Redefining-Indefinites Compress

Redefining Indefinites
Studies in Natural Language and Linguistic Theory
VOLUME 85
Managing Editors
Marcel den Dikken, City University of New York

Liliane Haegeman, University of Ghent, Belgium
Joan Maling, Brandeis University
Editorial Board
Guglielmo Cinque, University of Venice

Carol Georgopoulos, University of Utah
Jane Grimshaw, Rutgers University
Michael Kenstowicz, Massachusetts Institute of Technology
Hilda Koopman, University of California, Los Angeles
Howard Lasnik, University of Maryland
Alec Marantz, Massachusetts Institute of Technology
John J. McCarthy, University of Massachusetts, Amherst
Ian Roberts, University of Cambridge
For further volumes:

http://www.springer.com/series/6559
Carmen Dobrovie-Sorin • Claire Beyssade
Redefining Indefinites
Carmen Dobrovie-Sorin Claire Beyssade
Laboratoire de Linguistique Formelle Jean Nicod Institute
University of Paris 7 UMR 8129
UMR 7110-CNRS CNRS, ENS, EHESS
Paris Paris
France France
ISSN 0924-4670
ISBN 978-94-007-3001-4 ISBN 978-94-007-3002-1 (eBook)
DOI 10.1007/978-94-007-3002-1
Springer Dordrecht Heidelberg New York London
Library of Congress Control Number: 2012934950
© Springer Science+Business Media B.V. 2012

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and
executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this
publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s
location, in its current version, and permission for use must always be obtained from Springer. Permissions
for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to
prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Foreword
Redefining indefinites is a new, updated and thoroughly revised version of Définir

les indéfinis, published by CNRS Editions in 2004. The general structure of the
initial book, divided into eight chapters, has been preserved but each one of the
chapters has been revised for two main reasons.
On the one hand, a large body of new data and accounts has emerged in the years
that followed the publication of Définir les indéfinis. Since we aim at providing a
state-of-the-art treatment of each of the main issues discussed in the book, we were
led to incorporate these recent developments. On the other hand, our own account
has developed and we no longer defend the property-based analysis of indefinites
we assumed in 2004. Since then a large body of empirical and theoretical studies on
the distribution and interpretation of nominal expressions and bare nouns has been
published in a wide range of languages, with and without articles, which have drawn
our attention to previously unnoticed facts and differences between bare singulars
and bare plurals, which led us to revise our analysis. We now hold that weak
indefinites are not property-denoting elements but rather generalized quantifiers
over amounts.
This change of perspective and evolution of our line of thinking has led us to
revise all the chapters of the book, which have been affected to various degrees.
Chapter 2 has been entirely rewritten in order to incorporate two novel empirical
domains: (i) the comparison between argumental DPs and predicative DPs and
(ii) the contrast between bare plurals and bare singulars in Romance languages like
Spanish and Romanian. It is precisely these data that led us to abandon the property-
denoting analysis and move to an account in terms of generalized quantifiers over
amounts. Chapter 3, which in the 2004 version served to illustrate the property-
analysis of weak indefinites, as well as Chapters 4 and 5, which deal with the ambi-
guity of indefinites, were thoroughly revised, in line with our new proposal. The
analysis of generic plural indefinites proposed in Chapter 7 has been revised in order
to integrate an Individuation Constraint on Quantification. We have also revised the
Introduction and Chapter 1: in the interest of coherence, we found it necessary to put
forward our main hypothesis, according to which indefinites are existential generalized
v
vi Foreword
quantifiers over amounts. Finally, Chapters 6 and 8 have not undergone any major
conceptual modification.
We want to thank Anamaria Falaus for having translated those parts of Définir
les indéfinis that made their way into the present book and for having helped us with
formatting part of the manuscript and Susan Schweitzer for having reread the whole
of it. We have benefited from financial support from the Laboratoire de Linguistique
Formelle and from the ANR project Genius (ANR-08-JCJC-0069-01), directed by
Alda Mari.
Contents
1 Why Indefinites? ....................................................................................... 1

1.1 Typology of DPs ................................................................................ 2
1.1.1 Referential DPs ...................................................................... 3
1.1.2 Quantified DPs ....................................................................... 5
1.1.3 Indefinite DPs ......................................................................... 9
1.2 The Representation of Indefinite DPs ................................................ 11
1.2.1 Indefinites and Existential Quantification .............................. 11
1.2.2 Indefinites as Free Variables................................................... 12
1.2.3 Indefinites as Choice Functions ............................................. 13
1.2.4 Indefinites as Skolem Terms .................................................. 16
1.2.5 Indefinites and Properties ....................................................... 17
1.2.6 Indefinites as Existential Generalized
Quantifiers over Amounts ...................................................... 18
1.2.7 Conclusion.............................................................................. 18
1.3 Semantic Properties of Nominal Determiners ................................... 18
1.3.1 Conservativity ........................................................................ 19
1.3.2 Intersectivity ........................................................................... 21
1.3.3 Symmetry ............................................................................... 21
1.3.4 Proportional Determiners ....................................................... 22
1.3.5 Monotonicity .......................................................................... 23
1.3.6 The Semantic Characterization of Indefinites ........................ 24
1.4 The Interpretation of Indefinites ........................................................ 24
1.4.1 The Interpretation of Indefinites and Presupposition ............. 24
1.4.2 Distributive and Collective Readings ..................................... 26
1.4.3 Scope Ambiguities ................................................................. 27
1.4.4 Specific/Non-specific/Generic Readings................................ 28
1.5 Conclusion ......................................................................................... 29
vii
viii Contents
2 Bare Noun Phrases .................................................................................... 31

2.1 Bare Noun Phrases across Languages................................................ 31
2.1.1 An Overview of Crosslinguistic Variation ............................. 32
2.1.2 The Distribution of Bare NPs in Romanian,
Spanish and Catalan ............................................................... 34
2.1.3 The Syntactic Structure of Bare NPs ..................................... 36
2.2 Bare Plurals Are not the Plural Counterparts
of Singular Indefinites ........................................................................ 38
2.2.1 Opacity ................................................................................... 38
2.2.2 Scope ...................................................................................... 39
2.2.3 Aspect..................................................................................... 39
2.2.4 Anaphoric Relations ............................................................... 40
2.3 Count Bare Singulars Are not the Singular
Counterparts of Bare Plurals .............................................................. 41
2.3.1 Distribution ............................................................................ 42
2.3.2 Crosslinguistic Variation ........................................................ 45
2.3.3 Interpretation: Narrow Scope
with respect to Negation......................................................... 47
2.3.4 Conclusions ............................................................................ 49
2.4 The Semantics of Bare Plurals ........................................................... 50
2.4.1 Bare Plurals and Reference to Kinds ..................................... 51
2.4.2 Bare Plurals and Property Denotation .................................... 52
2.4.3 Bare Plurals and VP-level Existential Closure ....................... 58
2.4.4 Bare Plurals as Amount-Referring Expressions..................... 60
2.5 Existential Predicates and Entity Predicates ...................................... 63
2.5.1 Individual-Level and Stage-Level Predicates ......................... 63
2.5.2 Space Localization ................................................................. 65
2.5.3 Some Apparent Problems....................................................... 67
2.6 French Indefinites Headed by du/de la/des ........................................ 69
2.6.1 Bare Plurals and Bare Mass NPs............................................ 69
2.6.2 Parallelisms between du/de la/des French Indefinites
and Bare NPs in the Other Romance Languages ................... 70
2.6.3 On the Strong Reading of des Indefinites .............................. 71
2.6.4 Mass Nouns and the Impossibility of Individuation .............. 73
2.7 Bare NPs in Predicate Positions......................................................... 74
2.7.1 A Subclass of Nouns .............................................................. 75
2.7.2 Distributional Differences between
Singular Indefinites and Bare Singulars ................................. 77
2.7.3 Higgins’ Typology Revisited ................................................. 78
2.7.4 Semantic Composition ........................................................... 79
2.7.5 Copular Sentences Built with Singular
Indefinites as Equatives .......................................................... 81
2.7.6 Explaining the Contrasts between Bare
Singulars and Indefinite Singulars ......................................... 83
Contents ix
2.7.7 Modified Bare Nouns ............................................................. 89

