether or not we have a mental logic and gave three arguments against
saying we do. The first is that people, to a very striking degree, are
unable to perform well on certain 'easy' reasoning tasks. Suppose you
have four cards put in front of you on the faces of which you see the
symbols 'E', 'K', '4', and 7', and you are told to test the following hypo-
thesis by turning over as few cards as possible:
(11) If a card has a vowel on one side, then it has an even number
on the other.
Interestingly, in several experiments involving 128 subjects, only five got
the correct answer.
Johnson-Laird's second argument against the 'doctrine of mental logic'
is that people are rather imperfect at syllogistic reasoning tasks. Given
sentences (12) and (13), some 'highly intelligent' university students were
asked to identify what conclusions (from those in (14)) could validly be
drawn from these premises.
(12) All of the beekeepers are artists.

(13) None of the chemists are beekeepers.
None of the chemists are artists.
(14b) None of the artists are chemists.
(14c) Some of the chemists are not artists.
(14d) Some of the artists are not chemists.
(14e) None of the above.
g
None got it right.
More recently, Rips (1983) has reported on an interesting research
program in which people's deductive abilities are tested and evaluated
with respect to a computer model of propositional reasoning employing a
natural deduction system. Rips reports an overall success rate by sub-
jects in making validity determinations of about 50%, with actual suc-
cess rates ranging from about 17% to 92%. So, given (15L subjects were
asked to say whether or not (16) could validly by deduced.
(15) If Mary goes to Boston, then Sue will go to New York.
(16) If Mary goes to Boston and Sandy goes to Chicago, then Sue
will go to New York.
In this case, 58% got the correct answer.
Rips appears to assume that we have a perfect mental logic but that
we are not very reliable in using it. He assumes that errors result from
the failure to retrieve inference rules, the failure to recognize the
applicability of a rule, or the failure to apply a rule correctly. Johnson-
Laird notes that others (Henle 1962) who believe in the doctrine of
mental logic are inclined to account for errors in reasoning by saying
that they result from misunderstanding or forgetting premises or bring-
ing in irrelevant or unwarranted information, or the like. One clearly
pays a very high empirical price for hypothesizing that we do have a
mental logic.
Saying that we have a substantive validity-based reasoning capacity,
but that we are very often unable to use it properly, is empirically
empty unless an empirically motivated, predictive theory of how it is
that we go wrong is provided, for instead of giving us reliable empir-
ical evidence supporting the doctrine of mental logic, it offers a blanket
excuse for every past or future failure to demonstrate the existence of
this capacity. There is an alternative—that we have a nonlogical system
of reasoning (i.e. one without a substantive validity detection capacity),
perhaps along the lines of Johnson-Laird (1982a, 1982b), and that we are
relatively good at using this. A priori, I believe this more positive
approach to reasoning to be superior, for it offers the possibility of a
more principled account of reasoning than does the mental logic theory.
Johnson-Laird's third argument against the doctrine of mental logic is
that it is very hard to understand how we could have come by it in the
natural state. Could it be learned from experience? Given the manifest
inability of humans to reason reliably in experimental studies, how can
we assume that children will see enough positive examples of valid
reasoning from their parents and others to extract this abstract mathe-
matical knowledge from experience? There is, of course, the logical
possibility that this wondrous 'gift', a gift we don't seem to be able to
use very well, is innate. However, as Johnson-Laird has said (1982b: 10),
'positive arguments that the ability to reason is innate are as hard to
come by as positive arguments that it arises from divine intervention.'
Johnson-Laird concludes that reasoning is learned and is a skill and
that some people do it better than others. This seems to square rather
well with the facts. In Geis (1982), I took the line, based in part on
earlier work by Wason and Johnson-Laird (1972), that if we do not con-
trol the concept of validity as reasoners, then it would be unreasonable
to assume that we have control of this concept as language learners and
language users. The question arises, for instance, as to how children

might come to be able to make the sorts of semantic distinctions that
are discovered by linguists through our validity judgments. Some lin-
guists will, of course, claim that we have an innate capacity to make
just the sorts of distinctions that are required to develop a full-blown
truth-conditional semantic competence in a language. Moreover, some
of these linguists might say, to avoid the thrust of Johnson-Laird's
arguments, that this validity detection capacity is highly dedicated to
language. Fodor's (1983) theory of mental faculties would appear to lend
credibility to the thesis that this or that ability could be dedicated
specifically to language and thus be unavailable in other mental do-
mains. However, before we say that we have an innate validity detec-
tion device highly dedicated to language, which vanishes whenever we
test for it, we must first show that we have a validity detection device
in the first place. I argue that we have no such device.
The question is: do speakers have a conscious or unconscious, explicit
or implicit, capacity to somehow distinguish valid from invalid infer-
ences? Interestingly, as we have noted, even logically sophisticated
speakers sometimes have difficulty making validity judgments. What
can we say about the capacities of logically untutored speakers? Fox
and I (cf. Geis and Fox 1984) have run an experiment on a large number
of subjects in which we attempted to determine people's abilities to
make validity judgments about pairs of sentences. What we have learned
is that significant percentages of people are unable to distinguish valid
from invalid inferences when the latter are sustained by speech act
felicity conditions and generalized conversational implicatures. Sixty
percent of our subjects voted (cf. Geis and Zwicky 1971), for instance,
that (17) warrants (18).
(17) I'll give you 5 bucks if you mow the lawn.

(18) I won't give you 5 bucks if you don't mow the lawn.
Of course, they are wrong truth conditionally, but they would not be
wrong conversationally (cf. Bo6r and Lycan 1973) in many of the cir-
cumstances in which (17) might be uttered. In a stunning demonstration
of the strength of generic claims, we found that 58% of our subjects
incorrectly said (19a) entails (19b).
(19a) ABC paint contains water-repelling oils.

(19b) ABC paint doesn't absorb water.
Sixty-four percent of our subjects incorrectly claimed that (20a) entails
(20b).
(20a) ABC filters remove bacteria from your drinking water.

(20b) If you use ABC filters, your drinking water should be free of
bacteria.
And 82% of our subjects incorrectly said (21a) entails (21b).
(21a) John can solve algebra problems.
(21b) If 3ohn is given an algebra problem, he'll probably be able to

solve it.
It is clear from this that generic claims are construed as very strong
claims and (contra Stevens) can does not weaken claims. Indeed, it
seems to be a kind of intensifier.
These results obviously do not show that people have no truth-condi-
tional capacities. What they show is that even in an explicitly truth-
conditional task, people have great difficulty distinguishing between
entailments and inferences that are warranted by general pragmatic
considerations. In naturalistic circumstances (e.g. watching television),
they will be all the more likely to draw these sorts of inferences, of
course.
The fact that conscious validity judgments are unreliable doesn't prove
that we do not have a validity detection device. Much of language
comprehension is the result of some sort of automatic unconscious
activity. Perhaps our validity detection capacity operates at an un-
conscious level. Certainly, the recognition of the validity of some
inferences—recall (6) and (7)—is virtually automatic. Against the view
that we exhibit unconscious control of the validity-invalidity distinction
in language comprehension are three facts. First, as Morgan (1978) has
ably demonstrated, some nonlogical inferences are also automatic.
Morgan would say that recognition that (22) is a request requires no
calculation.
(22) Could you take out the garbage?
Second, there is much evidence that people go for the 'gist' of a sentence
(cf. Loosen 1981 and references therein) rather than for its literal mean-
ing in comprehension. Third, there is much evidence (cf. Harris, Dubit-
sky, and Bruno 1983 and references therein) that people don't always
distinguish assertions from inferences (including invalid inferences) in
comprehension tasks. In an experiment by Brewer (1977), for instance,
more people, on being exposed to (23a), recalled the inference (23b) than
recalled (23a) itself.
(23a) The safecracker put a match to the fuse.

(23b) The safecracker lit the fuse.
Thus, we may safely conclude that we do not exhibit a conscious or
unconscious control of the validity-invalidity distinction in language
comprehension.
Might we exhibit an otherwise unconfirmed control of the validity-
invalidity distinction in speech production, as opposed to speech percep-
tion? I think not. Speech production, viewed truth-conditionally, is
extraordinarily simpler than speech comprehension. The advertising
copywriter who constructs a sentence like (1) isn't necessarily lin-
guistically cleverer than the consumer who might be fooled by it. The
reason the copywriter may seem more clever is that going from facts to
sentences is easy. But imagining all of the ways in which this kind of
claim might deceive takes real imagination. Thus, I can see no reason to
believe that speech production presumes semantically more sophis-

ticated abilities than does speech comprehension.
Finally, let's consider the question of origins. The supposition that we
might be able to acquire a full-blown (albeit tacit) notion of validity as
an automatic concomitant of language acquisition is very hard to be-
lieve. What we learn when we learn language is that we must become
inference drawers, that we must draw inferences of all sorts—inferences
warranted by what words seem to mean, by felicity conditions, by rules
of conversation, by background knowledge, by politeness conventions, by
the multifarious aspects of context, and by extraordinarily varied com-
binations of all of the above. Many of these inferences that we draw in
ordinary life are not valid, of course, and no one is there to discriminate
those that are valid from those that aren't in any usefully systematic
way, either as we learn or as we use our languages. Thus, it is very
difficult to see how we could acquire a validity-invalidity distinction as
a by-product of acquiring language.
2. Methodological implications. There is solid empirical evidence

that logically naive people do not control the validity-invalidity distinc-
tion consciously, and there is evidence that this distinction does not play
a role in ordinary speech comprehension. Theoreticians regularly re-
spond at this point by noting that we linguists often assume the exist-
ence of tacit linguistic competences that speakers don't explicitly know
about. Speakers may not reliably identify what is and is not a noun
phrase, but this does not mean that speakers do not have a tacit know-
ledge of what is and is not a noun phrase. The problem is that this
putative parallel between syntax and semantics is quite specious.
The primary data for syntax come from grammaticality judgments,
many of which come from syntacticians themselves rather than from
naive informants. This is warranted, at least partially, by the following
facts about syntactic judgments:
(24) SYNTAX JUDGMENTS
(24a) Grammaticality judgments are normally automatic in char-
acter.
(24b) Grammaticality judgments presume no special skills beyond
the capacity to distinguish between two types of automatic
judgments—grammaticality judgments and meaningfulness
judgments.
(24c) We linguists do not get better at grammaticality judgments
to any significant degree over time.
(24d) Our grammaticality judgments are not substantially more
reliable than are those of syntactically naive speakers.
Indeed, field workers routinely rely on such speakers exclu-
sively for data.
Compare these properties with those of validity judgments.

(25) SEMANTIC JUDGMENTS
(25a) Validity judgments are often not automatic, but, instead,
require reasoning—sometimes very sophisticated reason-
ing. 1 1
(25b) Validity judgments presume analytical, essentially logical

skills, e.g. the skill to construct model-theoretic counter-
examples.
(25c) We obviously get better at making validity judgments with
experience.
(25d) We linguists are very much better at making validity judg-
ments than are logically naive speakers.
Now, facts (24a-d) tend to justify empirically our use of our own intui-
tions in research in syntax, at least to some degree; however, facts (25a-
d) render our use of our validity judgments empirically quite suspect in
semantic studies. Semanticists who want to say that they are studying
the semantic competence of speakers of the languages of the world must
come up with some sort of empirical justification for their data collec-
tion practices.
In my view, sentence-interpretation is partly automatic (see Fodor
1983) and partly a skill—simply listening or listening with a critical e a r -
depending on what sort of linguistic activity we are engaged in. And I
believe that our mental lexical representations are formed for use in the
automatic mode and, thus, consist of truth conditions or meaning postu-
lates that stipulate in a relatively narrow way the 'easy' entailments
these words sustain. Thus we will acquire a truth condition (or equiva-
lent meaning postulate) for jtf saying that 'If P, then Q' is true if P is true
and Q is true, and false if P is true and Q is false, but not the more
elaborate one according to which 'If P, then Q' is true if and only if P is
true and Q is true or P is false.
Interestingly, semanticists have the feeling that when they discover
some novel aspect of the truth conditions of some class of sentences
they are literally discovering something, something that is already in
their heads. I believe that it would be a mistake to take this impression
very seriously. I would argue that truth-conditional semantics is a study
of the truth-conditional limits of languages, and, thus, that semanticists
are not necessarily discovering what is already in their heads, but are
discovering what would be in their heads were they fully to develop their
semantic competence. An analogy will help make this point clear. If a
photographer unknowingly underexposes a piece of negative film and
does not compensate by overdeveloping the negative, a thin negative will
result, with perhaps only the scene's highlights being easily recognized.
This unprintable negative can be enhanced through chromium intensifi-
cation, a process whereby chromium is deposited on existing grains of
silver. The result is sometimes printable. Note that this enhanced
image was not there before intensification, but the resulting image was
constrained by what was there. In such a way do the semanticists dis-
cover—I would say develop (to alter the photographic metaphor slightly)
—the truth-conditional limits of their language knowledge.
Some will say that I have gone wrong because I have not made a
sentence-utterance distinction. As I've noted, the standard view is that
utterance meaning is a function of truth-conditional sentence meaning
and contextual pragmatic considerations—as if sentence meaning bears
the same kind of relationship to utterance meaning as phonemes do to
allophones. I believe that this view is fundamentally mistaken. Gazdar
(1979) notes that sometimes truth conditions cannot be defined without
reference to pragmatic considerations, that semantics is, in fact, not

autonomous of pragmatics. Let us reconsider how this might apply to a
sentence like (26), and thus, to (1).
(26) P kills infection-causing germs.
I believe we would say that (26) could happily occur as an assertion in a

Food and Drug Administration study of the efficacy of various anti-
septics if P proved to kill only five percent of infection-causing germs.
But, when a claim like (26) occurs in an advertisement, as in (1), it is
being used as supporting argument for an offer to sell P to those con-
sumers who need protection from infection-causing germs. In this case,
the pertinent truth condition for (26), and thus (1), is that it is true if
and only if it kills a sufficiency of infection-causing germs in conditions
of normal use to protect consumers from infection. It must kill
enough germs to satisfy the sincerity condition on offers (see Bach and
Harnish 1982) that the speaker have a warranted belief that his action,
service, or product, etc. will satisfy the hearer need or desire that
occasions the offer. In short, the truth-conditions of (26) depend cru-
cially on the speech act it is used to perform—a simple assertion or an
assertion used to entice a consumer to accept a commercial offer.
3. Practical implications. In my view, standards of truth in adver-

tising should be based on actual, rather than ideal speaker linguistic
abilities and, thus, I would argue that the narrow legalistic truth-condi-
tional theory of sentence meaning must be abandoned by those entrusted
to protect the public interest in favor of a holistic, pragmatically based
view. I believe also that what I have said forces us to the view that a
genuinely empirical theory of language understanding must be based on
experimentally sound data gathered from theoretically (including logi-
cally) naive speakers. It will not do to use ourselves as informants: we
know too much about logic and too little about human behavior. I be-
lieve, for these reasons, that pragmatics itself must also be based on
naive informant judgments, not ours. Pragmatic judgments are not
'right' or 'wrong' in the sense in which validity judgments are.
What, though, of semantic studies in which data are derived from
introspective validity judgments? Clearly, whether or not one sentence
entails another is an empirical matter, and semanticists are better than
most at making such judgments. I said earlier that truth-conditional
semantics is a study of the truth-conditional limits of our language. It
is, I believe, the study of the principles of valid reasoning with sen-
tences. It is my impression that much, if not most, serious reasoning is
done in connection with sentences, whether these are embedded in
advertising, in a political debate, in newspaper reports, in textbooks, or
in scholarly papers, etc., and I believe that people's ability to reason
could be significantly improved were we to provide students with natural
language semantic training, perhaps coupled with grammatical training.
People need to be able to reason successfully with generic sentences,
with conditional sentences (including counterfactuals), with nominaliza-
tions, and so on. Let us not, however, take this kind of truth-conditional
semantic analysis of English as an empirically correct theory of what is
already in the semantic part of the average speaker's head, or as an
empirically correct theory of how utterances should be understood in
conversation or in any other naturalistic context, e.g. watching a tele-
vision advertisement.
Notes
I am indebted to my colleagues Robert Fox, Wayne Cowart, David
Dowty, Robert Kraut, Brian Joseph, John Nerbonne, George Schumm,
and Arnold Zwicky for their comments on the views presented here.
1. I do not mention the product name, for I don't want what I say
about the language of (1) to be taken as a comment on the quality of this
product.
2. There are 1984 Listerine advertisements that use very similar
language, saying (if my short-term memory can be relied on) that the
product 'kills the germs that can cause plaque' and 'kills the germs that
can cause bad breath.' Right Guard advertisements of late February
1984 assert (same qualification) that this product 'kills the germs that
can cause perspiration odor.' I suspect lawyers tell advertisers to put
can in claims like these to render them too we$k to be assailed, which is
Stevens' (1978) view. This 'theory' of can is very far from the truth. See
note 14.
3. Gazdar (1979:2) says 'PRAGMATICS = MEANING - TRUTH CON-
DITIONS,' but notes in his last chapter that truth-conditions sometimes
appear not to be statable without reference to pragmatic conditions and,
thus, that semantics may not be autonomous of pragmatics. I argue for
this point of view in connection with sentence (1).
4. Though generative semantics is no longer with us, linguists con-
tinue to presume that it is sensible to talk about a linguistic level of
logical form (see Gueron and May 1984, Bosque and Moreno 1984, or
Solan 1984, for example). Just how substantive a logic is presumed in
these cases is not wholly clear, to say the least.
5. As Davis (MS) has argued, (8a) is true of normal refrigerators, but
(8b) would appear to be true only of refrigerators rigged to explode
unless opened. In Geis (1973), I showed (using Lycan's 1984 paraphrases
instead of my original ones) that (9a) and (9b) differ, as do (i) and (ii).
(i) I will leave in any situation in which you don't leave,
(ii) I will leave in any situation other than one in which you don't
leave.
Lewis (1973:35) argues against the view that (10a) entails (10b), reason-
ing thus: 'Suppose that Boris wanted to go, but stayed away solely in
order to avoid Olga, so the conclusion is false; but Olga would have gone
all the more willingly if Boris had been there, so the premise is true.'
6. Janet Dean Fodor (1982) notes that contemporary semantic re-
search presumes the existence of a substantive logical capacity.
7. The correct choices are 'E' and '7,' the latter being the difficult
choice.
8. The correct answer is (14d).
9. I have replaced Rips' proposition letters by sentences to illustrate
the sort of thing subjects would see. The correct answer is 'yes.'
10. Rips (1983) is trying to come up with such a theory, and I wish him
luck.
11. Recall Johnson-Laird's claim that 'thinking is a skill.'
12. I believe that the troublesome sentence (14) provides very nice
support for the view that language comprehension is automatic and
reflexive, as Fodor (1983) has argued. In Geis and Fox (1984), subjects
noted (incorrectly), by a two to one vote, that (i) does not entail (ii).
(i) John won't leave unless Bill doesn't leave,

(ii) If Bill leaves, John won't leave.
Interestingly, in a pretest of 19 subjects, one wrote stay over won't

leave, and then decided correctly that (i) does entail (ii). This extra-
ordinarily simple move doesn't seem to be available in automatic sen-
tence processing.
13. See Fodor, Fodor, and Garrett (1975) for a defense of the use of
meaning postulates in semantics.
14. The presence of can in claim (1) (contra Stevens) does not weaken
this claim. Recall the discussion of (21a). Note further that if can were
to signify mere 'possibility', the offer would be an infelicitous (because
insincere) offer, for what is required by the consumer is real protection
from infection. Since the offer appears to be a sincere offer, can cannot
signify mere possibility.
References
Austin, J. L. 1965. How to do things with words. Oxford: Oxford
University Press.
Bach, Kent, and Robert M. Harnish. 1982. Linguistic communication
and speech acts. Cambridge, Mass.: The MIT Press.
Barwise, Jon, and John Perry. 1983. Situations and attitudes. Cam-
bridge, Mass.: The MIT Press.
BoeY, S. E., and W. G. Lycan. 1973. Invited inferences and other
unwelcome guests. Papers in Linguistics 6.3-4.
Boer, S. E., and W. G. Lycan. 1976. The myth of semantic presupposi-
tion. Mimeograph. Bloomington: Indiana University Linguistics Club.
Bosque, Ignacio, and Juan-Carlos Moreno. 1984. A condition on quanti-
fiers in logical form. Linguistic Inquiry 15.164-67.
Brewer, W. 1977. Memory for the pragmatic implications of sen-
tences. Memory and Cognition 6.673-78.
Carlson, Greg N. 1979. Generics and atemporal when. Linguistics and
Philosophy 3.49-98.
Davidson, Donald. 1967. The logical form of action sentences. In: The
logic of decision and action. Edited by Nicholas Rescher. Pitts-
burgh: University of Pittsburgh Press. 81-95.
Davis, Wayne. Weak and strong conditionals. MS.
Fodor, Janet Dean. 1982. The mental representations of quantifiers.
In: Processes, beliefs, and questions. Edited by Stanley Peters and
Esa Saarinen. Dordrecht: D. Reidel. 129-64.
Fodor, Jerry A. 1983. Modularity of mind: An essay on faculty psy-
chology. Cambridge, Mass.: The MIT Press.
Fodor, J. D., J. A. Fodor, and M. Garrett. 1975. The psychological
unreality of semantic representations. Linguistic Inquiry 6.515-31.
Gazdar, Gerald. 1979. Pragmatics: Implicature, presupposition, and
logical form. New York: Academic Press.
Geis, Michael L. and Arnold M. Zwicky. 1971. On invited inferences.
Linguistic Inquiry 2.561-66.
Geis, Michael L. 1973. If and unless. In: Issues in linguistics: Papers in
honor of Henry and Renee Kahane. Edited by B. B. Kachru et al.
Urbana: University of Illinois. 231-53.
Geis, Michael L. 1982. The language of television advertising. New
Geis, Michael L., and Robert A. Fox. 1984. Mental logic and sentence
comprehension. To appear in: Proceedings of the 1984 Chicago
Linguistic Society Conference.
Grice, H. P. 1975. Logic and conversation. In: Syntax and semantics,
Volume 3: Speech acts. Edited by P. Cole and J. L. Morgan. New
York: Academic Press. 41-58.
Gueron, Jacqueline, and Robert May. 1984. Extraposition and logical
form. Linguistic Inquiry 15.1-31.
Harman, Gilbert. 1972. Logical form. Foundations of Language 9.36-
65.
Harris, R. J., T. M. Dubitsky, and Kristin Joe Bruno. 1983. Studies of
misleading advertising. In: Information processing research in adver-
tising. Edited by R. J. Harris. Hillsdale, N. J.: Lawrence Erlbaum
Associates. 241-62.
Henle, M. 1962. On the relation between logic and thinking. Psycho-
logical Review 69.366-78.
Jackendoff, Ray S. 1972. Semantic interpretation in generative gram-
mar. Cambridge, Mass.: MIT Press.
Johnson-Laird, P. N. 1982a. Formal semantics and psychology. In:
Processes, beliefs, and questions. Edited by Stanley Peters and Esa
Saarinen. Dordrecht: D. Reidel. 1-68.
Johnson-Laird, P. N. 1982b. Thinking is a skill (Ninth Barlett Memorial
Lecture). Quarterly Journal of Experimental Psychology 34A.1-29.
Lakoff, R. 1977. What you can do with words: Politeness, pragmatics
and performatives. In: Proceedings of the Texas Conference on
Performatives, Presuppositions and Implicatures. Edited by A. Rogers
et al. Arlington, Va.: Center for Applied Linguistics. 79-105.
Lewis, David. 1973. Counterfactuals. Cambridge, Mass.: Harvard
University Press.
Loosen, F. 1981. Memory for the gist of sentences. Journal of Psycho-
linguistics 10.17-25.
Lycan, William G. 1984. A syntactically motivated theory of condi-
tionals. Midwest Studies in Philosophy.
Morgan, J. L. 1978. Two types of convention in indirect speech acts.
In: Syntax and semantics, Volume 9: Pragmatics. New York: Aca-
demic Press.
Partee, Barbara Hall. 1982. Belief-sentences and the limits of seman-
tics. In: Processes, beliefs, and questions. Edited by Stanley Peters
and Esa Saarinen. Dordrecht: D. Reidel. 87-106.
Searle, J. R. Speech acts: An essay in the philosophy of language.
London: Cambridge University Press.
8* / Michael L. Geis
Solan, Lawrence. 1984. Focus and levels of representation. Linguistic

Inquiry 15.174-78.
Stevens, Paul. 1978. Weasel words: God's little helpers. In: Language
awareness. Edited by Paul Escholz et al. New York: St. Martin's
Press. 74-85.
Rips, Lance 3. 1983. Cognitive processes in propositional reasoning.
Psychological Review 90.38-71.
Thomason, Richmond H. 1970. Symbolic logic: An introduction. Lon-
don: Macmillan.
•SUBORDINATION1 IN FORMAL AND INFORMAL DISCOURSE
Sandra A. Thompson
University of California, Los Angeles
Important inroads have recently been made into understanding some of
the differences among various styles, registers, and modes of communi-
cation with respect to the way in which clauses in discourse are com-
bined. Studies on this subject include Tannen (1982a, 1982b), Ochs
(1979), Chafe (1982, 1984, to appear), Clancy (1982), Beaman (to appear),
and Mithun (1984). My paper is intended as a further contribution to this
discussion. I would like to look at some differences between formal and
informal discourse in terms of the way people have used the term 'sub-
ordination1, and to talk about the concept of 'subordination' itself.
My discussion takes as its starting point recent work of Chafe, parti-
cularly (1982) and (1984). Chafe (1982) discussed differences between
spoken and written languages in terms of a number of parameters, such
as fragmentation and integration. By 'integration' he meant 'the packing
of more information into an idea unit than the rapid pace of spoken
language would normally allow' (p. 39). As measures of 'integration', he
used such constructions as nominalizations, participles, complements,
relative clauses, attributive adjectives, and conjoined phrases, and he
found that formal written language had a larger proportion of all these
devices than informal spoken language.
Chafe (1984) has suggested that written and spoken language differ in
terms of the percentage of adverbial subordinate clauses they contain.
Mithun (1984) has also shown that languages differ in terms of the per-
centage of 'subordinate' clauses they contain in informal spoken dis-
course analysis samples, with English topping the list at roughly 34%,
while Mohawk and Kathlamet, two indigenous American languages, had
only about 6%, and Gungwinggu, an indigenous language of Australia,
showed only about 2%.
I think that this work is very important, and that it strongly suggests
that both within one language as well as across languages, there is varia-
tion in the way in which language users make use of formally marked
dependencies among clauses in communication. In what follows, I would
like to add to this discussion by comparing formal and informal English
86 / Sandra A. Thompson
discourse with respect to what have been called 'subordinate' clauses.

But first, I want to talk about the terms 'subordination' and 'subordinate
clause' themselves.
The term 'subordination' is in quotes in my title because I think it is a
misleading term. Its use encourages linguists and descriptive grammar-
ians to think that at least some languages have a category of clause
types which can be called 'subordinate'. In fact, Haiman and I (Haiman
and Thompson 1984) have recently made a plea for abandoning the notion
of 'subordination' in grammatical descriptions, at least for the time
being, and talking instead about various relationships which pairs of
adjacent clauses might have with each other, properties such as whether
they share a subject, whether one of them is reduced or not, whether one
of them is within the scope of the other, and so on. In this way, we felt,
more progress could be made toward understanding how clauses actually
combine in discourse than if we simply assume that everybody under-
stands each other when we use the term 'subordinate clause'.
We suggested that experience with the linguistic literature in recent
years shows that linguists have typically either taken the term 'sub-
ordination' as a primitive requiring no definition, or attempted to estab-
lish a set of criteria according to which a 'subordinate clause' can be
unmistakably identified. Both of these approaches to 'subordination'
have had unfortunate consequences.
If the term is taken as a primitive, then it is usually not made clear
just what clauses or clause types within a given language it is intended
to cover. At best, it can be assumed that it is meant to include all
clauses which are 'nonmain'. But then, we know of no attempt in the
literature to define 'main'.
Attempts to establish criteria, however, do not seem to have fared
much better. In fact, the traditional criteria for 'subordination'—includ-
ing dependence, reduction, backgrounding, and preposability, among
others—turn out to be at best ex post facto rationalizations of our own
(Western educated) 'intuitions', which means that they are all completely
circular. Moreover, these criteria are often inconsistent or language
specific. The reason for this, I think, is that 'subordinate clause' is not a
grammatical category at all. That is, there does not seem to be a single
function or even a group of functions that we can think of this 'category'
as having been designed, as it were, to serve. So the term 'subordination'
seems to be at best a negative term which lumps together all deviations
from some 'main clause' norm, which means that it treats as unified a
set of facts which we think is not a single phenomenon. For these rea-
sons, we have found it more fruitful to tease it apart into its component
parts and try to determine what are the discourse motivations that
underlie each of these components. That is, there seem to be a number
of properties that are involved in what people have heretofore called
'subordination', and a better understanding of clause-combining phe-
nomena in discourse can be achieved if we consider each of these sepa-
rately, and give up the idea of a category of 'subordinate clause' en-
tirely.
Chafe (1984) and Mithun (1984) seem to have used the term 'subor-
dinate' for something like 'dependent'; that is, for them, as for many
other linguists, a clause which is morphologically marked so that it
cannot stand by itself can be considered as 'subordinate'. I think that
'Subordination1 in formal and informal discourse / 87
this subcategory of 'subordinate' needs to be broken down into its com-

ponent parts too, since there seem to be different types of 'dependency',
doing different types of discourse work. My suspicion is that 'depend-
ence' may well need to be defined separately for each language, but, to
begin with, we can list a set of properties which may be useful for a
given language.
In the remainder of this paper, I would like to describe the research
that I have been doing on English in this area and to suggest that it has
promise for understanding differences among types of discourse in one
language as well as among languages. I emphasize that this is a descrip-
tion of ongoing research, which means that what I say must be taken
very much in the spirit of a progress report.
I want to make an important distinction between dependent clause
types which represent organizational options available to the language
user, and dependent clause types having to do with facts about the
grammar of English. This means that I have restricted my study to
clauses which do not play a grammatical role in constituency with some
noun, some verb, or some preposition. That is, I have not been con-
cerned with relative clauses, complement clauses, or clauses which are
the objects of prepositions. An example of each of these is given in (1).
(1) Not included in this study: clauses whose appearance is

governed by grammatical principles, that is, clauses which are
in constituency with:
(la) a noun (relative clauses):
Pineapples that don't smell good won't taste good
(lb) a verb (complement clauses):
I don't think that subordination exists
We want to learn Hausa
(lc) a preposition:
I didn't think of checking my shoe size
If we leave aside, then, at least for now, clauses which are in constit-
uency with some noun, verb, or preposition, we are left with essentially
those clause types typically called adverbial clauses, participials, and
nonrestrictive relatives, as shown in (2).
(2) Clause types which are included in this study:

(2a) adverbial clauses
(2b) participials
(2c) nonrestrictive relatives
That is, unlike the clause types given in (1), which are more or less
determined by the grammar of complementation and the pragmatics of
reference, the clause types in (2) have everything to do with how one
decides to convey and relate propositions. It is the clause types in (2)
that concern me here, since these could have conceivably been expressed
as nondependent clauses, whereas the clause types in (1) are almost
completely determined by the lexical items chosen. Chafe (personal
communication) has suggested that the items in (1) also differ from
those in (2) in that those in (2) are typically separate 'idea units' or
'intonation units' (or 'punctuation units' in writing), while those in (1) are
typically part of the same 'idea unit' or 'intonation unit'.
Now all of the clause types in (2) seem to be formally 'dependent1, in

the sense that they cannot represent independent utterances on their
own, but must occur with some other clause. Pragmatically, of course,
almost every utterance we use is dependent, since it requires context for
its interpretation, but what grammarians have been concerned about
when they use the term 'dependent' has generally been grammatical
dependency, not this kind of pragmatic dependency. And grammatical
dependency can be broken down into a number of components, as shown
in (3).
(3) Indicators of formal dependency

A. Nonfinite verb form (or no verb)
B. Nonrestrictive relative pronoun: which, who, where, such
as, etc.
C. Connector for adverbial adjunct:
Temporal connector: when, as, after, as soon as, etc.
Conditional connector: if, unless, as long as, etc.
Causal connector: because, since, as, etc.
Degree connector: so...that, as many...as, the more...the
more, etc.
Concessive connector: although, except, etc.
Means connector: by_
Purposive connector: so that
Manner connector: as if, as
This is perhaps not an exhaustive list, and it is certainly subject to
revision, but it is at least representative of the kinds of grammatical
properties which have led grammarians to talk about 'dependent' clauses
in English. Obviously, the list would be quite different for other lan-
guages; indeed, one of the points of Mithun (1984) was that for many
languages (perhaps in particular those without long written traditions)
such a list might be much shorter, and its devices employed much less
frequently.
My data base consisted of three types of short texts, as shown in (4).
As is evident, my study contrasts with much earlier work in not being
based primarily on narrative discourse data.
(4) Data base
1. Formal Written
Excerpts from the writings of Bertrand Russell and the
economist Thomas Sowell, UCLA administrative memos,
and short articles from the University of California Bulle-
tin and travel literature.
2. Informal Written
Letters to the editor, personal letters, and chatty pieces
about people in a radio program guide.
3. Informal Spoken
Monologues in two-party conversation.
For each of these three groups, out of roughly 300 clauses, I counted
how many were formally marked in one of the ways indicated in (3). An
example of each follows.
Example (5) is from the Formal Written data, a short memo from the
office of the Chancellor of UCLA. The relevant clauses are numbered
(remember that I am ignoring relatives and complements).
(5) Formal Written example (memo from UCLA Chancellor):
ALL UCLA EMPLOYEES:
1. In March 1982 we were informed that the Office of Federal
Contract Compliance Programs of the U.S. Department of
Labor was initiating a review
2. to determine if the campus, as a recipient of federal funds, is
in compliance with federal regulations regarding equal em-
ployment opportunity and affirmative action for women,
minorities, disabled persons and Vietnam era veterans.
3. The OFCCP has completed its off-site examination of UCLA's
Affirmative Action Plan and
4. will begin the on-site phase of its review on Monday, June 28,
1982.
5. As part of that review, OFCCP has asked that all employees
who wish to be interviewed or to comment about UCLA's
compliance with its obligations contact the OFCCP ofice.
6. Employee organizations,
7. if they wish to be interviewed,
6. may also contact that office.
8. The number to call is (213) 997-3185.
9. The last day for arranging these voluntary interviews is July
9, 1982.
10. Supervisors are encouraged to accommodate changes in work
schedules
11. in order to enable individuals to participate in interviews.
There are many interesting things one could say about this text,
including the pragmatics of making an offer which the offerer hopes no
one will take him up on! But our interest here is in the percentage of
clauses which show one of the dependence properties listed in (3). In
clause 2, there is one instance of property A, the nonfinite verb to
determine, and clause 7 has a conditional marker if, which is C in (3X
Finally, clause 11 has another infinitive verb form to enable. So three of
a total of 11 clauses are formally marked as being 'dependent' in the
sense I have defined.
Example (6) is from the opposite extreme, the Informal Spoken group.
It is from a conversation in which a young singer is talking about how it
feels to have been a hit in a New York City Opera production of Bern-
stein's Candide.
(6) Informal Spoken example (young opera singer talking):
1. I got this hysterical phone call from Walker Joyce and John
Rice about two weeks into the run of Candide.
2. They left a message on my machine, 'Well, golly, gee, Erie,
now that you're a star, could we borrow some money?'
3. I called them back
k. and it was wonderful to talk with them.

5. And Tom Detwiler came to the show.
6. But I think what it did for me was put me on the map as a
name that people now recognize.
7. When they see the name,
8. they can think, 'Oh yes, I read about her.1
9. I think it was a wonderful way for me to make a debut.
10. It couldn't have been better, in fact.
11. Though I'm glad that what I do next in New York are two
operas: the Fairy Godmother in Massenet's Cendrillon and
Mordonna in Handel's Alcina.
12. They aren't as big roles as Cunegonde,
13. but at least it will show the people of New York and the
critics that I do other things.
In this sample, two clauses are formally marked for dependency:

clause 7 is a temporal clause, C on our list, and clause 11 is a concessive
clause, also C on our list.
An illustration of the Informal Written style is taken from a personal
letter; example (7) is the body of a letter to me from Cousin Margaret,
who explains why she cannot come to visit.
(7) Informal Written example (letter from Cousin Margaret):
1. Your kind invitation to come and enjoy cooler climes is so

tempting
2. but I have been waiting to learn the outcome of medical
diagnosis
3. and the next 3 months will be spent having the main thumb
joints replaced with plastic ones.
k. Thumbs began to be troublesome about k months ago
5. and I made an appointment with the best hand surgeon in the
Valley
6. to see if my working activities were the problem.
7. Using thumbs is not the problem
8. but heredity is
9. and the end result is no use of thumbs
10. if I don't do something now.
11. I have heard Mother talk about the family arthritis
12. but one never expects to be included.
13. Writing has almost become impossible
14. so we had the typewriter serviced
15. and I may learn to type decently after all these years.
Of these 15 clauses, only two show any markings for dependency:
clause 6 is a purpose clause with a nonfinite verb, property A in (3), and
10 is a conditional, property C.
The totals for the samples I used are given in Table 1; each fraction
represents the number of clauses with indicators of formal dependency,
out of the total number of clauses.
The Table 1 figures need to be taken with a grain of salt since the
total number of clauses at this stage of my research is not very large
and since the categorization represented in (3) is still in the process of

being refined. But they are suggestive, and they are statistically signif-
icant at the .01 level: not only do they suggest that Written Formal
English has a higher proportion of clauses of the type I am here calling
'dependent' than either of the 'informal' styles, but the percentage of
such clauses is almost exactly the same for Spoken Informal as it is for
Written Informal, which confirms the findings of others working on this
problem, such as Chafe, Tannen, and Beaman: that in the area of clause
combining, the formal/informal distinction, rather than the written/spo-
ken one, may be most relevant.
Table 1.
Written Formal: Written Informal: Spoken Informal:

87/302 (29%) 61/306 (21%) 63/306 (21%)
Just as interesting here as the overall numbers, however, may be the

way in which they break down. That is, one can also look at how the
various components of dependency distribute themselves in the three
types of discourse I have been considering. These figures are given in
Table 2.
Table 2.
Written Written Spoken

Formal Informal Informal
Nonfinite verb
form (A) 24
28 (32%) 36 (59%) \ (18%)
Nonrestrictive
relative pronoun (B) 9 12
Connector for ad-

verbial adjunct (C) 59 (68%) 25 52 (82%)
Totals 87 (100%) 61 (100%) 63 (100%)
In Table 2, I have grouped the first two types of dependent clauses, A

and B, together and separated them from the adverbial clause types C,
so that we may compare the way they are deployed in the three types of
discourse I am examining. The most striking thing about Table 2 is that
the three types of discourse make radically different use of the three
types of dependency.
A comparison of the three columns, for example, shows that fully 82%
of the dependent clauses which appear in the Spoken Informal discourse
data (rightmost column) are of the adverbial clause type, while the
adverbial clauses account for only 41% of the dependencies in the Writ-
ten Informal discourse (middle column) and only 68% of the dependencies
in the Written Formal discourse (lefthand column). Extrapolation from
these figures suggests that almost all of the dependencies in Spoken
discourse occur in the adverbial clauses, with people giving causal, con-
ditional, and temporal comments on the events they are discussing. In
contrast, in the Spoken data, there are very few clauses (only three, in
fact) with nonfinite verbs (A in (3)).
The Written Formal discourse column indicates that a larger per-
centage, 32%, as compared to 18% for the Spoken data, were of one of
the types A or B, that is, nonfinite or nonrestrictive. Only 68% were ad-
verbial clauses.
The Written Informal samples (totals shown in middle column), how-
ever, were surprising in this respect. First, a large majority of the
dependencies, 59%, were of the nonadverbial clause type, that is, one of
those shown in A or B. To a certain extent, this could have been due to
the subject matter of the texts in this group. Most of the personal
letters and chatty pieces contained relatively few occasions for pro-
viding reasons, concessions, or, most notably, conditions for the events
being discussed. So this could account for the relatively low 41% figure
for adverbial clauses in the middle column.
But even more striking is the relatively large number of times that the
informal writers (as compared with the informal speakers) took advan-
tage of the option to use nonfinite verb clauses, or clauses with no verb
at all (Property A), or nonrestrictive relatives (Property B). As can be
seen in Table 2, of the total of 306 clauses in the Written Informal
group, 2k were nonfinite (A) and 12 were nonrestrictive (3): 36 of the
total 306. An example of property A, a clause with a nonfinite verb
form from a letter to the editor, is given in (8).
(8) Perhaps the hotel is clearing out its rooms to make way for
condo conversions.
Example (9), from a personal letter, shows Property B, a nonrestric-
tive relative clause.
(9) we were able to start within five minutes after the sound of
the gun, which nobody heard.
In contrast, the Spoken Informal column of Table 2 shows only 3
clauses of type A and 8 of type B; that is, 11 of the 306 clauses in the
Spoken Informal data were of these two types. An example of each of
these is shown in the speech of a man who is explaining that he resem-
bles a number of other people (example (10)). Clause 3 in example (10)
shows a nonrestrictive relative clause.
(10)
1. It's probably—easier no:w -for people to think that uh some-

one looks exactly like me,
2. because, anyone with uh, a beard, of this particular style and
cut—
3. which is - partly gray and partly black,
An example of a nonfinite verb form can be found later on in this
same monologue, as can be seen in clause 15 of example (11).
•Subordination' in formal and informal discourse / 93
(11)
14. There was a long moment,
15. having just come back from::some time:: in funny places like
China, an:: hhh! Guam and places like that...
So 36 of the 306 clauses in the Written Informal sample were non-

restrictives or contained nonfinite verbs, while in Spoken Informal, the
number of clauses of either of these two types, either with a nonfinite
verb, or nonrestrictive relatives, is quite small, only 11. This suggests
that in these two clause types the planning involved in writing may make
a difference: writers, even those writing somewhat informally, make
relatively free use of these particular dependency devices, but speakers
hardly ever do.
What do these data tell us then? Table 1 suggests that there is a
difference between formal and informal discourse in terms of how users
of English employ some of the dependent clause options available to
them. That is, the two rightmost columns, those representing somewhat
more informal discourse, show a smaller overall percentage of dependent
clause types than the leftmost column, which represents formal dis-
course (around 20% as compared to 30%).
But Table 2 suggests that there is also a difference between speaking
and writing in terms of these dependencies. The left and middle col-
umns, representing the written discourse types, contrast with the right-
most column, which represents spoken discourse, in the percentage of
nonfinite verb forms and nonrestrictive relative pronouns: the Spoken
discourse can be seen to have fewer such forms than the Written. This
shows that some of these options are primarily associated with writing,
namely, those by which language users create dependent clauses by using
nonfinite verb forms and nonrestrictive relatives.
What these data suggest to me is, first of all, confirmation for what
previous researchers, such as Tannen, Chafe, Seaman, and others, have
found. But they also suggest that my initial suspicion was correct: the
notion 'subordinate clause' cannot be assumed. We cannot continue to
use the term 'subordination', believing that other linguists will know
what we are talking about—partly because linguists do not all mean the
same thing by the term, and partly because important facts are missed if
we lump various types of 'nonmain' clauses into one group. We have to
break down the notion of 'subordination' into its component parts, and
try to determine what each of them is doing in different types of dis-
course.
Notes
I am grateful to Wally Chafe for much helpful discussion of the ideas
in this paper.
1. His research also shows that there do not seem to be large differ-
ences between written and spoken English, at least with respect to that
complements and infinitival complements.
2. This list is adapted from a similar one being used by Chafe in his
current research.
9* / Sandra A. Thompson
3. Tannen (personal communication) makes the interesting observa-

tion that this example, as is indeed the case for many of my examples in
the Written Informal group, represents not so much an informal genre as
an informal notion of formality, a kind of 'folk formality1, as she puts it.
Evidence for this includes Cousin Margaret's quasi-poetic cooler climes,
her use of nominalizations such as using thumbs and medical diagnosis,
and her one subject in clause (12). This suggests that my category
'Written Informal1 is perhaps somewhat more formal than its position
between 'Formal Written1 and 'Informal Spoken' would indicate. How-
ever, as I show further on, in many ways the two groups which I have
characterized as 'Informal' behave alike in contrast to the one which I
label 'Formal'.
References
Beaman, Karen, (to appear) Coordination and subordination revisited:
Syntactic complexity in spoken and written discourse. In: Tannen (to
appear).
Brugman, Claudia, and Monica Macaulay. 1984. Proceedings of the
Tenth Annual Meeting of the Berkeley Linguistics Society. Berkeley,
Calif.: Berkeley Linguistics Society.
Chafe, Wallace. 1982. Integration and involvement in speaking, writing,
and oral literature. In: Tannen (1982b).
Chafe, Wallace. 1984. How people use adverbial clauses. In: Brugman
and Macaulay (1984).
Clancy, Patricia. 1982. Written and spoken style in Japanese narra-
tives. In: Tannen (1982b).
Givon, Talmy, ed. 1979. Discourse and syntax. Syntax and semantics,
vol. 12. New York: Academic Press.
Haiman, John, and Sandra A. Thompson. 1984. 'Subordination' in uni-
versal grammar. In: Brugman and Macaulay (1984).
Mithun, Marianne. 1984. How to avoid subordination. In: Brugman and
Macaulay (1984).
Ochs, Elinor. 1979. Planned and unplanned discourse. In: Givoli (1979).
Tannen, Deborah. 1982a. The oral/literate continuum in discourse. In:
Tannen (1982b).
Tannen, Deborah, ed. 1982b. Spoken and written language. Norwood,
N.J.: Ablex.
Tannen, Deborah, ed. (to appear) Coherence in spoken and written
discourse. Norwood, N.J.: Ablex.
SPEAKING, WRITING, AND PRESCRIPTIVISM
Wallace Chafe
University of California, Berkeley
For the last several years I have been looking at differences between
spoken and written language, trying to establish a little more clearly just
what the differences between them are, and trying to identify the rea-
sons for these differences. Here I am going to discuss one of the rea-
sons, one which I have only recently begun to look into in any serious
way. I refer to the influence which normative or prescriptive grammar
has had on the differentiation between speaking and writing.
To a large extent, prescriptivism itself arose from the fact that
written language is visual rather than auditory, and thus has tended to be
more permanent than spoken language. The fact that it can be leisurely
examined has made it highly subject to reification, to treatment as an
object of critical comparison and analysis. People have tended to pay
attention to differences between the forms of language used by different
writers much more than they have paid attention to the presence of such
differences between speakers. The evanescence of speaking allows lin-
guistic differences to slip by unnoticed, whereas the permanence and
perusability of written language calls attention to such differences.
When human beings notice differences in the ways they do something,
they seem universally to react with a firm conviction that one of the
ways is the correct one, and that all the others are at the very least
mistaken, but probably also illogical or corrupt. This reaction is as good
a candidate as exists anywhere for an automatic human reflex. The
permanence and perusability of writing lead inevitably, by this reflex
reaction, to the notion that there is one kind of language that is best,
and many kinds that should be avoided. If speaking has the freedom and
fleetingness of a butterfly, writing pins language to a board. As soon as
this happens, it is natural to say 'this is the way a butterfly should be,
and other butterflies are monstrosities'.
The ultimate result may be the establishment of an academy whose
goal is to establish the right form of language and protect it against cor-
rupting influences. That is what happened in Italy, France, Spain, and
95
96 / Wallace Chafe
Sweden in the sixteenth, seventeenth, and eighteenth centuries. It did

not happen in England, although there was significant sentiment in favor
of it. A prescriptive attitude has nevertheless played a noticeable role
in shaping the English language. Its influence reached a climax in the
nineteenth and early twentieth centuries, particularly as it provided
guidance for schoolteachers in need of something to teach (Finegan
1980, Quinn 1980, Baron 1982).
My own interest has been in trying to pin down the ways in which
prescriptivism has contributed to the relation between speaking and
writing. For one thing, I have been trying to identify some of the specif-
ic differences between spoken and written English which are causally
attributable to it. At the same time I have been trying to identify the
kinds of historical developments and mechanisms which have been at
work. Contrary to what I first expected, I have learned that the course
of prescriptivism has not always been the same, nor have its results
always been of the same kind.
I am going to describe here three courses of development, each quite
different from the others. There is no implication that these are the
only developments that can occur, but they do illustrate well the range
of possibilities.
The chief kind of evidence I have for these developments that is new,
and not derived from what others have said, comes from a project
(funded by the National Institute of Education, Grant 80-0125) in which
3ane Danielewicz and I collected two kinds of spoken language (roughly
formal and informal) and two kinds of written language (also roughly for-
mal and informal) from each of 20 individuals, and subjected these
samples to an analysis in terms of a number of features hypothesized to
have different distributions in at least some kinds of writing as opposed
to some kinds of speaking. The four kinds of language are (1) conversa-
tional speaking, from dinnertable conversations; (2) academic speaking,
from academic lectures; (3) letter writing, from personal letters; and CO
academic writing, from academic papers. I have compared these sam-
ples in various ways elsewhere (Chafe 1982, in press, Chafe and Daniele-
wicz in press). Here I am going to discuss only a few examples of differ-
ences which appear to be attributable in one way or another to prescrip-
tivism.
There is one kind of development that seems to fit into this picture,
but that is not so much a matter of prescriptivism per se as a matter of
written language possessing a greater inertia than spoken when subjected
to the forces of linguistic change. We might say that speaking, because
of its evanescence, is intrinsically more prone to change, whereas writ-
ing, because of its permanence and perusability, is slower to accept
change. (Like everything I say here, this is an oversimplification, since
writing may also foster changes of its own kind.) The result is the
course of development which is represented schematically in Figure 1.
This figure should be interpreted as two time lines, one representing
spoken language and one written language, moving from left to right.
The vertical dimension represents the absence or the significant pres-
ence of some feature, as indicated on the left. Initially, this feature is
absent from both speaking and writing, as shown by the coincidence of
the two lines on the left. At some point it emerges in speaking, but
remains for a time absent from writing, as shown by the divergence of
Speaking, writing, and prescriptivism / 97
the two lines in the middle. During the middle period, the feature is one
which differentiates speaking from writing. Eventually, writing begins
to catch up, and eventually, the two lines come back together, with
writing as well as speaking making significant use of the innovation.
Figure 1.
Presence: Speaking
Absence: / Writing
A feature which probably fits this pattern is the use of the demonstra-
tive this (or its plural these) as an alternative to the indefinite article.
Here are some examples:
...the second thing I heard was this...woman's voice,
...and here's this divider and the...this big wooden thing,
...there's this...nun in the program,
...and he saw these two foreigners,
I do not know exactly when this usage began to be common in spoken
English, but it seems to have emerged in fairly recent times. So far as I
know, school grammars have not prescribed against it, although I am
acquainted with a few older people of a prescriptive turn of mind who
are put off by it. Both intuition and the evidence from our study suggest
that it has not yet made any significant inroads into careful, edited
written English. We found five occurrences of it per thousand words of
academic speaking, and no occurrences at all in either academic writing
or in letter writing. I do not want to suggest that it never occurs in
letters, however, since I recently received a letter in which there were
two occurrences in the first sentence. The situation pictured on the
right side of Figure 1 is in this case only a prediction: a situation in
which the indefinite demonstrative will in the future be used as freely in
all styles of written English as it is presently used in conversation. But,
except for the inertia of writing and the prescriptive reflexes of people
who are likely to become fewer as time goes on, there is nothing to pre-
vent such a development. This is a feature which is neither illogical,
productive of ambiguity, nor inelegant, so that it will be difficult to
keep it out of writing with arguments along those lines.
Figure 2 shows a very different course of development. It is, I think,
the kind of development typically associated with prescriptivism. In
Figure 2 there is some feature which is at first equally present in both
speaking and writing. It is then found to be undesirable, perhaps because
it is at odds with the grammar of Latin; perhaps because it is believed to
be illogical, inelegant, or productive of ambiguities; or perhaps because
it violates some artificially imposed pattern. Once legislated against,
this feature comes to have a reduced frequency in writing, as indicated
in the middle portion of Figure 2. Probably it goes on being used natu-
rally in the speech of most people, although attempts to avoid it in
speaking may become an index of pedantry. At this middle stage, then,
the presence or absence of this feature is diagnostic of written vs.
98 / Wallace Chafe
spoken language. The righthand portion of Figure 2 suggests that, with a

relaxation of the influence of prescriptivism, such a feature may sooner
or later regain its frequency in writing, with perhaps an eventual return
to the original lack of differentiation between speaking and writing in
this respect.
Figure 2.
Presence: Speaking
Absence: \ Writing /
One feature of this kind is evidently the use of shall and will. The
story of the prescriptions regarding these two auxiliaries is a long, com-
plicated one, and much of it was told by Charles Fries (1925). Fries not
only surveyed the origin of the peculiar rules which came to dictate
their use; he also traced their actual occurrences in 86 dramas from the
middle of the sixteenth century to the first quarter of the twentieth,
looking separately at their use in independent declarative statements,
questions, and subordinate clauses. I simplify the story here by focusing
on the use of I will as a marker of the future tense in declarative sen-
tences.
As in many such cases, the prescriptions regarding shall and will came
to be stated with precision only in the nineteenth century. Even then,
different grammarians said somewhat different things. While everyone
agreed that I shall was the proper choice for the simple future tense,
Brown (1851), for example, said that I will should be used for a 'promise,
command, or threat,1 whereas Alford (1864) associated I will on etymo-
logical grounds with the will of the speaker. It appears that the con-
sistent use of I shall for the future was actually not established until the
second half of the nineteenth century. This usage was thereafter care-
fully adhered to by many who learned to write during the period from
approximately 1850 to 1950, including, from the sublime to the ridicu-
lous, Edmund Wilson (e.g. 1971) and me. The evidence is fairly clear
that the prescription is quickly running its course at the present time,
with the incidences of I shall declining, and those of I will taking over.
In our data we found two instances of I shall vs. five of I will in academ-
ic writing. In letter writing there was only one instance of I shall vs.
seven of I will. I find that I shall has completely disappeared from my
own writing. It may not be long before I will has completely vanquished
its alternative, so the situation will have become that represented on the
right side of Figure 2.
Another feature which has followed the course schematically repre-
sented in Figure 2 is the use of prepositions at the ends of relative
clauses and question-word questions. The prohibition against this usage
is generally supposed to have begun with John Dryden. Citing the fol-
lowing two lines from Ben 3onson:
The Waves, and Dens of beasts cou'd not receive
The bodies that those souls were frighted from.
Dryden found fault with 'the Preposition in the end of the sentence; a
common fault with him, and which I have but lately observ'd in my own
writings' (1672, quoted from Bolton 1966:60). He then 'went through all
his prefaces contriving away the final prepositions that he had been
guilty of in his first editions' (Fowler 1965:473). Subsequent prescrip-
tivists were actually less dogmatic in this case than in others. For
example, Robert Lowth, who seems to have been about the only eigh-
teenth century grammarian to have paid much attention to final pre-
positions, said that their use
is an idiom, which our language is strongly inclined to: it prevails
in common conversation, and suits very well with the familiar style
in writing: but the placing of the Preposition before the Relative is
more graceful, as well as more perspicuous; and agrees much better
with the solemn and elevated style (Lowth 1775:164).
Interestingly, Lindley Murray (1795:122) quoted this opinion verbatim,
except that he amended it to read 'to which our language is strongly
inclined'. Alford (1888:118) was relatively permissive in this regard:
'There is a peculiar use of prepositions, which is allowable in moderation,
but must not be too often resorted to. It is the placing them at the end
of a sentence as I have just done in the words "resorted to".1
In spite of this fairly relaxed attitude among the best known pre-
scriptivists, our data suggest that final prepositions are even now signi-
ficantly avoided in more formal kinds of writing. Dryden's prohibition
was formulated more clearly and at an earlier date than the prohibition
against I will, and it continues to exert a stronger influence. We found
1.2 occurrences of final prepositions per thousand words of conversation,
1.7 occurrences in lectures, 1.1 in letters, but only 0.2 in academic
writing.
Figure 3 shows a third course of development, and it is one which
contradicts the notion that prescriptivism always increases the distance
between writing and speaking. Here there is a situation in which the
feature in question has always been absent from or at least has been rare
in spoken language. Initially, there is a stage at which the feature has
some currency in writing. There then develops a prohibition against it
which causes its decline in that medium, with the paradoxical result that
writing becomes more, not less, like speaking in this regard. As the
prohibition exerts less influence, the feature comes to be written more
commonly again, with a paradoxical increase in the differentiation of
spoken and written language.
Figure 3.
Presence : Writing
Absence: Speaking \ /
I have noticed two features that fit this pattern. Both are well known
as usages prescriptivists would like us to avoid. One is the split in-
finitive, the other the dangling participle. There is some question
100 / Wallace Chafe
whether split infinitives fit the lefthand side of Figure 3 or not. In a

study of prescriptive grammar in the eighteenth century, Leonard (1929:
95) remarked that
the most striking circumstances in this array of censured 1 con-
structions is that no mention whatever of the 'split infinitive was
discoverable, nor was the construction itself observed save once or
twice in the authors read. Apparently, it was both a discovery and
an aversion of nineteenth century grammarians.
His book includes the following example from George Campbell (written
in 1776):
to vigilantly attend to every illegal practice that were beginning to
prevail (Campbell 1963:370).
But, of course, it is quite possible that the construction was used more
frequently by writers who were not grammarians, and thus were not sur-
veyed by Leonard.
In any case, in the nineteenth century we find Henry Alford saying
with regard to the split infinitive, 'But surely this is a practice entirely
unknown to English speakers and writers. It seems to me, that we ever
regard the to of the infinitive as inseparable from its verb' (Alford
1888:133). Fowler (1965:579) has an amusing entry in which he says, 'The
English-speaking world may be divided into (1) those who neither know
nor care what a split infinitive is; (2) those who do not know, but care
very much; (3) those who know and condemn; (4) those who know and ap-
prove; and (5) those who know and distinguish.1 Later, in giving a set of
examples of split infinitives, he notes that because they 'are from news-
papers of high repute, and high newspaper tradition is strong against
splitting, it is perhaps fair to assume that each specimen is a manifesto
of independence' (Fowler 1965:581). In our own data we found no split
infinitives anywhere at all except in academic writing, where there were
three examples in 20,000 words:
individuals are called upon to verbally display some knowledge or
proficiency at recurrent communicative tasks,
we began to systematically audio and videotape these episodes
there is more dependence on context to indirectly express these
relationships
Two of these examples were from the same writer.
I do not mean to suggest that split infinitives never occur in speak-
ing. They are, however, rare, and usually restricted to cases in which
the splitting word is an intensifier like really or actually. The written
cases, on the other hand, usually involve a manner adverb, as in the
examples I have given. It is thus relevant that in our data, manner ad-
verbs themselves were only half as common in conversation as they were
in either academic writing or in letters. Both the use of a manner ad-
verb and the use of such a word to split an infinitive seem to require at
least a small amount of planning of the kind that is performed by writers

more easily than by speakers.
Split infinitives, then, may have occurred significantly in writing
before the nineteenth century (contra Leonard), but more evidence is
needed. Only in the nineteenth century did there arise an explicit pro-
hibition against them. Their subsequent avoidance by careful writers has
had the effect of making written language more like spoken language
from the point of view that both have failed to make much use of them
(the situation schematized in the middle of Figure 3). With the current
relaxation of prescriptive guilt, split infinitives may be entering careful
expository prose more often, though at present individual writers differ
in their tolerance of them. We may expect that they will gradually be-
come more common in writing, and if they do, the gap between writing
and speaking will paradoxically have been increased.
'Dangling' or 'unattached' participles may also follow the pattern of
Figure 3. Again I am unsure of their prevalence before the nineteenth
century, though there must have been something there for grammarians
to inveigh against. Alford, again, is explicit in this regard:
One of the commonest of newspaper errors is to use a participial
clause instead of one with a- verb, leaving the said clause pendent,
so that in the reader's mind it necessarily falls into a wrong rela-
tion. Thus we had in the Times the other day, in the description of
the York congress, assembled under the presidency of the Arch-
bishop: 'His Grace said, &o, and after pronouncing the benediction,
the assembly separated' (Alford 1888:191).
Prescriptivists have often been able to point in this way to the fact that
dangling participles, if taken literally, will lead to an unintended inter-
pretation. On the other hand, Brown (1851:605) was liberal enough to
allow that certain participles 'are frequently used in discourse so inde-
pendently, that they either relate to nothing, or to the pronoun I or we
understood; as, "Granting this to be true, what is to be inferred from it?"
. . . This may be supposed to mean, "I, granting this to be true, ask what
is to be inferred from it?'"
As was the case with split infinitives, views like Alford's reduced the
use of dangling participles in writing. And again the paradoxical result
was to make written language more like spoken, for dangling participles
(or for that matter participles of any kind) may never have been a com-
mon feature of spoken language. In our data we found only 3 occur-
rences of participles of any kind per thousand words of conversation, and
no more than 5 in lectures, as opposed to 10 in letters and 19 in academ-
ic writing. Although, lacking adequate data on earlier stages of spoken
English, we can never be entirely sure, in all probability the incidence of
participles in speaking was not significantly higher in centuries past than
it is now. The use of clauses built from participles rather than finite
verbs represents a strategy foreign to the normal pattern of spoken
English, with its preference for the chaining together of relatively
independent clauses.
Our materials contained four examples of dangling participles in
20,000 words of academic writing—hardly an impressive number—but
there were none at all in any of our other language samples. The four
from academic writing were:
102 / Wallace Chafe
Assuming a self-terminating search, the number of decisions should

increase with the number of target properties.
Also, objective semantic properties may be used, avoiding the pos-

sible displaced rehearsal problem of evaluative ratings.
Based on the research summaries so far, written language contains

many more complex syntactic structures than oral language, sug-
gesting that in written language ideas are related or combined in
ways not apparent in spoken language.
(The last example contains two.) If these examples can be taken as

symptoms of a trend, here too the current relaxation of prescriptive
standards has had the effect of widening, not narrowing, the gap be-
tween writing and speaking.
To summarize, I have introduced the question of how prescriptivism
has affected the relation between spoken and written language. I have
described three very different kinds of historical developments, which I
summarized schematically in Figures 1, 2, and 3.
The Figure 1 pattern, exemplified by the indefinite demonstrative
this, involves an innovation in spoken language whose introduction into
writing is resisted by the inertia of that medium, as well as by possible
prescriptive attitudes against it. There is thus a period during which
such a feature contributes to the gap between speaking and writing, but
after which writing may catch up, reducing or eliminating the gap.
The Figure 2 pattern, exemplied by the shall-will dogma and by final
prepositions, involves a feature which was once common in both speaking
and writing, whose incidence in writing was decreased by prescriptivism,
and which may subsequently be reintroduced into writing as the influ-
ence of prescriptivism wanes. Here the maximum differentiation be-
tween spoken and written language occurs during the period of maximum
prescriptive influence.
The Figure 3 pattern, exemplified by split infinitives and dangling
participles, presents the opposite picture. Here there was never a very
significant occurrence of the feature in spoken language. These are
features whose use is fostered by the extra planning time available to
writers, although, from the point of view of a prescriptivist, they might
be regarded as planning which has gone astray. During the period of
maximum prescriptive influence the gap between speaking and writing
was narrowed, not widened, and it is with the relaxation of prescrip-
tivism that the gap is now increasing.
There is a sense in which each feature affected by prescriptivism has
had its own history. On the other hand, there are sets of features which
do seem to fall into repeated patterns of the kinds I have discussed.
Studies of this sort can have some significance, not just for our under-
standing of certain differences between speaking and writing, but also
for our understanding of the mechanisms of language change. From the
small amount of work I have done in this area so far, I can report that it
is frustrating—in part because of the diversity of the written data, in
part because of the lack of satisfactory spoken data from earlier
periods. Even so, it is certainly possible to do considerably more than I
have done in this very preliminary study, and the results, as they come
in, are likely to be intriguing and rewarding.
References
Alford, Henry. 1888. The queen's English. A manual of idiom and
usage. Seventh ed. (Originally published in 1864.) London: George
Bell and Sons.
Baron, Dennis E. 1982. Grammar and good taste. Reforming the Amer-
ican language. New Haven and London: Yale University Press.
Bolton, W. F. 1966. The English language. Essays by English and Amer-
ican men of letters 1490-1839. Cambridge: Cambridge University
Press.
Brown, Goold. 1851. The grammar of English grammars. New York:
Samuel S. and William Wood.
Campbell, George. 1963. The philosophy of rhetoric. Edited by Lloyd
F. Bitzer. (Originally published in 1776.) Carbondale: Southern Illi-
nois University Press.
Chafe, Wallace L. 1982. Integration and involvement in speaking, writ-
ing, and oral literature. In: Spoken and written language: Exploring
orality and literacy. Edited by Deborah Tannen. Norwood, N.J.:
Ablex.
Chafe, Wallace L. (in press) Linguistic differences produced by differ-
ences between speaking and writing. In: The nature and consequences
of literacy. Edited by David Olson, Angela Hildyard, and Nancy Tor-
rance. Cambridge: Cambridge University Press.
Chafe, Wallace L., and Jane Danielewicz. (in press) Properties of
spoken and written language. In: Comprehending oral and written
language. Edited by Rosalind Horowitz and S. J. Samuels. New
Finegan, Edward. 1980. Attitudes towards English usage. The history
of a war of words. New York and London: Columbia University
Teachers College Press.
Fowler, H. W. 1965. A dictionary of modern English usage. Second ed.,
revised by Sir Ernest Gowers. (Originally published in 1926.) Oxford
and New York: Oxford University Press.
Fries, Charles C. 1925. The periphrastic future with shall and will in
modern English. Publications of the Modern Language Association 40.
936-1024.
Leonard, Sterling Andrus. 1929. The doctrine of correctness in English
usage 1700-1800. University of Wisconsin Studies in Language and
Literature, No. 25.
Lowth, Robert. 1762. A short introduction to English grammar: With
critical notes. London.
Murray, Lindley. 1795. English grammar. (Reprinted by The Scolar
Press, Menston, England, in 1968.)
Quinn, Jim. 1980. American tongue and cheek. A populist guide to our
language. New York: Pantheon Books.
Wilson, Edmund. 1971. Upstate. Records and recollections of northern
New York. New York: Farrar, Straus and Giroux.
SUBSTRATE AND UNIVERSALS IN THE TOK PISIN VERB PHRASE
Gillian Sankof f
In previous papers on various aspects of Tok Pisin syntax and morpho-
logy (Sankoff 1977a, b; 1979; Sankoff and Laberge 1973; Sankoff and
Brown 1976), I have tried to steer a course that founders neither on the
lofty cliffs of universals nor on the hidden substrate reef. There are
certainly those who see one or the other of these as a haven, and who
would deny the viability of any third course, on the grounds that these
two 'solutions' exhaust the universe of explanatory possibilities. In my
view, neither can constitute a full explanation of developments in Tok
Pisin, because neither explains the dynamics of the processes involved.
By this I mean something quite specific. Recourse to either substrate or
universals, at least as employed in the Creole literature, has generally
been little more than an exercise in pattern matching. One observes
how some part of the grammar of a Creole language patterns, and then
matches this pattern to some template, whether a putative universal or
a part of the grammar of some other language.
We need to do more than this. There may well be a 'natural semantax'
(Bickerton 1974) in the brain, and people no doubt also carry in their
brains the indelible structures of their first language grammar. But for
either (or indeed any) pattern to surface and be sustained in the lan-
guage, the discourse that it generates must prove viable in the everyday
conditions of its production. The abstract rules or patterns that are
usually thought of as generating discourse are in fact in a symbiotic
relationship with it, because they will be altered if what they produce
cannot survive in its natural environment: talk. Talk is in turn regu-
lated by both the microinteractional and the larger sociohistorical
context. Thus the reproduction of systems and subsystems in grammars
depends on both aspects of what I have called the 'social life' of language
(Sankoff 1980:xix-xxii).
This paper examines the grammatical subsystem that has been at the
heart of recent theoretical discussions of the nature of Creole languages,
i.e. tense, mood, and aspect marking (cf. Bickerton 1974, 1981; Muysken
1981, Givon 1982). In what follows, I steer a course that will throw
neither universals nor substrate out with the bath water, but that is not
104
Substrate and universals in the Tok Pisin verb phrase / 105
merely a compromise between the two. This course has been charted
elsewhere, notably in Sankoff and Brown (1976) and Laberge and Sankoff
(1979). It rests on the assumption that everything that becomes part of
the grammar of a language must first appear in discourse. A change in
discourse patterns will be reflected in a change in the probabilities
associated with the use of particular variants, which may lead to a
reorganization of grammatical categories and rules.
Grammaticalization processes are, of course, by no means limited to
pidgins and Creoles: what is special is that typically many such pro-
cesses are occurring at once. The result is that often we cannot rely on
relating something we are trying to establish in one part of the grammar
to some clearly established fact in some other part of the grammar, for
that second fact will also be something that is in transition between two
categories. We shall see an example of this in attempting to relate
subject clitics to the establishment of an auxiliary.
1. Universals and substrate in the sociohistorical context. The socio-

historical context in which Tok Pisin evolved is fundamental to under-
standing the problems Tok Pisin presents to both the substrate and the
universals lines of explanation. Tok Pisin is purportedly a bad candidate
for universals because it was clearly not an 'early creolized1 Creole
(Bickerton 1981:3-4). Though it had its roots in nineteenth century
plantations (Mosel 1980, MUhlhMusler 1978), it was not a 'plantation
Creole' but rather a plantation pidgin. Its formative period occurred in a
plantation system organized around relatively short-term indentured
labor rather than slave labor. Workers stayed a few years rather than a
lifetime. Thus the 'successive generations' of Tok Pisin speakers were
successive cohorts of mainly young single men, each being replaced by
another such cohort. Most of the children born to these men were born
back home, into families with a Papuan or Austronesian home language.
The fathers of such children may have introduced a little Tok Pisin into
some of their talk with them, but most of their mothers could not speak
it, with the result that those children who acquired it learned it in the
same context their fathers had: as young adults away at work.
Now although Tok Pisin was formed in the plantation context, where
speakers were removed from their home communities, it survived and
flourished back in the New Guinea villages that workers returned to
after their period of indenture was over. In rural areas near European
settlements, many people participated in the cash economy, where Tok
Pisin was the lingua franca. And even in the more remote rural areas,
initial recruitment of plantation laborers was usually concomitant with
bringing an area into European control, with ensuing involvement with
the civil authorities, the cash economy, and the missions. With the
exception of some of the missions (cf. Sankoff 1977c:288-89), Tok Pisin
was the lingua franca in all of these domains.
The creation of an infrastructure for Tok Pisin 'on the ground1, in the
places to which the plantation workers went home, has meant that the
vast majority of Tok Pisin speakers have over the past hundred years
spent most of their adult lives using Tok Pisin as a lingua franca in all
sorts of dealings with the world outside of their natal villages, but have
kept their first languages for home and local use (Sankoff 1971, Gilliam
1984). These would seem to be ideal conditions for the introduction of
106 / Gillian Sankoff
many substrate features into Tok Pisin. According to Bickerton, the
introduction of such features while the language was still at the 'pidgin'
stage would have blocked the development of a number of the putative
universals he finds in 'early creolized1 Creoles. But there is also a major
problem with invoking substrate as an explanation for anything to be
found in Tok Pisin grammar: the vast number and diversity of languages
spoken as first languages by speakers of Tok Pisin. Among the several
hundred languages spoken in Papua New Guinea, one could surely always
find at least two or three possible candidates for the source of anything
less specific than a lexical item. Limiting the search to the first lan-
guages involved at the time when a given feature got established in Tok
Pisin would reduce the number of possible substrate candidates, but
even so, one is left with a bewildering array of potential sources, and no
principled reason for choosing among them.
Three unsatisfactory possibilities come to mind: (1) Tok Pisin has
arbitrarily acquired feature x from language X, feature y from language
Y, and so on; (2) Tok Pisin has incorporated only those substrate fea-
tures with wide areal distribution; and (3) there is no unified grammar
of Tok Pisin, only a set of versions or dialects, each of which reflects a
different substrate, and which have enough surface resemblances to
permit communication. The first of these possibilities, sometimes
termed the 'cafeteria principle', is the least satisfactory and can, I think,
be seriously invoked only in the case of particular lexical items. The
problem with the second is that it is not sufficiently specific, but it
contains a grain of truth which I extract and refine in the concluding
section of this paper, where I also reexamine the third possibility.
2. The verb phrase in Tok Pisin. There is no current, unified analysis

of the Tok Pisin verb phrase. Such an analysis is complicated at least by
(1) the presence of so-called 'serial verb' constructions in Tok Pisin; (2)
the uncertain status of many of the forms which seem to mark tense,
mood, and aspect in Tok Pisin as modals, auxiliary verbs, full verbs,
particles, adverbs, etc.; and (3) the lack of a clear understanding of the
distribution and function of the so-called 'predicate marker' j_. The
present treatment is a first step toward unifying these questions.
There have been several attempts to analyze certain aspects of the
problem. Hall (1943), Laycock (1970), Mihalic (1971), and especially
Wurm (1971) all provide descriptions of each of the putative markers;
MUhlhclusler (1976) discusses as part of his treatment of the lexicon
those serial verb constructions analyzable as compounds. Woolford
(1977) tries to unify an analysis of MUhlhSusler's 'compound' type serial
constructions with other kinds of chaining, but her treatment is flawed
by not taking into account the fact that some of the verbs she deals with
should be analyzed as belonging to the auxiliary (cf. section 2.1). Final-
ly, Mosel (1980), in her examination of the extent to which Tok Pisin
exhibits substrate influence from Tolai, looks at both MUhlhMusler's
'compounding' chains and some of the tense, mood, and aspect markers,
but in two quite separate parts of her analysis. In addition, given the
gap in the literature on Tok Pisin, her treatment of tense, mood, and
aspect is necessarily perfunctory. Some of the aforementioned sources
also provide comments on_i, which is also the subject of Woolford (1981)
and Franklin (1980).
2.1 Serialization. Although verb serialization, or verb chaining, is a
feature not unique to pidgins and Creoles, it is often said to be charac-
teristic of them. In Tok Pisin, we find examples such as those in (1)
through (7).
(1) Em i laik go hukim pis.
3S want go hook fish
'He wanted to go rcatch fish"!
Ifishing.' J
(2) Em slip karai istap.
3S lie cry stay
'He was lying there crying.'
(3) Mipela igo swim igo.
1PL.EXCL go swim go
'We went swimming.'
CO Ol man kam sanap lukluk nau.
PL man come stand look then
'The people came and stood looking then.'
(5) Man ia kirap amarim em.
man DET get-up hammer 3S
'The man up and hammered him.'
(6) Na em i salim mi kam.
and 3S send IS come
'And he sent me.'
(7) Yu yet mekim kamap long mi.
2S INTENS make happen to IS
'You yourself made it happen to me.'
In my view, the verbs underlined in this set of examples should be
analyzed as having different statuses, and as I develop the analysis, I will
return to them at various points in this section and in the next two
sections. First, however, let us examine Woolford's proposal (1977),
which essentially treats them as all having the same status.
Woolford's analysis unites what she calls 'strictly defined' serial verbs,
i.e. 'strings of two verbs with an intermediate noun such that the noun is
the object of the first verb and the subject of the second' and 'other
sorts of verb chaining' (p. 176). Her first type is exemplified by sen-
tences such as (6) and (7), as well as by the second two verbs in (8)
(Woolford's (2), 1977:176). In citing Woolford's example sentences, I use
her glosses.
(8) Em i save salim smok igo antap long maunten.
he know send smoke to up to mountain.
'He sends smoke up to the mountain.'
Her mixed bag ('other sorts of verb chaining') contains two main types
of sentences. Sentences of the first type are like (9) (Woolford's (7),
1977:177), consisting of two intransitive verbs with the same subject, the
first a motion verb and the second what Woolford (1977:181) calls a
'directional verb'. The second two verbs of (3) also demonstrate this
pattern.
(9) Em i kalap i ££ daun long si.
he jump go down to sea
'He dived into the sea.1
The second type contains sentences like (10) (Woolford's (8), 1977:177)
and (11) (Woolford's (9), 1977:178). Though Woolford does not character-
ize them in any way, they seem to be chains of verbs with the same
subject that take either NP or sentential complements.
(10) Meri ia giaman pasim ai na em i lukluk.
woman pretend close eyes and she look
'The woman pretended to close her eyes and she looked.'
(11) Meri i holim pasim man bilong en.
woman hold fasten man of her
'The woman held her husband fast.'
Woolford proposes (1977:181) the phrase structure rule in (12) to
account for at least the two main types of sentences she discusses.
(12) VP—>V (NP) (VP)
Thus, (8) would be derived as in (81) and (9) would be derived as in (9').
(80 ^VP^ (9') VP
For chains of three or more verbs, Woolford's analysis gives us a right-

branching structure. This would apply reasonably well to a number of
verb chains found in my own data, e.g. (I1) and (13) through (16). The
nodes are labeled provisionally according to Woolford's analysis. Note,
however, that the verb save in (16) is labeled a habitual marker rather
than being glossed as 'know' as in Woolford's sentence, presented as (8).
(10 J. (13) •VP
VP V
i laik V VP i laik V, / \ P
go V NP go V
hukim pis paitim
'He wanted to go fishing' 'He wanted to go hit (him)1
VP (15) VP
V yp V VP
I YP I / \
i ken V
i ken V VP
go istap (o V
wokim
'He could go make (it)'
(16) ^^V
'He could go stay' P
V VP
mi save / \
(HAB) V VP
laik I
V
toktok
'I like to talk1
Now consider sentence (17). The negative, no, occurs in the topmost
VP node. Indeed, negation may occur only in this node, as illustrated in
(160.
(17) (160
i kam save
kisim (mi*) (no*) laik
(mi*) (no*) toktok

110 / Gillian Sankoff ate and universals in the Tok Pisin verb phrase / 7
At the very least, this means that we would have to write a surface
filter to rule out the application of no anywhere other than in the top-
most VP. We also need to limit the application of the subject clitics—-mi
(1st sing.), û_ (2nd sing.), yumi (1st pi. inclusive), ol (3rd pi.), and the
-pela forms (various duals and plurals) in the same way. Note that the
third person singular ^m is not a clitic (cf. sentence (17)); I return to the
subject of clitics in section 2.3.
Now notice the parallels in (I1) and (13) through (15). All consist of a
modal (laik 'want' or ken 'can'), followed by go 'go' or kam 'come', fol-
lowed by a main verb. This is sketched in (18). Example (16), where
save precedes laik, indicates that one might theoretically find sentences
with more in the preverbal position. Other examples, however, lead us
to order the three tentatively as follows: ken, save, and laik. And bin,
marking past, or past-before-past, must occur prior to all of them, e.g.
bin save wok 'had used to work' of bin go wok 'had gone to work', and so
forth.
(18)
hukim
toktok
paitim
stap
wokim
As (18) makes evident, it is clear that in all of these examples the

rightmost verb in the string is the main verb, j»o or kam somehow modi-
fies it, and ken, save, or laik has scope over the whole thing. In other
words, a right-branching structure is appropriate.
It was seen earlier that the placement of clitics and the negative
causes a problem which is perhaps minor and reparable for Woolford's
analysis. But we are in more trouble with a VP like (19), for which a
right-branching structure seems counterintuitive. In contrast with (I1)
and (13) through (18), where the rightmost verb is the main verb, here
the main verb is the leftmost, with both i go and the completive or
anterior marker pinis modifying it. The appropriate analysis would seem
to be a left-branching one, as in (19').
(19) (19')
ron
run
A.
N
x pinis
ANT
igo / igo
away pinis ron away
ANT run
•had run off
8 / Gillian Sa Substrate and universals in t h e Tok Pisin verb phrase / 1 1 1
A very similar sentence is illustrated in (20), but in this case the main
verb karim 'to carry1 is transitive, so a phrase marker is needed along the
lines sketched there—again, a left-branching structure. Now note that
Woolford's original sentence (8), rediagrammed here in its essential
particulars as (8'), is very much like (20); however, it lacks the comple-
tive or anterior pinis, and contains instead the preverbal habitual marker
save. I would suggest that (8") is a more correct representation of
sentence (8): the habitual save, as in Woolford's analysis, takes scope
over the whole predicate, but go is analyzed as a postverbal modifier
rather than a main verb.
(20)
pinis
ANT
karim supia
carry spear
'had carried the spear away1
(80 x\ (8")
save
HABIT
sahm smok
send smoke
Looking back to Woolford's other sentences, one sees that in none of
them is the rightmost verb unequivocally analyzable as the main verb.
Sentence (8) is a good model for quite a number of sentences in my
corpus where in a three-verb chain, the middle one is the main verb.
In summary, we have seen that Woolford's analysis fails to indicate
that the topmost VP node is in any way different from later ones. We
have also found that since the main verb may be modified both on the
left and on the right, an analysis is needed that incorporates this fact in
some way.
2.2 Tense, mood, and aspect. Taylor (1960) and Thompson (1961)
noted that a range of Creole languages had three preverbal particles: a
past tense marker, a potential mood marker, and a durative aspect
marker. Bickerton (1974) changed the labels and their definitions
slightly, and offered an interpretation in terms of cognitive universals to
replace Taylor's and Thompson's essentially historical accounts. In
addition, Bickerton made a basic distinction between stative and non-
stative verbs, claiming that the unmarked interpretation of a nonstative
will be simple past. The first marker then becomes 'anterior' rather than
'past1, since it moves the unmarked reading back one stage to simple past
for statives, and to past-before-past for nonstatives. The 'durative'
aspect marker was relabeled 'nonpunctual1, to cover both continuative
and iterative. Finally, the 'potential' mood marker was redubbed 'ir-
realis', referring to futures, conditionals, etc. Bickerton's schema is, of
course, an idealization, and as such does not account in a detailed and
specific way for every occurrence or nonoccurrence of these markers in

discourse. Givon (1982), for example, pointed out that many unmarked
verbs in Bickerton's Hawaiian texts have neither of the two readings that
Bickerton's schema allows for them.
I will not, however, dwell here on issues of accountability. Rather, I
want to see to what extent the putative pancreole basic categories
structure the Tok Pisin tense, mood, and aspect system. Though Bicker-
ton specifically stated that Tok Pisin need not conform to the system he
presented, it is tempting to see how far we can go with it. There are,
however, some major obstacles. In several important respects, Bicker-
ton's system is not sufficiently detailed for us to use, even in its most
recent (1982) version. For one thing, he says nothing about negation,
except to suggest that negative concord may be basic, despite its non-
occurrence in Hawaii. It does not, for that matter, occur in Tok Pisin
either. But Bickerton is silent on the issues of placement and scope.
Further, he tells us nothing about modals, perhaps because they are too
variable in the Creoles which he examined. In any case, there is no way
in which modals can be omitted from the heart of the Tok Pisin system.
Anterior and habitual markers, clearly part of verb modification, com-
monly take scope over them, as was noted in (16).
As far as the semantic categories themselves are concerned, Tok Pisin
has two candidates for anterior: bin, which occurs first after negative in
the verb phrase, and pinis, which occurs at the very end, even typically
after the object NP if there is one. Irrealis is handled by bai, which, like
k> in Papiamentu, occurs before the verb phrase (although as described in
Sankoff and Laberge 1973, it has been moving closer to the verb over the
past 60 or 70 years). Bickerton's third category, nonpunctual, is shared
by the habitual marker save, which occurs preverbally, and the post-
verbal istap or igo, indicating continuative aspect.
Now what about placement? Muysken (1981) claimed that, theoreti-
cally, one could postulate two possible orders for the TMA markers
within universal grammar, that is, either TMAV or VAMT. However,
only the first seems to be attested in Creoles. Muysken's two possibili-
ties for representing TMA in universal grammar are given in (21).
(21a) TMA / VP, VP—>V...
(21b) AMT / VP_,VP—*...V
Muysken's two possibilities for representing TMA in attested Creoles are
shown in (22).
(22a) S (22b)
T M
Tok Pisin, as we have just seen, seems to have both left- and right-
branching verb modifiers. The two possibilities for Tok Pisin that
correspond to IWuysken's suggestions are presented as (23) and (24), and
(25) gives a linear representation of the major elements and their
ordering.
(23)
T T
bin pinis
"M"
ken; save; laik A Â
g° igo
kam ikam
stap istap
(25) T-M A A T
r go (igo )
stap I istapr > . .
save laik V
l <V2> ).. ( pinis
v
ikenj J kam (ikam)
L kamap
kirap
A few comments about this schema are in order, though a more de-
tailed presentation will appear in Sankoff (in preparation). First, it is
quite possible, and indeed frequently the case, that verbs occur without
any of the TMA modification described here. Aspectual markers on both
left and right are the most basic and most frequent, as in (3), for exam-
ple; followed by modals; with the 'tense' markers bin and pinis clearly
constituting the least important and least frequently marked grammati-
cal category. Second, with the partial exception of bin_ and ken, all of
the forms appearing in (24) and (25) can also occur as main verbs. Third,
since, as has been explained, the irrealis marker bai does not appear in
the verb phrase, the markers labeled 'M' are the modals ken and laik
(both of which, however, sometimes indicate future and both of which
can doubtless be considered as irrealis), as well as the habitual marker
save. Lastly, I am considering the set of verbs Woolford referred to as
'directional' as being basically aspectual in character. When used after a
verb of motion, igo and ikam certainly add direction to their marking of
continuity, but igo in particular is basically, like istap, a marker of
continuity, as illustrated in (26), where stap is the main verb, making
postverbal istap infelicitous as a marker of continuity. Note that post-
verbal igo can be iterated to convey a longer time period.
114 /Gillian Sankoff
(26) Mipela stap igo igo.