2.7.8 The Argument Structure of Relational Nouns ....................... 91
2.7.9 Comparison with Other Approaches ...................................... 92
2.8 Conclusions ........................................................................................ 93
3 Existential Sentences................................................................................. 95
3.1 Constraints on Existential Sentences ................................................. 96
3.1.1 Existential Sentences Have
Property-Denoting Arguments ............................................... 96
3.1.2 Problems with the Property Analysis ..................................... 99
3.1.3 Our Proposal in a Nutshell ..................................................... 103
3.2 Existential Sentences in French ......................................................... 103
3.2.1 The Locative Existential Construction ................................... 104
3.2.2 The Eventive Construction ..................................................... 106
3.2.3 The Enumerative Construction............................................... 110
3.2.4 Conclusions ............................................................................ 111
3.3 Existential Sentences Cannot Have
Individual Variables as Arguments .................................................... 111
3.3.1 Heim’s (1987) Constraint ....................................................... 111
3.3.2 Quantified DPs in Existential Sentences ................................ 112
3.3.3 Existential Sentences Inside Relative Clauses ....................... 117
3.4 Conclusion ......................................................................................... 125
4 The Ambiguity of Indefinites: Towards a Denotational
Definition of the Weak/Strong Distinction .............................................. 127
4.1 Weak and Strong Indefinites .............................................................. 127
4.2 Weak Indefinites................................................................................. 129
4.2.1 Weak Indefinites as Individual Variables
Bound by Existential Closure................................................. 129
4.2.2 Weak Indefinites as Property-Denoting Expressions ............. 131
4.2.3 Weak Indefinites as Amount-Referring Expressions ............. 133
4.3 Strong Indefinites ............................................................................... 134
4.3.1 Quantificational Strong Indefinites ........................................ 134
4.3.2 Non-quantificational Strong Indefinites ................................. 136
4.3.3 The Two Strong Readings of Indefinites
and the Denotation of DPs ..................................................... 139
4.4 The Weak/Strong Distinction and Presuppositionality ...................... 139
4.5 Conclusions ........................................................................................ 141
5 Disambiguating Indefinites ...................................................................... 143
5.1 Disambiguating Indefinites: DP-Internal Factors .............................. 143
5.1.1 Lexical Specification of Plural Indefinites:
Partitivity, Contrastivity and Distributivity ............................ 143
5.1.2 Partitive Indefinite DPs .......................................................... 150
5.1.3 Modified Cardinals ................................................................. 151
5.1.4 Prepositional Accusatives:
Denotation Type and Specificity ............................................ 155
x Contents
5.2 Information Structure and the Disambiguation of Indefinites ........... 161

5.2.1 Indefinites in the Topic Position Are Presuppositional .......... 161
5.2.2 Non-topical Presuppositional Indefinites ............................... 163
5.2.3 Only Indefinites in Topic Positions Are Quantificational ...... 164
5.2.4 Indefinites at the Left Periphery ............................................. 165
5.3 Conclusions ........................................................................................ 169
6 The Scope of Indefinites............................................................................ 171
6.1 Scope: Current Analyses .................................................................... 172
6.1.1 Scope and Quantifier Raising ................................................. 172
6.1.2 Scope Ambiguity or Ambiguous Indefinites? ........................ 173
6.1.3 Intermediate Scope or Referential Dependency? ................... 174
6.2 Scope and Type of Denotation ........................................................... 177
6.2.1 On the Obligatory Narrow Scope of Weak Indefinites........... 177
6.2.2 Inverse Scope and Individual-Type Denotation ..................... 179
6.3 The Distributivity of Indefinites:
Quantification or Distributive Predication? ....................................... 181
6.3.1 The Distributivity of Inverse
Scope Specific Indefinites ...................................................... 181
6.3.2 The Quantificational Status
of Indefinites in Subject Position ........................................... 183
6.4 Referential Dependencies and Skolem Functions.............................. 184
6.5 Conclusions ........................................................................................ 189
7 Genericity, (In)Definiteness and Bare Nouns ......................................... 191
7.1 Generic Indefinites: Quantification over
Events and over Individuals ............................................................... 192
7.1.1 Quantification over Events and Indirect
Binding of Indefinites............................................................. 192
7.1.2 Adverbial Quantification over Individuals ............................. 193
7.1.3 Syntax-Semantics Mapping Rules ......................................... 194
7.1.4 Two Types of Generic Readings for Indefinites ..................... 195
7.2 Characterizing Sentences with Habitual Predicates ........................... 196
7.2.1 Adverbial Quantification over Events .................................... 197
7.2.2 Adverbial Quantification over Individuals ............................. 198
7.2.3 Proposal: Quantification over Individuals
Combined with Quantification over Times ............................ 198
7.2.4 Conclusions ............................................................................ 200
7.3 The Genericity of Singular Indefinites ............................................... 200
7.3.1 GEN and the Nomicity Constraint ......................................... 201
7.3.2 Pseudo-Generic Indefinites in Object Positions ..................... 203
7.4 The Genericity of French des-Indefinites .......................................... 205
7.4.1 The Individuation Constraint on Quantification..................... 205
7.4.2 The Generic Readings of Plural
Indefinites Built with Symmetric Nouns ................................ 209
Contents xi
7.4.3 The Pseudo-Generic Reading of Plural Indefinites ................ 212

7.4.4 Conclusions ............................................................................ 216
7.5 The Genericity of English Bare Nouns .............................................. 216
7.5.1 The Genericity of French Plural and Mass Indefinites
and the Genericity of English Bare Nouns............................. 216
7.5.2 Adverbs of Quantification and Kind-Predication ................... 218
7.6 Conclusions ........................................................................................ 220
8 Dependent Indefinites in Donkey Sentences ........................................... 221
8.1 Dependency and Donkey Sentences .................................................. 221
8.1.1 Indefinite DPs and Universal Quantification .......................... 221
8.1.2 Indefinite DPs, Free Variables
and Unselective Binding ........................................................ 223
8.1.3 Symmetric and Asymmetric
Readings of Donkey Sentences .............................................. 225
8.2 Dependent Indefinites ........................................................................ 227
8.2.1 Dependency on a Situation..................................................... 228
8.2.2 Dependency on a Quantified DP ............................................ 230
8.3 Dependency and Proportion ............................................................... 231
8.3.1 Symmetric and Asymmetric Readings ................................... 231
8.3.2 Weak and Strong Asymmetric Readings................................ 232
8.4 Dependency and Reference................................................................ 234
8.5 Conclusion ......................................................................................... 235
8.6 Appendix ............................................................................................ 235
8.6.1 Egli’s (1979) Solution ............................................................ 236
8.6.2 E-type Analyses ..................................................................... 238
Conclusion ....................................................................................................... 243
Bibliography .................................................................................................... 249
Author Index.................................................................................................... 261
Subject Index ................................................................................................... 265