1PL-EXCL stay CONT CONT
'We stayed on (and on and on).1
Preverbal stap also conveys continuous aspect, and preverbal go, kam,
kamap, and kirap are inceptive.
Though having both left- and right-branching verb modification may
seem unlikely, it is consonant with several other aspects of Tok Pisin
grammar, as section 3 shows. But before discussing these typological
issues, let us examine several facts about clitics and about the distribu-
tion of _i- that support the analysis presented here.
2.3 The placement of _i-. In a paper given at the 1977 Georgetown
University Round Table on Languages and Linguistics (Sankoff 1977b), I
analyzed two successive waves of cliticization of the subject pronouns in
Tok Pisin. The first such wave washed j_ (originally the English third
person subject pronoun) right up into the verb phrase, where it has
remained ever since, being widely referred to in the literature as a
'predicate marker'. I prefer to regard it as a subject marker attached to
the auxiliary.
The postverbal markers virtually always appear as ikam, igo, and
istap. (The latter two may be iterated up to four or five times to mark
extremes of continuity and duration.) Thus _i_- is obligatory with the
postverbal aspectual markers. Though in other environments _i- is vari-
able, some fairly strong trends can be discerned with respect to its
distribution. The second most likely place for it to occur is the begin-
ning of the verb phrase. On the other hand, it is very unlikely to occur
before a main verb if that verb already has preverbal markers, unless the
main verb is go, kam, or stap, as in (14). It never occurs between two
main verbs: cf. (10) and (11) where, for example, (11') would be un-
grammatical.
(II1) *Meri i holim i pasim man bilong en.
This supports MUhlhSusler's analysis (1976) of this kind of 'serialization'

as belonging essentially to the lexicon. All of these facts suggest that _i-
is placed before the auxiliary, not the main verb. The right-hand auxil-
iary elements are marked more assiduously perhaps than the left because
they occur in a place which is normally filled by something else, i.e. a
noun phrase.
When j_ became a subject marker within the verb phrase, its place as a
subject pronoun was filled by em, derived from the English third person
object pronoun him (perhaps also them). But whereas the tendency to
copy noun subjects with pronouns, which was responsible for the original
cliticization of i, had also resulted in the cliticization of the other
subject pronouns, this did not happen again with em, leaving it in the
subject noun phrase. Evidence for this analysis is to be found in the
distribution of bai. Though it is variable in all positions, bai generally
occurs to the left of all the subject pronouns, as in (27a) and (27b), but
more frequently occurs to the right of full noun phrases and of em, as in
(27c) and (27d). Em is alone among subject pronouns in showing this
distribution, which Laberge and I observed a decade ago (Sankoff and
Laberge 1973) but were unable to interpret at the time.
(27a) Bai mi go.
IRR IS go
'I will go!
(27b) Bai ol i go.
IRR 3PL SM go
'They will go'
(27c) Em bai i go.
3S IRR SiM go
'He will go'
(27d) Tupela meri bai i go.
two girls IRR SM go
'Two girls will go1
3. Left and right in Tok Pisin. Verbal auxiliaries are not the only
things in Tok Pisin to show mirror image or nearly symmetrical struc-
tures. Nouns can have determiners on the left or right or both, as in
(28); relative clauses can be marked on the left or right or both (cf.
Sankoff and Brown 1976, Bradshaw 1979a), as diagrammed in (29); and
quoted speech can be bracketed on either or both sides, as in (30).
(28a) wanpela man 'a man'
(28b) man ia 'the man'
(28c) ol man (ia) 'the men'
(29) NPia [S ia|
(30) Na boi i tok, 'Mi no bairn smok', em i tok olsem.
and boy SM say IS NEG buy cigarette 3S SM say thus
'And the boy said, "I didn't buy cigarettes", so he said.
As an SVO language, one would expect Tok Pisin to have all four of
these—tense and aspect markers, determiners, relative clause, and
quoted speech introducers—on the left, like Tolai. Mosel's (1980) sche-
matization of the verb phrase in Tolai, an SVO Austronesian language, is
shown in (31).
(31) SM TMX TM2 Vi V2 V3 V^
New Guinea's Papuan languages, however, are strongly SOV, and tend to
have all of these on the right. Kaluli, for example, is a Papuan language
with typical VAT ordering (Schieffelin, personal communication).
Tok Pisin has something for everyone. It is not the case, however,
that those speakers with Papuan first languages have a version of Tok
Pisin grammar that does all its modifying on the right, and conversely
for those speakers with Austronesian first languages. Tok Pisin has not
adopted the cafeteria approach to grammatical expansion, nor is its
grammar simply a relexification of a number of very diverse sub-
strates. Its left- and right-branching structures have some parallels,
though, with languages in other areas where word order shift or contact
between SVO and SOV languages is occurring. Bradshaw (1979b) has

pointed to left- and right-bracketed negatives, relative clauses, and
serial causatives in a number of New Guinea languages; serial verbs and
left- and right-bracketed relative clauses also occur in a number of West
African languages. Such amalgams of 'typically SVO1 and 'typically
SOV features are not, however, established in a pidgin language by fiat,
overnight, by one or a number of speakers relexifying their native cate-
gories. Indeed, most of the features of Tok Pisin syntax that have been
analyzed in any detail have been shown to have taken many decades to
evolve. The amalgam, then, has involved forging new closed-class items
and negotiating where to put them during a period when the grammatical
glue of the original target, or superstrate, has dissolved. This has been
done through talk, and has required some common metalevel understand-
ings about communication, which I will call discursive practices. And
such discourse patterns are, I argue, the most likely things for New
Guineans of diverse language backgrounds to share. Along with some
widely held prosodic and semantic features, discursive practices are
better candidates for areal features than are the specifics of morphology
or syntax. What Grimes (1972) calls 'outlines and overlays' in narrative
conventions; rhetorical conventions used in the speech act known in Tok
Pisin as rabisim ('rubbishing', or shaming); the copious use of sentence-
final particles for asserting, seeking agreement, and so on; and clause-
chaining patterns whereby only the last clause is finite, are found in one
New Guinea language after the other.
My scenario for Tok Pisin runs as follows. It adopted SVO order not
because, as Bickerton would claim, it 'needed' word order for case mark-
ing, but rather because of the clear and almost invariant English pat-
tern. The open-class lexical items, largely of English origin, that eroded
into closed-class items through ever-increasing use as glue, slowly
became part of the grammar. They owed their initial placement to their
placement in English, but some of them (like bai) subsequently began
migrating to a place more befitting their emerging status. English
words, however, were also pressed into service to do rhetorical work:
reinforcing, asserting, questioning whether reference had been achieved,
and so on. Much of this work, in New Guinea languages, is done by
exploiting utterance-final position. Thus English words like the adverb
here or the conjunction ^r_ that typically occur at the right-hand end of
constituents or utterances, were enlisted, first on an ad hoc basis and
then as members of the regular army.
As I pointed out several years ago (Sankoff 1979), in every area of Tok
Pisin grammar that has been studied carefully, the grammar has been
reinvented, not copied. Routinization of discursive practices provides
the mechanism. And there is much that is shared about such practices,
across a wide variety of substrate languages in Papua New Guinea.
Substrate, then, enters at a metalevel, through discourse rather than by
the relexifying of grammatical specifics. And in a sense, universals have
to enter via the same quotidian path. Categories like the semantics of
determiners or of tense, mood, and aspect may typically involve some
basic, putatively universal characteristics, but they will interact with
other elements in response to processing constraints involving things like
ordering and memory, the input constraints provided by the superstrate
model, and the subtle shaping of socially transmitted discursive
practices.
Notes
This paper was begun in Vancouver, where I was a visitor in the Lin-
guistics Department at the University of British Columbia, and finished
in Sao Paulo. I thank my colleagues at UBC, especially David Ingram,
Derek Nurse, Clifton Pye, and Michael Rochemont, and the student
members of my seminar, for many helpful discussions. Talmy Givon,
Bambi Schieffelin, and Fernando Tarallo contributed support, encour-
agement, and many fine insights. None of them, of course, is responsible
for any errors that remain. I am grateful also to the many Papua New
Guineans who helped me learn Tok Pisin and who worked with me during
a number of visits to Papua New Guinea between 1966 and 1977. Lastly,
I want to thank my mother, Marjorie Topham, the staff of the Lilliput
Day Care Centre at UBC, and Ruth Elisabeth Lopes Moino of Sao Paulo
for looking after Alice so well while I was working on this paper.
1. Superstrate influence is clearly not a serious contender as far as
the morphology and syntax of Tok Pisin are concerned.
2. Note Benveniste's Latin motto: Nihil est in lingua quod non prius
fuerit in oratione (1966:131).
3. In the late 1960s, I frequently heard Buang fathers joking with their
small children in Tok Pisin. It seemed that what made this joking espe-
cially funny was the use of Tok Pisin itself.
4. Even today, more than twice as many men as women speak Tok
Pisin (Sankoff 1977c:294).
5. Tok Pisin was used as a lingua franca in German New Guinea.
According to Firth, 85,000 would be a conservative estimate of the
number of New Guineans 'who went as indentured labourers to planta-
tions in German New Guinea from villages within that colony' (Firth
1976:51).
6. Mosel (1980) appears to be correct in proposing Tolai in particular,
and the Tolai-Patpatar language group in general, as the most important
potential substrate source during the formative period of Tok Pisin's
development.
7. Most authors who have done comparative work on Creoles, includ-
ing Bickerton, have noted this feature. Bickerton's (1981) explanation
for it in terms of a 'need' to mark oblique cases is, I think, totally mis-
guided and appears to neglect the existence of prepositions in all known
Creoles.
8. Woolford, however, explicitly rejects an analysis of a sentence like
(10) as involving any kind of reduced sentential complement.
9. Though the markers occur only when expressing the particular
semantic function with which each is associated, their ordering with
respect to each other and to the main verb is always fixed, thus (Tense)
(Modal) (Aspect) Verb.
10. Since the point of Woolford's (1977) analysis was to provide a
phrase structure for serial verbs, rather than concentrating on details of
negative and clitic placement, it is not clear how she would have solved
the problems their placement would pose.
11. Bradshaw's (1979b) seminal paper on some of the consequences of
SOV and SVO contacts in the Huon Gulf area of Papua New Guinea was
an important influence in my thinking about the Tok Pisin verb phrase. I
do not, however, agree with his view (1979a) that the double-bracketed
118/Gillian Sankoff
relative clauses in Tok Pisin have their origin in the syntax of any other
language(s).
References
Benveniste, Emile. 1966. Problemes de linguistique generale I. Paris:
Gallimard.
Bicker ton, Derek. 1974. Creolization, linguistic universals, natural
semantax and the brain. University of Hawaii Working Papers in
Linguistics 6.3:125-41.
Bickerton, Derek. 1981. Roots of language. Ann Arbor, Mich.: Karoma
Publishers.
Bradshaw, Joel. 1979a. The origins of syntax in syntax: Tok Pisin
relatives reconsidered. University of Hawaii Working Paper in Lin-
guistics.
Bradshaw, Joel. 1979b. Causative serial constructions and word order
change in Papua New Guinea. Paper presented at the annual meeting
of the Linguistic Society of America, Los Angeles, December.
Firth, S. 1976. The transformation of the labour trade in German New
Guinea, 1899-1914. Journal of Pacific History 11.51-65.
Franklin, Karl J. 1980. The particles _i- and ria in Tok Pisin. Kivung
12.2:134-44.
Gilliam, Angela M. 1984. Language and 'development' in Papua New
Guinea. Dialectical Anthropology 8.303-18.
Givon, Talmy. 1982. Tense-aspect-modality: The Creole prototype and
beyond. In: Tense-aspect, between semantics and pragmatics. Edited
by Paul J. Hopper. Philadelphia: John Benjamins.
Grimes, Joseph. 1972. Outlines and overlays. Lg. 48.513-24.
Hall, Robert A., Jr. 1943. Melanesian pidgin English: Grammar, texts,
vocabulary. Baltimore: Linguistic Society of America.
Laberge, Suzanne, and Gillian Sankoff. 1979. Anything you can do. In:
Discourse and syntax. Edited by Talmy Givon. New York: Academic
Press. 419-40.
Laycock, D. C. 1970. Materials in New Guinea pidgin (coastal and
lowlands). Pacific Linguistics D-5.
Mihalic, Rev. F. 1971. The Jacaranda dictionary and grammar of Mel-
anesian pidgin. Port Moresby: Jacaranda Press.
Mosel, Ulrike. 1980. Tolai and Tok Pisin. The influence of the sub-
stratum on the development of New Guinea pidgin. Pacific Linguistics
Series B.
MUhlhSusler, Peter. 1976. Growth and structure of the lexicon of New
Guinea pidgin. Ph.D. dissertation, Australian National University,
Canberra.
MUhlhMusler, Peter. 1978. Samoan Plantation Pidgin English and the
origin of New Guinea pidgin. Pacific Linguistics Series A.54:67-119.
Muysken, Pieter. 1981. Creole tense/mood/aspect systems: The un-
marked case? In: Generative studies on Creole languages.
Dordrecht: Foris Publications. 181-99.
Sankoff, Gillian. 1971. Language use in multilingual societies: Some
alternate approaches. In: Sociolinguistics. Edited by J. Pride and J.
Holmes. Harmondsworth, Middlesex: Penguin. 33-51.
Sankoff, Gillian. 1977a. Creolization and syntactic change in New

Guinea Tok Pisin. In: Sociocultural dimensions of language change.
Edited by M. Sanches and B. Blount. New York: Academic Press.
119-30.
Sankoff, Gillian. 1977b. Variability and explanation in language and
culture: Cliticization in New Guinea Tok Pisin. In: Georgetown
University Round Table on Languages and Linguistics 1977. Edited by
Muriel Saville-Troike. Washington, D.C.: Georgetown University
Press. 59-73.
Sankoff, Gillian. 1977c. Multilingualism in Papua New Guinea. In: New
Guinea area languages and language study, vol 3: Language, culture,
society and the modern world. Edited by S. A. Wurm. Fascicle 1, 265-
307. Pacific Linguistics C-40.
Sankoff, Gillian. 1979. The genesis of a language. In: The genesis of
language. Edited by K. Hill. Ann Arbor, Mich.: Karoma Publishers.
23-47.
Sankoff, Gillian. 1980. The social life of language. Philadelphia:
University of Pennsylvania Press.
Sankoff, Gillian, (in preparation) The Tok Pisin verb phrase.
Sankoff, Gillian, and Penelope Brown. 1976. The origins of syntax in
discourse: A case study of Tok Pisin relatives. Lg. 52.631-66.
Sankoff, Gillian, and Suzanne Laberge. 1973. On the acquisition of
native speakers by a language. Kivung 6.32-47.
Taylor, D. 1960. Language shift or changing relationship? UAL 26.144-
61.
Thompson, R. W. 1961. A note on some possible affinities between the
Creole dialects of the Old World and those of the New. In: Proceed-
ings of the Conference on Creole Language Studies. London: Mac-
millan.
Woolford, Ellen. 1975. Variation and change in the ± predicate marker
in New Guinea Tok Pisin. Paper prepared for the 1975 International
Conference on Pidgins and Creoles, Honolulu.
Woolford, Ellen. 1977. Aspects of Tok Pisin grammar. Ph.D. disserta-
tion, Duke University.
Wurm, S. A. 1971. New Guinea Highlands Pidgin: Course materials.
Pacific Linguistics D-3.
THE PRAGMATICS OF REFERENTIALITY
Talmy Giv6n
University of Oregon, Eugene
1. Reference and existence. The treatment of referentiality (and
coreference) in linguistics began as a by-product of a purely logical
approach, whereby—to use the most extreme logical positivist formula-
tion—terms have reference if there exists some entity in the so-called
Real World (cf. Carnap 1958 or Russell 1905, inter alia). Thus, in (la),
the term the king of France does not refer, because there is no king in
France, while the queen of England does refer because there is a queen
in England.
(la) The king of France is bald,

(lb) The queen of England is bald.
A more realistic linguistic approach would concede that what is at issue
is not whether terms refer to individuals existing in 'the Real World', but
rather whether they refer to those existing in some 'universe of dis-
course'. Thus, in (2), both (2a) and (2b) are equally referring expressions,
regardless of the likelihood that in the so-called Real World the subject
of (2a) does not exist while that of (2b) does.
(2a) There was once a unicorn. The unicorn loved lettuce.
(2b) There was once a rabbit. The rabbit loved lettuce.
By using the term 'universe of discourse', one could, of course, perpe-

tuate the logic-bound tradition of viewing linguistic expressions as
entities on their own (i.e. propositions or their concatenations), rather
than as communicative acts performed by some speaker for a particular
purpose. One consequence of adopting a more pragmatic approach to
the problems of reference is that 'existing in some universe of discourse1
in fact readily translates into 'meant by the speaker to have reference in
some universe of discourse'. While this intensional reformulation of
'have reference 1 is indeed pragmatic—rather than logico-semantic—in
some obvious sense, it nonetheless retains residual logico-semantic
120
The pragmatics of referentiality / 121
characteristics. The chief one of these is that the referential properties
of terms can be determined strictly within the atomic proposition in
which they are embedded, and without recourse to a wider extraproposi-
tional context.
2. Reference and propositional modalities. The logico-semantic

tradition is virtually silent about the systematic connections revealed in
language between referential properties of NPs and the propositional
modalities within whose scope those NPs reside. A brief summary of
these regular correlations follows. Modern logicians recognize four
major propositional modalities (cf. Carnap 1947), shown in (3).
(3a) Necessary truth (true analytically, by definition)

(3b) Factual truth [true synthetically, as facf]
(3c) Possible truth [possibly true]
(3d) Nontruth [falsf]
In human language it is possible to show that these same modalities may
be redefined in terms of the communicative contract between speaker
and hearer, as respectively (cf. Giv6n, 1982a):
(4a) Presupposition [true by agreement/convention;

speaker assumes that (i) the
hearer will not challenge this
proposition, and (ii) no evidentiary
support is required)
(4b) realis-assertion [strongly asserted as true; speaker
assumes that the hearer may
challenge this proposition, and is
prepared to back it up with evi-
dentiary support^
(4c) irrealis-assertion [weakly asserted as true; speaker
is offering the proposition as
hypothesis, expecting strong
challenge and having little evi-
dentiary support]
(4d) NEG-assertion [strongly asserted as false; with
similar stipulations as in (4b)3
For the purpose of their systematic effect on the referentiality of
NPs within their scope in human language, the four propositional modal-
ities just given may be grouped into two larger macromodalities—fact
and nonfact, as shown in (5).
(5a) FACT: presupposition and realis-assertion

(5b) NONFACT: irrealis-assertion and NEG-assertion
The correlation between propositional modalities and the referentiality
of NPs under their scope can be then summed up as shown in (6) (cf.
Givon 1973, 1983a:Ch.ll).
122 / Talmy Givdh
(6) NPs may be nonreferential only when they are under the
scope of some NONFACT modality. Otherwise they must be
referential.
There are many sources in human language for the NONFACT modality,
some taking the entire proposition under their scope, others taking only
portions (such as the object or a sentential complement) under their
scope. At this point I will not enumerate all of those, but rather illu-
strate the generalization (6) with a few representative examples, using
only indefinite NPs.
(7a) FACT, presupposition:

Mary regretted that she saw an eagle.
QDMary saw a particular eagle, one that the
speaker is committed to assert: 'It does existD
(7b) FACT, realis-assertion:
Mary saw an eagle.
CDMary saw a particular eagle, one that the
speaker is committed to assert: 'It does exist']
(7c) NONFACT, irrealis-assertion:
Mary will see an eagle.
(i) QpMary will see a particular eagle, one that the
speaker is committed to assert: 'It does existO
(ii) Cz> Mary may see some member of the type 'eagle1,
though the speaker does not have any particular
one in mind]
(7d) NONFACT, NEG-assertion:
Mary did not see an eagle.
Cl)The speaker does not have in mind any particu-
lar eagle that Mary didn't see, rather, he asserts
that she did not see any member of the type
'eagle']
When I first discussed the systematic correlation between proposi-

tional modalities and referentiaiity properties of NPs in human language,
one of the strongest supports I observed for such a formulation was the
distinct morphological systems marking referential vs. nonreferential
arguments in many human languages. In some, this is the major distinc-
tion within the article system, disregarding the further contrast of
definiteness. Thus, consider examples (8)-(10) from Bemba (GivoYi 1973).
(8) FACT, REALIS-ASSERTION:

(8a) a-a-somene ici-tabo [jREF]
he-past-read REF-book
f(i) 'He read the book' [DEF] \
l(ii) 'He read a book' [INDEP])
(8b) *a-a-somene ci-tabo QNONREF]
he-past-read NREF-book
(9) NONFACT, IRREALIS-ASSERTION:

(9a) a-ka-soma ici-tabo (REF]
he-future-read REF-book
((i) 'He will read the book' (DEg I
\(ii) 'He will read a particular book' [iNDEf] /
(9b) a-ka-soma ci-tabo [NONREFJ
he-future-read NREF-book
'He will read some—unspecified—book'
(10) NONFACT, NEG-ASSERTION:

(10a) ta-a-a-somene ici-tabo [REQ
NEG-he-past-read REF-book
'He didn't read the book' (DEFj
(10b) ta-a-a-somene ci-tabo [NONREFJ
NEG-he-past-read NREF-book
'He didn't read a/any book' [INDEFJ
In some other languages, such as Spoken Israeli Hebrew, Chinese,

Turkish, Sherpa, Persian, all Creoles, and many others, the numeral 'one'
is used to mark referential-indefinite arguments, while nonreferential
indefinites are most commonly left unmarked. Thus consider example
(11), from Hawaii English Creole (Bickerton 1975, 1981).
(lla) He read one book [FACT, REFJ

he read-past REF book
'He read a book'
(lib) *He read book [FACT, *NONREFj
(lie) He go-read one book [NONFACT, REF]
he future-read REF book
'He will read a particular book'
Old) He go-read book [NONFACT, NONREFJ
he future-read book
'He will read some—unspecified—book'
(lie) He no-read book [NEG, NONREE)
he NEG-read-past book
'He didn't read a/any book'
(llf) *He no-read one book [*NEG, REFj
The facts as described so far seemed rather clean and uncontroversial
at the time, so that once 'semantically referential' was recast in terms
of the speaker's referential intent, morphology in one language after
another seemed to cooperate and thus uphold such a formulation (see
e.g. discussion in GivoYi 1975/1979). Soon after, however, the ground
suddenly shifted, when I noticed that in Hebrew one could in fact get
semantically referential indefinite NPs to appear without the numeral
'one' under the scope of incontrovertibly FACT modalities. To illustrate
this possibility, consider (12a-b) (originally from Givon 1976/1981; see
further discussion in Givon 1982b, 1983a:Ch. 11).
(12a) ...az yatsati la-rexov ve-nixnasti le-xanut sfarim

so I-exited to-the-street and-I-entered in-shop-of books
'...So I went out and entered a bookstore
124 / Talmy Givdn
ve-kaniti sefer-xad she-himlitsu li alav meod,
and-I-bought book-one that-they-recommended to-me on-
rt verv
and I bought a book that was recommended to me highly,
ve-ratsti habayta ve-hitxalti likro oto

and-I-ran home and-I-started reading rt
and I ran home and started reading rt,
ve-ze_haya beemet sefer metsuyan...

and-it was truly book excellent
and rt was indeed an excellent book...'
(12b) ...az yatsati la-rexov ve-nixnasti le-xanut sfarim
so I-exited to-the-street and-I-entered in-shop-of books
'...So I went out and entered a bookstore
ve-kaniti sefer ve-axar-kax ratsti habayta

and-I bought book and-after-that I-ran home
and bought a book and then I hurried home
ve-hitraxatsti ve-axalti ve-halaxti lishon...

and-I-washed and-I-ate and-I-went to-sleep
and washed and ate and went to sleep...'
Semantically, both (12a) and (12b) involve a realis-assertion—FACT—
about buying 'a book1, so that logically that particular book must have
existed. However, the referential-indefinite marker 'one' marks that
book only in (12a), where the particular referential identity mattered in
the subsequent discourse, and where 'book' keeps recurring as topic in
the subsequent discourse. In contrast, in (12b) the specific referential
identity of the book seems not to matter; rather, the speaker 'did some
book-buying' and then went home, with the particular book that was
bought never mentioned again.
In assessing what such facts mean (GivoVi 1982b), I suggested that
semantic referentiality is merely a special case of a more general phe-
nomenon, pragmatic referentiality. It is, of course, not an accident that
entities that tend to individuate strongly and have easily recognizable
'semantic' referential identity also tend to be more important—and
recurrent—in discourse. Similar generalizations have been noted by
Hopper and Thompson (1980, 1983). Thus, as I demonstrate shortly, the
majority of arguments that persist longer in discourse are also semanti-
cally referential. But such a correlation is not absolute, and two types
of crucial nonoverlaps are of great interest.
(i) As in (12b), the semantically referential NPs are nonetheless
marked by the characteristic morpheme that tends to mark
nonreferential NPs, just in case they are nonrecurrent and
unimportant for the subsequent discourse.
(ii) Equally interesting are instances where semantically nonref-
erential NPs are nonetheless marked by the characteristic
referentiality marker 'one' in those rare cases where they
are of great thematic importance in the subsequent dis-
course.
In section 3 I report briefly on work done in two text studies, by Coore-
man (1982, 1983) and Rude (1983), working on the discourse-conditioning
of the antipassive construction in two ergative languages, Chamorro and
Nez Perce, respectively.
3. Semantic referentiality and discourse persistence. It has been
common knowledge that objects in the antipassive construction tend to
be nonreferential, low on individuation (i.e. plural), low on topicality (i.e.
indefinite), etc. (see e.g. discussion in Hopper and Thompson 1980;
Kalmar 1979, 1980; Heath 1976, inter alia). What Cooreman has shown
in her text-based study of Chamorro is that objects in the antipassive
construction are lowest in topicality as compared to objects of the more
common transitive ergative construction or to the patient-subjects of
the passive. 'Topicality' was measured both anaphorically and cataphori-
cally, and objects of the antipassive tend to be both primarily indefinite,
i.e. have no anaphoric antecedence in discourse, and lowest in persist-
ence in subsequent discourse. The comparison is summarized in Table
1. The measure of referential distance (RD) expresses anaphoric topi-
cality, with a high value meaning low topicality. The measure of topical
persistence (TP) expresses cataphoric topicality, with a high value
meaning high topicality.
Table 1. Topicality properties of direct objects in Chamorro
discourse.
Object type Average RD Average TP
in ERG-clause 0.81
in AP-clause 20.00^ 0.00
in IN-passive 1.38 2.00
in MA-passive 3.33 iM
Cooreman (1982) also notes that the patient in the antipassive construc-
tion does not appear at all—i.e. is 'deleted nonanaphorically'—70 percent
of the time, and further, that the remaining 30 percent are predomi-
nantly semantically nonreferential.
Almost identical results are reported for Nez Perce by Rude (1983).
The comparison between the topciality measurements of patient-objects
in the ergative-transitive, antipassive, and passive constructions are
given in Table 2.
Table 2. Topicality properties of direct objects in Nez Perce

discourse.
Object type Average RD Average TP
in ERG-clause 5723 2753
in AP-clause 13.86 0M
in PASSIVE-clause ZM L71
In the Nez Perce text studies by Rude (1983), out of the total of 51
objects in the antipassive construction, 30 (or 59 percent) were indefi-
nite, mostly full NPs (29 out of the 30), and most of those semantically
nonreferential. While the percentages are not identical to those in
Chamorro, the general direction is the same.
126 / Talmy Givon
4. Pragmatic referentiality and major participants in Krio narra-
tives. Krio is an English-based Creole language spoken in Sierra Leone.
Its article system conforms to the normal Creole pattern described by
Bickerton (1975, 1981), with a contrast between di 'the' (or dis 'this' and
_â_ 'that') marking definites, wan 'one' marking referential indefinites,
and the zero form marking nonreferentials (see example (11) for the
Hawaii Creole illustrations). In this study, I have analyzed four stories
taken from Hancock (1972). They are all short stories; the first, a joke,
is given in section 4.1. In these stories, major participants are intro-
duced for the first time as either REF-INDEF NPs marked by wan or
names, or relational-definite NPs ('his wife', 'her brother' etc.). I have
chosen only stories where at least some of the major participants are
introduced as REF-INDEF NP, i.e. with wan. For each story, I divided
all NPs/topics into two groups, major and minor participants. The major
ones in these stories are either wan-marked or relational definites. The
minor 1
participants are either relational definites ('his neck1, 'the win-
dow ) or zero-marked, with the majority of the latter being semantical-
ly NONREF, but an interesting minority being semantically referential.
Since the first story is short, I may as well illustrate the text analysis
methodology used in this study by reproducing the story in its entirety.
*.l Story 1.
Na Kamfoh Wi Kam foh!
'It's Camphor That We Came for'
wan-dey-ya, tu-pikin denh gram sen denh na wan mamf inh shap
one-day-here two child their grandma sent them to one woman her shop
'One day the grandmother of two children sent them to one woman's
shop'
fo g6 bdi kcimfoh. Wey denh rich dey, di mamf aks denh
for go buy camphor when they reach there the woman ask them
'to go and buy some camphor. When they arrived there, the woman
asked them'
weytin denh kam foh; na-dey di pikm-denh sey, denh don-foget.
what-thing they come for TOP-they the child-they say they done-forget
'what they came for; so those children said that they had forgotten.1
di mamf kol ol di tin na di shap; owri, siminji, etc., denh til no memba.
the woman call all the thing LOC the shop butter, clove, etc. they still
NEG remember
'The woman called (the names of) all the things in the shop: butter,
cloves, etc., but they still couldn't remember.1
di pikfn-denh jfs dey-say, denh kam foh..., denh kam foh...
the child-they just PROG-say they come for they come for
'The children just kept saying, they came for...they came for...'
na-fnh di mami taya, en sey mek denh go owm en aks denh granf
TO-him the woman tired and say make they go home and ask their
grandmother
'Well, so that woman got tired, and she told them they should go home
and ask their grandmother'
weytin denh kcfm foh. As denh go owm nomoh, na-fnh wan di pikih-denh
memba,
what-thing they come for as they go home nomore TO-him one the child-
they remember
'what they were supposed to buy. As they went home, one of the child-
ren remembered,
en ron go bak na di shap en sey, 'Na-kamfoh wi kam foh, Ma!1

and run go back LOC the shop and say 'FOC-camphor we come for,
Ma'ame'
'so he ran back to the shop and said: "It's camphor that we came for,
Ma'ame!1"
The recurrence pattern of all participant NPs in this story is given in
Table 3.
Table 3. NP marking on first appearance in discourse, discourse

importance status, and recurrence in subsequent discourse
for Krio story 1.
NP and its marking Semantic Importance Total occur-
at 1st occurrence REF-status in discourse rences in
discourse
tu pikin Kbh-lNDbh major topic 18
'two children'
wan mamf REF-INDEF major topic 6
'a woman'
wan di pikin REF-INDEF major topic 3
'one of the children'
denh granf REL-DEF minor, related 2
'their grandmother' to major
inh shap REL-DEF minor, related 3
'her shop' to major
kamfoh NONREF minor, but 2
'camphor' thematically
important
owri NONREF minor 1
'butter'
siminji NONREF minor 1
'cloves'
As one can see, the correlation between the use of wan (or _tu) to intro-
duce an NP into the discourse and the discourse status of that NP is
rather striking. Only major participants, ones likely to recur through the
story more frequently, are introduced by the numeral-article. And all
the bare-stem NPs are both semantically nonreferential and
128 / Talmy Givon
pragmatically unimportant in the discourse—and thus recur little if at all
after their initial entry into the discourse.
<L2 Story 2. Wan Brokow Ows 'A Tumbledown House.' The second
story is somewhat longer. Table k presents an analogous analysis of its
correlations. In this story we meet one more marking device of Krio,
the suffixal -ya(-so) 'here...so', which is used to mark only important
participants in the discourse when they are reintroduced, as full definite
NPs, after a certain gap of absence.
Table *. NP marking on first appearance in discourse, discourse
for Krio story 2: Wan Brokow Ows 'A Rundown House1.
discourse
a speaker; major
T REF-DEF
wan dey dey REF-INDEF major 13 (22)
'one there1 (pronoun)
wan patikla wan 1
'a particular one
di trit REF-DEF minor 3
'the street'
wan eykuru dog REF-INDEF minor 2
'a mangy dog'
yayam NONREF minor 1
'food'
di doti-boks DEF-REF minor 1
'the trash cans'
di oda pat DEF-REF minor, referring 1
'the other side' to 'street'
ows REF-INDEF minor, prelude to 2
'houses' major theme
denh ows-ya REF-DEF minor, prelude to
'those houses-here' major theme
broko-ows denh NONREF minor, but at- 1
'rundown houses-they' tributive of
major theme
di wol-denh REF-DEF minor, part of 1
'the walls' major
di winda REF-DEF minor, part of 2
'the window' major
winda-blain NONREF minor, part of 1
'curtains' major
da fain-fain gadin REF-DEF minor, part of 1
'that beautiful garden' major
di geyt REF-DEF minor, part of 1
'the gate' major
di do REF-DEF minor, part of 1
'the door' major
Table 4 (continued).
(I) (2) 13J W
di step REF-DEF minor, part of 1
'the steps' major
ol di bowd REF-DEF minor, part of 1
'all the boards' major
gras-gras NONREF minor, attrib- 1
'grassy' utive
maskita REF-INDEF minor 1
'mosquitos'
gows NONREF minor 1
'ghosts'
eni noiz NONREF minor 1
'any noise'
spirit NONREF minor 1
'spirits'
titi NONREF minor, attrib- 1
'girl' utive of major
wan owl mami NONREF minor; attrib- 1
'an old lady' utive of major;
appear at thematic
peak
wan mi fambul 1 NONREF minor; attrib- 1
'a family member utive of major;
appear at thematic
peak
wan big noiz REF-INDEF minor; thematic 1
'a big noise' peak of story
Key: U) NP and its marking at first occurrence; {2) semantic REF-
status; (3) importance in discourse; (4) total occurrences in discourse.
Of the two major topics/participant NPs in this story, one is the

speaker, coded as 'I1; the other, 'house' is coded as REF-DEF by the
numeral wan. A host of minor participants are introduced as definite
parts of a major participant, the house, and one may argue that in fact
those constitute recurrences of the whole i.e. 'house' (cf. Linde 197*0.
Three types of interesting exceptions to the Creole morphological mark-
ing rules can be observed.
(i) Semantically referential NPs coded by bare-stem form. Two out
of the total of nine bare-stem NPs—'houses' and 'mosquitos'—are seman-
tically referential. However, they are thematically minor participants
and low in individuation (cf. Hopper and Thompson 1980, 1983)—in fact,
plurals.
(ii) Semantically nonreferential NPs marked with the numeral wan.
There are two instances of these: 'an old woman' and 'a family mem-
ber'. Both are similes predicating the major topic of this story, 'house1.
And both come at the very thematic peak of the narrative, just prior to
the old house crashing down.
(iii) Thematically nonrecurring—minor—NPs marked with wan. There
are two of these, 'a mangy dog1 and 'a big noise', and one of them—the
latter—appears at the thematic peak of the story.
130 / Talmy Givon
These exceptions—with only 'a mangy dog1 remaining problematic —in

fact illustrate the underlying pragmatic nature of marking referentiality
by the numeral wan in Krio:
(a) referential indefinites that are unimportant in the discourse
do not get marked by wan in spite of their semantic refer-
entiality;
(b) nonreferentials that are of central thematic importance in
the narrative are marked by wan regardless of their semantic
status as nonreferential;
(c) finally, although there is a strong correlation, statistically,
between discourse importance of participants and their fre-
quency of recurrence in the discourse, the bottom line is that
thematic centrality overrides considerations of mere fre-
quency.
One last point concerns the use of the suffixal -ya(-so) in this story. It
appears only twice, in pan denh ows-ya 'among those houses' and dis ows-
ya 'this house'. Its appearance is thus limited to the central character or
to the group of which the central character is member. As will be seen
later on, in the two other—longer—stories, -ya(-so) is used with much
higher frequency to reintroduce central characters into the discourse.
*.3 Story 3. Tif Tif, God Laf 'A Thief Steals, God Laughs'. The third
story is still longer. The distribution of its NP participants, their mor-
phological marking, their semantic referential status, and their discourse
status and overall frequency, are all given in Table 5.

for Krio story 3: Tif Tif, God Laf 'A Thief Steals, God
Laughs'.
NP and its markine Semantic Importance Total occur-
discourse
Major participants:
wan man REF-INDEF major 74
'a man'
wan fam REF-INDEF major 12
'a firm'
inh bos onh pikin sef REL-DEF major 29
'the boss's own son'
plenty moni REF-INDEF major 18
'lots of money' (mass noun)
wan kobod REF-INDEF major 5
'a cupboard'
Minor participants:
inh wok REL-DEF minor, related 1
'his work' to major
inh bodi REL-DEF minor, part of 1
'his body' major
U) •12) W
inh main RfcL-DEF minor, part of 1
'his mind' major
inh grachuiti REL-DEF minor, related 1
'his gratuity to major
inh at REL-DEF minor, part of 1
•his heart' major
di sai REF-DEF minor.loca- 1
'the place' tional 8
son sai REF-INDEF minor 1
'some place'
son denh big man-denh REF-INDEF 9 minor 1
'some of the important
men1
maneyja NONREF minor; attribute 1
'manager' of major
grachuiti NONREF minor 1
'gratuity'
meseynja NONREF minor; attribute 1
'messenger' of major
pey NONREF minor 1
'pay'
penshon NONREF minor 1
'pension'
bris NONREF minor 1
'breeze'
wowk NONREF minor 1
'work'
ol sai REF-INDEF minor 1
'every place'
no sai NONREF minor 1
'nowhere'
ol man NONREF minor 1
'everybody'
Key: U) NP and its marking at first occurrence; (2) semantic REF-
status; (3) importance in discourse; (4)"total occurrences in discourse.
Of the five major participants in this story, three are introduced

initially with wan, a fourth with the quantifier 'plenty', since it is a mass
noun, and the fifth is relationally definite, thus related to 'firm' via 'the
boss'. Not a single minor participant is introduced by wan.
All minor participants appear only once in the discourse. They are
further divided into two major types: (1) relational-definite referential
ones, largely related to either major participants or to the location, thus
indirectly to the major participant 'firm'; or (2) zero-marked NPs, over-
whelmingly nonreferential semantically, and only one marginally refer-
ential ('everywhere'). These facts again underscore the use of wan to
introduce only referential arguments that are of major importance for
the discourse.
132 / Talmy Givdn
Only two NPs are marked by the suffixal -ya-so. Both are major
participants, the messenger (dis meseyna-ya-so, dis man-ya-so) and the
money (dis moni-ya-so). The main participant in the story, the messen-
ger, appears 13 times in the story as full definite NP. Of these, 12
occurrences are marked with -ya-so. The money, the other major
participant, occurs 10 times in the story as full definite NP. Of these,
only 2 are marked with -ya-so. The correlation between the use of -ya-
so to reintroduce a topic back into the discourse and the thematic
importance of that topic is again rather striking.
bA. Story 4. Weytin Du No foh Tros Uman 'Whatever you do, Don't
Trust a Woman'. Our fourth story is the longest. The same analysis done
for the first three stories is presented, for this story, in Table 6.
for Krio story 4: Weytin Du, No foh Tros Uman 'Whatever
You Do, Don't Trust a Woman'.
discourse
Major participants:
wan big man REF-INDEF major 66
'an old man'
inh wef REL-DEF major 29
'his wife1
wan vileyj REF-INDEF major 5
'a village'
wan oda vileyj REF-INDEF major, related 1
'another village" to major
tri jownk REF-INDEF major 3
'three braids'
tri minin REF-INDEF 12
'three meanings'
di vileyj inh chif REL-DEF major, related 14
'the village's chief to major
wan switat REF-INDEF major 11
'a lover'
wan pikin REF-INDEF major 17
•a child'
wan lapa REF-INDEF major 6
'a wrapper'
Minor participants:
dis pikin inh pipul denh REL-DEF minor, related 2
'this boy's relatives' to major
son di ia na inh eyd REF-INDEF minor, part of 1
'some of the hair on his head' major
inh eyd REL-DEF minor, part of 2
'his head' major
ol man REF- minor, related I

'all the men1 GENERAL to major (village)
blok REF-INDEF minor 2
'cell'
inh nek REL-DEF minor, part of 3
'his neck1 major
dis man inh sikrit REL-DEF minor, related 2
'this man's secret' to major
uman NONREF minor, but related 2
'woman' to major theme
slip NONREF minor, but related 2
'sleep' to major theme
pipul-pikin NONREF minor, but related 2
'other people's child' to major theme
Key: (1) NP marking at first occurrence; (2) semantic REF-status; (3)
importance in discourse; (4) total occurrences in discourse.
The correlation between morphological marking with wan and the

importance of an NP participant in discourse is again obvious in this
story. In most cases this is also reflected in the frequency of occurrence
of the NPs. In one exceptional case—wan oda vileyj 'another village1—
wan participates in a fixed expression for which there is no easily acces-
sible morphological alternative, given that oda vileyj could only mean, as
indefinite, the plural 'other villages', but never the singular 'another
village'. Another apparent exception, 'three braids', with only three
occurrences in the story, is thematically of critical importance.
Of the bare-stem NPs, two are technically semantically referential,
'all the men' and 'cell1. The first may actually be considered anaphoric-
referring, and thus definite. It is also of obviously low individuation (cf.
Hopper and Thompson 1980, 1983). And the second, 'cell', is clearly
unimportant in the story, so that its thematic unimportance overrides its
semantic referentiality, and it is not marked by wan.
The distribution of the uses of -ya-so as marking reintroduced definite
NPs in this story is given in Table 7.
Of the 10 major participants listed in Table 7, 7 get subsequently
marked upon reintroduction as full DEF-NPs—i.e. when reintroduced into
the discourse following a considerable gap of absence or a thematic
break—by -ya-so, at least once. Not a single minor participant ever gets
reintroduced as full NP with -ya-so. Finally, of the 7 participants re-
introduced at least once with -ya-so, 6 were introduced originally by a
numeral, either wan or tri. While these correlations are not absolute,
they are nonetheless again striking. They illustrate how the grammar, of
at least this Creole language, not only marks major participants in a
special way upon first introduction into the discourse, but also upon
subsequent reintroductions.
13* / Talmy Givon
Table 7. Distribution of the use of -ya-so in marking reintroduced

full DEF-NPs in Story k.
Participant NP Importance Total occurrences:
status As full With -ya-so
DEF-NP
denh tri jownk-ya-so important 1 1 UUU % ;
'those three braids'
di tri minin-ya(-so) important 10 1 (40 96)
'those three meanings'
dis man-ya-so inh wef important 10 2 (20 %)
'this man's wife'
dis switat-ya-so important 2 1 (50 96)
'this lover'
dis man-ya-so important 18 8 (W 96)
'this man'
na inh dis chif-ya-so important 8 1 (12 %)
'it was this chief
dis lapa-ya-so important k 1 (25 96)
'this wrapper1
5. Discussion
5.1 Semantics vs. pragmatics of referentiality. It is perhaps not an
accident that in a broad way, the history of the treatment of referen-
tiality by philosophers, logicians, and linguists follows a similar course as
the history of the treatment of propositional modalities. For the latter,
one began with strictly deductive approaches to isolated propositions or
truth relations between them in terms of only two modalities—true and
false. By Kant's time, it was already clear that two more modalities
must be contemplated in order to account for the more complex facts of
human language and human cognition. The mode of 'truth' was thus split
into analytic truth and synthetic truth, and this is very clearly an ante-
cedent—once one recognizes 'definition' or 'rules' as products of a com-
municative pact between speaker and hearer—to the later distinction
between presupposition and assertion, respectively. Somewhere between
Kant and Peirce 'synthetic truth1 was split into factual truth—our realis-
assertion, and possible truth—our irrealis-assertion. And again, although
the logicians adopted this as a purely semantic-logical distinction, it is
easy to see how in human language what is really involved is the com-
municative contract in terms of evidentiality, challenge, and their
interaction with subjective certainty (Givon 1982a).
In a similar vein, purely logico-deductive approaches first recognized
only the problem of 'existence', mapping onto one Real World. Variables
bound by the existential quantifier referrred to some entities existing in
that Real World, while those bound by the universal quantifier did not
likewise refer, but rather 'pertained to classes/types'. The introduction
of the Universe of Discourse to replace the Real World as locus where
terms found their reference was, in fact, as I have argued earlier, a
move toward making reference a matter of the speaker's referential
intent although still within the scope of atomic propositions—and thus
retaining residual 'semantic' coloration. What I have suggested here is
that if one is to account for the referential properties of natural lan-
guage, one has no choice but to take the next step and recognize that
the speaker's referential intent is not restricted to whether he 'means an
actual entity in the discourse universe', but rather, whether that entity
is important enough thematically in the communication so that its
unique referential identity actually mattered. What I have broadly
traced here is the historical progression given in (13).
(13) OBJECTIVE REFERENCE IN 'THE' REAL WORLD—*
SPEAKER'S INTENT OF REFERENCE IN SOME
DISCOURSE »
SPEAKER'S INTENT OF IMPORTANT
REFERENCE IN THE DISCOURSE
In human language, at least so far as the evidence of structural coding
seems to suggest, the relevant context for determining reference is not
the Real World, nor even the Universe of Discourse per se, but rather
the thematic organization of that universe of discourse. Within the
discourse universe, then, entities are considered 'referential' only if the
nature of their participation in the thematic organization makes their
unique referential identity important enough. Otherwise, they are
consigned to the same nondistinct grab-bag of masses, plurals, objects of
habit, groups or types, all of which may or may not exist at some meta
level, but whose individual identity does not matter in this particular
thematic context.
5.2 Correlations between physical, semantic and pragmatic referen-
tiality. With all that has been said here, one cannot fail to recognize the
rather privileged position that semantically referential entities occupy
within the pragmatics/thematics of reference. As noted, the vast ma-
jority of entities that were marked by wan in our Krio narratives were
indeed semantically referential. And conversely, the vast majority of
entities that were marked by bare-stem (zero) morphology were seman-
tically nonreferential. One could also go on and note that—if the dis-
course is about the normal themes of human affairs, everyday life and
the nuts-and-bolts of human existence and struggle for survival—a very
large frequency of overlap exists between semantic referentiality in the
universe of discourse and physical referentiality in the Real World.
Neither of these facts is accidental. And neither takes away even one
iota from the contention made earlier, namely, that reference in human
language is essentially a pragmatic-thematic matter. Human discourse
does not have to revolve around human affairs in the Real World. But
the fact that such discourse is both ontogenetically and phylogenetically
privileged is well motivated by the evolution and use of our communica-
tion system as a tool for survival in this particular world. However, the
pragmatics of perspective, saliency, and the assignment of thematic
importance in a particular task-context governs our use of language. In
other words, communication is never task-neutral or context-free. And
the systematic 'exceptions' in the Krio linguistic code, miniscuie as they
are in terms of frequency, nontheless illustrate how, when semantics
and pragmatics are in conflict, the coding system goes with pragmatics.
136 / Talmy Givon
5.3 Possible psycholinguistic and neurological correlates of pragmatic

referentiality. Almost all human languages make a coding distinction
between a referential noun introduced into the discourse for the first
time (REF-INDEF) and one already introduced previously, by whatever
means (DEF). In addition, as we have seen, many languages—perhaps
most—also code nouns in a special way if they are likely to be themati-
cally important in the discourse. While thematic importance is not a
totally objective and measurable property, it is nonetheless real.
Further, quite often—as noted here—it correlates with straightforward
measures of frequency of occurrence in the text. Perhaps the most
universal device crosslinguistically for coding pragmatically important
entities is that of 'naming1. Tagging a participant by name is a way of
filing it in the permanent memory—but in a particular fashion, thus
making it instantaneously available in a way that is independent of lower
level thematic contexts. In other words, it tags the participant/entity as
belonging to a higher thematic level/unit.
In a recent experimental study, Anderson et al. (1983) show that the
processing of referents that belong to a higher—more general—thematic
level in the narrative is easier, faster, more efficient. Such referents
are thus clearly 'more available1, as compared with referents pertaining
to lower, smaller, and more limited thematic contexts. Such results
point to an exciting field of empirical study of discourse, whereby the
grammatical coding of referents in discourse as 'more important 1 will be
shown to have processing consequences, in terms of making such refer-
ents more readily available to the language processor.
Notes
1. For a complete list, see Givon (1973, 1983a:Ch.8)
2. The average 20 (clauses), for referential distance measurements,
represents terminating the count at 20 clauses, i.e. either RD larger
than 20 clauses or—as is most likely here—no anaphoric reference at
all. The choice of 20 is motivated by considerations that are not im-
portant here (see GivoYi 1983b), but some limit has to be imposed arbi-
trarily to avoid dealing with infinity.
3. See Linde (197*0 for the definiteness of parts-of-whole in loca-
tional expressions.
^. I have used the same transcription as in Hancock (1972), but added
hyphenations of grammatical morphology as well as lexical stress
marks. The traditional Krio writing system renders all grammatical
morphemes as separate words.
5. When topics/participants remain in the discourse continuously, they
are marked as pronouns or occasionally, and in these Krio texts only
following the conjunction en_ 'and', as zero anaphora. The -ya-so marker
never appears with pronouns, but only with full NPs. It thus involves the
reintroduction of important participants. For the systematic, cross-
linguistic description of the role of full definite NPs vs. pronouns or
zeros, see Giv<5n (1983b).
6. One may, of course, propose that 'a mangy dog' was thematically
important in terms of the narrator's choice in describing the utter deso-
lation of the scene, which is the crucial backdrop for visiting the old
house, the major topic in the narrative.
7. One may argue that in both exception cases here, the nonreferen-
tial NP was attributive of the central participant in the discourse, the
old house.
8. This could also be an anaphorically based reference to 'firm', in
which case it should be counted there. But see also Linde (1974).
9. Given the context of 'firm' and normative assumptions—culturally
based—about men working in the firm, this may be also counted as
relational-definite and thus containing some anaphoric reference to a
major topic.
10. The justification for considering discourse about human actions
and doings as in some sense the ontologically prime mode of language
requires no further comment. The generalizations made by Hopper and
Thompson (1980, 1983) with respect to how concreteness, physical sa-
liency, individuation, compactness, and agentiveness of nouns correlate
with their so-called 'discourse manipulability' and their prototype prop-
erties qua nouns, all pertain to taking this fundamental assumption for
granted. Similarly, the statistical correlations recorded by both Coore-
man (1982) and Rude (1983) concerning the low discourse topicality of
objects that are semantically nonreferential were all recorded in this
'basic' discourse mode.
References
Anderson, A., S. C. Garrod, and A. 3. Sanford. 1983. The accessibility
of pronominal antecedents as a function of episode shifts in narrative
text. Quarterly Journal of Experimental Psychology 35A.427-40.
Bickerton, D. 1975. Creolization, linguistic universals, natural seman-
tax and the brain. MS. Honolulu: University of Hawaii.
Bickerton, D. 1981. Roots of language. Ann Arbor, Mich.: Karoma.
Carnap, R. 1947. Meaning and necessity. Chicago: University of
Chicago Press.
Carnap, R. 1958. Introduction to symbolic logic and its applications.
New York: Dover.
Cooreman, A. 1982. Transitivity, ergativity and topicality in narrative
discourse: Evidence from Chamorro. In: Giv<5h (1983b: 427-89).
Giv6n, T. 1973. Opacity and reference in language: An inquiry into the
role of modalities. In: Syntax and semantics, vol. 2. Edited by J.
Kimball. New York: Academic Press. 95-122.
GivoVi, T. 1975/1979. Negation in language: Function, pragmatics,
ontology. WPLU no. 18, Stanford University. Reprinted in Givdn
(1979).
Givon, T. 1979. On understanding grammar. New York: Academic
Press.
Givon, T. 1976/1981. The development of the numeral 'one' as an indef-
inite marker. Folia Linguistica Historia 2.1: 35-53.
Givon, T.I982a. Evidentiality and epistemic space. Studies in Language
6.1.: 23-49.
Givon, T. 1982b. Logic vs. pragmatics, with human language as the
referee: Toward an empirically viable epistemology. 3ournal of
Pragmatics 6.1.: Kl-133.
Givo'n, T.I983a. Syntax: A function-typological introduction, vol. 1.
Amsterdam: J. Benjamins.
138 / Talmy Givon
Givon, T., ed. 1983b. Topic continuity in discourse: A quantitative

cross-language study. Typological studies in language, vol. 3. Am-
sterdam: J. Benjamins.
Hancock, I., ed. 1972. The Journal of the Krio Literary Society 1.1.
Heath, J. 1976. Antipassivization: A functional typology. Berkeley
Linguistics Society, vol. 2. Berkeley, Calif.: Berkeley Linguistics
Society.
Hopper, P., and S. Thompson. 1980. Transitivity in grammar and dis-
course. Lg. 56.4: 251-99.
Hopper, P., and S. Thompson, (in press) The communicative basis for
lexical categories. Lg.
KalmaV, I. 1979. Case and context in Inuktitut (Eskimo). Ottawa:
National Museum of Man (publication no. 49).
Kalmar, I. 1980. The antipassive and grammatical relations in Eskimo.
In: Ergativity: Toward a theory of grammatical relations. Edited by
F. Plank. New York: Academic Press.
Linde, C. 1974. The linguistic encoding of spatial information. Unpub-
lished Ph.D. dissertation, Columbia University.
Rude, N. 1983. Topicality, transitivity and the direct object in Nez
Perce. MS. University of Oregon, Eugene.
Russel, B. 1905. On denoting. Mind 14.479-93.
WHITHER RADICAL PRAGMATICS?
Jerrold M. Sadock
University of Chicago
1. Pragmatic theory in general. A pragmatic account of a linguistic
datum is one which refers the fact not to independent structural or
semantic principles, but to principles concerning the use of the lan-
guage. Whereas the units of structural syntax and semantics are ab-
stractions like 'sentence' and 'proposition', 'verb phrase' and 'predicate',
those of a pragmatic theory are factual entities like 'utterance' and
'conversational contribution'. The principles that are invoked in syntac-
tic and semantic accounts have no independent existence outside the
language system, and include, for example, statements like 'PRO must
be ungoverned' and 'variables must be properly bound1. But pragmatic
principles, such as the principle that one should not do more than is
necessary to achieve a certain goal, or that one should consider the
feelings of others, would seem to exist regardless of the existence of a
linguistic system.
Thus there is a sense in which a pragmatic account of a fact of lin-
guistic behavior is a deeper explanation than is an account of that fact
in terms of syntax or semantics. A pragmatic treatment reduces a
phenomenon in one sphere to general principles of another sphere; syn-
tactic and semantic treatments do not. When we explain the chemical
behavior of an element in terms of quantum-mechanical physics, we
have, in a real sense, explained it more deeply than when we point to the
position of the element in the periodic table.
Putting it somewhat differently, pragmatic accounts, because of their
nonlinguistic nature, are capable of genuinely simplifying the account of
natural language behavior by eliminating—not just renaming—structural
features that would otherwise be required to model the facts.
To take a simple case, consider the well-known pragmatic treatment
of exclusive or_ in English (Horn 1976, Gazdar 1979). It is a fact that O£
can be used to indicate something like 'exactly one of...1 or something
like 'at least one of...1 Now since there is no doubt that use can reflect
meaning (we use alligator to refer to alligators only because it means
'alligator'), and since ambiguity, both lexical and structural, is an
139
1*0 / Jerrold M. Sadock
undeniable fact of natural language, one observationally adequate theory

of or_ is that it is ambiguous, an accidental phonological merger of two
distinct words with distinct meanings, and hence uses.
But a highly satisfying account of the bifurcated use of or_ can also be
given, relying on only one meaning (identical to the weaker, inclusive
sense) and pragmatic principles with an independent motivation. In
particular, an account can appeal to the very general notion that where
the language provides expressions of the same type and degree of com-
plexity which differ in that one is more contentful than the other, then
not using the more contentful expression in a context that provides no
apparent reason for not using it can indicate that the user believes the
stronger expression to be false. So, since and and or_ are expressions of
the same type, and and is stronger and no more complex that or, saying
P or Q where I might well have said P and Q can come to indicate rough-
ly the same thing as P or Q and not (P and Q), i.e. exclusive negation.
A serious question that needs to be asked about this or any similar
account is whether we really have simplified anything by invoking it, or
whether, in fact, the same constructs have just taken up residence in
another locality under assumed names. In this particular case, that does
not seem to be so, because as it turns out, exactly the same pragmatic
mechanism is needed, even if we adopt the ambiguity theory of or_.
Suppose or_ is ambiguous. Still it is the case that when the word is
used, there rarely is any unclarity as to how it was intended. We must
therefore have some pragmatic means of calculating the intended mean-
ing of an ambiguous sentence used in a particular context. It turns out
that the pragmatic sketch of the range of understandings of ^r_ that I
have given is exactly what we would need to explain how hearers will
choose the right meaning of a presumably ambiguous or_ (and how
speakers will be able to calculate whether their utterance will be taken
in the intended fashion).
Suppose we are discussing 3ohn's moral character and you say, 'Well,
he either smokes or drinks.1 Supposing still that or_ is ambiguous, then
what you have said is either (1) or (2), and it is my job to figure out from
the context which one you intended.
(1) (John smokes) V inc (John drinks)
(2) (John smokes) V ex (John drinks)
How can we account for the fact that you are likely to be taken as
having intended (2) in this context? In general, we will take an utter-
ance as having as much relevant content (or just 'relevance 1 , cf. Sperber
and Wilson 1983) as is consistent with the context. Now we know that
(2) is stronger than (1), that the stronger statement would be relevant in
the context of this discussion of John's habits, and finally that there is
nothing in the form of utterance that would give an indication that the
weaker reading was intended (in particular, that the longer unambiguous
form and/or has not been chosen). So we conclude that (2) was intended.
But note that this account of the disambiguation of the sentence
contains all the details of a pragmatic account that would be sufficient,
in and of itself, to account for the way the sentence is understood with-
out the need for a postulated ambiguity. So the account that makes use
of ambiguity is seen clearly to be one that ought to be sent to Mr.
Occam's shop for a trim.
Whither radical pragmatics? / 1*1
2. Radical pragmatics. Because of the hope that pragmatic accounts

hold out for actually explaining (and thereby doing away with) structural
assumptions and principles, there has been an understandable tendency
during the last ten years or so to render as many facts in pragmatic
terms as possible, and, in the process, to simplify the structural system
perhaps to the point of invisibility. The big idea in radical pragmatics is
that many of what were conceived of as essentially inexplicable struc-
tural facts of language are really reflections of the fact that natural
language is used by real speakers in real contexts to accomplish real
goals. The hope is that by taking into account the appropriate aspects of
the use of expressions, the structure becomes otiose and can be dis-
pensed with in explaining the primary linguistic datum, namely, that a
certain expression can be used on a certain occasion with a certain
effect.
Pragmatic accounts of natural language data are, in other words, quite
seductive; so what I wish to do here is be a linguistic Ann Landers and
try to answer the question, 'How do we know when we've gone too far?'
Like the columnist, I believe that it is not only possible to go too far, but
that many young people nowadays are doing so.
Two fairly distinct traditions of pragmatic explanations are in evi-
dence today, one based squarely on Grice's (1975) notions of the role of
cooperativity in language use, and the other based more on discourse-
functional notions reminiscent of the thinking of the London School of
Firth and Halliday, and the Prague School of Firbas and Sgall. In gen-
eral, a Gricean pragmatist is not averse to the employment of discourse
notions in his/her accounts, nor is a functionalist opposed to the invo-
cation of Grice's maxims; but the brunt of the explanatory effort does
differ between these two schools. A characteristic difference between
the two modes of explanation is that the result of Gricean reduction is
ordinarily an impoverishment of the semantic system of the language,
while reduction in the functionalist mold is usually at the expense of the
syntactic system. But both camps can be indicted for failure to exercise
moderation in their pragmatic proselytizations. In an effort to be fair, I
will discuss one example of excessive pragmatism from each.
Whereas some extreme structuralists can be criticized for adopting an
unjustifiably negative view of function in the description of natural
language data, pragmatists of both stripes can be chided for the opposite
error: the radical structuralist says that there is nothing of value to be
gained by considering the functions that natural languages serve; the
radical pragmatist takes the same dim view of structure.
But Saussure (1966) was right: human languages have structure.
Therefore there are correct and incorrect descriptions of their struc-
tures. While it is true in a sense that a purely formal account of the
structure of a natural language is 'merely a recapitulation of the data',
or 'merely a statement about correlations to be found in the data', it
does not follow that no differences of any kind attach to differences in
formal description. It is a true fact about English t h a t Q G andQ)Dare
in complementary distribution, but it would be just plain wrong to de-
scribe English as presenting a significant generalization here by, say,
making ChTJ and {j)3 allophones of the same phoneme. It is a fact that
the regular third person nominal plural morpheme in English and the
regular third person singular verbal inflection are phonologically
identical; but it would be wrong to claim that this is a principled fact of
English by, say, postulating a morphological analogue of the liver fluke,
a motile morpheme that swims from one host to another.
A language system is not simply a collection of facts, but rather a
collection of generalizations obtaining among these facts. To describe a
language formally in such a way as to make one generalization is not to
do exactly the same thing as to describe it in such a way as to make
another, for in the two cases we are describing different systems. And
to describe structure without making any generalizations is the equiv-
alent of claiming that there is no structure at all.
3. Function and structure. The radical pragmatic stance is that once

we know the 'why1, we need no longer be concerned about the 'what1,. To
pursue a biological metaphor that has become rather popular lately, the
radical pragmatist would be like an anatomist who, realizing that birds
fly, loses all interest in the structure of their wings. Since fly they
must, he says, then wings they must have and those wings must be as
they are, else the birds could not fly. But this thinking is faulty. Bats
fly with very different wings, and helicopters and rocket ships fly with-
out benefit of wings. In fact, to predict the wingedness of birds, we
must bring autonomous structural considerations to bear. Independent
structural knowledge of the articulation of animal appendages rules out
the possibility of birds with rotors, and independent structural knowledge
of their digestive systems makes jet-propelled birds at least highly
unlikely. The appropriate point of view is to examine structure in the
light of functional requirement and function in the light of structural
requirement, and this holds of linguistics every bit as much as it does of
biology.
In fact, one of the strongest motivations for the independent study of
linguistic structure is exactly that it is needed to support any reasonable
functional theory. The structural features of natural language are what
support the raiments of natural language function. Ann Landers would
probably advise us to keep our raiments on.
t
t. Radical Griceanism. Let me now present a case in the tradition of
Grice where it is clear that a pragmatic explanation can be taken so far
as to undermine its own structural foundations.
Consider the elegant and convincing treatment of the understanding of
scalar expressions originally given by Horn (1976) and elaborated by
Gazdar (1979). The fact that a scalar expression, e.g. a cardinal number
expression, can be variably understood as indicating 'at least n' or 'exact-
ly n' was claimed by Horn not to be the result of a systematic ambiguity
in such expressions, but rather the result of Gricean rules of pragmatic
inference. In particular, Horn argued that such expressions are lower
bounded semantically, that is by entailment, but upper bounded only
pragmatically, that is by implicature. Thus (3) is false if John makes
$10,000, but only misleading in some contexts of use if he makes
$20,000.
(3) John makes $15,000.

Whither radical pragmatics? / 143
Horn also observed that the direction of the scale on which such
expressions lie can be pragmatically adjusted. To break 100 in golf
means to have a score of less than 100, while to break 100 in bowling
means to have a score of more than 100. An utterly pragmatic interpre-
tation of this situation (not, incidentally, Horn's ) would be that the
natural numbers have no meaning in and of themselves, but are merely
ordered with respect to each other by entailment. Thus three, for
example, would not mean 'three', but any expression containing it would
imply the truth of a similar expression with two substituted for three,
but not one with four substituted for it.
On this theory, the natural number words we have in English would
mean hardly anything at all. They would achieve their effective ability
to communicate information about cardinality in an entirely pragmatic
way. Such a result might offend our sense of propriety, but what is
really wrong with it is that it cannot work. There is simply not enough
conventional content left in the number words for any pragmatic theory
to use as input. On a theory that gives the natural number words a
semantic lower bound, the precise use of these words (e.g. Three is the
square root of nine.) can be modeled as a contextual setting of an in-
determinate upper bound equal to a semantic lower bound. But if there
is no precise upper or lower bound, it is not at all clear how we could
ever account for the precise use.
Surprisingly, there is actually an advantage to the outrageously prag-
matic theory of number words in that it treats them differently from the
inexact quantifiers like some, many, most, and aH. With these we find
exactly the same upward dual understanding as with the numerical
quantifiers (some being variably 'some only' or 'at least some'), but we do
not find any scale reversal. It does not seem possible to use some, for
example, in such a way as to implicate 'at most some'.
In order both to capture the correct range of contextual under-
standings of the various quantifiers and to account for the asymmetry
between the exact and inexact ones, it seems to me that a less radical
pragmatic theory than even Horn's might be required. His theory of the
inexact quantifiers can remain, but I suggest that the exact quantifiers
have exact meanings. The pragmatic principle involved in their inter-
pretation as one-way unbounded is one of loose-speaking, the same
principle that allows us to describe France as hexagonal (see Austin
1962). A speaker using three to indicate 'three or more' would then be
conveying less than his words imply, rather than more.
5. Radical functionalism. The second case I want to describe is an
example of overboard functionalism. It has to do with the treatment of
Eskimo sentence types and case marking. There are four sentence
patterns that we will be immediately concerned with which are tradi-
tionally called the intransitive, the transitive, the antipassive, and the
passive. These are illustrated in that order by (4)-(7).
(t*) Angut sulivoq.

man work-3sg
'The/A man works.1
1** / Jerrold M. Sadock
(5) Angutip nannut takuai.

man-erg bears see-3sg/3pl
'The man sees the bears.1
(6) Angut nannunik takuvoq.
man bears-inst see-3sg
'The/A man sees bears.'
(7) Nannut angummit takusaapput.
bears man-abl see-passive-3pl
'The bears were seen by a man.'
There is clearly a strong association in this language between the form

of a sentence and discourse-functional dimensions such as old infor-
mation-new information, and theme-rheme, as the definiteness of the
noun phrases in the glosses in these examples is meant to suggest. Ivan
Kalmar, in a study of the North Baffin Island dialect (Kalmar 1979),
suggests that these notions are all we need to understand the fundamen-
tal facts of Eskimo grammar, and provides the following (mostly prag-
matic) flow chart as a scheme for selecting sentence type in Eskimo. In
this chart, what I have called the transitive Kalmar calls the 'ergative',
and what I have called the antipassive, he calls the 'accusative', reflect-
ing his belief that the ergativity of Eskimo is an artifact of functional
concerns and that the antipassive is really another sort of transitive
clause. For the same reason, he caUs the case that I have labeled in-
strumental in (6) the accusative case.
(8)
INTRANSITIVE
CLAUSE
one
PROPERTY/RELATION
PASSIVE CLAUSE
arguments? theme
two
VOICE
ERGATIVE CLAUSE
given
rheme
ERGATIVE/
ACCUSATIVE
Ipatient?
ACCUSATIVE
CLAUSE
While Kalmar is surely onto something, his form-follows-function
doctrine leads him to the claim that the very obvious syntactic prin-
ciples of Eskimo do not, in fact, exist. However, the following are
among the structural facts that characterize not only the four sentence
patterns illustrated, but the whole language. (1) Every clause contains
exactly one absolutive case noun phrase and at most one ergative case
noun phrase. It may contain any number of oblique case arguments
which are always optional. (2) The Eskimo verb agrees with its abso-
lutive and ergative arguments only, thus producing two sentence pat-
terns, an intransitive with single agreement and a transitive with double
agreement. (3) The language is ergative in case marking. (4) There are
a large number of relation-changing affixes in the language, including
those that detransitivize transitive verbs and those that transitivize
intransitive verbs. Whatever else these do, they do not produce excep-
tions to (l)-(3).
Kalmar's functional treatment misses these facts entirely and in fact,
he specifically denies (2). Any or all of these strictly structural facts of
the language could be different and Kalmar's scheme would come
through unscathed. For example, associations between clause type and
discourse function could simply be scrambled in a diagram like (8), if the
facts warranted it.
But, of course, the facts could not warrant it. We sense that it is not
accidental that the oblique case passive agent and the oblique case
antipassive patient are nonthemes, in Kalmar's terminology. Nor does it
seem accidental that these arguments are optional, whereas the abso-
lutive arguments in these patterns and in the transitive are obligatory.
There is, in other words, a rigid structure to this language. What we
see, then, is that the road connecting form and function is not a one-way
street as the radical functionalists seem to imply. Not only does form
follow function, but function also follows form, a fact that cannot be
captured without recognizing the essential independence of the struc-
tural layer of language.
Having stated these structural verities explicitly, it becomes possible
to observe that one functional principle accounts for roughly what
Kalmar's purely functional scheme does, but does it both better and
more generally. This is the principle that arguments with which the verb
agrees are, ceteris paribus, more topical (thematic) than other argu-
ments which could stand in contrast with them.
First of all, this generalization extends to the other sentence patterns
of the language, some of which Kalmar worries about, but cannot deal
with. To take just two examples, the structure-dependent functional
principle directly handles the fact that the incorporated object in (9) is
nontopical, and that the promoted object in (10) is much more topical
than the nonpromoted argument in (11). But the direct matching be-
tween discourse properties and surface form that Kalmar proposes will
not extend to these since (9) and (11) are not antipassive (i.e. accusative)
clauses in any obvious sense.
(9) Kaali illoqarpoq.

Karl house-have-3sg
'Karl has a house.'
(10) Kaalip meeraq oqarfigaa.
Karl-erg child speak-to-3sg/3sg
'Karl spoke to the child.1
146 / Jerrold M. Sadock
(11) Kaali meeqqamut oqarpoq.

Karl child-allat speak-3sg
'Karl spoke to a child.1
Furthermore, Kalmar's scheme is empirically inadequate in several
ways. For one thing, the discourse-functional differences among sen-
tence types disappear where the grammatical system of the language
forces a choice of one form. For example, factive nominalizations in
-neq are formed from the intransitive stems only, and thus for transitive
verbs must be formed from the antipassive clause rather than from the
transitive. Thus, whereas (12) is odd compared to (13) because unique
entities like nature, the sun, and so forth are automatically topical in
Eskimo (just as they are automatically definite in English), the nomi-
nalization of (12), namely (14), is not odd because it is the only possible
form.
(12) ??Qallunaat pinngortitamik mingutittipput.

Danes(abs) nature-inst make-dirty-antip-3pl
'The Danes despoil nature. 1
(13) Qallunaat pinngortitaq mingutippaat.
Danes(erg) nature(abs) make-dirty-3pl/3sg
'The Danes despoil nature.'
(14) Qallunaat pinngortitamik mingutittinerat
Danes(erg) nature-inst make-dirty-antip-nom-3pl
'the Danes' despoliation of nature'
Kalmar also observes that there are clear exceptions to his principles
in his own very brief corpus. Examinations of longer texts would surely
turn them up in great numbers. Consider the following two snatches of
connected text from a Greenlandic children's book (Schwaerter 1961).
(15) Ullut ilaanni Suulut karsikoq nassaaraa.

days one Suulut-erg old-box find-3sg/3sg
'One day Suulut finds an old box.'
Aamma qisualunnik nassaarpoq.
also wood-piece-inst find-3sg
'He also finds some pieces of wood.1
(16) Ullut ilaanni Suulut allunaasamik nassaarpoq.
days one Suulut string-inst find-3sg
'One day Suulut finds a piece of string.'
Aamma nassaaraa kikiak.
also finds-3sg/3sg nail
'He also finds a nail.'
Both of these pairs of sentences consist of one transitive and one anti-
passive. In (15) the transitive is first and in (16) the antipassive pre-
cedes. But there is clearly no difference in terms of givenness between
the patients in each pair. What this shows is that the pragmatics of case
and transitivity, being pragmatics and not grammar, is much more subtle
than a functionally driven account of natural language could accommo-
date. While there are clearly pragmatic ramifications to the choice of
sentence type in this language, the choice is not absolutely mandated by
the context. Sometimes the choice is motivated merely by a desire to
avoid repetitious sentence patterns and this is precisely what I suspect is
going on in the little Greenlandic texts cited here.
6. Conclusion. What I have tried to show here is, first, that the
grammar of a language, while clearly suited to the purposes for which its
speakers want it, is essentially independent from function; and second,
that the recognition of this fact is essential to a correct functional
description of the language. To the extent that the goal of radical
pragmatics is the elimination of independent structural descriptions, it is
misguided and will eventually prove unworkable. To the extent that its
goal is a detailed and correct account of the intricate symbiosis that
characterizes the association between structure and function in natural
language, it is an important and exciting linguistic enterprise that de-
serves to flourish.
Notes
1. Other examples that might have been discussed include Atlas
(1977), Kempson and Cormack (1981), Garcia (1979), and my own Sadock
(1976).
2. For a recent defense of this time-honored fact, see Newmeyer
(1983).
3. The anatomical metaphor is employed by both radical structuralists
and radical functionalists. Compare Chomsky (1977:86):
How does it Dhe heart] grow in the individual from the embryo
to its final form? The answer is not functional: the heart does not
develop because it would be useful to carry out a certain function,
but rather because the genetic program determines that it will
develop as it does.
Let's go back to linguistics: here comparable remarks can be
made...suppose that someone proposes a principle which says: The
form of language is such-and-such because having that form per-
mits a function to be fulfilled—a proposal of this sort would be
appropriate at the level of evolution (of the species, or of lan-
guage), not at the level of acquisition by an individual...
with Givon (1979:5):
Imagine an anatomist describing the structure of the human body
without reference to the functions of various organs. But this is
precisely what happened in transformational-generative linguis-
tics...an attempt has been made to describe the structure of human
language...without reference to natural explanatory parameters.
The value of my own use of this metaphor should be considered in the

light of its apparent flexibility.
4. Horn theorizes that the quantifiers remain lower bounded, but the
high end of the scale becomes the low end.
5. This is a perfect example of the pragmatic paradox: the less some-
thing means in fact, the more it signifies in practice. The more precise
quantifiers have greater contextual latitude (one-way unboundedness and

scale reversal) than the less precise ones (one-way unboundedness only).
6. John Richardson (1984) presents some cogent arguments along
somewhat different lines. In particular, he points out Kalmar's (and by
extension, other radical functionalists1) grave misunderstanding of the
role of optionality in a structural theory.
7. 'One' and 'two' in this diagram refer to the number of semantic
arguments in the clause. Kalmar uses 'theme' to refer to "what the sen-
tence is about', and 'rheme' to indicate the comment made about the
theme.
8. There are also clear pragmatic correlates of the choice of passives
and actives in English, as shown in various studies such as Givon (1983).
But these studies also show that the correlation is not perfect and that it
is, in other words, a use of the structure of the language rather than a
substitute for it.
References
Atlas, Jay D. 1977. Negation, ambiguity, and presupposition. Linguis-
tics and Philosophy 1.323-36.
Austin, J. L. 1962. How to do things with words. London: Oxford Uni-
versity Press.
Chomsky, Noam. 1977. Language and responsibility. New York: Pan-
theon Books.
Cole, Peter, and Jerry Morgan, eds. 1975. Syntax and semantics, vol.
3: Speech acts. New York: Academic Press.
Garcia, Erica C. 1979. Discourse without syntax. In: Syntax and se-
mantics, vol. 12: Discourse and syntax. Edited by T. Givon. New
Gazdar, Gerald. 1979. Pragmatics: Implicature, presupposition, and
logical form. New York: Academic Press.
Givon, Talmy. 1979. Understanding grammar. New York: Academic
Press.
Givon, Talmy, ed. 1983. Topic continuity in discourse: A quantitative
cross-linguistic study. (Typological studies in language, vol. 3.)
Amsterdam: John Benjamins.
Grice, H. P. 1975. Logic and conversation. In: Cole and Morgan, eds.
(1975). 43-58.
Horn, Laurence R. 1976. On the semantic properties of logical opera-
tors in English. Ph.D. dissertation, Yale University. Reproduced by
Indiana University Linguistics Club, Bloomington.
Kalmar, Ivan. 1979. Case and context in Inuktitut (Eskimo). (National
Museum of Man Mercury Series paper no. 9) Ottawa: National Mu-
seums of Canada.
Kempson, Ruth, and Annabel Cor mack. 1981. Ambiguity and quanti-
fication. Linguistics and Philosophy 4.259-310.
Newmeyer, Frederick J. 1983. Grammatical theory: Its limits and
possibilities. Chicago: University of Chicago Press.
Richardson, John. 1984. An Arctic afternoon: Pragmatic correlates of
grammatical processes in Eskimo. University of Chicago course
paper.
Whither radical pragmatics? /
Sadock, 3errold M. 1976. Larry scores a point. Pragmatics microfiche,

fiche I A.
Saussure, Ferdinand de. 1966. Course in general linguistics. Edited by
Charles Bally and Albert Sechehaye in collaboration with Albert
Reidlinger, trans, with an introduction by Wade Baskin. New York:
McGraw-Hill.
Schwaerter, Adolf. 1961. Suulut. Godthaab: Det Gr^nlandske For lag.
Sperber, Dan, and Deirdre Wilson. 1983. Relevance: Foundations of
pragmatic theory. MS, working draft.
A PSYCHOLOGICALLY AND SOCIALLY PLAUSIBLE THEORY
OF LANGUAGE STRUCTURE
Richard Hudson
University College, London
1. An introduction to word grammar
1.1 Background. In this paper I want to do two things. First, I shall

tell you something about a linguistic theory called word grammar, which
I have been working on for the last few years—since finishing the text-
book on sociolinguistics (Hudson 1980) which some of you may know.
Then I shall take a particular problem that faces those with an interest
in both sociolinguistics and the theory of language structure, and explain
how word grammar helps to solve it. The fact that word grammar is
useful for the sociolinguistically inclined is not unconnected with the
fact that I started developing the theory just after I had finished working
on the sociolinguistics textbook. I was tired of having to criticise
theories of language structure for being asocial, and decided to try and
do something about it.
However, I do not want to give the impression that word grammar is
especially designed for sociolinguists, because it is meant as a stab at
the true nature of language, and I for one do not believe that truth for
sociolinguists is different from truth for psycholinguists or for linguists
interested solely in language structure. Admittedly, they are all in-
terested in a different part of the truth, but their findings must be
compatible with one another, so that they could in principle be fitted
together into one giant jigsaw puzzle, whose bits all mesh. So word
grammar is meant to be a theory of language structure which will ac-
commodate everything I know as a linguist, plus the things I've read by
sociolinguists, psycholinguists, and cognitive psychologists. This is what
I mean when I describe it as 'psychologically plausible' and 'socially
plausible'—at least, it is not blatantly at odds with what psychologists,
psycholinguists, and sociolinguists believe.
I shan't be able to tell you much in this paper about the theory, but
you can make up your own mind about its plausibility by looking at a
book-length account which is due out soon (Hudson 1984). Meanwhile, I
150
A plausible theory of language structure / 151
shall try to persuade you that the effort of reading the book will be
worthwhile. To be honest, I must make it clear that the book is mainly
about language structure, so it has a chapter on morphology and lexical
relations, a longer one on syntax, and an even longer one on semantics.
The first chapter says something about psychology, and the last one is a
brief discussion of the structure of utterances, but it doesn't develop
these areas in any detail. On the other hand, I have a fair idea about
how they might be developed, given enough personpower, and I'm not
aware of any fundamental problems waiting for those who try to develop
them. So the book and the theory are primarily aimed at linguists, and
the most well-developed suggestions are in the area of language
structure. This is just because I am a linguist, and language structure is
what I am most interested in—and what I have thought and written about
most.
I hope you will agree that at least the more mainstream theories of
language structure are not very plausible either psychologically or
socially—I have in mind theories as diverse as government-and-binding,
relational grammar, generalised phrase-structure grammar and
Montague grammar. Of course, their proponents would say that they do
not care, because they were only concerned with linguistic structure,
and from that point on I should have to concentrate on showing them
that even with their restricted aims, word grammar could do better—
which it could. But it seems to me very shortsighted to restrict one's
aims in this way, knowing that sooner or later the crunch will come and
all the details of your theory, which you have worked out with such
loving care, may need to be changed drastically. For example, it is all
very well to say that you are not interested in words like hullo, but you
should at least have some idea of how they fit into your theory of lan-
guage structure, even if you leave the details to be worked out by
others.
What I have just said about the well-known mainstream theories is
equally true of the little ripple that I caused in the mainstream by a
theory called daughter-dependency grammar, which I described in a book
some years back (Hudson 1976). This was a nontransformational theory
which made much use of arbitrary syntactic features. It is hard to think
of parallels for syntactic features of this kind outside language, so they
are suspicious from the point of view of psychological plausibility; but
there are good internal linguistic reasons for avoiding them too, namely,
that if a feature only relates to a single fact, then it is always better
and easier to state that fact directly, rather than use the clumsy device
of a feature to mediate it. And daughter-dependency grammar was not
very promising from the social point of view, either, because it made no
provision for referring to the nonlinguistic context, as one would have to
do in dealing with words like hullo.
Word grammar is a radical alternative, by any standards. You may not
like radical alternatives, so this is not meant to impress you. It is just a
fact. Word grammar has no transformations, no phrase structure, no dis-
tinction between the rules and the lexicon, or between semantic and
pragmatic structures, or between sentences and utterances. It is called
word grammar because most of the grammar is about words, taken
either as particular lexical items, or as general types (such as 'noun' or
'plural'); but there are no references in the grammar to phrases, clauses,
152 / Richard Hudson
or sentences. In spite of this, we can write word-grammar accounts of

relative clauses, unbounded dependencies, extraposition, and so on—all
the tricky areas of grammar against which linguists like to test their
theories. At the moment you just have to take my word for it, but if you
read the book, you will find a full account.
1.2 Instantiation. One of the main features of word grammar is that

generalisations are made in terms of a hierarchy called instantiation
hierarchy. For example, dog is an instance of a 'noun', and 'noun' is an
instance of a 'word'—nothing very radical there, of course. The inter-
pretation is not very controversial, either: certain generalisations about
distribution and inflections are made in relation to the category 'noun',
and we say that dog is a noun because these generalisations are also true
of dog. So we can say that if X is an instance of Y, X has the properties
of Y among its properties (along with other properties, such as being
pronounced /dog/ and being used to refer to things that bark). I call
'noun' the 'model' and dog the 'instance', and I find it is helpful to
connect the instance and its model in a diagram by means of a double-
shafted arrow, as in Figure 1. You can think of the double shaft as a
kind of tube for conveying information from the model to its instance,
because the instance takes some of its properties from its model, but not
vice versa. In the terminology of some cognitive scientists, the instance
inherits properties from the model.
Figure 1.
word \ noun \ dog
The theory becomes more controversial when we add the word

'normally' to the principle that an instance has all the properties of its
model. This proviso is needed because the properties of the model can
be overridden by properties of the instance. For example, the inflec-
tional peculiarities of an irregular noun will override the normal rules
for making nouns plural, as in the case of goose. Provided we specify in
the grammar that the plural of goose is geese, the normal way of
forming plurals will not apply. Linguists have known this for thousands
of years, and have even given the principle a name—the Proper Inclusion
Precedence Principle (Pullum 1979:82). The effect is that we can say
that goose is a noun, although it does not share all the properties which a
typical noun has—it is not as typical a noun as dog. You can see that the
category 'noun' acts like the 'prototypes' studied by Rosch and her col-
leagues, so that we could rank words for 'nouniness', in true Rossian
fashion (Ross 1973). Some nouns are more typical than others regarding
their inflections; others are more typical regarding distribution—e.g.
today is a noun, but it can occur in places where nouns typically cannot
occur. And so on.
Even more controversial is the claim that the relation of a particular
occurrence of the word dog to the stored word dog is the same as that of
the latter to the category 'noun1. Suppose someone says My dog is bigger
than your dog. This utterance contains two instances of the word dog, so
it seems fairly reasonable to say that the relation between each of these
instances and the stored word dog is that of instantiation;1but that is the
relation we have1 assumed between stored dog and 'noun , and between
'noun' and 'word . So unless you can find some fundamental difference
between these two cases, we have to accept that the relation is the
same. That may seem easy to agree to, but if you do agree, then you
have lost the boundary between sentence and utterance, because the
logical relation between an utterance of an expression and the stored
representation of that expression is a relation which is also found
between stored elements. Let us call the two occurrences of dog in My
dog is bigger than your dog 2 and 7, respectively (to show their position
in the utterance). Figure 2 shows how they fit into the pattern started
in Figure 1. Of course, as we should expect, their properties include
those of the stored word dog, plus others (e.g. the properties of being
uttered as second and seventh words of this utterance).
Figure 2.
word =^noun
If we take the step that I have just described, then the stored
representations for linguistic expressions can be seen as normalised
models for uttered expressions. This is an important point because
utterances have just the properties that sociolinguists are interested in—
they occur at particular times and places, they are said by particular
people to particular other people, and their speakers have purposes and
see their utterances as fulfilling particular social functions. In the
mainstream tradition of linguistics, all these properties have been simply
left out of the normalised stored representations, whereas other proper-
ties such as pronunciations have been left in. No general principles are
ever offered to explain why this particular distinction is made. It is not
because pronunciations are always as specified in the normalised repre-
sentation; they often are not (e.g. the normal ji at the end of in_ is often
represented by a quite different pronunciation as a result of assimila-
tion). Nor is it because words are never restricted in relation to these
situational properties; for example, good morning is restricted to use
before lunchtime. I conclude, then, that unless good reasons can be
found for excluding such information from the permanently stored repre-
sentation of a word, it may be stored.
13 Companionship. The relation of instantiation is often called the

'isa' relation by cognitive scientists, because dog is a noun, and so on.
We can match this relation with a 'hasa' relation, which I call compan-
ionship—the relation between one entity and another which may (or
must) occur with it. This is the relation in language between a word and
the words with which the grammar allows it to occur. For example, the
subject is a companion of the verb, and so are the various objects and
other complements, because all these relations are directly sanctioned
by the grammar. So, in cows eat grass slowly, cows, grass, and slowly
are all companions of eat, but grass and slowly are not companions. We
can say that the verb has a subject and an object, and that is why I call
this the 'hasa' relation.
The companion relation is asymmetrical, so we can link companions by
an arrow pointing from the 'haver' to the 'had1—from the verb to its
subject and object, for example. This relation is well known in
traditional grammar, and is called dependency; and it is because we can
exploit dependency that we have no need for phrase structure. When
used in syntax, arrows can be left without further labels (except in the
cases of subjects and other preverbal elements), since we only need to
show the direction of dependency. Figure 3 is an example of a de-
pendency structure, combined with the instantiation structure which
shows which lexical items the words are instances of.
Figure 3.
subject
arefully
In semantic structure we need companions too, to link arguments to

their predicates, and generally speaking, the semantic companion
relations match the syntactic ones; there are enough exceptions, though,
to make it clear that the two levels are distinct. For example, the
semantic element corresponding to tread has the referent of its subject
as a companion, and the referents of people and feet in the last example
are linked as companions. In semantics, I think it is clear that we need
to distinguish different kinds of companions, but it does not matter
whether we call the distinct categories semantic roles, semantic func-
tions, deep cases, or theta roles; the idea is the same behind all these
different names. For example, we should label the subject-based seman-
tic companion of tread 'actor', but this would be the wrong label for the
referent of feet in relation to people. There is a partial semantic
structure in Figure 4, where a star attached to a number is the name for
the semantic element corresponding to the word with that number; so,
for example, 1* stands for the referent of the first word, which is
people.
Figure *.
actor
One of the main uses of the companion relation is that it allows us to
impose restrictions on potential companions, in order to show which
combinations of words are possible and which are not. This applies both
to syntactic and to semantic components. For example, we can require
subjects to precede the verb, or to be in the nominative case, or
whatever the facts demand; and we can require actors to be animate,
and to provide the energy used in the action, or again whatever the
correct analysis is. These restrictions can be imposed in relation to very
general categories, like 'verb' or 'action', and in effect they provide a
definition of the companion types concerned. This is important because
it means that there need be no undefined labels in a word-grammar
analysis, in contrast with many other theories where semantic roles, in
particular, are left extremely short of definition.
3ust to make the point clear, take the category 'subject'. Suppose we
want to require subjects to be nouns and to precede their verb; we can
show this diagramatically, as in Figure 5. By the way, I should explain
that there are two notational systems to choose between in word
grammar; if you do not like diagrams, you can use algebraic formulae. I
think diagrams are easier for the reader, but formulae are certainly
better for the writer. Figure 5 shows that a verb takes a companion
called 'subject1, which must be an instance of 'noun' and which must
precede the verb (this is shown by the 'is less than' sign, which refers to
the numbers that are used to name words, and that get larger with
time). Since this is a property of the general category 'verb', it will
automatically be inherited by any instance of 'verb', so there is no need
to state separately for a particular verb, such as tread, that it takes a
subject with these properties.
Figure 5.
.word>
Since I shall be using the companion relation as applied to semantic

structures, I shall give a semantic example as well. Suppose we take
'action' as a particular instance of 'event', in which there is an actor. We
could say that an event must, by definition, have a time and place, so by
inheritance an action must have these companions too. Further down
the instantiation hierarchy, we could say that a deed is an action, in
which the actor is a person. All this information is contained in Figure
6. So if we say that the meaning of tread is a kind of deed, then we
automatically predict that it will have a doer who is also an actor, and
who is a person and an energy source, and that it will have a time and
place. We could even require the actor to be the referent of the verb's
subject, as it normally is.
At the moment, I do not want to dwell on the linguistic advantages of
having word meanings structured in this way. Instead, I should like to
point out that Figure 6 could quite reasonably be taken not just as an
account of a part of semantic structure, but as a hypothesis about part
of general cognitive structure~i.e. of the mental structures which we

use when we process our experience, nonlinguistic as well as linguistic.
Suppose I hear a creaking floorboard. I interpret this bit of experience
as an instance of treading, which automatically means it is a deed, so
there must be a human actor, and there must be a place (namely, on the
floorboard). We need not worry about the order of operations in this bit
of processing; the point is simply that I could not take it as an instance
of treading without including the other elements in the analysis.
Figure 6.
place=^x < Place
nergy-source
person
2. Help for the sociolinguist
2.1 The problem. I have now given you enough information about
word grammar to be able to move to the main point of this paper: how
word grammar can help bridge the gap between theories of language
structure and the findings of sociolinguistics. The problem is that we
know all sorts of things about the ways in which words interact with the
situations in which they are used, but none of the mainstream theories
provides a slot into which this information can be put, in contrast with
the rich provision that they make for information about syntax,
morphology, phonology, and semantics. It may not be a coincidence that
the same is true, more or less, of conventional dictionaries, which
provide just ad hoc labels like 'slang' or 'greeting'.
Let us suppose that someone wanted to write a grammar for my com-
petence, and that he came to the word cookie. Now, as you probably
know, we in Britain use biscuit instead of cookie, but we know that
Americans use cookie, so this is part of our competence, in the sense of
'knowledge of linguistic expressions'. It certainly affects our linguistic
behaviour, because we do not use the word, although we know it; and, of
course, the reason why we do not use it is because we know that it is
used only by Americans, and we are not Americans. But how do you fit
this information about cookie into, say, a transformational lexical entry
for cookie? The best you could hope for would be some kind of uninter-
preted diacritic feature like & American], but without an interpretation,
that might as well not be there. But this problem is surely ridiculous
because the fact to be stated is so straightforward: all we want to say is
that the word is used only by Americans.
As you are all well aware, this problem is the thin end of a very large
wedge indeed, and we could make similar points in relation to register
differences in vocabulary (e.g. tr^_ versus attempt), regional and social
variation in pronunciation, politeness forms, greetings, and many other

phenomena. What they all have in common is that they require some
kind of reference to situational elements such as speaker, addressee,
time, place, and genre of speech activity. Even if you are not into
sociolinguistics, you still have an interest in solving this problem because
the same is true of words with deictic meaning, and speech act cate-
gories.
What we need, then, is some kind of principled account of the struc-
ture of speech situations, or rather, of our mental representations of
speech situations. Once we have this, we can combine it with our
account of mental representations of word structures, and we shall be
able to say just what we want. Of course, I am not the first person to
say that we need an account of speech situations, and there are a
number1 of analyses already on the market to choose from—for example,
Hymes list of 16 components of speech (Hymes 1972). But these are not
principled, in the sense that they could be fitted together with a lin-
guistic analysis as part of a unifying theory. So at best we need to adapt
them before using them in a generative grammar.
2.2 The word-grammar solution. I think word grammar provides the
basis for a solution. First, ask yourself what kind of a thing a word is.
Admittedly, we are used to thinking of words as static objects or
patterns, because of writing; but when you think of spoken language and
ask what a word is, the answer is clear: it is an event. But we have just
been discussing events, in our classification of knowledge, and related
the general category to more specific categories, 'action' and 'deed1. In
terms of this system of categories, we can improve on the description of
words as events: they are deeds. This means that we can join up Figure
2 and Figure 6, via the link in Figure 7, so that automatically 'word'
inherits all properties of 'deed1: every word has a human doer, a time,
and a place.
Figure 7.
d e e d = i word
This immediately gives us three of the elements we are looking for:

speaker, time, and place. So if we want to impose some restriction on
the speaker of a word, such as being American, we can do it in just the
same way as we did in requiring that a doer should be a person. You can
see how the restriction on cookie would be stated in Figure 8, and Figure
9 gives the meaning of now, by using the variable x. inherited from
'event'.
Figure 8.
cookie doer
22£L A z<, American
Figure 9.
To bring in the addressee, we have to refer to a more precise model

for 'word1, namely, 'communicative deed', which is in turn an instance
both of 'deed' and 'communication1, where 'communication' would include
static forms of communication like writing and pictures as well as
dynamic forms. We can assume that, by definition, a communication has
a source and a goal, or addressee, so 'word' inherits this companion from
'communication'. The details are not too important at the moment; the
main point is that by linking words to more general cognitive categories
for kinds of event, we can provide a principled theoretical basis for our
analysis of the structure of the situation.
Let me finish with a few words about speech acts. As I have already
pointed out, there is an interesting connection between semantics and
situational analysis, because we referred to general categories like
'deed', 'action', and 'event' in our analysis of situation structure, but
there is no obvious reason why we should not take these concepts as the
meanings of English words deed, action, and event, respectively. So we
have actually killed two birds with one stone: we have found a theoret-
ical framework for a cognitive analysis of situation structure, but at the
same time we have given ourselves the basis for semantic analysis of
three English words. My assumption is that we have only a single con-
ceptual structure corresponding to any given bit of experience; the
alternative would be to assume that we have one cognitive representa-
tion for processing deeds themselves, and a different one for processing
the word deed. We would need very strong evidence for such a perverse
assumption.
Now much the same seems to me to be true of speech-act categories.
I assume that we classify 1particular bits of speech in terms of categories
like 'invitation', 'promise , and so on; presumably, such categories are
further subdivisions of 'communication'. But we also have words like
invitation and promise to deal with, and we need to provide some kind of
conceptual analysis of their meanings. The logic of what I have just said
about deeds and so on is that the meaning of the word invitation is the
concept 'invitation' which I described as a speech-act category, and
similarly for promise and so on. If this is true, then we can see speech-
act theory as a branch of lexicography, concerned with the meanings of
words like invite and promise. Once we have worked out what these
words mean, we can be sure that at least these meanings exist, as con-
ceptual categories, in our minds, and the natural assumption to make is
that these are the concepts that we use in distinguishing one speech act
from another. If so, then there can be no conflict between truth-con-
ditional semantics and speech-act semantics, because the latter is one
particular area of the truth-conditional analysis of word meaning.
3. Conclusion. The main point that I have tried to make is that word
grammar is the kind of theory of language structure that sociolinguists
and their friends need, because it provides a view of language into which
our knowledge of situations of utterance can be integrated in a very
satisfactory way. I could have made similar points about the attractions
of word grammar for those interested in language acquisition, where I
think there is very clear evidence that the child's understanding of the
situation of utterance plays a crucial part (e.g. Barrett 1984). But even
if you are not keen on sociolinguistics or language acquisition work, word
grammar should help you in your work on language structure, and I hope I
have said enough to make you want to find out a bit more.
References
Barrett, Martyn D. 1984. Early semantic representation and early
word-usage. In: Semantic development. Edited by S. A. Kuczaj and
M. D. Barrett. New York: Springer.
Hudson, Richard. 1976. Arguments for a non-transformational gram-
mar. Chicago: University of Chicago Press.
Hudson, Richard. 1980. Sociolinguistics. Cambridge: Cambridge Uni-
versity Press.
Hudson, Richard. 1984. Word grammar. Oxford and New York: Black-
well.
Hymes, Dell. 1972. Models of the interaction of language and social
life. In: Directions in sociolinguistics. Edited by John Gumperz and
Dell Hymes. New York: Holt, Rinehart and Winston. 35-71.
Pullum, Geoffrey K. 1979. Rule interaction and organisation of a
grammar. New York and London: Garland.
Ross, John R. 1973. A fake NP squish. In: New ways of analyzing
variation in English. Edited by Charles-James N. Bailey and Roger
Shuy. Washington, D.C.: Georgetown University Press. 96-140.
THE MAKING AND BREAKING OF CONTEXT
IN WEST TEXAS ORAL ANECDOTES
Richard Bauman
University of Texas at Austin
Linguistically and sociolinguistically informed approaches to context
have had a significant effect on the study of oral narrative in recent
years, to the considerable enrichment of our understanding of the pat-
terns and functions that organize the place of narrative in social life.
But this has been a recent development; traditionally, for most folk-
lorists and anthropologists interested in oral narrative, a concern with
context has directed attention to the institutional context of expressive
forms, in the Malinowskian tradition, or to the context of cultural mean-
ing—what one has to know about the ethnographic particularities of a
society in order to understand what is going on in its stories—a more
Boasian concern (Bauman 1983).
As early as 1926, to be sure, Malinowski exhorted students of oral
narrative to attend to the situational context within which stories are
actually told (Malinowski 1926), but the call was never fully taken up
until comparatively recently. Under the impetus of the ethnography of
speaking, however, and the performance-centered approach to oral
forms that developed under its stimulus, the analysis of storytelling as
situated verbal activity—as part of the process by which the ongoing
conduct of social life is verbally accomplished—has burgeoned (Bauman
1977). Such studies have involved as well the examination of the ways in
which oral narratives are contextualized by the surrounding discourse
within speech situations (e.g. Bauman 1981), and the ways in which
stories may be collaboratively accomplished in conversational interac-
tion (e.g. Roemer 1980).
I have invested a lot in the promotion and practice of these analytical
approaches and I continue to believe strongly in their productiveness,
both for narrative studies and for linguistics. They have become rea-
sonably well established as part of our analytical repertoire, however,
and I would prefer to use this opportunity to advance another line of
language-centered contextual analysis that is as yet less fully developed
in narrative studies.
160
The making and breaking of context in West Texas oral anecdotes / 161
In a paper presented at the 1981 Georgetown University Round Table
on Languages and Linguistics, William Labov (1982) examines the dynam-
ics of speech actions and reactions, as portrayed in narratives of per-
sonal experience, as a means of elucidating relations of speech and ac-
tion in the social world reported by such narratives. Although he is
ultimately—and I think most usefully—concerned with the performance
and effects of speech actions in social life, Labov's use of narratively
framed accounts of social interaction requires of him a consideration of
the ways in which speech actions are endowed with social meaning by
their contextualization in the narratives themselves; the narrative
discourse creates a context for the reported discourse and renders it
meaningful.
For present purposes, the point I draw from Labov is that insofar as
acts of speaking are of focal importance in certain kinds of narrative, an
understanding of the ways in which these speech acts are contextualized
within the narrative can enhance our understanding both of how speaking
operates and is understood to operate in social life, and of how narra-
tives are constructed.
My interests in this regard are stimulated also by Bakhtin's suggestion
(1971:195) that:
the orientation of the word among words, the various perceptions of

other speech acts, and the various means of reacting to them are
perhaps the most crucial problems in the sociology of language
usage, any kind of language usage, including the artistic.
Of the various forms Bakhtin identifies by which speech may be oriented
toward another utterance, he is most interested in reported speech. I
share Bakhtin's interest in reported speech as social process and artistic
resource, but with special reference not to written, but to oral narra-
tives.
I propose to examine in this paper two oral narratives I have re-
corded—a sample of a larger corpus—in which reported speech is not
only an artistic device but the very esthetic focus of the story; the
reported speech is the maximally reportable act recounted in the narra-
tive (Labov 1982). In generic terms, the narratives I examine are anec-
dotes; the anecdote may be defined as a short, humorous narrative, pur-
porting to recount a true incident involving real people (B^dker 1965:26,
Botkin 1949, Rohrich 1977:6-8, Taylor 1970). The characteristic formal
features of the genre include a focus on a single episode and a single
scene, and a tendency to limit attention to two principal actors. As a
corollary, perhaps, of this last feature, anecdotes also tend to be heavily
dialogic in construction, often culminating in a kind of punchline, a
striking, especially reportable statement rendered in direct discourse.
That is to say, quoted speech is a significant stylistic feature of the
genre; accordingly, the anecdote would seem to offer itself as an es-
pecially apt focus for investigation of the formal and functional role of
reported speech in oral narrative.
The stories under study were recorded in 1972 from a West Texas
rancher, then 70 years old, whom I will call (pseudonymously) Caswell
Rogers. Here are the texts we will examine; the titles are my own,
supplied for convenience in referring to specific stories.
162 / Richard Bauman
NOT THAT YOUNG
CR: Jack was a good worker and a good cowboy too. He was...
Mrs. R.: And a good drinker...
CR: ...but he was a little heavy on the bottle. And the old man
...uh...he wanted him to work all the time, and Jack just didn't
see any use in that, workin' so much.
One day we were workin1 cattle and they had a pole cor-
ral...a pretty hot day and there's a bit of shade right around
the edge of this pole corral.
Ol1 Man Trimble came out and said, 'Say,1 said, 'come here,
sit down and rest a little bit,' says, 'those boys out there are
younger'n you are.'
I sat down there and Jack was doin' 'bout twice much work
as anybody else out there, and I knew the ol' man and he
was...been havin' a pretty hard time, so I told him...well, I
thought it would help Jack a little. I said, 'Now you see Jack
is doin' twice as much work as anybody out there.'
The ol' man said, 'Yeah, he sure is, and he's a good hand. I
try to help him and every time I try to help him, why he gets
off on one of these big sprees,' and said, 'I just can't help a
fella like that.'
I said, 'Well, Mr. Trimble, Jack is young,' I said, 'probably
you was young one time.'
He said, 'Hell, yes, but not that young!'
PASTURE FULL
Int.: What's that story, that Lawr-... that. ..when you and
Lawrence were on the jury and...and Shorty Hammond was
being tried for cattle rustling?
CR: Oh, they's tryin' Shorty. Shorty was always into something,
and he'd stole four calves from a fella, so they was tryin' 'im.
Lawrence and I was a prospective jury—they'd already picked
the jury, and...but we had to stay there.
So, uh, 'bout the time they picked the jury, why, Shorty
decided, or his lawyer decided, maybe he better plead guilty;
that might be the easiest way out, because they had the
evidence against him. So he decided to plead guilty, so the
Judge put him on the stand and got to questioning him and
said, 'Mr. Hammond, is it true that you stole those calves from
Mr. Bales?' (Stole 'em from Ira Bales).
Said, 'yes.1
'Well, Mr. Hammond, don't you know that's wrong to steal
cattle?'
Shorty said, 'yes.1
'Well, why did you do it? 1
He said, 'Oh, I got drunk, so I didn't know what I's doin'.'
Said, 'I do that every time I get drunk.1
Lawrence said, 'Aw, that's no excuse.' Said, 'I'd have a
pasture full if I stole cattle every time I got drunk.'
Let us attend first to the structure of the stories. In both, we may
observe, the maximally reportable act—that is, the point of the story, is
an instance of quoted speech. This bit of reported speech always occurs
at the end and brings the narrative to closure. Goffman has noted
(1974:559) that in informal talk, 'tales told about experience can (and
tend to) be organized from the beginning in terms of what will prove to
be the outcome. 1 Taking our lead from this observation, we may prof-
itably examine the organization of the highly end-oriented stories before
us in terms of the way in which they set up the climactic reported
utterance, that is, in terms of the way in which they accomplish the
creation of a context for the quoted speech that brings them to
closure. As we shall see, however, the punchlines of these stories are
reflexive: they loop back to reconstitute, or rekey (Goffman 1974:79-
81), what has come before. In this process, the antecedent portion of
the narrative, which has built up a context for the punchline, is itself
recontextualized. Accordingly, we need to determine both how the
punchline is set up and how the portion of the story antecedent to the
punchline is made available for such subsequent rekeying.
We may begin by observing that these are stories about morality-
proper and improper behavior, responsible and irresponsible action, and
attitudes toward them. The moral tenor of the stories is introduced
from the beginning; the first piece of narrative business that is per-
formed in these texts is the introduction of the central actors by refer-
ence to the problematic, morally loaded attributes that will make for
the focal conflict of the story. While the establishment of character
adumbrates the moral tension of the story, the character attributes that
are introduced are not bound to the event recounted in the narrative but
are antecedent to it, elements of character by which the individuals por-
trayed are more generally known in the community. While the central
actors are presented from the beginning in terms of morally colored at-
tributes, the initial section of the stories also serves to bring onto the
stage all other interactants in the narrated event, who may be impli-
cated in a variety of ways in the central moral conflict.
In addition to the introduction of the dramatis personae, the other
function performed by the opening section of the narratives is the set-
ting of the scene for the narrated event to follow. This involves the
establishment of the time and place of the central encounter and the
occasioning acts or circumstances that bring the dramatis personae into
the interaction that will in turn set up the concluding punchline (cf.
Chafe 1980:42, Colby 1973:654, Labov and Waletzky 1967:32).
I draw the introduction of the central actors and the setting of the
scene into a single section because they are not always separate and
sequentially ordered in the stories. While an element of character is
always presented first, not all the dramatis personae are necessarily
brought forward before the narrator moves to the setting of the scene.
Elements of time, place, or occasioning action may intervene before all
the principal actors are finally introduced.
In 'Not That Young,' the dimension of moral conflict is made explicit
in the introduction of Jack and 'the old man,' who is his grandfather.
Jack is a drinker—publicly known as one, as witness Mrs. Rogers' inter-
jection—and while he is capable of good work, his moral worth is com-
promised by the fact that he often chooses to go off on binges instead.
Thus, because the old man wants him to work all the time and Jack goes
off on sprees, his drinking brings about conflict within the family, a
serious problem. The old man is a cattleman, a cattle owner, which
implies a certain economic substance and status, while Jack, who simply
works for him, goes off on sprees. The moral contrast between the two
is strongly implicit. Mr. Rogers himself is the third character in the
story, and one of the two central interactants.
The scene of the story is set by references to time ('One day'), place
(Mr. Trimble's cow-lot, made up of pole pens), and occasioning action.
The latter, which functions to bring together the principal actors in the
narrative, is the 'working' of Mr. Trimble's cattle: the branding, ear-
marking, dehorning, inoculation, and castration of his new calves.
In the 'Pasture Full' story, Shorty is identified at the beginning as a
cattle thief, the moral valence of which needs no comment in this cattle
ranching community. His character as a trouble maker is amplified by
identifying him as someone who is 'always into something,1 a euphemism
for always doing things that cause trouble. He is, in short, identified as
a disruptive person. Lawrence is set up as the morally contrasting
figure. That he is a prospective member of the jury for Shorty's trial
implies that he is a respectable, upstanding member of the community.
The scene is set in the courtroom and the presence of the interactants
is accounted for: Shorty as defendant, Lawrence and Mr. Rogers as pro-
spective jurors required to remain after the jury has been selected. The
particle so_ marks a sequential movement toward the central narrated
event: 'So, uh, 'bout the time...' begins to situate the narrative event
temporally and leads into the further occasioning act represented by 'so
he decided to plead guilty.1 The next sp_ ('so the judge put him on the
stand') marks the transition from the setting of the scene to the onset of
the narrated event itself.
Let us summarize. We have seen that the initial section of both of
these anecdotes is devoted to the fulfillment of two complementary
functions: the introduction of the dramatis personae and the setting of
the scene for the narrated event to follow. To this point in our examina-
tion of the narratives:
(1) the central actors have been introduced in terms of morally
weighted attributes that they bring with them to the nar-
rated event, and most of the secondary characters have been
introduced as well;
and a variable combination of the following scene-setting functions have
been accomplished:
(2) the narrated event has been situated in place;
(3) the dramatis personae have been brought onto the scene of the
narrated event by certain occasioning actions or circumstan-
ces;
(4) the story has been situated in time by the use of time markers
such as one day;
(5) in one of the stories (and others in my corpus) a process of
narrative sequentiality has been set in motion toward the
onset of the narrated event, marked by the sequential parti-
cle so.
We are ready, then, to turn to a consideration of the narrated event
itself, the sequence of actions and reactions that is actually replayed for
us in the narrative and toward which the introductory section has led.
We have already noted that the essential part of the narrated event is
the conversational encounter that culminates in the punchline, but this
portion of the narrative may include other elements as well. Themati-
cally, the narrated event implicates a moral offense, which then pro-
vides the focus for the conversational encounter that concludes the
narrative. The narrated event concludes with a dialogic exchange cul-
minating in the quoted speech of the punchline.
The moral offense around which the narrated event revolves has al-
ready been adumbrated in the opening section of the story in which the
principal actors are presented in terms of particular morally weighted
attributes. In the two stories before us, the morally offensive actions
have taken place antecedent to the narrated event, namely, Shorty's
stealing of the cattle in 'Pasture Full' and Jack's going off on sprees
when he should have been working in 'Not That Young.'
In 'Pasture Full,' the narrated event consists wholly of the reported or
quoted speech of the trial proceedings. Part of the account is very
summary; 'the Judge put him on the stand1 may not be readily apparent
as reported speech, but it is in fact a summary of the verbal routine by
which the defendant is called to the stand. The next action, 'got to
questioning him,' is also summary, but here the speech act involved is
explicitly named. From this point on, the story consists entirely of
quoted speech with its associated framing devices.
'Not That Young1 is somewhat more complicated. After the opening
act of quoted speech, in which Mr. Trimble calls tMr. Rogers over to sit
down, followed by Rogers' response to the summons, the story momenta-
rily shifts away from the conversational interaction as Mr. Rogers tells
us more about the background of the exchange to follow. The essential
point here is that Mr. Rogers is a spokesman for Jack's position in the
conflict between Jack and his grandfather; Jack is not a direct inter-
actant in the reported encounter, but his present behavior within the
frame of the narrated event does influence the conversational inter-
action between Mr. Trimble and Mr. Rogers. Accordingly, we are given
an account of Jack's relevant action within the narrated event. This
additional information is to account for Mr. Rogers' motivation for
taking the tack he does in the ensuing dialog with Mr. Trimble. The
information is reserved for this point in the narrative because it does not
become relevant until Mr. Trimble calls him over. Only then does his
idea of smoothing things out between the old man and his grandson come
into play.
We arrive, then, at the conversational encounter itself, the core of
these anecdotes. As we move into the conversational encounter, an im-
portant shift takes place in the presentational mode of the stories, a
shift from the recounting of circumstances and actions—telling about
them—to replaying the actions, reenacting them to a degree by ostensi-
bly repeating what was done in the original past event of which the nar-
rative is an account. In the terminology of classical rhetoric, this may
be seen as a shift along the continuum from diegesis to mimesis, from
telling to showing. The great majority of the reported turns at talk in
the stories is rendered as direct discourse. The central encounter in 'Not
That Young1 has five turns at talk, all in direct discourse, while 'Pasture
Full' has seven turns, all once again in direct discourse. To be sure, the
mimetic closeness with which the original dialog is replayed is atten-
uated by the quotative devices that frame the direct discourse, but the
retention of the tense of the original quoted utterance—a basic feature
of direct discourse—enhances the sense of reenactment by transposing
the past into the present. And in 'Pasture Full,' the framing devices fall
away and the quoted speech is left to stand on its own.
As suggested earlier, even direct discourse is kept at a remove from
full reenactment of the purported dialog of the original event by the
quotative devices with which it is framed. Reported speech, especially
quoted speech, involves special problems of communicative manage-
ment, because the narrator is actually speaking for other people in
addition to himself. Accordingly, there is a need for ways of marking
the difference between the voice of the narrator in the present story-
telling context and the reported speech of the actors in the original
event being reported (one of whom, of course, can be the person who
later tells the story, but in a different voice), and of marking speaker
change within the conversational dialog that is the core of the narrated
event. The quotative frames are an important means of accomplishing
these tasks. An essential constituent of the quotative frames is the
verbum dicendi, the verb of saying. In the stories before us, the verba
dicendi used to introduce the quoted speech are all said, with a single
exception—one instance of told, as a false start in 'Not That Young.'
In addition to the verba dicendi, there is a further range of devices
which serve to organize the reported speech of these stories. In fact,
the organizing system for reported speech in these texts is highly redun-
dant, with multiple devices operating concurrently to indicate who is
speaking and when. Most of these can be illustrated from the dialog in
•Pasture Full1:
...so the Judge put him on the stand and got to questioning him
and said, 'Mr. Hammond, is it true that you stole those calves from
Mr. Bales?' (Stole 'em from Ira Bales).
Said, 'yes.'
'Well, Mr. Hammond, don't you know that's wrong to steal cattle?'
Shorty said, 'yes.'
•Well, why did you do it?'
He said, 'Oh, I got drunk, so I didn't know what I's doin'.' Said, 'I
do that every time I get drunk.'
Lawrence said, 'Aw, that's no excuse.' Said, 'I'd have a pasture
full if I stole cattle every time I got drunk.'
First of all, reported speech may be attributed. The speaker may be

identified by a variety of means: by pronoun (he said), by name (Shorty
said), or by some other identifying term (the Judge...said). Similarly,
reported speech may be addressed to an identified addressee, as in Well,
Mr. Hammond... Third, the beginnings of quoted utterances may be
marked by particles such as Well, Oh, or Aw, that tend to occur only in
this initial position, while the ends of these utterances are marked by
transitional pauses that are longer than those that may occur within
quoted utterances. And fourth, the conversational encounter may be or-
ganized by recognizable conversational structures and routines, such as
the routine of courtroom interrogation in the passage above, made up of
question and answer adjacency pairs. This routine serves so well in

assisting us to differentiate among speakers that two of the quoted turns
at talk are replayed without recourse to any quotative frames at all.
While cohesive ties such as this may be found throughout the replayed
dialogs in these stories, helping to give coherence to the interactions
portrayed, cohesion assumes special stylistic importance when we get to
the punchline that brings the narratives to closure: the last lines of the
stories are saturated with anaphoric ties to the lines that precede them
to a far greater degree than at any other point in the stories. A demon-
stration of these ties (based on Halliday and Hasan 1976) will open the
way to a more extensive discussion of the role of the punchline in these
anecdotes, toward which1 much of our analysis has been directed.
In 'Not That Young, the patterns of cohesion show up especially
clearly when we note that the penultimate line is itself made up of two
segments, marked by the repetition of the quotative frame, I said. If we
break the line down into the two resultant shorter segments, we find
that the punchline has strong anaphoric ties to both of them, intensifying
the sense of cohesion across the concluding lines of the dialog.
I said, 'Well, Mr. Trimble, 3ack is young.11
I said, 'probably you was young one time.
He said, 'Hell, yes, but not that young!'
An inventory of the cohesive ties between the punchline and what pre-
cedes it would include the following:
(1) the parallelism of the opening quotative frame: I said (2x)/He
said;
(2) the rhyme of the particle that opens the quoted utterance:
HelhWell;
(3) the assent marked by yes, probably to both parts of Mr. Rogers'
statement, but certainly to the suggestion that he 'was young
one time';
(4) the contradiction marked by the adversative conjunction but,
qualifying the assent of the yes that precedes it;
(5) the contrastive demonstrative not that, qualifying the assent
concerning his earlier youth still further;
(6) the lexical repetition of the key word young.
Like the penultimate line in 'Not That Young,' the quoted utterances
in both concluding turns at talk in 'Pasture Full' are divided into two
segments, each introduced by the quotative frame. Here, though, de-
spite the fact that both lines are relatively long, the patterns of co-
hesion extend across the whole lines.
He said, 'Oh, I got drunk, so I didn't know what I's doin'.1 Said, 'I
do that every time I get drunk.1
Lawrence said, 'Aw, that's no excuse.' Said, 'I'd have a pasture
full if I stole cattle every time I got drunk.1
The ties include:
(1) the parallelism of the initial quotative frame;

(2) the parallelism of the particles Oh and Aw that open the
quoted utterances;
(3) the demonstrative reference that's which points back to the
preceding line;
(^) the word no, which repudiates an aspect of the preceding line;
(5) the naming of the illocutionary force of Shorty's preceding
statement, namely, excuse;
(6) the repetition of the medial quotative frame, said;
(7) the substitution of stole cattle for that in Shorty's statement;
(8) the parallelism of the entire concluding phrase, varying only
the tense of get;
(9) the lexical repetition of drunk which is tied to both uses of the
word in the penultimate line.
Let us examine the effect of the strong cohesion between the punch-
lines of the stories and the preceding lines that set them up. It would
not be too strong to say that density and multiplicity of the ties between
them set the concluding lines of the stories off to a degree from the rest
of the text and give them the quality of a kind of closing couplet. The
multiple repetitions, parallel constructions, and other forms of anaphora
contribute directly to the strong sense of closure that is achieved by the
punchlines (Smith 1968:158-71). When the punchline is spoken, we know
that the story is complete simply in formal terms.
The cohesion between the punchline and the line that comes before it
is such that the punchline impresses us as a transformation of the pre-
ceding line—similar to it and modeled upon it, but transforming it in the
process. The punchline contains within it two voices—its own and that
of the preceding speaker upon which it has wrought a transformation.
Most important, this double-voicedness is a correlative on the formal
level of the double-voicedness of the punchline on the level of meaning
as well. That is, the formal nature of the punchline is a correlative of
its fundamentally ironic function. To demonstrate this, we need to
examine the content and social interactional structure of the reported
conversational encounters that lead up to, and culminate in, the punch-
lines of the stories.
We may begin by recalling that these anecdotes are moral stories; the
conversations are about moral issues: decorum, responsibility, work,
order. How are these issues developed in the interactions between the
principal actors?
In 'Not That Young,' we find that the story is artfully constructed
from the very beginning in terms of a set of paired contrasts between
youth and age, and working and avoiding work. In the narrated events,
these contrasts are evoked by the old man, Mr. Trimble, when he calls
Mr. Rogers over to rest in the shade while Jack, who is "younger'n you
are' and doing 'twice as much work as anybody else out there' continues
to work. While Mr. Rogers, knowing of the friction between grandfather
and grandson, takes the opportunity to speak on Jack's behalf, Mr.
Trimble, from his place in the shade, criticizes Jack, who is working
especially effectively, for occasionally going off on sprees; this is one of
the major ironies of the story. Rogers, in turn, attempts to account for
Jack's irresponsibility by attributing it to his youth, the same youth that
enables him to work so vigorously. Young men will be irresponsible at
times, but at least this is compensated for in Jack's case by his ability to
do the work of two men when he does work. Then, using the rhetorical
strategy of identification, Rogers points out that Mr. Trimble was young
once too, and here Mr. Trimble is caught, for he was known in the com-
munity for his own escapades and was fully as young as Jack. What is
more, people knew it.
From this vantage point, the latter half of Mr. Trimble's response, 'but
not that young,' appears on the surface as an after-the-fact gesture at
upholding the position he has been taking as a spokesman for sobriety
and responsibility, but it is understood by Mr. Rogers as an acknowledg-
ment by the old man of the inconsistency of his attitude toward Jack. In
explaining the story to me, Mr. Rogers made clear that he sees Mr.
Trimble in the punchline as in effect stepping back from the line he has
sustained throughout the exchange, and rereading the situation and his
own stance toward it. The punchline amounts to a concession that he
knows he is taking the situation too seriously; it objectifies the situation
and assumes a relativist stance. Not that work and responsibilty aren't
valid ideals—they are—but one has to recognize that life doesn't—per-
haps shouldn't—always work that way.
'Pasture Full1 works to similar effect. In this story, the interaction of
the narrated event is organized in terms of the routine of interrogating a
defendant in a court of law. There are clear contrasts operating in the
story between the agents of morality and the legal order, namely, the
judge and the jury, and the morally culpable defendant, a confessed
thief.
The interaction proceeds with the judge doing the questioning and
Shorty, having decided to plead guilty, doing the remedial work of ac-
knowledging the wrongness of his act. Ironically, though, what Shorty
offers as a mitigating factor in his guilt actually casts him in a light that
is worse yet: not only is he a thief, but a drunkard as well. We have
already established the strong tension surrounding drunkenness in this
region.
Right after Shorty's confession of his double moral lapse comes the
response by Lawrence that serves as the punchline. Lawrence, as a
prospective juryman, aligned with the agents of morality and order
against the hapless defendant, says 'Aw, that's no excuse. I'd have a
pasture full if I stole cattle every time I got drunk,' suggesting, though
with some joking exaggeration, that he often gets drunk himself. At one
level, this statement plays on the ambivalence felt in this region toward
alcohol; Lawrence likes to drink, sometimes even to excess, but he
doesn't let himself get out of control like Shorty. Most important,
however, his statement shifts the moral alignment in the courtroom:
Lawrence is now identified with Shorty, as someone who gets drunk,
even while denying the validity of Shorty's excuse, which is, of course,
truly 'no excuse.1 By his words, he has broken the moral alignment that
has prevailed and rearranged it, reframed the situation. Again, he does
not deny that Shorty is guilty and behaved wrongly—he just sets up a
relativistic alternative concerning what can be morally acceptable in
real life, regardless of ideal public standards.
Now that our analytical path through these anecdotes has led us to an
understanding in formal and functional terms of the efficacy of the
punchlines that bring them to closure, we can see more clearly just how
tightly structured these stories are. The punchline is the crucial ele-
ment, the point of the story. But the punchline in turn depends closely
upon the line that precedes it and on the social interactional and sub-
stantive thrust of the entire replayed conversation that constitutes the
core of the narrated event. All of these elements are rooted in a par-
ticular moral tension that is the subject of the conversational inter-
action and gives the punchline its ironic and relativistic impact; this
tension is adumbrated from the very beginning of the story in the intro-
duction of the dramatis personae. Thus, from the introduction of the
principal actors, to the setting of the scene that brings them to the
central encounter, to the conversational interaction between them, to
the punchline that caps it off, the parts of these anecdotes constitute a
markedly tight structure.
In both cases, the anecdotes achieve their effect by rekeying the
situation, recontextualizing it, overturning the apparent direction of the
interaction and the moral alignments and attitudes that have seemed to
control it, and establishing an ironic alternative, not as a substitute but
as a coexistent perspective. The effect of the punchline is to that
extent subversive, a breakthrough both on the part of the one who is
reported to have spoken it, and on the part of the narrator, into a kind of
skepticism and relativism that takes pleasure in refusingjto take ideal,
normative moral expectations too seriously—a 'comic'corrective,1 in
Burke's apt phrase, 'containing two-way attributes lacking in polemical,
one-way approaches to social necessity' (1937:213).
This, in fact, is the essence of much humor. Indeed, upon examina-
tion, these stories may seem to have some basic affinities with other
humorous expressive forms. The punchline in many narrative jokes built
upon reported speech, for example, works by reframing what has come
before it (Sherzer 1982). In all, the 'successful subversion of one form by1
another completes or ends the joke, for it changes the balance of power
(Douglas 1968:365). To carry the correspondences still further, tra-
ditional verbal jokes also represent a form of reported speech. They are
often introduced by reference to the person from whom they were
learned, as in 'Wanna hear a joke my sister told me?' (cf. Sacks 1974),
and insofar as they are known to be in oral tradition, they are in a sense
reported out of the abstract collective voice of tradition. And again, as
noted, many narrative jokes employ fictional reported speech both as
stylistic device and as punchline.
Given all these correspondences between jokes and other humorous
routines and the anecdotes of our West Texas storyteller, what are the
differences among them? Part of the answer to this question is sug-
gested by a consideration of the context in which these stories occur.
They are conventionally told in a variety of small group sociable settings
in which the conversation deals with the members of the local com-
munity and the surrounding region. In recent years, they have figured
most prominently during visits by the narrator to members of his family
who have moved away from home, or on occasions when those relatives
have come back on visits of their own. Conversation on such occasions
often involves catching up on the people of the community—births,
deaths, marriages, divorces (in recent years), and other significant ac-
tivities. The community is small and the ranching and farming region
around it is rather thinly populated. People are still identified in terms
of residence ('Cal lived back this side of Johnny's1) and kin connections
('Ms. Brown, you know was one of 'em—Joe Bob's mother'), with the lat-
ter providing one of the major organizing principles by which successive
people are brought up for mention and discussion. Of course, kinship is
not the only salient social feature by which people are known; as in most
small, traditional, agrarian communities, the personal and social identity
of individuals is also defined in part by their actions and experiences,
elements of their local social biography. It is here that narrative comes
into play; stories are the major means by which such actions and exper-
iences are memorialized and given expression. Thus, the mention of a
given individual may evoke a story about him or her, either a personal
narrative in which the teller figures with that individual, or a third
person narrative about the individual in question that has been told to
the narrator by someone else at varying degrees of remove from the
original event. That is, the chain of transmission may be of varying
length, but there is always a sense of locality and familiarity about the
dramatis personae of the stories—they are all known personally or in
terms of their connections within the community: kinfolk, neighbors,
friends.
Because these stories are about known and familiar individuals and
constitute a part of their social biographies, they are densely indexical
in a concrete social sense. That is, part of their meaning derives from
the great complex of indexical associations that they evoke—the individ-
uals portrayed, other known aspects of their lives and characters, and
potentially all the other people in the community, including those pre-
sent at the storytelling event, with whom they are linked by the kinds of
social and communicative ties that give cohesion to the conversations in
which the stories are told.
To be sure, these stories, like all literature used as equipment for
living (Burke 1941), have a certain metaphorical as well as metonymic
meaning, as a kind of extended name or label for the recurrent social
problem situations they portray: the embarrassment occasioned by pub-
licly visible immoral behavior, the damaging of someone else's property
through careless incompetence, and so on. And to extend the Burkean
perspective still further, the stories also convey an attitude toward such
situations and a strategy for dealing with them. The attitudes will vary
depending upon the situation, but there is always an attendant unease
about the public moral conflict the stories portray, and the favored
strategy that emerges from the stories is to alleviate the resultant
tension by ironically transforming the ongoing situation into something
else. It is here that we see the importance of the crucial bits of quoted
speech that bring the stories to closure; what is highlighted in these
anecdotes is the transformative capacity of speech. Bakhtin maintains
that 'The speaking person in the novel is always, to one degree or anoth-
er, an ideologue, and his words are always ideologemes...It is precisely as
ideologemes that discourse becomes the object of representation in the
novel' (1981:333, italics in the original). So too in these anecdotes, but
the ideology to which the last speaker gives voice is ultimately ironic
and skeptical (White 1973:37), showing how the normative pressures of
morality that lead to social tension may be evaded by those with the
verbal wit to do so.
Traditional punchline jokes, as many have pointed out, have the same
subversive potential. Unlike our anecdotes, however, jokes are not at all
rooted in community; they are anonymous, impersonal, generalized.
Indeed, if one were inclined toward speculation, one might suggest that
the modern punchline joke, which emerged as a recognized form only in
the nineteenth century (Rflhrich 1977:4, 8), might have evolved out of
the punchline anecdote under the social conditions of the modern indus-
trial era. Anecdotes of the kind we have been examining thrive in the
intimate social environment of the small local community, whereas jokes
belong preeminently to the impersonal milieu of urban industrial society
(Rtthrich 1977:9). As imaginative products ungrounded in a known com-
munity of real individuals, jokes can only be metaphorical and specula-
tive in their relationship to actual experience. They tell us in abstract
terms about how structures might fall apart or be overturned, while the
true anecdotes are told to keep us aware of the vulnerability of life as it
really _is and the capacity of speech both to make this vulnerability
apparent and to bring it under control.
Notes
1. Concerning the transcriptions: my representation of spoken lan-
guage is, frankly, intended to have more expressive than linguistic ac-
curacy in a strictly formal sense. I am more interested here in the
narratives as oral literature than as dialectological data. No words have
been added or deleted (ellipsis indicates pauses, not deletions), no gram-
matical constructions 'corrected 1 , no eye-dialect introduced, but I have
been concerned to convey that this is a record of language in a spoken,
not a written mode, and to preserve something of the quality (however
vague and impressionistic that term may be) of oral discourse. To this
end, I have selectively employed a variety of devices, some of them in
themselves conventions for representing oral speech in print, some of
them attempts to capture certain features of local pronunciation as em-
ployed by the speakers. Above all, I would emphasize that no pejorative
connotation of any kind is intended by the mode of presentation I have
employed (cf. Preston 1982).
2. I am using the term 'mimetic' here solely to identify a presenta-
tional mode, with no claims implied concerning the degree of actual
correspondence between the original event and its representation in
narrative discourse. In the process, I am begging some very large issues,
for an excellent discussion of which see Sternberg (1982).
References
Bakhtin, M. M. 1971. Discourse typology in prose. In: Readings in
Russian poetics. Edited by L. Matejka and K. Pomorska. Cambridge,
Mass.: MIT Press. 176-96.
Bakhtin, M. M. 1981. The dialogic imagination. Transl. by C. Emerson
and M. Holquist; edited by M. Holquist. Austin: University of Texas
Press.
Bauman, R. 1977. Verbal art as performance. Repr. ed. 1984. Prospect
Heights, 111.: Waveland Press.
Bauman, R. 1981. 'Any man who keeps moreYi one hound'll lie to you1:
Dog trading and storytelling in Canton, Texas. In: And other neigh-
borly names. Edited by R. Bauman and R. D. Abrahams. Austin: Uni-
versity of Texas Press. 79-103.
Bauman, R. 1983. The field study of folklore in context. In: Handbook
of American folklore. Edited by R. M. Dorson. Bloomington: Indiana
B^dker, L. 1965. Anecdote. In: International dictionary of regional
European ethnology and folklore, vol. 2. Copenhagen: Rosenkilde and
Bagger. 26-27.
Botkin, B. 1949. Anecdote. In: Standard dictionary of folklore, my-
thology, and legend, vol. 1. New York: Funk and Wagnalls. 56.
Burke, K. 1937. Attitudes towards history, vol. 1. New York: The New
Republic.
Burke, K. 1941. Literature as equipment for living. In: The philosophy
of literary form. Baton Rouge: Louisiana State University Press.
293-304.
Chafe, W. L. 1980. The deployment of consciousness in the production
of narrative. In: The pear stories. Edited by W. L. Chafe. Norwood,
N.3.: Ablex. 9-50.
Colby, B. N. 1973. A partial grammar of Eskimo folktales. American
Anthropologist 75.645-62.
Douglas, M. 1968. The social control of cognition: Some factors in joke
perception. Man 3.361-76.
Goffman, E. 1974. Frame analysis. New York: Harper and Row.
Halliday, M. A. K., and R. Hasan. 1976. Cohesion in English. London:
Longmans.
Labov, W. 1982. Speech actions and reactions in personal narrative. In:
1981. Edited by D. Tannen. Washington, D.C.: Georgetown Univer-
sity Press. 219-47.
Labov, W., and 3. Waletzky. 1967. Narrative analysis: Oral versions of
personal experience. In: Essays on the verbal and visual arts. Edited
by 3. Helm. Seattle: University of Washington Press for the Amer-
ican Ethnological Society. 12-44.
Malinowski, B. 1926. Myth in primitive society. London: Kegan Paul,
Trench, Trtlbner.
Preston, D. R. 1982. 'Ritin' fowklower daun 'rong: Folklorists1 failures
in phonology. Journal of American Folklore 95.304-26.
Roemer, D. 1980. Interjected routines as metanarrative commentary.
Working Papers in Sociolinguistics 68. Austin: Southwest Educational
Development Laboratory.
Rohrich, L. 1977. Der Witz. Stuttgart: Metzler.
Sacks, H. 1974. An analysis of the course of a joke's telling in conversa-
tion. In: Explorations in the ethnography of speaking. Edited by R.
Bauman and 3. Sherzer. New York: Cambridge University Press.
337-53.
Sherzer, J. (in press) Puns and jokes. In: Handbook of discourse ana-
lysis. Edited by T. Van Dijk. London: Academic Press.
Smith, B. H. 1968. Poetic closure. Chicago: University of Chicago
Press.
17* / Richard Bauman
Sternberg, M. 1982. Proteus in quotation-land: Mimesis and the forms

of reported discourse. Poetics Today 3(2).107-56.
Taylor, A. 1970. The anecdote: A neglected genre. In: Medieval
literature and folklore studies. Edited by T. Mandel and 3 . Rosen-
berg. New Brunswick, N.J.: Rutgers University Press.
White, H. 1973. Metahistory. Baltimore: 3ohns Hopkins University
Press.
ENTER TEXTUALITY: ECHOES FROM THE EXTRATERRESTRIAL
Thomas A. Sebeok
Indiana University
Jorge Luis Borges once observed that every writer of fiction creates
his own precursors by modifying each of his reader's conceptions of the
past. The same retrospective axiom surely constrains every other sort
of text fabricant, from scientist to, in the case at point, cinematic
auteur. For purposes of this presentation, a 'text' will be regarded as
any significant object, or, more technically, as any coherent string of
signs. Although the internal simplicity or complexity such a string may
display is not at issue here, multistranded strings tend to be more fas-
cinating than single filaments, and syncretic aggregations—of which film
offers a conspicuous example—even more so.
The Janus-faced concept of 'intertextuality', which, in contrast with
Borges' dictum, works as much prospectively as it does in a retrograde
scape—by extension of M. Baxtin's original, although hardly precise,
formulation of heteroglossia, dialogism, and polyphony—denotes ways in
which works of art (especially of literature) are produced in response not
to social reality but to previous works of art and the codes and other
conventions governing them. In Kristeva's reformulation (1969:146),
'tout texte se construit comme mosa^que de citations, tout texte est
absorption et transformation d'un autre text. A la place de la notion
d'intersubjectivite s'installe celle d'intertextualite...1 Intertextual codes,
which Barthes characterized (1974:10) as a 'mirage of citations', are
ultimately insubstantial, every reader having become a more or less
representative embodiment of a vague, generalized intertextuality; as he
went on to write in 1970: 'This "I" which approaches the text is already
itself a plurality of other texts, of codes which are infinite or, more
precisely, lost (whose origin is lost).' To the degree to which a work of
art is 'intertextual1, it becomes distorted, opaque even, a darkly specular
reflection of actuality—as, for instance, a myth. It becomes a lattice of
signposts, regressing into, effectively, infinity, and thus capable of
sustaining many alternative interpretations. Yet it may become a
dialectical (vs. sequential) tool for furthering the study of typological
goals far more tellingly than the more indeterminate conception of
175
176 / Thomas A. Sebeok
'influence', and, as such, may assume various shapes, including, notably,

allegorical.
I once argued (Sebeok 1979:xiii), modifying Samuel Butler's famous
brocard about a hen and an egg, that a sign is only a sign's way of making
more signs. I learned only recently that B. F. Skinner (1983:30) made an
analogous claim, 'that a poet is only a literary tradition's way of making
more of a literary tradition, and...that a scientist is only science's way
of making more science.' It is in line with these views that all texts
must be regarded as 'intertextual', although, of course, variations, by
definition, do occur on all levels. These are the mechanisms for in-
creasing information, the measure of novelty, with the passage of the
ages.
In 1859, Darwin instructed the readership of his Origin of Species that
'all the living forms of life are the lineal descendants of those which
lived long before the Silurian epoch', and we now have an impressive
body of evidence that the direct-filiation hypothesis, as it has since
come to be called (Margulis 1981:37), applies universally to the patterns
of evolution of all animals and plants, indeed of all genomes; (true, the
evolution of eukaryotes from prokaryotes seems to require the invoca-
tion of an additional principle of symbiosis, but, even so, the combined
effects lead to the formation of precisely the same phylogenies). Note
that the concept of 'intertextuality', with its actual or implied groping
for ultimate codes whose origins are lost, postulates a body of discourse
assumed to be already in place and thereby allows for the creation of the
works in posse, envisaging the signifying practices of texts to be pro-
duced by some future weaver of yet another text. In fact, it is a tacit
extension into the cultural sphere of the canon of direct filiation, so well
recognized in the domain of our biosphere. If, as Pierce taught, man is a
sign, or a text (Sebeok 1979:60f., 1981:6), then each and every one of us
is the 'intertextual product1, in quite a literal sense, of myriad prior
biochemical developments and innumerable precursors—minimal self-
replicating entities, leading to polygenomic cells—that populated the
Earth since the common ancestors of life arose, more than three billion
years ago, from organic compounds on the surface of our planet, organ-
izing themselves into macromolecular systems of signs capable of stor-
ing and transmitting information.
In this paper, I propose to examine Steven Spielberg's artful and en-
trancing transfiguration of the archetypally familiar boy-and-his-pet
story (the everyday carnal) into a sci-fi fantasy (the mythic supernal), as
enacted in E. T. The Extraterrestrial, and attempt to decipher how the
unfamiliar trappings, transcendental reverberations, and subliminal
currents thereof come into focus in consequence of a consistent applica-
tion of intertextual techniques and effects (the de"ja-vu, as well as the
de'ja-lu, and even the deja-entendu).
The march of variegated visual and auditory metonyms—disembodied
footsteps and jangling keys, unoccupied moving swings and toy soldiers,
ominous breathing and musical notes, Reese's pieces and clinical charts—
which, together, and more, constitute the organizing pith of this picture,
its energetic armature, if you will, is punctuated by Elliott's and E.T.'s
ligamentary emblematic geranium: it wilts, but, in the end, it re-efflo-
resces in its pot. The tropes are carefully balanced out on multiple
planes, one of which is cast in a botanical idiom. The visitors, who touch
Enter textuality: Echoes from the extraterrestrial / 177
down on a California forest clearing, densely begirded by tall trees and
lush foliage, are back, we learn, on a plant-gathering field trip. As the
crew of this peaceable expedition is about to be captured by deperson-
alized but ill-boding government agents, the aliens take off in their
pumpkin-shaped spaceship (the film's texture is ingrained with carnival-
esque images of, particularly, Halloween and Christmas), inadvertently
abandoning one of their kind. The focal meaning of the geranium was
underscored by Kotzwinkle (1982:183), who wrote a novel based on
Melissa Mathison's screenplay, as he ends his book with this pregnant
sentence: 'He went into the misty light, with his geranium.'
The marooned 'old botanist' (an epithet sometimes used to charac-
terize E.T.) is left shuffling through the grass as the ship's hatch shuts,
petals folding inward. He finds his way to the vegetable garden of
Elliott and his family, rounded out by a teenage brother, a younger
sister, and their mother, lately singled (dad having taken off for Mexico
with Sally, a girlfriend). The lonesome alien, yearning for his home,
three million light-years distant, and the fatherless (i.e. 'alienated') boy
are united by a natural bond of companionship.
The movie is obviously a direct, although undeclared, continuation of
Spielberg's Close Encounters. Kael (1982:119) remarked that 'that's
partly because this film has the sensibility we came to know in that
picture, and partly because E.T. himself is like a more corporeal version
of the celestial visitors at the end of it.' The link, moreover, is non-
verbally signaled by John Williams' variations on an unmistakable theme,
the scores serving to enhance both scripts as well as those of the tripar-
tite Star Wars saga released so far.
With palpable deliberation, Spielberg intersows self-references to his
own and kindred previous works. During the Halloween scene, among the
costumed children fanning out to go haunting, there is a little figure
dressed up as Yoda, the gnostic gnome introduced in The Empire Strikes
Back. E.T., an 'old monster' now in ghostly disguise, turns his head in
recognition of the Jedi, along with the complicit members of the audi-
ence, as a colleague who has, as it were, strayed into this picture from a
neighboring set.
Jaws puts in a brief synecdochical appearance in the guise of a scoop
in Elliott's fish tank, while the tub of live frogs he liberates during
biology class to prevent the chloroforming and dissection of these Ker-
mit-like creatures, themselves reminiscent of the eye-boggling appari-
tion of the extraterrestrial ('Are you a Muppet?' Elliott's sister, Gertie,
asks), evokes the episode featuring Marion, the heroine in the snakepit,
from Raiders of the Lost Ark, underlined by a comic shot of a school-
mate of Elliott's standing as if she were paralyzed in midst of the pan-
demonium of escaping frogs, crawling over hands and feet. The climac-
tic big chase sequence, the children and the space-goblin fleeing military
vehicles on bicycles that sometimes fly high in the sky, by application of
a low-level antigravity formula, recalls several high-powered cliff-
hanger pursuits from the same 1981 film and similar tricky effects from
still earlier Spielberg pieces. The contrast between the sophisticated
military vehicles and other lethal equipment at the disposal of the
uniformed members of the search party—their space-travel gear, espe-
cially their gleaming head masks, bring to mind the terror of Darth
Vader in carapace and his undifferentiated metallic troops—with the
kids' ordinary yet bewitched bikes, is neatly captured as Elliott and 'the
ancient fugitive' soar in the dusk against the moonlight. This remarkable
shot is another icon of Halloween, and the one that instantly made me
think of the Good Witch from The Wizard of Oz, with which the film
shares still further implicit but deep figurations (as it does, if more
tangentially, with Meet Me in St. Louis). In preparing for Halloween, by
the way, Mary, Elliott's mother, actually plays the Good Witch: looking
beautiful, like a star-creature, she carries a star-wand to touch trick-or-
treaters on the head.
E.T. shares paradigmatic qualities perhaps most explicitly with J. M.
Barrie's Peter Pan, passages from which are read aloud in the movie.
Tink's dying and getting well again if children do and declare that they
believe in fairies—E.T. himself avers his belief—prefigures the extra-
terrestrial's own revival.
Such intertextual lineages are rooted, I believe, in Spielberg's sensibi-
lity; they could easily be traced out further: to Robinson Crusoe, for
example (Kotzwinkle 1982:125), or Humperdinck's Hansel and Gretel
(again, the Reese's Pieces), and selected passages from the imagination
of Jules Verne and H. G. Wells. Although Spielberg supplied the idea for
the picture, the screenplay was developed by three writers, among them
Melissa Mathison, who was one of the scenarists for Carroll Ballard's
1979 film, The Black Stallion. She is probably responsible for the em-
phatic development of the boy-and-his-pet theme, sparked by Walter
Farley's 1941 novel, which inseminated numberless sequels. She also
seems fond of lonesome boys who are bereft of their father, a motif that
is autobiographically congenial to Spielberg as well.
The fatherless ten-year-old Elliott and his forsaken unearthly familiar
are, however, much more than friends. In a profound sense, they are
identical, as the boy's very name, ElliotJT, insinuates. Communication
between the two of them progresses rapidly from sign language to Earth
language: 'the language center of his marvelous brain came fully on, a
thousand stored tongues reappearing, as reference and cross-reference
took place, so that Earth's language could be viewed in the round. He
grasped its fundamentals, and then its delicate edges' (Kotzwinkle
1982:75). It continues via two-way telepathic connections. When E.T.
gets plastered on a six-pack of beer at home while his telepathic sender
is in full force, Elliott suffers bizarre effects in school, along with the
loaded source. When E.T.—who, by the way, is constituted of DNA like
all the rest of us—becomes mortally sick, Elliott does so in analogous
fashion, as reflected by electronic scanners and displays. Kotzwinkle
(1982:167) comments: 'The boy and the monster were linked somehow, as
if the monster were feeding on the child's life. The child came in and
out of consciousness, hallucinating, babbling, sinking under again. I'd cut
the cord that ties them, thought the doctor, if only I knew where and
what it was.' E.T., of course, recovers ('Do you believe in fairies?1), and
so does Elliott; the wilted geranium blooms again, simultaneously.
The 'intertextual' pedigree of E.T., traced out so far to literary,
filmic, musical, and autobiographical lines of descent, crisscrossing but
controlled in a highly craftsmanlike manner by the omnipresent director,
must be regarded as a pattern that works only skin-deep. Notwithstand-
ing that these interwoven elements of its surface structure may suffice
to convey the gratification derived from wonted up-beat narrative
Enter textuality: Echoes from the extraterrestrial / 179
constituents distanced by exotic trappings, they do not, by themselves, I
think, account for either the genuinely tearful cathartic enchantment
many viewers report experiencing during this movie, or its record-break-
ing box office success. These must be ascribed to the archetypally
subliminal religious infrastructure in which the film is soaked. It is thus
turned into a calculated yet unpretentious allegory.
Before ascending to this level of inspection, however, it might be
worth pointing briefly to yet another reason for E.T.'s lachrymose al-
lurement. The physical design and traits of E.T. enable us to descry him
as an idealized bundle of what ethnologists call biological 'releasers'.
Although, to be sure, E.T. is not actually an animal, he is, as mentioned,
a very special pet of Elliott's, exemplifying to perfection Morris'
(1967:230) first law of animal appeal, which states that 'The popularity
of an animal is directly correlated with the number of anthropomorphic
features it possesses.' E.T. does so, in superoptimal degree.
A preternatural creature descends to the Earth from his unearthly orb,
inscribed with a delicate gothic design ('as if an enormous Christmas
tree ornament had fallen from the darkness...the greatest heart-light the
world has ever seen'—Kotzwinkle 1982:182). He is found in a tool shed,
arguably a suburban Southern California rendition of a manger. He is
adopted and protected by a family, the mother of which is named Mary;
(it is, incidentally, worth noting that Mary is the sole adult female in the
picture, like Princess Leia Organa was in Star Wars). The visitor pro-
ceeds to perform magical deeds, like making mundane balls float in the
air and rotate in orbit. He executes wonderful healings and levitations.
Indeed, his presence is referred to several times as 'miraculous'.
Love is the message E.T. brings to Earth, and, among the Earthlings, it
is the children he attracts as his immediate disciples. He is hounded and
persecuted by sinister legions of faceless pursuers, undergoing suffering,
death, resurrection, and, at last, ascension, to his home. Holding onto
his geranium, he embraces Elliott, and delivers an intricate wave-sign of
benediction, 'to release the child from the narcosis of the stars' (Kotz-
winkle 1982:182). 'I'll be right here,' he finally pronounces, as his lumi-
nous fingertip glows over the boy's chest.
We, who are left behind, can look forward to a Second Coming.
Postscript. This essay was finished before the release of Ron
Howard's entrancing comedy, Splash. While its lyrical surface sheen has
seemingly nothing in common with E.T., a detailed analysis of the under-
lying structure reveals that this film is almost a point-for-point trans-
form of the earlier masterwork. As the extraterrestrial descends from
outerspace, so the Thalassic creature ascends from below, both to mas-
ter language in identical fashion. Both plots are about all sorts of love,
although Allen's is consummated both sexually and romantically with the
ravishing mermaid, whom he finally joins (after a compulsory chase
scene), forever after, in her maritime abode. Although space will not
allow me to particularize my argument here, the dense 'intertextuality'
between these two movies needs at least to be pointed out, and, else-
where, to be explored much further.
Note
1 • Several auditors of this lecture have called my attention to a scene