Introduction
There are two types of studies that can be found in the rich literature on the syntax
and semantics of nominal phrases (DPs). On the one hand, we have a large number
of empirical studies, which identify, describe and analyze in detail language specific
facts, such as the expression of quantity, bare nouns or DPs headed by de in French.
On the other hand, we have more theoretical studies that seek to isolate the formal
properties characterizing the syntactico-semantic behavior of DPs (such as the
availability of generic readings, or the licensing of discourse anaphora). The theo-
retical work deserves credit for bringing methodological rigor into an empirical area
that is particularly hard to grasp in view of the diversity of attested interpretations.
However, such studies have the problem of narrowing down the area of investiga-
tion to a fragment of a given language, leaving for future research many other
attested data, which although well described, still resist formalization.
Our investigation of indefinites attempts to bring together empirical and theoretical
approaches: we develop a formal analysis of DPs in context, on the basis of lan-
guage facts. The examples we provide come mainly from English and Romance
(especially French and Romanian, although we also include data from Spanish and
Italian). We have systematically confronted our proposal to other existing theories.
We have attempted to observe the principle of compositionality and to reduce as
much as possible the postulation of ambiguity.
For each one of the phenomena under consideration, we offer an overview of the
recent literature and lay out our own hypotheses and the remaining open questions.
Our analysis is formal insofar as it relies on formal representations but it is not
technical. We have made use of the core assumptions underlying the Discourse
Representation Theory (indefinites are represented as variables and their
quantificational force is due to the operators in their context of occurrence), without
however getting into the details of an implementation, in a specific framework. We
favored conceptual clarity, argumentation and empirical generalizations over the
technicalities of formal details. The data and the analyses pursued here can be for-
malized in different theoretical frameworks.
The typology of nominal phrases is one of our main interests in this book.
The research carried out in the 1980s, in the semantics of discourse (Kamp 1981;
xiii
xiv Introduction
Heim 1982) and in the theory of generalized quantifiers (Barwise and Cooper 1981;
Keenan and Stavi 1986; Westersthal 1989), brought new insights into the study of
DPs and made it clear that the traditional split between definite and indefinite DPs,
as well as the logicians’ distinction between quantified and referential expressions,
could not capture the complexity of the empirical facts. We aim to provide an appro-
priate characterization of the class of so-called indefinite DPs, which are to be
distinguished from both referential and quantificational expressions.
The issues we explore and the questions we raise are extremely diverse and are
situated at the interface of morphology, syntax, semantics and pragmatics. For
example, we wonder whether there are syntactic positions that force or preclude the
occurrence of certain DPs and whether these restrictions are related to any specific
semantic property. In other words, we want to find out whether there are any correla-
tions between syntactic position, semantic type and interpretive effects of DPs. The
most influential line of thinking in the literature, in the spirit of Montague, posits the
existence of a homomorphism between syntax and semantics: argumental syntactic
positions can only be occupied by elements which have an argument-type of denota-
tion, i.e., either entity (type e) or quantificational type (<<e,t>,t>). This principle
automatically rules out property-denoting elements from argument positions.
Higginbotham (1985, 1987) challenged this widely assumed homomorphism and
proposed to analyze certain indefinite DPs as property-denoting elements (of type <e,t>).
This hypothesis was taken up in the 1990s, in particular by McNally (1995), van
Geenhoven (1996) or Dobrovie-Sorin (1996), who exploit the hypothesis that indefinite
DPs denote properties in order to shed new light on existential ‘there’ sentences or on
existential bare NPs. Van Geenhoven’s work on West Greenlandic brought in novel
empirical facts, which were shown to be typologically related to data in a wide variety of
languages, with and without articles. These highly significant data further supported
the reconsideration of the syntax/semantics homomorphism and led to an increasing
number of studies relying on what came to be labeled ‘semantic incorporation’, i.e.,
the semantic composition of property-denoting arguments.
One of the recurrent questions in our study concerns the existence of property-
denoting DPs in non-predicative positions. We will provide a negative answer and
propose instead that weak indefinites in argument positions need to be analyzed as
existential generalized quantifiers over amounts.
A different set of questions concerns the distinction between so-called strong and
weak determiners and DPs. Is this a semantic or a pragmatic distinction? Does
it overlap with the difference between existential and partitive readings of
indefinite DPs?
A third set of questions relates to scope phenomena. Although there are cases
where the co-occurrence of two DPs gives rise to ambiguous sentences, with differ-
ent readings arising from different scope configurations, this does not always hap-
pen. We seek to predict instances of ambiguity on the basis of the semantic type of
DPs, as well as the distributivity effects induced by certain determiners but not by
others.
Most of the answers we provide rely on the distinction between two types of enti-
ties that we take to be central for the analysis of DP denotations: individualized
Introduction xv
objects, i.e., entities that are distinct from each other, and amounts, i.e., non-individuated
entities, which share the property of verifying a certain measure. This measure can
be precise or vague, e.g., 300 g of butter and three books refer to amounts consisting
in a determined quantity (300 g and three) of butter and of books, respectively,
whereas butter and books refer to any amount/quantity, i.e., an indeterminate quan-
tity of butter and books, respectively. Assuming the distinction between individual-
ized objects and amounts, we develop a new account of weak indefinites, as
existential generalized quantifiers over amounts. We argue that this hypothesis
solves a large number of problems faced by other analyses and offers interesting
novel research perspectives.
The book is divided into eight chapters. We first introduce the reasons why we
need to postulate the existence of a third type of DP, in addition to referential and
quantificational ones and provide a summary of the main results of the theory of
generalized quantifiers (Chap. 1).
The next two chapters study in detail the denotation of bare NPs (Chap. 2) and
the properties of existential constructions (Chap. 3). We point out the problems
of analyses of indefinites as property-denoting elements and instead defend an
account that relies on the distinction between two types of entities, individuals and
amounts and treats weak indefinites as existential generalized quantifiers over
amounts, which combine with existential predicates identifying an amount with an
individual. In Chaps. 4 and 5, we focus on the distinction between strong and weak
DPs, due to Milsark (1977). On the basis of a detailed analysis of indefinite DPs in
context, we show that their interpretation is not a mere instance of underdetermina-
tion but rather a genuine ambiguity, each reading being assigned a distinct type of
representation at the level of Logical Form.
Chapter 6 deals with the scope of indefinites. If we are right in claiming that
indefinite DPs are non-quantificational, the issue of scope should not arise. We show
how the notion of dependency captures the relevant facts and how dependency rela-
tions are constrained by the denotation of DPs.
Chapter 7 is devoted to generic indefinite DPs. Moving beyond the widely
accepted hypothesis that the generic readings of indefinites depend on the occur-
rence of quantificational adverbs (generally, always, often, rarely), we identify the
constraints that govern the genericity of singular and plural des-indefinite DPs in
French.
Finally, in Chap. 8, we return to the interpretation of dependent indefinites. In
sentences like Every farmer who owns a donkey beats it and If a farmer owns a
donkey, he beats it, the indefinite DP a donkey is not to be interpreted as a quantified
expression but rather as a referential expression, whose reference is dependent on
the quantifier introduced either by a quantified DP (every farmer) or by the condi-
tional operator if. We show that the representation of dependent indefinites as
Skolem terms can both capture the interpretive properties of these sentences
(such as the asymmetric readings, which were problematic in the framework of
the Discourse Representation Theory) and explain why a pronoun coindexed with
a dependent indefinite but not c-commanded by it can have a bound variable
interpretation.
Chapter 1
Why Indefinites?
In this introductory chapter we will be concerned with the reference, or denotation,

of nominal phrases, as well as with the relations between their syntactic properties
and their denotation – their semantic type. We wish to know (a) what semantic
types of nominal phrases are universally possible, (b) whether there are correlations
between syntactic positions and semantic types, (c) whether there are correlations
between semantic types and scopal interpretations.
The denotation of nominal phrases depends on the denotation of their parts,
namely on lexical items such as nouns, adjectives and on items that are more
grammaticalized in nature, such as number inflections, cardinal and ordinal
numbers and articles of various sorts, e.g., definite, indefinite or demonstrative.
Because they seem to be in complementary distribution with each other, it is
currently assumed that the various types of articles occupy the same designated
position inside the nominal projection, dubbed Det(erminer). It has also been
observed that Determiners are crucial for both the syntax and the semantic com-
position of nominal projections and therefore nominal phrases are now viewed
as Determiner Phrases (DPs), i.e., as maximal projections of the Determiner
itself, rather than as NPs (Nominal Phrases), a label that is now used to refer to
a lower syntactic constituent, which comprises the noun together with its
modifiers, e.g., adjectives.
The logical tradition, which distinguishes between referential DPs (proper
names, definite expressions) and quantified1 DPs (every man, each man), groups
indefinite DPs together with quantified DPs, since they can be said to involve
existential quantification. Relatively recent work (Heim 1982; Kamp 1981), which
developed rapidly in the framework of the Discourse Representation Theory (DRT),
has established that indefinites do not behave like quantified DPs. The need to
distinguish between indefinites and quantified DPs arises when we address questions
1
We will not differentiate between the terms quantified and quantificational.
C. Dobrovie-Sorin and C. Beyssade, Redefining Indefinites, Studies in Natural 1

Language and Linguistic Theory 85, DOI 10.1007/978-94-007-3002-1_1,
2 1 Why Indefinites?
(b) and (c) above. There are indeed syntactic positions that are accessible to
indefinites but not to quantified DPs:
(1) a. John is a nice man.
b. *John is each man I’ve loved.
c. *John is every man I’ve loved.
The above contrast shows that indefinites can function as nominal predicates2
whereas quantified DPs cannot do so. Moreover, there are also argument positions
that allow only indefinites (see (2a), (2b) and (2c)), to the exclusion of quantified
DPs (see (2d), (2e)):
(2) a. There was (a/one) book on the table.
b. There were (three/sm)3 books on the table.
c. There arrived (a/three/sm) student(s) last night.
d. *There was each book on the table.
e. *There arrived each student last night.
It has also been observed that indefinites and quantified DPs do not generate the
same scopal interpretations:
(3) a. Someone believes that Mary read a (certain) book.
b. Someone believes that Mary read every book.
In (3a), there are two interpretations, (3a¢) and (3a″), whereas in (3b), every book
cannot have wide scope; (3b″) is not a paraphrase of (3b).
(3) a¢. Someone believes that Mary read a book.
a″. There is a book that someone believes Mary read.
b¢. Someone believes that Mary read every book.
b″. For every book, there’s someone who believes that Mary read it.
1.1 Typology of DPs
Since indefinites can be assimilated neither to referential DPs nor to quantificational