where E.T. hides from the children's mother in a toy-filled closet. As
she opens the door, he blends so well with the dolls that she fails to
notice him. I am told that this sequence is intended to recall a mirthful
pause in Chaplin's 'The Circus', which opened in 1928, but which I have
never seen. The sequence is briefly described by Huff (1951:213) thus:
'Charlie freezes into a wooden statue among other figures on a Noah's
Ark, as the cop dashes past.' The principle involved, in both cases, is, of
course, that of Poe's 'The Purloined Letter'.
References
Barthes, Roland. 1974. S/Z. New York: Hill and Wang.
Culler, Jonathan. 1981. The pursuit of signs: Semiotics, literature,
deconstruction. Ithaca, N.Y.: Cornell University Press.
Huff, Theodore. 1951. Charlie Chaplin. New York: Henry Schuman.
Jefferson, Ann. (in press) Intertextuality. In: Encyclopedic dictionary
of semiotics. Edited by Thomas A. Sebeok et al. 3erlin: Mouton.
Kael, Pauline. 1982. The pure and the impure. The New Yorker, June
14.
Kotzwinkle, William. 1982. E.T. The extra-terrestrial. New York: G.
P. Putnam's Sons.
Kristeva, Julia. 1969. Semiotike: Recherches pour une semanalyse.
Paris: Seuil.
Margulis, Lynn. 1981. Symbiosis in cell evolution: Life and its environ-
ment on the early earth. San Francisco: W. H. Freeman.
Morris, Desmond. 1967. The naked ape: A zoologist's study of the
human animal. New York: McGraw Hill.
Sebeok, Thomas A. 1979. The sign and its masters. Austin: University
of Texas Press.
Sebeok, Thomas A. 1981. The play of musement. Bloomington: Indiana
University Press.
Skinner, B. F. 1983. Origins of a behaviorist. Psychology Today
17.9:22-33.
ON THE PRAGMATIC 'POETRY1 OF PROSE:
PARALLELISM, REPETITION, AND COHESIVE STRUCTURE
IN THE TIME COURSE OF DYADIC CONVERSATION
Michael Silverstein
The University of Chicago
The natural, public habitat of linguistic structure is in the intersub-
jective complexities of communication. This is the realm of inter-
actional pragmatics. Regularities in such communication that can be
formally described might be termed pragmatic structures; many linguists
have been seeking to account for them within linguistics, or to explain
why one should not or even cannot. My discussion is aimed at demon-
strating one such kind of structure by example, a 'poetic' structure of
linguistic forms that develops coherence in the realtime work of dyadic
conversation. Such poetics of prose has fundamental importance, I
believe, for clarifying some of the issues emerging in debates concerning
the integrity of grammatical form, the varieties of meaning, and the
contextualization of language use. Perhaps this relevance will be more
vivid when cast in terms of current linguistic controversy over self-
styled 'functionalist' and 'formalist' approaches.
For linguistics in recent years, the sentence has been the hero for the
orthodox, whose functional and formal autonomy are to be celebrated—
or, for nonbelievers, the villain whose grammatical riches, only appar-
ent, are to be reattributed to other phenomena. Assumed to be a tran-
scendent unit of abstract theory, the autonomy of the sentence is gener-
ally defended as being tantamount to the autonomy of linguistics as a
science with pretensions to real 'explanation' in its own terms. (Note the
echoes of Saussure and of Bloomfield.) Assumed to be a self-contained
totality of theoretically relevant structures, the formal integrity of the
sentence in consistent structural terms is seen to mark the most signifi-
cant dividing line of a structural hierarchy of inclusiveness, and hence of
types of scientific accountability for language. Assumed to be a syntac-
tic structure of uniquely determinate (or at least decidable) proposi-
tional value, the sentence condenses and interrelates numerous function-
al speculations: on the modeling in logic of some hypothesized system of
individual cognition; on its structured mirroring in intersubjective rep-
resentational forms (for example, in surface syntax); and on the depend-
181
182 / Michael Silverstein
ence of distinctly verbal communication and hence of linguistics on

isolating some unifunctional meaningfulness in an abstract logical
model. The sentence, then, is—like the taxonomic phoneme of yore, to
some—at a triple intersection of epistemological, formal, and functional
assumptions. No wonder the frequently heated debates have been so
concerned to attack or to defend its specification in particular!
Here, the problem of linguistic form is approached in a manner some-
what different from the 'text grammatical', 'functional', or 'discourse
grammatical 1 points of view, as these have been represented in much
contemporary work. My demonstration here is based on distinguishing,
first, principles of sentence-grammatical form, possibly equivalent to
the transcendent ('second-order' mental), autonomous ideal structures
that the model-fitting strategies of orthodox syntacticians seek to
characterize. Let us agree that, however partial in specifying public,
surface-structure forms, these principles define the sentence as a
grammatical entity. These principles, however ultimately minimal,
would determine something of the recursive structure of unboundedly
complex, logically relevant ideal forms said to underlie the 'conceptual
creativity' of syntax. This level is not much in evidence in the following
conversational material.
Second, we should identify surface forms of what we might call
sentence-scope, as these unfold in realtime discourse. By the first
assumption, these are at least partly instancing or realizing the prin-
ciples of syntax, and at most partly something else. Syntax in the ear-
lier sense almost certainly cannot specify everything about such sen-
tence scopes; however, what is syntactic in the strictly autonomist-
formalist-logicist sense must be definable on evidence only within the
bounds of sentence-scopes. Anything larger as the data-base would
threaten the triplex of assumptions that divide abstract, transcendent,
formal sentence syntax from actual realtime, immanent, functionally
multiplex discourse.
To be sure, syntacticians have now 'clarified' their positions to mean
that most likely certain realtime, perhaps cognitive-psychological, or
even actor-strategic 'functional' principles operate with respect to
production/reception of some aspects of the ultimate shapes of sen-
tence-scopes as occurring performance units. (The interactions between
these and autonomous syntax are decidedly vague, however.) But strict
syntacticians do not countenance formal, abstract, and transcendent
principles of the very same order as syntactic ones determining struc-
tures larger than the sentence-scope in occurring discourse.
The problem of formal ordering, on which my demonstration centers,
illustrates the opposition rather nicely. For, in the public domain,
discourse unfolds in a time-bound, linear 'sequentiality'; ever since
Saussure's postulate on the matter (Q916J 1960:103), this sequentiality
has been explicitly differentiated from the concatenational 'syntag-
maticity 1 of the transcendent domain of grammar. Syntagmatic order of
elements is, of course, a common syntactic principle, one even in our
own English language relating to Agent-Patient relationships. Compare
also enclisis, as in Wackernagel's Law in Indo-European, and similar
positional restrictions on specific classes of syntactic constituents.
But such regularities seem to be based on qualitative principles of
relative order. They can always be formulated with respect to some
On the pragmatic 'poetry' of prose / 183
particular level of syntactic/morphological 1constituency—and hence, to
this degree, are always 'structure dependent , note—in terms of which an
ordinal (or positional) rather than cardinal (measured), and local (within
constituency-defined scope) rather than global (defined on the entire
scope), principle can be formulated. By contrast, as Jakobson (1960:358)
so brilliantly and oracularly formulated it, 'poetic function [pf language
pragmatical projects the principle of equivalence from the axis of selec-
tion into the axis of combination.' This means that formal units, as units
of a poetic order of language structure, depend for their very definition
and special kind of meaningfulness—'equivalence' or any of its variants,
including antonymy, in the Jakobsonian idiom—on the fact of their
cardinal combinatory positions within the sequential order of actual,
linear discourse. The fundamental poetic principle, then, is sequential
measure, or meter, in terms of which the linear signal can be measured
off into units of whatever sort.
In the most highly developed poetry—to be distinguished, of course,
from poetic principles as these apply to language in general (Jakobson
1960:359)—the aesthetic form generally consists of multiply overlaid,
hence both cyclic and hierarchy-generating, measurements, frequently
at several different structural planes of propositional linguistic form
simultaneously, e.g. several phonological measures concurrent with
morphosyntactic and lexical ones. However, meter, in the sense of the
basic fact of poetic organization, need not rest on specifically phono-
logical measure; rather, anything in linguistic form can serve as the
basis for poetic pragmatic structure, such form being defined with
respect to the metric principle(s). The data here, for example, illustrate
such poetic structure in the referential-and-predicational sentence-
scope forms of a developing dyadic conversation.
To follow the demonstration, one must understand a few principles
and concepts from our heritage of semiotic functionalism. First, I will
be using some reasonably uncontroversial proposals about English surface
grammatical constituency within the sentence-scope, with no ultimate
claim implied about their reducibility/autonomy as sentence-grammat-
ical. Second, I will be using the concept of parallelism, involving a
measured repetition that serves as the metered frame for positional
variability. This very elementary type of poetic structure (see Jakobson
1966; Bricker 1974; Fox 1977) can be developed in a purely sequential
fashion, continued or not whenever a significant unit boundary is
reached, and hence nicely merges the notions of local and global meas-
urement. It is perfect as a form developed through strategic contingen-
cy. Third, I will make reference to the theme/rheme status of linguistic
units (see Firbas 1964; Danes' 1970). These can be identified in the
meter-framed variable portions of sentence-scope within the poetic
structure. Hence, these are relative to English surface structure, in-
cluding stress-level phenomena, within the sentence-scope stretch of
discourse.
The use of these few tools gives us the ability to understand several
aspects of how this conversation is integrated as a developmental struc-
ture in time, an 'inter'-subjective, publicly negotiated creation in which
poetic organization plays a central role both in cohesion or 'texture'
(Halliday and Hasan 1976:1-30) and in what have been termed 'speech
acts' (Searle 1969) and illocutionary/perlocutionary effects (Austin
18* / Michael Silverstein
Q962 1975). The mutually entailing cooccurrences of forms—special,

text-internal indexical relationships that make text part of the context
of occurrence of any sign—induced by the way forms have positions in
the induced or constituted poetic organization privilege the construai of
'topics', denotata that, once introduced, can be maintained as true
referents with pro-forms or even zero explicit linguistic forms. The
parallelisms and repetitions of the material that remains in poetic
sequence contribute the determining structure, independent of any
sentence-level contribution to surface sentence-scopes. In this view, the
syntactic problem of 'recoverability'—or reconstitution of grammatical
and referential identity—of empty, absent, or deleted elements of
would-be sentence-scopes, is an artifact of the grammar-level per-
spective on the way poetic organization determines a discourse structure
of topics and comments.
In the material that follows, in fact, it is interesting to see that as
conversation moves forward in its linear time frame, certain momenta-
rily ambiguous pro-units and 'deletions' are resolved by implication as
part of the conversational work much later than at the point at which
they first occur. But that is as we would expect, since the sentence-
scope fragments that occur in parallelistic structures develop certain
thematic elements by rhemes that have linear ordering relationships.
Conversation moves through a specification of its unique and topical
referential content from among whatever is the residual set of possibil-
ities at the moment of ambiguity, in due course the specification being
the explicit thematic base for a rhematic form that serves as comment
on the topical referent.
This linear progression of syntactic poetics played out in time
achieves the semblance of what some would reconstruct as 'indirect
speech acts' (Searle 1975). Rarely in this conversation, as in the rest of
our experience outside of formal occasions, does anyone issue an explicit
metapragmatic formula of canonical 'performative' shape, or even of
recognizably conventionalized sort. What we find, rather, is that the
poetically established sequence of thematic/rhematic elements sets up a
dynamic, analogical quasi-syllogism of concepts in realtime. The playing
out of this quasi-logic is differentially contributed to by the two partici-
pants, and shows differential cohesion within vs. across interlocutors.
From such a public, emergent structure, not necessarily from the propo-
sitional interpretations of sentences implied by sentence-scopes plus
vague canons of so-called implicature, can any interpretability of speech
act 'moves' (or any equivalent actor-centered reconstitution of intents in
a psychological philosophy) proceed on an empirical basis.
Were we to view abstract sentence grammar as an immanent, rather
than transcendent, cognitive psychological mechanism—a 'first-order
mental structure1—as some are unfortunately wont to do, we would
constitute the problem of discourse coherence in terms of the tension
between the hierarchical organization of information, as revealed in
syntactic structures, and the apparent linearity of realtime-bound pro-
ductive/receptive discourse performance. But there is no direct tension
of this sort. For to make this leap is to forget the triple 'decontext-
ualization1 that sentence-syntax constitutes vis-a-vis public linguistic
data. For example, the hierarchical 'chunking' of conceptual informa-
tion, whatever its cognitive reality, is not merely reflected in hier-
On the pragmatic 'poetry1 of prose / 185
archical forms, such as super-/subordinate syntactic structures of sur-

face sentence-scopes, which are at a completely different order of
phenomena, regardless of the fact that the superordinate and subordi-
nate portions of sentence-scopes occur in certain language-particular
ordering relationships. Poetic pragmatics, I would maintain, mediates
among these orders. It is a transcendent structure in exactly the same
sense as syntax, but it is developed in realtime through principles that
are inherently linear, yet abstract and formal. Along this poetic dimen-
sion of structure, some true understanding of immanent 'functional'
constraints may be forthcoming.
Let me illustrate this in an example. The data I discuss come from
the set of dyadic conversations induced, videotaped, transcribed, com-
puter-stored, and in several respects analyzed by my colleague Starkey
Duncan and his associates at The University of Chicago, as material for
studies of nonverbal interpersonal turn-taking behavior (see Duncan and
Fiske 1977 for documentation ).
The transcript as organized and presented here needs some explana-
tion. I have organized the contributions of the two subjects, labeled A
and B, into two parallel columns. I have labeled the explicit questions
(Q) and responses (R) in pairs, with a numerical subscript index of its
position within the total conversational sequence of each respective
speaker. Thus Q A £ labels A's sixth explicit question form in the tran-
scribed conversation. The temporal dimension is represented from left
to right on every line of transcript, and vertically down the transcript
for sequential utterances. Every line represents talk no earlier than the
line above it. When there is an overlap in the conversation, this is
indicated by putting braces two lines high around the overlapping seg-
ments; material appearing on the resulting upper line of a two-line brace
is simultaneous with that on the lower line of the next occurring brace
according to the temporal dimension conventions. Thus, in (6), ...cation
changed L7Q0msec3 a... spoken by B is simultaneous with ...overwhelming.
spoken by A. I give a section number, in the left-hand margin, to each
explicit Q—R sequence, or to each whole poetic structure of a subject,
plus its associated material. Speech is continuous, and silences in the
range of ca. 20-30 msec are indicated with an em-dash; long pauses are
indicated by specifying their approximate length in brackets.
Linguistic material is indicated in a close-to-standard English ortho-
graphy, indicating only unusual stress and intonation-height charac-
teristics, i.e. those distinctive in degree and/or placement given surface
syntax. Since participants refer to each other with the personal deictics
2 and you, these have been indexed with subscripts, _i = subject B, _j_ =
subject A. Unambiguously restorable material is, in a few cases, neces-
sary for the transcript in English orthography to make sense; this is
indicated in square brackets. Also indicated, by Qj is material that,
relative to the parallelistic structure of sentence-scopes, would be
reconstituted as missing constituent(s) by a sentence-syntactic view of
sentence-scope completeness. Finally, according to the method of
maximizing parallelism and poetic structure in a determinate earlier-to-
later (i.e. realtime) fashion, I have used outline-style indentation to
various margins to highlight units of equivalent relative status in the
poetic structures that develop.
...there...
(1) Hu'tih, An' n how do you. like Chicago compared [to 0]
n
y
did you. go to school there 6v uh, (wa
A7 '
I. did go to school there
i
I. went to school h£re Also, ffim

Oh , uh-huh
um, so I. came back kind of
I
Oh, uh-huh
(2) Q._ An' you. w€nt to undergraduate 0 h€re 6r
0 In Chicago fit, uh, Loyola
(3) Oh Oh Oh Oh Oh I 'm an 6Td Jesuit boy myself., unfor tuna tely|
(4) Oh fire ya. 0 n Where'dl you. g6 [to
3 g J
Bl
„R 0 [at] Georgetown, down fin Washington
A1 ~ '
0:h yâhj, yeah
(It's) too bad
I.— [710 msec]
(5) Did you. finish 0
Qa9 J
Um Y£ah W611 this Is my. second y£ar here
J
Oh, uh- huh
, uh, I. don't know,
It was nice
I* sorta enJ6yed it —
This place is really really — di different —

I. mu1 —
Yeah
I, must aAy
But, uh, T710 msec]
I. don't know,
I. I. enjoyed the education there

<X2> And it rfially was g6od
It 9
(6) I . think
Jesuit edufeation changed [700 msec] a
overwhelming
|
16t in the lfist five or six y£ars
An' I, think
I i Just caught (v
l^ caught the tail end of uh — 6t the — really 61d "<.

sch6ol —
O
i-h
"V
It was Q
fto
I. — was really *
8S
I . mean — '***
'cause Ij Ij did my. dndergraduate work jjy
I. finished that like four years a [690 msec] go ^

five years ago V}
<
And — n
I ± think rt
3'
Ndw — pr6bably Loyola is a 16t different
an' a lot better [690 msec]
y'. know
a 16t more — variety of courses
being offered et cetera
find —
ydu. know
— I,...
The two participants in this excerpt are males in their 20s not before
acquainted; they have been asked to sit facing each other and to con-
verse. Subject A is a graduate student in the Law School, B in the
School of Social Service Administration. Their conversation moves
through several phases, from (i) mutual introductions, to (ii) self-con-
scious mutual excuses, through (iii) trying to find mutual topics of dis-
course. In searching for topics, A and B start with mutual prior ac-
quaintances; it turns out that each had previously participated in a
similar conversational dyad with a female student known to the other
(relative participant gender was a variable in the study). Their search
continues in question (Q) and response (R) sequences about prior, cur-
rent, and future experiences. We enter during the first of these, when A
has been asking about where B came from 'before1, i.e. before coming to
graduate school at The University of Chicago. Up to the point at which
we enter, it has been exclusively A's role in the interaction to ask expli-
cit questions, B's role to respond to them.
At the point we enter, it appears that A had just been trying to find
out at what school B had been, but had gotten, instead, a response about
where B had lived. (Up to this point, the ambiguity of the form Chicago
as denoting both place and institution has allowed this room for maneu-
vering.) B's introducing his prior residence, Iowa, prompted several
sequences of 'Trivial Pursuit' about the state, its towns and cities, geo-
graphical loci last referred to by B with the pro-form there, as indicated
on the transcript. A then begins anew to clarify B's origin in the sense
of establishing old school ties, making the conversational seam in an
ingenious way.
In (1), A brings the conversation's informational structure back to the
main rank of questions-responses instanced some turns back (A: Where
did you come from before? B: Urn, Iowa. I lived in Iowa.) A poses his
sixth such question in the conversation (labeled QA^), HOW do you: like
Chicago compared C 3 ? which is a conventionalized conversational
follow-up, complete with stressed lfke, to the piece of information on
where the interlocutor had lived before, taking Chicago as the name of a
place, the deictic here and now/after, as compared with the place Iowa,
the deictic there and then/before. Note that A's question lacks com-
pletely any syntactically expected coda, ...to there/...to Iowa, which is
of course fully recoverable from the thematic status of Iowa/there in
preceding discourse.
Immediately, without any pause to give B the turn to respond, A
focuses on the now-locational reading of Chicago vs. Iowa (= there).
Formulating it with the stressed rheme theVe, A once more asks his
question about prior education in the seeming guise of clarifying the
rheme like Chicago compared C Z) with a parallel rheme of an almost
parenthetical nature (By the way, did you:...), stressing and thereby
making focal the recoverable topic there in an ambiguous—and thus
perhaps 'politely indirect'—yes/no and either/or structure: Q ^ Did
go to school theVe <5r...? Note that the utterance indexically presupposes
that B went to school somewhere as asked. The material that might
complete the disjunction introduced by 6r_ is, of course, entirely unim-
portant to reestablishing the question on the table (though leaving room
for the possible response that B had lived in Iowa for some other rea-
son). B responds now to the pointedly defocused (and hence, for polite
conversation, more compulsive) specificity of the predicate go to school

with its emphatic, confirmatory form, including the stressed auxiliary do_
(in past tense). Abstracting syntactically from the assertorial dfd go, we
find that B's utterance R^j is otherwise the precise syntactic counter-
part to A's question Qpj> cnanging only the personal deictic from you^ to
Jj, because of the shift of deictic origin relative to the speech-event
roles. We should also observe that the form I| went to school there is
also precisely parallel to B's earlier remark at the same information-
rank, I: lived in Iowa, formally contrasting go to school with live and
specifying it further. (In the instance, the form there has its precise
interpretability from the long-standing thematic status of Iowa/there,
maintained by both A and B in exactly analogous syntactic positions
since an earlier part of the conversation.)
The participants have both been maintaining reference to Iowa with
the form there. There vs. the implicit deictic origin here (included in
the denotation of Chicago so far) is now to be established as a major,
explicit contrast in rhematic position. Speaker B continues his response,
reintroducing Chicago vs. Iowa as two geographical loci in the final,
contrastive portion of an entirely parallel sentence, Ij went to school
he*re lisp. Observe that B's new information is in the contrast of loca-
tives, and hence he uses a sequence of stresses in the rhematic portion,
here also, and the nonassertorial went instead of did go, as in his pre-
vious utterance. This is followed by some space-filling pause forms (urn
urn...) during which A indicates his registration of the utterance turn
(Oh, uh-huh), and then B continues. He gives a summary statement that
establishes a time sequence, so L came back kind of. With its normal
stress peak on the rhematic final portion, the verbal came back, this
utterance unit concludes with a modalizing parenthetical, kind of. In
this way, the deduced conclusion is highlighted, in a way that, for exam-
ple, Ij came back here, kind of would not do, except with extra heavy
stress on the phrasal verb came back. Given the perspectival deixis of
the verb, one can, of course, come back only if here—i.e. as established,
Chicago—both preceded and followed there—i.e. as established, Iowa. B
begins to make this coming back more precise, in the context of ques-
tions about going to school, saying Ij waca..., but A interrupts, register-
ing B's utterances with Oh, uh-huh and then continuing to pursue his line
of questioning about prior education.
In this eighth question-and-response sequence, A uses a form that is
again parallel to the utterances in Qpj—RQJ> now uninverted in order,
asking for a confirmation of a particular deduction about the time
sequence of B's here-and-there-and-here education that B has just set
up. The operative predicate, in parallel utterances, for most of the
interaction has been go to school, and A continues to use this, asking
more particularly if undergraduate school is to be paired with the first
location here of the sequence entailed by RQ-/' The utterance an' you^
we*ht to undergraduate heVe 6*r has form parallel both to A's previous
question and to B's responses; it deletes the repeated noun school and
uses merely its would-be modifier undergraduate—this has ben the focus
of A's questioning all along—and uses a final stressed heVe in the maxi-
mally rhematic position; for B's having gone to undergraduate school
here, i.e. at or in Chicago, is one way of understanding coming back. So
note that there is a certain implicative continuity of this confirmatory
question, following as it does upon the came back rheme. But its form is
cast into the proper syntactic mold of the unfolding structure of paral-
lelisms in utterances that establish the specific here and there reference
points.
The next utterance, B's, labeled R^g, is at once a perfect structural
reply to the question and the continuation/conclusion to B's last, inter-
rupted syntactic unit in (1). In Chicago cit, uh, Loyola specifies two
things and at last definitively disambiguates thereby locational and
institutional referents. First, it specifies with in Chicago that Chicago
is indeed now intersubjectively to be taken in the geographical-locative
sense, in keeping with the sequence of heres and theres just developed.
Second, it confirms that going to (undergraduate) school here can no
longer be taken to be Chicago, The University of (which would be a_t
Chicago). Third, it specifies the long sought-after information, jk—note
the contrastive and disambiguating stress--uh—note the pregnant hesita-
tion—Loyola. (As we shall see, throughout the rest of this portion of the
interaction, as indeed beyond it to the conclusion of the videotape, B
demonstrates considerable tension about this datum.) So this phrase is
properly a syntactic conclusion either to the recoverable I: went to
undergraduate school... or to the earlier, interrupted 1^ waBJ... Given
the tight parallelisms determining the sequences of syntactic forms, it is
irrelevant for us to try to decide/determine if one or the other may have
ben 'intended1; the seeming structural ambiguity of cohesion—or the
double determination of the syntactic fragment by rich parallelism—does
not affect its information content in the flow of the discourse.
At this rhematic news of Loyola, subject A jumps in, further to intro-
duce himself, making the discovered commonalty between B and himself
an intersubjective reality. I:'m an old Jesuit boy myself, he remarks,
with mock-humorous qualification unfortunately uttered simultaneously
with B's response to this revelation. B's response is the inverted rhetor-
ical question form of discovery, Oh are ya:? deleting the entire re-
coverable predicate noun phrase of A's utterance, but otherwise perfect-
ly parallel with it, stressing the finite verb. B then continues without
interruption to deliver the expected information question, Qgi, begin-
ning interactional segment (4), Where'd you- g6c 3? Observe that the by
now well established to undergraduate school is~not necessary to the
explicit signal to keep the sense of gc> here the specific one in use, i.e.
'attend'. The syntactic form is thus precisely parallel to A's questions
and B's responses in (1) and (2), with a preposed WH-form. A's answer
here, R^l* is the perfect counterpart to B's earlier R^g, moreover, with
a discourse chiasmus in the explicit material. Deleting all would-be
Subject and Predicate material except the rhematic essentials, A's
utterance is1 equivalent to 'D: went to (school at)3 Georgetown, down in
Washington. Note the mirror-image syntactic symmetry between Rgg
and R^|*. in Chicago at, uh, Loyola vs. pat) Georgetown,...in Washington.
This exchange, in the rhetorical figure of a syntagmatic cross
(X-Y...Y-X), is in fact the crux of the interaction. Up to this point, only
A has posed questions. At this point B, having once asked his first, now
keeps the interrogator's role for the rest of the transcribed conversa-
tion. A's volunteering of his own personal information in (3) shifts the
roles by letting B assume the questioner's role in (4). Indeed, after B
registers the response to this with the assenting O:h ye'ah, yeah of
recognition, A begins to formulate another comment, and hesitates,
leaving a silence of some 710 milliseconds, relatively long for this inter-
action. (He seems to be waiting for B to pose a question, having ex-
changed interaction roles with him.) Before he desists, A begins the
syntactic outline of a comment he will fully develop, it turns out, in
segment (5), a discourse structure with finite clauses of form rt Q>e]
^ j and clauses with Subject h in a complex and regular
pattern.
But, A having hesitated, B offers his next question, Qg2 Did you:
finishC 3 ? where the syntactic completion with either a quasi-object or
a locative again is fully recoverable in any of a number of ways: Did
you: finish... school/undergraduate school/at Georgetown [now equal to
A's thereTwe should observe]/down in Washington/there/etc. All these
are possible sentence-level completions fully in keeping with the
thematic cohesion of the particular syntactic unit in Q$2> kut ^ e r e again
it is irrelevant to the structure of discourse at this point to make a
unique syntactic determination.
Observe how, whatever the 'correct' conclusion to B's question Qg2»
the suppressed reference sets up an equivalent there for A, i.e. where he
was before or then, parallel to the there vs. here distinction that has
been the emerging contrast for subject B in the conversation thus far.
So, in answer to B's question, A replies in R a 2 affirmatively to having
finished there, and then he goes on to prepare for the continuation of the
utterance he began but interrupted in his earlier turn. We'll thus is my
second year here, he remarks. This introduces in maximally rhematic
position the explicit contrast between a current here, which can only be
the institution, The University of Chicago, and the former there,
Georgetown University, of two years earlier and more. (It should be
noted too that this affirmative utterance is precisely the syntactic
parallel of A's own earlier question Q ^ , not included in our sample
transcript, Is this your^ first year here, or? Note additionally the con-
sistency of question style, a blend of (optionally inverted) yes/no ques-
tion and disjunctive (either)/or question, exactly as in Q/^j and Q^g
discussed earlier.)
6h, uh-huh, B registers, finishing the last syllable as A continues.
And—perhaps A rushes so as not to lose the opportunity through turn-
transition to express what he hesitated on before—L don't know / It was
nice / I: sorta en|6yed it. Let us term this sequence of three clause
structures with no explicit connectives a | - b | - c j . After an unstressed
misfire (L), A continues with a fourth unconnected affirmative c l a u s e -
let us call it d|—with a heavy stress on both initial deictic topic and
final rhematic portion, thus: Thus place is really re'ally—di different—
emphasizing the contrastive value of both.
Then comes a parenthetical. The structure of this parenthetical is
interesting, since it appears to demonstrate in its misfired first portion,
L mu'—, something of the compulsive rhythmic or metrical nature of
sentence-scope discourse units, considered in phonological terms. The
expected parenthetical stress contour, when achieved, would of course
have predicate stress, h must say or, as here, L must say. A begins to
use a discourse unit stressed initially on L, just like units aj and Cj,
which form, with this unit—let us term it e—a pattern of alternating
sentence-scope discourse units. After B's intercalated registration of
On the pragmatic 'poetry1 of prose / 193
A's discourse so far, Yeah, A repairs the stress contour by redoing the
full-clause parenthetical, 1- must sSy.
Let us look a bit more closely at the sequence a^-bj-c^-d^-e. It is a
sequence of full, simple clauses, every odd-positioned clause beginning
with L. Read together, these give a rather hedged judgment about
personal experience, I- don't know, I- sorta enjoyed it, L must say. The
form it here clearly refers to something in the past, perhaps the same
thing as the rt of clause b j , It was nice, if we wish to make a cohesive
totality of a^ through e. On the other hand, out of context, the sen-
tence-scope bj could possibly include the initial form it just as a place-
filler for the predicate be nice. Just from the sequence bj-Ci, on which
such cross-sentence local cohesion relations are definable, it is not clear
how to take the subject it of bj in relation to the object It of c j ; they
may be coreferent, they may not be.
However, once we look at the overall structure of the sequence of
poetic units, things are clarified. Units bj and dj have a structure with
parallel be + [Evaluative!]^;* t>i with past tense was and dj with present
tense _is. Having already established his prior there and his current here,
A is clearly contrasting these explicitly in topical position in bj vs. d j , it.
vs. thfs place, corresponding to the deictic tense differentiation, was vs.
h>. This leaves no room for doubt as to what A is driving at. The con-
trast of the evaluative adjective nice in bj with really really different in
dj shows that different here is a value-laden index of speaker disappro-
bation, heard in colloquial usage in some parts of the country (This sure
tastes/smells/feels/etc, dftferent means n n 'I don't really like the way
this tastes/smells/feels/etc.')
Now A continues his development with a contrastive But, uh—com-
pare the And, uh with which he began this whole sequence—and then
pauses for a significantly lengthy time before plunging ahead. I- don't
know, he continues, using a unit we can term a2 to emphasize its poetic
identity with a j , t- b enjoyed the education there, a unit parallel to Cj
and so termed C2 here. It is this unit that finally clarifies the reference
of object It in Cj, when we consider it in its poetic position and in light
of the contrastive rhematic stress on the education. (Observe: What
was it you liked / Why'd you like the place? I (sorta) enjoyed the educa-
tion there.) Jjt in c j and there in C2 seem to be coreferential.
And it really was gcfod, continues A, with a unit we can label b 2 . This
unit parallels the form of b^, It was nice, with the addition of the adverb
really from d j . Observe that, as bj is to dj in the earlier sequence, bj
about the past there and dj about the present here, dj having the incre-
mental really on its evaluative adjective, here the position of really is
shifted to the evaluative of b2« And then A concludes with a unit d2
that contrasts with b2 just as dj contrasts with b j . Since bj and b2 are,
in essence, the same, ultimately d2 contrasts with dj by three pairs of
parallelisms: this place vs. rt, js vs. was, and really really different vs.
not overwhelming, whence different is (by contrastive poetic logic) not
not overwhelming or overwhelming. The ultimate message of contrast in
dj vs. d 2 is that 'this place' (The University of Chicago) is what it/there
(Georgetown University) is not, really really different, i.e. overwhelm-
ing. Things have gone from good to bad.
Observe, by the way, that B has already begun segment (6), following
upon the explicit form of A's C2-b2, as A is uttering the last unit of (5),
/ Michael Silverstein
do. Observe also how B pauses during the utterance of overwhelming

simultaneous with his own speech.
Subject A has thus developed a sequential structure of the form a j -
bi-Ci-di-e-a2-C2-b2~d2) with an internal chiasmus -bi-Ci- : -C2-b2~
within the basically bipartite structure of repetitions. The there of then
in the past tense is characterized as good, nice, and enjoyable; the here
of now in the present tense is characterized by contrast as different,
overwhelming. And all this has been achieved not with explicit logical
connectives, as with principles of poetic parallelism, determining a
verse-like structure for privileged rhetorical contrasts and cohesive
anaphoric continuity. Hence, in the transcript, the various correspond-
ing units are aligned according to the overall structure, and the chiasmus
is noted with a marginal Greek X. A complete structure, it forms a
segment by itself.
In (6), B takes up a commentary on the portion of A's turn (5) that has
preceded his entry into speaking. A's turn culminated, then, at di in the
rheme re*ally really dffferent. (For note that a2~C2~b2 essentially re-
capitulates the content of the first 'verse', making some of the refer-
ences clearer.) B proceeds to give a reading of the opposition set up
between then and now as enjoyable vs. different that is of a piece with
the hesitation and avoidance we noted in his introduction of his own
undergraduate educational experience prior to being at The University of
Chicago. He proceeds to fashion a discourse about the then and now of
his own it/there, Loyola University, in a sequence of units that form a
remarkable continuation of the style of A in (5), independent declara-
tives structured by parallelism and rhythmic repetition, containing but a
single quasi-logical connective ('cause), and that only in an explanatory
subsection.
I: think, B begins, overlapping with A's last sentence, and mirroring
A's 1: don't know of a^ 2* L e t u s c a ^ t n * s u n i t a 'l> s o a s t o emphasize the
structural parallelism! B further mirrors A's education there (i.e. at
Georgetown) of C2, the last referent mentioned in the rhematic portion
of A's utterances, in making the thematic seam to his own conversa-
tional contribution. And, for the rhematic portion of his observation, he
mirrors the rheme of A's culminative d j , (be) different, only in its pro-
cessual equivalent: changed, including even an equivalent of the in-
tensive really r6ally, namely, a I6t. Given the processual formulation of
the 'difference' between then and now, this time shift needs an interval
formulation, such as that B supplies in his utterance. Thus: Jesuit
education changed...a lot in the last five or six years. Notice how the
time interval, the last five or six years, is highlighted as the focal por-
tion of the rheme, and how, in so doing, the stressed last brings us up to
the conversational now. As an utterance unit, the whole is a statement
about an it/there but making rhematic continuity with d j ; it has proper-
ties akin to A's bj 2 anc* t o A's d j . Let u s term it b'j/d 1 .
An' I| think, B continues, introducing an experience of his own in his
old school, I| just caught/I^ caught the tail end of uh—of the—really old
school. Such a structure mirrors A's clauses with personal deictic Sub-
ject in cj 2» anc* n e n c e can be termed c ' j . If we termed Ij think a'|, An'
I: think would be An' + a ' j , or a " j . Thus, the sequence so far in (6) is a ' j -
yd'V'
At this point B demonstrates an interesting disfluency, going into an

'it' statement, then going into a 'Ij' statement, then attempting to repair
with the clarifying frame 1^ mean —, structurally equivalent here to L
think of a'j. B continues to clarify the time interval separating then and
now with another personal statement constituting his then experience of
having caught the tail end of the old school. He says, 'cause f| I| did my
undergraduate work Ij finished that like four years a--by comparison
with b'j/d', we see that this falls short of the interval that B had himself
set up there for the process of 'change', and there is a long, perhaps
computational pause--...go five years ago—the numbers now being per-
fectly to scale.
This a'2~c'2 sequence is then followed by And—1^ think, parallel to the
second sequence-opener he used, and hence a"2, and then by a complex
'it' statement about Loyola University, B's it/there, bringing the whole
structure back to explicit use of the predicate be different of A's dj in
(5). But here, the predicate glosses change—as it turns out, for the
better—rather than glossing overwhelming, as in A's earlier segment.
This b'2/d' unit is itself highly complex in internal parallelistic form,
illustrating the fact that the poetic subsequences are hierarchically
embedded in other, superordinate ones.
N6w—probably Loyola is a 16t different, starts out the segment, with
an undifferentiated b^/d' parallelistic form. But it continues with a
parallel to the last, rhematic element, an1 a lot better, explicitly clinch-
ing the equivalence of [now] different jjsc, through change] = better
Qhan then]. After a long pause, B adds a parenthetical to introduce
further explication, y': know (compare A's own parenthetical in (5) e, h
must say), followed by another parallel unit. A 16t more—variety 01
courses being offered et cetera, continuing in the comparative form,
gives the specific sense of 'different' = 'better'.
With the and—, B continues with a perfectly parallel stressed conjunc-
tion, such as that linking the lot different-lot better-lot more units. But
this is not quite correct from a discourse structure point of view, and
after a hesitation, B frames what he has said so far with another paren-
thetical, you: know, which demonstrates the essence of the kind of
composition such interaction consists of. For this now serves not so
much as a parenthetical, an afterword-boundary for something just
uttered, as a prospective framing device for what follows. B fashions in
this way what we can regard as a formal seam, leading on to the next
section, in which he will speak of his dissatisfaction with the Loyola of
there-and-then, in pointed contrast to the idyllic memory of A's George-
town, his there-and-then. From the point of view of B's discourse so far,
A d k know hhere llooks
And—you: k liklike a ffailed
i l d unit
it off form
f L^nd)c
L^d) Qnj [[a lot
lt
[jComparJ^j-C]* you. kn<5w. But from the point of view of how it figures
in introducing further discourse, it is precisely equivalent to the various
a' and a" structures, with the substitution of you: know for 1^ think.
Observe how the parenthetical y': know contrasts with the presentational
y6u: know, just as the unstressed an' of the comparative phrases con-
trasts with the stressed and like those joining major sequence seg-
ments. Hence, the very parallelistic structure internal to b^/d', slightly
violated, becomes the seam to the next major structure B will develop.
Having distinguished a view of then and now in which it/there gets
better, B will go on to distance himself from there-then, i.e. Loyola in
terms of dissatisfaction, confirming for us the underlying tension sur-

rounding the comparison of Loyola to Georgetown developed in the
interactional structure.
In (5), then, A has introduced a very tight structure of apb^-c^-dj-e;
ao-Co-^-do. In (6), B responds with an equally tight structure, of the
form a ' l - b y d ' - a ' ^ - c ' i ; a' 2 -c' 2 -a"2-b' 2 /d'. The last unit, b' 2 /d f , is itself
internally complex, consisting in its final, rhematic section of d"-b"^-
e'j-b'^/e^j where d" is Loyola is a lot different, the b" units are the
equivalent comparative evaluations (cf. the bi 2 units of A's disquisi-
tion), and the e1 units are the parentheticals, the last poetically ambig-
uous in metric structure. B will then go on to emphasize the contrasts
to A's Cj 2 units in his next section.
We must emphasize once again the distinctiveness of the organization
of the linguistic material we have analyzed. In this, as in any dyadic
conversation, we have discovered a poetic organization that has its
seemingly adirectional, atemporal, 'structural' aspect—once we describe
it—and also its seemingly directional, temporal, 'functional' aspect—
which we as well as the interlocutors seem to understand. The atempor-
al or 'structural' looks hierarchical, almost constituency-like; the tem-
poral or 'functional' looks transformative, accomplishing the semiotic
work. But the structure identified is transcendent in exactly the same
sense as syntax, though as native speakers in a certain tradition we may
not have any intuitions about it in the same fashion as we do for the
latter. And the function, about which we have many intuitions, seems to
involve interpersonal social acts, accomplished in some realm of beliefs,
intents, thoughts, rather than being merely a time-bound formal struc-
ture, in which we are participants, but to which we are not really par-
ties.
One's 'heroes'—and 'villains'—may be a matter of unexamined, and
ultimately irrelevant, perspectival assumptions.
Notes
I am grateful to Jerrold Sadock and 3ames McCawley for comments

on oral presentation of an earlier draft of this paper, and to Anthony
Woodbury for extensive commentary on its earlier written form.
1. These recent debates with syntactic formalism have been uncon-
vincing, I think, at least partly because they have not been paying atten-
tion simultaneously to all three areas of assumptions surrounding the
inviolability of the sentence-scope boundary for true grammar. For
example, one of the concerns of work professing a 'text grammatical'
and similarly global text-orientation to larger-than-sentence scopes of
language (cf. van Dijk 1977; Givon 1979), has been to extend the analysis
characteristic of sentence-scope syntax to stretches 'higher', as it were,
in a constructed hierarchy of forms (cf. also earlier tagmemic work). As
orthodox syntacticians have been quick to point out, the coherence, or
putative well-formedness, of stretches of language 'above' the sentence
may well not be formulable in strictly comparable autonomous, and
abstract terms. Rather, such entities as the topical period, the para-
graph, and the whole text seem to be content units—from the perspec-
tive of referential-and-predicational analysis of language—consisting of
causal or temporal event sequences, narrative scenes and plots, etc. The
relationship of these content units to formal abstract sentence struc-

ture, or even to the propositional schemata associated with sentence-
scope chunks of language, is quite problematic (see Morgan and Sellner
1980).
Similarly incomplete have been attempts at various types of psycho-
logical or communicational reductive functionalism—demonstrating that
within-sentence formal units are the local exemplars of speaker-oriented
and/or hearer-oriented human information-processing strategies, dis-
ambiguation requirements, etc. The long known alignment of
theme/rheme, old/new information-coding, and topic/comment of the
surface sentence-scope, a chunk of discourse, with the Subject/Predicate
distinction of abstract sentence structure, is an intriguing unmarked
situation in many languages, to be sure. But, as orthodox grammarians
have pointed out (see Morgan 1982; Newmeyer 1983), teleological, actor-
centered, one-factor explanations of these and other language-particular
formal facts or universal statistical tendencies do not (some would argue
cannot in principle) constitute reduction of syntactic categories and
regularities in the normal sense. We might note, then, correspondence
of grammar-determined portions of surface sentence-scopes with func-
tion-driven distinctions, both 'localizable1, as it were, in the same seg-
ments. But such facts need something more like multivariate statistical
models of competing functional requirements, given structure in the
normal sense, than alternatives to it, the argument goes, playing upon
the residual 'arbitrariness' of much of sentence-syntax as a still func-
tion-independent or unmotivated realm of structural regularity in lan-
guages.
But the house of autonomous formal sentence-propositional grammar
is hardly inwardly calm and ordered. It is not difficult to understand in
the terms of these debates the recent history of orthodox syntax. On
the one hand, portions of sentence-scope facts have been ceded to
functionalists and text-theoreticians, while on the other, ever more
abstract autonomous principles are being formulated that redefine, in
effect, what is still contained in grammar in the orthodox sense. Never-
theless, no real theory has been forthcoming of just how the multiplicity
of now-recognized factors interact, making this, mutatis mutandis, as
empty a 'modularized' structural-functionalism (as opposed to a reduc-
tive one) as one might imagine.
I doubt that real clarity will emerge until all sides of the discussion
recognize the full set of dimensions of the problem of isolating sentence
syntax, and become more honestly semiotic (see Silverstein 1984 for
some proposals).
2. The computer-generated transcript available to me includes, in
addition to segmental linguistic forms, for each participant, transcrip-
tion of head and hand kinesics, and paralinguistic features such as
smiles, laughs, drawl, etc., all displayed against a time measure in
milliseconds, for synchronization. The data are used with the permission
of Starkey Duncan, to whom I am grateful for encouragement.
3. To say that full sentences are, or can potentially be, 'understood'
from what then emerge as sentence-fragments, is only to say that sen-
tence-scopes with certain grammatically specifiable forms are recon-
structible 'off line1, as it were, by a kind of secondary reflection upon
the occurring material (with contextual cues), plus various default
prescriptions based on hypotheses about the way sentence-scopes code

propositionality. This reconstructive procedure gives the illusion that
decontextualization is equivalent to abstraction (in the syntactician's
canonical sense). From this confusion, sentences can then be viewed as
pure, complete forms of ideal theory, while what from this perspective
appear as sentence-scope fragments are merely contextualized reduc-
tions. Decontextualization, abstraction, and idealization are not simple
equivalents; this is not what is intended in sentence-syntax, nor here,
though mistakenly identifying them has generated much confusion in the
literature about ideal speaker/hearers and the like.
4. We may add that there seems to be something of an asymmetric
and manipulative quality to A's interactional style, culminating at this
point. Were we interpreting the interaction, rather than giving a formal
analysis based solely upon syntactic form in discourse context, we would
see this as a game of One Upmanship. A, the Law School student, has
led B, the lower-status Social Services student, to reveal his background
in some detail. He may well be responding in an aggressive manner to
the noticeable hesitancy with which B, the student at the academically
and professionally lower-ranked division of The University of Chicago,
has at first avoided as long as possible taking the 'where-did-you-come-
from-before' line of questioning as directed to schooling. At last, having
been worked into this corner, he very hesitantly reveals that his under-
graduate school was not indeed The University of Chicago (one continu-
ing possibility for the reference of here), but Loyola University of Chi-
cago. But A reveals first off only that he, too, has gone to a Jesuit
institution (as an undergraduate, presumably, else the formal parallelism
established in the interaction is a misfire). His mock humor here has a
slightly aggressive connotation, as he sets B up to ask him which Jesuit
institution. A will now be able to increase his situation-relevant status
even more, when he reveals his undergraduate institution.
5. Compare Irvine's (1974) discussion of Wolof (Senegal) dialogic
greetings, in which the roles of initiator and respondent have status and
caste connotations, and entail certain further relations of bestowal
action. One of the strategic moves of greeters is to shift their dialogic
role from the one to the other by certain interesting formulaic repeti-
tions, hesitations, and readjustments.
References
Austin, J. L. D963 1975. How to do things with words. 2nd edition.
Cambridge, Mass.: Harvard University Press.
Bricker, Victoria R. 1974. The ethnographic context of some traditional
Mayan speech genres. In: Explorations in the ethnography of speak-
ing. Edited by Richard Bauman and Joel Sherzer. Cambridge: Cam-
bridge University Press. 368-88.
Danes', Frantis*ek. 1970. One instance of Prague School methodology:
Functional analysis of utterance and text. In: Method and theory in
linguistics. Edited by Paul L. Garvin. Janua linguarum, Series maior,
40. The Hague-Paris: Mouton. 132-46.
Dijk, Teun A. van. 1977. Text and context: Explorations in the seman-
tics and pragmatics of discourse. Longmans Linguistics Library, 21.
London: Longmans.
Duncan, Starkey, and Donald Fiske. 1977. Face-to-face interaction.
Hillsdale, N.J.: Lawrence Erlbaum Associates.
Firbas, Jan. 1964. On defining the theme in functional sentence anal-
ysis. Travaux linguistiques de Prague, 1. L'Ecole de Prague
d'aujourd'hui. Prague: Editions de l'Acade"mie Tche"coslovaque des
Sciences. 267-80.
Fox, James J. 1977. Roman Jakobson and the comparative study of
parallelism. In: Roman Jakobson, Echoes of his scholarship. Edited
by Daniel Armstrong and C. H. Van Schooneveld. Lisse: Peter de
Ridder Press. 59-90.
Giv6n, Talmy, ed. 1979. Discourse and syntax. Syntax and semantics,
12. New York: Academic Press.
Halliday, Michael A. K., and Ruqaiya Hasan. 1976. Cohesion in Eng-
lish. English Language Series, 9. London: Longmans.
Irvine, Judith T. 1974. Strategies of status manipulation in the Wolof
greeting. In: Explorations in the ethnography of speaking. Edited by
Richard Bauman and Joel Sherzer. Cambridge: Cambridge University
Press. 167-91.
Jakobson, Roman. 1960. Closing statement: Linguistics and poetics.
In: Style in language. Edited by Thomas A. Sebeok. Cambridge,
Mass.: MIT Press. 350-77.
Jakobson, Roman. 1966. Grammatical parallelism and its Russian
facet. Lg. 42.399-429.
Morgan, Jerry L. 1982. Discourse theory and the independence of
sentence grammar. In: Georgetown University Round Table on
Languages and Linguistics 1981. Edited by Deborah Tannen. Wash-
ington, D.C.: Georgetown University Press. 196-204.
Morgan, Jerry L., and Manfred B. Sellner. 1980. Discourse and lin-
guistic theory. In: Theoretical issues in reading comprehension.
Edited by Rand J. Spiro, Bertram C. Bruce, and William E. Brewer.
Hillsdale, N.J.: Lawrence Erlbaum Associates. 165-200.
Newmeyer, Frederick J. 1983. Grammatical theory: Its limits and its
possibilities. Chicago: University of Chicago Press.
Saussure, Ferdinand de. Q91(0 1960. Cours de linguistique geneYale.
5th ed. Paris: Payot.
Searle, John R. 1969. Speech acts, An essay in the philosophy of lan-
guage. Cambridge: Cambridge University Press.
Searle, John R. 1975. Indirect speech acts. In: Speech acts. Edited by
Peter Cole and Jerry L. Morgan. Syntax and semantics, 3. New
Silverstein, Michael. 1984. The functional stratification of language
and ontogenesis. In: Culture, communication, and cognition: Vy-
gotskian perspectives. Edited by James V. Wertsch. New York:
Cambridge University Press. 207-37.
THE POLITICS OF POLITENESS: SOCIAL WARRANTS
IN MAINSTREAM AMERICAN PUBLIC ETIQUETTE
Thomas Kochman
University of Illinois at Chicago
This paper begins with two propositions: first, there is a distinct
political character to mainstream American public protocols; and sec-
ond, this political character is often lost on mainstream Americans who
tend to see the standards they set for social interaction as impartial,
that is, serving the political interests of all participants equally. I want
to deal briefly with the second proposition before turning more attention
to the first.
Mainstream Americans do not see themselves as being politically
shrewd, or disingenuous, when they ask members of minority groups who
have felt themselves aggrieved to stop being angry as a prerequisite to
negotiations aimed at discussing the nature and/or resolution of those
grievances. Rather, mainstreamers simply see themselves as invoking
what society has established as conventional procedures for handling
disputations and disagreements. Such procedures require that individual
negotiators on all sides remain calm and 'rational' (i.e. unemotional) in
presenting their views. They also require that negotiators be
predisposed to acknowledge at the outset that the parties to the
negotiation may not be altogether correct in their respective assess-
ments of the situation. Furthermore, mainstream Americans also be-
lieve that (1) there are multiple sides to an issue, (2) no side has a mo-
nopoly on the truth, (3) the more firmly one side believes that it is right,
the less likely it is to be flexible enough (to be predisposed) to acknowl-
edge the 'truth' in the other side's position, and/or to agree to accept
compromise as the proper resolution of differences. Moreover, to the
extent that one side or the other is angry or otherwise emotional, its
perspective on matters is also likely to be commensurately distorted.
All of these views are applied a priori to mainstream social inter-
actions, such as negotiating sessions, rather than being empirically
tested for their accuracy. Thus, mainstreamers do not actually consider
whether the thinking of opposing parties really does become fuzzy or
incoherent when they behave emotionally, or even—in a relative sense
200
The politics of politeness / 201
for any one person—more fuzzy or incoherent. Nor do they actually
examine qualitatively the 'truths' being advocated by opposing sides to
consider whether one side may well have a greater claim on reason,
human feeling, and/or justice than the other (Marcuse 1965:85 calls this
nonpartisan stance 'abstract' or 'pure' tolerance.) And because these
mainstream views are not tested empirically, they can hardly be said to
be 'scientific 1 , but rather must be classified as part of mainstream
American cultural folklore or ideology. Thus, we find folk sayings which
encode such views: 'People are less rational when they are emotional',
'Compromise, or "agreeing to disagree", is the proper way to deal with
differences that cannot be resolved in any other way.'
Yet, upon closer examination, one might see that mainstream Ameri-
can protocols might not, in fact, be all that impartial. Rather, as with
most established systems, such protocols can be seen as serving the
interests of those who have installed them, even if the individuals whose
roles are regulated and behavior defined within and through the system
do not see the self-interest within that system. After all, participants
need not be aware of their complicity within a system for the system to
have its effect. Indeed, it may be better both for them and for the
system if they are not. For how can they maintain a sense of themselves
acting in 'good faith' if they know that by asking aggrieved parties to
stop being angry as a prerequisite to negotiation, they may well, in
effect, be requesting that the aggrieved parties weaken their bargaining
stance, thus preparing them to accept less than they might otherwise be
willing to accept if they were to remain angry?
Moreover, one might see a political strategy in mainstreamers asking
aggrieved parties to assume a posture of nonanger before having been
given a substantive reason for doing so (such as a change in the condi-
tions that produced the anger). The assumption of such a posture would
allow mainstreamers to gain a pacified adversary as a prerequisite to
negotiation, although from their own perspective of self-interest, this
might well have been regarded as the desired consequence of negotia-
tion. Mainstreamers1 requests for pacification as a prerequisite to
negotiation, however, free them from having to make any substantive
concessions to get it. For example, some years ago a White House
official said, with regard to the American Indian Movement's takeover of
Wounded Knee, 'The White House will not negotiate while guns are
pointed at federal officers at Wounded Knee' (Christian Science Monitor,
April 9, 1973). One might well wonder why the federal government
would have any further need to negotiate if they were able to get the
American Indians to surrender their arms as a prerequisite to negotia-
tion, since that result may well have been the motivation for the govern-
ment's agreement to negotiate with the Indians to begin with.
Or, one might also see a political motivation behind the mainstream
promotion of compromise, or 'agreeing to disagree1, as the proper way to
deal with 'differences'. For do not 'compromise' and 'agreeing to dis-
agree' ultimately establish a priority of peace over conviction, truce
before 'truth'? And is not 'truce' much easier for established societies to
manage than the disorder that is often produced from commitment by
political adversaries to irreconcilable 'truths'?
In any event, I am less concerned for the moment about mainstream-
ers' awareness (or lack of awareness) of the political character of their
202 / Thomas Koch man
public protocols than I am about the actual political nature of those

protocols. Consequently, what I would like to do here is to develop and
document further the proposition that mainstream American protocols
have this political bent, above and beyond what I have already said. I am
going to do this by examining the notion of social warrants within the
framework of mainstream American public etiquette. This is an espe-
cially useful way of revealing both the cultural and social norms of
mainstream American communication (or for that matter, of any socio-
cultural group), and their distinctive political character.
Social warrants. The term 'social warrants' is especially apt in con-

noting the entitlements to which individuals can lay claim in social
interaction, since these entitlements bear directly upon the degree of
leverage that individuals can hope to muster in establishing parity for
their point of view, or, more generally still, in influencing the course and
outcome of encounters and transactions in which they are participants.
These entitlements themselves have a social basis—consequently, the
term 'social' warrant. However, when people typically consider the
social basis for such entitlements, they look to ascribed characteristics
of age, sex, and family position, both apart from, and in conjunction
with, achieved characteristics, such as rank, title, or social position.
Rarely do people see social warrants or entitlements in terms of the
cultural norms that society has set to regulate social interaction. Yet,
social warrants do have a cultural basis. Although culturally defined,
etiquette systems exist ultimately to govern social interaction. For
example, they do more than define parameters for proper and improper
conduct: they also contain within them the means to command conform-
ity with these norms. In socially symmetrical public transactions, these
means are in the form of social warrants; in the context of social inter-
action, such warrants are the entitlements that enable individuals to
claim respectful consideration for themselves from others with whom
they are interacting.
Unlike entitlements that derive from an individual's social position,
however, these culturally based entitlements are conditional upon situ-
ated performance, and the very pattern of this performance serves to
maintain the standards that society has set for social interaction. And
unlike other social warrants, which individuals can be said to own per-
manently, these warrants accrue to individuals only so long as they
deserve them—and they continue to do so only as long as they behave in
ways that society has established as appropriate. As long as people
deserve and receive each other's respectful consideration, their sense of
themselves as bearers of social warrants tends to recede. (This may be
yet another reason why mainstream Americans tend to be unaware of
the political character of their public social interaction.) But should
individuals forfeit respectful consideration for themselves, it is likely
that they will become immediately aware of their loss of political lever-
age, because this loss will be translated interactionally into their lost
capacity to influence further the course and outcome of the encounter
or transaction which engages them. Thus, it is in the self-interest of
those who would wish to maintain their political leverage not to forfeit
the social warrant that grants them that leverage. And it is in protec-
tion of this warrant that people often work to behave in ways that
conform to society's prescriptions of how to act.
The political character of mainstream American public etiquette is
expressed not only in the form of the leverage that it generates in any
given transaction. This political character is also expressed, in a larger
social sense, in the kind of behavior that the norms wish ultimately to
produce. Thus, there is a political motivation behind those mainstream
norms that discredit presentations that are argumentative, confronta-
tional, and emotional; presentations that reflect or display the degree to
which an individual believes in what s/he says; or presentations that
might rely upon force of individual personality or verbal skill in order to
succeed. Such presentations are defined as either irrelevant to, or
threatening to undermine, the social basis upon which decisions are
reached in mainstream public social life. The reason such presentations
are so defined is that one concern of mainstream public life is the main-
tenance of a social order in which power is by nature administrative,
invested in office, rank, and social position, or, to borrow Hacker's
(1979:52) phrases, 'corporate' rather than 'corporeal1. With respect to
this corporate nature of power, individuals are socialized to act in ways
that would effectively enlist their cooperation in the official exercise of
administrative power.
Mainstream American social etiquette also acts as a principal social-
izing agent: it promotes mainstream American communication and dis-
credits that of other social or cultural groups (e.g. the lower classes or
ethnic minorities), at least insofar as their patterns of behavior are
incompatible with the way power is promoted and exercised in main-
stream social life. Consequently, I would argue that mainstreaming in a
cultural sense, that is, as part of Americanization, is first and foremost
a process of political socialization, concerned with behavior that relates
to people's willingness or ability to cooperate with the official exercise
of established administrative power, and with other matters that direct-
ly or indirectly relate to that power. Principal among such forms of
behavior are those concerned with individual allegiance, especially
divided allegiance and matters that deal with the priorities that indivi-
duals have set as to what, and whose, interests they shall serve, and with
what measure of zeal and devotion.
I do not have time to develop all of these matters here. Indeed, I have
yet to establish adequately my first proposition that along with its rules
of conduct, mainstream American public etiquette generates a strong
political bent. Consequently, I would like to show this by looking more
closely at the specific etiquette governing the mainstream American
notion of 'showing consideration for the feelings of others', especially as
this notion is realized in mainstream public occasions and/or trans-
actions. I am going to do this by considering the cultural significance of
such expressions and terms as 'showing consideration', 'feelings', and
'others' to show both the distinctive character of the mainstream Amer-
ican cultural pattern, and its fit and function within the larger main-
stream public contexts.
Showing consideration. Public consideration is typically shown nega-
tively by avoiding actions or statements that others might find offen-
sive. This is opposed to more positive demonstrations of consideration,
e.g. actively doing things that might give another pleasure or satis-
faction, that one might show in more private personal contexts. But in
204 / Thomas Kochman
public, consideration is shown by keeping one's behavior at the non-

arousal level: not only by not doing or saying things that might prove
offensive, in terms of content, but also by not providing a stimulus so
powerful that it might arouse others beyond the level that their defenses
can comfortably manage. As Slater said (1976:115) about mainstream
American public occasions generally: 'Clothes must be drab and incon-
spicuous, colors of low intensity, smells nonexistent...sounds should be
quiet, words should lack affect. 1
Thus, polite conversation avoids issues or topics about which indivi-
duals are likely to have strong feelings: traditionally, sex, religion, and
politics. Consequently, polite conversation is also a way of showing
consideration for other people's feelings, that is, not saying or doing
anything that might unduly excite or arouse. The 'gentleman's agree-
ment' (though hardly confined to adult males) is and was, 'You don't do or
say anything that might arouse my feelings, and I won't do or say any-
thing that might arouse yours.' 'Considerate' topics are therefore safe
(nonarousing) topics. Like Trevor Pateman's notion of 'idle discourse',
polite conversation is static rather than dynamic: 'Once begun it aims
only to stay where it is' (Pateman 1975:40). Ultimately and essentially,
then, mainstream consideration is a form of protection, not really of
feelings, but rather of sensibilities, as will be argued next.
Feelings. I want to consider now the term 'feelings' as part of the

mainstream American notion of 'showing consideration for the feelings
of others'. This term is somewhat of a misnomer if one tends to equate
feelings with emotions, since emotions are precisely what mainstream
Americans do not show consideration for in everyday public social inter-
action. On the contrary, the public expression of emotions such as
anger, joy, and sorrow is supposed to be muted, if they are to be ex-
pressed publicly at all. Again, I think this is primarily because of their
potential to arouse and involve others, and in so doing, actively to
charge and heat up public occasions and events. Emotions in this regard
are double edged and, consequently, also doubly frowned upon, connoting
the existence of an already aroused state, but in that state, also posses-
sing a capacity to arouse others. Thus, a person often feels compelled to
apologize to others when an emotion (such as grief or joy) erupts beyond
his or her ability to contain it. And the reason one feels so compelled is
that social etiquette considers such a public expression of emotions to be
an imposition on other people's consideration, where even among known
others—as at the workplace, for example—their expression often needs
to be justified. ('Other people's consideration1 is here being limited to
avoiding undue arousal, but not to showing commiseration or other
empathetic involvement.) In fact, showing strong emotions in public
may well forfeit for those individuals the social warrant that entitles
them to receive respectful consideration for their feelings—their feel-
ings here obviously not being charged emotions. But not only are indi-
viduals not entitled to receive respectful consideration from others when
they express strong emotions; they actually risk being accused of being
inconsiderate of other people's feelings. The result is that they then risk
forfeiting the consideration of others for their own feelings.
But if not emotions, what are the feelings for which mainstream
Americans show consideration? I would argue that these feelings are
inactive or inert states, which is to say, unaroused states. And again,

public consideration is shown for these feelings by not activating them in
any potent way. And what are these feelings qualitatively? That is hard
to say. Being inactive and uncharged, they are also undifferentiated and
difficult to identify. Because of these qualities, I prefer to call them
'sensibilities1; I can then reserve the term 'feelings' for those that are
more actively charged, that is, emotions (even though I recognize that
the term 'sensibilities' may also connote more than just a state of in-
active, undifferentiated feelings). I also believe that the term 'sens-
ibilities' characterizes the internal state accompanying the interactional
role that mainstreamers occupy when they are deemed most deserving of
respectful consideration from others. This last point becomes clearer in
the next section, where I discuss the mainstream cultural meaning of the
term 'others'.
Others. Upon hearing for the first time the expression 'showing con-
sideration for the feelings of others', one might be likely to construe the
word 'others' as referring to another person, that is, someone other than
oneself. In this regard, individuals are often taught to consider them-
selves 'selfish' if they 'place their own feelings before those of others. 1
But a closer examination of this expression reveals an apparent contra-
diction in that interpretation. For are not we ourselves also 'others' in
other people's consideration, just as they are 'others' in ours?
And what about behavior that we consider socially unacceptable, as in
the case of a young man playing a radio loudly on a bus? Are we not
then also the 'others' who should be receiving respectful consideration,
even as we are also 'ourselves'? And if it is for 'ourselves' that we
request that the radio be turned off, are we not then also being 'selfish'?
Objectively speaking, yes, but culturally speaking, no. We are not being
'selfish' because social etiquette has determined that 'we' are the 'others'
in that situation, not the young man who played the radio loudly. Con-
sequently, this example shows that oneself can also be an 'other' and
moreover, as an 'other', can insist upon receiving respectful considera-
tion for oneself, even preemptively, without being considered 'selfish1.
But what does this do to our first interpretation, in which the term
•others' referred to other people, not to oneself, and which defined as
'selfish', the placing of one's own feelings before those of another?
Clearly, that interpretation needs to be qualified. For as we have dis-
covered, the term 'other' (or for that matter, 'self') does not refer social-
ly to a person per se, that is, a 'who', but in reality to a 'what': a social
category or status, for which individuals must qualify, which they must
continue to deserve, and for which they can also become disqualified, or
can forfeit, by behaving improperly.
But that, of course, does not explain why mainstream etiquette con-
sidered 'us', the nonradio-playing bus passengers to be, preemptively, the
'others' in that situation, or why the behavior of the radio's owner dis-
qualified him from that category. This explanation must come from a
closer look at the process of accommodation that is expected within
mainstream public etiquette. According to such etiquette, individuals in
their interactional role as assertors are expected to mute or otherwise
moderate their level of self-assertion to a point which individuals as
receivers can tolerate. That is, individuals as assertors are socially
206 / Thomas Koch man
accountable in this regard to individuals as receivers, but not vice

versa. A failure by assertors to acknowledge such accountability by
modifying their behavior accordingly makes them 'selfish'. Thus, it was
the interactional role that each party (assertor or receiver) played on
the bus that allowed us to make the determination that, even though
both parties were promoting their own self-interest, only the young man
was being 'selfish'.
This relationship between interactional role and the attribution of
selfishness exists for the following reasons. While receivers always
qualify as 'others' in mainstream American public etiquette, assertors
continue to qualify as 'others' only if the level of their self-assertion
conforms to public norms respecting 'good taste'. Recall that in main-
stream public etiquette, 'good taste' invariably requires the application
of individual self-restraint, since powerfully arousing presentations are
generally considered to be in 'bad taste'. And individuals who do not
show adequate self-restraint in their assertiveness risk forfeiting consid-
eration for their sensibilities. Once that happens, those individuals'
sensibilities are regarded as no longer deserving public protection. Thus,
an adult passenger on the bus, in brusque and hostile fashion, may order
the young man to 'turn his damn radio off. Note that this other passen-
ger can do this without being considered 'rude', since one can only be
'rude' to people who still retain the status of 'other'. But, culturally
speaking, one cannot be 'rude' to those who have forfeited the social
warrant that grants them the right to claim respectful consideration
from others. Note, however, that a mainstream sense that 'the punish-
ment should fit the crime' may still be operative; but because this last
notion falls within the mainstream American sociocultural realm of 'fair
play', not 'politeness' per se, I will leave it for another time.
It seems that one can almost claim that the role of 'other' is always
that of receiver in mainstream social interaction. Thus, even when
assertors have not disqualified themselves as 'others', as when they keep
their self-assertiveness within proper bounds, their sensibilities still
deserve protection. But the social frame which shows that assertors
have not forfeited consideration for themselves—that they retain their
respective statuses as 'others'—requires that they be placed on the
receiving end of an interaction, for it is essentially as receivers that we
demonstrate the concept of 'consideration' within the mainstream notion
of 'showing consideration for the feelings of others'. And because it is
essentially only as receivers that individuals can and do receive consid-
eration for their feelings, it again seems more appropriate to call the
internal state that is being protected by such 'consideration', 'sensibili-
ties' rather than 'feelings'. To show consideration for feelings, that is,
emotions, others would also have to show consideration for the role of
assertor, since it is this role that people play when they are expressing
their feelings.
As I have stated, mainstream culture does not show such considera-
tion. Black culture, by way of contrast, does. Consequently, black
social etiquette is obliged to promote a process of accommodation
whereby receivers also accommodate assertors, for example, by granting
them culturally sanctioned outlets for an unrestrictive expression of
those feelings. In short, the kinds of feelings that we show are directly
tied to the kind of role that we assume in social interaction. And the
kinds of feelings that receive 'consideration' are directly tied to the
status of that role within the culture in which those feelings are dis-
played.
A further point needs to be made to show that within mainstream
etiquette, there is a sociocultural bias, or predisposition, in favor of low-
keyed assertive public behavior. Note that individuals forfeit considera-
tion for themselves as 'others' only when their behavior becomes more
intense than cultural norms allow, that is, when they are being irrespon-
sibly self-assertive. But what happens when individuals are being ir-
responsibly nonassertive, as when they refuse to participate in a group
discussion, notwithstanding the need to hear from everyone? Under
those circumstances, individuals do not forfeit consideration for them-
selves as 'others'; they are still entitled to receive respectful considera-
tion for their sensibilities, because according to the prevailing rule
governing social etiquette, they have a 'right' not to speak if they do not
want to do so. Thus, those who might otherwise be able to influence a
person's behavior in a situation where he or she is behaving irresponsibly,
might find that they only have a preemptive social warrant vis-a-vis that
individual only when his or her behavior is in the direction of being
irresponsibly active, but not when it is being irresponsibly passive. In
short, we can order a person to 'keep quiet', but not to 'talk'. Main-
stream American society grants social warrants to others only for the
purpose of generating a quiet public climate, but not to induce more
forceful self-assertive behavior, even when such would clearly be in the
interests of producing more responsibly active social participation.
The role and function of individual sensibilities. Individual sensibili-
ties have a vital role in maintaining social standards for public behavior,
for socialization works only to the extent that individuals internalize
social norms as to what constitutes proper and improper behavior. In-
deed, this shared sense of what is, and what is not, appropriate is what is
conveyed by individual sensibilities. And, as already pointed out, sensi-
bilities can function as regulators and custodians of public order in social
interaction only if the role in which sensibilities are brought to bear-
namely, the receiver role—is granted a preemptive social claim for
consideration in public social interaction.
However, simply granting preemptive status to the receiver role in
order to enable sensibilities to exercise their custodial function does not,
in and of itself, ensure the production of only low-keyed presentations.
It is also necessary to keep sensibilities relatively sensitive and/or
susceptible. This is where the reciprocal nature of the process of ac-
commodation, discussed earlier, makes its contribution. Assertors'
reduction of the power of their self-assertion to the level that can be
withstood by their receivers, ensures—through the reciprocal nature of
accommodation—that the sensibilities of receivers will remain suscep-
tible, precisely because they are not being confronted with those more
forceful presentations to which, through repeated exposure, they might
well become inured. This, in turn, might well have the effect of making
them less personally sensitive and/or susceptible, and consequently, also
more socially tolerant of more forceful public behavior. Thus, just as
low defense produces low offense within an etiquette system that re-
quires that assertors moderate the level of their self-assertion to the
208 / Thomas Kochman
level that receivers can comfortably manage, so, too, does low offense
maintain low defense by moderating the level of assertive behavior only
to the level that receivers can already comfortably manage.
And so mainstream American etiquette accomplishes socially what
mainstream American society wishes to accomplish politically: public
behavior is kept sufficiently low-keyed so that established authority can
exercise its power with a minimum of resistance and/or risk, no doubt a
consummation devoutly to be wished by mainstreamers in societies all
over the world.
Notes
1. Of course, the ascribed or achieved qualities also have a bearing on
the extent to which individuals, in fact, do or do not forfeit considera-
tion for themselves, especially in socially asymmetrical situations where
one side has prerogatives that the other side does not, and where con-
sequently there may also be present a different accountability for one's
behavior. In this paper, so as to reveal the particular bias that only the
behavioral norms themselves generate (reflected in the way social
warrants are maintained or forfeited), I assume that prerogatives grow-
ing out of ascribed or achieved status are equal for all social inter-
actants, or, said another way, that the risk or accountability to the
norms regulating social interaction is the same for all participants.
2. In different cultures, status as an adult 'other' may itself have to
be earned, rather than being considered a 'given'. An American white
middle-class couple who lived in New Zealand for five years made an
interesting distinction between life there and in America which bears
upon this point. They said that in America you are presumed to be 'all
right1 until/unless you prove yourself otherwise. In New Zealand, you are
assumed not to be all right until you prove yourself otherwise. In the
American system (at least for white middle-class adults), respectful con-
sideration for oneself, or status as a qualified 'other', is given. One is
innocent until proven guilty. Thus, the American etiquette system can
be regarded as one in which others disqualify themselves. In the New
Zealand system (and probably the English one), respectful consideration
for oneself must be earned. There, one is presumed guilty and must
prove oneself innocent. That etiquette system may well be regarded as
a more severe system of qualification for 'other' status, and, consequent-
ly, as one in which some individuals, in the eyes of others, may never get
to be quite 'all right1, a situation similar to that of children and members
of minority groups in the United States.
3. Receivers, of course, may accommodate forceful self-assertive-
ness, but they are not required by mainstream public etiquette to do so.
Consequently, that they should do so is seen as a matter of individual
choice. And, because receivers are not required to accommodate im-
moderate assertor behavior, only receivers can get public credit for
being 'tolerant1, tolerance here being understood as an individual's will-
ingness to endure behavior that society would consider to be in bad
taste. But assertors would not be seen as being especially tolerant of
receivers who, for example, were not being properly attentive to what
the assertors were saying. Nor could assertors hold receivers to be
socially accountable for that behavior since, even though it falls within
the general category of social etiquette, it makes no impact on public
order. Consequently, such a matter of etiquette would not become an
issue for the courts. But publicly cursing a policeman would—and, in
fact, did, in the case of Lewis v. City of New Orleans (40 U.S. Law Week
3614), cited in Haiman (1972a). (See also Haiman 1972b.)
4. This should not imply that the status of the feelings being 'con-
sidered' derives from the status of the role. Rather, it may be the
opposite: the status of the role derives from the kinds of expression that
people manifest while in that role.
References
Hacker, Andrew. 1979. Two 'new classes' or none? Society 16.2:49-54.
Haiman, Franklyn. 1972a. The fighting words doctrine: From Chaplin-
sky to Brown. Iowa Journal of Speech 3.1:3-31.
Haiman, Franklyn. 1972b. Speech v. privacy: Is there a right not to be
spoken to? Northwestern University Law Review 67.153-99.
Kochman, Thomas. 1981. Black and white styles in conflict. Chicago:
University of Chicago Press.
Marcuse, Herbert. 1965. Repressive tolerance. In: A critique of pure
tolerance. Edited by Robert Paul Wolff, Barrington Moore, Jr., and
Herbert Marcuse. Boston: Beacon. 81-123.
Pateman, Trevor. 1975. Language, truth and politics. Nottingham:
Stroud and Pateman.
Slater, Philip. 1976. The pursuit of loneliness. Revised edition. Bos-
ton: Beacon.
TALK AND ITS OCCASION: THE CASE OF CALLING THE POLICE
Don H. Zimmerman
University of California, Santa Barbara
0. Introduction. Erving Goffman has proposed (1974:500) that 'utter-
ances—whether1 formal or informal—are anchored in the surrounding,
ongoing world . The degree to which they are anchored varies, to be
sure, for utterances inhabit and are responsive to a domain of face-to
face conduct—the 'interaction order' (Goffman -1983b)—and may be only
'loosely coupled' to some more encompassing scene. The point of Goff-
man's observation is nevertheless clear: whatever the relevant world,
talk is to some extent part of it, and this connection cannot be ignored
when we consider language as it is deployed in the conduct of both
everyday and extraordinary affairs in society.
Though it bears repeating, this insight is not exactly new. Linguists
concerned with semantic issues have come to recognize the limits of
reliance on semantic structures alone for the interpretation of utter-
ances. While they can be assigned a reading, utterances taken in isola-
tion do not provide sufficient clues to the meaning and use of similar
locutions fully situated in a live setting. Natural languages like English
cannot profitably be made to behave like formal languages, and thus
recourse is necessary to the circumstances and manner of utterance
production in order to settle—if so definite a term is possible—matters
of meaning and function. One must, in short, appeal to context—indeed,
to the social context, the occasion of speaking.
To speak of context is, of course, to speak of many things: the pre-
ceding or surrounding discourse—the 'co-text' (Brown and Yule 1983:46);
the situated identities of participants and their knowledge of the world,
particularly of the social world, brought to 'focal consciousness' (Goff-
man 1983b) by the play of those identities—the list is potentially lengthy
and the selection of contextual features for consideration somewhat
arbitrary. Here, those features are selected which appear to figure in
the organization of a particular instance of talk as a social activity, both
'internally' as discourse and 'externally' as action, the springs of which
originate in and ultimately rebound on a 'surrounding world'.
210
Talk and its occasion / 211
A small segment of the world is examined in this paper—citizen phone
calls to an emergency number for police or medical assistance. These
calls, developed over a sequentially ordered and organizationally struc-
tured series of utterances and turns, constitute a recurrent social occa-
sion, 'calling the police'.
Inspection of the calls reveals a definite underlying organization with
distinct segments, each of which performs specific functions. For
example, the opening of the call, like telephone call openings more
generally (cf. Schegloff 1979), provides the site and marshals the tools
for routinely addressing and resolving the 'identity' issue posed in part by
the lack of visual access in this mediated form of face-to-face contact.
The initial alignment of situated identities (e.g. 'citizen'-'police') thus
achieved projects a framework within which structurally relevant under-
standings of subsequent utterances can be achieved.
Other aspects of the organization of the call to be discussed here
include the complaint/remedy (or request/response) bracket—an adja-
cency pair organization in which the first and second pair parts are
routinely separated by a type of insertion sequence (Schegloff 1972),
termed here an 'interrogative series', which addresses response contin-
gencies relevant to the police (cf. Merritt 1977; Maynard and Wilson
1980). The calls also exhibit a closing sequence which, like the opening,
is closely fitted to a particular type of transaction—a request for ser-
vice—as well as reproducing the essential features of conversational
closings noted by Schegloff and Sacks (1974). The overall shape of the
calls appears to be quite comparable to that of calls to other types of
service-oriented organizations, suggesting that the issues and contingen-
cies involved in calling the police are general to a whole class of tele-
phone transactions (cf. Frankel 1977, 1981; Merritt 1977).
The contextual features important to an analysis of the production,
comprehension, and consequentially of these packages of talk point to a
world outside the discourse, indeed, a world brought into it by callers'
descriptions of distant events (cf. Whalen and Zimmerman forth-
coming). But description is language at work, and its workplace is,
among other things, an occasion brought into being, in part, by the talk
itself. This paper outlines the work talk does in fashioning the occasion
of calling the police.
Finally, the overall organization of the calls and their contextual
features described here, along with their observed similarity to 'service
calls' in general, is viewed as the contingent accomplishment of un-
noticed but nevertheless skilled work by callers and service personnel.
That is, the calls' evident orderliness is not the product of following a
known-in-common prespecified format or plan. Rather, the organiza-
tion emerges as the working through of issues—both internal to the
discourse and resident in the affairs to which the discourse is ad-
dressed—using culturally distributed procedures as ordinary and perva-
sive means of getting on with and getting through the call.
1. The setting. The 'Mid-City' police department is located in a
large metropolitan area in the midwestern United States. At the time of
the research, the department used a centralized, two-tiered computer-
based dispatching system, offering the public an emergency number to
call for reports of trouble, or requests for police or emergency medical
assistance.
212 / Don H. Zimmerman
Civilian employees, henceforth 'complaint-takers' (CT), answered calls
distributed via a rotary system and entered information derived from
their interaction with callers (C) into a computer terminal. While the
system provided prompts for the required information, it could be en-
tered in any order. The information acquired was then transmitted
electronically to dispatchers (sworn officers) who had the responsibility
of sending police units to the scene of the trouble. Ambulance and
paramedic assistance were sent by an ambulance dispatcher who was
contacted by the complaint-taker by phone.
The task of the complaint-taker was to collect and codify the infor-
mation necessary to dispatch the police or ambulance to the scene of a
reported incident. This generally involved ascertaining and coding the
nature of the reported problem, that is, its character as 'policeable
business1, e.g. a 'burglary', 'domestic', and so forth; its location, and
other pertinent information such as the presence of weapons or numbers
of persons involved in the incident. The demand for information of this
type in large part makes up the 'contingencies' of police response dis-
cussed later.
A small number of calls are transferred to some other office (e.g. to a
particular precinct at which crime reports are to be filed, or to some
division such as homicide or internal affairs). With few exceptions,
incoming calls are processed and forwarded electronically to the dis-
patcher for ultimate disposition.
The 125 calls examined in this paper represent a two-hour (11PM-
1AM) segment of tape from the first weekend of the month, which
complaint-taker folklore designates as busier than usual since it is
payday for many, with a consequent increase in drinking and drinking-
related incidents. There was also a rainstorm this night, which increased
the number of accidents and reports of (false) burglary alarms from
security organizations. It is worth noting that the volume of 'business'
on this evening was a feature attended to by at least one complaint-
taker, as revealed in a personal call also recorded on the tapes.
(1) [MCE 21-26:3<36

1 CT: Hello
2 C: Hello what ur ya doin'
3 CT: Oh god (.) busy as shit
k C: Is it still.
5 CT: Yes. Between storms, accidents (.) you name it it's goin'
on.
Thus, the discussion to follow deals with a collection of calls that may
most directly reflect the activities of late night, weekend hours, as well
as the contingencies of weather. Different troubles—or at least, dif-
ferent frequencies of occurrence for particular problems—may charac-
terize different time periods. The interests of this paper, however, are
in the achieved organization of the calls which, for reasons to be devel-
oped further on, is assumed to be largely independent of their specific
content, rendering the selectivity of the collection less problematic than
it might otherwise be. Moreover, other studies of citizen calls to the
police in Canada and in the northeastern United States provide examples
of calls with essentially the same organization as that discussed here (cf.
Sharrock and Turner 1983, Meehan 1983).
2. Calling the police. While telephone calls to a police emergency
number have their own specificity—they are usually requests for assist-
ance of the type that the police in particular have the responsibility to
provide—they are, nevertheless, telephone calls. Moreover, even first-
time callers of the police will probably have little difficulty in working
their way through such calls, suggesting that some rather general social
organizational principles bear on the management of this particular kind
of encounter. In this regard, it appears that calls to the police are one
member of a class of telephone transactions here termed 'service
calls'. The alacrity and precision with which they are generally con-
ducted is likely the consequence of the application by both complaint-
taker and caller of methodic procedures appropriate to calls of1 this type,
suitably adapted to the specific feature of a 'call to the police .
The interactional work of a call, then, addresses issues posed in the
first instance because it is a telephone encounter. Such work is also
responsive to issues that emerge because placing the call is by that very
act a 'request' for service of a particular sort. Moreover, in some but by
no means all of the calls, caller makes a fairly explicit request for
assitance, as in (2).
(2) (MCE 21-21:283
C: Yes if ya gotuh squad car could ya send one over tuh
Beach an Owen Avenue.
CT: What's the problem there.
But just as commonly, callers simply report a trouble.
(3) [MCE 21-23:32]
C: Yes, I'd like tuh report uh loud party.
CT: Where is it ma'am.
If the caller requests assistance, as in example (2), the result will be a
query from the complaint-taker concerning the nature of the problem
which, obviously, is not required in the case illustrated by example (3).
The 'request' format explicitly projects the remedy or response to an as
yet unvoiced 'complaint1 to an organization whose business it is to know
about and respond to it, concerning some state of affairs that is in need
of redress.
What is being made focally relevant in the opening segment of the call
is the division of labor in our society with regard to matters of social
control. In Sharrock and Turner's terms (1978:187), a socially organized
resource—police power—is being mobilized to deal with a problem that
others cannot or choose not to deal with by other means. This mobiliza-
tion raises certain issues, i.e. the policeable nature of the problem and
its urgency (see the discussion of the 'contingencies of response', Section
2.3). The point to be noted here is that the hearability of an utterance
as a 'complaint' draws on its location within an institutional framework—
/ Don H. Zimmerman
that of policing—which touches the interactional realm through the
organization of the call. That organization provides a place, just 1after
the alignment of identities in the opening sequence, the 'first topic slot
(Schegloff and Sacks 1974, Schegloff 1979), which is where the reason
for the call is ordinarily provided.
Figure 1 presents this proposed general organization; it consists of
five major components which, it is suggested here, characterize not only
calls to the police but the broad class of service calls that occur in this
society.
Figure 1.
Opening/Identification
Complaint/Request
Interrogative series
Remedy/Response
Closing
In terms specific to the subset of service calls addressed in this paper,
the police complaint-taker and citizen-caller must (1) accomplish a
proper 'opening1, that is, align their respective identities and thus project
the nature of the call; (2) provide and/or elicit a reason for the call, e.g.
by making a 'complaint' or a 'request' for assistance; (3) arrive at a
mutually acceptable description of the reported trouble, including the
caller's stance toward or involvement with it (Whalen and Zimmerman
forthcoming) and an adequate formulation of the trouble's location—a
process involving what is called here an 'interrogative series', ordinarily
initiated and directed by the complaint-taker; (4) the offering of a
'remedy' or 'response' to the complaint or request for assistance; and (5)
the achievement of 'closing', that is, a coordinated exit from the call.
Example (4) displays this organization rather starkly.
(4) QrfCE 21-9:12]
1 CT: Mid-City Emergency.
2 C: Um yeah (.) somebody jus' vandalized my car,
3 CT: What's your address.
4 C: (Gives addressU
5 CT: Is this uh house or an apartment.
6 C: Ih tst uh house
7 CT: Uh-your las' name.
8 C: Minsky
9 CT: How you spell it.
10 C: M-i-n-s-k-y
11 CT:Wull sen' someone out to see you.
12 C: Than' you.
13 CT: umhm bye.
14 C: Bye
The opening segment is accomplished in turn 1 and in the first com-

ponent (Um yeah) of turn 2; the complaint is delivered in the second
component of turn 2, with the interrogative series initiated in turn 3 and
running through turn 10. (Here, and in segments 5, 6, and 14, the
interrogative series is indented for expositional purposes.) The
complaint response occurs in turn 11, closing the complaint-response
bracket and serving also as a 'pre-closingf move (Schegloff and Sacks
1974), with turns 12 through 14 occupied with the closing.
A virtually identical organization is evident in a call from C (caller) to
an operator (O) at a county animal control agency, reproduced in exam-
ple (5). The organization of the opening (lines 1 and 2), the complaint
(line 2), the interrogative series (lines 3 through 16), the remedy (line
17), and the closing (lines 19-20) are all remarkably comparable. The
indented utterance pair in lines 8 and 9 is a nested insertion sequence
(Schegloff 1972) within the interrogative series.
(5) (JAS 2:83:DS
1
1 O: Coun animal control?
2 C: Yes Uhm (.) There's uh an injured cat at the corner of
Jackson an1 Matthews?
3 O: Is it alive?
4 C: Mm yeah I think so r-cuz
L
5 O: Ok
6 C: it's bleeding from the mouth
7 O: Which corner? Near the hospital?
8 C: Uh no It's like uh Y'know where the flower
stan1 is?
9 O: Uh huh
10 C: Jus' there" In thuh gutter
11 O: Your name?
12 C: Sandy Hillman?
13 O: Wher're y'calling from
14 C: Home
15 O: Number?
16 C: 869-8606
17 O: Okay I'll get thuh man on call t'check it out
18 C: Thanks
19 O: Bye
20 C: Bye
A call to an airlines reservation number (A) provided in example (6)
differs only in the feature that caller's request for assistance involves
the provision of information, with the response (the information deliv-
ery) in lines 11-13 and 15 (caller's line 14 being, perhaps, an acknowl-
edgment of the receipt of information to that point and an invitation to
continue). Note that caller in line 16 requests further information,
resulting in a brief interrogative series (lines 17-18) and a further re-
sponse (line 19) before the call is closed.
(6) Q.L: Ai£]
1 A: Federated Airlines Reservations 3oanne

2 C: Hi Could you tell me what your cheapest fare would be
from Portland to Washington, D.C.?
3 A: Mmhm? Your travelling one way or round trip?
C: Round trip
5 A: And when are you planning on travelling
6 C: Uh probably in the summer sometime
7 A: Okay So it would be after the fifteenth of June
you'd be leaving?
8 C: Yes
9 A: All right About how long 'r your planning on stay-
ing
10 C: Um two or three weeks?
11 A: Okay Okay I can tell you right away the lowest rate that
we can offer um would be our Super Saver fare (.) You
would have to make your reservation at least a week
before you travel Okay for the
12 C: Mhm
13 A: lowest fare (.) you'd have to stay at least seven days (.)
and not more than fourteen days
1* C: Okay
15 A: 'Kay An1 that's going to be a round trip of three ninety-
nine
16 C: 'Kay an' what is your cheapest fare one way
17 A: On a one way?
18 C: Yeah
19 A: That would be (.) three oh nine
20 C: Okay Thank you very much
21 A: Uh huh bye bye
Examples (4) and (5) are typical of the service calls collected to date,
and point quite clearly to the strong kinship among calls of this type.
The operators working these diverse numbers do not appear to be given
specific training on how to organize a call. Rather, they are instructed
in organization policy and goals and the type of information needed to
advance these aims. It seems likely, then, that the common discourse
features these calls exhibit are achieved by the use of methodic proce-
dures generally available not only to the operator but also—at least in
part—to the caller, the call's actual features being an interactional
accomplishment. The achieved structures of this concerted effort are
considered next.
2.1 Opening identification. Openings, like closings, bear upon 'the

distinctive characteristics of various "types" of conversations' (Schegloff
1979:25). They provide a place where 'the kind of conversation this is', is
indicated, negotiated, and ratified.
In face-to-face conversations, one may speak of processes prior to a
conversational opening, a 'pre-beginning' in Schegloff's terminology
(1979:27), where the issue of identification is salient. The identity of
the person with whom one is about to speak determines how the
conversation will be initiated (i.e. the components of initial turns at

talk) as well as what 'type' of conversation will be projected.
Identification is both 'categorical' and 'recognitional', the latter refer-
ring to the recognition of 'particular, known others' (Schegloff 1979:25)
in face-to-face encounters. Identification is ordinarily accomplished
through visual means, and has both cognitive and social status, the latter
when identification is 'prospectively reciprocal' (Schegloff 1979; cf. also
Schiffrin 1977) and displayed in some form (e.g. a mutual turning toward,
approach, etc.).
In telephone conversations, identification/recognition cannot routinely
occur prior to the initiation of a conversation, nor can it be a condition
for its occurrence. Thus, a distinctive feature of telephone call openings
is the identification/recognition 'problem' and its solution; callers and
answerers can be shown to orient to the problem and to attempt to
resolve it in and through, and in some instances parallel with, other
activities in their talk.
Space does not permit a detailed review of the management of the
identification/recognition issue in casual telephone contacts (Schegloff
1979). It is sufficient here to point out that in calls to service agencies
in general, and to Mid-City police as a case in point, answerers employ
categorical self-identification. Three different forms of categorical
self-identification were employed by complaint-takers in the Mid-City
Police Department in their opening turn: (1) Mid-City Emergency; (2)
Mid-City Police and Fire; (3) Emergency Center. These designations
may be considered to be the first component in answerer's first turn; in
Mid-City, they are the only components. Meehan's (1983) 'Big City'
police sometimes employed one or more additional suborganizational
identifiers. Example (6) illustrated the practice of using three com-
ponents (Federated Airlines Reservations Joanne on line 1). Thus, the
answerer's opening can vary in the depth and detail of the identifica-
tion. The pattern of the callers' response to complaint-takers' opening is
exemplified in example (7).
(7) CMCE 21-2:Q

1 CT: Mid-City emergency.
2 C: Yes I'd uh like tuh report uh disturbance...
The caller's first turn typically offers a yes or other acknowledgment
marker, which may be characterized as confirmation that caller has
reached the number intended and as a preface for additional components
which will ordinarily complete the alignment between complaint-taker
as an 'agent of the police1 and caller as 'citizen/complainant'.
That callers are oriented to answerer's self-identification and its
possible implications for the call is evident in the fact that each an-
swerer variant carries with it its own problematic: the use of the term
emergency can draw responses such as Oops, this isn't an emergency,
while the use of the pair Police and Fire can evoke the 'switchboard'
request, Police, please. Note here that caller withholds an acknowledg-
ment if such an alignment issue does occur, but in its place addresses
that issue, as shown in (8).
(8) UACE 21-25:33

2 C: Can I 've the police, please.
3 CT: This is police.
k C: Oh uh there's some loud music over on...
(9) ©CE 21-8:1(3
1 CT: Mid-City emergency ((silence)) Hello?

2 C: Hullo Is thuh ((unintelligible)) sherf's deparmen
3 CT: Uh yes (an) what
k C: Is this thuh Marlin Count-j-sheriff depar
L
5 CT: No: it's the Mid-City police.
6 C: Uh I would like tuh make uh (hones') phone call.
Caller in example (8) apparently hears complaint-taker's opening as a
'switchboard' situation—the term emergency, as Meehan (1983:82-88) has
suggested, is by design rather broad in its coverage, and caller specifies
the particular 'emergency' organization she wishes to contact. In exam-
ple (9), caller does not respond to complaint-taker's opening, leading to
her Hello? (line 1) to determine if there is anyone on the line, and call-
er's confirmatory Hullo. Caller's first turn is devoted to checking out
the identity of answerer and is followed by complaint-taker's mistaken
answer to caller's query, quickly corrected by her repeat-request (what)
in line 3. The confusion is resolved in line 3 with the caller offering a
preface to his subsequent complaint.
Other types of service calls illustrate the same sort of identification
issue. In a call to a pottery shop, Caller (C) encounters from Answerer
(A) an unexpected personal rather than categorical identification.
(10) [AR: Pottery}
1 A: Irene
2 C: Uhm (.) let's see is this the pottery retail place?
3 A: Uh yes uh huh
4 C: Uhm I was wondering if...
A partially cut off identification from Answerer leads to the same

'identity checking1 from Caller.
(11) (XL: Train Reservations: 2]
1 A: 'svations Miss Ghent speaking

2 C: Is this Train Reservations?
3 A: Yes it is
A call, then, does not proceed to further business until the type of call
projected by complaint-taker's opening is 'recognized' by the caller, that
is, reconciled with the caller's sense of who the recipient of the up-
coming complaint or report should properly be. While the point cannot
be pursued in detail here, it seems evident that the care taken by
participants in crafting an opening, that is, in aligning their situated

identities, provides a working framework which commits participants to
the nature of the occasion summoned up by the initiation of the call and
thereby provides for the presumptive hearability of utterances as rele-
vant to the purposes of that occasion of talk (cf. Meehan 1983:88-97).
2.2 Complaint/request. The complaint-request component (line 2 in
examples (4), (5), (6), and (7)) is ordinarily the next event following the
caller's acknowledgment of complaint-taker's categorical self-identifica-
tion. The examples cited to this point display a 'single utterance format1
for delivering the complaint (Sharrock and Turner 1978:175-77) although
the report of trouble can be delivered in a 'multiple utterance' format,
as in example (12).
(12) [MCE 21-7:8-9^
2 C: This is thuh Kit-Kat Club on ten one three oh Williams?
3 CT: Mmhm
4 C: and thuh laundrymat (.) Jim's laundrymat?
5 CT: Mmhm
6 C: it's down the street here a bit (.) tst anyway it's lef open.
It's wide open.
7 CT: Mmhm
8 C: An it's supposed to be locked at nine a clock so I don' know
if somebody broke in there or what's goin' on
9 CT: We'll get somebody there.
In this call, caller provides a categorical self-identification (Kit-Kat
Club) and a location delivered with a question contour, or 'try-marker'
(line 2), which solicits acknowledgment or recognition and is followed by
complaint-taker's recognitional Mmhm in line 3. The continuation of
caller's utterance in line 4 then identifies another establishment, again
offered with a try-marker (Jim's laundrymat?) followed by caller's
specification of its location relative to that of caller's establishment
(lines b and 6). The trouble (it's wide open) is then reported in lines 6
and 8, with complaint-taker's continuer (Schegloff 1982) in line 7 and
followed immediately in line 9 by complaint-taker's promise of assist-
ance.
The example just described, example (12), is one of the very few calls
in which the caller, building her designedly multiple-utterance turn with
the collaboration of the complaint-taker, provides a connected account
of the candidate policeable trouble (i.e. the situation that is potentially
a problem for police to handle) and its location without any interrogation
(and hence, without any insertion sequence). The caller's use of try-
markers in lines 2 and 4 elicits acknowledgment—not only of understand-
ing but of the prefatory character of the utterance to that point, an
utterance which is displaying both how the caller might have come to
know of the trouble, and why she might be interested, as a context for
the report (cf. Whalen and Zimmerman forthcoming). The presentation
of the complaint (including locational information) is directly followed
by the remedy, or response to the complaint, which occurs at the turn-
transition beginning on line 9.
There are two points to be stressed here. The first is that the com-
plaint or request for assistance—the reason for the call—is hearable as a
matter requiring response or remedy (Sharrock and Turner 1978:174-
75). In those calls where the citizen's 'complaint package' (Meehan
1983:101-8) is developed without intervention by the complaint-taker,
this response is found in the next turn adjacent to the reported trouble,
as in example (12). The other point is that other talk—sometimes a good
deal of it—routinely intervenes between the initiation of a complaint and
the delivery of a remedy, as in example (13), where both location and the
problem are elicited by a course of questioning inserted between the
request/response bracket.
(13) [UCE 21-24a:33
1 CT:
CT: Mid-City Emergency.
2 C: Yeah we'd like ya tuh send an ambulance out
3 CT: Where to.
C: Uh: sixty- uh sixty five and uh: ((background voice))
um ((background voice))
5 CT: Tuh where?
6 C: Eighty second=
7 CT: =Sixty five what Sixty five an what
8 C: Ninety second. Sixty fifth street an ninety second
9 CT: Ninety second Avenue?
10 C: Yeah
11 CT: What's thuh problem.
12 C: Uh there's been un accident.
The work accomplished by this elicitation defines the next major compo-
nent of overall organization—the interrogative series.
2.3 The interrogative series. It is by now fairly clear that in a call to
a 'service agency' like the police, caller's request (or complaint) routine-
ly elicits responses which are not responses to the request itself, but to
the issues that it raises. This was evident, for example, in (6), in which
the reservation desk had to know caller's travel dates to determine the
cheapest fare.
Of particular interest here is the fact that making a request engages
an organization rather than simply an individual, and thus, varying with
the circumstances of the call, encounters contingencies of response
which are evident to the organizational personnel receiving the request,
but are perhaps unknown or only vaguely perceived by caller. Thus, a
complaint or request routinely involves some processing, that is, some
course in which its features—many of which have yet to be made evi-
dent—are fit to the requirements of organizational response (cf. Merritt
1977; Frankel 1977, 1981).
In their study of calls to the police, Sharrock and Turner (1978:175)
speak of a 'police-initiated fact-seeking sequence' dealing with matters
preliminary to police response. In their corpus, as in the Mid-City
materials, a series of question-answer pairs characterize this 'fact-
seeking sequence' or interrogative series, which appears to function as
an insertion sequence between a complaint and its remedy. Schegloff
(1972:76-79) discusses insertion sequences in terms of embedded or

nested sequences of questions and answers, as in the following adapted
from Schegloff (1972:78-79).
Q base SI: Are you coming tonight?
Ql S2: Can I bring a guest?
Al SI: Sure.
A base S2: I'll be there.
As evident in this example, the nested Q/A pair deals with a contingency
of response to Q base, namely, whether S2 can bring a guest or not.
Additional nested Q/A pairs could conceivably be inserted, ultimately
motivated by Q base and coming to rest in A base. There appears to be
nothing in the logic of the insertion sequence, however, to prohibit the
deployment of a series of chained (as opposed to nested) Q/A pairs
addressed to a set of more or less discrete contingencies of response (cf.
Merritt 1977).
Example (14), taken from Maynard and Wilson (1980:301-5), provides
an interesting illustration of an insertion sequence consisting of just such
a series of Q, A pairs in the context of sentencing procedures in a muni-
cipal court.
(14) QlSS: 301-300
1 P.D.: Your Honor, we request immediate sentencing and
2 waive the probation report.
3 Judge: What's his record:
4 P.D.: He has a prior drunk and a GTA Uirand Theft
5 Nothing serious. This is just a shoplifting case.
6 He did enter the K-Mart with the intent to
steal.
7 But really all we have here is a petty theft.
8 Judge: What do the people have?
9 D.A.: Nothing either way.
10 Judge: Any objections to immediate sentencing?
11 D.A.: No
12 Judge: How long has he been in?
13 P.D.: Eighty-three days.
14 Judge: I make this a misdemeanor by P.C. article 17 and
15 sentence you to ninety days in County Jail, with
16 credit for time served.
The circumstances in this interrogative series of example (14) are dif-
ferent, of course; it is face-to-face, and involves three participants. It
is, moreover, a request from one sector of an organization, represented
by the Deputy Public Defender (P.D.), to another sector (the Judge) over
a matter that is also the professional concern of a third, the Deputy
District Attorney (D.A.). Nevertheless, this brief interaction is inte-
resting because the P.D.'s request also engages an organization and its
contingencies of response (although the latter are undoubtedly known to
the P.D. and figured in his request) which are elicited by the Judge.
Requests (or complaints) that are followed by insertion sequences like

the interrogative series can thus be seen to occur in a variety of organi-
zational contexts. These episodes of questioning address a series of
contingencies associated with the particular organization's response to a
request for its service and appear to be understood as so addressed by
persons initiating the request. That is, most callers in the Mid-City
corpus appear to recognize what the person on the other end of the line
is doing and grant it legitimacy, a comprehension and acceptance which
facilitates their delivery of a response.
It is in this moment between the request and its response that two
rather major issues are addressed and, in most instances, resolved with-
out marked trouble. First, in the case of the call to the police, caller
must furnish an account or description of some occurrence so that
complaint taker can hear the incident as policeable trouble. Descrip-
tions are answerable to criteria of adequacy that vary by socially struc-
tured circumstances (e.g. gossip, courtroom testimony, ethnographic
reportage, etc.) and are thus particularly vulnerable to contingencies of
response (Whalen and Zimmerman forthcoming, Sharrock and Turner
1978, Frankel 1981). The second issue involves the achievement of an
adequate locational formulation (usually in terms of specific addresses
or street intersections). In Mid-City, for example, complaint-takers
required street addresses or intersections in order to process a com-
plaint. This information was entered into their computer and verified.
The race, sex, number, and location of persons involved in a reported
incident may also be asked for, as well as whether weapons are present.
The completion of the interrogative series is marked by the com-
plaint-taker's delivery of a promise of assistance, i.e. the remedy or
response to the complaint. This promise has structural equivalents in
other service calls, e.g. the reservation clerk's provision of the requested
fare information (example (6)) and the judge's sentencing (example
(14)). It is important to point out that the promise to send assistance is
adjacent to the closing sequence, and itself can be analyzed as a pre-
closing move (Schegloff and Sacks 1974, Zimmerman and Loseke 1981).
Its role as an initator of closing—particularly the closing of a telephone
call—furnishes a powerful structural motive to withhold it until the
information deemed necessary for the police reponse is elicited.
2.4 Remedy/response and closing. The fourth segment of the call is
the remedy/response move by the complaint-taker in the form of a
promise of assistance, e.g. Wull sen' someone out to see you. As already
suggested, utterances of this general format following the interrogative
series are hearable both as the response to the initial complaint and as a
pre-closing move. That complaint-takers and callers are oriented to this
type of utterance as a response to the complaint, and hence as the
occasion to terminate the call, is indicated by the fact that complaint-
takers will 'abort' its production by an abrupt cut-off (indicated in exam-
ple (15) by the hyphen), followed quickly by the reinstitution of question-
ing if they are uncertain of some item.
(15) Dl-16a:21-22D
13 CT: We'll get somebod- are you sure it's Southeast and not
Northeast.
Likewise, callers will withhold acknowledgment of the promise of assist-
ance to pursue some feature of the complaint presumably not yet dealt
with to their satisfaction, as example (16) shows.
(16) C21-7:9-9j
9 CT: We'll get somebody there.
10 C: _We do not have thuh owner so you must have it on file at
thuh Northside station.
Upon receiving the promise of assistance, callers typically provide
some form of acknowledgment (as in example (0, where caller says
thank you (line 11)). The promise thereby furnishes the response to the
caller's complaint and initiates a pre-closing move completed by caller's
acknowledgment (cf. Schegloff and Sacks 1974). It is upon the comple-
tion of this pivotal sequence that a terminal-exchange (Schegloff and
Sacks 1974) ensues by which the call is brought to a fairly rapid conclu-
sion.
Calls to the police—like other service calls—have a single purpose:
they have an expectably focused matter to bring to the attention of the
police, and then project by their very initiation the expectation of police
response to that matter, i.e. the dispatch of a squad car. It is this
feature of the1 call that provides for the status of the promise, or 'action
announcement by the complaint-taker (we'll get somebody there) as the
initial turn in the terminal sequence.
Like exit from ordinary conversations and ordinary calls, exit from
police calls is managed with reference to speaker's rights to resume or
initiate various lines of talk, the actual exchange of the farewells being
preceded by a closing section in which speaker willingness to forego
further talk is coordinated and displayed.
3. Conclusions. If talk is indeed anchored in the world—a world which
includes the social world—then the nature of this connection surely
cannot be ignored when considering the meaning of utterances. It would
miss the point, however, to treat this proposal as yet another call to
respect 'the context'. Every utterance is interpreted in some context,
however impoverished. Even the contemplation of isolated utterances
contrived by the linguist, alluded to in the opening remarks, draws on a
context formed by the existence of a scholarly community sharing many
of the presuppositions about language, speakers, and the world, and
which is disposed to treat such an enterprise as sensible. The issue is not
whether context is considered, but how, from what perspective, and with
respect to which features?
Something else is implied then, namely, that talk for and between
persons is social in character, that it occurs within and with respect to
social occasions, occasions which are coupled to a greater or lesser
degree to some larger institutional arena or some series of such arenas.
22^ / Don H. Zimmerman
The talk is not independent of its. occasion, and merely set down in it,
but rather is coconstitutive of it. A framework is progressively erect-
ed for the comprehension of what is said and, in being said, done. Talk,
by bringing into focus a particular type of occasion, summons up the
context of its interpretation as well.
Calling the police was the case in point here. These calls were viewed
as instances of a more general class of service calls, displaying a defi-
nite achieved organization. The term 'achieved' was employed to
emphasize a particular view of the stable, recurrent features exhibited
by these calls. The organization of these calls was treated as a
situational accomplishment of callers and service operators' applying
general interactional skills and specific social knowledge to tasks and
issues posed by the nature of the telephone encounter as such, as well as
the contingencies involved in securing and providing the particular
service in question.
Thus, callers and operators—complaint-takers, in the case of the Mid-
City police—have to achieve an opening, which involves, among other
things, the alignment of identities appropriate to the occasion and
consequential for further inference and action in the call. A request for
service—which can be accomplished in a number of ways—must be made
and understood as such, and a course of work initiated, in the form of an
insertion sequence, through which the contingencies of providing (or
promising to provide) the service requested, can be addressed and satis-
fied. A closing attuned to the specificities of the call must be jointly
accomplished. Although the circumstances displayed by each police call
were varied, each was managed in light of its individuality and with
respect to the resolution of the issues posed by the mediated encounter
and the jointly acknowledged purpose of the call. The interactional
order (e.g. caller vis-a-vis complaint-taker) and the institutional order
(e.g. citizen-complainant vis-a-vis the police) are articulated within the
unfolding organization of telephone exchange.
Two further points bear emphasis here. First, if the lineaments of a
social occasion and its linkages to a larger institution and context are
brought to focus in these calls, then both the occasion and its relevant
institutional site are as much the achieved outcome of encounters as the
context for interpreting the discourse occurring within them. From the
vantage point of the completed call (or any species of discourse), the
dynamic reflexive nature of what transpires may easily be overlooked.
For that matter, the importance of a practical comprehension of utter-
ances by participants sharing the responsibility for managing the out-
come of the encounter may also be overlooked, and this raises the sec-
ond point.
The production of most discourse is fundamentally interactive (cf.
Schegloff 1981). Each 'move' in a discourse establishes some kind of
understanding of the previous move, and of the larger context, and is
subject to acceptance, modification, or reconstitution, depending upon
the subsequent move by another that it motivates. These 'scenic inter-
pretations' are structurally relevant in that they are requisite for ad-
dressing and accomplishing the purposes of the encounter, for working it
through, as it were. The issue for interactants is to understand enough
to proceed (which is a principle of practical interpretation), rather than
understanding for its own sake (a principle of theoretical interpretation).
3.1 A final note. The problem of context, of the relationship of talk
to its occasion, can barely be posed, much less addressed, in one investi-
gation or in a short paper. It does seem likely, however, that the issues
involved will require some concession to sociological sensibilities, as
well as recourse to sociologically sensitized studies of talk in actual
settings. It appears particularly strategic to examine those places in
society where talk is patently a major instrument for the accomplish-
ment of the setting's work, as in the case of calling the police.
Notes
This paper is based on work undertaken with several persons. I should

like to acknowledge and thank Donilene Loseke, Deirdre Boden, Marilyn
Whalen, Lynda LaFontaine, Terri Cook, 3udy Blum, and Audrey Rice.
Thomas P. Wilson provided helpful comments. Deborah Schiffrin also
gave particularly helpful suggestions. Responsibility for this paper, and
most particularly for its shortcomings, is mine. Thanks are also due
Mary Wiemann for her patient and efficient preparation of the manu-
script.
The research reported here was supported by the Center for the Study
of Violence and Anti-Social Behavior (Grant No. 34616-03). Portions of
the analysis reported here were begun while I was guest professor at the
Center for Survey Research, Mannheim, West Germany, June-3uly 1981.
1. In this paper, language behavior, or talk, is viewed as a form of
social interaction.
2. This is not to say that service calls do not employ explicit for-
mats. As pointed out further on, specific items of information may be
required by the organization in question, and the elicitation may be
guided by written or computer-prompted instructions (cf. Frankel
1981). The point stressed here is that the overall structure common to
classes of this type is not achieved by reliance on an explicit format.
3. The calls were obtained by Donilene Loseke during a brief period of
field observation at Mid-City. The analysis reported here had its roots
in a co-authored, unpublished paper (Zimmerman and Loseke 1981). This
section, and other parts of the study, borrow from portions of that
earlier work which drew heavily on Loseke's field notes and observations.
k. With two exceptions, the 26 complaint-takers employed by Mid-
City were female. All dispatchers were male. The sex of callers was
imputed by transcribers from informal assessment of voice qualities.
Accordingly, all pronouns referring to persons are used in a specific
rather than generic sense.
5. Segments from transcripts of those calls are presented in simpli-
fied form. The notation (.) indicates a very brief pause; words enclosed
in parentheses indicate transcriber uncertainty.
6. All transcript segments are identified as to their source. 'MCE'
denotes 'Mid-City Emergency1; the next digits, e.g. 21-26, designate tape
channel number and call number ; the last digits indicate page number of
the transcript. Segments from other sources are denoted or identified
by similar coding.
7. In speaking of service calls, reference is intended to those tele-
phone contacts in which callers have recourse to a telephone 'service
port' (Goffman 1983b) for the purpose of securing some assistance or
service, whether in the form of simple information (such as determining

bus or train schedules or the cost of some commodity), or of accomplish-
ing some transaction (such as ordering from a catalog, making a travel
reservation), or securing the promise of assistance (such as arranging for
the plumber to come to the house or for the police to be dispatched).
See Merritt (1977) and Frankel (1977), for a discussion of service trans-
actions.
8. Meehan (1933:81-82) reports a range of openings involving up to
three components: (1) an organization identifier; (2) a subunit identifier;
and (3) answerer's organizational identity. Components (2) or (3) are
deletable, whereas (1) is not. Thus, actual calls to the Big City police
may be answered with the organization identifier alone:
[BC W:23
P: Big City Police Emergency
Or calls may combine component (1) with other components:
CBC 18-23
P: Big City Police Emergency three thirty five
I3C 32-23
P: Big City Emergency Dispatcher McGrew
Meehan provides a perceptive and useful discussion of the relevance of a
call's opening to the nature and problematics of the business to be trans-
acted within it. His analysis of the Big City calls and the analysis re-
ported in this paper are remarkably convergent, although done indepen-
dently and without knowledge of the other. Space precludes discussing
Meehan's findings here.
9. The relationship between complaint-taker's opening and that of the
caller is fairly complex, and has been simplified here for reasons of
space. For example, in some calls, caller's first component consists of a
frame, This is, followed by a categorical self-identification, usually a
business or organization. It should be noted, moreover, that the identi-
ties involved in the opening, particularly the identity of the caller, are
subject to transformation over the course of the call. Callers may
become 'victims', 'passers-by', 'neighbors', 'interested parties', etc. The
negotiation of identities in calls to the police is the subject of another
paper in progress.
10. The term 'withhold' might refer to the fact that when identifica-
tion issues emerge, acknowledgment is absent and talk directed to the
issue proceeds directly. A stronger sense is intended, however, as sug-
gested in (i) and (ii).
(i) [MCE 17-5:10£)
1 CT: Mid-City police an' fire
2 C: Cu Yeah uhm I'd like tuh
(ii) CMCE 17-6:691
1 CT: Mid-City police an' fire
2 C: Hi oh I want thuh police
In (i), caller's opening word is self-interrupted, or at least is left
unfinished, and is followed by an acknowledgment token. This suggests
an orientation to the appropriateness of the latter's occurrence as a first
component of the turn. Acknowledgments can be 'retracted' by use of
'surprise' markers like oh_ in (ii), followed by an utterance 'explaining' the
contravention of caller's original expectations. Acknowledgment is thus
not merely present or absent but accountably present or absent.
11. Saying does not of itself necessarily make it so. In the present
case, calling the police, while it initiates a state of talk, also brings into
play an apparatus which was already 'there', e.g. the police organization,
its division of labor, the occupational role of 'complaint-taker', policy,
procedure, accumulated experience, etc. But the call itself is what
achieves the instantiation and integration of these elements into the
recurring occasion, 'calling the police'.
References
Brown, Gillian, and George Yule. 1983. Discourse analysis. Cam-
bridge: Cambridge University Press.
Frankel, Richard M. 1977. Between client and server: Aspects of the
organization of conversational exchange during service. Unpublished
Ph.D dissertation, The City University of New York.
Frankel, Richard M. 1981. 'I wuz wondering-uhm could Raid uhm affect
the brain permanently d'y'know?': Record keeping in the context of
communicative interaction. Paper presented at the annual meeting of
the American Sociological Association, Toronto, Canada.
Goffman, Erving. 1974. Frame analysis: An essay in the organization
of experience. New York: Harper Colophon.
Goffman, Erving. 1983a. The interaction order. American Sociological
Review 48.1-17.
Goffman, Erving. 1983b. Felicity's condition. American Journal of
Sociology 89.1-51.
Maynard, Douglas W., and Thomas P. Wilson. 1980. On the reification
of social structure. In: Current perspectives in social theory, vol.1.
Edited by Scott G. McNall. New York: 3AI Press. 287-322.
Meehan, Albert T. 1983. For the record: Organizational and inter-
actional practices for producing police records on juveniles. Unpub-
lished Ph.D. dissertation, Boston University.
Meehan, Albert T. 1984. Assessing the police-worthiness of citizen's
complaints to the police: Accountability and the negotiation of
'facts.' Unpublished paper.
Merritt, Marilyn. 1977. On questions following questions in service
encounters. Language in Society 5.315-57'.
Schegloff, Emanuel. 1972. Notes on a conversational practice: Formu-
lating place. In: Studies in social interaction. Edited by David Sud-
now. New York: Free Press. 75-102.
Schegloff, Emanuel. 1979. Identification and recognition in telephone
call openings. In: Everyday language: Studies in ethnomethodology.
Edited by George Psathas. Boston: Irvington. 23-78.
Schegloff, Emanuel. 1982. Discourse as an interactional achievement:
Some uses of 'uh huh1 and other things that come between sentences.
In: Georgetown University Round Table on Languages and Linguistics
1981. Edited by Deborah Tannen. Washington, D.C.: Georgetown
Schegloff, Emanuel, and Harvey Sacks. 1974. Opening up closings. In:
Ethnomethodology. Edited by Roy Turner. Harmondsworth, England:
Penguin. 233-64.
Schiffrin, Deborah. 1977. Opening encounters. American Sociological
Review 44.679-91.
Sharrock, W. W., and Roy Turner. 1978. On a conversational environ-
ment for equivocality. In: Studies in the organization of conversa-
tional interaction. Edited by Jim Schenkein. New York: Academic
Press. 173-98.
Zimmerman, Don H. and Donilene Loseke. 1981. Citizen calls to the
police. Unpublished paper.
Zimmerman, Don H., and Marilyn Whalen. (forthcoming) Telling
trouble: Citizen calls to the police. In: Language in institutional
settings. Edited by Richard M. Frankel. Norwood, N.3.: Ablex.
IDEALIZATION IN SOCIOLINGUISTICS:
THE CHOICE OF THE STANDARD DIALECT
Alan Davies
University of Edinburgh
0. Introduction. In this paper, I argue that idealization is as necessary
in sociolinguistics as it is in linguistics. In so doing, I assume that no-
tions of the Speech Community and of the Standard Language are part of
that idealization. I discuss a recent British example of the ongoing de-
bate over Standard and Dialect in education; I conclude that on educa-
tional issues, sociolinguistics may advise but not judge.
Before I begin, I suggest four 'truisms' about language to set the scene.
(Truisms, of course, are other people's false ideas.)
0.1 All languages are equal. This is a frequent statement in intro-
ductory linguistics textbooks but it has an Orwellian penumbra about it.
It means that all languages (all language 'codes') have equivalent (not the
same) devices like grammars and vocabularies, and that they are all ca-
pable of doing the same things, e.g. expressing intimacy, poetry; telling
jokes, writing novels; teaching science, engineering, medicine; running
governments, operating air traffic controls and computer systems; main-
taining genealogies and oral traditions. But surely this is a statement of
linguistic potential; in terms of sociolinguistic suitability—that is, choos-
ing a language for a particular function, use, or set of uses—languages
are certainly not equal. Of course, they can develop, but at any point in
time they are not equally developed. An example is languages that
currently lack writing systems, and therefore swiftly become less and
less developed, while those already developed in terms of domain of use,
in terms of types and tokens of use (e.g. English), snowball on to greater
and greater developments.
0.2 Political change favours dialectal maintenance. The idea here

presumably is that revolutions are perpetrated on behalf of minorities or
underdogs, and that, therefore, their interests and identities gain in
strength and respectability. It is, of course, more complicated: the Irish
language served as a symbol for Irish independence, but when Ireland
229
230 / Alan Davies
achieved independence the language was, in effect, abandoned. It was
the Irish example against which Saunders Lewis (1962) warned Welsh na-
tionalists. Make sure of the language first, he urged, before aiming for
political freedom. Indeed, it has been suggested (and the low vote in
Wales in favour of devolution (11%) is evidence for this) that in an in-
dependent Wales, the Welsh language would get far shorter shrift than it
presently does from London. France illustrates the same trend of aban-
donment of minorities. At the time of the French Revolution, the Con-
vention decided that only one language, French, was revolutionary, and
that others (Basque, Provencal, etc.) were living remembrances of feu-
dalism and therefore counter-revolutionary. France has a well-known
tradition of monolingual policies both in France itself and in its colonies.
Minority languages and dialects do not survive revolutions.
0.3 Different language functions require different language codes.
The argument behind this is that there is a linguistic virtue in having
more than one dialect code (the bidialectal, tridialectal...policy) avail-
able in order to carry out a wide range of functions. Now there may be
other grounds for a bidialectal policy, but it is not the case that differ-
ent functions (home, work, school, religion, etc.) need different codes,
though of course they may require different uses of the same code.
Confusion about code and function did disservice to Bernstein's work. In
any case, my first truism, 'all languages are equal', and my third cannot
both be true.
0.* Speech communities have unique membership. This is the identity
argument again, that in language and dialect, as in life, we have to
choose. But things are not like that. Indeed, if this statement is true,
then no one can be bilingual or bidialectal—which is manifestly absurd.
The fact surely is that speech communities, like other social groups,
overlap, and link up in a multiplicity of ways; we are, most of us, mem-
bers of more than one speech community.
1. Standard or dialect: An educational choice. The standard-vs-
dialect issue is not an old problem which is now resurfacing in contem-
porary contexts of varieties of world English and maintenance of ethnic
minority languages. Rather, it is an old solution to the permanent edu-
cational problem of the best medium for curriculum efficiency. Educa-
tional systems have typically assumed, not always correctly, that the
speech community is a pretheoretical political primitive and that it is
not the business of education to promote or create potential speech
communities. Fishman (1977) reminds us that if we accept a pluralistic
model, as advocated, for example, by Kjolseth (1973), then we imme-
diately face the practical standard-dialect issue: which Arabic? which
Chinese? which Italian? which Punjabi? In some speech communities,
the issue may be resolved by choosing a superposed standard, e.g. Urdu
in place of 'standard' Punjabi or English in place of 'standard' Tok Pisin.
Such a choice is curiously easier than attempting to resolve competing
claims among home dialects.
Educational choices about language, the medium for writing and for
speech (which may be different), and the selection of foreign language
options, are neither linguistic nor sociolinguistic choices. Although
Idealization in sociolinguistics / 231
linguistic and sociolinguistic ideas and research inform those choices,
they do no more; they provide evidence about distribution, attitudes,
elaboration, and codification which are taken into account in the court
of educational decision—a court which is primarily a political one. In
that court, what determines the final decision is an educational philos-
ophy about the kind of society we wish ours to be, with the recognition
that for the advantage of pluralism in dialect and language maintenance,
there is the price of fragmentation, nonintegration through one standard,
and lack of proficiency for many in that standard. On the other hand,
for the advantage of efficiency, of proficiency in and integration
through one standard, there are the disadvantages of alienation, overall
cultural loss, and possible cognitive dissonance.
Of course, the choices that educational systems make will attempt to
secure as many advantages and as few disadvantages as possible for all
members, but in my view there is no escape from making these choices
on the basis of somewhat idealized models (getting the best for most)
and making provision for individual exceptions and special cases on an ad
hoc basis. As always, revelations about special problems--e.g. the deaf,
the illiterate, the minority language—take us back to a reexamination of
the provision for the majority, and to our general educational philosophy.
If our choice falls on minority dialect and language maintenance, then
the majority must be involved.
My concern here is with the debate over institutionalised language
variety. This debate has recently focussed on the issue of which dialect
to use for language maintenance in bilingual education programmes (Tosi
1982, Rosen and Burgess 1980, Trueba and Barnett-Mizrahi 1979).
Cummins' (1979) interdependency hypothesis provides a psychological
rigour to this debate, but if indeed he is right about the necessity to cog-
nitive/academic language proficiency of the home language, then our
conclusion can only be pessimistic since many, perhaps most, children in
the world's schools learn to read in a superposed code.
2. Language vs. code: Enter linguists. The language/code debate
began a generation ago in the 1950s, when a connection was made be-
tween language deprivation and disadvantage—a development of earlier
attempts to connect intelligence to educational disadvantage. The
issues in the language debate seem to be as follows: children in certain
social and ethnic groups (e.g. the working class and blacks) are dis-
advantaged, and they do not do as well as they should or could in educa-
tion. Such children also employ different language varieties. They are
thus seen to suffer from deprived language.
Through the late fifties and sixties, in work such as that of Bereiter
and Engelmann (1966) and of Bernstein (1971), and in the Head Start pro-
grams and the Educational Priority Areas, the theory was developed and
remedial action was taken to provide enriched language. At the same
time, there was the more linguistic debate as to the nature of the de-
prived language—was it a language deficit (the groups in question speak-
ing an impoverished variety) or was it a language difference (their
variety was not impoverished, just different)? Notice that I have de-
liberately used the term 'variety', not 'dialect', because I wanted to avoid
committing myself as to whether this argument was about code (i.e. the
structure of language) or about the use of code. In general, those who
232 / Alan Davies
saw the argument as being about code supported the deficit view; those
who saw it as being about the use of code supported one or other version
of the difference view.
It is at this point that linguists entered the argument, largely in order
to present specialist advice on the nature of those alternative minority
varieties, which I will here call dialects. A united front is presented by,
for example, Labov (1969), Trudgill (1975), maintaining that nonstandard
dialects, i.e. those of the working class, the blacks, and, by implication
(explicitly so in Trudgill's work) the Scots, are equal, fully structured,
and in no way impoverished. In this view, Standard English in not a
superior or primary dialect, but is one dialect among others. In switch-
ing to this sociolinguistic view, it is as though the obviousness that is
accorded to the necessary idealization in linguistics (dealing only with
well-formed, decontextualised, written sentences) is forgotten. Socio-
linguists, whose very data are contextualised language and language
variety, can easily forget that they too are describing and analysing data
at a distance, not the authentic data of a realistic speech community,
but partly contextualised, partly destandardised data. Otherwise, they
could not analyse, they could not study. Reflection, analysis, and aca-
demic research must always deal with somewhat idealized data tran-
scribed according to some imposed system, analysed according to some
other system, etc., while judgements and pragmatic decisions about
educational questions and problems must deal with realities.
2.1 Sociolinguistic idealization. I now want to take further the ques-
tion of idealization in relation to the problem of data. It was axiomatic
in linguistics (as distinct from sociolinguistics) that the data of actual
speech were not amenable to analysis since they were too variable,
•fairly degenerate in quality1 (Chomsky 1965:31), consisting of 'fragments
and deviant expressions of various sorts'. Linguistics has therefore
chosen to describe a more idealized form of data ['langue, competence 1 ),
and, within this area, has achieved success.
The sociolinguist is very much concerned with destandardising and
contextualising, i.e. de-idealizing the linguistic data, and with a wider
definition of data, one which may permit a correlation of linguistic form
with social function. So what are the sociolinguist's 'primary data'?
Not, obviously, linguistic competence, nor the 'degenerate' data of ac-
tual speech, but something in between. If it is 'communicative com-
petence' as suggested by Hymes, etc., this is presumably some kind of
idealization of both language and social structure. Bell (1976) suggests
that it is 'socially meaningful behaviour within a given society'. There
is, further, the problem of the degree to which the data are already
idealized by the very process of collection; there is also the problem of
the relationship between data and theory.
Bell (1976) argues, as far as the data-theory relationship is concerned,
that sociolinguistics can only be inductive (like structural linguistics but
unlike transformational grammar). Therefore, its task for the moment is
to discover, first for the specific speech communities and ultimately for
the universal notion 'speech community1, the system which contains 'the
set of community norms, operating principles, strategies and values
which guide the production and interpretation of speech, the community
ground rules for speaking' (Bauman and Sherzer 1974:7).
As for the idealization problem, we have already observed that it is
impossible to collect raw data. Collection and transcription already im-
pose partial analysis: a problem that cannot be avoided by any activity
which attempts to describe nondiscrete events in terms of discrete units
of analysis. The sociolinguistic view seems to be that a level of ideal-
ization lower than that in linguistics can and should be accepted. Labov
(1972:203) has argued that the ungrammaticality of everyday speech is 'a
myth with no basis in actual fact', that 'the great majority of utter-
ances—about 75%—are well-formed sentences by any criterion' and that
if you take ellipsis and universal editing rules into account, you are left
with less than 2% of ungrammatical or ill-formed sentences. If that is
so, then introspection is less appropriate a method than empirical in-
vestigation. Labov further argues that what is of interest to the socio-
linguist is the vernacular in which the speaker's utterances are well-
formed because he is not as preoccupied with his speech as he might be
in formal contexts. To some extent, then, we can accept that the pri-
mary linguistic data of sociolinguistic concern may be less idealized, i.e.
closer to the reality which the description is intended to model.
The basic model with which sociolinguistics works is that of the
Speech Community. Of course, the Speech Community is not homo-
geneous, but neither is it variable in an unlimited way; its variability
must be assumed to be systematic and not individually based. The
Speech Community is an idealization in sociolinguistics analogous to the
Standard language idealization. What sociolinguistics emphasizes in its
concern with variability is parts/groups/minorities vs. wholes/states/
majorities; but idealization away from individual behaviour still takes
place. This holds true even for such radical sociolinguistic approaches as
ethnomethodology, where the data are transformed by a residue of 'com-
mon-sense operation' (Labov 1969:201, quoting Garfinkel). What this
means is that within the majority group/minority group view, sociolin-
guistics is/can be just as normative as the standard language view: if
indeed a language is a dialect with an army, arming the dialect does not
make it less normative than the language it seeks to challenge.
3. Language vs. code: The reaction of educators. I want now to pre-
sent the opposing views in the language/code debate, as they have been
dramatically represented in recent publications. What I shall suggest is
that the apparent permissiveness of the linguistic views has led to a
stern rebuttal from an educational standpoint, an antithesis which,
bizarrely, seems to say many of the right things but for the wrong rea-
sons.
Trudgill (1975) makes a strong case for the use of nonstandard dialects
in education. But I will quote one of his more recent publications (1983)
to make the point that we are dealing with an issue of present concern.
Trudgill writes (1983:193): 'In educational circles this contrast between
Standard English and the non-standard dialects is currently the focus of
some considerable debate. To what extent, the question has been asked,
are we justified in continuing to encourage and reward the use of Stand-
ard English in British schools?' It appears that he is referring to spoken
English since he does say (pp. 194-95): 'It is certainly true that all read-
ing materials are written in Standard English and that many children
learning to read have also to cope with a new and different dialect." But
23* / Alan Davies
even here he is ambivalent, suggesting that even Standard written Eng-

lish does not need to be 'taught' since the 'differences' (from what? other
varieties?) are 'not sufficient to cause great difficulties, and most
children appear to become skilled at "translating" as they go along, at a
very early stage' (p. 195). Trudgill points out (p. 205), quite correctly,
that 'Grammatical forms which are most typical of working-class dia-
lects have low status, because of their association with groups who have
low prestige in our society,' and comments (p. 206): 'If children suffer
because of attitudes to non-standard dialects it is the attitudes that
should be changed and not the dialects'. Trudgill also suggests that
society is becoming more relaxed about accents, as witness a wider
range of accents among radio and television presenters, announcers,
etc., and sees this as a sign that our attitudes have relaxed, the im-
plication being that if attitudes to accent can change, so can attitudes
to dialect. But I query whether society's attitudes to accent have
changed. Macaulay's account of self-stigmatizing Glaswegian children
(Macaulay and Trevelyan 1975) still rings true. We now accept a wider
range of accents on television and radio because we now have a much
wider variety of programmes, and on the entertainment slots (the ma-
jority of the output), regional (not strong regional) and class (again not
very marked) accents are, as it were, licensed. But on the 'serious' news
and comment slots, the presenters (including the ethnic talking heads)
use modified RP or Scots.
Trudgill proposes slightly different treatments for various minority
groups. For working-class children, he says that society's attitudes must
change. For West Indian or Caribbean children, he suggests (1983:194)
that what would be helpful is a recognition, especially by teachers, that
some WI children in British schools may be faced with what is perhaps
best described as a semi-foreign language problem; that is, while they
have a problem, it is a problem not with Standard English but with
British English as a whole. This is an odd position to take in view of the
fact that most 'Caribbean' children now in our schools were born in the
UK. And for Scottish children, Trudgill says (1983:190-91) that 'while
social dialect continua ranging from local dialect to Standard English are
found in much of England, in lowland Scotland...there is discontinuity
because of the greater linguistic differences involved...Many Scottish
children...are well aware that they have one dialect for school and
another for other situations.' Trudgill concludes that while 'translation'
is possible from nonstandard to standard dialect, and children learn early
how to do it, the 'problem' must be a question of social attitudes. The
only reason, he claims, for teaching Standard English is that it is socially
advantageous, i.e. not linguistically or cognitively so. To dismiss linguis-
tic advantage in this way implies either that children do not need access
to the range of written material, most of which is in Standard English, or
that they will acquire Standard English anyway—in which case, what is
the fuss about? It isn't linguistic, says Trudgill, it's social attitudes; it's
a matter of giving prestige to the children's home speech, and not deni-
grating it explicitly or by implicitly denying it a school role. He is pre-
sumably (it is not clear) talking here only about spoken, not written
English. In the case of writing, there are, of course, two major prob-
lems. First, there are no accepted conventions, in most cases, of how to
write down dialect forms; second, in the absence of sophisticated
linguistic training, trying to distinguish between correct and incorrect
dialect forms would be an impossible burden on teachers. How would
one write, for example: mi asks di man fi put mi moni Una him pakit 'I
asked the man to put my money in his pocket' (Sutcliffe 1982)?
Trudgill is game enough to quote in his recent book some of the fierce
attacks made on him, such as: 'Some poor children already suffer from
progressive teachers anyway who think it wrong to make them read.
They are now threatened with a rash of Trudgills who won't correct their
grammar. Yet nothing could penalise the working-class more than to be
denied the right to knowledge' (Sunday Telegraph, 28 November 1975;
Trudgill 1983:198). Whatever Trudgill has actually written, however
subtle his qualifications, I think it understandable that he should be
popularly interpreted in this way.
Let me turn now to his antagonist, dohn Honey, who earlier in 1983
published Number 3 in the Black Paper series (Honey 1983); it is one of
the Kay-Shuttleworth Papers on Education, published by the National
Council for Educational Standards.
Honey's premises are to me rebarbative: he considers that standards
have declined, that civilisation is collapsing and the barbarians are at
the gate. It's the 'law and order' issue in language education. 'It is a
serious matter', he says, 'that our educational system...continues to turn
out...an annual crop of total illiterates' (p. 3). The responsibility, he
claims, lies not only with the schools, but with 'a group of specialists in
linguistics' (yes, you, Trudgill) propagating the notion 'that for schools to
foster one variety of English is contrary to the findings of the science of
linguistics' (p. 3). 'What we are dealing with', he tells us, is 'the theory
of functional optimism', which seems to be a way of representing the lin-
guistic view that all languages are equal. I have to say that Honey's
strictures against linguistics are often ill-informed. He uses as a clinch-
ing argument against the 'all languages are equal' position that you
cannot translate an academic paper on physics from English into Masai.
But nobody ever said you could! Such flatulence does not help his cause;
as I said earlier, he does have a cause, alas! Again, he shows his lack of
understanding on the same point when he remarks: 'we have not been
given any evidence that all languages or dialects have a grammatical
structure of equal complexity1 (p. 17). But what on earth does 'equal
complexity' mean? How would you show that English and French, say,
have equal complexity?
However, when he comes to his 'cause1, he has my attention (Honey
(1983:21-22):
For schools to foster non-standard varieties of English is to place
their pupils in a trap. To persuade such speakers that their par-
ticular non-standard variety of English is in no way inferior, no less
efficient for purposes of communication, but simply different, is to
play a cruel trick. All the evidence we have suggests that listeners
filter the messages they receive from utterances of other speakers
in accordance with perceptions of those speakers which are heavily
influenced by the standard or non-standard nature of the language
of the utterance in question. Quite apart from this filter mech-
anism, we know that the use of non-standard English has the power
to distract from the speaker's intended message. The 'adequacy for
236 / Alan Davies
communication' of any language or dialect is therefore, at least in

part, a function of those sociolinguistic factors which we have seen
dictating the acceptability of such a language or dialect in specific
situations, for specific functions, and specific audiences.
It is difficult, says Honey (correctly), to change attitudes to dialects.
Furthermore, the local dialect is not necessarily romantic; it can be
given a sentimental value which puts limits in advance on children's
'ability to express themselves...outside their immediate subculture, and
to slam the door on any real opportunity for social mobility' (pp. 2^-25).
And he blames linguists for their support of linguistic diversity (1983:27):
To sociolinguists [sicf] like me, for whom all the speech forms of
the dialects and other varieties of English in contemporary Britain
are a source of fascination and joy, it is hard to face the sad but
true fact that in a plural society the handicaps of disadvantaged
groups can be increased by promoting linguistic diversity, as they
can be reduced by fostering greater linguistic uniformity.
Honey favours a bidialectalism (or bilectalism) approach—a policy
'which is designed to foster the child's use of his own non-standard lan-
guage while giving him competence in a second dialect—standard Eng-
lish—whose use he will need for certain specific functions' (p. 30).
t
f. Conclusion: Bridging the gap? In this final section I want to ask
whether these two positions ((1) Standard English does not need to be
taught, even for writing (Trudgill), (2) Standard English only should be
recognised in schools (Honey)) are really as far apart as the polemic sug-
gests. Let me first quote Richards (1978:55-56):
The task of promoting Standard English is not the mammoth un-
dertaking it is sometimes made out to be. Nor need its practice
prevent the use of dialect in conversation where its adequacy for
conveying meaning is not in question...Such problems as do arise are
comparatively small in relation to those of countries in which the
dialects are so deviant from the standard that they almost con-
stitute other languages; or where within the country, more than one
language is in common use. Attempts to promote dialects as alter-
natives to Standard English can only exacerbate what problems
there are. To contemplate such action in an area as small as the
British Isles is altogether ludicrous.
And on the problem of alienation, Richards says: 'The danger lies in be-
lieving the problem to be solely educational and thus expecting a cure to
be effected through this agency alone' (ibid.).
Whether or not linguistics is normative (and Haas 1982:3 argues that it
always is: 'It is well known that all the descriptive linguistic disciplines
owe their origin to demands for their normative application'), sociolin-
guistics certainly studies language in normative institutions, and educa-
tion is in essence normative. Societies need standard languages in order
to function as societies; as social agencies, schools have to accept and
teach standard language if only because one of the primary purposes of
education is literacy. What gets lost in the language/code debate is,

paradoxically, the individual. Although society as a whole requires Stan-
dard English, individuals may well gain and grow through sympathy for,
and attention to, their home language variety—which in many cases is
not the spoken standard. (For children beginning school, of course, it is
never, or very rarely, the written standard.) Between an individual idio-
lect and the social standard there are certain dialects which are non-
standard. To promote these as a school policy is (and I agree with
Richards) ludicrous, an essentially divisive procedure; but to accept
them is not ludicrous. On the contrary, to attempt to eradicate spoken
dialect forms from individual speech (as against a bidialectal policy of
adding standard forms) is a vain and hopeless policy: it will not work.
The goal in schools must be decided educationally, not linguistically: it
must surely be to promote production and understanding of written
Standard English, and understanding of spoken Standard English; and it
must tolerate production, where appropriate, of non-Standard dialect
English in the spoken form. Milroy (1981:18), in a recent study of Bel-
fast, reminds us that 'the standard is actually an ideal, a norm which
speakers may have in mind...perhaps there is no speaker of English who
actually speaks the standard in all its levels (grammar, vocabulary,
phonology) at all times'. And he quotes normal code switching in Low-
land Scots:
33 to me - 'Dae ye ken ocht aboot thon stane abune the hoose...'

33 (later to my wife) - 'I was askin yer man (i.e. husband) if he knew
anything aboot that stone above the house?'
A recent discussion by Coates (1982) recognises that standardisation is

an attempt to solve and not to create a problem. Coates makes the ex-
cellent point (1982:41) that standardisation has an inevitability and that
'it symbolises integrative aspiration towards the social group whose norm
is represented by distributions of SPFs1. (SPFs are Standard Pronuncia-
tion Features.) 'For those involved in education', he says, 'standardisa-
tion can be seen as a litmus paper for a pupil's self-identification with
the demands of school and a willingness to meet them practically even
at the possible cost of distancing him or herself from the (language of
the) peer group.' He then helpfully distinguishes features 'for which
there is no shift within a given community from those for which certain
individuals display shift.1 An example of the first might be intervocalic
glottal stop and of the second, rhotacism (+ /r/). This is a most helpful
approach. 'My plea is', he says, 'that thought should be given before a
linguistic feature is condemned or corrected' (p. 47). Exactly! And as a
coda, 'One may applaud or regret the impact of standard accents as one's
conscience dictates; acknowledging that it is there, we should evaluate
in an understanding way the linguistic tensions that it gives rise to in
real people' (p. 47). Amen to that!
In summary, the argument is more about identity than about language.
Language teaching needs to steer wisely between Scylla and Charybdis.
If the child's or student's home dialect or language is ignored or scorned,
s/he may suffer alienation. If the home dialect or language is given
official status—as the medium of spoken interaction—there is the danger
of functional divergence and eventual diglossia for those bidialectal
238 / Alan Davies
groups for whom the written language, for example, may thus become
fossilised, static, not amenable to change, and therefore not something
with which they can identify.
Sociolinguistics idealises just as linguistics does, but less so. To put it
another way, sociolinguistics is a wider linguisitics, still idealising. One
aspect of that idealisation is the notion of standard language. My argu-
ment has been that sociolinguists must advise that for teaching the writ-
ten language, the standard dialect is the only choice. Anything else
would be trahison de clercs.
Thus, as students of language, linguists, sociolinguists, language
teachers, we may (no one can stop us) use our knowledge of language to
inform our judgements and our advice on language in education. But the
decision is really an educational one which has to do with balancing two
apparently conflicting values, those of society and those of the individ-
ual. Balance demands cost, on both sides, and it is the measure of a just
society that the cost should be one we are prepared to pay.
Notes
1. Current developments of teaching materials for 'language aware-
ness1 (Hawkins 1984) may be one way of creating that involvement.
2. Trudgill goes further, recommending that such nonstandard dialects
be used in education (to be usually interpreted as LI in English teaching).
3. This itself raises questions about generality since sociolinguistics is
presumably concerned to generalize across societies while, of course,
using data from particular societies.
References
Bauman, R., and J. Sherzer, eds. 1974. Explorations in the ethnography
of speaking. London: Cambridge University Press.
Bell, R. T. 1976. Sociolinguistics. London: Batsford.
Bereiter, C , and S. Engelmann. 1966. Teaching disadvantaged children
in the pre-school. Englewood Cliffs, N.J.: Prentice-Hall.
Bernstein, B. 1971. Class, codes and control, vol. 1. London: Rout-
ledge and Kegan Paul.
Coates, R. 1982. How standard is standard? In: Languages for life.
Edited by T. Pateman. Brighton: University of Sussex.
Chomsky, N. 1965. Aspects of the theory of syntax. Cambridge,
Mass.: The MIT Press.
Cummins, J. 1979. Linguistic interdependence and the educational de-
velopment of bilingual children. Review of Education Research 49.
Fishman, Joshua A. 1977. Standard versus dialect in bilingual educa-
tion: An old problem in a new context. Modern Language Journal
61.7. Reprinted in Trueba and Barnett-Mizrahi, eds. (1979).
Haas, W. 1982. Standard languages: Spoken and written. Manchester:
Manchester University Press.
Hawkins, E. 1984. Language awareness. London: Cambridge University
Press.
Honey, J. 1983. The language trap. Kay-Shuttleworth Papers on Educa-
tion, No. 3. Kenton, Middlesex: National Council for Education Stan-
dards.
Kjolseth, R. 1973. Bilingual education progress in the U.S.: For as-

similation or pluralism? In: Bilingualism in the South West. Edited by
J. A. Turner. Tucson: University of Arizona Press.
Labov, W. 1969. The logic of non-Standard English. In: Georgetown
University Round Table on Languages and Linguistics 1969. Edited by
James E. Alatis. Washington, D.C.: Georgetown University Press. 1-
43.
Labov, W. 1972. Sociolinguistic patterns. Philadelphia: University of
Pennsylvania Press.
Lewis, Saunders. 1962. The fate of the language (Tynged yr laith). BBC
radio speech of 13 February. Trans, by G. Aled Williams. In: Pre-
senting Saunders Lewis. Edited by A. R. Jones and G. Thomas. Car-
diff: University of Wales Press. 127-41.
Macaulay, R. K., and G. Trevelyan. 1975. Language, education and em-
ployment. Edinburgh: Edinburgh University Press.
Milroy, 3. 1981. Regional accents of English. Belfast: Blackstaff
Press.
Richards, J. 1978. Classroom language: What sort? London: Unwin.
Rosen, H., and T. Burgess. 1980. Languages and dialects of London
school children. London: Ward Lock.
Sutcliffe, D. 1982. British Black English. Oxford: Blackwell.
Tosi, A. 1982. Between the mother's dialect and English. In: Language
and learning in home and school. Edited by A. Davies. London:
Heinemann.
Trudgill, P. 1975. Accent, dialect and the school. London: Edward Ar-
nold.
Trudgill, P. 1983. On dialect. Oxford: Blackwell.
Trueba, H. T., and C. Barnett-Mizrahi, eds. 1979. Bilingual bicultural
education and the professional. Rowley, Mass.: Newbury House.
LANGUAGE AND THE LAW: REFERENCE, STRESS,
AND CONTEXT
Ellen F. Prince
1. On linguists as expert witnesses. The presence of linguists as
expert witnesses in courtroom proceedings has grown in the past few
years from a newsworthy rarity to a not uncommon event (see, for
example, Levi 1983). At the present time, however, it is still not as
commonplace, compared, for example, with the presence of physicians
or civil engineers as expert witnesses. The reason for this disparity is, I
believe, twofold.
First, and most obvious, it is probably the case that most lawyers
simply do not think of calling upon linguists, due to the fact that their
participation is still a relative novelty. Second, and more important, is
the fact that the domain of the linguist—language—is viewed by the
courts, both implicitly and explicitly, in a qualitatively different way
from those of the physician or engineer, and this difference militates
against the use of linguists as experts. The difference, as I see it, is
basically as follows. Domains such as medicine or engineering are seen,
both by the layman and by the courts, as clearly outside the ken of the
average individual; there is no sense, among any sector of the popula-
tion, that an ordinary person of average intelligence and education will
have reliable intuitions about whether, for example, the ingestion of a
certain chemical will have an adverse effect on a patient who has a
certain disease, or whether the substitution of one grade of concrete for
another in the construction of a building will impair the structural
soundness of that building. In contrast, the domain of language is, in a
manner of speaking, in everyone's ken; we do not need an advanced de-
gree to know what some English sentence means.
From a linguist's point of view, this position is on the surface unim-
peachable. After all, is it not that very competence that we are in the
business of studying? However, a consideration of some of the questions
that arise in legal proceedings reveals that a linguist's expertise is often
very much in order. First, certain questions call not for linguistic com-
petence so much as for what we call linguistic 'metacompetence': the
240
Language and the law / 241
knowledge that linguists have about linguistic competence. For exam-

ple, the competence of any normal speaker will enable him/her to assign
an interpretation to an utterance in a context, but it is a linguist's meta-
competence that is required to determine how the interpretation might
be affected as the context is varied, say, by positing different tacit
assumptions that the hearer may have. Second, in our society at least,
laymen, including the courts, are susceptible to holding certain folk
theories of language which may be brought to bear on their analyses,
along with their linguistic competence. Thus, in a sense, the situation is
not that laymen use linguistic competence while linguists use meta-
competence; rather, both make use of metacompetence, but it is only
the linguist's that is subject to scientific inquiry and testing.
In what follows, I am going to recount a situation in which I served as
an expert witness and which, I believe, illustrates the need for a lin-
guist's expertise both to show how an interpretation of a text depends on
its interaction with its context, and also to counter a deeply entrenched
folk theory held by society and incorporated into legal practice: namely,
the notion that the written language is the 'true' language and, conse-
quently, that a written transcript of an oral discourse is a perfectly
adequate representation of that discourse—if not the best representa-
tion.
2. The case. In February 1980, a United States District Court held

grand jury proceedings investigating police corruption in a state capital.
Among those testifying under oath was the former chief of police of the
city, henceforth the Chief. Two months before that, in December 1979,
the FBI had covertly taped a personal conversation between the Chief
and the then assistant chief of police, henceforth the Assistant Chief,
but this tape was not transcribed until November 1981, or 21 months
after the grand jury proceedings.
Three days later, on the basis of a comparison of the transcript of this
tape with a transcript of the grand jury proceedings, an indictment was
filed charging the Chief with having committed perjury during the grand
jury proceedings. At this point, I was asked by the defense lawyer to
look at the materials and see if there was anything that I, as a linguist,
could contribute to the case. The materials given to me were the tran-
script of the grand jury proceedings, a copy of the tape made of the per-
sonal conversation between the Chief and the Assistant Chief, and the
FBI transcript of that tape. I examined them and found out that, indeed,
there was much to be said from a linguistic point of view.
2.1 The problem. The utterance for which the Chief was indicted for
perjury is his response to the district attorney (D.A.) in (1).
(1) D.A.: And you are not familar with any other police officer
on your force accepting money from anyone?
Chief: I am not.
The motivation for this indictment was the revelation, in the transcript
of the covertly taped personal conversation, that the Chief had been told
by the Assistant Chief at that time that the latter had received some
money from someone, henceforth the Gambler. Therefore, the argument
2*2 / Ellen F. Prince
was that since the Chief knew that the Assistant Chief, a police officer
on his force, had accepted money from someone, his answer in (1) was
false.
I shall now present the analysis that I presented in court.
2.2 The analysis. Linguistic theory construes the problem of inter-

preting a linguistic utterance not as trying to discover what the sentence
'really means', but as discovering what understandings of the utterance
competent and cooperative language-users may predictably construct,
following the various linguistic principles and strategies that they have
(unconsciously) at their disposal. The basis for this point of view is that
a sentence in vacuo may 'mean' very little, and the interpretation of an
utterance, i.e. a sentence at some time, in some place, by someone, to
someone, varies with the linguistic and situational context in which it
occurs and with the beliefs, assumptions, and hypotheses of the inter-
preter. Fortunately, this variation is neither arbitrary nor idiosyncratic:
language-users operate in a rule-governed way, employing various con-
ventional principles and strategies as^ they go about constructing an
understanding of a linguistic utterance.
Let us now consider the question shown in (1), the crucial question in
this case. Out of context, this question could presumably be answered in
a variety of ways. Consider, for example, the possible answers in (2).
(2a) Yes, Joe Smith received $25 as a wedding gift from his aunt
in Peoria. (Joe Smith is a member of the police force.)
(2b) Yes, my wife and I gave our son $50 for his 25th birthday.
(2c) Yes, Sam Jones won $5 in the instant lottery.
But, presumably, if one could have answered (2a), (2b), or (2c), and gave
instead the answer in (1), that alone would not be considered to consti-
tute perjury. The reason is simply that the answerer must construct an
understanding of the question that is appropriate to (his understanding
of) the context. The prior context, in this case, consists of two ques-
tions and the answers given to them; thus we have a subtext consisting
of three question-answer pairs.
(3) Ql: Now, all the names that I have just read to you, I am
going to ask one more time. Are you familiar with any
of those individuals ever making any payments to any
police officers on the police force, either while you
were the Captain of Detectives or the Chief of Police?
Al: No, sir.
Q2: Have any of those individuals ever paid you any money?
A2: No, sir.
Q3: And you are not familiar with any other member on your
force accepting money from anyone?
A3: I am not.
Let us now analyze the prior context to see how it influences the under-
standing of Q3 that a cooperative answerer would presumably construct.
Language and the law / 2*3
2.2.1 Question 1. Question 1 explicitly evokes a set of previously
named individuals—those local gamblers whose names have just been
read to the defendant—and asks whether any member of this set made
any payments to any police officers during a certain time interval.
The only serious potential for difference in understanding here, given
a single, unambiguous prior list of names, concerns the interpretation of
make any payments and any police officers. Making a payment, or
paying, is conventionally understood not as a synonym for giving money,
but as a special case of it. That is, one makes a payment for some-
thing. On the other hand, it is conceivable that one may construe the
giving of money to a policeman by a nonintimate in his district as some-
thing approaching a payment, even if there is no explicit discussion of
goods or services rendered, just so long as the potential for such a ren-
dering is salient.
The potential for difference in understanding any police officer re-
lates to whether the Chief understands himself to be included in the set
of police officers evoked. On the one hand, he might construe that he is
so included, in accordance with the definition of police officer and the
fact that he fits that definition. On the other hand, he might invoke the
linguistic convention of referring which says that speakers should refer
to coparticipants (i.e. speaker and addressees) explicitly if they mean to
refer to them at all. Thus it is bizarre and misleading, though not false,
for A to say to B the sentence in (4) if someone is meant to refer to A or
B.
(4) A: Someone has been spending a lot of time with your wife.
That is, if the someone is A, he has not lied; but, by not saying J, he has
misled B into thinking 'not-A.1 Likewise, if the someone is B, A still has
not lied, but, by not saying you, he has misled B into thinking 'not-B.1
(The situation is actually more complex: even if someone is a third
party, if A thinks B knows that party, A is conventionally obligated to
indicate who it is, and the lack of such an indication leads B to infer that
he does not know the individual in question.) With respect to Ql, then,
the Chief may construe any police officer to include him, following a
technical, definitional understanding; or, he may construe any police
officer to exclude him, following the linguistic conventions of referring.
Thus, there are four possible understandings of Q l .
(5) Ql: PAY
+Service... -Service...
ANY +Chief Qla Qlb
POLICE
OFFICER -Chief Qlc Qld
(6) Ql: During a particular time interval T, has any of a pre-

viously mentioned set of individuals I given money...
Qla. ...to you or any other police officer in exchange for
something?
Qlb. ...to you or any other police officer?
Qlc. ...to any police officer other than you in exchange for
something?
/ Ellen F. Prince
Qld. ...to any police officer other than you?
As will become relevant later, the Gambler from whom the Assistant
Chief claimed to have taken money was not among the set I, and the
time at which the Assistant Chief claimed to have taken the money was
not during T.
2.2.2 Question 2. Now consider Q2. Have any of those individuals

ever paid you any money? It is a very basic, perhaps universal, conven-
tion of language that one does not ask for information one already has.
This convention is known as the Gricean Maxim of Quantity (Grice 1975).
This is not to say that one cannot do so: it simply says that, without any
explicit reason for thinking the contrary, a cooperative answerer as-
sumes that new information is being requested when one is asking
him/her a question. Thus the understanding of Q2 will presumably vary
according to how the answerer understood Q l . Put simply, the answerer
will try to find some difference between Ql and Q2. Since the same
morpheme, pay, is used in both questions, it is highly unlikely that a
difference will be found to lie in the sense of that term. That is, how-
ever pay was understood in Ql, it will presumably be understood the
same way in Q2. Likewise, the set of individuals (I) referred to in Ql is
explicitly referred to again in Q2 (any of these individuals) and cannot
account for the difference being sought.
One possible difference lies in the possible understanding of ever,
especially for individuals who had chosen Qla or Qlb. However, how the
word was actually stressed by the speaker is relevant, if not crucial. If
ever is phonologically destressed in Q2, it would normally be understood
as anaphoric, i.e. as referring back to something already mentioned, in
this case presumably the time interval T referred to in Q l . If, on the
other hand, ever is phonologically stressed, it would generally be under-
stood as nonanaphoric and would presumably be understood as 'at any
time at all1 (see Ladd 1978). Consider examples (7) and (8), where A is a
job interviewer and B is a job applicant.
(7a) A: What was your grade point average in college?