DPs, we are led to assume that there are three distinct semantic types of DPs: refer-
ential, quantified and indefinite. We must now propose definitions and formal
representations for each of the three classes and establish the criteria that will allow
us to place a particular DP in one or another class.
2
It is currently assumed that DPs associated with the copular verb be function as predicates. This
issue will be examined in detail in Chap. 2.
3
sm is the weak (i.e., unaccented) form of some.
1.1 Typology of DPs 3
1.1.1 Referential DPs
Proper names and definite expressions constitute the simplest type of DP with
respect to denotation: they refer to an individual, an element of type e, where e is an
abbreviation for entity. Demonstrative DPs may also be analyzed as denoting
individuals, even though, in this case, the reference is not “rigid” – that is to say,
stable across different contexts (possible worlds or situations) – but rather variable
according to the utterance context.
In examples (4), John denotes the individual who carries this name and the
predicates respectively denote the properties of singing, being a doctor or being
handsome, or extensionally, the sets of individuals who are characterized by these
properties. The truth-value of the sentence is calculated by checking whether the
individual denoted by John is a member of the set of individuals that are singing in
(a), that are doctors in (b), or that are handsome in (c). The sentence is true or false
depending on whether or not membership is confirmed.
(4) a. John is singing.
b. John is a doctor.
c. John is handsome.
Formally then, a sentence such as (4c) is represented as in (4¢c), where j is an

individual constant referring to John and Handsome is a one-place predicate that
denotes the set of all that is handsome.
(4¢) c. Handsome (j)
According to (4¢c), the denotation of John is handsome, which is a syntactically

complex constituent, is calculated from the denotations of its parts, handsome and
John, by applying the rule of functional application4: handsome is a function that
takes John as an argument and yields the truth value ‘true’ if the denotation of John
belongs to the denotation of handsome and ‘false’ if the denotation of John does not
belong to the denotation of handsome.
Following the current use in formal semantics, we assume two primitive
denotation types, individuals (type e) and truth-values (type t), which respectively
correspond to referential DPs and propositions. All the other syntactic constituents
denote functions. Leaving aside intensionality, a one-place predicate such as hand-
some denotes a function of type <e,t>, i.e., a function that applies to the domain of
individuals (De) and returns values in the domain of truth-values (Dt = {0,1}).5 In
set-theoretical terms, a function of type <e,t> denotes a set of individuals.
4
Most semantic models assume at least one additional rule, predicate modification. But it is
primarily for reasons of simplicity of exposition that this choice is made, since technically, one can
limit oneself to functional application (cf. Heim and Kratzer 1998: 61ff).
5
The domain of truth values, notated Dt, is composed of two elements, true and false, notated 1 and
0 respectively.
The same type of analysis is often assumed for definite descriptions, e.g., (5) can
be represented as in (5¢):
(5) The student is crying.
(5¢) Cry (the student)
It remains to be shown how we can get from a noun phrase with internal
structure (determiner + noun) to a simple type, e, while obeying the principle of
compositionality. According to Russell’s analysis of definite descriptions, the noun
N denotes the set of individuals that have the property of being N and the definite
article is represented as the iota operator, whose function is to extract from this set
the only element that it contains. The iota operator is thus a function of type
<<e,t>,e>, i.e., a function that applies to a property or a set of individuals
(of type<e,t>) – such as student in (6) – and returns an individual – here, the
student in question, of type e.
(6) the student
(6¢) ix student (x)
Proper names as well as definite or demonstrative DPs display remarkable

semantic properties. Semantics, which deals with the truth-values of sentences, or,
to be more precise, with their truth-conditions, can predict the type of logical
relations, e.g., equivalence, entailment or contradiction, that hold between proposi-
tions. Certain laws of logic apply to natural language sentences. We mention two
of these here:
– the law of contradiction, according to which “P and not P” is a contradictory
proposition, i.e., a proposition that is always false (cf. (7a));
– the law of the excluded middle, according to which “P or not P” is a tautological
proposition, i.e., a proposition that is always true (cf. (7b)).
(7) a. The neighbor’s son is more than 30 years old and the neighbor’s son is
not more than 30 years old.
b. Either the neighbor’s son is more than 30 years old or the neighbor’s
son is not more than 30 years old.
It should be observed that these laws are verified for sentences built with
referential DPs (e.g., proper names, definite or demonstrative DPs) but not with
indefinite or quantified DPs in subject positions. Indeed, contrary to (7a), (7c) is not
a contradictory sentence. As for (7d), it is not tautological and thus illustrates the
case of a sentence that does not verify the law of the excluded middle.
(7) c. Two children came and two children did not come.
d. Each student is present or each student isn’t present.
We conclude that DPs such as two children or each student must denote some-
thing other than an individual, something other than an element of type e.
1.1.2 Quantified DPs
First-order logic makes use of two quantifiers, " and $, which allow us to formalize
expressions such as every, each, a and even cardinals6 as well as modified cardinals
such as two, exactly two, at least two…:
(8) a. All men are mortal.
a¢. "x (H(x) → M(x))
b. A man came.
b¢. $x (H(x) Ù C(x))
c. At least two men came.
c¢. $x $y (H(x) Ù H(y) Ù x ¹ y Ù C(x) Ù C(y))
In (8), the natural language sentences a, b and c are paired with their represen-
tations in predicate calculus, a¢, b¢ and c¢. The Ù sign marks conjunction and
the → sign marks entailment. H(x) signifies ‘x is human’, M(x) ‘x is mortal’ and
C(x) ‘x came’.
The two quantifiers of predicate calculus do not allow us to account for all the
quantifiers found in natural languages. Indeed, proportional expressions such as
most or more than half, which cannot be reduced to Boolean combinations of
existentials and universals, cannot be represented in first-order logic. The study of
natural language quantifiers has led to the development of representations that are
designed for natural language quantifiers. In what follows we will present the tripar-
tite structure representations inspired by the work of Lewis (1975) and the theory of
generalized quantifiers, stemming from the work of Montague.
1.1.2.1 Tripartite Structures
Tripartite structures were introduced by Lewis (1975), who observed that adverbs of
quantification (sometimes, often, always, never, etc.) can quantify not only over
times and events but also over individuals. Thus, sentence (9a) has the same inter-
pretation as (9b), which may be represented as in (9¢), where never and no are trans-
lated in the same way as NO:
(9) a. A student is never stupid.
b. No student is stupid.
(9¢) NO x (x a student) [x is stupid]
We are interested here not in the interpretative flexibility of adverbs (known as

“unselective binding”), nor in the algorithms that allow the derivation of (9¢) from
6
The formal representation of cardinals (see (8c¢)) requires first-order logic with identity.
(9a) but rather in the tripartite representation proposed by Lewis, which was extended
to quantified DPs (cf. Heim 1982). Tripartite representations make clear an impor-
tant property of natural language quantifiers: independently of its syntactic status
(adverb or determiner), a quantifier has scope over the minimal sentence to which it
belongs. The quantifier thus appears, in Logical Form, in the initial position of the
tripartite structure.7 The second part of the structure (often called the ‘restriction’)
indicates the domain of quantification, which depends on the syntactic context in
the case of adverbs but is fixed for determiners: the nominal constituent that is the
sister of the determiner appears in the restriction.8 The third part of the tripartite
structure (often called the ‘(nuclear) scope’) is filled with the predicate obtained by
abstracting over the position of the quantified DP.
Tripartite representations can be obtained by assuming two rules that map
syntactic configurations onto LF representations: (i) the quantified DP is adjoined to
the sentential node that immediately dominates it and its original position is filled
by an empty category interpreted as a variable9; (ii) the determiner of the moved
quantified DP raises out of the DP and adjoins to the sentential node. By applying
these two rules the examples in (10) can be represented as shown in (10¢):
(10) a. Each professor came.
b. Some students are sad.
(10¢) Quantifier Restriction Scope
a. each (professor (x)) [x came]
b. some (student (x)) [x is sad]
(10¢a) reads as follows: ‘for each individual x, if x is a professor, then x came.
1.1.2.2 Generalized Quantifiers
Determiners can also be analyzed as denoting the relation between the two sets
corresponding to the restriction (the NP that is the sister of Det) and the scope
(the predicate obtained by abstracting over the position of the quantified DP). The
determiner each, for example, denotes the relation of inclusion between two sets:
(10a) is true if and only if the set of professors is a subset of the set of individuals
who came. Some denotes the relation of intersection: (10b) is true if and only if the
intersection of the set of students and the set of sad individuals is not empty.
7
If the same sentence contains more than one quantified DP, their relative scope may vary
(see Chap. 6).
8
This mapping may be affected by contrastive stress.
9
The need for such a rule is generally assumed in models of the syntax-semantics interface: see
Montague’s rule of “Quantifying in” or May’s (1985) “Quantifier Raising”. However, it is not
clear whether this rule obeys general constraints on syntactic movement, as claimed by May
(1985). Cf. Chap. 6.
This analysis of determiners, called the relational analysis, involves a relaxation