(7b) B: 3.3, a B+.
(7c) A: Did you ever get an F?
(7d) B: No.
(7e) A: Did you ever make Dean's List?
(7f) B: No.
(7g) A: Did you EVER receive any scholastic distinction?
(7h) B: Yes, in high school I was class valedictorian.
(8a) A: What was your last job?
(8b) B: I was a piano teacher at the Powell School.
(8c) A: Did your students come for lessons one at a time or did
you ever teach a large class?
(8d) B: I taught them just one at a time.
(8e) A: Did you EVER teach a large group?
(8f) B: Yes, in 1975 I taught music appreciation at the Lee
School. There were about 30 students in each class.
6 / Ellen F. Prince Language and the law / 24.5
Note that ever in (7c) and (7e) is destressed and is normally understood
as being anaphoric to the time interval previously mentioned: in college.
That is, CG would not be taken to have lied in (7d) if he had, in fact,
received an F in high school. In (7g), however, ever is stressed, the
stress marking its nonanaphoricity, and the high school episode men-
tioned in (7h) becomes relevant. Likewise, in (8c), destressed ever is
taken to refer to the time interval previously evoked: last job, while in
(8e) the stressing of EVER releases it from anaphoricity and induces the
understanding 'at any time at all1.
Thus, depending on how Ql was understood and on how ever is under-
stood—which in turn depends to a very large extent on how it was ut-
tered—we find the following four plausible understandings of Q2.
(9) Q2: Has any individual in I given money to you...

Q2a: ...in exchange for something during T?
Q2b: ...in exchange for something at any time?
Q2c: ...during T?
Q2d: ...at any time?
(See the Appendix for a list of the possible understandings.) Note that
two routes are redundant, i.e. Qla —> Q2a and Qlb —} Q2c, in that no
new information is requested in Q2. Conceivably, such redundancy
might prompt a hearer to reject these understandings of Q2 for those in
which ever is nonanaphoric, even if it is destressed; however, little is
known about how language-users deal with such infelicities. Another,
perhaps more plausible, possibility is that a hearer, arriving at the
redundant understandings of Q2a and Q2c, might remove the redundancy
by backtracking and reanalyzing Qla or Qlb as Qlc or Qld, respec-
tively. It should be noted that such a reanalysis would not affect the
truth-value of the answer given to Ql and so, presumably, would not be
mentioned. Thus, there is a slight statistical edge in the likelihood of
the understandings of Qlc or Qld being chosen. In any event, if the
hearer had chosen, immediately or via reanalysis, Qlc or Qld, and if
ever is destressed in Q2, then a choice of Q2a or Q2c is more likely than
a choice of Q2b or Q2d. That is, there is no reason to change the under-
standing of ever so long as there is something new asked for. Thus,
assuming ever is destressed, Q2a and Q2c are the most likely under-
standings of Q2.
If, on the other hand, ever is stressed, the situation is quite different:
Q2b and Q2d are highly favored.
2.2.3 Question 3. Finally, Question 3 is highly complex: Now, all the
names that I have just read to you, I am going to ask one more time.
Are you familiar with any of those individuals ever making any payments
to any police officers on the police force, either while you were Captain
of Detectives or the Chief of Police? First, there is the matter of the
change from pay to accept. This may reflect merely a change in point
of view, e.g. give/take, in which case accept is understood as the inverse
of pay. On the other hand, for those who had the more restricted under-
standing of pay (i.e. Qla, c; Q2a, b), accept may be broader or it may
not.
Second, there is the matter of the understanding of any other police

officer. This is a necessarily anaphoric term, in that other indicates
that the individual in question is different from some previously evoked
individual or set of individuals. There are two possible candidates for
that previously evoked individual or set thereof: either you or any police
officers on the police force (during T) in Ql. (Note that, since the latter
has two possible understandings, there are in fact three candidates for
that previously evoked individual or set thereof. Due to Q2 and to the
Referring Principle mentioned earlier, however, one of these—the one
where any police officers on the police force is understood as excluding
the Chief, leaving any other police officer to refer to the Chief—is
extraordinarily unlikely.) In any event, any other police officer can be
plausibly understood as referring either to a police officer other than the
Chief, or to a police officer other than those on the police force during
T.
Third, and perhaps the most crucial, is the understanding of anyone.
Here again there is a problem of stress. If anyone is destressed, its most
plausible understanding is anaphoric, i.e. anyone among the previously
evoked set I. If it is stressed, however, it is not likely to be construed as
anaphoric and then would include anyone at all. Consider examples (10)
and (11).
(10) A: Have you hired someone yet?

B: I've interviewed 12 people but I don't like anyone.
A: I think you don't like ANYONE.
(11) A: How was your high school reunion?
B: Oh, lots of people showed up, but I didn't remember any-
one.
A: Good grief! If you don't remember your old classmates,
you won't remember ANYONE!
Note that destressed anyone in (10B) is anaphoric to the set of 12 people
interviewed, while the stressed ANYONE in A's response is nonana-
phoric, understood as 'anyone at all'. Likewise, destressed anyone in
(113) is anaphoric to the set of people who showed up, while stressed
anyone in A's response loses its anaphoricity, being understood as 'anyone
at all'. Notice how bizarrely redundant A's response would be in both
(10) and (11) if anyone were destressed.
Thus there are three binary variables in Q3, leading to eight possible
understandings. Furthermore, as no new time interval is stated, one
would infer that the same time interval that held in Q2 still holds. But
Q2 could be understood as involving one of two different time intervals,
T and any time at all. Thus Q3 has 16 possible understandings, shown in
(12). (See also Appendix.)
(12) Q3A: Has, during time interval T,...

Q3Aa: ...any I given any police officer other than you money
in exchange for something?
Q3Ab: ...any I given any police officer other than you
money?
Q3Ac: ...any I given any police officer not on the force
during T money in exchange for something?
Q3Ad: ...any I given any police officer not on the force
during T money?
Q3Ae: ...anyone at all given any police officer other than
you money in exchange for something?
Q3Af: ...anyone at all given any police officer other than
you money?
Q3Ag: ...anyone at all given any police officer not on the
force during T money in exchange for something?
Q3Ah: ...anyone at all given any police officer not on the
force during T money?
Q3B: Has, at any time at all,...
Q3Ba: ...any I given any police officer other than you money
Q3Bb: ...any I given any police officer other than you
money?
Q3Bc: ...any I given any police officer not on the force
during T money in exchange for something?
Q3Bd: ...any I given any police officer not on the force
during T money?
Q3Be: ...anyone at all given any police officer other than
you money in exchange for something?
Q3Bf: ...anyone at all given any police officer other than
you money?
Q3Bg: ...anyone at all given any police officer not on the
force during T money in exchange for something?
Q3Bh: ...anyone at all given any police officer not on the
force during T money?
Note once again that these are all merely theoretical possibilities, if
we ignore stress. If EVER was stressed in Q2, then Q3B is favored over
Q3A; if ever was destressed in Q2, then Q2A is favored. Likewise, if
ANYONE was stressed in Q3, then Q3Ae-h and Q3Be-h are favored; if
anyone was destressed, then Q3Aa-d and Q3Ba-d are favored. The inter-
action of these two variables is shown in (13).
(13) Favored understandings of Q3, depending on stress:

ever EVER
anyone Q3Aa-d Q3Ba-d
ANYONE Q3Ae-h Q3Be-h
In any event, we have found 4 plausible understandings for Ql, 4 for Q2,
and 8 for Q3, allowing for 128 theoretically possible routes, some of
which would likely be favored by a hearer seeking nonredundancy, and
some favored according to how certain items were stressed.
2.3 Understanding of take. In order to determine whether the Chief's

response to Q3 was consistent with his beliefs, one must, among other
things, make some hypothesis about what his beliefs in fact were. In this
respect, part of his conversation with the Assistant Chief, covertly
taped two months before the grand jury proceedings, seems to shed light
on his beliefs about the Assistant Chief's taking of money.
At the trial, the judge did not permit me to present this section,
stating that the personal conversation was not 'relevant' to the perjury
issue—in spite of the fact that it was this conversation that led to the
perjury indictment. However, as it was included in my report to the
lawyer, which report the judge read, I include it here.
Following is a corrected transcript of the pertinent part of the con-
versation, where C = Chief and A = Assistant Chief.
(1*) C: What the fucking hell's the MATTER with you?

A: I don't know. Whew...I don't know. I don't know what they
got. (Tape cut) I didn't want to say nothing yesterday, I
didn't want to upset you. So when X says—I called X
last night: so he called me early this morning: and he
said, Your phone must be tapped: and they're watching
your place. He said, when he came out, they—he had
somebody in the car with him: and he seened somebody
pull in: and uh I don't know whether he said HE followed
THEM or THEY followed him when he left there. I don't
know.
C: They can't have anything.
A: I don't know.
C: What the hell can THEY have? You never took anything.
A: I did, yeah.
C: From what?
A: From Y up there—
C: Yeah, but you went to the track with that money.
A: Well, yeah, but you know, I used to stop in there and—
C: That's right but—
A: He'd give me: like ten dollars, twenty dollars, something
like that usually—
C: So, what the hell?
A: a week or a couple of weeks. There was never really a n y -
serious uh—
C: You never did him any favors.
A: No.
The crucial line here is You never took anything. If one assumes that
one's statements in private conversation reflect one's beliefs—an as-
sumption which presumably underlies the government's covert taping of
private conversations such as this one—one must infer that, at this point
in time, the Chief believed that the Assistant Chief had not taken any
money.
The next question is whether there is evidence that he changed his
belief. As it appears here, there is none. First, the transfer of money in
question appears to be one that he already knew about, since he volun-
teers that the Assistant Chief 'went to the track with that money.' Thus
he either lied when he said You never took anything or else he did not
consider this transfer of money to the Assistant Chief to constitute
taking. Second, and equally important, is the statement You never did
him any favors. By a very basic principle of linguistic interaction, the
Gricean Maxim of Relation (Grice 1975), utterances are intended to be
relevant to their context. Thus one infers that this statement is

intended to be relevant to the issue at hand, i.e. whether the Assistant
Chief took money from Y. Since the Chief has been arguing that he did
not, the relevance of this statement appears to be as evidence for that
position. Taking money, from the Chief's point of view as it is repre-
sented here, crucially involves the notion of receiving money in ex-
change for something.
2A Finding. It is my understanding of the case that the Chief, in

order to have perjured himself by answering Q3, had to have understood
Q3 as Q3Bf: ...anyone at all given any police officer other than you
money?, rather than as one of the other 15 possible understandings.
That is, I am assuming he believed the following: Assistant Chief, a
member of the police force during T, was given money at some time not
in T, by someone not in I. If both ever and anyone were stressed, these
are plausible understandings. However, if neither or only one were
stressed, other understandings are more plausible.
3.0 Discussion. As I have tried to show, a finding of perjury in this

case hinged largely on how two words uttered by the D.A. during the
grand jury proceedings were stressed. Interestingly, the proceedings
had, in fact, been taped and so the answer to this question should have
been available. However, as I was told, such proceedings are taped only
to ensure that a transcript can be made; once the transcript is finished,
the tapes are destroyed, the transcript serving as the official record of
the event. I should add that there are no fine-tuned legal conventions
for indicating stress or intonation in the transcripts.
A courtroom anecdote may be relevant here. . In the course of my
testimony, I read the examples shown in (7)-(8) and (10)-(ll) and, notic-
ing a distraught look on the court stenographer's face, I said jokingly to
her, 'Hope you can handle this.' As soon as court was adjourned, and
even before the jury had left the jury box, the stenographer jumped up
and said to me something along the lines of: 'Oh, it's impossible to
transcribe that! They tell you just to punctuate it as you hear it, but it's
not like punctuation.' The stenographer had obviously learned what
linguists know but what the courts have a real problem with: written
language and oral language are different, and the former cannot be
taken to be simply a representation of the latter.
In conclusion, the Chief, who was, I am sorry to say, a most distasteful
character, was acquitted after a very brief deliberation. Given that the
only other witness for the defense was a clergyman testifying to his
philanthropy, I infer that my testimony played an important role in the
acquittal. If the courts could adopt a more enlightened attitude about
language, perhaps by consulting linguists when deciding on general
practices such as keeping records of speech events, it is possible either
that such a trial would never take place (with a large saving to the
taxpayer) or that it would lead to a conviction. But I guess we have
progressed if a linguist's testimony can prevent a conviction where the
evidence is missing.
250 / Ellen F. Prince
Appendix
Possible understandings of Ql-3

Ql: During a particular time interval T, has any of a previously men-
tioned set of individuals I given money...
Qla: ...to you or any other police officer in exchange for something?
Qlb: ...to you or any other police officer?
Qlc: ...to any police officer other than you in exchange for something?
Qld: ...to any police officer other than you?
Q2: Has any individual in I given money to you...

Q2a: ...in exchange for something during T?
Q2b: ...in exchange for something at any time?
Q2c: ...during T?
Q2d: ...at any time?
Q3A: Has, during time interval T,...

Q3Aa: ...any I given any police officer other than you money in exchange
for something?
Q3Ab: ...any I given any police officer other than you money?
Q3Ac: ...any I given any police officer not on the force during T money
Q3Ad: ...any I given any police officer not on the force during T money?
Q3Ae: ...anyone at all given any police officer other than you money in
exchange for something?
Q3Af: ...anyone at all given any police officer other than you money?
Q3Ag: ...anyone at all given any police officer not on the force during T
money in exchange for something?
Q3Ah: ...anyone at all given any police officer not on the force during T
money?
Q3B: Has, at any time at all,...
Q3Ba: ...any I given any police officer other than you money in exchange
for something?
Q3Bb: ...any I given any police officer other than you money?
Q3Bc: ...any I given any police officer not on the force during T money
Q3Bd: ...any I given any police officer not on the force during T money?
Q3Be: ...anyone at all given any police officer other than you money in
exchange for something?
Q3Bf: ...anyone at all given any police officer other than you money?
Q3Bg: ...anyone at all given any police officer not on the force during T
money in exchange for something?
Q3Bh: ...anyone at all given any police officer not on the force during T
money?
Notes
I should like to thank 3oshua D. Lock, Esq. for his confidence and co-
operation, without which this work would most certainly never have been
done. I am grateful also to Tony Kroch and Debby Schiffrin for their
help.
1. As was pointed out to the court, the term 'cooperative' is a tech-
nical one following the Gricean model, indicating that the individuals are
engaged in rational, goal-directed behavior and are assuming the same of
their coparticipants. It does not, of course, suggest that they are being
friendly, nice, or affable in any way.
2. See, among others, Horn (1972), Bolinger (1977), Clark and Havi-
land (1977), Kuno (1978), Morgan (1978), Gazdar (1979), Green (1980).
3. Webster's Third International Dictionary defines payment and pay
as follows:
payment: 1. the act of paying or giving compensation : the
discharge of a debt or obligation. 2. something that
is paid : something given to discharge a debt or
obligation or to fulfill a promise. 3. archaic:...
pay: v.i.: 1. to give a recompense : make payment : discharge
a debt or obligation. 2. to make suitable return for
expense or trouble : be worth the effort or pains : be
profitable. 3. to be amiss or afoot.
v.t.: 1. (obsolete)... 2a. to satisfy (someone) for services
rendered or property delivered : discharge an obliga-
tion to : make due return to. b. to engage for
money : HIRE. 3a. to give in return for goods or
service. b. to discharge indebtedness for :
SETTLE, c. to assume the charge of. d. to make
any agreed disposal or transfer of (money). 4. to
give or forfeit in expiation or retribution. 5a. to
make compensation for : make up for : RECOM-
PENSE, b. to make retaliation for—usually with
back, c. to requite (someone) according to what is
deserved : get even with—usu. with back, d. (ar-
chaic)... 6. to give, offer, or make freely or as fit-
ting. 7a. to return value or profit to. b. to bring in
as a return. 8. to slacken (as a rope)...
4. Note, of course, that this is simply a probability of likelihood and
not a statement about what is possible or which one is 'better'.
5. Working with a Sanyo transcriber, I found an average of 12 substan-
tive mistakes per typewritten page in the FBI transcript of the conversa-
tion. This was analogous to the situation I had found in two previous
cases where FBI transcripts of private conversations were used. More-
over, in at least one instance in the present case, a mistake introduced a
sense of culpability that the spoken text did not contain. That is, the
Chief was recounting to the Assistant Chief that all the local gangsters
had been calling him since he came home from the hospital; one, in fact,
had said, 'Do you need anything?', to which he had responded, 'No, I don't
need nothing.' This was transcribed as: 'Do you know anything?', 'No, I
don't know nothing.' The conspiratorial flavor of the transcript version
was no doubt responsible for the question-answer being quoted in the
local newspaper during the trial. It should be noted, of course, that
tapes of private conversations are generally of very poor quality; the
jury is, therefore, generally given a transcript to read along and it is
highly unlikely that they will detect any errors in it. Thus there is a
grave responsibility to provide as accurate a transcript as possible, a
responsibility which, in my opinion, the government has not appreciated.
252 / Ellen F. Prince
6. Of course, the notion that private social conversation is for the

purpose of transmitting objective information and, therefore, that the
speakers are committed to the truth of the propositions they are ex-
pressing, is not supported by studies of social conversation. See, for
example, Goffman (1967), among others. However, it is a fact that
utterances made in covertly taped private social conversations have
figured prominently in Federal court cases as evidence of the utterer's
beliefs and knowledge, analogous to utterances made under oath. The
seriousness of this problem cannot, in my opinion, be overestimated.
References
Bolinger, D. 1977. Meaning and form. London: Longmans.
Clark, H., and S. Haviland. 1977. Comprehension and the given-new
contract. In: Discourse production and comprehension. Edited by R.
Freedle. Hillsdale, N.3.: Erlbaum. 1-40.
Gazdar, G. 1979. Pragmatics: Implicature, presupposition, and logical
form. New York: Academic Press.
Goffman, E. 1967. Interaction ritual. Garden City, N.Y.: Anchor
Books.
Green, G. 1980. Linguistics and the pragmatics of language use: What
you know when you know a language...and what else you know. Center
for the Study of Reading. Urbana: University of Illinois.
Grice, H. P. 1975. Logic and conversation. In: Syntax and semantics,
vol. 3. Speech acts. Edited by P. Cole and 3. L. Morgan. New York:
Horn, L. 1972. On the semantic properties of logical operators in
English. Ph.D. dissertation, University of California, Los Angeles.
Kuno, S. 1978. Generative discourse analysis in America. In: Current
trends in textlinguistics. Edited by W. Dressier. Berlin and New York:
de Gruyter. 275-94.
Ladd, D. R. 1980. The structure of intonational meaning. Bloomington:
Indiana University Press.
Levi, 3. 1983. Language and the law: A bibliography. Bloomington:
Indiana University Linguistic Club.
Morgan, 3. L. 1978. Toward a rational model of discourse comprehen-
sion. In: TINLAP-2. Edited by D. Waltz. New York: Association for
Computing Machinery.
PERSONAL, GROUP, AND COUPLE IDENTITIES:
TOWARDS A RELATIONAL CONTEXT FOR THE STUDY
OF LANGUAGE ATTITUDES AND LINGUISTIC FORMS
Howard Giles
University of Bristol
Mary Anne Fitzpatrick
University of Wisconsin-Madison
It is a central argument of this paper that although scholars have
explored both personal and group identities as salient contextual para-
meters of language attitudes and behaviors, a crucial point on the indi-
vidual-group continuum has been neglected. While it is true individuals
have relatively enduring personal and group identities, we also know that
individuals have dyadic and/or couple identities as well. Just as individ-
uals define themselves and are defined by others as upper class and
female or Welsh and Catholic, so too they define themselves as husband,
lover, and girlfriend. Such relational identities, while of some import
for social actors, are largely ignored by language scientists. To begin
our exploration of dyadic identities, we have chosen to focus on the
couple, particularly the marital dyad. Marriage is an important 'nomos-
building1 institution in our society (Berger and Kellner 1975) because it
creates for people the sort of order in which they can experience their
worlds as making sense. It is through conversations with one another
that married partners construct a shared reality and subsequently define
their identities. Through these marital conversations, a couple identity
is not only built but also kept in a state of repair and ongoingly refur-
bished. As Berger and Kellner (1975:226) so eloquently phrase it:
The nomic instrumentality of marriage is concretized over and over

again, from bed to breakfast table, as the partners carry on the
endless conversation that feeds on nearly all they individually or
jointly experience.
To place our emerging interest in couple identities within the frame-

work of our past work, we will first survey some of the research
253
254 / Howard Giles and Mary Anne Fitzpatrick
conducted on the relationships between social meanings, particularly

personal and group identities, and various linguistic forms.
Personal identities. Within the study of personal identities and lan-

guage, a ubiquitous finding has been that listeners rate standard ac-
cented speakers on audiotape more favorably across competence (or
status-related) traits than their nonstandard counterparts (Ryan and
Giles 1982; Giles and Edwards 1983). Recently, and in line with other
workers (e.g. Romaine 1980; Ryan and Bulik 1982), we have been exam-
ining some contextual riders to this, and other, effects. For instance, in
one study, we found that the typical status upgrading of British standard
RP (Received Pronunciation) was attentuated significantly when the
sociophysical setting for the investigation was an evening youth club
rather than the usual, more formal, classroom setting (Creber and Giles
1983). Perhaps more surprising (see, however, Myers and Lamm 1976;
Taylor and Royer 1980) was another of our findings that the status
connotations of RP were accentuated when informants were asked to
discuss their speaker evaluations with each other for approximately 90
seconds before making their ratings (Giles, Harrison, Creber, Smith, and
Freeman 1983). In a further study, we found a tendency for Welsh
bilingual respondents to make evaluative distinctions between standard
RP English and nonstandard Welsh-accented English on status traits
when the experimental procedure was conducted in English, but not when
it was in the Welsh language. When the rating task was in the latter
language, however, listeners would make social attractiveness compari-
sons between standard and nonstandard voices that would not be appar-
ent in an English judgment setting (Price, Fluck, and Giles 1983). Hence,
not only can the status connotations of a standard variety be diminished
or exaggerated, depending on the nature of the context, but the evalu-
ative criteria brought to bear in them can also vary.
Another ubiquitous finding in the language attitudes domain has been
the positive linear relationship between speech rate and perceived
competence (e.g. Brown 1980; Stewart and Ryan 1982; Street, Brady, and
Putnam 1983); that is, the faster you talk, the more competent you
sound. The effect was obliterated, however, when the rating task was
moved out of its social vacuum. In this instance, allowing listeners
access to information that a speaker was taped while helping a naive
audience comprehend an unfamiliar topic, made him sound just as intel-
ligent and competent as when he was heard to talk much faster (Brown,
Giles, and Thakerar 1983).
But not only does contextual knowledge affect one's evaluations of
linguistic forms; it can also influence people's perceptions of those very
same features. Guided in part by Snyder (1981), we found that telling
people the speaker they had just listened to on tape was a high status
person induced them to rate him as having spoken at a faster rate and
with a more standard accent than those provided with no such informa-
tion who, in turn, rated him more positively on these linguistic dimen-
sions than did a third group provided with low status information
(Thakerar and Giles 1981). This phenomenon, which we subsequently
labelled the 'retrospective speech halo effect' (Ball, Byrne, Giles,
Berechree, Griffiths, Macdonald and McKendrick 1982), suggests that
what we think someone should sound like can sometimes be more
Personal, group, and couple identities / 255
influential in determining our linguistic judgments about them than what
can be measured objectively as such.
Furthermore, this rather robust phenomenon, which we recently
replicated in Australia (Ball et al. 1982), has implications beyond mere
cognitions to actual speech behaviors. In this respect, another vein of
our research has been concerned with showing how and why individuals
converge their linguistic and vocal forms toward each other in coopera-
tive or integrative frames of mind, or when they desire each other's
approval (see, for example, Street and Giles 1982; Giles 1984). There is,
of course, no problem here if there is isomorphism between where peo-
ple's speech styles actually are objectively, and where others believe
them to be subjectively. Recently, we found that people converge
toward where they believe others to be linguistically rather than where
they are in physical reality (Thakerar, Giles, and Cheshire 1982). In this
study, members of dyads who had each agreed about the relative sta-
tuses of their partners were required to talk about a problem task they
had just solved together. At the commencement of their discussion,
they were judged to possess highly similar accents and speech rates.
During the course of their conversations, however, lower status partners
in the dyads standardized their accents and talked faster while the
higher status speakers nonstandardized their accents and talked slower.
Although, in objective linguistic terms, this might seem like 'divergence',
in actual fact, data were gathered indicating that speakers believed they
were moving toward each other linguistically. This process of so-called
'linguistic divergence but psychological convergence' (Thakerar et al.
1982) has been noted independently (albeit in different conceptual terms)
in a number of cultural settings (see Beebe 1981; Bell 1982; Caporael
1981). These mutual apparent divergences could be potentially the
interpretive breeding ground for communication misattributions and
misunderstandings (see Giles and Bourhis 1976; Platt and Weber 1984).
In sum, then, we do not just passively listen to someone's voice; we
actively construct and perhaps reconstruct our impressions of it accord-
ing to incoming contextual data (cf. Street and Hopper 1982; Roloff and
Berger 1982) and in ways that have implications for our own linguistic
forms. Yet who is not there can be just as, if not more, important a
vocal and linguistic determinant of one's self-presentations as those
physically present. On occasions, many more significant others get to
hear about a message's meaning and form than ever heard the message
at the time of delivery. Hence, speakers can accommodate their speech
styles strategically to absent influential people who are known to pluck
the grapevine of highly active social networks.
Group identities. Given our disciplinary backgrounds, we have been

fascinated, as the foregoing attests, by the cognitive mechanisms that
mediate language attitudes and linguistic forms. For instance, we have
long recognized the linguistic implications of there being no necessary
overlap between investigators' allocations of respondents to ethnic group
X and the latter's feelings of group belongingness on that occasion, nor
any inevitable direct correspondence between assigning speakers to
context Y and the informants' subjective assessment of that situation.
Hence, taping a very elderly person's linguistic forms in a supposedly
formal context can be less predictive of his or her language behaviors if
this person feels middle aged and construes the social atmosphere as
very relaxed and informal. Obviously, sociolinguists such as Fishman
(1966), Hymes (1967), Labov (1966), and Gumperz (1982) have articulated
this actor's-eye-view for some time, particularly as it can relate to
social identity. Yet, such a cognitive stance has not been realized in any
really extensive and sophisticated manner in either methodologies or
theories. Whilst recognizing the potential, conceptual sterility of arti-
ficially dichotomizing language and context, we attempted recently to
fill these voids by proposing models of how linguistic forms can be
mediated by participants 1 cognitive representations of the social situa-
tions they are in and their group identities at that time. We also dis-
cussed some of the ways in which language behaviors in social inter-
action can also, reciprocally, mould interlocutors' cognitive representa-
tions of the context and their group identities (Giles and Hewstone
1982). This was achieved by recourse to work in the social psychologies
of social situations on the one hand (Wish 1978; Forgas 1983) and inter-
group relations on the other (Tajfel and Turner 1979). Space precludes
any real discussion of these models, but we would like to highlight one of
these mediating cognitive constructs, viz. the interindividual inter-
group continuum.
Tajfel and Turner (1979) argued that whilst on many occasions our
interactions are fully determined by our personal attributes (the so-
called 'interindividual' pole), on other occasions social interaction can be
almost exclusively dependent on our social group memberships, be they
academic discipline, gender, age, class, etc.; this is the 'intergroup' pole.
Studies referred to in the previous section can be typified at the inter-
individual pole where personal identities were our focus. Let us now
expend some time talking about language forms at the intergroup pole
where our group identities assume contextual salience and de-individua-
tion can occur (Turner 1982). Precisely how much of our everyday social
behavior is located near this extreme pole is, of course, an empirical
question (see Stephenson 1981). Nevertheless, perhaps much of what is
included under the generic rubric of 'interpersonal communication' (e.g.
male-female interactions, young-old encounters) could arguably and
profitably be reinterpreted in intergroup terms.
'Ethnolinguistic identity theory' is our current theoretical approach for
studying language behaviors at the intergroup pole (Giles and Johnson
1981), the core of which depends heavily on Tajfelian principles of social
identity (Tajfel 1978, 1982). One important proposition of our perspec-
tive is that if a speaker experiences a strong sense of group belonging-
ness whilst subjectively defining an interaction as an 'intergroup' one, he
or she will wish to achieve a positive group identity in that very context.
The realization of the positive-negative affect associated with group
identity comes through making 'intergroup 1 comparisons between the
position of your own group and that represented by an outgroup speaker
on valued dimensions such as relative power, resources, capabilities,
etc. Much of ethnolinguistic identity theory and the empirical research
attending it has been concerned with articulating the conditions neces-
sary, and strategies used, for achieving a positive identity. Given that
language forms can be important dimensions of social group member-
ships, particularly class and ethnic ones (Giles 1977; Giles and Saint-
Jacques 1979), one strategy for achieving such a positive group identity
would involve linguistic differentiation from outgroup speakers. This

process of so-called 'psycholinguistic distinctiveness' (Giles, Bourhis, and
Taylor 1977) can be manifest by language code-switching, emphasizing
an ingroup dialect, slang, or argot; diverging phonological, grammatical,
and discourse structures, etc. We have argued that the extent of this
ethnolinguistic differentiation, either in the long term of community-
wide sociolinguistic change or in the shorter term cases of interpersonal
divergences, is dependent on a variety of intergroup, contextual vari-
ables (Giles and Johnson 1981). Two of these are now mentioned briefly.
The first is the status relations existing between in- and outgroup as
perceived by the individuals themselves, as well as the nature of the
social comparisons made by them. Put more concretely, we argue, and
we have data to support this (Johnson 1984), that the more individuals
identify strongly with their ethnic group and the more they perceive any
status differences existing between them as illegitimate and potentially
changeable (see Turner and Brown 1978), the more likely it is that forms
of ethnolinguistic differentiation will occur. A second set of variables
we deem important are referred to collectively as 'vitality' factors
(Giles et al. 1977) and were formulated out of a dire need to place social
psychological processes in language studies in their proper macrosocio-
structural contexts (see also Ryan, Giles, and Sebastian 1982). Vitality
factors were envisaged as forming three independent clusters of vari-
ables: the relative sociohistoricolinguistic status of groups in contact;
their demographic profiles in terms of birth rates, proportions, emigra-
tion patterns; and the institutional support their cultures and languages
enjoy. We have argued that members of ethnic groups cognitively repre-
sent such societal forces (e.g. Bourhis, Giles, and Rosenthal 1981; John-
son, Giles, and Bourhis 1983). Yet, not only do we have evidence that
they actually can do this, but that in- and outgroups construe these
forces differently, and attach different cognitive weights to them (Bour-
his and Sachdev 1984; Giles, Rosenthal, and Young in press). Further-
more, we argue that the more of these vitality factors individuals con-
strue in their group's favor, the more likely their perception of societal
forces can bolster their sense of ethnic identification, and consequently,
the more likely ethnolinguistic differentiation will ensue in a wide array
of intergroup contexts (Johnson 1984).
Recently, we have begun to explore how the areas of ethnic language
attitudes, on the one hand, and second language acquisition, on the
other, can be theoretically enriched by recourse to these mediating
cognitive contextual variables. Despite the fact that both domains pay
lip service to them, we nonetheless believe that they are operating in
'intergroup vacua'. Space again precludes a development of these ideas
but, suffice it to say, we can proffer concrete propositions, based on an
analysis of individuals' intergroup belief structures, concerning what
language attitude profiles certain listener-judges will emit and why
(Ryan, Hewstone, and Giles 1984), and what levels of proficiency minor-
ity group members will exude in the dominant group's language, and why
(Giles and Byrne 1982; Ball, Giles, and Hewstone 1984).
Couple identities. The discussion so far has provided a flavor of how

contextual factors, often cognitively mediated at the individual and
intergroup levels by personal and group identity processes, can influence

linguistic forms and attitudes. Interestingly, we have, as has main-
stream literature in the language sciences, passed by a crucial level of
analysis, viz. that involving contexts where relational identities are
paramount. It is our view that cognitive processes attending 'couple
identities 1 (intimate heterosexual, as well as others) can mediate the
social meanings of linguistic forms; our methods and theories seem to
have bypassed such a relational context analysis. Indeed, Bolton (1961)
has shown how psychologically meaningful it is for partners in a hetero-
sexual relationship to arrive at a consensus of their being a 'couple'.
Spiegelberg (1973) provides an interesting analysis of how this is mani-
fest linguistically, for instance, by statements such as we feel..., we
did... (see also Rausch, Barry, Hertel, and Swain 1974), while Hopper,
Knapp, and Scott (1981) have explored the fascinating world of intra-
couple talk through pet names, abbreviations, and neologisms. Whilst
again our field is hardly devoid of dyadic and relationship constructs and
measures (see, for example, Berger and Kellner 1975; Bochner and
Krueger 1979; Brown and Levinson 1979; Krueger 1982), it is neverthe-
less the case that the predominant unit of analysis is the individual. It
may well be that we are missing important social meanings in our studies
of language behaviors if we neglect to probe how a sense of couple
identity defines what we listen to and how we speak (see Bradac 1983).
Relatedly, Wiemann and Kelly (1981) argued that interpersonal com-
petence and empathy are not skills inherent in individuals' communica-
tive behaviors as much as in their jointly negotiated relationship
performance.
As an initial exploration into the empirical world of couple identity
and language attitudes, we designed and conducted an experiment to be
discussed in due course. It was our belief that just as people have speech
stereotypes of competent individuals as fast talking, standard-accented
speakers, so too, people may have speech stereotypes of how different
couple types sound. Moreover, and as we demonstrated earlier, stereo-
typed beliefs may color listeners' rememberings of such talk. In this
vein, and building upon various assumptions of labeling theorists (e.g.
Garfinkel 1956; Kitsuse 1962; Lofland 1966), Snyder (1981:184), from a
'cognitive bolstering1 perspective, stated:
An individual having adopted stereotyped beliefs about a target

will: (1) remember and reinterpret past events in the target's life
history in ways that bolster and support these current stereotyped
beliefs; and (2) will act upon these current stereotyped beliefs in
ways that cause the actual behavior of the target to confirm and
validate the individual's stereotyped beliefs about the target.
Hence, the prime aim of our admittedly embryonic investigation was to
determine whether the 'retrospective speech halo effect' found operating
at the individual level in the first section of this paper could also func-
tion relationally. More specifically, would providing listener-judges with
contextual information about a couple's state of marital adjustment, on
the one hand, or their relational identity, on the other, evoke sufficient
stereotyped communication beliefs about them that would influence
perceptions of their talk as a couple? The empirical enterprise of
eliciting linguistic judgments of couples as relational units seems an

important innovation in the language attitudes domain, conceptually and
methodologically, as it has archetypally adopted the individual as its unit
of analysis (see reviews in Ryan and Giles 1982). Before proceeding to a
description of our study, let us survey briefly the research traditions
which led to the formulation of specific hypotheses concerning the con-
tent of these relational speech stereotypes which might color linguistic
reconstructions of couple talk.
Scholars in a variety of disciplines, including historians, sociologists,
psychologists, and clinicians, have pursued the search for the causes of
the success or failure of a marriage, each offering a variety of explana-
tions for marital satisfaction and stability. Nevertheless, there emerges
a surprisingly consistent viewpoint: that is, in our culture, subjectively
experienced contentment in a marriage is the primary determinant of
whether it will remain stable or intact (Lewis and Spanier 1979). Fur-
thermore, the communication that takes place between a husband and
wife leads to this contentment or satisfaction. Thus, a whole tradition
of research has sprung up, investigating the linguistic/com municational
correlates of marital disturbance. Dozens of studies now attest (be it in
self-report, informal observation by case studies, laboratory role-play-
ing, etc.) that maritally nonadjusted couples exude more negative affect,
verbally, vocally, and visually, as well as portray inconsistencies
between these channels. Less information (particularly personal) is ex-
changed, fewer problem-solving acts are attempted, and their interper-
sonal conflicts are dealt with in a mutually less accommodating manner
(e.g. Lederer and Jackson 1968; Mishler and Waxier 1968; Altman and
Taylor 1973; Gottman 1982; Krueger and Smith 1982; Noller 1982).
Figure 1 displays some of these differentiating items. It was our hypo-
thesis that laypeople may well possess cognitive representations of
marital communication in accord with at least some of the findings, such
that these stereotyped beliefs could potentially bias listeners' linguistic
judgments of an unfamiliar couple's talk in the direction of their known
marital adjustment-nonadjustment. Hence, items appearing in Figure 1
were included as dependent measures in our study to follow.
Another approach to this research area has been that of Fitzpatrick
and her associates (Fitzpatrick 1977, in press; Fitzpatrick and Best
1979), which empirically types marital relationships in terms of their
levels of interdependence, their ideologies of relationships, and their
communication patterns. More specifically, this typology emerged after
a series of large-scale pilot tests and finally after an analysis of over
1,000 married partners' responses to a questionnaire (the 'Relational
Dimensions Instrument': RDI) which includes such items as: 'We tell each
other how much we love or care about each other1 (agree—disagree) and
'The ideal relationship is one which is marked by novelty, humor and
spontaneity'. Three ways of defining a relationship arose (seemingly
independent of demographic correlates), viz. 'traditional 1 , 'separate', and
'independent'. About 60% of couple partners agree on the basic
dimensions of their marriage in terms of interdependence, ideology, and
communication (i.e. the 'pure1 types), while the other 40% comprise the
'mixed' varieties (e.g. 'traditional-separate').
Figure 1. Some communicational correlates of marital adjustment-

nonadjustment also serving as predictions for the present
investigation.
Items Couple type:
Adjusted Nonadjusted
Marital adjustment la +
Self-disclosed 2a +
Talked frequently 2a +
Stopped communication
less often 2a +
Accurate interpretations
of spouse meaning 3a +
Generally positive 4b +
Generally negative 4b +
Cross-complaining
sequences 4b +
First reciprocated positive
remarks, then stopped 4b +
Interaction patterned 4b +
Discussing a problem, wife
uses a negative tone 5b +
Key. The letter a_ refers to findings from self-report studies, while b_
indicates behavioral observation studies. Numbers refer to the study in
question, viz. 1, Spanier (1976); 2, Kahn (1970); 3 Bienvenu (1970); 4,
Gottman (1979); 5, Noller (1980). Plus sign (+) indicates this item is
evoked by the couple type. Minus sign (-) indicates that item is not
clearly apparent.
Very briefly, let us outline the attributes of these couples. Tradi-

tionals hold conventional ideological values in relationships, favor a high
degree of interdependence between marital partners, and report that
they are expressive with their mates. Independents hold fairly noncon-
ventional values about relational and family life, favor interdependence
yet push for autonomy, and report that they are extremely expressive
with their mates. Separates vacillate between conventional and noncon-
ventional ideologies of family life, are not companionate with their
mates, and report a distinct lack of expressiveness in their relation-
ships. Importantly, research by Fitzpatrick and others suggests that not
only do self-reports of couples' communicative behaviors differentiate
meaningfully between the types, but that their actual exchange of
messages can be observed objectively as predictive of the typology in
both casual and conflict resolution situations (see Fitzpatrick in press;
Fitzpatrick, Vance, and Witteman 1984). Figure 2 displays a fair number
of these differentiating behaviors. Therefore, as earlier, it was our
further hypothesis that laypeople may well hold cognitive representa-
tions of marital communication in accord with at least some of the
findings relating to Fitzpatrick's typology, such that these stereotyped
beliefs could feasibly bias listeners' linguistic judgments of an unfamiliar
couple's talk in some settings in the directions of their relational identi-
ties (cf. Wilmot and Baxter 1983). Hence, items appearing in Figure 2
were also included as dependent measures in the study to follow.
Figure 2. Some communicational correlates of Fitzpatrick's typology

also serving as predictions for the present investigation.
Items Couple types:
Tradi- Inde- Separate
tional pendent
Showed affection la ++ + +
Marital adjustment 2a ++ + +
Agreed with each other 2a ++ +
Took partner into account 3a ++ +
Self-disclosed 3a ++ +
Husband (sounded) mas-
culine 4a + + ++
Husband (sounded) fem-
inine 4a - +
Wife (sounded) feminine 4a + +
Wife (sounded) masculine 4a - +
Talked frequently 5b + +
Cooperated in conflicts 5b + -
Negotiated verbally 6b + -
Flexible communication 7b + _
Vocal emotional tones
negative 8b - +
Each partner tried to con-
trol the conversation 9b - ++
Complementary in communi-
cation style 10b - +
Key. The letter a refers to findings from self-report studies, while ]3
indicates behavioral observation studies. Numbers refer to the study in
question, viz. 1, Fitzpatrick (1976); 2, Fitzpatrick and Best (1979); 3,
Fitzpatrick (1977); 4, Fitzpatrick and Indvik (1982); 5, Fitzpatrick,
Fallis, and Vance (1982); 6, Fitzpatrick, Tenney, and Witteman (1983); 7,
Fitzpatrick (1983); 8, Sillars, Pike, Redman, and Jones (1983); 9, Wil-
liamson (1983); 10, Fitzpatrick (in press b). Plus sign (+) indicates that
this item is evoked by the couple type, and double plus sign (++) indicates
this is extremely so. Minus sign (-) indicates this item is not clearly ap-
parent, and double minus sign (—) suggests it is definitely not evoked.
To reiterate, the research questions prompting this exploratory re-

search were: (1) would contextual information pertaining to marital
adjustment-nonadjustment and Fitzpatrick's three 'pure' couple types
have enough psycholinguistic reality to bias retrospectively listeners'
judgments of a couple they had just heard talking on audiotape? and (2)
would these reconstructions be in the direction of the linguistic and
communicational profiles displayed in Figures 1 and 2?
Participants in our investigation were 244 undergraduates enrolled in
an introductory communication class at a mid western university. They
were assigned randomly to one or other of our ministudies, viz. the
marital adjustment, or the typology experiments. All participants
listened to the same 90-second conversation between a young married
couple concerning the purchase of a used automobile. The speaking roles
of husband and wife were equalized to the extent that they engaged in
the same number of floor switches and used the same number of words.
In addition, the number of agreements and disagreements used by each
spouse was not signifcantly different.
This conversation was pretested to determine whether the conversa-
tion itself revealed any information about the type of marriage of the
conversants. Thirty-seven individuals, drawn from a pool of subjects
similar to those who would be participants in the experiment proper,
listened to the tape and completed 17 questions drawn from the RDI. On
1^ of these, the pretesters saw the couple as not significantly different
from the mid-points on these scales. In other words, we were satisfied
that this conversation was emotionally neutral, although the remaining
three items suggested listeners viewed the couple as nonconflict-ridden
and traditional.
All participants in the main study listened to the stimulus tape in the
knowledge that they would be required to answer questions about it sub-
sequently. Immediately after having listened to the taped couple, and
just prior to making a series of ratings along 9-point scales, listeners
were provided with typewritten information about the couple (cf. Ball et
al. 1982). In the marital adjustment ministudy, participants received
descriptions of the couple as being either very satisfied or very dis-
satisfied with their marriage (see Spanier 1976), viz.:
3ohn and Susan want to stay together very much and have few in-
tense or serious conflicts. They have never discussed separation or
divorce. They are very satisfied with their marriage and tend to
kiss one another every day.
John and Susan do not want to stay together very much and have
frequently discussed separation or divorce. They are very dis-
satisfied with their marriage and tend not to kiss one another every
day.
In the typology ministudy, participants were provided with one or other
of the following three paragraphs representing the basic marriage types
(viz. traditional, independent, and separate, respectively).
John and Susan have very strong traditional values on marriage and
family life. They share almost all aspects of their lives with one
another. They have very regular daily time schedules and do not
feel the need for private space in their home away from one an-
other.
John and Susan have very strong nontraditional values on marriage
and family life. They share many, but not all, aspects of their lives
with one another. They have irregular daily time schedules and feel
the need for private space in their home away from one another.
John and Susan outwardly have very traditional values about mar-
riage and family life but often doubt these values. They share few
aspects of their lives with one another. They have very regular
daily time schedules and feel the need for private space in their
home away from one another.
Recall now that all listener-judges had listened to exactly the same
tape-recording of this couple talking. The only difference between the
five subgroups of listeners was in the contextual information provided
them concerning 3ohn and Susan. As can be seen from the foregoing,
arguably few explicit details were provided raters about the couple's
linguistic and communicational habits when talking together; this was
particularly so for the typology ministudy. Having read this brief infor-
mation, all participants were instructed to rate the language and com-
municational behaviors of the couple specifically as they had just heard
them talk on tape on that particular occasion. These ratings were
evinced by means of a 40-item questionnaire derived from Figures 1 and
2, and also included a large number of items relating to social evaluation
(see Table 1 for a synopsis of most of these scales).
Let us now survey the results. A correlation matrix for the 40 de-
pendent measures for all participants was submitted to a Principal
Components Analysis to determine the number of factors utilized by
raters. Cattell's Scree test suggested that five factors appeared to
define this data set, accounting for 51% of the variance. The relatively
small amount of variance accounted for here may well have been due to
the very heavy cognitive demands placed on our respondents; that is,
assessing a more or less bland conversation by means of kQ nine-point
rating scales. Obviously, future work would profit from invoking an
evaluatively less arduous judgmental task. A subsequent factor analysis
using squared multiple correlations in the diagonals of the matrix, fol-
lowed by a varimax rotation, yielded five interpretable factors which
could serve as dependent measures, in addition to the individual item
data, in further analyses (see Table 1).
The first factor was labelled 'social evaluation1, and defined by such
terms as: 'Would you like this couple as close friends?', 'Would you go to
them for advice?', etc. The second factor was labelled 'negative com-
munication style', and included items which viewed the couple as using
negative voice tones with one another, engaging in cross-complaining
sequences, etc. The third factor was labelled 'open communication style'
and included such items as sharing thoughts and feelings, showing affec-
tion, and taking one another into account. The fourth factor was la-
belled 'cooperative communication style', and included items suggesting
that the couple generally agreed with, and were positive about, each
other. The fifth factor was labelled 'sex stereotyped voices' (see Smith
1980) and included items asking for evaluations of the masculine or
feminine nature of the voices of the husband and wife on tape. (In
subsequent data analyses, this factor was split into a masculine and
feminine vocal judgment factor in order not to sum evaluations of the
husbands and wives into one index.) Thus, the first factor that emerged
was an overall evaluation of the dyadic identity presented by our couples
while the other four factors dealt with the make-up of that identity.
Respondents appeared to have little trouble in evaluating these couple
identities or in seeing relationships among the component parts of these
identities.
/ Howard Giles and Mary Anne Fitzpatrick
Table 1. Factors and item loadings.

Factor Loading
U) Social evaluation (eigenvalue = 6.79; 48.2% variance)
Would you like to have a marriage like theirs? .79
Would you like them as close friends? .76
Would you go to them for advice? .73
Would you discuss issues as they do? .69
Do you think they are really compatible? .57
Do you respect this couple? .56
Do they seem a typical married couple to you? .52
It is clear that this couple does not talk much
together. -.50
Do you think their marriage will last? .47
(2) Negative communication style (eigenvalue = 2.68;
19% variance)
This couple was generally negative toward one another. .61
When she was discussing a problem with her husband,
the wife had a negative tone. .58
Each member of this couple tried to control the
conversation. .57
This couple engaged in 'cross-complaining sequences'. .55
This couple was quick to reciprocate a negative remark
of the spouse. .53
This couple initially reciprocated positive remarks of the
spouse, but later in conversation stopped doing so. .46
The vocal emotional undertones of this couple were
negative. .45
This couple interrupted one another fairly frequently. .37
(3) Open communication style (eigenvalue = 1.97;
14% variance)
This couple took into account each other's feelings. .65
This couple shared their thoughts and feelings with
one another. .63
Overall, this couple shows a high degree of marital
adjustment. .57
This couple showed their affection for one another. .51
This couple verbally negotiated with one another. .48
(4) Cooperative communication style (eigenvalue = 1.54;
10.9% variance)
This couple was generally positive toward one another. .43
This couple agreed with one another on most matters. .41
This couple used a neutral tone of voice. .40
Overall, this couple used far more agreement than
disagreement statements. .38
The interaction between this couple was extremely
'patterned' in that one remark can easily be
predicted from another. .38
This couple gave in to one another. .33
Factor ..__.._. Loading
(5) Sex stereotyped voices (eigenvalue = 1.12; 7.9%
variance)
Husband type:
The husband in this marriage sounded masculine. .56
The husband in this marriage sounded feminine. -.57
Wife type:
The wife in this marriage sounded feminine. .63
The wife in this marriage sounded masculine. -.57
Let us now examine the findings relating to the marital adjustment

ministudy. Given that Hotelling's T on this particular data set differen-
tiated ratings of listeners in the adjusted and nonadjusted conditions
(p<.01), one-way ANOVAs were computed for each of the six factors as
well as for the individual items. As can be seen from Table 2, a signi-
ficant difference emerged on the social evaluation factor (i.e. factor 1),
showing that listeners were more favorably disposed toward the adjusted
couple. Table 2 also pinpoints the individual items of social evaluation
which were also significantly disposed in this direction.
Table 2. Mean ratings and F values for listeners' judgments in the
marital adjustment ministudy.
Dependent measures Mean ratings* Fs Ps
Adjusted Nonadjusted
couple couple (df=l,108)
Social evaluation 3.65 3.33 4.22 .04
Open communication 5.46 5.04 2.19 .14
Snowed affection 3.18 2.18 6.38 .01
Marital adjustment 5.63 4.62 6.29 .01
Generally positive 6.12 4.93 8.70 .003
Generally negative 2.45 3.27 4.04 .04
Cooperated in conflict 7.16 6.40 4.84 .03
Made each other laugh 1.16 0.53 5.92 .01
Agreed with other 6.59 5.82 4.83 .03
Talked to frequently 4.98 3.85 4.41 .04
Went to for advice 2.98 2.03 4.90 .03
Really compatible 4.49 3.62 3.79 .05
Marriage will last 5.29 3.33 20.71 .0001
*The higher the mean rating, the more listeners perceived the de-
pendent measures in question.
Although the overall mean values for open communication (factor 3) in

Table 2 were not significantly different statistically, an examination of
the means on the individual items indicates that listeners viewed the
maritally adjusted as having a more 'open' (as well as 'cooperative')
communication style (cf. Table 1).
Moving along to the typology ministudy, where an overall MANOVA on
this particular data set differentiated ratings for the three couple types
(p<.01), similar ANOVAs provided somewhat more extensive differences
than the foregoing. As Table 3 indicates, both the factors of 'social
evaluation1 and 'open communication1 were significantly different for the
couple types. An examination of the individual items specifies this
trend. Here, the traditionals were perceived to have shown more
affection, were seen to be more adjusted, were generally more positive
and hence less negative, cooperated in conflict, self-disclosed more, and
were less likely to stop positive reciprocity cycles in their communica-
tion than the independents, who in turn were more 'open' and 'coopera-
tive' (cf. Table 1) than were the separates. In addition, our respondents
evaluated the traditionals significantly more favorably than the indepen-
dents, who were in turn judged more positively than the separates.
Table 3. Meaning ratings and F values for listeners' judgments in
the marital adjustment ministudy.
Dependent measures Mean ratings*, couple type F F
Tradi- Inde- Sepa- (df=
tional pendent rate 2,131)
Social evaluation 4.13 3.14 2.71 3.06 .05
Open communication 6.21 5.66 4.89 14.35 .00
Showed affection 3.68 3.45 2.06 9.95 .0001
Marital adjustment 6.45 5.50 5.09 6.60 .002
Generally positive 6.30 6.13 5.40 3.59 .03
Generally negative 2.06 2.05 2.81 3.06 .05
Cooperated in conflict 7.58 6.87 6.75 3.54 .03
Took partner into
account 6.58 6.00 5.62 3.31 .04
Self-disclosed 6.34 5.95 4.28 11.45 .0001
Stopped positive
reciprocation 3.19 3.45 4.19 3.36 .04
Flexible communi-
cation 3.64 4.66 2.85 5.36 .005
Patterned communi-
cation 6.53 5.68 7.00 5.10 ,007
Respected this couple 5.11 4.53 4.00 2.92 .05
*The higher the mean rating, the more listeners perceived the depend-
ent measure in question.
In general, the differences emerging as statistically significant across

both ministudies, although not numerous, are nonetheless striking and in
line with our predictions. Interestingly, items not formally predicted on
the basis of Figure 1 to differentiate adjusted and nonadjusted percep-
tions did emerge in Table 2 (e.g. 'made each other laugh'). Similarly,
items not formally predicted to differentiate between the three couple
types from Figure 2 also appeared in the perceptual profile of Table 3
(e.g. 'stopped positive reciprocation'). It appears therefore that speech
stereotypes associated with different relational types may be concep-
tually distinct from the objective (and self-report) communication
measures which differentiate them (see Moscovici and Hewstone 1983).
Nevertheless, there is a good degree of isomorphism between the two.
An exception emerged with respect to findings on the measures of flex-
ibility and patterning of communication styles in Table 3. In line with
previous observational research on the couple types (see Figure 2), the
separates were viewed as the most rigid in their communication with the
spouse. The predicted differences for the traditionals and the inde-
pendents, however, did not occur. Indeed, the traditionals were viewed
by our respondents as more rigid and patterned in their communication
than were independents. In the couple realm, this may be akin to1 the
phenomenon of 'psychological convergence but linguistic divergence in-
troduced in the first section of this paper (see Thakerar et al. 1982).
The patterns which emerge in the stereotype of the various relationships
do not reflect what occurs linguistically when the interaction of actual
couples is analyzed. Although psychologically our respondents link the
conservative ideological orientation in marriage and family life of the
traditionals to a rigid communication style, linguistically such rigidity
does not occur in traditional dialogues with a spouse. Similarly, the
independents are not as flexible as they seem to our respondents.
Respondents did converge, however, on questions concerning the fre-
quency of certain communicative behaviors (e.g. self-disclosures, gener-
ally positive). Only on questions concerning interaction did reactions
shift away from our predictions. Our respondents, and perhaps couples
themselves, may not notice and store interaction patterns but rather
respond to the frequency of occurrence of specific communicative
behaviors. When respondents, and again couples themselves, evaluate
dyadic encounters, they may remember only first-order acts (he said/she
said) and not the complex strings of interaction favored by observational
researchers. The relationships between self-reports, behavioral frequen-
cies, and interaction patterns in the minds of those evaluating couple
identities may differ from actual relationships among these factors in
couple communication. Such questions will be of obvious import for
future work in this realm.
It is our belief that if the taped conversation had been less 'bland1,
with the couple evincing a little more interpersonal conflict and tension,
retrospective speech halo effects may well have permeated other judg-
mental dimensions as well (e.g. the negative communication factor). We
also think that, had the contextual information regarding couple types
been introduced prospectively rather than retrospectively, similar find-
ings would have emerged (see Snyder 1981). Indeed, our previous re-
search on this phenomenon at the individual level (Thakerar and Giles
1981; Ball et al. 1982) supports such a contention. Therein, listeners had
an opportunity of hearing the stimulus tape a second time, with the
chance to modify their ratings in an additional phase of these experi-
ments. On both occasions, respondents re-rated in a manner identical to
their original, and biased, linguistic judgments. Obviously, future work
should orient itself to discover the communicational and social environ-
ments which attenuate and accentuate the relational retrospective
speech halo effects found herein. Moreover, it would be interesting to
determine whether superimposing linguistic and communication be-
haviors in Figure 2 onto an emotionally neutral conversation between
spouses would induce raters to see them along the lines of Fitzpatrick's
typology (see Bradac 1983); our guess is that they would.
Two important findings emerge from this study for our present pur-
poses. First, listener-judges appear able to form linguistic judgments
along an interesting set of constructs (see Table 1) of a couple as a
relational unit. This is a far cry from the individual speaker/monologue

paradigm of language attitudes with which this paper started. Indeed, it
is our contention that very often we do not listen to the linguistic forms
and meanings of partners of a cohesive dyad as separate individuals; we
process their message 'as a couple' (cf. Pearce and Conklin 1979). Se-
cond, what we think we know about couples can lead us sometimes to
hear them as such. Or rather, this contextual information can evoke re-
lational speech stereotypes which admittedly, within certain limits, may
form a template for constructing and reconstructing our linguistic im-
pressions of them (cf. Snyder 1981).
The foregoing study was presented as a mere illustration of the type
of research that is needed if we are to explore the social meanings of
linguistic forms in a relational context. Obviously, we have just
scratched the surface empirically, and a whole plethora of more con-
sequential studies loom on the horizon, enabling us to tap the complexity
and richness of a couple identity approach to language behaviors. Yet,
our findings on how couple identity can cognitively mediate our rela-
tional language attitudes allow us the luxury of speculating briefly about
how language forms and other social behaviors are likewise affected.
When few objective standards are available for comparison, individuals
compare themselves to similar others to see how they are faring
(Festinger 1954; Suls and Miller 1977). Tajfel (1974) has extended such
interpersonal comparisons to the intergroup level. We hereby suggest
that intercouple comparisons are also prevalent and psychologically in-
formative. Hence, we track the linguistic and nonlinguistic behavior of
other couples to help us to define the state of our own relationship, or
even relationships. Our data seem to indicate, furthermore, that in-
dividuals eschew as standards of evaluation or comparison any couple
judged to be having any type of relational difficulty. At the same time,
however, such a comparison may cognitively bolster our own relation-
ship's sense of worth and vitality. Yet, it may be that once a couple is
judged by members of a social network as experiencing relational stress,
they are cut off from social contact. When the public couple identity is
damaged, internal relational difficulties may be further exacerbated,
because the couple is cut off from their external social support net-
works.
There are good reasons for believing that such a postulated couple
comparison process is utilized extensively to help us define the oft-
quoted 'meaningfulness' of our relationships, and 'where we are at'. As
Berger and Bradac (1982) point out, while there are literally hundreds of
terms for describing individuals—and we would add for describing
groups—there are few for couples. This lack of categorical language for
describing relationships may only add to the problems we have in de-
creasing cognitive uncertainty about our relationships. Numerous re-
searchers (e.g. Duck and Sants 1982) remind us that such uncertainty is
high not only at the beginning of a relationship but at many phases
throughout the development and dissolution of a relationship. Ragan and
Hopper (1984) proffer the intriguing suggestion that certain fiction
provides us with the fodder for our supposed idiosyncratic 'love talk';
after all, who drills us into our 'sweet nuthins'? Feminists and others
interested in the construction of social reality through talk often warn
of the difficulty of not having terms to describe interpersonal
experiences. Consider the ontological dilemma of an unhappily married

individual who has only exceedingly happy TV and real life couples
against whom to compare his or her own relationship. We can only
speculate that the ramifications of these comparisons for the satisfac-
tion and stability of his or her own evaluated relationship are negative
ones.
In our thinking about couple identities, we have been influenced by
two major theoretical perspectives: the Meadian (1934) perspective
which holds that other individuals' reactions to us can in time determine
our own view of our personal identities; and self-presentational theory
(Goffman 1959; Tedeschi 1981) which sees people actively attempting to
manage and influence others' impressions of them. Obviously, an im-
portant means of achieving such ends is via verbal strategies (Tedeschi
and Reiss 1981), vocal patterns (Giles and Street in press), and com-
municator styles (Norton 1983). We would desire explicitly to extend
these analyses also from their personal bases into the couple identity
arena. In other words, in parallel with couple comparison processes, we
feel that others' reactions to our own relationships can in time deter-
mine our views of them (see Milardo, Johnson, and Huston 1983). But we
can be highly active in determining the nature of this process by pre-
senting what 'we', at least, regard as a valued couple identity through
language behaviors. Such dyadic presentations can be fundamental to
our couple identity in many social networks. This 'couple talk', as im-
plied earlier, will be a function of the type of relational identity the
partners wish to espouse. For instance, managing a couple image of 'still
being madly in love after all these years' could be effected through
expressed egalitarian roles, shared involvements, and joint achieve-
ments; mutual expressions of trust and integrity, use of private codes
and taken-for-granteds, overt nonverbal indices of affection, an atten-
tive, empathic communication style, politeness and conversational
synchrony, etc., etc. Needless to say, effecting a couple presentation of
marital distress, or of 'independence', would be differently managed,
linguistically.
Such impression management occurs in good times and also in bad
ones. Duck (1982) implicitly discusses couple presentation, albeit of a
grossly different ilk, at the end of a relationship. His so-called 'grave
dressing phase' allows couples to effect a public, socially appropriate
scenario where often mutual blame for the relationship is accorded.
This allows both partners to disengage from the couple identity to a
personal identity as well. Negotiating the change from a couple identity
to a personal one may present difficulties for both relational outsiders
and for insiders. Often, we see and know others only as members of a
couple. When we meet them unexpectedly in another context, we find it
difficult sometimes to operate communicatively at a personal identity
level (Wiemann, personal communication). Communication therefore
might be awkward at times until interactants have negotiated which
identity is to predominate in a given interaction. Furthermore, inter-
acting without one's spouse or partner physically present does not neces-
sarily imply that the relationship is not paramount when one is talking to
a third party. In the same way as a cognitive definition of group mem-
bership allows one to act in accord with it thousands of miles away from
other group members (Giles and Johnson 1981; Turner 1982), so, too, we
can effect couple talk when separated from our partner. In such in-
stances, couple identity could mediate individual linguistic choices
through 'we...'-related topics, references to 'our' activities, and an 'us'
focus manifest in expressedly shared attitudes and couple- (rather than
self-) disclosures.
Finally, the importance of recognizing the powerful mediating in-
fluence of couple identity is also evident in its capacity to induce dif-
ferent dimensions of contextual construing. And on this note, we have
returned full circle near to where we started at the outset of this paper.
For example, the situational parameters defining a supposedly formal in-
teraction between unknown, different individuals will of necessity be
quite different if these strangers define the context as an intergroup
one, and different again if they negotiate a romantic or even sexual re-
lationship (Rands and Levinger 1979; Forgas and Dobosz 1980).
Conclusion. Our message has been this. It is true that context affects
language attitudes and linguistic forms, and examples were drawn from
our work in this respect. However, we advocate expending more re-
search effort on exploring the ways in which personal attitudes and
perceptions of others' speech and the situation we are in (not to mention
a host of other construals, such as subjective norms; see McKirnan and
Hamayan 1984; Giles and Street in press) cognitively mediate the social
meanings of language behaviors. Yet, an important development beyond
this approach is to extend our understanding of the linguistic implica-
tions of the contextually based notion of group identity. We have argued
that cognitive mediators of ingroup identification, as well as intergroup
belief structures, have important linguistic correlates and consequences.
Finally, we envisaged a 'relational vacuum' in current language research
and have begun flushing this out empirically in a judgmental vein whilst
speculating about its significance in production processes. Exciting
prospects are therefore ahead, not least of which will ultimately be
mapping out how people shift linguistically back and forth between their
personal, group, and couple identities sequentially, and conceivably, how
they maintain these different identities simultaneously by means of dif-
ferent linguistic features and communicational devices. As we have
emphasized elsewhere on a number of occasions, a social psychological
approach to the study of meaning, form, and context is but one cog in
the language and society wheel. There are many others. Sometimes,
like a figure ground illusion, a social psychological analysis comes to the
fore; at other times (and often), it is in the faded background of explana-
tory power. As you may have surmised from this paper, or from a gen-
eral reading of the social psychology of language and communication
science more generally, we are desperately in need of linguistic input
and sophistication in our analyses of language data and in our theoretical
models. But we do believe that through its methodologies, techniques,
instruments, and theories, a social psychological complement has the
potential for linguistic applications too. The time is ripe for us to move
beyond routinized multidisciplinarian service to a truly interdisciplinary
context for language study, not only in form, but also in our shared
meanings.
Notes
Further information on all aspects of this study, including more de-
tailed statistical data, can be obtained from the second author. We are
grateful to Diane Badzinski, Scott Broetman, Jane Byrne, and Hal Wit-
teman for their invaluable assistance in material preparations,
recordings, data collection, and analyses. This investigation was de-
signed and the data collected when the first author was Brittingham
Visiting Professor in Communication Arts, University of Wisconsin-
Madison. We are also grateful to James Bradac, 3ane Byrne, Dorothy
Krueger, Tony Mulac, and John Wiemann for their helpful and construc-
tive comments on an earlier draft of this paper.
1. Since this was our initial exploration into this particular domain,
only the 'pure' types were examined. Future research will consider
'mixed' (stimulus) types as well as explore the implications of varying
listener-judges' own relational identity types and marital experiences.
2. In order to determine which ratings were meaningfully and sig-
nificantly different from each other within each of the two ministudies,
appropriate statistical tests were applied to the raw data. These were
Hotelling's T , analysis of variance (ANOVA), and multivariate analysis
of variance (MANOVA).
References
Altman, I., and D. Taylor. 1973. Social penetration: The development
of interpersonal relationships. New York: Holt, Rinehart, and Win-
ston.
Ball, P., J. Byrne, H. Giles, P. Berechree, J. Griffiths, H. Macdonald, and
I. McKendrick. 1982. The retrospective speech halo effect: Some
Australian data. Language and Communication 2.277-84.
Ball, P., H. Giles, and M. Hewstone. 1984. The intergroup theory of
second language acquisition with catastrophic dimensions. In: The
social dimension: European developments in social psychology. Edited
by H. Tajfel. Cambridge: Cambridge University Press.
Beebe, L. M. 1981. Social and situational factors affecting the com-
municative strategy of dialect code-switching. International Journal
of the Sociology of Language 32.139-49.
Bell, A. 1982. Radio: The style of news language. Journal of Commun-
ication 32.1:150-64.
Berger, C. R., and J. J. Bradac. 1982. Language and social knowledge.
Baltimore: Edward Arnold.
Berger, P., and H. Kellner. 1975. Marriage and the construction of
reality. In: Life as theater: A dramaturgical handbook. Edited by D.
Brisset and C. Edgley. Chicago: Aldine.
Bienvenu, M. 1970. The measurement of marital communication.
Family Coordinator 19.26-31.
Bochner, A. P., and D. L. Krueger. 1979. Interpersonal communication
theory and research: An overview of some inscrutable epistemologies
and muddled concepts. In: Communication Yearbook 3. Edited by D.
Nimmo. New Brunswick, N.J.: Transaction.
Bolton, C D . 1961. Mate selection as the development of a relation-
ship. Marriage and Family Life 23.234-40.
Bourhis, R. Y., H. Giles, and D. Rosenthal. 1981. Notes on the 'Sub-

jective Vitality Questionnaire1. Journal of Multilingual and Multicul-
tural Development 2.145-55.
Bourhis, R. Y., and I. Sachdev. (1984) Subjective vitality perceptions
and language attitudes in Hamilton. Journal of Language and Social
Psychology 3.97-126.
Bradac, J. J. 1983. The language of lovers, flovers, and friends: Com-
municating in social and personal relationships. Journal of Language
and Social Psychology 2.l4l-62.
Brown, B. L. 1980. Effects of speech rate on personality attributions
and competency evaluations. In: Language: Social psychological
perspectives. Edited by H. Giles, W. P. Robinson, and P. M. Smith.
Oxford: Pergamon.
Brown, B. L., H. Giles, and J. N. Thakerar. 1983. The effects of accent,
speech rate and context on the attributions of a speaker's personality
characteristics. Mimeo. Psychology Department, Brigham Young
University, Provo, Utah.
Brown, P., and S. Levinson. 1979. Social structure, groups and inter-
action. In: Social markers in speech. Edited by K. R. Scherer and H.
Giles. Cambridge: Cambridge University Press.
Caporael, L. 1981. The paralanguage of caregiving: Baby talk to the
institutionalized aged. Journal of Personality and Social Psychology
40.876-84.
Creber, C , and H. Giles. 1983. Social context and language attitudes:
The role of formality-informality of the setting. Language Sciences
5.155-61.
Duck, S. 1982. A topography of relationship disengagement and dissolu-
tion. Personal relationships 4. In: Dissolving personal relationships,
vol. 2. Edited by S. Duck. London: Academic Press. 141-62.
Duck, S., and H. Sants. 1982. On the origin of the specious: Are per-
sonal relationships really interpersonal states? Journal of Social and
Clinical Psychology 1.27-41.
Festinger, L. 1954. A theory of social comparison processes. Human
Relations 7.117-40.
Fishman, J. A. 1966. Language loyalty in the United States: The main-
tenance and perpetuation of non-English mother tongues by American
groups. The Hague: Mouton.
Fitzpatrick, M. A. 1976. A typological approach to communication in
relationships. Unpublished Ph.D. dissertation, Temple University,
Philadelphia.
Fitzpatrick, M. A. 1977. A typological approach to communication in
relationships. In: Communication Yearbook 1. Edited by B. Rubin.
New Brunswick, N.J.: Transaction.
Fitzpatrick, M. A. 1983. Predicting couple's communication from
couple's self-reports. In: Communication Yearbook 7. Edited by R.
Bostrom. Beverly Hills, Calif.: Sage.
Fitzpatrick, M. A. (in press a) A typological approach to marital inter-
action: Recent theory and research. Advances in Experimental Social
Psychology.
Fitzpatrick, M. A. (in press b) A typology of relationships. Journal of
Communication.
Fitzpatrick, M. A., and P. Best. 1979. Dyadic adjustment in traditional,
independent, and separate relationships: A validation study. Com-
munication Monographs 46.167-78.
Fitzpatrick, M. A., S. Fallis, and L. Vance. 1982. Multifunctional coding
of conflict resolution strategies in marital dyads. Family Relations
31.61-70.
Fitzpatrick, M. A., and J. Indvik. 1982. The instrumental and expressive
domains of marital communication. Human Communication Research
8.195-213.
Fitzpatrick, M. A., B. Tenney, and H. Witteman. 1983. Compliance-
gaining in marital interaction. Paper presented at the International
Communication Association Conference, Dallas.
Fitzpatrick, M. A., L. Vance, and H. Witteman. (1984) Interpersonal
communication in the casual interaction of marital partners. Journal
of Language and Social Psychology 3.81-96.
Forgas, J. 1983. Language, goals and situations. Journal of Language
and Social Psychology 2.267-93.
Forgas, J., and B. Dobosz. 1980. Dimensions of romantic involvement:
Towards a taxonomy of heterosexual relationships. Social Psychology
Quarterly 43.290-300.
Garfinkel, H. 1956. Conditions of successful degradation ceremonies.
American Journal of Psychology 61.420-24.
Giles, H., ed. 1977. Language, ethnicity and intergroup relations. Lon-
don: Academic Press.
Giles, H., ed. 1984. The dynamics of speech accommodation. Inter-
national Journal of the Sociology of Language 46.
Giles, H., and R. Y. Bourhis. 1976. Black speakers with white speech—a
real problem? In: Proceedings of the 4th International Congress of
Applied Linguistics, vol. 1. Edited by G. Nickel. Stuttgart: Hochschul
Verlag.
Giles, H., R. Y. Bourhis, and D. M. Taylor. 1977. Towards a theory of
language in ethnic group relations. In: Language, ethnicity and inter-
group relations. Edited by H. Giles. London: Academic Press.
Giles, H., and J. L. Byrne. 1982. An intergroup model of second lan-
guage acquisition. Journal of Multilingual and Multicultural Develop-
ment 3.17-40.
Giles, H., and J. R. Edwards, eds. 1983. Language attitudes in multi-
lingual settings. Journal of Multilingual and Multicultural Develop-
ment 4 (2 and 3).
Giles, H., C. Harrison, C. Creber, P. M. Smith, and N. H. Freeman.
1983. Developmental and contextual aspects of British children's lan-
guage attitudes. Language and Communication 3.141-46.
Giles, H., and M. Hewstone. 1982. Cognitive structures, speech and
social situations: Two integrative models. Language Sciences 4.187-
219.
Giles, H., and P. Johnson. 1981. The role of language in ethnic group
relations. In: Intergroup behavior. Edited by J. C. Turner and H.
Giles. Chicago: University of Chicago Press.
Giles, H., D. Rosenthal, and L. Young, (in press) Perceived ethnolin-
guistic vitality: An Anglo/Greek-Australian setting. Journal of
Multilingual and Multicultural Development.
Giles, H., and B. Saint-Jacques, eds. 1979. Language and ethnic rela-
tions. Oxford: Pergamon.
Giles, H., and R. L. Street, Jr. (in press) Communicator characteristics
and behavior: A review, generalizations, and model. In: Handbook of
interpersonal communication. Edited by M. Knapp and G. R. Miller.
Beverly Hills, Calif.: Sage.
Goffman, E. 1959. The presentation of self in everyday life. Garden
City, N.Y.: Doubleday.
Gottman, J. 1979. Marital interaction. New York: Academic Press.
Gottman, J. 1982. Emotional responsiveness in marital conversation.
Journal of Communication 32.108-20.
Gumperz, J. J. 1982. Language and social identity. Cambridge: Cam-
bridge University Press.
Hopper, R., M. Knapp, and L. Scott. 1981. Couple's personal idioms:
Exploring intimate talk. Journal of Communication 31.23-33.
Hymes, D. 1967. Models of the interaction of language and social set-
ting. Journal of Social Issues 23.8-28.
Johnson, P. 1984. Language and ethnicity: Welsh/English context.
Ph.D. thesis, Bristol University.
Johnson, P., H. Giles, and R. Y. Bourhis. 1983. The viability of ethno-
linguistic vitality: A reply to Khan and Husband. Journal of Multilin-
gual and Multicultural Development 4.255-69.
Kahn, M. 1970. Nonverbal communication and marital satisfaction.
Family Process 9.449-56.
Kitsuse, J. I. 1962. Societal reactions to deviant behavior: Problems of
theory and method. Social Problems 9.247-56.
Krueger, D. L. 1982. Marital decision-making: A language-action
analysis. Quarterly Journal of Speech 68.273-87.
Krueger, D. L., and P. Smith. 1982. Decison-making patterns of
couples: A sequential analysis. Journal of Communication 32.121-34.
Labov, W. 1966. The effect of social mobility on linguistic behavior.
Social Inquiry 36.186-203.
Lederer, W. J., and D. D. Jackson. 1968. The mirages of marriage.
New York: Norton.
Lewis, R. A., and G. Spanier. 1979. Theorizing about the quality and
the stability of marriage. In: Contemporary theories about the
family: Research-based theories, vol. 1. Edited by W. Burr, R. Hill,
F. I. Nye, and I. R. Reiss. New York: Free Press.
Lofland, J. 1966. Deviance and identity. Englewood Cliffs, N.J.: Pren-
tice-Hall.
McKirnan, D. J., and E. J. Hamayan. (1984) Speech norms and attitudes
toward outgroup members: A test of a model in a bicultural context.
Journal of Language and Social Psychology 3.21-38.
Mead, G. H. 1934. Mind, self and society. Chicago: Chicago University
Press.
Milardo, R. M., M. P. Johnson, and T. L. Houston. 1983. Developing
close relationships: Changing patterns of interaction between pair
mates and social networks. Journal of Personality and Social Psychol-
ogy 44.964-76.
Mishler, E., and N. Waxier. 1968. Interaction in families. New York:
Wiley.
Moscovici, S., and M. Hewstone. 1983. Social representations and social
explanations: From the 'naive' to the 'amateur' scientist. In: Attribu-
tion theory: Social and functional extensions. Edited by M. Hew-
stone. Oxford: Blackwell.
Myers, D. G., and H. Lamm. 1976. The group polarization phenomenon.
Psychological Bulletin 83.606-27.
Noller, P. 1980. Misunderstandings in marital communication: A study
of couples' nonverbal communication. Journal of Personality and
Social Psychology 39.1135-48.
Noller, P. 1982. Channel consistency and inconsistency in the commun-
ications of married couples. Journal of Personality and Social Psy-
chology 43.732-41.
Norton, R. 1983. Communicator styles: Theory, applications, and
measures. Beverly Hills, Calif.: Sage.
Pearce, W. B., and F. Conklin. 1979. A model of hierarchical meanings
in 'indirect responses'. Communication Monographs 46.75-87.
Platt, 3., and H. Weber. 1984. Speech convergence miscarried: An
investigation into inappropriate accommodation strategies. Interna-
tional Journal of the Sociology of Language 46.131-46.
Price, S., M. Fluck, and H. Giles. 1983. The effects of language testing
on bilingual preadolescents' attitudes towards Welsh and varieties of
English. Journal of Multilingual and Multicultural Development 4.149-
61.
Ragan, S. L., and R. Hopper. (1984) Ways to leave your lover: A
conversational analysis of literature. Communication Quarterly.32.
Rands, M., and G. Levinger. 1979. Implicit theories of relationship: An
intergenerational study. Journal of Personality and Social Psychology
37.645-61.
Rausch, H., W. Barry, R. Hertel, and M. Swain. 1974. Communication
conflict and marriage. San Francisco: Jossey-Bass.
Roloff, M., and C. R. Berger. 1982. Social cognition and communica-
tion. Beverly Hills, Calif.: Sage.
Romaine, S. 1980. Stylistic variation and evaluative reactions to
speech: Problems in the investigation of linguistic attitudes in Scot-
land. Language and Speech 23.213-32.
Ryan, E. B., and C. M. Bulik. 1982. Evaluations of middle-class
speakers of standard American and German-accented English. Journal
of Language and Social Psychology 1.51-61.
Ryan, E. B., and H. Giles, eds. 1982. Attitudes toward language varia-
tion: Social and applied contexts. Baltimore: Edward Arnold.
Ryan, E. B., H. Giles, and M. Hewstone. 1984. Language and intergroup
attitudes. In: Attitudinal judgment. Edited by J. R. Eiser. New
York: Springer Verlag.
Ryan, E. B., H. Giles, and R. J. Sebastian. 1982. An integrative per-
spective for the study of attitudes toward language variation. In:
Attitudes toward language variation: Social and applied contexts.
Edited by E. B. Ryan and H. Giles. Baltimore: Edward Arnold.
Sillars, A. L., G. R. Pike, K. Redman, and T. S. Jones. 1983. Communi-
cation and conflict and marriage: One style is not satisfying at all.
In: Communication Yearbook 7. Edited by R. Bostrom. Beverly Hills,
Calif.: Sage.
Smith, P. M. 1980. Judging masculine and feminine social identities