of semantic composition, since it assumes operations of semantic composition other
than functional application. However, as pointed out by Heim and Kratzer (1998:
149–150), the relational analysis of determiners is translatable into a functional
analysis. The relational analysis can thus be used as a convenient abbreviation of the
functional analysis (the details of which will be given below), which relies solely on
functional application.
The work of Montague has shown that the semantic composition of quantificational
DPs may rely on functional application provided that we invert its orientation, that
is, we need to invert the relation between subject and predicate. Thus, in sentences
with referential subjects, the predicate denotes a function that takes as its argument
the denotation of the subject DP (cf. (11a)). In sentences with quantificational sub-
jects (see (11b)), it is the subject DP that denotes a function, which takes as its argu-
ment the denotation of the main predicate (type <e,t>) and returns a truth value
(type t): given this orientation of functional application, the predicate remains of
type <e,t> but the subject DP is analyzed as an expression of type <<e,t>,t>. In (11),
• notates functional composition.
(11) a. The functional analysis of referential DPs, e.g., definite expressions or
proper names:
a predicate of type <e,t> applies to an individual of type e
<e,t> • e returns t by functional application
b. The functional analysis of quantificational DPs:
the DP takes as its argument the predicate:
an element of type <<e,t>,t> applies to a predicate of type <e,t>
<<e,t>,t> • <e,t> returns t by functional application
Quantificational DPs thus denote functions of type <<e,t>,t>. Put in set-theoretical

terms, quantificational DPs denote sets of properties or, equivalently, sets of sets of
individuals. Thus, the quantified DP all children in (12) denotes the set of properties
that all children have. Sentence (12) is true if and only if this set contains the prop-
erty of singing.
(12) a. All children sing.
b. "x (child(x)) [sing(x)]
Given that quantified DPs have a denotation of type <<e,t>,t> and that NP con-
stituents (N or N + Adj) denote a property (type <e,t>), determiners denote a func-
tion of type <<e,t>,<<e,t>,t>>. However, we can avoid dealing with relatively
complex functions of this type by resorting to the relational analysis of determiners
(according to which determiners denote relations between two sets of individuals,
one denoted by the noun and the other by the verb phrase), briefly presented at the
beginning of this section.
Barwise and Cooper (1981) have extended Montague’s analysis of quantifica-
tional DPs to all DPs: they use the formal theory of generalized quantifiers (a part of
mathematical logic due to Mostowski 1957) as a framework for the investigation on

quantification in natural language. According to the theory of generalized quantifiers,
all DPs can be treated as generalized quantifiers, i.e., as denoting sets of sets. Thus,
even referential DPs, e.g., proper names, can be analyzed as generalized quantifiers:
John, for example, denotes the set of sets containing John and can thus be inter-
preted as the set of properties that John has. There exist then at least two distinct
formal ways of representing proper names: as constants, or as sets of properties, that
is, as abstractions over properties:
(i) j
(ii) lP (P(j))
These two representations can be related to each other by Partee’s (1987) rules of
semantic type-shifting: a referential DP, which originally denotes an individual
(type e), can be analyzed as a generalized quantifier, therefore as an expression of
type <<e,t>,t> by applying a type raising rule.
A terminological point must be made here: it is usual to analyze DPs as made up
of a determiner followed by a noun. The class of determiners is defined distribution-
ally, grouping together those expressions situated to the left of the common noun
(leaving aside adjectives modifying the noun): definite and indefinite articles and
other indefinite determiners, demonstratives and possessives. Within the class of
determiners, there is a subclass of expressions called quantifiers, such as each, all
and most. This use of the term ‘quantifier’ differs from Barwise and Cooper’s (1981)
‘quantifier’, which is synonymous to ‘generalized quantifier’ and constitutes the
denotation of the DP as a whole. So, in the vocabulary of the generalized quantifier
theory, determiners and quantifiers are distinct, whereas in standard terminology,
quantifiers are a subclass of determiners.
The use of generalized quantifiers has the advantage of capturing the syntactic
parallelism found in natural languages between referential DPs and quantified DPs.
Thus, although the syntactic structures of the sentences in (13) are identical (DP
VP), their first-order logic formalizations in (13¢) differ from one another. Individual
constants and variables are notated with lower case letters and predicates are notated
with capitals (S and H respectively correspond to the predicates sleep and be
human).
(13) a. John sleeps.

b. A person sleeps.
c. Everyone sleeps.
(13¢) a. S(j)
b. $x (H(x) ∧ S(x))
c. ∀x (H(x) → S(x))
The structural parallelism between (13b) and (13c) is maintained in the formal-
izations (13″b–c), which use generalized quantifiers. In these formulae, the DP is
analyzed as a function that takes a property as its argument, whence the notation lP.
In (13″b–c), the DPs denote generalized quantifiers that are applied to the property
denoted by the predicate sleep.
(13″) b. a person (sleeps) lP ($x (H(x) ∧ P(x))) S

c. everyone (sleeps) lP (∀x (H(x) → P(x))) S
If we assume that John can also denote a generalized quantifier, the analysis of
(13a) becomes parallel to that of (13b–c):
(13″) a. John (sleeps) lP (P(j)) S
An important advantage of the generalized quantifier theory is that it allows us to

represent natural language quantifiers such as most, which cannot be formalized in
first-order logic.
A further advantage of the generalized quantifier theory is that of unifying the
treatment of DPs, while at the same time allowing us to distinguish between differ-
ent types of DPs, depending on the nature of their determiners (see Barwise and
Cooper 1981), who rely on the properties of generalized quantifiers in order to
redefine the distinction between definites and indefinites and Milsark’s (1977) dis-
tinction between weak and strong determiners.
1.1.3 Indefinite DPs
At the beginning of the 1980s, Kamp (1981) and Heim (1982) departed from the
logical tradition, according to which indefinite expressions are quantificational and
proposed instead that the quantificational force of indefinites is not intrinsic to these
expressions but rather is contributed by certain elements of their context, e.g.,
adverbs of quantification, analyzed as unselective quantifiers by Lewis (1975).
This analysis of indefinites was meant to solve the problems raised by the so-
called donkey sentences (see (14)), in which indefinites take universal readings and
therefore cannot be represented as existential quantifiers.10
(14) a. If Pedro owns a donkey, he beats it.
b. Every farmer who owns a donkey beats it.
(15) a. ∀x [(donkey(x) ∧ own(p,x)) → beat(p,x)]
b. ∀y∀x [(farmer(y) ∧ donkey(x) ∧ own(y,x)) → beat(y,x)]
In order to account for this type of example, one needs to choose between two
possible analyses of indefinites: (i) they are ambiguous, sometimes translating as
10
A detailed analysis of donkey sentences will be provided in Chap. 8.
Table 1.1
Type of DP Referential Quantificational Indefinite
Type of denotation e <<e,t>,t> ??
existential quantifiers and sometimes as universal quantifiers; (ii) they are not
quantificational. Although their implementations differ, both Kamp and Heim chose
the latter option. According to Heim’s LF implementation, indefinites supply a free
variable and a condition on the domain of the variable.11
Thus, an indefinite DP such as a donkey is analyzed as supplying an individual
variable x and a condition on that variable, ‘donkey(x)’, its quantificational force
being contributed by some element of the context. Thus, the universal force that can
be observed in (14b) comes from the quantified DP every farmer, which binds both
a donkey and the anaphoric pronoun it, which is coindexed with a donkey. In other
contexts, the quantificational force of indefinites is supplied by adverbs of
quantification. In those if-clauses that lack overt adverbs of quantification, a GEN
operator (viewed as a covert adverb of quantification and paraphrasable by gener-
ally) is supplied by default (Farkas and Sugioka 1983). Indeed, (14a) can be roughly
paraphrased as (16):
(16) Generally, if Pedro owns a donkey, he beats it.
Turning now to the existential readings of indefinites, they appear in those con-
texts in which the indefinite appears in a matrix clause (as in (17a)) or in the nuclear
scope of a tripartite quantification (but crucially not in the restriction) (see (17b)).
(17) a. A man entered the room.

b. When Pedro is alone, he often buys a cat.
According to the analysis proposed by Kamp and Heim, indefinite DPs are not
quantificational expressions and may be viewed instead as referential terms. They
nevertheless differ from other referential terms such as proper names, which refer
rigidly: whereas proper names can be treated as constants, indefinites supply
variables.
Summarizing, three classes of DPs can be distinguished from the point of view
of their denotation: referential DPs, which are entity-denoting expressions (type e),
quantificational DPs (type <<e,t>,t>) and indefinite DPs. The semantic type of
indefinites is an open question (Table 1.1).
The types of denotation indicated here are primitive: in other words, referential
DPs are ‘born’ as type e expressions and quantificational DPs are born as generalized
11
In Kamp’s discourse representation structures, an indefinite DP contributes a discourse referent
and a condition on that discourse referent. Although technically different, free variables and dis-
course referents are comparable at the stage of evaluation.
1.2 The Representation of Indefinite DPs 11
quantifiers. The primitive type e can be ‘lifted’ by type-shifting, yielding a generalized

quantifier, a type of denotation that can thus be assumed for all DPs, as explained in
Sect. 1.1.2.2 above.
1.2 The Representation of Indefinite DPs
In this section we will briefly present the various analyses of indefinites that can be
found in the literature. It should be observed right from the start that indefinite DPs
are ambiguous, allowing both ‘strong’ and ‘weak’ readings in the sense of Milsark
(1977), an issue to which Chaps. 4 and 5 are dedicated. Although most authors did
not mention it explicitly, it is the strong readings of indefinites that were covered by
the traditional analysis in terms of existential quantifiers, as well as by the more
recent analyses originating from Kamp (1981), Heim (1982, 1997a) and Winter
(1997) or Skolem terms (Steedman (2003, 2006)). These accounts cannot be
extended to cover weak indefinites, which came into focus when bare NPs (i.e.,
nominal constituents lacking a Determiner) started to be seriously investigated
(Diesing 1992; McNally 1995a; van Geenhoven 2006). Diesing’s analysis relying
on VP-level existential closure and Van Geenhoven’s analysis relying on property
denotation are the most influential accounts of weak indefinites, which we will be
led to discard in Chaps. 2, 3, 4 and 5. We will instead propose an account relying on
generalized quantifiers over amounts, presented in Sect. 1.2.6 below.
1.2.1 Indefinites and Existential Quantification
Predicate calculus makes use of two quantifiers: the existential quantifier notated $
and the universal quantifier ∀. The former is used to represent particular judgments
such as A man is ill (to be distinguished from singular judgments of the type Socrates
is mortal) and the latter is used to represent universal judgments such as Every man
is mortal.
It is currently assumed that in natural languages it is the determiners that express
quantity and therefore determiners are assumed to correspond to the logical quantifiers.
Thus the determiners every, each and all, which are used in universal judgments, are
represented as the universal quantifier, whereas the indefinite article translates as the
existential quantifier. The existential quantifier can be used to represent not only sin-
gular indefinites but also cardinal indefinites such as two men or three men:
(18) a. A man came.

b. Two men came.
(18¢) a. $x (M(x) ∧ C(x))
b. $x $y (¬(x = y) ∧ M(x) ∧ C(x) ∧ M(y) ∧ C(y))
The logical formula ‘$x P(x)’ can be paraphrased as ‘there is an x that verifies
the property P’. As shown in (18¢b), for the representation of cardinal indefinites,
we need to use predicate calculus with identity, which allows us to distinguish and
therefore to count the elements of the domain.
As explained in Sect. 1.1.2.2. above, the generalized quantifier theory treats all DPs
as denoting generalized quantifiers, i.e., sets of sets of individuals (type <<e,t>, t>).
Within this framework, expressions of the type each N, every N and all N are repre-
sented as in (19a) and DPs headed by the indefinite article as in (19b):
(19) a. lQ ∀x (N(x) → Q(x))

b. lQ $x (N(x) ∧ Q(x))
Indefinite DPs are thus generalized quantifiers that differ from other generalized
quantifiers insofar as their representation contains an existential quantifier.
1.2.2 Indefinites as Free Variables
As explained in Sect. 1.1.3 above, according to DRT, singular indefinite DPs supply
free variables that range over individuals. A DP such as a student in (20a) is associ-
ated with the discourse representation structure (20b). The nominal element
(student in (20a)) functions as a predicate that restricts the range of the variable: in
(20b), the range of x is the set of students.
(20) a. a student
b.
x
Student (x)
In intuitive terms, (20a) and (20b) say that a DP such as a student refers to any
individual having the property of being a student. In other words, a random indi-
vidual is extracted from the set of students.
According to Heim (1982), indefinite DPs contain no element corresponding
to the existential quantifier – in particular, the existential quantifier does not
correspond to the indefinite article and is therefore absent from the representa-
tions in (20) but is introduced by rules of “existential closure” (see (22) below
and Chap. 2).
It is possible to extend Heim’s (1982) analysis from singular indefinites to plural
indefinites, by allowing variables to range not only over atomic individuals but also
over plural individuals (Link 1983). The two types of variables are represented by
lower case and capital letters, respectively.
(21) a. two students

a¢. X ∧ Students (X) ∧ Two (X)
b. sm students
b¢. X ∧ Students (X)
According to this analysis, a plural indefinite refers to a plural individual (made

up of several atomic individuals), whose cardinality may or may not be specified.
Cardinals (one, two, etc.) function as predicates indicating the cardinality of the
plurality: the determiner denotes the property of being one, or two, etc. The repre-
sentation in (21b¢) does not contain a cardinality predicate, because sm refers to an
indefinite quantity.
According to the original version of DRT, the variables supplied by indefinites
become bound by existential closure, by applying one of the two rules stated in
(22a–b) (see Heim (1982: 140–152)):
(22) a. Adjoin an existential quantifier to the nuclear scope of every quantifier.

b. Adjoin an existential quantifier to T(ext).
In later analyses inspired by DRT (Diesing 1992; Kratzer 1988, 1995), these two
rules were dispensed with and a different rule of existential closure was introduced,
which applies at the level of the VP. It is important to stress that (22)a–b on the one
hand and VP-level existential closure on the other hand were not designed for the
same type of indefinites: whereas (22)a-b were meant to respectively account for
specific indefinites and dependent indefinites, VP-level existential closure was pro-
posed for the analysis of existential bare plurals and more generally for weak indefinites
(see Chaps. 4 and 5 for the distinction between strong and weak indefinites).
1.2.3 Indefinites as Choice Functions
Another type of analysis is due to Reinhart (1997a) and Winter (1997), who propose
that certain indefinites be represented as existentially quantified choice functions.12
In the following representations, f is a choice function and [[student]] is a notation
for the denotation of student, i.e., the set of atomic students. Similarly, two students
and sm students respectively denote the set of sums of two students and the set of
sums of students of unknown cardinality:
(23) a. a student translates as f([[student]])

b. two students translates as f([[two students]])
c. sm students translates as f([[students]])
12
According to Reinhart, a subset of indefinite DPs does not behave like generalized quantifiers
and should be represented by means of choice functions. Winter, on the other hand, maintains that
all indefinites should be represented by choice functions.
A choice function maps any non empty set onto an element of that set. It is
therefore a function of type <<e,t>,e>, which applies to the property denoted by
the nominal predicate (of type <e,t>) and yields an individual (of type e) that has
that property. So the choice function applied to the property student returns a
student in the case of (23a), a plural individual corresponding to a sum of two
students in the case of (23b) and a plural individual of undetermined cardinality
in the case of (23c).
In the context of this analysis, the contribution of the indefinite is to introduce a
variable over choice functions that gets bound by existential closure, which may
apply at various points of the representation. Sentence (24) may be represented as
in (24¢a), which can be glossed as in (24¢b).
(24) A student is absent.

(24¢) a. $f (be-absent (f(student)))
b. There exists a choice function and the student that this function
chooses is absent
One advantage of choice functions is that they can account for the scope ambi-
guities characteristic of indefinites without resorting to covert movement at the level
of Logical Form. Example (25) illustrates this point. (25¢a) represents the interpreta-
tion of (25) in which the indefinite outscopes the universal quantifier. (25¢b) accounts
for the interpretation in which the universal quantifier scopes over the indefinite.
(25¢c) and (25¢d) are the tripartite representations that correspond respectively to
(25¢a) and (25¢b).
(25) Every woman read a book.