from content-controlled speech. In: Language: Social psychological
perspectives. Edited by H. Giles, W. P. Robinson, and P. M. Smith.
Oxford: Pergamon.
Snyder, M. 1981. On the self-perpetuating nature of social stereotypes.
In: Cognitive processes in stereotyping and intergroup behavior.
Edited by D. L. Hamilton. Hillsdale, N.J.: Lawrence Erlbaum Associ-
ates.
Spanier, G. B. 1976. Measuring dyadic adjustment: New scales for
assessing the quality of marriage and other dyads. Journal of Mar-
riage and the Family 38.15-28.
Spiegelberg, H. 1973. On the right say 'we1: A linguistic and phenomen-
ological analysis. In: Phenomenological society. Edited by G.
Psathas. New York: Wiley.
Stephenson, G. M. 1981. Intergroup bargaining and negotiation. In:
Intergroup behavior. Edited by J. C. Turner and H. Giles. Chicago:
Stewart, M. A., and E. B. Ryan. 1982. Attitudes toward younger and
older adult speakers: Effects of varying speech rates. Journal of
Language and Social Psychology 1.91-109.
Street, R. L., Jr., R. M. Brady, and W. B. Putman. 1983. The influence
of speech rate stereotypes and rate similarity on listeners' evaluations
of speakers. Journal of Language and Social Psychology 2.37-56.
Street, R. L., Jr., and H. Giles. 1982. Speech accommodation theory: A
social cognitive model of speech behavior. In: Social cognition and
communication. Edited by M. Roloff and C. R. Berger. Beverly Hills,
Calif.: Sage.
Street, R. L., Jr., and R. Hopper. 1982. A model of speech style evalua-
tion. In: Attitudes towards language variation: Social and applied
contexts. Edited by E. B. Ryan and H. Giles. Baltimore: Edward
Arnold.
Suls, J. M., and R. L. Miller. 1977. Social comparison processes. New
York: Wiley.
Tajfel, H. 1974. Social identity and intergroup behavior. Social Science
Information 13.65-93.
Tajfel, H., ed. 1978. Differentiation between social groups. London:
Academic Press.
Tajfel, H., ed. 1982. Social identity and intergroup relations. Cam-
bridge: Cambridge University Press.
Tajfel, H., and J. C. Turner. 1979. An integrative theory of intergroup
conflict. In: The social psychology of intergroup relations. Edited by
W. C. Austin and S. Worchel. Monterey, Calif.: Brooks Cole.
Taylor, D. M., and L. Royer. 1980. Group processes affecting anticipat-
ed language choice in intergroup relations. In: Language: Social
psychological perspectives. Edited by H. Giles, W. P. Robinson, and P.
M. Smith. Oxford: Pergamon.
Tedeschi, J. T., ed. 1981. Impression management theory and social
psychological research. New York: Academic Press.
Tedeschi, J. T., and M. Reiss. 1981. Verbal tactics of impression man-
agement. In: Ordinary language explanations of social behavior.
Edited by C. Antaki. London: Academic Press.
Thakerar, 3. N., and H. Giles. 1981. They are—so they spoke: Noncon-
tent speech stereotypes. Language and Communication 1.255-61.
Thakerar, 3. N., H. Giles, and 3. Cheshire. 1982. Psychological and
linguistic parameters of speech accommodation theory. Advances in
the social psychology of language. Cambridge: Cambridge University
Press.
Turner, 3. C. 1982. Towards a cognitive redefinition of the social
group. In: Social identity and intergroup relations. Edited by H.
Tajfel. Cambridge: Cambridge University Press.
Turner, 3. C , and R. 3. Brown. 1978. Social status, cognitive alterna-
tives and intergroup behavior. In: Differentiation between social
groups. Edited by H. Tajfel. London: Academic Press.
Wiemann, 3. M., and C. W. Kelly. 1981. Pragmatics of interpersonal
competence. In: Rigor and imagination: Essays from the legacy of
Gregory Bateson. Edited by C. Wilder-Mott and 3. H. Weakland. New
York: Praeger.
Williamson, R. 1983. Relational control and communication in marital
types. Ph.D. thesis, University of Wisconsin-Madison.
Wilmot, W. W., and L. A. Baxter. 1983. Reciprocal framing of relation-
ship definitions and episodic interaction. Western 3ournal of Com-
munication 47.205-17.
Wish, M. 1978. Dimensions of dyadic communication. In: Nonverbal
communication. Edited by S. Weitz. New York: Oxford University
Press.
COMMUNICATIVE COMPETENCE REVISITED
John 3 . Gumperz
University of California, Berkeley
In this paper I seek to develop arguments first made in 'The linguistic
bases of communicative competence 1 , presented at the 1981 Georgetown
University Round Table on Languages and Linguistics (Gumperz 1981).
The earlier paper made an initial attempt to explore some of the theo-
retical consequences that arise when we analyze discourse coherence
from a speaker-listener oriented interactive perspective. The question
can be put as follows: what does an interactive approach to communi-
cation, in which problems of understanding are studied not in terms of
meanings inhering in a given text or stretch of discourse, but rather as
outcomes of inferential judgments made in the course of situated pro-
cesses of conversational exchange, imply for our theories of communica-
tion and language use?
Conversing, as we all know, is a cooperative activity that involves
active participation and coordination of moves by two or more partici-
pants in the joint production of talk. There are good reasons to believe
that such coordination presupposes verbal abilities and types of knowl-
edge that are not as yet fully understood. Neither the theoretical lin-
guists' grammatical analyses, nor the commonly accepted sociolinguistic
studies that seek to formulate rules of language usage covering what can
be said when and under what circumstances, can account for this knowl-
edge. Nor do notions of schema or script defined in static terms as
extralinguistic knowledge of the world explain how such information
enters into discourse understanding.
When we look at problems of understanding from a participant's per-
spective, we see that what we must explain are on-line processing stra-
tegies. Conversationalists employ strategies in inferring the contextual
presuppositions about what is expected in an encounter. Related strate-
gies are also used in segmenting the stream of talk into information or
idea units, in determining the transition-relevant points for turn taking,
and in integrating what is said at various points in time into coherent
themes. Empirical investigations of conversational exchanges have led
to the discovery that in making the judgments relevant to these tasks,
278
Communicative competence revisited / 279
participants depend on their own perception of stylistic and prosodic
signalling cues that have hitherto not been seen as having semantic
import, and that are, thus, not ordinarily covered in sentence-level
linguistic analysis. These features of speech performance are processed
in accordance with contextualization conventions that retrieve schemat-
ic information and make it available as an input into the interpretive
process.
My claim is that the capacity to contextualize, and thus make sense of
what is heard in terms of what is already known, is governed by cogni-
tive abilities that share many of the characteristics of grammatical
competence. They are conventional in nature, learned as part of the
everyday language socialization processes, and once internalized, they
are usually employed automatically without conscious reflection. In
principle, therefore, these processing abilities should be analyzable by
in-depth qualitative methods similar to those employed in grammatical
analysis. But conversational processes also have special characteristics
of their own that derive from the very nature of conversations as multi-
party interactive performances. These properties require us to look at
the multiplicity of linguistic signs involved in conversing from a differ-
ent perspective.
To begin with, although it is true that all conversations are governed
by general and in large part universal organizing principles, these princi-
ples operate in a manner that is quite different from the operation of
all-or-none categorical grammatical rules. Conversational principles, as
Levinson (1983) has argued, are defeasible; that is, they do not deter-
mine what counts as an utterance in a language, or what can, or cannot,
be said or understood. On the contrary, they act as guidelines or stand-
ards of evaluations that give rise to expectations which, when violated,
generate the implicatures on which rests interpretation of so much of
what a speaker intends to convey.
Second, the phonetic, prosodic, and stylistic cues that participants in
face-to-face encounters rely on in contextualizing their performances,
are typically quite fleeting or transitory in nature. Since the relevant
information is not coded in lexical form, it becomes hard to retrieve
after the event. It cannot easily be communicated in contexts other
than those in which it originally occurs, so that recalling what was
actually perceived at any one time for the purpose of later analysis and
preservation of information in the form of adequate transcripts, presents
a major problem. In fact, until about ten years ago, when unobtrusive
means for recording everyday talk first became available, we simply did
not have the data necessary for systematic investigation.
Thus, while theory suggests good reasons to believe that in-depth
methods of analysis patterned on those employed in the study of spoken
language can yield insights into conversational processes not otherwise
obtainable through quantitative correlational techniques, the phenomena
to be studied also present by their very nature serious empirical and
analytical problems. It seems clear that although conversationalists
build on phonological, syntactic, and semantic knowledge in contex-
tualizing what they hear, contextualization conventions are as distinct
from sentence-level linguistic rules as phonology is from syntax, and
syntax from semantics. They must therefore be analyzed on their own
terms.
280 / John 3. Gumperz
What, then, does a qualitative, participant-centered approach to the
study of conversational phenomena entail? In arguing for methods of
analysis patterned on those employed in linguistics, I mean to suggest
that the analytical goals must parallel the linguist's concern with struc-
tural determinants of grammaticality. In other words, conversational
analysis should focus on the conditions that make possible shared inter-
pretation, rather than seeking to predict correct or appropriate usage.
For this reason, I proposed that the notion of communicative compe-
tence be redefined as: 'The knowledge of discourse processing conven-
tions and related communicative norms that participants must control as
a precondition to being able to enlist and sustain conversational coopera-
tion.' In what follows I would like to discuss what this approach to
communicative competence implies for the study of the issues of mean-
ing, form, and use in context that form the subject of this Round Table,
and for enabling us to integrate into the study of discourse processing
and discourse understanding the ethnographer's insights into culture and
cultural variability and context.
Let me begin with some background. The notion of communicative
competence was first proposed by Hymes in the context of the 1960s
debate on the limits of formalization in linguistics. In order to highlight
the role that cultural and linguistic variability play in speaking, Hymes
argued that communication in the sense of engaging in meaningful
interaction with others is a function of membership in a speech commu-
nity, and not simply a matter of grammatical competence. Speech
communities are human collectivities held together by shared history
and long-term participation in networks of relationship. Ethnographic
evidence shows that such communities are frequently not identical with
collectivities defined by control of grammatical rules (Gumperz 1972).
We cannot therefore assume that our studies become more socially
relevant by correlating sentence-level linguistic categories with inde-
pendently determined social variables. The extent to which culture is
shared in relation to speaking must be empirically determined by exam-
ining the norms, values, and ecological constraints on behavior as they
function in human groups.
Yet no matter how we select the population units to be studied, pat-
terns of language usage are never quite uniform. Apart from individual
performance factors, language usage varies with boundaries of class,
gender, ethnic allegiance, power relationships, and education. All these
are factors which, in the rapidly changing societies of today, are often
quite resistant to measurement. Even in small communities that appear
as relatively homogeneous, economically and socioculturally uniform
entities to the outsider, detailed ethnographic studies tend to reveal
sharp divisions among members who do, and members who do not, share
a sense of local identity. Thus, models of analysis that assume the
existence of clear, stable social boundaries will necessarily have diffi-
culty in specifying the social motivations of language behavior.
There are at least two interrelated and partially independent dimen-
sions of variabilities that we must account for in the study of cultural
sharing: the societal and the contextual. A major concern of ethno-
graphy of communication has been to clarify the relationship between
linguistic and sociocultural categories through comparative studies along
both of these dimensions. Given the facts of variability, a human
community defined in purely geographical terms cannot be the unit of
analysis for this comparative endeavor. Ethnographers of communica-
tion therefore chose to focus on speech events, culturally defined units
of interaction bounded in time and space, such as ritual performances,
ceremonies, public meetings. Such culturally sanctioned units of social
interaction stand apart from everyday talk, and thus form a convenient
starting-point for studying the interplay of social and linguistic aspects
of communicative behavior. To account for the role of culture and
social norms in speaking, it seemed useful to treat events as if they were
miniature social systems, that is, as if one could speak of norms of
conduct, constraints on choice of communicative content and on roles
that participants can play, which function like the norms of human
communities. The goal of speech event analysis was to formulate rules
of appropriateness showing how speaker's choice from a range of stylis-
tically alternate expressions relates to the norms governing behavior in
such events.
The data obtained in this type of study have been important in provid-
ing basic background information on hitherto little understood sources of
variability in speech form, as well as in speech function, and in demon-
strating that the analysis of linguistic form cannot ultimately be ab-
stracted from cultural considerations. But the information we have so
far tells about the role of cultural factors in what Hymes calls the
communicative economy of human groups. It specifies the knowledge
that is potentially available, but does not tell us what aspects of this
knowledge are actually used in communicative situations, and how they
are brought into the communicative process to affect the interpretation
of what transpires. If the claim that cultural knowledge forms part of
communicative competence is to have anything more than metaphorical
significance, we have to find ways of looking at verbal data that enable
us to work out empirical procedures for testing hypotheses about how
our understanding of what is said at any one point is affected by cul-
ture. For this reason we need more detailed insights into how verbal
encounters actually work.
By far the largest and most exhaustively described body of informa-
tion on conversation comes from the work of sociologists working in the
ethnomethodological tradition. Ethnomethodologists were the first to
look at conversation as social action. Their studies provide empirical
evidence to show how the sequential organization of turns at speaking
can serve to constrain conversationalists' ability to make themselves
understood, to introduce, establish, and change topics, and otherwise to
affect the course of a conversation. But their goals were quite different
from the ethnographer's concern with showing that sociocultural factors
are important in communication. Their work must be understood within
the context of the internal sociological debate over the empirical valid-
ity of commonly accepted social science generalizations. Symbolic
interactionists, for example, tended to limit their analysis of human
encounters to categories of roles, statuses, and motives, derived from
theory alone, without raising the question of the relationship of theory
to practice. To bring out the limitations of this type of a priori theo-
rizing, verbal data were collected to provide empirical evidence of how
social control was exerted in everyday interaction.
Since speech act analysis could be seen as an instance of such a priori
theorizing, its social premises were clearly unacceptable. Instead,

ethnomethodoiogists focused on everyday talk which lacked the formal
constraints that characterized the ritual and ceremonial occasions that
speech act analysts tended to study. As Schegloff (1981) puts it:
Quite aside from whatever individual cognitive or processing
achievements might be involved, the production of a spate of talk
by one speaker is something which involves collaboration with the
other parties present and that collaboration is interactive in char-
acter and interlaced throughout the discourse. That is, it is an
ongoing accomplishment rather than a pact signed at the beginning
after which discourse is produced entirely as a matter of individual
achievement. The character of this interactional accomplishment
is at least in part shaped by the sociosequential organization of
participation in conversation, for example, by its turn taking organ-
ization...
I want to argue here that while the sociosequential organization of
speakers' moves is as basic to conversational analysis as clause bound-
aries are to syntax, conversational analysts have gone only part way to
giving us an understanding of conversational processes. What they have
done is to give us some of the basic constraints that affect the working
of conversational processes in societies of all kinds. But these con-
straints, like the linguist's grammatical rules, are in large part universal
and thus abstracted from actual situations of usage. To see how com-
munication works on the ground, we cannot look at organizational fea-
tures of communication alone, but we must look in detail at the actual
processes of conversational inference, that is, the situated interpreta-
tions that participants make of each other's moves at any one point in
time. Let me illustrate with an example of the type commonly analyzed
by conversational analysts.
(1) Making arrangements for lunch (Levinson 1983:316-17).
1. R: Why don't we all have lunch

2. C: Okay so that would be in St Jude's would it?
3. R: Yes (0.7)
4. C: Okay so:::
5. R: One o'clock in the bar
6. C: Okay
7. R: Okay?
o.
O C: Okay then thanks very much indeed George
9. R: = AU right
10. C: //See you there
11. R: See you there
12. C: Okay
13. R: Okay // bye
14. C: Bye
This exchange constitutes the concluding portion of an encounter, the

rest of which is not reproduced here. R proposes a luncheon meeting and
C accepts. They agree on a time and place and then say good-bye. What
readers of the transcripts are likely to find odd is the form of the
leave-taking and the amount of time it takes. The substantive portion of
the event is completed in six turns while the departing formalities
extend over eight turns, more than half of the total sequence. C's Okay
in turn 6 is followed by R's Okay, pronounced with question intonation.
This evokes a formulaic expression of gratitude on C's part, presumably
an allusion to some favor obtained. R's acknowledgment in 9 is followed
by a simultaneous exchange of see you there, followed by another set of
Okay and a set of bye.
Levinson (1983:316-17), in discussing this example, suggests that it
might reflect a common general schema for closing sections, consisting
of four basic stages: (1) a closing down of some topic, such as the
making of arrangements in turns 1-5; (2) one or more passing turns (turns
6 and 7); (3) a typing of a call followed by a further exchange of pre-
closing items; (4) a final exchange of terminal elements. Yet neither
Levinson in his discussion of ethnomethodological approaches to conver-
sations, nor Schegloff and others who have done the primary research,
deal with the theoretical status of such notions as schema and nego-
tiation. Nor do they raise the question of how findings from the empir-
ical study of situated conversational exchanges can be integrated into
what we know about discourse understanding.
Most reports of conversational analysts' findings take the form of
empirical statements about patterned features of language usage occur-
ring in a variety of conversational texts that can be shown to have a
significant effect on the course of an interaction. The most frequently
cited examples of such regularities are the so-called adjacency pairs:
question-answer, greeting-acknowledgment, offer-acceptance, and the
like. In characterizing the regularities which underlie the occurrence of
such pairs, Sacks and Schegloff (1973) suggest that they reflect conver-
sational rules such as the following: having produced the first part of
some pair, the current speaker must stop speaking and the second speak-
er must at that point come up with a second part. Yet the status of such
rules is not made clear. Does the schema of example (1) reflect such a
rule? Given Schegloff's view of conversation as an 'ongoing accomplish-
ment', which in large part is created by the local sequences of moves and
countermoves, we would not want to say that we are dealing with a fixed
script that applies to sequences of all kinds. Such a script would hardly
require negotiation of transitions. Moreover, sequences such as conver-
sational closings, extending as they do over a number of moves, are
context- and to some extent culture-specific; thus they contrast with
adjacency pairs, which occur in conversations of all types.
A more reasonable position, and one that does not inherently conflict
with the conversational analysts' basic arguments, would be to assume
that schemata such as the one Levinson suggests specify knowledge
acquired through previous communicative experience of what could
potentially happen in situations like the one at hand. The actual on-line
processing and production strategies through which the closing sequence
is produced could then be regarded as the outcome of local inferential
processes to which schematic knowledge is one of several inputs. A
closer examination of both the form and the content of the moves and
countermoves that make up the exchange provides some insights into the
workings of this inferential process.
28* / John 3. Gumperz
After R's confirmation of the meeting in turn 3, the encounter is
interrupted by a pause. C then continues the interaction with Okay
so::; R's answer in 5 indicates that he interprets this as a request for
more information. C's Okay confirms this interpretation. When R
follows with a questioning Okay?, C counters with his formulaic expres-
sion of gratitude, which is then confirmed by R's all right. What do R
and C have to know in order to respond in such a way as to have their
moves understood? Seen in purely referential perspective, turns like
Okay so::: and Okay can perhaps be regarded as contentless. But the
form in which they are spoken—that is, both the elongated vowel and the
question intonation used—does have communicative import. They can be
paraphrased as queries like: is there anything else to be said? In turn 5
the answer is affirmative, whereas in turn 8 it is negative.
In arguing that message form communicates, I do not mean to say that
features such as prosodic contours have meanings. As Bolinger (1972)
points out, intonation contours are not meaningful elements in the
linguist's sense of the word. What we are talking about here are con-
text-bound judgments; such judgments build on schematic knowledge
generated through previous talk, as well as knowledge of contextualiza-
tion conventions, grammatical and semantic knowledge, and participants'
understanding of the sociosequential positioning of the utterance at
hand. These are matters of discourse-level interpretations, not matters
of meaning as such.
Examples (2) and (3) provide some further insights into the linguistic
and interactional nature of these interpretive processes.
(2) Secretary (B) talking to recently appointed research assistant

(A), who has just entered the building.
1. A: Good morning.
2. B: Hi 3ohn.
3. A: Howdi.
k. B: How're ya doin1.
5. A: Fine... ah... do you know... did you get anything back on
those forms ah... you had me fill out?
6. B: Hm... like what?
7. A: I wondered if they sent you a receipt or anything or a
copy ok.
8. B: LYou mean your employment forms?
9. A: Yea.
10. B: Yea, I kept a copy. Why is there a question?
11. A; 'Cause I just left it with the Anthropology Department.
12. B: Oh, that's O.K., they'll just send them over to L and S and
they'll send them on.
13. A: O.K., because the chairman wasn't there and...
14. B: Oh, she called me later... I guess it was you. She said
that a student had been over and that she was just going
to hold on to the forms and then send them. And she sent
you on, you know... She told you to go ahead and leave
them and not wait around.
15. A: I just wanted to make sure they're O.K.
16. B: Oh yea. Don't worry about it.
The opening of this encounter is an example of what ethnomethodo-

logists have called a presequence. The second of the two greeting
exchanges is treated as a bid for further talk. The remainder of the
interaction is typical of informal office enquiry or information se-
quences. A wants to know if his employment forms have been processed
and B answers by telling him something about what the procedures are.
Yet, even though the exchange, when seen in its entirety, seems famil-
iar, the way in which the two participants go about carrying out the
activity makes it seem highly unlikely that either of them has ap-
proached the encounter with a predetermined plan or script. On the
contrary, the interaction moves forward step by step. Each move is
replied to and evaluated before the next one is initiated.
The initial exchange through which entry into the encounter is nego-
tiated also sets the tone for what follows. When B counters A's rather
formal good morning with an informal hi, A follows suit and B continues
in the same informal mode. It appears that they have agreed upon a
tone of informality which is then maintained throughout the encounter.
Note that while sequential organization is crucial in the negotiation
process, this informality and the indirectness it entails also make de-
mands on the participants' ability to infer information not directly
reflected in the lexical content of what is said. After A's somewhat
vague enquiry in line 5, B asks for more clarification. When A's reply is
similarly vague, she guesses that he is referring to his employment
forms. Her guess is confirmed and she assumes he wants to know if she
kept a copy of the forms in her office. Her misunderstanding is cleared
up through further indirect inferences without overt reference to the
need for repair. A's statement that the chairman was absent then
evokes a narrative description of what happened with the form.
Throughout the interaction, A has pursued a strategy of not providing
detailed information and relying on A's ability to guess. Why he should
do this is outside the scope of this, analysis. Yet what is important is
that speaker and listener are able to make these indirect inferences and
have them understood, and that they do this by appealing to shared
knowledge of previous interactions and to organizational knowledge of
the workings of the university administration.
(3) An excursion (Falk 1979).
1. P: Yeah^vhere'd you go,v

where'd you all go.
v
act N
2. J: Well, y'wanna tell em, where we went?
3. R: We went uh... almost everywhere, except that it was
fbggy, for half the time we were over theje
4. M: Yeah, you don't sêm too enthusiastic about it.
r
5. R: Oh it was a good trip, yeah, it was, yâh.
6. 3: Well, 'it was a great trip, except that it was a foggy
day. and we started out by going to Twin Peaks,., at 9:30
in the morning. ..on a foggy day. you know what we saw,
we saw fog,,up to here.
7. R: *- We saw a fire down the hill.
8. 3: C A fire
down the hill, that's right.
9. R: Hfhere was a (f ) was on fire.
10. P: That was good ( ).
11. J: '-And then after the fog, and it was windy,
we., decided to go to the beach, we went to the beach
and it was, we saw some sea lions, though,
L
12. R: There were=
13. J: Then we had=
R: =about threê.
3: =ho.. hot dogs on a stick.
1*. M: Lun_huh.
15. R: SSme of them, some of them, had hot dogs on a st^ck.
16. M: ( ) Hi G. (G. enters)
17. 3: Hi G.. . and uh, we uh.. , talked ( ) about the tourists
there, and we le^ft, and went to the Palace,., of the
Legion of Honor, rfght?
18. R: *- Again, for a view of the Golden
Gate.
x
[
19. J: For a view, of the Golden Gate Bridge, and there=
20. G: Aaaaah.
3: =was no view.
The following transcription conventions

x
are used in this example:
\ low falling contour x
high falling contour
, low rising contour high rising contour
(The rpreceding accent marks are doubled for extra loudness.)
register shift upward ^register shift downward
aee. accelerated tempo "^latching
( ) unintelligible speech
The content of this exchange readily identifies it as an informal
narrative. Two ESL instructors, 3 and R, have taken a group of students
on an excursion to San Francisco to see the sights and are telling their
colleagues about the trip. Yet the structural characteristics we nor-
mally associate with narrative are nowhere overtly marked. There is,
for example, no overt introduction, no initial summary statement of
what the trip was about. In response to the question as to where they
went, R answers we went almost everywhere, and then goes on to say
that while it was foggy it was a good trip. J then makes the transition
into the substantive part of the narrative by first repeating R's state-
ment, and then shifting to slow rhythm and starting to list the things
they did. The exchange has some of the characteristics of a picaresque
narrative, inasmuch as the action moves forward by the way things are
said, rather than by what is said. The trip is described in a manner that
seems unusually vague and lacking in detail. Mention is made of some of
the sites visited (such as Twin Peaks, the beach, and the Palace of the
Legion of Honor), of what they saw (a fire, fog, sea lions) and of what
they ate (hot dogs on a stick). There is no attempt to explain what is
interesting about all this. Considering the fact that speakers and audi-
ence are all local residents, the list sounds strangely uninformative. Yet
no one queries the account or asks for more detail. On the contrary, in
line 10, after the third mention of the fire, P remarks that was good.
All this seems at first to be a violation of Gricean principles of rele-
vance. It is only in lines 13-17, where we in J's statement we had hot
dogs on a stick is corrected by R to some of them, that we begin to see
the significance of this apparent violation of Gricean principles of
relevance: even though J and R speak in the first person plural, the
story is not being told from the narrators' perspective, but from the
students' perspective. There is nothing in the surface form of the de-
scription to indicate this.
Note, also, that the narrative is not one person's performance: it is
jointly produced by J and R. The transcript reveals many instances of
what Falk (1979) has called duetting, where two speakers complete each
other's thoughts so that their utterances together make up a single turn
at speaking. For example, R's we saw a fire... in line 7 is latched to J's
...up to here in 6. That is, the two follow each other without the
rhythmic interval that normally marks transitions between turns. The
same is true for J's line 8 and R's lines 9 and 12. Lines 17-19 are
similarly cooperative, although there is no latching. Furthermore, J and
R regularly employ parallel syntactic constructions: their lexical
choices are stylistically congruent, with one person often anticipating
what the other has to say. It would be difficult to imagine that two
individuals could be as successful at managing this type of cooperation,
and entering into the spirit of what the other intends to convey, without
shared frames of reference and some level of shared culture.
The three examples illustrate the role that turn-taking and negotiation
of situated interpretations play in everyday encounters. But they also
show that conversing involves more than context-free principles of
sequentiality that can be applied by anyone who knows the language.
Participants do not enter into the negotiation process relying only on
their command of grammar and lexicon. The ability to enter into con-
versations, that is, to fit one's contributions into locally established
themes and lines of argumentation, depends on largely indirect inferen-
tial processes which draw on knowledge of the world to make sense of
what is said.
Culture enters into this process in two ways. It is an integral com-
ponent of what discourse analysts call schematic knowledge. Although
288 / 3ohn J. Gumperz
we customarily think of schemata as ways of organizing factual

information in terms of basic conceptual structures, conversational
analysis shows that assumptions about norms, interpersonal relationships,
and interactive or communicative goals are also involved. When seen in
this perspective, schemata come to take on forms that bear great
similarity to the ethnographer of communication's speech events, in that
they reflect values and experiences acquired through participation in
culturally bounded networks of social relationships. The difference is
that they do not predict action or determine judgments of what is appro-
priate. They must be seen as cognitive constructs that give rise to
expectations about what is to be looked for in an encounter, which in
turn enter into our interpretations of what transpires.
Cultural background and culturally channeled interactive experience
also determine the acquisition of the contextualization conventions by
virtue of which the choice among various performance features acquires
situated significance, for example, the choice between hi_ and good
morning in example (2), the prosodic contour of okay in (2), the pausing,
utterance latching, and pitch register and tempo alternations that mark
all three conversations.
It must be emphasized that what we are talking about here are not
context-free meanings. Interpretations can be seen as the outputs of
inferential processes, which have as their input knowledge of lexicon,
schemata, and control of contextualization conventions. Note that while
grammatical knowledge and control of the basic lexicon is shared, other
conventions entering into conversational inference are differentially
distributed in accordance with social and occupational boundaries, and
communicative experience. It is the interplay or tension between con-
text-free, clause-level linguistic competence, on the one hand, and
situation and subculturally specific knowledge of the discourse level
conventions that enter into communicative competence, on the other
hand, and the implication of this interplay for learning and understanding
in everyday communicative situations, that promise to become one of
the most fruitful areas for future sociolinguistic research.
References
Bolinger, D.L., ed. 1972. Intonation. Harmondsworth, England: Pen-
guin.
Falk, 3. 1979. The duet as a conversational process. Ph.D. dissertation,
Princeton University.
Gumperz, 3.3. 1981. The linguistic bases of communicative compe-
tence. In: Georgetown University Round Table on Languages and
Linguistics 1981. Edited by Deborah Tannen. Washington, D.C.:
Georgetown University Press. 323-34.
Gumperz, 3.3., and D. Hymes, eds. 1972. Directions in sociolin-
guistics. New York: Holt, Rinehart and Winston.
Levinson, S. 1983. Pragmatics. Cambridge: Cambridge University
Press.
Sacks, H., and E. Schegloff. 1973. Opening up closings. Semiotica
7.4:289-327.
Communicative competence revisited / 2X9
Schegloff, E. 1981. Discourse as an interactional achievement: Some
uses of 'uh-huh-1 and other things that come between sentences. In:
1981. Edited by Deborah Tannen. Washington, D.C.: Georgetown
PHONOLOGICAL STYLE IN BILINGUALISM:
THE INTERACTION OF STRUCTURE AND USE
Susan Gal
Rutgers University
1. Various types of massive language change, such as pidginization,
creolization, language acquisition, and language death have become
particularly interesting for recent linguistic theories, not only because
of what they promise to reveal about human cognitive processes, but
also because they occur as responses to equally great changes in the
social use of linguistic varieties (Halliday 1973, Hymes 1971, Slobin
1977). My aim in this paper is to clarify the complex relationship be-
tween linguistic structures and social uses by looking at how one of these
factors changes in relation to changes in the other. In a bilingual com-
munity undergoing language shift, as one of the languages expands in
use, the other language is utilized for ever fewer communicative tasks in
an ever narrower range of social contexts and in ever fewer role rela-
tionships. Such a community, therefore, provides a partially controlled
situation in which to examine how speakers' knowledge of a linguistic
variety is affected by limitations on the opportunities and motivations
for its use. Thus this study differs from most of the research on the
effects of bilingualism on linguistic structure because it does not inves-
tigate the influence of the two linguistic systems on each other (inter-
ference), but explores instead the effects on linguistic structure of
different patterns and contexts of use.
I propose to show that, in the first-learned but less used language,
stylistic variation in phonology is often substantially simplified during
language shift. However, in studying the interdependence of use and
structure it is just as important to outline the social and communicative
constraints as to describe the formal linguistic consequences or cor-
relates. Therefore, I look to speakers' patterns of language use and of
social interaction, as well as the social meanings of their linguistic
choices, in order to understand why for some bilinguals phonological
style is narrowed in their first-learned language while for other bilin-
guals it is not.
290
Phonological style in bilingualism / 291
Emphasizing differences between bilinguals within a single community
allows me the opportunity to point out yet another way in which the
quasi-theoretical notion of the 'native speaker' is problematical. The
'native speaker' has long been a useful theoretical idealization which
fuses the idea of exposure to a language from birth with the quite dif-
ferent concept of a natural or complete command of a language. No
matter how common the cooccurrence of these two phenomena actually
is, it is important to emphasize that conceptually they are distinct and
not inseparably mated. The language one acquires in early childhood is
not necessarily the same as the language one later learns to speak with
facility and expressive power. The two are united or divorced by histor-
ical contingencies. It is these, and not any linguistic factors, which
determine whether or not speakers have access to the social contexts
and institutions in which phonological styles gain their rhetorical effect-
iveness. Fusing the two allows us to see language largely as a cognitive
phenomenon and thus slight the role of history and social structure in the
formation and loss of linguistic skills by individuals and communities.
Another aim of this paper, then, is to highlight the disjunction veiled by
the 'native speaker' concept by showing stylistic reduction in a language
that is learned first, used daily, but is limited in its social functions.
'Stylistic reduction' implies a definition of style. This is a thorny
issue, even in the relatively straightforward and narrow case of phono-
logical variants (see Romaine and Traugott 1981, Irvine 1979). Here I
have taken a strategic approach which implies that in attempting to
gauge a speaker's stylistic repertoire the analyst is required to
understand the meanings, intents, and effects of choices between
referentially equivalent phonological variants. It is not enough to note
the components (e.g. setting, participants, topic) of the situations' in
which variants occur.
While many broadly stylistic strategies which are well documented for
bilinguals, such as code-switching, will not be considered directly in this
paper, the phonological styles which are the focus here have an import-
ance of their own. It should be remembered that a reduction in phono-
logical styles is not simply dispensable icing missing from the language
cake. From the point of view of linguistic structure, studies by Labov
(1972a) and others on sound change show that stylistically significant
variation is centrally involved in the structural rearrangement of phono-
logical systems. From the point of view of the speaker, such variation,
while referentially neutral, is necessary in conveying essential social
meanings about the identity of the speaker and the nature of the social
situation. Further, Gumperz's work (1982; Blom and Gumperz 1972)
indicates that such variation can also be used to express momentary
communicative intent, so that in a community that uses them, speakers
without phonological styles may, in some circumstances, sound less
subtle verbally and risk social misunderstanding.
Recent studies of language death, working within a broadly functional-
ist framework, have adopted a form of the research strategy I have
outlined here. They have selected and compared speakers within a single
bilingual community who know the nearly obsolescent language to dif-
ferent degrees. The works of Dorian (1981) on East Sutherland Gaelic,
Hill and Hill (1977) on Nahuatl, and Dressier (1972, Dressier and Wodak-
Leodolter 1977) on Breton have been exemplary in examining the
292 / Susan Gal
grammatical and lexical structures used by speakers of such languages.

But, although Hill and Hill (1978) have explored changes in honorifics,
there has been little attention paid to the possibility of changes in the
variable, stylistic part of the phonological system.
The few studies which have addressed this issue are variationist anal-
yses of bilingual repertoires. The most relevant work is Lavandera's
(1978) paper showing that in their second language, Spanish, a commu-
nity of Italian migrants in Buenos Aires fails to exploit one of the most
salient phonological variables. These fluent bilinguais use only one
variant of a phonological variable which, in its several realizations,
marks socioeconomic status and formality of the speech event in the
speech of Spanish monolinguals. Lavandera argues convincingly that
since these bilinguais have available the strategy of switching between
languages to express nonreferential meanings that monolinguals must
express through phonological styles, their phonological styles remain
undeveloped in their second language. However, it is important not to
assume from this that we are dealing with a structural complementarity
between two kinds of stylistic expression, such that where one exists the
other will not survive. Rather, I suggest that such complementarity is
but one of several sociohistorical possibilities. The differentiation of
phonological styles is absent in some cases of bilingualism, even in a
first language, as I will show. And language-switching is used by such
bilinguais to express social meanings. However, my evidence suggests
that the simple availability of language-switching as a stylistic device
does not in itself entail lack of phonological range. In fact, some bilin-
guais exploit both language-switching and phonological variation for
stylistic effects. The difference between those who do and those who do
not is their location in the community's social structure and the conse-
quent role and meaning of each language in their daily communication.
2. The Hungarian-German bilingual neighborhood in which this study

was done is located in a town within Austria very near the border of
Hungary. For about 400 years it has been one of three Hungarian-speak-
ing enclaves completely surrounded by German-speaking villages. But
only in 1921 did the area actually become part of a German-speaking
polity. At that time the territory, which had previously been part of
Hungary, was attached to Austria. Today, although the town itself is
largely German-speaking, the neighborhood of about 2,000 people is
bilingual, in a variety of ways. Both German and Hungarian are in daily
use within the Hungarian section of town. All bilinguais have Hungarian
as a home language; most read German newspapers. Hungarian reading
material is extremely scarce. All listen to German radio and TV. Many
older people also listen to Hungarian radio, but although Hungarian TV is
available at a slight expense only some musical and sports events attract
large numbers of bilingual viewers.
Importantly, speakers are divided by education: those over about 35
have Hungarian, those under 35 German elementary schooling. From
1921 until 1955 the local elementary schooling was conducted in Hun-
garian but with German lessons. By the early 1950s some bilingual
parents were sending their children to a German-language school. After
1956 all schooling was conducted in German and optional weekly Hun-
garian lessons were not instituted until the late 1960s. Another political
note of lingustic importance is that the Hungarian border was virtually
closed to traffic in labor, goods, and tourists from the end of World War
II until the late 1960s. Only in the last decade has the crossing into
Hungary been relatively free of red tape.
The material discussed in what follows is part of a larger participant-
observation study that also included systematic collection of information
about the language usage patterns and social networks of a sample of
speakers chosen to represent a range of age, occupation, and network
type within the bilingual neighborhood. The phonological styles of 20 of
these speakers will serve to illustrate some of the patterns found in the
community. The usage of these speakers does not exhaust the variation
in phonological styles in the bilingual neighborhood, but it does represent
two major patterns. Since it is the social experiences of speakers that
are to be related to their linguistic strategies and abilities, it will be
essential here to outline the social histories of these speakers. I will
concentrate on two factors: (1) their social identities as indexed by age,
occupation, and some aspects of their social networks; and (2) their
patterns of choice between German and Hungarian.
The 20 speakers can be divided into two age groups, with men and
women represented in each. The 10 older speakers range in age from 69
to 84 years. The younger speakers range from 23 to 40 years. This age
difference provides an accurate indication of acquisition history and
occupation. All speakers learned Hungarian first at home. The younger
group started learning German by the age of five, either in nursery or
from parents and siblings. The older group started to learn German by
about the age of 12 or 15, usually by being sent to a nearby German
village to serve as farmhand or domestic servant. The older people are
farmers, now mostly retired, and each has been or still is active in some
way in the local church or government. Those in the younger group all
come from farming families, but have a range of occupations: farmer,
housewife, several waitresses, several salespeople, a hairdresser, chauf-
feur, bookkeeper, postal worker. The important thing about these most-
ly service occupations is that they do not differ enough in income,
conditions of work, or prestige to create barriers of status or social
network among the young people. Many of the occupations require
exclusive use of German on the job. These young people also take a
relatively active role in public life, mostly in church activities or in
clubs and volunteer rescue squads.
What separates the younger group from the older most decisively is
their pattern of language use. Those in the younger group use Hungarian
only to their parents (and this not always) and to those in their grand-
parents' generation. With their siblings, age-mates, work-mates,
spouses, and children they use German, even if these people are them-
selves bilinguals. In sharp contrast, those in the older group use Hun-
garian to everyone except, of course, monolingual German speakers with
whom they must interact in service encounters or in the state bureau-
cracy, and sometimes with their own grandchildren.
All speakers sometimes engage in conversational language-switching,
using German remarks in an otherwise Hungarian interaction for com-
municative effect. All the cases of such language-switching which I
have analyzed in1 this community can 1be interpreted as the juxtaposition
of 'low, in-group and 'high, out-group codes, to allow inferences of, for
29* / Susan Gal
instance, interpersonal authority or expertise, or to deliver the final
winning retort in a hostile exchange. These interpretations depend on
the fact that within the bilingual neighborhood Hungarian now tends to
carry connotations of peasant agriculture and the past, while German is
decidedly the more prestigious language related to mobility out of
agriculture into wage labor and education (see Gal 1979).
The present source of prestige for German is both economic and
political. Although older people remember a time when Hungarian was
the langauge of the cultural and political elite and was used in schools
and offices, the language of the town's well-to-do business and profes-
sional stratum is now German, and German is also the language now
supported by virtually all educational and other state institutions. The
current difference in prestige between the two languages is reinforced
by a widespread fear of Hungarian communism and by the substantial
difference in standard of living between Austria and Hungary which
occasional tourist trips to Hungary make evident to everyone. In addi-
tion, there is the often hostile reaction of German monolinguals to the
use of Hungarian in their presence and even to Hungarian accents. For
the younger group, which has moved out of agriculture and which has no
experience of Hungarian as a prestige language, the use of Hungarian is
much more limited, and more limited in value, than for the older group.
In sum, the two age groups contrast in their history with the languages
and in the range of social relationships in which they now use Hun-
garian. They could as well be labeled the broad users and the narrow
users.
3. Before the speakers can be compared, it is necessary to describe
their linguistic varieties and the methods used in this study to assess
speakers' stylistic ranges. Although interviews in German were also
conducted, I restrict the analysis here to the Hungarian portion of the
repertoire.
The local forms of Hungarian constitute a relatively divergent dialect
of Hungarian which nevertheless shares many features with other west-
ern Hungarian dialects, but also shows the centuries-long influence of
German. A likely example of the latter is the frequent loss of the
voiced/voiceless distinction in final stops, matching the lack of this
distinction in the surrounding German dialects (Keller 1961). An exam-
ple of the former is the virtually universal use of a long high front vowel
[i] where standard Hungarian has a long mid front vowel OD (Imre 1971).
Many of the differences between standard Hungarian and the local forms
show no variation between speakers and between situations. They are
easy to spot on a single visit, sound regional to the urban Hungarian, and
identify the speaker as provincial.
More important for my purposes, however, were those features of the
local Hungarian dialect which seemed to carry social meaning for the
speakers themselves: what Labov (1972a) has called 'markers'. The
candidates were those local forms which had been mentioned in earlier
descriptions of the dialect but which I did not hear in my first weeks of
field work, when my initial contacts were limited to church-related
public events and interviews with older speakers. This indicated that
some local forms are not ordinarily used by older speakers in speech to
outsiders such as myself or to occasional visitors who speak Hungarian. I
first noticed the local forms of these variables in conversations I over-
heard between family members and neighbors. This is, of course, the
classic situation in many European dialects where the more standard
forms take on, by association with their urban users and their contexts
of use, connotations of publicness, outsideness, and the particular kind of
prestige associated with education and social mobility (Trudgill and
Chambers 1980). Notice that there is a parallel here with some of the
contrasting connotations of German and Hungarian.
In attempting to record systematically, I first wanted to create a
situation in which the seriousness, publicness, and on-recordness or
outsideness of the circumstances would give speakers the motiviation—
indeed, the interactional responsibility—to present themselves as people
aware of the wider world. The interviews were arranged well ahead,
the microphone was placed in front of the speaker, who was asked to use
Hungarian. The speaker and I were usually alone. The topics included
opinions and information about current public issues and policies, the
high divorce rate, child-rearing ideals, the significance of ethnicity, the
consolidation of land, the mechanization of farming, as well as life
histories. But what distinguished these events from other interactions
more than anything else was the distribution of rights and duties among
participants. I asked all the questions and selected topics; interruptions
and overlaps were extremely rare, as were disagreements. I had the
powerful role of generally defining the situation. They, on their part,
had consented to the interview and, thus, to consciously presenting
themselves convincingly as spokespersons, even authorities about public
or historical affairs, to a foreigner. I dwell on these fairly obvious
aspects of the interview situation in order to convince you that these
sessions were, indeed, occasions on which speakers would have several
overlapping strategic reasons to use standard forms.
There are indications that speakers did respond in this way to the
constraints of the occasion. As one might expect, metalinguistic aware-
ness was high during interviews for young as well as old speakers.
People corrected themselves if they started to use dialect words which
are widely known to be restricted in distribution, or recent German loans
that are otherwise in common use. People sometimes declined to answer
some questions and jokingly or irritatedly rebuffed some inquiries, but no
one overturned the structure of the situation itself—for instance, by
taking over the questioner's role, or changing the topic. Sensitivity to
the overlapping constraints was also apparent in a negative way when
one speaker, an elderly woman, announced at the start of the interview
that although other people would probably talk szfp 'pretty' to me under
these circumstances, she was not about to try. She did not, and thus was
an exception among older speakers, while at the same time explicitly
affirming the general social expectation.
To tap the use of local forms was more difficult for me, largely be-
cause of my identity as an outsider and my standard speech. All of these
barriers were lifted somewhat on my third and fourth visits. But from
the first, I tried a variety of tactics which would neutralize my pre-
sence: leaving the tape recorder on a kitchen table or sideboard and
walking out; asking young people to take my tape recorder when visiting
their relatives and friends; and mostly, of course, hanging around so that
conversations which developed around me did not always include or focus
296 / Susan Gal
on me. Needless to say, these tapes show very different interactional
rights and obligations among participants. Quite other identities than
local authority were linguistically claimed and defended. There are the
interruptions, overlaps, arguments, laughing, and desultory talk common
among people who have known each other for their entire lives and meet
frequently.
Many situations I was able to tape were neither interviews nor such
family-centered conversations. But I am isolating these two types of
events because they are circumstances in which, for strategic as well as
normative reasons, different linguistic forms would seem to be appro-
priate. Although subtler uses of style shifting also occurred within
situations, the global contrast between these two kinds of interaction
was particularly strong and therefore revealing for my purposes.
*. I compared older and younger speakers by examining their use of
three phonological variables. In Figure 1, a line indicates a single
speaker, connecting his/her percentage of local forms in the interview
with his/her percentage of local forms in the conversation. 'I1 indicates
interview and 'C refers to a tape of several family members and/or
neighbors in their kitchens or porches, in which I was either not present
at all or was barely spoken to. None of these tapes have the
question/answer structure and role distribution which characterize the
interviews.
Variable 1 in Figure 1 is (gy/dzs) and its voiceless counterpart (ty/cs),
written here in standard Hungarian orthography. In each pair the first is
the standard value, a palatal stopCj-, c], and can occur in initial, inter-
vocalic, and final positions. The local forms are more affricates than
stops and are strongly fronted to alveolar position \ji^ , t / 3 . Because in
final position standardCjO is sometimes softened to a glide by urban
standard speakers of Hungarian as well as bilinguals, this was also count-
ed as a standard form. In addition, when in final position and followed
across the word boundary by the palatoalveolar fricatives CO>C3D > or
affricates, the palatal stops regularly assimilate to the fricative's place
of articulation for all speakers. These environments were therefore
omitted from the count, as was the indefinite article which ends with a
palatal stop (egy) but has its own particular alternations. No further
linguistic constraints were evident.
Notice that on Variable 1, the older speakers go from 30-50% local
forms in the interviews to more than double that in the conversations,
while the younger people show virtually no increase. What is more, the
percentage of local forms the young people use in the interviews is often
as high as the percentage that the older people use in conversations.
Variable 2 is the separable verbal prefix el in its two realizations. The
prefix means 'away' or 'off and is also a marker of completed action.
The local form includes an initial glide with systematic dropping of CO
before consonants: standard Q.G o r Q J , local p£U or QaeG. The older
people show a clear situational difference, often going from virtually no
dialect forms in the interview to 50-90% local forms in the conversa-
tions. The young people make no systematic contrast between situa-
tions. For some young people the percentage of local forms in the
interviews is as high as the percent used by many older people in conver-
sations.
Figure 1. The percent of local forms for three phonological variables
in the Hungarian of bilinguals.
1.1 Variable 1. gy/dys, ty/cs
Younger speakers Older speakers
i
—*—n
- —
i ir 55=
b:—;
! c
< • - —
—i— r
1,
i
I C ,
1.2 Variable 2. Separable verbal prefix el/je
1.3 Variable 3. The definite article a/e
t
tr \
——
—1
L' ! -• ---
Ho
r -*
T ,
—— _
f-
I i l
1 i r
298 / Susan Gal
Variable 3 is the definite article (a/e) which is realized as the low
back vowel D?D orC?2 before vowels in the standard form, and is raised
and fronted to CO locally. The situational pattern of the older people is
clear. Among the young people, none makes a marked distinction be-
tween situations and some use as high a percentage of dialect forms in
the interview as the older speakers use in conversations.
Another way to look at the quantitative distinctions between the two
groups of speakers is to compare the amount of difference between
interview and conversation for each group for each variable. If the old
people distinguish the situations in their phonological styles and the
young people do not, then the difference between the scores for the two
situations should be considerably greater for the old than the young. For
variable 1, the young group's range of difference between situations is 1-
7%, the older group's 13-48%. For variable 2, the young group's range of
difference between situations is 1-14%, the older people's 28-78%; for
variable 3 it is 1-28% for the young people and 39-66% for the older
speakers. In none of the three variables do the ranges of the two groups
overlap. These differences are strong enough to make statistical tests
of significance unnecessary.
Several other variables show the same pattern: the stylistic differ-
ences in phonology which are clear among the older speakers do not
appear systematically for the younger speakers. We can see from their
variable use of both local and standard forms that the young people know
the forms themselves. Contrary to my earlier expectations, and the
results reported by Lavandera (1978), their usage does not completely
eliminate one stylistic variant. Rather, what they lack is the pattern of
shifting styles which conveys both the tone of the situation and the
presentation of self. If there is a regularity and significance to these
young people's choice among Hungarian variants, it is difficult to track
down. No interactions that I taped showed systematically higher or
lower frequencies of one of the variants, and thus far, closer analysis of
particular interviews and conversations has failed to yield specific topics
or discourse contexts in which even a briefly sustained shift to local or
standard forms is clearly evident and interpretable. Rather, it seems
that the distinction itself has lost significance.
To understand this, it is important to recall that the young people are
speakers who do not use Hungarian with age-mates at all. Since the
social network of peers is a powerful force in the maintenance of phono-
logical norms (Labov 1972b, Milroy 1980), it is particularly significant
that for these speakers there is no social network of Hungarian-speaking
peers which could maintain separate stylistic norms within Hungarian.
Furthermore, there is, for them, no social institution (such as school or
state bureaucracy) which would demand that they actively use a distinc-
tion between the social significance of standard and local forms. The
symbolic contrasts most salient to them, reinforced by school as much as
by their own work lives, are contrasts among variants of German and
those between the two languages. The latter they exploit frequently in
conversational code-switching. Judging by a certain reluctance and the
complaints expressed in some interviews that 'I can't bring it out' proper-
ly in Hungarian (nem tudom rendesen ki hozni magyarul), I suspect that
these young speakers feel frustrated having to use Hungarian alone.
The speech of the older generation remains as a possible source for
learning the meanings of the Hungarian variants. However, as I have
argued elsewhere (1979), the moral and economic power of the older
generations to enforce their own norms of language use has been serious-
ly deteriorated by the postwar decline in the value of land and agricul-
ture (as, indeed, has their power in other social domains). The younger
generation does not look to the older bilinguals but, rather, to the in-
dustrial-bureaucratic and German monolingual world for its models.
Thus, not only is the young people's access to the use of standard forms
in Hungarian limited by the narrow range of contexts in which they hear
and actively use Hungarian, but their motivation to learn the social
meanings of the variants is attenuated by widespread social changes.
5. With these contrasting patterns I have suggested, on the one hand,

some of the systematic relations between language usage patterns and
communicative motivations of speakers, and, on the other hand, the
stylistic range in phonology which speakers exploit for communicative
effect. Clearly, the patterns cannot be explained by interference be-
tween the two languages. Rather, what I have indicated is the kind of
constraint or limit on phonological style that operates on a language
even when it is learned first, remains in daily use, is spoken frequently,
but is severely limited in its social functions.
Under normal circumstances, when speaking with other bilinguals in
their accustomed manner, young people who do not use a stylistic range
in Hungarian do not experience embarrassment or communicative diffi-
culty. Their repertoire and that of their listeners includes German, and
switching to German is available to them as a meaningful way of
indicating authority, publicness, and other social meanings. However, I
suggest that the flattening or simplification of phonological style is
linked not simply to their bilingualism and the availability of language-
switching for stylistic effect, but rather, to the lack of social contexts,
networks, and institutions which would support stylistic norms and
meanings in Hungarian. It is the political and economic developments of
the postwar years that have deprived many young speakers of the social
contexts in which the phonological variants of Hungarian are meaning-
fully distinguished and in which those variants are rhetorically signif-
icant. The contrasting case of the older speakers highlights this. They
are also bilinguals and make use of conversational language-switching in
some communicative contexts. Yet, they systematically and meaning-
fully use phonological style as well. The juxtaposition of these cases,
found within a single community, suggests the way in which linguistic
skills are dependent on specific historically located social contexts.
Speakers' knowledge of a linguistic variety is affected by the socially
constructed limitations on the opportunities and motivations for its use.
Notes
My thanks to Susan Dannenbaum, Dee Garrison, Judith Gerson, and

Carol Nackenoff for their suggestions and to James Boster, John Gum-
perz, Jane Hill, and especially Kit Woolard for their criticism of an
earlier version of this paper. The research was funded by the National
Science Foundation (BNS 80-05889), whose support is gratefully
300 / Susan Gal
acknowledged. This paper is drawn from a longer report dealing with
some of the same issues (Gal 1984).
1. Dressier noted briefly that 'stylistic shrinkage' is a phenomenon of
decaying language (1972:454) and Dressier and Wodak-Leodolter (1977)
mention 'monostylism' as characteristic of some Breton speakers,but do
not go on to explore or document these changes. See also Lambert and
Freed (1982) for related research on language loss.
2. Segalowitz and Gatbonton (1977) discuss the stylistic range in the
second language of adults first learning that second language. Their
work is thus about nonfluent speakers, while the research reported here
concerns fluent speakers using their first language. Ma and Herasim-
chuk's (1971) careful work is not comparable for another reason. They
contrast reading and speaking as their measure of style, a strategy I
decided not to use (see note 4). Nevertheless, Ma and Herasimchuk's
study lends some indirect support to the viewpoint presented in this
paper since they found limitations in stylistic range in some, but not all,
of their demographically defined bilingual groups.
3. The study is based on a total of 20 months of field work: 11
months in 1974, the rest in 1979, 1981, and 1982.
4. The recent introduction of Hungarian instruction in the early
grades is the exception. The Reformed Church also continues use of
Hungarian.
5. Following Labov's (1972a) early work, reading passages and word
lists are often used to tap speakers' phonological styles, especially in
urban dialect surveys, on the assumption that reading evokes the high-
prestige standard forms. I decided against this method in the present
study for a number of reasons. Recent work on literacy indicates that
reading, if it is done at all, is a speech event with a particular place in
the communicative economy of a community. Thus, the evaluation of
reading and the kind of pronunciation deemed appropriate to it may well
vary crossculturally, as with any speech event. One cannot assume that
reading of word lists, for instance, will contrast maximally with pronun-
ciation in 'casual' speech events in all communities. Milroy (1980:99-
105) showed convincingly that in working-class Belfast, word lists were
read with more vernacular pronunciation than that used in interviews.
She suggests that there are norms for reading pronunciation in Belfast
which have little to do with pronunciation norms in nonreading speech
events. Similarly, when I asked speakers to read a folk tale collected in
the community, spelling pronunciation swamped all differences on most
variables. What is more, for the agricultural segment of the neighbor-
hood I am describing here, and for those young people who did not attend
Hungarian-language schools, reading in Hungarian, in most cases, is a
problematic activity done rarely and haltingly, if at all. It may not
reflect these speakers' nonreading repertoires.
6. Approximately 40 minutes of tape were examined for each speaker,
about 20 minutes in each of the two situations. The average number of
tokens per person per variable was: 305 for variable 1 (range 96-651), 36
for variable 2 (range 9-62), 233 for variable 3 (range 84-446). There was
a reliability of 90% on the coding judgments for these and several other
variables in a 30-minute sample of tape coded twice, about a month
apart. Two further checks of different 30-minute samples showed 81%
and 92% reliability.
References
Blom, 3an-Peter, and 3. 3. Gumperz. 1972. Social meaning in linguistic
structures: Code-switching in Norway. In: Directions in sociolin-
guistics. Edited by 3. Gumperz and D. Hymes. New York: Holt,
Rinehart and Winston. 407-34.
Dorian, Nancy. 1981. Language death: The life cycle of a Scottish
Gaelic dialect. Philadelphia: University of Pennsylvania Press.
Dressier, Wolfgang. 1972. On the phonology of language death. Chi-
cago Linguistic Society Papers 8. 448-57.
Dressier, Wolfgang, and Ruth Wodak-Leodolter. 1977. Language preser-
vation and language death in Brittany. International 3ournal of the
Sociology of Language 12.33-44.
Gal, Susan. 1979. Language shift: Social determinants of linguistic
change in bilingual Austria. New York: Academic Press.
Gal, Susan. 1984. Contraction and expansion of phonological styles.
MS.
Gumperz, 3. 3. 1982. Conversational code-switching. In: Discourse
strategies. Edited by 3. Gumperz. London: Cambridge University
Press. 59-99.
Halliday, Michael. 1973. Explorations in the functions of language.
New York: Elsevier.
Hill, 3ane, and Kenneth Hill. 1977. Language death and relexification in
Tlaxcalan Nahuatl. International 3ournal of the Sociology of Language
12.55-69.
Hill, 3ane, and Kenneth Hill. 1978. Honorific usage in modern
Nahuatl. Lg. 54.1:123-55.
Hymes, Dell. 1971. Introduction, Part III. In: Pidginization and creoli-
zation of languages. Edited by D. Hymes. London: Cambridge Uni-
versity Press. 65-90.
Imre, Samu. 1971. A felstfoVi, nyelvjara"s The FelstfoV dialect .
Nyelvtudomanyi EVtekeze'sek 72. Budapest: Akademiai Kialdo.
Irvine, 3udith. 1979. Formality and informality in communicative
events. American Anthropologist 81.4:773-90.
Keller, R. E. 1961. Upper Austrian. In: German dialects: Phonology
and morphology. By R. E. Keller. Manchester: Manchester Univer-
sity Press. 200-47.
Labov, William. 1972a. On the mechanism of linguistic change. In:
Sociolinguistic patterns. By William Labov. Philadelphia: University
of Pennsylvania Press. 160-82.
Labov, William. 1972b. The linguistic consequences of being a lame.
In: Language in the inner city. By W. Labov. Philadelphia: Univer-
sity of Pennsylvania Press. 255-96.
Lambert, Richard D., and Barbara F. Freed, eds. 1982. The loss of
language skills. Rowley, Mass.: Newbury House.
Lavandera, Beatriz. 1978. The variable component in bilingual perfor-
mance. In: Georgetown University Round Table on Languages and
Linguistics 1978. Washington, D.C.: Georgetown University Press.
391-409.
302 / Susan Gal
Ma, Roxana, and Eleanor Herasimchuk. 1971. The lingusitic dimensions
of a bilingual neighborhood. In: Bilingualism in the barrio. Edited by
3. Fishman et al. Bloomington: Indiana University Press. 3^7-46**.
Milroy, Lesley. 1980. Language and social networks. Baltimore: Uni-
versity Park Press.
Romaine, S., and E. Traugott. 1981. The problem of style in socilin-
guistics. Paper presented at the Winter Meeting of the Linguistic
Society of America, New York.
Segalowitz, Norman, and Elizabeth Gatbonton. 1977. Studies of the
non-fluent bilingual. In: Bilingualism: Psychological, social and
educational implications. Edited by P. Hornby. New York: Academic
Press. 77-90.
Slobin, Dan I. 1977. Language change in childhood and in history. In:
Language learning and thought. Edited by 3. Macnamara. New York:
Trudgill, Peter, and J. K. Chambers. 1980. Dialectology. Cambridge:
Cambridge University Press.
SOME COMMENTS ON THE SUBJECTIVE LEXICON
George A. Miller
Princeton University
When you learn a language, several kinds of learning occur: pronun-
ciation, vocabulary, grammar, usage, each characterized in a separate
body of linguistic theory. All of these—phonology, syntax, semantics,
pragmatics—describe things that a language user has to learn, and that a
skillful speaker knows.
Now, what people know and how they learn it are basic questions for
cognitive psychology; when those questions concern what people learn
and know about languages, we call it psycholinguistics. In principle, a
psycholinguist should be interested in all these kinds of learning. In fact,
however, lexical learning is generally considered uninteresting. That
evaluation is a mistake. There are as many fascinating puzzles in the
lexical as in any other aspect of language.
To know the vocabulary of a language is to know the sounds and mean-
ings of the basic words in the language. I acknowledge the technical
difficulties with this formulation—the problems of determining what a
word is, what a meaning is, and what makes a word important—but they
are not my present concern. However those problems are solved, what it
means to know a lexical unit eventually comes down to one or both of
two criteria:
(1) The 'receptive criterion': knowing what the word means when
you hear it used;
(2) The 'productive criterion': using the word naturally and appro-
priately in sentences that express your own thoughts.
The receptive criterion. The receptive criterion involves recognizing
the word and retrieving its meaning from lexical memory. Since the
operational test is to give (or at least recognize) the word's definition,
retrieval from lexical memory is usually likened to looking up the word
in a dictionary. Therefore I shall focus my comments about the recep-
tive criterion for word knowledge on the implicit assumption:
303
30* / George A. Miller
(3) The 'receptive assumption1: A language user's lexical knowl-
edge is organized into independent lexical entries, the way
a printed dictionary is.
Of course, with a printed dictionary you have to use alphabetical re-

trieval, whereas you can get access to your own lexical entries either by
way of their pronunciation or by way of the ideas they express. Except
for that, however, a printed dictionary is generally viewed as an explicit
theory of lexical knowledge.
Not until I had thought about (3) for some time did I realize how
improbable it is. Lexicographers who compile dictionaries are not
thinking about psychology—how wonderful that they should manage to
produce psychological theories! The way you go about writing a diction-
ary is not at all the way you would go about constructing a psychological
theory. And it is certainly not the way a child goes about learning a
vocabulary.
I want to consider some differences between objective and subjective
dictionaries; I introduce my comments by considering briefly the growth
of vocabulary in children.
The growth of vocabulary. Recent studies of cognitive development
have not totally ignored vocabulary, but I think it is fair to say that the
principal interest in it has been as a convenient window on conceptual
development. Interest in the words themselves has been secondary.
It was not always so. During the 1930s and 1940s, no doubt stimulated
by the observation that vocabulary is the most dependable part of any
intelligence test, psychologists devoted considerable attention to the
growth of vocabulary in children. It was found that the first word ap-
pears between 12 and 18 months of age, then new words begin to appear
more and more rapidly until by the age of three a parent can no longer
tell whether the child knows a particular word. Thereafter, vocabulary
grows so rapidly that dictionary-sampling techniques are required to
track it.
Several careful, cross-sectional studies of vocabulary growth were
conducted back in the days before transformational generative grammar
was invented. The results that I have most faith in were collected by
Mildred Templin and published in 1957. Accoding to Templin, a six-year-
old child of average intelligence knows 13,000 words and the average
eight-year-old child knows 28,300.
In order to appreciate these numbers, it helps to convert them into
words learned per day. Between the ages of six and eight years, an
average child learns 21 words per day. Children of superior intelligence
probably learn words at twice that rate.
This estimate may seem high, but all the other studies of vocabulary
growth that I have seen give even higher rates. Apparently, a broad and
rapid learning process goes on during childhood and psychologists know
very little about it. At least two things do seem clear, however: first,
young children are very good at acquiring vocabulary; second, since
nobody teaches children 21 words every day, they must be very good at
educing meanings from context.
In 1974, Susan Carey and Elsa Bartlett invented a way to observe this
rapid learning in more detail (Bartlett 1977; Carey 1978; Carey and
Some comments on the subjective lexicon / 305
Bartlett 1978; Miller 1977). They used color names. First, they es-
tablished that a group of three-year-old children did not know the color
olive; most called it green, some called it brown. Carey and Bartlett
decided to teach the children a nonsense name for olive, a name that
they would not have heard anywhere but in our nursery school. So they
painted one tray olive and another tray blue and asked each child casual-
ly, 'Hand me the chromium tray. Not the blue one, the chromium one.1
The child would pause, perhaps point to the olive tray and ask, 'This
one?1 A week later, with no further guidance the children were again
asked to name the colors. This time, when olive was presented, they
paused. They couldn't remember the nonsense name for it, but now they
knew it wasn't green or brown. A single exposure was enough to begin a
reorganization of their color lexicon.
Carey thought she saw a two-step learning process. First, children
notice a new word and assign it to an appropriate category. After
hearing chromium just once, they had assigned it to the semantic field of
color names. Children are good at keeping these fields separate, even
when they don't really know what the words mean.
The second step is to work out the distinctions among words within a
category; that may take a long time. When you calculate that an aver-
age child learns 21 words a day, it does not mean that the learning
process is complete in one day. A child will be alert to and working out
the relations among many words at the same time, but exactly how many
are being learned at any one time is not known.
One thing is reasonably sure, however. Children could not memorize
21 arbitrary, unrelated facts every day. Words can be learned rapidly
because they are not unrelated, because they form conceptually related
patterns. A vocabulary is a coherent, integrated system of concepts. In
other words, a feature that tends to be ignored in objective dictiona-
ries—the organization of words into semantic fields—is what makes rapid
vocabulary learning possible. If we hope to understand this process, we
must have some reasonably definite characterization of lexical
organization.
Lexical organization. Lexical memory must be organized in (at least)

two ways, phonologically and semantically. A variety of common obser-
vations—that it is easy to think of words starting with the same sound or
of words that rhyme, that confusions in memory and slips of the tongue
often occur between words that sound alike, and so on—all support the
assumption of a phonological organization. But an equal array of com-
mon observations—that people associate words with similar meanings,
that similar meanings are used to define new words, that relations of
entailment between words are easily recognized, and so on—all support
the assumption of a conceptual organization, an organization into se-
mantic fields. Children exploit both kinds of organization in their effort
to learn words, but my present interest is primarily in the conceptual
organization of the subjective lexicon.
Most printed dictionaries of English recognize specialized semantic
fields, although the recognition is buried under the alphabetical ordering
and very difficult to use. For example, technical senses of a word will
be marked medicine or jurisprudence or physics or whatever. If you
gathered together all those senses with a given marking, you would have
306 / George A. Miller
a reasonably good glossary of technical terms in that specialty. Un-
fortunately, however, with a printed dictionary there is no easy way to
retrieve all and only those senses bearing a particular meaning. And no
English language dictionary that I know of uses markers in this way to
distinguish among nontechnical semantic fields. What they do use, of
course, are synonyms. If you trace out the network of cross-references
provided by synonyms and antonyms, you can get an idea of the relevant
semantic field. But this is tedious work with a printed dictionary,
whereas it comes immediately with the subjective lexicon. That is to
say, the subjective lexicon seems to be organized more like a collection
of dictionaries on many different subjects.
Semantic fields are sometimes explained as a natural reflection of the
semantic decomposability of lexical concepts. For example, I have
argued (Miller 1972, Miller and Johnson-Laird 1976) that the semantic
field of motion verbs can be defined as including all those verbs whose
semantic decomposition includes the primitive (or atomic) concept of
change-of-location-over-time. In addition to motion, however, some
motion verbs indicate the method of moving—walk versus run, for
example; some require knowledge of the speaker's location—come versus
go; many indicate the direction of motion—rise, fall, approach—or the
medium through which the motion occurs—swim versus fly_. You can
tease out about a dozen different components, so that each one of the
verbs of motion looks like a particular package of semantic concepts,
and the semantic field consists of many other related but slightly differ-
ent packages of concepts.
Not every shared concept can serve as the nucleus of a semantic field,
however (Miller and Johnson-Laird 1976). Although the set of verbs that
share a concept of movement do form a semantic field, the set of verbs
that share a concept of causation do not. Move, kill, give, and tell, for
example, are all causative verbs, but they clearly do not form a seman-
tic field. Johnson-Laird and I speculated that some concepts, like
motion or possession, can provide a nucleus around which a semantic
field can develop, whereas other concepts, like cause or instrument, are
used to elaborate concepts within many different semantic fields.
Learning words from context. Sometimes you can satisfy the recep-
tive criterion when you do not really know a word, because the context
makes the meaning apparent. Indeed, context is a major source of the
information used in learning the meanings of many new words.
In 1950, Werner and Kaplan reported an experiment designed to reveal
how well children could guess the meaning of a new word after hearing it
used in a succession of sentences. After hearing A corplum may be used
for support, Corplums may be used to close off an open space, A wet
corplum does not burn, and so on, the children offered their opinions of
what a corplum was. Children eight years old seemed to have difficulty
disentangling the new word from the context in which it occurred, but
13-year-old children could usually solve the problems.
The question raised by Werner and Kaplan was ignored until 1981 when
two workers in Amsterdam, M. M. van Daalen-Kaptjens and M. Elshout-
Mohr, reopened it. What they contributed was an insightful analysis of
protocols that they collected from college students who were asked to
think out loud while they solved such problems.
The process of transforming several contexts of use into an acceptable
definition was called 'decontextualization1 by Daalen-Kaptjens and
Elshout-Mohr. They claimed that it requires at least two steps: first, a
reformulation of the context into a sentence about the unknown word,
e.g. to reformulate The painter used a corplum to stir his paints, which
is about a painter, into a sentence like Corplums can be used to stir
paints, which is a sentence about corplums; second, the transformation
of this reformulated information into an aspect of the meaning, e.g.
Corplums can be implements.
Subjects who could perform both steps generally succeeded in formu-
lating an acceptable definition; subjects who performed only the first
step were not as successful. Daalen-Kaptjens and Elshout-Mohr specu-
lated that their successful subjects used the first context, plus their
general knowledge, to select a schematic model, then used successive
contexts to narrow down the exact properties of the model. Less suc-
cessful subjects seemed to have difficulty because they did not carry out
the second step, the translation of contextual information into further
aspects of the model.
The preferred way to write a definition is to give the name of the
superordinate class to which the concept belongs, and to follow that with
a relative clause that differentiates this particular instance of the class
from all other instances. This strategy is not always available but when
it is, it seems to lead to clear, simple definitions. The general format is
as follows.
(4) Definitional format: 'An X is a Y that . . .,' where the phrases
following that provide distinguishers.
College students have far more world knowledge to draw upon than do
young children, but it seems to me that the Carey and Bartlett three-
year-olds and the Daalen-Kaptjens and Elshout-Mohr college students
were both trying to impose format (4) on the meanings of new words via
a two-step path: first identify a general category, later work out the
distinguishing particulars. No doubt there are exceptions to this rule,
but a substantial fraction of word learning probably does follow some
such pattern.
A more psychometric approach has been adopted by Sternberg and
Powell (1983), who found that high school students' scores on a learning-
from-context task correlated about 0.6 with IQ, vocabulary, and reading
comprehension scores. They suggest that context is valuable insofar as
it provides information about certain general aspects of the target
word's meaning. Indeed, the aspects that they list resemble the kind of
meaning components that 3ohnson-Laird and I used to characterize
motion verbs, and that others have developed for other kinds of semantic
decomposition. I take it, therefore, that Sternberg and Powell share my
own intuition:
(5) The 'contextual learning assumption1: An ability to perform
conceptual decompositions is valuable for learning word
meanings from context.
If assumption (5) is correct, it suggests how semantic fields and

leaming-from-context might work together in some eventual theory of
vocabulary growth.
These comments indicate some of the differences between objective
and subjective lexicons: differences in organization, use, and method of
construction. In view of these differences, I think it is important to
reconsider the all too pervasive assumption (3) that lexical retrieval
from a dictionary is a good model for lexical retrieval from personal
memory.
The productive criterion. The productive criterion for knowing a word

is less easily applied than is the receptive criterion, because the investi-
gator must wait for the speaker to introduce the word—which may entail
a very long wait indeed. However, school teachers have the authority to
reduce this wait by assigning pupils the task of using particular words in
sentences. Indeed, this assignment is frequently used as part of the
students' training in 'dictionary skills'.
Polysemy. As dictionaries are written, many different senses are

distinguished and a reader is expected to be able to tell which one is
appropriate in any given context. Most intelligent adults are able to
solve this problem, but it is easy to show that children have great diffi-
culty with it. For example, Deese (1967) reported on one teacher of
seventh grade English who gave her pupils the assignment of looking up
certain words in the Webster's Collegiate Dictionary and using them in a
sentence. Here is a slightly simplified summary of the definitions that
the children found when they looked up the word chaste:
chaste: 1: innocent of unlawful sexual intercourse. 2: celibate.