(25¢) a. $f [∀z (woman(z) → read(z, f(book)))]
b. ∀z [woman(z) → $f (read(z, f(book)))]
c. $f ∀z (woman(z)) [read(z, f(book))]
d. ∀z $f (woman(z)) [read(z, f(book))]
In (25¢a) and (25¢b), the function f applies to sets of books. While remaining
in situ, the DP a book outscopes the universal quantifier of every woman in (25¢a)
due to the fact that the choice function variable, f, is bound by an existential quantifier
with wide scope. In (25¢b), on the other hand, the point of insertion of the existential
quantifier is lower.
We may compare this approach to the traditional approach in logic, where the
relative position of quantifiers (existential and universal) accounts for the difference
in interpretation: (26a–b) correspond respectively to the wide and narrow scope
interpretations of the indefinite:
(26) a. $x ∀y (woman(y) → (book(x) ∧ read(y,x)))

b. ∀y $x (woman(y) → (book(x) ∧ read(y,x)))
The superiority of a choice function analysis becomes clear when the property
denoted by the head noun of the DP is empty.13 To illustrate this, take the case of a
conditional sentence with an indefinite in the antecedent clause:
(27) Max will be furious if I invite a philosopher.

(27¢) a. $f [invite(I, f(philosopher)) → furious(Max)]
b. $x [(philosopher(x) ∧ invite(I, x)) → furious(Max)]
Let us suppose that there are no philosophers in the world. Then the representa-
tion (27¢b) is true, since an implication is true when its antecedent is false:
philosopher(x) is always false, so the antecedent is false and therefore the proposi-
tion (27¢b) is true, counter to intuition.
In contrast, the choice function analysis makes it possible to account for the intu-
ition that in a world without philosophers, (27) is not true but, at best, undetermined.
For, according to Reinhart (1997a), a choice function is only defined if the set denoted
by the property that the choice is based on is not empty. As a result, (27¢a) is not true:
since there is no function f such that f chooses a philosopher, f(philosopher) is not
defined and thus the antecedent of the conditional is not either.
Let us now briefly compare the choice function approach with the DRT account,
where an indefinite DP introduces an individual variable that is bound by the exis-
tential quantifier. The specificity of the choice function analysis is that the descrip-
tive content of the indefinite DP is not displaced: in both (25¢a) and (25¢b), the
expression f(book) appears in the second argument position of the predicate read.
Compare the formulas in (28¢), corresponding to the DRT-type analysis, where the
predicative expression is raised to the same level as the existential quantifier:
(28) Every woman read a book.

(28¢) a. $x [book(x) ∧ ∀y (woman(y) → read(y,x))]
b. ∀y (woman(y) → ($x (book(x) → read(y,x))))
Example (29) shows that this move of the DP from its surface position is neces-
sary. If we did not move the condition book(y) from its surface position, we would
obtain the representation (29¢b), which is not satisfactory:
(29) Each professor rewarded every student who read a book he had
recommended.
(29¢) a. ∀x [prof(x) → $y (book(y) ∧ recom(x,y) ∧
(∀z (student(z) ∧ read(z,y)) → reward(x,z)))]
13
In the literature, this point is known as the empty restriction problem or the Donald Duck
problem.
(29¢) b. ∀x [prof(x) → $y (∀z (student(z) → ((book(y) ∧ recom(x,y) ∧

read(z,y)) → reward(x,z)))]
(29¢b) is not adequate, since it is verified in a world where the professors have
not rewarded any students, not even those who read recommended books. Indeed,
y can be instantiated with anything other than a book in order for (29¢b) to be
verified. In (29¢a) this problem is avoided because the predicative condition
book(y) has been moved out of the antecedent of the conditional.
In the case of (29), this movement is not problematic. But moving the descrip-
tive content of the DP may pose a problem in the case of plural indefinites. This
is illustrated in example (30), due to Ruys (1992):
(30) If three relatives of mine die, I will inherit a house.

(30¢) a. $f [(die(f(relatives)) ∧ |(f(relatives))| = 3) → $y (inherit(I, y) ∧
house(y))]
b. [$X (relatives(X) ∧ |X| = 3 ∧ (∀x ∈ X (die(x) → $y (inherit(I, y) ∧
house(y)))))]
Sentence (30) may receive an interpretation where the indefinite three relatives
of mine takes narrow scope with respect to if: I will inherit a house once any group
made up of three of my relatives passes away. But (30) can also mean that there is a
group of three of my relatives such that if all three die, then I will inherit a house.
This second interpretation, where the indefinite takes wide scope, is well captured
by (30¢a), which uses choice functions, but poses a problem for the DRT-type analy-
sis as (30¢b): (30¢b) is not able to dissociate wide scope and a distributive interpreta-
tion of the plural indefinite. The advantage of (30¢a) is that it can account for the
collective reading, even if the predicate die is a distributive predicate, ultimately
applying to individuals.
1.2.4 Indefinites as Skolem Terms
Building on Farkas (1997a, b, 2001) and Steedman (2003, 2006) proposed that
indefinites should not be represented as variables but rather as referential
expressions (type e) modeled as Skolem terms, which can be either (a) depen-
dent terms obtained by applying a Skolem function, e.g., ‘the mother of’ or
‘the daughter of’ to all the variables bound by a universal quantifier in whose
scope the indefinite occurs or (b) constant terms (whenever there are no univer-
sal quantifiers C-commanding the indefinite). Steedman’s proposal resembles
the choice function analysis insofar as it assumes that indefinites are analyzed
in situ.
In (31a), the indefinite in the object position is represented as a dependent

Skolem term notated sk53(x), referring to the entity14 that is obtained by applying the
constant Skolem function sk53 to each variable x bound by the universal quantifier.
In other words, the individual referred to by somebody is not picked up randomly
but instead its referent depends on the variable bound by the universal quantifier.
The difference between (31a) and (31b) is that in the latter, somebody translates as
a constant, notated sk95, without x. This notation is meant to account for ‘widest
scope’ indefinites, which according to Steedman refer to ‘specific individuals’,
which ‘have scope everywhere’ (pp. 14, 37)15:
(31) Everybody loves somebody.

a. ∀x[person (x) → (person (sk53(x)) ∧ loves (sk53(x)), x)]
b. ∀x[person (x) → (person (sk95) ∧ loves (sk95, x)]
It should be stressed that the Skolem function proposed by Steedman is a con-

stant (as indicated by the numerical index attached to sk (from Skolem)) in both
(31a) and (31b), i.e., regardless of whether the indefinite is interpreted as specific
(taking widest scope) or as dependent on another quantifier. Under Steedman’s
analysis then, existential closure is dispensed with.
But note that Steedman does not discuss bare nouns. For this type of indefinite
expression, Diesing’s (1992) VP-level existential closure might well be needed.
1.2.5 Indefinites and Properties
Nominal expressions lacking a determiner (currently referred to as ‘bare NPs’) can

be grouped together with indefinites because in at least part of their distribution they
take existential readings. According to a quite influential proposal, existential bare
NPs are property-denoting expressions (type <e,t>) not only when they function as
predicates but also when they occur in argument positions (McNally 1995a; van
Geenhoven 1996; Dobrovie-Sorin 1997a, b; Ladusaw 1994; Chung and Ladusaw
2003). Property-denoting arguments rely on a particular rule of semantic composi-
tion, known as ‘semantic incorporation’ (cf. Chap. 2).
14
Steedman insists that Skolem terms do not denote functions (a constant Skolem function is
merely part of their representation) but are referential terms (type e) that are assigned structured
representations.
15
According to Steedman (2006) the scope of indefinites is always assigned in situ and depends on
‘generalized Skolem term specification’, which is an “anytime” operation, in the sense that it can
apply at any point in a derivation: a Skolem term is specified as a constant if the specification
operation applies as soon as the DP constituent has been formed, i.e., before the DP combines with,
e.g., some quantified expression; it is specified as a dependent Skolem term if the specification
operation applies only after the indefinite DP combines with the main predicate and other argu-
ments of that predicate.
Table 1.2
Indefinites analyzed as Notation References
Existential generalized lP $x (N(x) Ù P(x)) Montague (1974)
quantifiers
Free variables x Ù N(x) Kamp (1981) and Heim (1982)
Choice functions f(N) Reinhart (1997a) and Winter (1997)
Skolem terms Sk53 (x) Steedman (2006)
Property-denoting expressions N Van Geenhoven (1996), McNally (1998),
and Dobrovie-Sorin (1997a, b)
Several authors have proposed that the ‘weak’ readings (in the sense of Milsark
(1977)) of indefinites can also be analyzed as relying on property denotation
(see Ladusaw 1994; Dobrovie-Sorin 1997a, b; McNally and van Geenhoven 1998
and Chaps. 4 and 5, below). This would explain, among other things, their ability to
appear in existential there constructions (see McNally 1998).
1.2.6 Indefinites as Existential Generalized Quantifiers