3: pure in thought and act, modest. 4: severely simple in design
or execution, austere.
With that lexical entry in mind, let me read you Deese's report (1967) of
what some children did with it:
Here are some sentences written by these youngsters after they had
looked up the word chaste. You will have to admit that they are all
consistent with at least one of the senses supplied by the dic-
tionary.
1. 'The amoeba is a chaste animal.1 Evidently, the youngster who
wrote this sentence is following that part of the entry that
says chaste means simple in design. . .
2. 'The plates were still chaste after much use.' Here the notion
of being unstained seems to be critical.
3. 'The milk was chaste.' Evidently, the sense of pure is meant

here.
Such examples demonstrate that dictionaries are reference books, not

teaching instruments. If you already know something about the word, a
dictionary will remind you of it. But if you are totally innocent, you
cannot trust a dictionary as your sole source of lexical information. Not
only do dictionaries make too many distinctions among meanings—they
often do not make the right ones.
At Princeton, Patricia Gildea and I are following up on Deese's project
and trying to extend it. We have made contact with local school
teachers who are interested in teaching dictionary skills and have found
them most cooperative. They have allowed us to see their students'
worksheets, and we are beginning to build up an inventory of mistakes
that children make when they use a dictionary. We have not gotten far
enough with this project to draw any generalizations. At the moment we
are puzzling over why a child might think that two boxers engaged in
fisticuffs were betrothed, or why the person who stuck his digit in his
ear put a number in his ear, and similar confusions. As yet, however, we
have not collected enough examples to make any guess about the sources
of the children's difficulties.
Along similar lines, Julia Jorgensen and I tried to analyze what you do
when you look up an unfamiliar word in the dictionary (Jorgensen 1984).
It is an intellectually challenging task. The continuity of your reading is
interrupted, but the context in which the word occurred must be kept in
mind. Once the word is found in the dictionary, you may need to choose
among alternative lexical entries on the basis of part of speech; then,
within the right entry, you may have to choose among several alterna-
tive senses on the basis of context. That is to say, the context of the
original passage must be compared with a succession of contexts sug-
gested in the dictionary until a best guess can be made as to the in-
tended sense.
In order to study this process in a simplified form, Jorgensen (1984)
developed a questionnaire based on 40 words (20 frequent words and 20
relatively rare ones). For each word, phrases defining two different
senses, with a sentence illustrating each sense, were taken from a
children's dictionary. The words were then replaced in the illustrative
sentences by nonsense syllables, and subjects were asked which defini-
tion of the nonsense syllable was most appropriate in the context pro-
vided by the sentence.
For example, one item presented the sentence:
The snow is hattay with the windows.
Then two senses for hattay were given: (1) 'level, flat, smooth'; (2) 'at
the same level'. Subjects were asked to choose the better of the two
senses of the word as it is used in the illustrative sentence.
The questionnaire was given to 20 Princeton undergraduates and to 20
fourth grade children. College students identified the correct sense on
83% of the items, fourth graders on 63%: chance, of course, was 50%.
The fact that adult performance was not perfect suggests that the illus-
trative sentences found in dictionaries leave something to be desired.
That the children made even lower scores confirms the difficulty they
have with such context-matching tasks.
After going through the questionnaire, the subjects then went through
it a second time, but this time trying to guess what the real words were
for which nonsense syllables had been substituted. (For the example just
given, the original word was even.) College students guessed right 59%
of the time for high frequency words, but only 24% of the time for low
frequency words; the comparable figures for children were 18% and
5%. Moreover, the ones they guessed correctly were not always the
same ones that they had gotten right in the context matching task—so
their success could not be attributed to seeing through the nonsense
substitutions.
What interested us most, however, was the consistent and reliable
difference between high and low frequency words. Since we had sub-
stituted nonsense syllables, we expected that they would all be treated
like very low frequency words. The most likely explanation of the
difference is that dictionaries do not define high and low frequency
words the same way. Definitions for the high frequency words seemed
better, at least in this particular sample.
Which raises the interesting question of how to evaluate the quality of
a proposed definition.
The quality of a definition. The problem of definitional quality arises

repeatedly in attempts to experiment with lexical materials. Do
children have difficulty using dictionaries because the definitions are
poor? Do lexicographers write better definitions for senses that are
represented by a greater number of instances of use? If you wanted to
improve the quality of dictionary definitions, how would you evaluate
your work? And so on. Obviously, it would be very useful to have some
convenient way to assess the quality of definitions.
One possible solution is the subjective rating scale. Sternberg and
Powell (1983) found that judges could agree in rating the quality of
definitions written by high school students. But such ratings are hardly
sensitive enough to evaluate the work of professional lexicographers,
and, lacking further specification, critical questions about the rater's
criterion are left open: what is a good definition supposed to be good
for?
Jorgensen (1984) proposed the following method for evaluating defini-
tional quality. First, a nonsense syllable was inserted into a definition
and subjects were requested to write sentences using the nonsense syl-
lable in the intended sense. Then the original word was restored in these
sentences and the resulting sentences were submitted to a second group
of subjects for evaluation. A good definition was taken to be one that
enabled a person to write an acceptable sentence using the word in the
intended sense. Otherwise said:
(6) The 'productive criterion of definitional quality': A good

definition is one that provides the information required to
satisfy the productive criterion for knowing the word.
Note that (6) is a much more modest requirement than the traditional
criterion of synonymity:
(7) The 'synonymity criterion of definitional quality': A good
definition is one that is substitutable for the word it defines
in all contexts without altering truth values.
An operational test for (7) might challenge subjects to pair words with
their definitions—a task that should be easy if the words and their def-
initions were synonymous. Experience suggests that this challenge
becomes very difficult to meet when the words involved are at all simi-
lar in meaning.
A still different criterion might require that the definition be useful
for helping someone—a child, perhaps—learn the meaning of an un-
familiar word:
(8) The 'pedagogical criterion of definitional quality1: A good
definition is one that provides the information required to
add the word to one's vocabulary.
This criterion would impose different requirements on definitions in-
tended for learners of different sophistication. For a given learner,
however, a definition satisfying (8) should also satisfy (6), although the
reverse would not necessarily hold.
I have no firm favoritism for any one of these alternatives, but I do
feel a need to reserve the possibility that the quality of a definition
might depend on the use you plan to make of it, and that definitions good
for one purpose might not be optimal for some other.
In the time available I have been able to do little more than hint at
the variety and complexity of lexical problems worthy of theoretical and
experimental investigation. I hope, however, that this small sample will
be enough to revive your interest in this often neglected suburb of
linguistic studies.
Note
This talk was prepared with the help of grants to Princeton University
from the Alfred P. Sloan Foundation and the Spencer Foundation.
References
Bartlett, E. 3. 1977. The acquisition of the meaning of color terms: A
study of lexical development. In: Proceedings of the Stirling Confer-
ence on the Psychology of Language. Edited by P. Smith and R.
Campbell. New York: Plenum.
Carey, S. 1978. The child as word learner. In: Linguistic theory and
psychological reality. Edited by M. Halle, 3. Bresnan, and G. A.
Miller. Cambridge, Mass.: MIT Press.
Carey, S., and E. 3. Bartlett. 1978. Acquiring a single word. Papers and
Reports on Child Language Development 15.17-29. Department of
Linguistics, Stanford University.
Daalen-Kaptjens, M. M. van, and M. Elshout-Mohr. 1981. The acquisi-
tion of word meanings as a cognitive learning process. Journal of
Verbal Learning and Verbal Behavior 20.386-99.
Deese, 3. 1967. Meaning and change of meaning. American Psycholo-
gist 22.641-51.
Jorgensen, 3. 1984. Polysemy and psycholexicology. Ph.D. thesis,
Princeton University.
Miller, G. A. 1972. English verbs of motion: A case study in semantics

and lexical memory. In: Coding processes in human memory. Edited
by A. W. Melton and E. Martin. Washington, D.C.: Winston.
Miller, G. A. 1977. Spontaneous apprentices: Children and language.
New York: Seabury Press.
Miller, G. A., and P. N. 3ohnson-Laird. 1976. Language and perception.
Cambridge, Mass.: Harvard University Press.
Sternberg, R. J., and 3. S. Powell. 1983. Comprehending verbal com-
prehension. American Psychologist 38.878-93.
Templin, M. C. 1957. Certain language skills in children: Their devel-
opment and interrelationships. Minneapolis: University of Minnesota
Press.
Werner, H., and E. Kaplan. 1950. Development of word meaning through
verbal context: An experimental study. Journal of Psychology 29.
251-57.
A SONG WITHOUT MUSIC AND OTHER STORIES:
HOW COGNITIVE PROCESS CONSTRAINTS INFLUENCE
CHILDREN'S ORAL AND WRITTEN NARRATIVES
Marilyn Shatz
University of Michigan
Narratives as a kind of language use exhibit a remarkable range of
length, complexity, style, and content. The minimal unit of a narrative
is the relation of two events or actions, as illustrated in (1), a story by a
three-year-old (reported in Sacks 1972).
(1) The baby cried. The mommy picked it up.
More elaborated instances can have introductory segments, evaluative
and resolving statements, and recapitulations or specific ending state-
ments (see Labov 1972). Narratives can be about personal experiences,
reported events, or fantasy.
Children's ability to deal with narrative language in comprehension
and production tasks has recently received much attention from cogni-
tive and developmental psychologists. Researchers have investigated
children's understanding of story structure (e.g. Brown 1975, Stein and
Glenn 1979), their story retellings (McNamee 1979), and their story
creations (Pitcher and Prelinger 1963). Findings from these and other
efforts show that even quite young children have some understanding of
story structure (see Mandler 1983 for a review). Moreover, they use
some of the formal devices characteristic of narrative segments. For
example, four-year-olds studied in my laboratory produced introductory
statements in their narratives. They used phrases like one time, one
day, once upon a time, and my favorite—once of a little time. Applebee
(1978) reports that children as young as two years of age recognize such
phrases as appropriate to the narrative register.
Yet, young children's narrative skills are far from mature. The fact
that many children are reluctant even to attempt the narrative task
when requested to do so suggests that narration is a difficult mode for
them. One researcher has reported only a 50% success rate in story
elicitation with two-and-one-half-year-olds (Ames 1966). In my data
313
314 / Marilyn Shatz
with four-year-olds, about the same success rate was achieved. My

elicitation situation was quite informal and unconstrained, with children
often encouraged to elaborate on topics they themselves had introduced
into a casual conversation with a parent or an experimenter they knew
well. Nonetheless, the task regularly proved to be more than the children
wanted to handle. When encouraged to elaborate, children would typ-
ically respond, 'I forgot' or 'I don't know'. However, it is unlikely that
memory was the real problem, since some children on occasion overcame
their hesitations and actually went on to narrate an experience, or they
did so after a more explicit question from the listener. Examples (2) and
(3) illustrate this hesitant behavior in a four-year-old girl.
(2) Experimenter: And what did you do in the sand?
Child: I forgot. Dig. You dig fishes and stones.
(3) Child: He (Uncle Timmy) plays it (music) from a guitar.
Experimenter: Tell me about Uncle Timmy's guitar playing.
Child: No.
Experimenter: Why not?
Child: 'Cause it's a secret.
What, then, makes narration a difficult task? A hint comes from
Scollon and Scollon's (1983) discussion of the oral tradition of the Atha-
bascans living in Alaska and Canada. Scollon and Scollon report that
Athabascan stories, although preserved from generation to generation,
are rarely told the same way twice. Instead, any particular telling of a
story is modified to take into account the characteristics of the audi-
ence. The Scollons' claim, then, is that oral narratives are no less a
communicative task than is conversation. They require perspective-
taking and adjustment skills just as other forms of talk do. If we accept
this claim, then we can see that narrative production is not merely a
problem of memory or imagination or knowledge of narrative structure.
It involves these but it also involves the ability to create a sequence of
utterances that provides an appropriate format for a specific listener.
Of course, just as with rules of conversation, the rules governing narra-
tive performances are to some extent culture-dependent. Not all of the
constraints governing Athabascan story-telling pertain in other cul-
tures. Nonetheless, the general principle does hold: narration is a
complex communicative skill requiring attention to listener characteris-
tics.
Accommodating to heavy cognitive workloads. If children were
egocentric, that is, unaware of differences in perspective among listen-
ers, then one might expect them to plunge into the narrative task with-
out concern for the communicative aspects of it. This is not the case.
We know from my work, and others' as well, that children do take ac-
count of the age or knowledge level of their listeners and adjust their
speech accordingly in a variety of explanation and description tasks (see
Shatz and Gelman 1973; Shatz 1983, for a review). As is shown later in
this paper, children know that narratives too require some tailoring to
the level of the listener. Now, the point to be stressed is that the
knowledge that a narrative requires adjustment is one of the compli-
cating factors making the narration task so difficult for young children.
A song without music and other stories / 315
It is one thing to know something about the formal devices and topics
that are appropriate to narrative. It is another to know which ones are
appropriate to which listeners. Part of the reluctance of preschoolers to
narrate, I would argue, is a result of the cognitive overload they experi-
ence when they are asked to communicate in a register that makes all of
the communicative demands on them that other modes do and that has,
in addition, its own set of structural characteristics that are not yet well
controlled.
The argument that I am making is consonant with the cognitive pro-
cess theory I advanced several years ago to account for children's vari-
able performance on other sorts of communication tasks, for example,
those involving referential and explanatory skills (Shatz 1978). The
essentials of the argument are that a child has limited resources to carry
out any task. A complex task can be divided into component skills neces-
sary for adequate performance on the task. Each skill can be controlled
by information-handling techniques—devices for spreading resources
farther than they would otherwise go. For example, knowledge of social
conversational conventions can be considered such a technique. One can
initiate and carry on a conversation for several minutes with little effort
if one knows the conventional forms for doing so. Much more effort
would be involved if one had to search out and create new openings each
time an interaction was attempted. Nonetheless, some degree of cogni-
tive work is involved in the accessing and use of any skill or technique.
The better learned and more Dracticed the technique, the lower the
workload value it has; that is, the fewer resources it takes to execute.
Moreover, there is additional work involved in a complex task, which
requires identifying the subskills, selecting the information-handling
techniques, and integrating their outputs. Subskills are not independent
of one another in that they draw from the same resource pool in order to
be accessed and executed. Thus, in one context, a given component skill
may be displayed with ease because the other components required in
that context are well learned. However, the same skill may not be well
executed or even displayed at all in a context where the resources are
absorbed by other more central and high-cost skills. The potential for
exhibiting any particular ability, then, must be considered with regard to
the context of its display.
Figure 1 is not a process model of complex task execution, but in-
stead graphically illustrates the central constructs of the theory and
their relations. Especially noteworthy is the variation in the amount of
resources different information-handling techniques (IHT) draw from the
central resource pool, and the assumption that selection and integration
functions also draw resources from the same pool. Not clearly apparent
in Figure 1 is the system's potential to select one of several possible
information-handling techniques to control a subskill with relative
success, depending on the resources available. For example, one tech-
nique may do the job passably but not superbly, but use fewer resources
than the technique of choice in less resource-demanding circumstances.
Thus, the theory allows trade-offs between the elegance, precision, or
appropriateness with which a particular subskill is carried out and the
amount of resources devoted to its execution.
316 / Marilyn Shatz
1% Constructs of limited resource-cognitive process theory.
Execute complete task

How can the cognitive process theory be used to understand better
children's early narrative performances? Several aspects of my data on
children's narratives are especially compatible with the limited re-
sources view. First, an examination of some early narratives shows that
children recruit a variety of information-handling techniques from other
kinds of language use to assist in the successful execution of their narra-
tives. While these techniques may not be strictly appropriate to a narra-
tive style, they are well learned, and hence cheap in terms of re-
sources. Thus, they can be executed even in a fairly difficult context.
Moreover, narratives can be quite variable with regard to style, and
children may be encouraged by this flexibility to borrow devices from
other sources.
Second, I noted that children do adjust their narratives according to
their listeners, although sometimes they cannot meet the various
demands that both listener adjustment and narrative style exact. Third,
it is also noted that when first using a narrative style, children are
reluctant to stray far from what they know best—their own personal
experiences. It seems that a well-known topic is less costly to express in
a novel format than venturing far into fantasy is. Finally, an examina-
tion of early written attempts at narration by one child shows that
working in a new mode is resource-costly, requiring other concomitant
component skills to operate at a high level of efficiency. That is, the
transition from oral to written narrative is a drain on resources; the
nascent skills of spelling and writing encourage a return to well-learned
forms and topics of narratives common in younger children's oral produc-
tions. This return to earlier, easier forms is an example of the selective
trade-offs the limited resource system can make when confronted with
novel complex tasks taxing its capacity.
In sum, I will show that children do have considerable knowledge of
what constitutes an appropriate narrative. However, integrating all the
components of that knowledge into a smooth, mature performance on
any given occasion is a difficult task for the novice. Instead, the child's
solutions often combine more and less appropriate topics and devices in
order to express their fragile narrative ability.
Oral narratives. I turn now to some examples of narratives to illus-

trate the points I have just made. My first examples are oral narratives,
each told by four-year-olds to a familiar listener in the casual conversa-
tional context of looking at magazine pictures. Before the story illus-
trated in (4), the female narrator and an adult experimenter had been
discussing the ocean and beach depicted in a picture.
(4) Child: Guess what. Once we went—The last time

we went down there (to the beach)~over
the weekend, guess what happened to us.
Experimenter: What?
Child: We had a hurricane. It was in the evening
and—and the house was shaking...'cause it
was really old. Guess what. The next day
when we went to the beach, the—the
water was only half way up the beach.
Of note is the child's use of the conversational attention-getter guess
what?, not only to begin her narrative, but also as a marker of each
segment of the story. Uses of phrases such as guess what or know what
are common in children's conversations. The use of guess what makes
very clear that there are three segments to the story. Guess what
operates much like the phrase and then in the narratives of slightly older
children. Of the seven four-year-olds who produced narratives for me,
only one used and or and then to mark the segments of his stories, four
used guess what or you know what, and one child produced both kinds of
markers.
The child who produced the more conventional narrative markers, and
and and then, seemed more sophisticated on several counts than most of
the other children. He seemed to have an easier time with the story-
telling mode generally, telling more stories to all varieties of listeners
than 12 of the 14 children in my sample. He also was the only child to
produce a reasonably coherent—and quite extended—retelling of a story,
as illustrated in (5).
(5) I went on the subway to the movies to watch Pinocchio on the

movies. And his father been eaten up by a whale but he was
still alive.*** Well, he was in a boat and then the whale
dived out of the water and then dived and caught the boat
because it was in the sky and it was straight going to the
boat. And Pinocchio went diving down and he used a rock.
***They jumped on a raft and the raft broke apart and they
swimmed on the shore. And Pinocchio's father named Ge-
petto laid on the sand. And Pinocchio was laying on the
water. And the whale died because he bumped hisself on the
rocks.
I suggest, then, that conversational devices customarily used to gain

attention in an interaction get recruited to a different function in nar-
ratives. The children seem to know that marking the segments of a
narrative is reasonable. In lieu of control of the appropriate markers,
they borrow forms from another aspect of language use they know well.
318 / Marilyn Shatz
The result is a narrative with the outline of well-marked structure, but a
deceptively conversational style.
It might be argued that perhaps children using such forms are not
really producing narratives, but rather protonarratives—the relation of
two or more experienced events in sequence without the structure or
marking of narrative. I would argue this is not the case for several
reasons. The first is that phrases such as once, one day, and one time
appear right along with the conversational devices, but such time ad-
verbials occur only in narrative contexts and nowhere else in the child-
ren's speech. Another argument concerns the content of the narratives
in which the conversational phrases occur. You know what and the like
are not limited to the recounting of personal experiences. Examples (6)
and (7) illustrate that such devices occur in third person fantasy and in
first person fantasy as well. In both of these stories, the placement of
the marker is crucial: it sets off the resolution statement for the whole
narrative. Yet another indication that these devices are more a part of
the child's attempt at structuring his or her story and less an attentional
device is the fact that the devices occurred in stories directed to highly
attentive adults as well as to peers and older two-year olds. Thus,
attentiveness of the listener did not seem to be a criterion for their use.
(6) One day when the bear saw something in Tuney and Looney,
went downstairs and they saw their rhinosores. And one time
the rhinosore ate them, and they— *** Tuney is three and a
half and Snoofy is three and a half, and Goody is four and a
half. Ya know how they got away from him? They holded
hands and they runded downstairs all the way to their house,
and the rhinosore didn't get them.
(7) There was a big sea monster comin' onto the land. The sea
monster was getting all the—he was getting all the dead fish
for—to eat the crabs. *** I was way down there with my
father. My sister Annie and my mother were down there. I
was taking a walk. Soon I saw a sea monster get close to get
the crab in their teeth. You know what he did? The crab did
pinch me. We brought rubber bandaids, so I got a bandaid on
my feet.
However, this is not to say that listener characteristics have no

influence on the choice of devices. The four-year-old who produced the
sea monster story was later encouraged to repeat the story to one of our
least mature two-year-old listeners, who was barely verbal. As can be
seen by a comparison of (7) and (8), the story—and its structure—change
quite radically. Not only does the marker of the resolution disappear, so
does the resolution.
(8) I saw a sea monster. Wynne, there was a sea monster. But
Wynne, a crab (unintelligible) in the water, in and out. He ate
fish also. He was green all over. He really was green. See
big, big waves. Big rocks.
Interestingly, there are several other phrases in this retelling that seem
to function as attention-getters and are age-appropriate: the use of the
child's name and the directive see. It might appear that the narrator
was switching to a strategy of simply describing the picture he had been
shown. Unfortunately, in our picture, there weren't any big, big waves
or big rocks to be seen. Apparently, the task of creating a coherent
narrative for such a limited listener with the devices at his disposal was
a bit more than this four-year-old could manage.
Nonetheless, that the child was adjusting for his listener is confirmed
by a quantitative comparison of the two versions of the story. In Table
1, the child who told the sea monster story is Subject 1. It can be seen
that the mean length of utterances (MLU) in the story directed to the
adult was considerably longer than that directed to the two-year-old.
Moreover, even though each utterance in the story was more elaborated,
there were also more utterances directed to the adult. There were only
three children who were willing to tell the same story to both an adult
and a younger child, and each of the subjects shows pretty much the
same pattern of adjustment: more sentences and longer sentences to
adults. These are the same kinds of adjustments found in the conversa-
tional speech of these children (Shatz and Gelman 1973).
Table 1. Number of utterances and mean length of utterances for
stories told to both adult and two-year-old listeners.
Type of listener No. utterances MLU
in story in words
Subject 1 Adult 9 9.22
Two-year-old 8 5.00
Subject 2 Adult 5 8.6
Two-year-old 3 6.3
Subject 3 (a) Adult 10 7.3
(2 stories) Two-year-old 6 7.3
(b) Adult 8 14.5
Two-year-old 9 10.1
It should be noted, however, that narrative language in all our sub-

jects generally involved longer MLUs that did spontaneous conversation
about toys, regardless of listener. Again, this suggests that children's
narratives did differ somewhat from their more conversational styles. It
also suggests that the narrative task places more of a burden on the
child's sentence-making ability than does conversational speech. Thus,
yet a third factor, in addition to unfamiliar structural demands and
perspective-taking, adds to the difficulty of the narrative task—namely,
the increased use in narrative of more complex sentence structure.
Indeed, the use in narrative style of sequences of long, nonelliptical
sentences seems to be at odds with the directive style children typically
address to younger children—full of short, elliptical utterances, such as
see? or don't. Little wonder, then, that the task of telling stories to
two-year-olds proved difficult for the children.
One way that some children had of handling the demands of the
narrative task was to stay in familiar territory so far as content was
concerned. Four-year-olds' narrations were more often about personal
320 / Marilyn Shatz
experiences than about fantasy. Example (4) about the hurricane is a

typical example. Narrations of personal experience conform well to a
conversational interaction style. Moreover, the multiple memory cues
the child has for the order of events must make sequencing the story
relatively easy in comparison to retelling a nonexperienced story or
creating a sequence from scratch. Stories that are partly fantasy but
set in the context of real-world personal experience and told from a
first-person perspective are the second most frequent kind of story.
Again, the first person style should be the most familiar to the child, and
keeping either events, locations, or the cast of characters firmly
grounded in the familiar should ease the burden of production. The sea
monster tale (7) is an example of first person fantasy.
In my sample of four-year-olds' stories, there were no stories that
were completely original third-person creations. The only examples of
third person fantasy were the retelling of the Pinocchio movie (5) and
several more original efforts by the child who told the story of the
rhinoceros (6). Even she, however, relied on conventional characters
from television or story books to create her list of characters. As can
be seen in example (9), she also borrowed bits of well-known stories and
tucked them in among her original events. Thus, the danger of being
eaten and going to grandmother's house is more reminiscent of Little
Red Riding Hood than of Goldilocks' adventures.
(9) Once of a little time Goldilocks and hims friends went out on
a story show. They had a picnic on the story show. And then
the story—and then it went talkervingle *** Whaley get-there
were two whaleys. One was a bad one and one was a good
one. And the bad one wanted to eat Snoofy, and Tuney and
Rooney and Goldilocks. And then they all run in the water to
their Grandmother's house. And then the shark whale can't
get her. The whale holded the door so he—and he gotted back
and he was holding it with hims little shark pointey.
This example confirms that the use of familiar devices, characters, and
events functions like a prosthetic device, allowing children to stretch
their limited narrative abilities into sequences of discourse that are
clearly recognizable as narrative—chaotic though they may be.
Written narratives. As we turn to written narratives, it is worth
noting that children just a few years older than those I have been report-
ing on are considerably more willing and able to tell stories. They use
more standard narrative devices and they regularly tell and retell fanta-
sies. What about their first written stories? The data to be considered
come from only one child, but a particularly interesting one, since she
taught herself to read when she was five, before any formal training.
Having mastered reading, she set about writing, painstakingly practicing
letters and words and inventing nonstandard spellings as do many self-
taught preschoolers who read (Read 1975). Then, shortly before formal
schooling began, she started writing stories. Often these first attempts
were no more than two or three sentences long. Many were merely
descriptions of her daily life, and in that sense questionably narratives.
Yet, they were undeniably sequences of connected propositions, if not

actions or events. She called them stories, and she meticulously marked
most with the words The End in the appropriate place. Examples (10),
(11), and (12) illustrate some of these early efforts. Spelling has been
corrected, but grammar and punctuation are intact. Example (12) is
particularly interesting because it is in the first person, but obviously is
a fantasy of sorts.
(10) I like to be inside in my house when it is cold out there. My

house is hot and sometimes it is cold in my house.
(11) I love mommy. I love dad. IVIy mom and dad love me so we
can have fun.
The end.
(12) I am a snowflake floating down onto the ground. I can be a
snowman.
The end.
As the year progressed, the stories became longer and more elabo-
rate. Some began to get titled. Interestingly, however, these stories
show the same reliance on external supports that the younger children's
oral stories did. One of the most creative uses of external supports was
a story that recruited a nursery rhyme. It was titled 'A Song without
Music1 and is reproduced as example (13).
(13) A SONG WITHOUT MUSIC

A tisket a tasket a green and yellow basket. I wrote a letter
to my love and on the way I drop it. A boy picked it up and
put it in his pocket he took it home and drop on the way and
I picked it up. I took it to my love and went back to see my
ma. A tisket a tasket a green and yellow basket I write a
letter to my love and on the way I lost it. I went to my
loves and I made another note and went back home to see
my mommy working. My mommy said lets try to play the
song on the piano.
Another interesting aspect* of these more elaborate stories is that

they contain more errors of grammar and punctuation than the shorter
ones. Typically, the shorter stories were in the present tense. The
later, longer stories start out in the past tense and then often fluctuate
back and forth between present and past. Apparently, the child knew
that stories were told in the past, but it was difficult to maintain that
aspect of narrative style while doing all the other work required for a
longer story. This is aptly illustrated in the child's attempt to express in
narrative style a common experience in her family—the removal of ticks
from the adventuresome family dog (see example (14)). Similar switch-
ings of tense occurred in the longer oral narratives by four-year-olds
(see example (9)). Whereas the use of present tense can be a highlighting
device in narrative, it is questionable whether young children intend it as
such. Its rather arbitrary sprinkling among verbs marked for past in
example (14) and the increased frequency of present tense in longer over
shorter stories suggests instead that the mixing of tenses may be a
consequence of cognitive overload, not intentional highlighting.
322 / Marilyn Shatz
THE FAMILY
Once upon a time there were two children one was a boy and
one was a girl. The boys name was George, and the girls
name was Ann, and they live together of course. They lived
with their mother and father. And they even live with their
dog, which name is Toto. My mother and father have to
take his ticks out because he always gets ticks in his skin.
One night when my brother and my father and I were playing
Sorry, my dog came over to my father and my dog said I
have two ticks. And my father said tick-Toto and he did
have two ticks. And my brother Geoffrey said thanks for
telling us Toto that you had two ticks. Then we finished the
game and we had supper. Then I went right to bed and fell
asleep.
The end.
Within a year's time, and after considerable practice at the art of

story-writing, our young writer was indeed able to write a more coherent
narrative. Example (15) is of a third-person fantasy, with a title, a
proper beginning, a resolution, and a conclusion.
(15) WHY IT IS RAINING TODAY

One day the whole school had a big jar of water. They all
drank it all up. Their cheeks were very fat. Someone said
let's laugh. So they all laughed and the water went up in the
air and came down and poured and poured til there were
puddles everywhere.
The end.
One might quibble that the context is still a familiar one, or that the
referent for the school has not been properly established. Admittedly,
there were still a few things left to improve on in second grade. Yet,
the obvious progress in the child's written work over the course of the
year suggests that the knowledge expressible in oral narratives only
gradually transferred to the written mode. In the interval, some of the
devices I had observed four-year-olds using to make oral narratives a
manageable task again appeared as crutches in the written domain. As
the mechanics of writing and punctuation became better learned and
more practiced, much of what could be done orally began to show itself
again in the newer mode. The written narrative is the oral task taken to
a still higher level of complexity, requiring the intercoordination of the
nascent skills of writing and spelling with much that was learned in the
oral mode.
Summary. Both oral and written narration are complex tasks requir-
ing a period of practice at integrating the component skills of the task
before smooth performance can be achieved. Unfortunately, story-
telling is not quite like tennis, where one can practice strokes indepen-
dently from footwork or body stance, and after practicing all parts
separately, can work at putting them together. It is not very easy to
eliminate content and work on structure, or vice versa. Children make
modest attempts to do something like that when they narrate personal
experiences or when they borrow well-known devices from other regis-

ters to support their primitive narrative structures. Such strategies do
ease the processing burden of having to produce a novel speech genre.
Parents and teachers of young children apparently also help out by
providing conversational structure to carry the children through a nar-
rative. McNamee (1979) reports that listeners regularly ask questions
like What happened next? or And then what did he do? to help the child
structure and resolve narratives.
One implication of the arguments made here is that teachers of young
children might assist them in learning how to produce written narratives
by providing prosthetic devices for them. Especially for children in the
early primary grades, teachers might provide the children with the
written propositions of a story and let the children practice providing
the narrative devices structuring them into a story. Alternatively, they
could provide the structure and let the children produce the events.
My final point concerns future research on narrative development. I
have presented only a small amount of data from a small number of
children. Nevertheless, the data I have are quite consonant with other
reports in the literature—for example, the emphasis in early narratives
on personal experience. I have offered a cognitive process explanation
to account for these common characteristics found in children's narra-
tives. Yet, the universal processing constraints that all children—and
adults, for that matter—labor under should not be allowed to obscure
possible individual differences in the creative means developed to sur-
mount those limitations. We need to know more about how these means
develop and display themselves as the child overcomes processing limita-
tions. Longitudinal studies of the development of narrative production
in the same children as they transit from the oral to the written mode
should be particularly revealing.
Note
1. Asterisks indicate interruptions by listener.
References
Applebee, A. N. 1978. The child's concept of story. Chicago: The

Ames, L. B. 1966. Children's stories. Genetic Psychology Monographs
73.337-96.
Brown, A. L. 1975. Recognition, reconstruction, and recall of narrative
sequences by preoperational children. Child Development 46.156-66.
Labov, W. 1972. Language in the inner city: Studies in the Black Eng-
lish vernacular. Philadelphia: University of Pennsylvania Press.
Mandler, 3. 1983. Representation. In: Cognitive development, vol. III.
Edited by 3. H. Flavell and E. M. Markman. Handbook of child psy-
chology, 4th ed. Edited by P. H. Mussen. New York: John Wiley.
McNamee, G. D. 1979. The social interaction origins of narrative
skills. The Quarterly Newsletter of the Laboratory of Comparative
Human Cognition 1.63-68.
Pitcher, E. G., and E. Prelinger. 1963. Children tell stories: An anal-
ysis of fantasy. New York: International Universities Press.
324 / Marilyn Shatz
Read, C. N. 1975. Children's categorization of speech sounds in Eng-
lish. Research Report No. 17. Urbana, 111.: National Council of
Teachers of English.
Sacks, H. 1972. On the analyzability of stories by children. In: Direc-
tions in sociolinguistics. Edited by 3. 3. Gumperz and D. Hymes. New
York: Holt, Rinehart and Winston.
Scollon, R., and S. B. K. Scollon. 1983. Cooking it up and boiling it
down: Athabascan story retellings. In: Coherence in spoken and
written discourse. Edited by D. Tannen. Norwood, N. 3.: Ablex.
Shatz, M. 1983. Communication. In: Cognitive development, vol. III.
Edited by 3. H. Flavell and E. M. Markman. Handbook of child psy-
chology, 4th ed. Edited by P. H. Mussen. New York: 3ohn Wiley.
Shatz, M., and R. Gelman. 1973. The development of communication
skills: Modifications in the speech of young children as a function of
listener. Monographs of the Society for Research in Child Develop-
ment 38 (5, serial no. 152).
Stein, N. L., and C. G. Glenn. 1979. An analysis of story comprehension
in elementary school children. In: New directions in discourse pro-
cessing, vol. 2. Edited by R. Freedle. Norwood, N. 3.: Ablex.
CLARIFICATION AND CULTURE
Elinor Ochs
University of Southern California
0. Introduction. For some time now, scholars from several fields have
been grappling with the concept of the social event or social activity and
its importance. A primary concern has been the ways in which the
stream of behavior is divided and organized by members of a social
group. This concern has been articulated by a long list of scholars,
including Bateson (1972), Goffman (1974), and Minsky (1975), in their
discussions of event frames; ethnosemanticists (cf. Frake 1961, 1969), in
their discussions of event domains; Wittgenstein in his discussion of
language games (1958); Prague school linguists Uakobson 1960); and
ethnographers of speaking (Hymes 1974, Bauman and Sherzer 1975,
Gumperz 1983, Duranti in press), in their discussions of speech events;
cognitive psychologists, in their discussions of event schemata (Piaget
1929, Flavell 1977), event scripts (Schank and Abelson 1975, Nelson
1981, Nelson and Greundel 1981), and the ecological validity of experi-
mental tasks (Cole and Means 1981).
Certain discussions have been directed more specifically at the ef-
fects of mental representations of events (whether they be called
frames, schemata, scripts, or domains) on the production and interpre-
tation of behavior. Much of the work in artificial intelligence, for
example, concerns the role of knowledge of event goals in the interpre-
tations of particular behaviors (Grosz 1972). In another field, inter-
pretive anthropologists such as Gumperz have indicated (Gumperz 1983)
that interactions may break down when participants have vastly differ-
ent conceptualizations of the event taking place. When speakers from
different social groups interact, they may fail to understand how one
another's actions relate to the overall 1goal of the interaction, creating
what Gumperz has termed 'cross talk . Phenomenologically oriented
sociologists have been arguing for some time that even members of the
same social group do not always concur on their understandings of what
is going on between them. In the phenomenological perspective, partici-
pants of an interaction usually negotiate and cooperatively define and
construct the events taking place.
325
326 / Elinor Ochs
Still another concern in the study of events has been the impact of
participation in events on social, emotional, and cognitive develop-
ment. We have learned from several decades of intense research that
children bring biologically based capacities and dispositions to their
interactions with the world. Most influential have been Freud's discus-
sions of the role of instinct and impulse in emotional development (1960,
1965), and Piaget's argument that the child is an active agent in his
intellectual development, constructing action schemata from reflexes
and logic from action schemata. Of course, these same scholars have
stressed the impact of experience. Freud emphasized that construction
of one's ego and superego is influenced by life's experiences, and Pia-
getian research has emphasized that children construct knowledge
through their interactions with objects and persons in their environ-
ment. Freud's concern with the impact of social experience on one's
concept of self has been taken up by numerous social scientists, includ-
ing George Herbert Mead (1956), who proposed that one's sense of self is
influenced by the roles one habitually assumes in social interactions.
That is, one's sense of self is to a large extent a social construction,
constructed in and through participation in social activity. Currently, a
number of developmental psychologists have combined Piagetian models
of event schemata with cognitive science notions that knowledge is
organized in terms of event representations (cf. Bretherton 1984, Nelson
1981). This work suggests that children's understanding of objects,
persons, actions, states, and roles is a dimension of their understanding
of events at any one point in developmental time. Children display their
understanding of events through pretend activities, elicited retellings,
and descriptions of events.
In addition to these approaches, the Vygotskian school of Soviet psy-
chology, also called the sociohistorical or sociocultural approach, has
developed the idea that intrapersonal psychological processes emerge
not only in but through interpersonal ones, i.e. through social activities
(Vygotsky 1978, Luria 1976, Leontyev 1981, Wertsch 1980, in press,
LCHC 1981). In contrast to other approaches, this school has empha-
sized the role of knowledgeable persons in facilitating the acquisition of
higher order cognitive functions. Leontyev wrote, for example, 'The
individual, the child, is not simply thrown into the human world; it is
introduced into this world by the people around it, and they guide it in
that world' (1981:135). Further, the sociohistorical school has empha-
sized the role of society and culture in organizing activities. Vygotsky,
Luria, and Leontyev have all stressed the point, first, that activities
vary in content and structure across societies; and second, that this
variation has impact on members' cognitive skills.
In Europe and the United States, the Soviet approach has influenced
the work of scholars such as Bruner (1975), Cazden (1981), Cole (Cole
and Griffin 1980, Scribner and Cole 1981), Goody (1977), Greenfield
(Greenfield and Smith 1976), Griffin (Cole and Griffin 1980), Scribner
(Scribner and Cole 1981), and Wertsch (1980). This orientation is evident
in their research on the impact of literacy and schooling. The well-
known research of Scribner and Cole (1981), for example, indicates that
the development of cognitive skills within an individual is not so much
the effect of literacy per se but rather the effect of participating in
particular types of literacy activities. For example, participation in
Clarification and culture / 327
literacy activities characteristic of European schooling enhances the
development of hypothetical reasoning, whereas participation in literacy
activities characteristic of Koranic schooling does not. I have used the
term 'enhances' rather than 'determines' in discussing the effects of
participation, because we know that many factors influence participa-
tion in an activity. For example, individuals may involve themselves or
direct their attention to the activity to varying extents (Wentworth
1980). Further, early life experiences in literacy events differ (Heath
1983, Scollon and Scollon 1981, Michaels 1981), and these differences
affect children's participation in classroom literacy events at school.
That is, primary socialization experiences influence secondary socializa-
tion experiences.
All of this research on activities and events has important conse-
quences for understanding the relation between language, thought, and
culture. The sociohistorical approach in particular implies that not only
literacy activities but language activities in general have an impact on
social, emotional, and cognitive development. Along with anthropolo-
gical approaches (such as ethnography of communication, ethnoseman-
tics, and interpretative sociolinguistics), the sociohistorical approach
suggests that we need to examine closely the organization of language
activities, including the verbal means used to achieve goals, the sequen-
tial organization of verbal means, and the contexts in which goals,
means, and sequential orders are taken up by language users, and relate
these organizational patterns to cognitive skills and to systems of belief
and social order.
Further, this body of research calls for a reconsideration of the notion
of linguistic relativity. Let us consider again Sapir's classic statement
on this topic (quoted in Mandelbaum 1949:162).
It is quite an illusion to imagine that one adjusts to reality essen-

tially without the use of language and that language is merely an
incidental means of solving specific problems of communication or
reflection. The fact of the matter is that the 'real world' is to a
large extent unconsciously built up on the language habits of the
group...We see and hear and otherwise experience very largely as
we do because the language habits of our community predispose
certain choices of interpretation.
Sapir here speaks of language habits. Whorf spoke of fashions of speak-

ing. What is needed is to strip the linguistic relativity hypothesis of its
undesirable deterministic elements and preserve Sapir's notion that
language habits predispose certain choices of interpretation. The notion
of predisposition is akin to phenomenological views that experiential
frames influence construction of interpretations. It is also akin to the
sociohistorical view that habitual participation in language activities
enhances the emergence of certain psychological skills. We want to
make certain that we allow for creativity and individual difference in
our reconsidered theory of linguistic relativity. We want to say that
persons are oriented to ways of viewing the world through habitual
participation in language activities, but this process is open-ended.
World views developed through verbal interactions can be transformed
through further participation in language activities. The extent to which
328 / Elinor Ochs
such transformations occur depends on personal and social conditions,
but for most, socialization is a lifespan process.
While many would accept this interpretation of linguistic relativity,
opinions will vary concerning the scope of its application. Most accept
the idea that there is variation in literacy and school language activities
and that this variation has consequences for the acquisition of skills and
conceptual orientations. But what about earlier verbal activities that
infants and young children experience? What about just ordinary con-
versational activities in the child's social environment? Do these activ-
ities vary cross-culturally and does this variation have an impact on the
child's understanding of the world?
These are important questions to pose. Everyday, nothing-special sort
of conversational activity is the kind of social behavior that those in
cognitive science would call unscripted. It is not represented in memory
as a spatially or temporally bounded activity like going to the grocery
store or eating out at a restaurant or going to the doctor. It is ubiqui-
tous. Indeed, informal conversation is the most basic of verbal activities
and as such, it is the critical sociolinguistic context for the socialization
of knowledge and skills. If we want to know if language activities have
an impact on psychological development, then a most reasonable place
to look is the organization of everyday conversational discourse.
My own interest within this area is the relation between participating
in routine everyday conversational activities and the acquisition of
cultural knowledge. Over the past several years, Bambi Schieffelin and I
have worked collaboratively and independently, comparing, across sever-
al societies, coversational interactions in which children participate. We
have found that conversational activities involving small children vary in
ways that systematically relate to beliefs, values, and social order (Ochs
1982, in press; Ochs and Schieffelin 1982, in press; Schieffelin and Ochs
1983; Schieffelin 1979, in press). We have suggested that children ac-
quire sociocultural knowledge through exposure to and participation in
everyday, run-of-the-mill verbal exchanges. As Bateson (1972) has
noted, novices abstract from an event not only information specific to
that event but more general information concerning roles, relationships,
emotions, self, tasks, causality, temporal and spatial relationships, and
other dimensions of the sociocultural environment. Our orientation is
compatible with the sociohistorical approach and with phenomenoiogical
approaches to socialization such as those provided by Cicourel (1973),
Giddens (1976), and Wentworth (1980).
1. Clarification
1.1 Clarification and epistemology. I would like now to turn to a type

of conversational activity that is pervasive in the daily lives of all of us,
namely, the activity of clarification. Making clear our own and others'
behaviors is surely a universal endeavor, necessary for social order and
survival. I would like to put forward several suggestions concerning the
nature of clarification exchanges cross-culturally and their role in the
socialization of world view.
First, I would like to suggest that while clarification is a universal
activity, the manner in which clarification is accomplished varies cross-
culturally. Preferences for accomplishing clarification are embedded in
local principles of social order and local epistemologies. More specifi-

cally, I would like to suggest that both the conditions under which clari-
fication takes place (what gets clarified, who participates in the activity
of clarification, in which roles), and the discourse procedures speakers
prefer to use, index members' views of knowledge, particularly members'
views on the limits of knowledge (what can be known) and the paths to
knowledge (how knowledge is acquired). Another way of looking at this
is to say that when members engage in the activity of clarification, they
display and construct tacit guidelines and principles for creating knowl-
edge. These guidelines and principles in turn are tied as well to local
theories of meaning, of learning, and of self. I am interested in those
cases of clarification where the participants are caregivers and young
children. As caregivers involve infants and small children in clarifica-
tion exchanges, they are displaying and constructing with them more
general, socially valued epistemologies. In Sapir's terms, the language
habits of their communities predispose children to view knowledge in a
certain light. In the following discussion, I examine patterns of clarifi-
cation between children and caregivers and their relation to folk episte-
mologies in two societies: American White Middle Class (WMC) and
Western Samoan.
1.2 Structure of clarification sequences. Let us now try to formulate

a working definition of the activity of clarification. A clarification
sequence contains a verbal or nonverbal behavior that is seen as unclear
by at least one participant to an interaction. Unclarity may involve both
surface expression and/or underlying meaning. An utterance, for exam-
ple, may be unintelligible because it has been poorly articulated, because
it has not been heard, and so on. On the other hand, even when the
surface form of an utterance is intelligible, its meaning may not be
clear.
Using the terminology of conversation analysis, we can say that the
unclear behavior is a trouble source for some participant and that the
clarification sequence attends to the work of repairing or attempting to
resolve that trouble (Schegloff, Jefferson, and Sacks 1977). That is,
clarification is a goal of at least one participant.
As a type of repair sequence, the clarification sequence has the struc-
tural options that have been noted for repair sequences generally.
Clarification may be self-initiated (i.e. initiated by the party who pro-
duces the unclear behavior) or other-initiated. Attempts to clarify may
also be carried out either by self or by other.
13 Other-initiated clarification in two societies. I would like now to

turn to one of these structural varieties, namely, those clarification
sequences in which clarification is other-initiated. This type of clarifi-
cation sequence has received considerable attention in the language
acquisition literature, because transcripts are laced with children's
utterances and nonverbal behaviors that are followed by caregivers'
initiations of clarification, as illustrated in examples (1) and (2).
330 / Elinor Ochs
(1) Jordan is a l4-month-old male infant, being served his lunch.
(From Golinkoff 1983:58-59.)
1. Jordan: (Vocalizes repeatedly until his mother turns
around.)
2. Mother: (Turns around to look at him.)
3. Jordan: (Points to one of the objects on the counter.)
4.—) Mother: Do you want this? (Holds up milk container.)
5. Jordan: (Shakes his head 'no'.)
(Vocalizes, continues to point.)
6.—> Mother: Do you want this? (Holds up jelly jar.)
7. Jordan: (Shakes head 'no'.)
(Continues to point.)
8, 9, 10, 11. (2 more offer-rejection pairs.)
12.-* Mother: This? (Picks up sponge.)
13. Jordan: (Leans back in highchair, puts arms down, tension
leaves body.)
1^. Mother: (Hands Jordan sponge.)
(2) Allison is 16 months, 3 weeks old. (From Bloom 1973:152-53.)
1. Mother: What do you see?
2. Allison: (A leans forward; looking in bag)
Pig/
(A stands up)
3 . - -> Mother: What?
4. Allison: Pig/
5 . - -> Mother: Play? Is that what you're saying? Play?
6. Allison: oh/ pig/ — /
For some researchers, these caregiver responses have been taken as
evidence that caregiver speech facilitates the acquisition of grammar
(Cross 1977, 1981). For others, these responses have been treated as a
means by which caregivers are able to sustain communication with a
young baby or child (Brown 1977, Snow 1977).
Western Samoan and WMC caregivers both initiate clarification of
children's verbal and nonverbal behavior. However, the set of proce-
dures used by rural Western Samoan caregivers to initiate clarification is
a subset of those used by WMC caregivers.
1.3.1 WMC clarification strategies. WMC caregivers rely heavily on

two related strategies for initiating clarification. The first strategy is
to exhibit minimal or no grasp of what the child has said or done and to
rely primarily on the child to resay or redo the unintelligible utterance
or gesture. Let us call this strategy the 'minimal grasp strategy 1 . This
may be accomplished indirectly by the caregiver expressing nonunder-
standing through a quizzical expression or through a verbal statement
such as I don't understand, I can't understand what you are saying, and
the like. Or the caregiver may directly ask the child what he said or to
supply a piece of what he said, using WH interrogatives such as What?
Who? He went where?, and so on. This type of clarification request is
illustrated in example (2), line 3. The caregiver may also request or
order the child to resay or redo through utterances such as Say it again
sweetie, Show me another time, Could you say it once more? and so on.
A second strategy of WMC caregivers is to articulate a guess at what
the child's unclear utterance or gesture could be ox could mean. Let us
call this strategy the 'expressed guess strategy'. In contrast to the
minimal grasp strategy, here it is the caregiver who attempts a reformu-
lation of the unclear act. The child is asked to validate or confirm the
caregiver's guess. This strategy is illustrated in example (1), lines 4, 6,
and 12, and in example (2), line 5. In the case of disconfirmation, the
child may resay or redo his utterance or gesture and the caregiver may
continue to supply alternate guesses, as illustrated in example (1).
The speech act of guessing covers a range of uncertain knowledge. A
caregiver or any speaker may formulate a guess when she is not at all
certain of her knowledge. In interactions with infants, this is often the
case. Caregivers often find themselves articulating wild guesses at what
the infant could be signalling. On the other end, caregivers and others
may formulate guesses when they are fairly certain of what an infant is
saying or doing. In these cases, the caregiver is using the guess to make
sure of, or to double check, her understanding.
Not only WMC caregivers but WMC speakers generally prefer con-
structions that display the most of what they have understood of the
problematic utterance. That is, speakers show a preference for using
the strongest form they can in initiating repair of another's utterance
(Schegloff, personal communication). For example, speakers prefer
specific interrogative pronouns (Where? Who?, etc.) over the weaker
construction Huh?, and prefer partial repetitions plus an interrogative
pronoun (He went where?, etc.) over the interrogative pronoun on its
own. Relevant to our concerns here, the preference of WMC speakers is
for the expressed guess strategy over the minimal grasp strategy, where
conditions of hearing and understanding permit.
1.3.2 Samoan clarification strategies
1.3.2.1 Caregiver strategies. In Western Samoan households, care-

givers prefer strategy 1 (the minimal grasp strategy) but not strategy 2
(the expressed guess strategy) to initiate clarification. They use quiz-
zical expressions, statements of nonunderstanding, WH questions, and
other directives to elicit from the child a reformulation of all or part of
the unclear utterance or gesture. An example of this strategy is pro-
vided in (3).
(3) Maselino (4 years) is with Sililo, (16 years), Olagi (5 years),

and his mother's brother's wife, Atoa.
Mas (to Sil): Mai Liaga le kusia sou igoa le kegi

'Uliana said they are not going to write your name
(on the list of workers) for the gang.'
Sil: E aa?
'What?'
Olagi: // ((laughs))
Mas: //mai Liaga le kusia e sau igoa le kegi e e.
'Uliana said that they are not going to write your
name (on the list of workers) for the gang (warning
particle).'
332 / Elinor Ochs
Atoa (to Sil): Mai e aa? ( )
'She said what? 1
Sil: Mai le kusia so'u igoa le kegi.
'She said they are not going to write my name (on
the list of workers) for the gang.1
In the corpus of interactions that we recorded, we did not find cases in

which caregivers (either sib caregivers or adult caregivers) formulated
an explicit guess at what an unclear utterance or gesture of the child
may be. This dispreference was also manifest when members of the
family or others would listen with me to recordings of children's unclear
utterances. Almost everyone found my own enterprise of explicitly
guessing at a garbled or telegraphic utterance of a child puzzling and not
worth the time.
This does not mean that these caregivers and others listening to and
watching children do not guess silently. However, silent guesses differ
from expressed guesses. First, expressed guesses make explicit a possi-
ble proposition. Expressed hypotheses, conjectures, and speculations all
commit their makers tentatively to the possibility that some state of
affairs may hold. Second, expressed guesses elicit the involvement of
the original speakers who produce the enigmatic utterances in the pro-
cess of understanding, whereas the silent guess does not. In the case of
caregiver-child interaction, the expressed guess of the caregiver gives
the child a role in the assignment of meaning; the child is given veto
power, so to speak, over the caregiver's understanding. In the expressed
guess, then, meaning is negotiated before it is assigned. In the silent
guess, any negotiation of meaning that may occur takes place after the
caregiver's initial assignment of meaning.
The Samoan preference for repetition is manifested more generally in
situations in which instruction is taking place. As in many societies,
Samoans rely heavily on repeated, often passive, observation of behav-
iors as a means of transmitting and acquiring knowledge and skills.
Dance practice, for example, consists of one person modelling entire
dances over and over in front of novices, who imitate the dance move-
ments or watch to one side. As Samoan caregivers engage young chil-
dren in clarification sequences, they are then socializing them into
broader, socially valued methods of education, namely, that the path to
knowledge is through repeated exposure—through listening and watching
over and over.
In WMC society, repetition of information is also an important strat-
egy in the transmission and acquisition of new information. However,
the tradition of clarifying thought through Socratic, dialogic methods is
also strong in this society. In the Socratic method, knowledge is pursued
through formulating and pursuing initial hypotheses, that is, through
laying out for others explicit guesses. WMC caregivers who initiate
clarification of children's utterances or gestures through yes-no inter-
rogatives or other forms of guessing are socializing children into this
socially valued procedure for gaining knowledge, just as when they elicit
resayings or redo ings, they are socializing them into the alternative
procedure whereby knowledge is enhanced through repeated
observations.
2. Social rank. Goody (1978) has noted that among the Gonja of
Northern Ghana, the use of questions is socially constrained. In adult-
child interactions, questions are appropriate speech acts of adults but
not of young children. In Samoan society, the speech act of guessing is
also affected by social status.
Samoan society is highly stratified. Rank is assessed in terms of
political title (e.g. chief, orator, and positions within each of these
statuses), church title (pastor, deacon, etc.), age, and generation, among
other variables. Titled persons have higher rank than untitled persons
and older, higher generation persons have higher rank than younger
persons (Mead 1930, Shore 1982). Among the demeanors associated with
distinctions in social rank is that of perspective-taking. Lower ranking
persons are expected to assume the perspective of higher ranking per-
sons more than higher ranking vis-a-vis lower ranking parties in a social
situation. Lower ranking persons are expected to notice and anticipate
the wishes of higher ranking persons. They stand in a service relation to
those of higher status. As I have noted elsewhere (Ochs 1982), young
Samoan children are socialized early in their lives to a sociocentric
perspective. As infants, they are often held and fed facing outward
toward others in a group. When they begin to speak, much time and
effort is devoted to instructing the young child to notice others and to
repeat their personal names. In Samoan society, sib and parental care-
givers work hard to get children, even before the age of two years, to
take the perspective of others. This demeanor is a fundamental com-
ponent of showing respect, a most necessary competence in Samoan
daily life.
The process of communication is affected by these social expectations
concerning perspective-taking. It is obvious that communication re-
quires degrees of perspective-taking by all participating parties, i.e.
degrees of what has been called 'intersubjectivity 1 (Trevarthen 1979).
In Samoan interactions the extent to which parties are expected to
assume the perpective of another in assigning a meaning to an utterance
of another varies with social rank. In speaking to those of lower rank,
higher ranking persons are not expected to do a great deal of perspec-
tive-taking to make sense out of their own utterances or to make sense
of the utterance of a lower ranking interlocutor. Higher ranking per-
sons, then, are not expected to clarify and simplify for lower ranking
persons. For example, caregivers are not expected to simplify their
speech in talking to young children (Ochs 1982). And exactly the reverse
is expected of lower ranking persons. Lower ranking persons take on
more of the burden of clarifying their own utterances and the utterances
of higher ranking interlocutors.
Of the two clarification strategies discussed earlier, the 'expressed
guess1 strategy involves more perspective-taking than the 'minimal grasp'
strategy. One reason why we do not see caregivers making explicit
guesses at what their charges are saying is that such a response demands
an orientation that is generally inappropriate to the social role of care-
giver. Only in situations in which a small child is speaking on behalf of
someone of high status (e.g. when the child is a messenger) is this degree
of perspective-taking expected. Typically, when very small Samoan
children produce unintelligible utterances, they are disregarded or
addressed with a construction indicating noncomprehension and directed
33* / Elinor Ochs
to redesign their utterances to meet the communicative needs of
others. Through such procedures, children develop early in life a sensi-
tivity to the demands of their social environment and communicative
skills to meet them.
Looking at transcripts of interactions across many contexts (adult-
adult, adult-child, child-child), I have found few instances of explicit
guessing. Of those instances located, most occur in interactions among
peers and a few occur in interactions in which a higher ranking person
has produced an unclear utterance. While guessing appears across sever-
al speech activities in peer interaction, when a lower ranking person
directs a guess at a higher ranking person it is situationally con-
strained. As audience to personal narratives, gossip, or speeches of
higher ranking persons, lower ranking persons do not typically guess
explicitly at the meaning of their utterances. However, when a higher
ranking person directs the lower ranking person to do something, then he
may clarify by directing a guess to the speaker. I emphasize that this
strategy is not very frequent. It is generally dispreferred for lower
ranking persons to guess at the utterances of higher ranking persons.
The expectation is that the lower ranking person should be attending
(and therefore not need to clarify on grounds of not having heard the
utterance) and should understand. In multiparty situations, lower rank-
ing persons may get out of this bind by directing to a co-present peer a
guess as to part or all of the utterance of the higher ranking person.
This strategy is illustrated in example (4), in which a group of boys of
differing ages are playing on the beach, pretending to be preparing a
meal. In their play, the older boys direct the preparation and the
younger boys carry out the directives (just as in daily life). The oldest
boy (Boy 1) directs a younger one (Sesi, Boy 3) to make saka 'boiled
taro'. The younger boy then turns to a boy close to his age and requests
confirmation of his understanding of what was said.
(4) Boys Playing on Beach

Boy 1: Sole, alu Sesi fai saka ee!
'Mate, go Sesi to make saka (emph. particle)! 1
Ke iloa fai~
'You know how (to make it)'
Boy 2: ((hums))...eli ma'a
((hums))...'dig stones.1
Boy 3: Fai mai 'Fai saka'?
'He said 'Make saka'?
Boy 1: Sole, alu oe e e (pause) koli mai ulu.
'Mate, you go to twist off and fetch down breadfruit.'
3. The trouble- In addition to social rank, the nature of the trouble or
the object of clarification is an important variable constraining the use
of explicit guessing in Samoan interactions. In the WMC caregiver-child
interactions observed, the clarification sequences pursue at least two
major goals. One is to clarify what the child has just said or done, that
is, to obtain an output that is intelligible. A second goal is to assign a
reading to that output that is compatible with the child's intended mean-
ing. In all speech, but particularly in children's speech, utterances may
have several meanings. In WMC caregiver-child interactions, a major
problem is to sort out which meaning is the 'correct' one, where correct-
ness is based on the caregiver's assessment of the child's intentions (what
Grice 1968 calls 'utterer's meaning1). So important is the understanding
of the child's intentions that caregivers will check with the child if their
understanding of the child's intended meaning is correct or not. This job
is accomplished through the expressed guess. In guessing, the caregiver
displays a tentative reading before a final interpretation. The child has
an option, indeed is directed, to influence the caregiver's understanding
of some particular utterance or action before a meaning is assigned.
In so doing, WMC caregivers are conforming to a cultural theory of
communication in which speakers' personal intentions are critical to the
interpretation of an utterance or action. Certain philosophical theories
of meaning, such as that of Searle (1969) and Grice (1968) articulate the
system of knowledge that underlies this folk theory. In the work of both
Searle (following Austin 1962) and Grice, the issues taken up focus on
the relation between convention and intention, locutionary and illocu-
tionary meaning, sentence and utterer's meaning, evaluating the relative
importance of each in a theory of meaning and language use.
Recently, several sociolinguists and anthropologists have discussed
this orientation to meaning in relation to cultural beliefs and orienta-
tions (Duranti 1984, Kochman 1983, Ochs 1982, Rosaldo 1982, Shore
1977, 1982). All of these discussions have focused on the concept of
person that emerges from language behavior and from folk and academic
theories of meaning. The emphasis on personal intentions in Anglo
society and scholarship is tied to a cultural ideology in which persons are
viewed as individuals, i.e. coherent personalities, who have control over
and are responsible for their utterances and actions.
Personal intentions are important in a vast range of situations. Mem-
bers of Anglo WMC society seek to clarify an individual's personal inten-
tions for a range of purposes. For example, members of this society
usually base their assignments of responsibility and sanctions on the
speaker/actor's particular intentions behind an utterance or action. This
society distinguishes, for example, between inadvertent and planned
behaviors, and between accidental and purposeful behaviors. In legal and
other contexts, if it is established that a negatively valued behavior was
consciously intended, then sanctions1 are usually more severe than if the
speaker/actor 'didn't mean to do it or could not help doing it or other-
wise was not in control. Note that establishing intentionality is not
always critical to sanctioning. In many situations, members of this
society say 'It doesn't matter whether you meant it or not.' The impor-
tant point is that in Anglo-American WMC society, what a person means
or meant to do or say is an important cultural variable. For this social
group, what a person means to do is distinguished from what he does.
This orientation leads members to take seriously, and to pursue the
establishing of, individual's motivations and psychological states.
This concern with and emphasis on personal intentions is not matched
in other societies. In societies such as American Black working class
(Kochman 1983), Ilongot (Rosaldo 1982), Ifaluk (Lutz 1982), and Samoan
(Duranti 1984, Ochs 1982), the consequences of an utterance or action
play an important role in assigning meaning.
In certain accounts, the emphasis on consequences takes the form of
focusing on the social ramifications of a behavior (rather than on
336 / Elinor Ochs
speaker/actor's intentions). Lutz (1982), for example, notes that the

Ifaluk focus on the 'wake' of an action. In Ochs (1982), I have discussed
the primacy of consequences of an action in Samoan evaluations of
actions. In Samoan households, children will be sanctioned according to
the negative effects of their behaviors. This is also the case in the legal
arena, where actions are assessed almost exclusively in terms of social
and economic losses and disturbances. In the context of assessing mis-
deeds, in Samoan society, the focus is much less on personal intentions
behind an utterance/action. In this context, it is not terribly important
if the wrongdoer did something by accident, inadvertently, or on pur-
pose. Indeed, Samoans see persons as not in control of their misdeeds
(Shore 1977, 1982). Samoan children may try to get out of punishment
by denying that they did that culpable act, but they do not try to worm
out by saying I didn't mean it, It was just an accident, I did it by mistake,
I couldn't help it, I didn't do it on purpose, as do WMC children almost by
routine.
Other accounts, following a more phenomenological approach to
communication, have focused on the importance of the hearer's role in
the assignment of meaning. Kochman (1983), for example, has com-
mented that for this community of Black speakers, very often the per-
locutionary effect on the hearer takes precedence over speaker's in-
tended meaning. Indeed, here as in other societies such as Samoan
society, speakers often leave ambiguous what is meant, waiting to see
how a hearer will take it up. In this sense, meaning is in the hands of the
audience more than in those of the speaker; the audience has the final
word.
Taking these accounts altogether, we might propose that we have
found a variable in terms of which societies contrast. There are soci-
eties like the WMC in the United States that focus primarily on the
personal sources of utterances/actions and other societies—such as the
Ilongot of the Philippines, the Ifaluk of the Caroline Islands, the
Samoans, and the working-class 31acks in the United States—that focus
primarily on the social consequences of utterances/actions.
This distinction, however, is too simplistic. For example, there are
theories supported by scholars within the WMC society in the United
States that argue against the primacy of personal intentions in establish-
ing meaning. Sociohistorical theories of meaning such as that held by
Bakhtin (Volosinov 1973), deconstructionist theories within literary
criticism, and hermeneutic perspectives (e.g. Gadamer 1976) are alive
and popular within this country. (Notice, however, that these traditions
stem from scholarly lines outside the United States.) This observation
and ethnographic observations of Samoan interaction suggest that within
each society, both orientations persist. The difference between soci-
eties lies in the contexts in which these two orientations prevail, the
relative importance given to each of them, and the frequency with which
these orientations mark social interaction.
In Samoan society, personal intentions are a focus of concern in a
restricted set of contexts, primarily when the speaker/actor is of high
social status and/or of higher social rank relative to the hearer/audi-
ence. For example, Shore (1977, 1982) and Duranti (1981) have noted
that in the context of political meetings of titled persons, only high
chiefs and high status orators are entitled to voice personal opinions. In
this sense, high status speakers in this context are treated more as
individuals than are others present, and their personal intentions are
attended to. In addition, when a higher ranking person orders a lower
ranking person to carry out some action, personal intentions of the
speaker are also of primary importance. The lower ranking party cannot
assign his own interpretation but rather must grasp that intended by the
higher ranking speaker.
Where the speaker is of low status and/or of lower rank than the
hearer, then his or her personal intentions tend to assume low priority in
assigning meaning and the interpretation of the higher ranking hearer
takes precedence. Notice that whether the higher ranking party is
speaker or hearer, that high party controls meaning.
Given that explicit guessing is tied to the pursuit of speaker's inten-
tions, it is somewhat understandable, in light of the foregoing comments,
that we would observe very little explicit guessing directed to lower
ranking speakers. The personal intentions of lower ranking speakers,
such as children talking to caregivers, do not 'count' in the same way as
do those of higher ranking speakers. It would be particularly improbable
for caregivers to direct guesses at infants, since at this early point in
life, infants are seen neither as personalities nor as conversational
partners (Ochs 1982).
While the two perspectives on meaning are variable within WMC and
Samoan society, the two have different contextual distributions and
salience in each of the two societies. That theory of meaning which
Holquist (1983) calls the 'personalist' view of meaning (the view that 'I
(the speaker) own meaning') is far more salient in WMC society than in
traditional rural Western Samoan communities. When WMC caregivers
attend very carefully to the unclear gestures and utterances of their
infants and young children, when they explicitly guess at what the child
means, they are socializing children into this prevailing view of meaning
in which personal intentions are of primary importance. The absence of
explicit guessing by Western Samoan caregivers is tied to the restricted
relevance of this theory of meaning to Samoan social life, in particular
to its inappropriateness in a wide range of contexts, including those in
which children communicate with caregivers. Samoans generally display
a strong dispreference for guessing at what is going on in another per-
son's mind. This dispreference has reflexes in a range of verbal activi-
ties and accounts for the rarity of activities such as test questions,1
riddles, and guessing games of the 'Twenty Questions' and 'I Spy
variety. These activities are not part of traditional instruction settings
nor are they common in informal adult-child, adult-adult, or child-child
interactions. (They appear mainly in the context of formal classroom
instruction in Christian churches and Western-oriented public schools.)
Western Samoan caregivers1 behaviors, then, are congruent with tradi-
tional Samoan theories of knowledge, including their theories of learning
and their theories of meaning.
Notes
I am grateful for the helpful comments of Elaine Andersen, Yigal
Arens, Niko Besnier, Alessandro Duranti, and Edward Finegan, and for
338 / Elinor Ochs
the long discussions with Emanuel Schegloff on earlier drafts of this
research paper.
1. I am indebted to E. Schegloff for providing this term.
2. This strategy is roughly comparable to the notion of 'candidate
understanding1 within the paradigm of conversation analysis (Schegloff,
personal communication).
References
Austin, J. L. 1962. How to do things with words. Oxford: Oxford
University Press.
Bateson, G. 1972. Steps to an ecology of mind. New York: Ballantine.
Bauman, R., and 3. Sherzer. 1975. Explorations in the ethnography of
speaking. Cambridge: Cambridge University Press.
Bloom, L. 1973. One word at a time. The Hague: Mouton.
Bretherton, I. 1984. Symbolic play. New York: Academic Press.
Brown, R. 1977. Introduction. In: Talking to children: Language input
and acquisition. Edited by C. E. Snow and C. Ferguson. Cambridge:
Cambridge University Press. 1-27.
Bruner, 3. 1975. The ontogenesis of speech acts. Journal of Child
Language 2:1-19.
Cazden, C. 1981. Performance before competence: Assistance to child
discourse in the zone of proximal development. Quarterly Newsletter
of the Laboratory of Comparative Human Cognition 3.5-8.
Cicourel, A. 1973. Cognitive sociology. London: Macmillan.
Cole, M., and M. Griffin. 1980. Cultural amplifiers reconsidered. In:
Social foundations of language and thought. Edited by D. Olson. New
York: W. W. Norton and Co.
Cole, M., and B. Means. 1981. Comparative studies of how people
think. Cambridge, Mass.: Harvard University Press.
Cross, T. 1977. Mothers' speech adjustments: The contribution of
selected child-listener variables. In: Talking to children: Language
input and acquisition. Edited by C. E. Snow and C. Ferguson. Cam-
bridge: Cambridge University Press. 151-88.
Cross, T. 1981. The linguistic experience of slow language learners.
In: Advances in child development. Edited by Nesdale et al. Univer-
sity of Western Australia.
Duranti, A. 1981. The Samoan fono: A sociolinguistic study. Pacific
Linguistics, series B, vol. 80. Canberra: The Australian National
University, Department of Linguistics, R.S. Pac.S.
Duranti, A. 1984. Intentions, self, and local theories of meaning: Words
and social action in a Samoan context. MS. Laboratory of Compara-
tive Human Cognition. University of California, San Diego.
Duranti, A. (in press) Sociocultural dimensions of discourse. In: Hand-
book of discourse analysis. Edited by T. van Dijk. London: Academic
Press.
Duranti, A., and E. Ochs. 1982. The social organization of other-cor-
rection in Samoan verbal interaction. Paper presented at the Ameri-
can Anthropological Association Annual Meetings, Washington, D.C.
Flavell, J. 1977. Cognitive development. Englewood Cliffs, N.J.:
Prentice-Hall.
Frake, C. 1961. The diagnosis of disease among the Subanun of Min-
danao. American Anthropologist 63.1:113-32.
Frake, C. 1969. A structural description of Subanun religious be-
havior. In: Cognitive anthropology. Edited by S. Tyler. New York:
Holt, Rinehart and Winston.
Freud, S. 1960. The ego and the id. New York: Norton.
Freud, S. 1965. New introductory lectures on psychoanalysis. New
York: Norton.
Gadamer, H-G. 1976. Philosophical hermeneutics. Translated and
edited by D. E. Linge. Berkeley: University of California Press.
Giddens, A. 1976. New rules of sociological method. New York: Basic
Books.
Goffman, E. 1974. Frame analysis. New York: Harper and Row.
Golinkoff, R. 1983. The pre-verbal negotiation of failed messages:
Insights into the transition period. In: The transition from
prelinguistic to linguistic communication. Edited by R. Golinkoff.
Hillsdale, N.J.: Lawrence Erlbaum. 57-78.
Goody, E. 1978. Towards a theory of questions. In: Questions and
politeness. Edited by Esther Goody. Cambridge: Cambridge Univer-
sity Press. 17-43.
Goody, J. 1977. The domestication of the savage mind. Cambridge:
Greenfield, P., and 3. H. Smith. 1976. The structure of communication
in early language development. New York: Academic Press.
Grice, H. P. 1968. Utterer's meaning, sentence-meaning, and word-
meaning. Foundations of Language 4.1-18.
Grosz, B. 1977. The representation and use of focus in dialogue under-
standing. SRI International Technical Note 151. Menlo Park, Calif.:
Stanford Research Institute.
Gumperz, J. 1983. Discourse strategies. Cambridge: Cambridge
University Press.
Heath, S. 1983. Ways with words: Language, life, and work in com-
munities and classrooms. Cambridge: Cambridge University Press.
Holquist, M. 1983. The politics of representation. Quarterly News-
letter of the Laboratory of Comparative Human Cognition 5.1:2-9.
Hymes, D. 1974. Foundations in sociolinguistics: An ethnographic
approach. Philadelphia: University of Pennsylvania Press.
Jakobson, R. 1960. Linguistics and poetics. In: Style in language.
Edited by T. Sebeok. Cambridge, Mass.: The MIT Press. 350-77.
Kochman, T. 1983. The boundary between play and nonplay in black
verbal dueling. Language in Society 12.3:329-38.
LCHC CLaboratory of Comparative Human Cognition, University of
California, San DiegoJ. 1981. Culture and cognitive development. To
appear in: L. Carmichael's manual of child psychology: History,
theories, and methods. Edited by W. Kessen. New York: John Wiley.
Leontyev, A. N. 1981. Problems of the development of mind. Mos-
cow: Progress Publishers.
Luria, A. R. 1976. Cognitive development: Its cultural and social
foundations. Cambridge, Mass.: Harvard University Press.
Lutz, C. 1982. The domain of emotion words on Ifaluk. American
Ethnologist 9.1:113-28.
3*0 / Elinor Ochs
Mandelbaum, D. G., ed. 1949. Selected writings of Edward Sapir.
Berkeley/Los Angeles: University of California.
Mead, G. H. 1956. On social psychology. Chicago: University of Chi-
cago Press.
Mead, M. 1930. The social organization of Manu'a. Honolulu: B. P.
Bishop Museum.
Michaels, S. 1981. Sharing time: Children's narrative styles and dif-
ferential access to literacy. Language in Society 10.3:423-42.
Minsky, M. 1975. A framework for representing knowledge. In: The
psychology of computer vision. Edited by P. Winston. New York:
McGraw-Hill.
Nelson, K. 1981. Social cognition in a script framework. In: Social
cognitive development. Edited by 3. Flavell and L. Ross. New York:
Nelson, K., and 3. Gruendel. Generalized event representations: Basic
building blocks of cognitive development. In: Advances in develop-
mental psychology, vol 1. Edited by M. Lamb and A. Brown. Hills-
dale, N.3.: Lawrence Erlbaum.
Ochs, E. 1982. Talking to children in Western Samoa. Language in
Society 11:77-104.
Ochs, E. (in press) Culture and language acquisition: Acquisition of
communicative competence in a Western Samoan village. New York:
Ochs, E., and B. Schieffelin. 1982. Language acquisition and socializa-
tion: Three developmental stories. In: Working Papers in Sociolin-
guistics. Edited by R. Bauman and 3. Sherzer. Austin, Texas: South-
west Educational Laboratory. To appear also in: Culture theory.
Edited by R. Schweder and R. Levine. New York: Cambridge Univer-
sity Press.
Piaget, 3. 1929. The child's conception of the world. London: Rout-
ledge and Kegan Paul.
Rosaldo, M. 1982. The things we do with words. Ilongot speech acts
and speech act theory in philosophy. Language in Society 11:203-37.
Schank, R.C., and R.P. Abelson. Scripts, plans, goals and understand-
ing. Hillsdale, N.3.: Lawrence Erlbaum.
Schegloff, E., G. Jefferson, and H. Sacks. 1977. The preference for
self-correction in the organization of repair in conversation. Lg.
53.361-82.
Schieffelin, B. (in press) How Kaluli children learn what to say, what to
do, and how to feel. New York: Cambridge University Press.
Schieffelin, B., and E. Ochs. 1983. Cultural perspectives on the transi-
tion from prelinguistc to linguistic communication. In: The transition
from prelinguistic to linguistic communication. Edited by R. Golin-
koff. Hillsdale, N.3.: Lawrence Erlbaum.
Scollon, R., and S. Scollon. 1981. Narrative, literacy, and face in inter-
ethnic communication. Norwood, N.3.: Ablex.
Scribner, S., and M. Cole. 1981. The psychology of literacy. Cam-
bridge, Mass.: Harvard University Press.
Searle, 3. 1969. Speech acts. Cambridge: Cambridge University Press.
Shore, B. 1977. A Samoan theory of action. Doctoral dissertation.
University of Chicago.
Clarification and culture / 3*1
Shore, B. 1982. Sala'ilua: A Samoan mystery. New York: Columbia
University Press.
Snow, C. E. 1977. The development of conversations between mothers
and babies. Journal of Child Language 4.1-22.
Trevarthen, C. 1979. Communication and co-operation in early infan-
cy: A description of primary intersubjectivity. In: Before speech:
The beginnings of interpersonal communication. Edited by M.
Bullowa. Cambridge: Cambridge University Press. 321-47.
Volosinov, V. N. 1973. Marxism and the philosophy of language. New
Vygotsky, L. 1978. Mind in society. Cambridge, Mass.: Harvard Uni-
versity Press.
Wentworth, W. 1980. Context and understanding: An inquiry into
socialization theory. New York: Elsevier.
Wertsch, J. 1980. The significance of dialogue in Vygotsky's account of
social, egocentric and inner speech. Contemporary Educational Psy-
chology 5.150-62.
Wertsch, J. (in press) Cognitive development theory: A Vygotskian
perspective. Cambridge, Mass.: Harvard University Press.
Wittgenstein, L. 1958a. Philosophical investigations. Oxford: Black-
well.
Wittgenstein, L. 1958b. The blue and brown books. Oxford: Blackwell.
( ontents
Ruth M. Kempson. Pragmatics, anaphora, and logical form

Laurence R. Horn. Toward a new taxonomy for pragmatic inference
William Labov. Intensity
Michael L. Geis. On semantic and pragmatic competence
Sandra A. Thompson. 'Subordination* in formal and informal discourse
Wallace Chafe. Speaking, writing, and prescriptivism
Gillian Sankoff. Substrate and universals in the Tok Pisin verb phrase
Talmy Givon. The pragmatics of referentiality
Jerrold M. Sadock. Whither radical pragmatics?
Richard Hudson. A psychologically and socially plausible theory of lan-
guage structure
Richard Bauman. The making and breaking of context in West Texas oral
anecdotes
Thomas A. Sebeok. Enter textuality: Echoes from the extraterrestrial
Michael Silverstein. On the pragmatic 'poetry' of prose
Thomas Kochman. The politics of politeness
Don H. Zimmerman. Talk and its occasion: The case of calling the police
Alan Davies. Idealization in sociolinguistics
Ellen F. Prince. Language and the law: Reference, stress, and context
Howard Giles and Mary Anne Fit/patrick. Personal, group, and couple
identities
John J. dumper/. Communicative competence revisited
Susan Gal. Phonological style in bilingualism
George A. Miller. Some comments on the subjective lexicon
Marilyn Shatz. A song without music and other stories
Elinor Ochs. Clarification and culture
Georgetown University Press ISBN: 0-87840-119-9

Meaning, Form, and Use in Context: Linguistic Applications: Deborah Schiffrin

Uploaded by

Copyright:

Available Formats

Meaning, Form, and Use in Context: Linguistic Applications: Deborah Schiffrin

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

Meaning, Form, and Use in Context: Linguistic Applications: Deborah Schiffrin

Uploaded by

Copyright:

Available Formats

What is the title of the book discussed in the text?

What is the title of the book discussed in the text?

What is the new title format for the series discussed?

What is the new title format for the series discussed?

Meaning, Form, and Use

Georgetown University Press, Washington, D.C. 20057

Copyright (§) 1984 by Georgetown University Press

Form and Function

Meaning and Form

Contexts: Institutional and Interpersonal

productive session. Before concluding, however, I am pleased to note

The theme of this year's Georgetown University Round Table on

models, and constructs: linguistics, anthropology, sociology, psychology,

Finally, I want to thank Susan M. Hoyle for her tremendously able

(6) John's house is a mess. The roof needs mending.

(8) Every accompanist needs the singer to be quiet in the opening

(11) Either my friend hasn't bought a Jaguar, or the car'U be in the

(E) Similar to this is the property of uniqueness. This problem is

1. Utterance interpretation is an inferential, largely deductive exer-

utterance, in (13) by some previous representation in the same

(A) The phenomenon of bridging cross-reference

Yet we also preclude examples such as (16).

(16) *A woman grinned at every man^. Hej smiled back.

This is a paradigm case where we have to manipulate information about

(18) His} mother likes John^.

terms of anaphoric dependency. So within the theory, the relation

(20) By hisj bed each child i kept a teddy bear.

(22) Each child^ kept a teddy bear by his: bed.

which permits its interpretation to be subsequent to that of the quanti-

Chafe, Wallace L. 1976. Givenness, contrastiveness, definiteness,

generating Grice's conversational maxims and the schema for pragmatic

2. The maxims of conversation. In his ground-breaking work on

Quality: Try to make your contribution one that is true.

builds in Quantity j in defining Quantity2. In section 3, I have

3. Minding our Qs and Rs.

(la) The Q Principle (Hearer-based):

1-sided: 'at least 3'

This analysis, foreshadowed a century ago by DeMorgan (1847:100) and

*. Q-based vs. R-based inference: Some early skirmishes. Like the

(3a) It is possible that John solved the problem *

An application of quantity in the latter case would generate the opposite

(4a) X is meeting a woman this evening >

5. Politeness rules, pragmatic competence, and maxim clashes. A

Crucially for us, the Deference or Aloofness maxim, Rule 1, is an upper-

participants in a talk-exchange tend not to draw what we have been

[Interlocutors] regularly provide less information than is required by

In fact, however, as Keenan makes clear in her discussion, Quantity

6. Quantity vs. informativeness. The most detailed and careful

(7a) some*—> not all

But, Atlas and Levinson point out, there is a substantial class of

(8) The inference from Informativeness: What is communicated

(8a) if p then q—>if -p then -q (Geis and Zwicky's 1971 'invited

(8b) The Principle of Informativeness: 'Read as much into an

Atlas and Levinson (1981:42) formulate their informativeness-based

The key notion here is the restriction of a more general predicate to a

(8d) The secretary smiled. —>The female secretary smiled.

7. Negation as implicatum-canceller: Q-based vs. R-based infer-

(10a) He didn't eat 3 carrots—he ate k of them.

(lla) She wasn't able to solve the problem.