over Amounts
In this book we will show that the property analysis of indefinites, which we had
ourselves adopted in Dobrovie-Sorin and Beyssade (2004), is confronted with several
problems. We will therefore put forward a new analysis for weak indefinites, which
crucially relies on the distinction between two types of entities: individualized enti-
ties and amounts, notated ei and ea, respectively. The core proposal to be developed
in the various chapters of the book will be that weak indefinites are existential gen-
eralized quantifiers over amounts, i.e., expressions of type <<ea,t>,t>.
1.2.7 Conclusion
The following table summarizes the analyses of indefinites that can be found in the
literature (Table 1.2).
In the various chapters of this book we will present arguments based on lexical,
syntactic and contextual criteria, which allow us to choose among the possible
representations of indefinites.
1.3 Semantic Properties of Nominal Determiners
Assuming the need to distinguish between referential DPs, quantified DPs and
indefinite DPs, we must specify the criteria that enable us to place a random DP
into one of these three classes. Referential DPs can be easily spotted, even if their
1.3 Semantic Properties of Nominal Determiners 19
determiners are not all of the same form. Here we will assume the usual list: proper
names, definite and demonstrative DPs (i.e., DPs with definite or demonstrative
articles), possessive DPs (i.e., DPs with possessive adjectives).
Referential DPs
a. proper names: Mary, John, Smith
b. definite DPs: the book, the girl
c. demonstrative DPs: this book, that girl
d. possessive DPs: my book, my girl
The remaining DPs must be classified as either indefinite or quantified DPs. We

will place DPs with indefinite articles (a girl) among the indefinite DPs and univer-
sal DPs (every girl, each girl, all girls) among the quantified DPs. But what should
we do with the rest: two girls, three girls, most girls, 20% of the girls, certain girls,
a few girls, exactly three girls, etc.?
Several semantic tests and semantic properties, such as conservativity, intersec-
tivity and symmetry, can be used in order to classify DPs (Barwise and Cooper
1981; Keenan and Stavi 1986). To illustrate these properties, sentences will be rep-
resented as tripartite structures of the form Det A B corresponding to the syntactic
sequence Det NP VP obtained from a subject – predicate configuration DP VP by
shifting DP into Det and NP: A is the set of x’s such that NP(x) is true and B the set
of x’s such that VP(x) is true. The determiner relates a noun phrase and a verb
phrase.
1.3.1 Conservativity
A determiner Det is said to be conservative if and only if (a) and (b) have the same
truth conditions:
(a) Det A B
(b) Det A (A ∩ B)
This means that in order to determine whether Det A B is true, one need only
look at set A and the intersection of A and B, i.e., (A ∩ B); those elements that are
neither A nor B, as well as those that are B but not A, are irrelevant.
(32) a. All cats are grey.

b. All cats are grey cats.
(32¢) a. (Most/Some) children sing.
b. (Most/Some) children are children who sing.
(32″) a. No child sings.
b. No child is a child who sings.
To see whether the proposition all cats are grey is true, one does not need to
identify non-grey non-cats (e.g., white mice) or grey non-cats (e.g., grey mice).
As shown by Keenan and Stavi (1986), the majority of natural language deter-
miners are conservative. There are however certain elements that might be analyzed
as determiners (due to their prenominal position) but that are not conservative (cf.
Gamut 1991, vol II, p. 246):
(33) a. Only men smoke.

b. Only men are men who smoke.
It seems clear that (33a) and (33b) are not true in the same worlds. To evaluate
(33a), it is not sufficient to look at the set of men and the set of men who smoke, we
must also consider the set of non-men smokers. The non-conservativity of only-DPs
is not problematic for Keenan and Stavi’s generalization, since there are reasons to
believe that only is not a determiner: it never combines with a common noun but
only with a maximal nominal projection (a proper name, a common noun preceded
by a determiner or a bare plural).
(34) a. Only John

b. Only the neighbor *Only neighbor
c. Only girls *Only girl
Many is another expression that is not conservative and yet it is traditionally

analyzed as a determiner:
(35) a. Many Scandinavians have won the Nobel prize.

b. Many Scandinavians are Scandinavian Nobel prize winners.
Clearly (35a) and (35b) do not have the same truth conditions: (35a) is evaluated
with respect to the number of Nobel prizes and (35b) is evaluated with respect to the
number of Scandinavians.
If many is assumed to be a determiner, then it must be admitted that conservativ-
ity does not characterize all natural language determiners but only the extensional
ones (cf. Keenan and Stavi 1986).16 Extensional determiners are those that do not
vary according to context.17 This is not the case with the relative use of many.
16
An alternative solution is to say that many takes a third argument, one that corresponds to the
norm. This argument must be contextually instantiated. Once the norm is determined, many
becomes conservative.
17
Formally, a determiner Det is said to be extensional if and only if for all A, B, E and E¢ such that
E¢ Ê E and Det A B is true in E, then Det A B is also true in E¢.
1.3.2 Intersectivity
A determiner Det is said to be intersective if and only if (a) and (b) have the same
truth conditions:
(a) Det A B
(b) Det (A ∩ B) B
In other words, to verify a proposition of type (a), we must check whether Det
individuals that are A and B are B. To determine the truth value of the sentence,
one need only consider individuals that verify both A and B. One need not know
the set A in its entirety: specifically, there is no point in considering those A’s that
are not B.
Certain determiners, such as cardinals, are intersective:
(36) a. Two students arrived late.

b. Two individuals that are students and that arrived late arrived late.
Besides cardinals, the following determiners are also intersective: a, some,

several.
1.3.3 Symmetry
A determiner Det is said to be symmetric if and only if (a) and (b) have the same
truth conditions:
(a) Det A B
(b) Det B A
It can be demonstrated (cf. Szabolcsi 1997: 21) that, for conservative determin-
ers, symmetry goes hand in hand with intersectivity: if a determiner is conservative
and intersective, then it is also symmetrical (if Det A B, Det A (A∩B) and Det
(A∩B) B have the same truth conditions, then Det B A does so too) and vice versa,
if a determiner is conservative and symmetric, then it is also intersective (if Det A
B, Det A (A∩B) and Det B A have the same truth conditions, then Det (A∩B) B
does so too).
This generalization can be illustrated by considering again example (36a): two,
which is intersective, is also symmetric. (36a) and (36c) have the same truth
conditions.
(36) a. Two students arrived late.

c. Two individuals that arrived late are students.
1.3.4 Proportional Determiners
A determiner Det is said to be proportional when the truth conditions of Det A B

depend on the relation between the cardinality of (A∩ B) and that of A. Many, most,
20% of, all are examples of proportional determiners. Such determiners are non-
intersective: to verify whether a certain proportion of A is B, it is not enough to
compare the elements of (A∩B) and those of B. One must also know the cardinality
of A, the number of individuals that verify A. Thus, (37a) and (37b) do not have the
same truth conditions:
(37) a. Most students arrived late.

b. Most individuals who are students and who arrived late arrived late.
Imagine that there are six students and that only two arrived late. Two out of six
is not much. In this model, (37a) is false. And yet (37b) is true.
Proportional determiners are not symmetric either: (37a) may be true while (37c)
is false. This would be the case if, out of six students, five arrived late in addition to
20 other people who are not students.
(37) c. Most individuals who arrived late are students.
Proportional determiners should be distinguished from partitives such as two of

the, which are intersective (cf. (38a)) and symmetric (cf. (38b)).18
(38) Two of the students arrived late.

a. Two of the people that are students and that arrived late arrived late.
b. Two of the people who arrived late are students.
Every can be classified as a proportional determiner, since totality can be viewed

as a specific proportion, i.e., 100%.19
18
Here, we are not taking presuppositions into account.
19
This point is still under debate. Thus, Keenan (1996) notes that, unlike other proportional deter-
miners, every is not necessarily defined as a relation between the cardinality of (AÇB) and the
cardinality of A. Knowing the cardinality of A is not necessary to verify that Every A is B, since it
is enough to check whether the set of A’s that do not verify B is empty.
1.3.5 Monotonicity
1.3.5.1 Monotone Increasing with respect to A
Det is monotone increasing with respect to A if, for all A¢ greater than A (e.g.,
A = linguist and A¢ = researcher), Det A¢ B is true if that Det A B is true:
(39) a. Many linguists are French.

b. Many researchers are French.
(40) a. Few linguists are French.
b. Few researchers are French.
(39a) entails (39b), therefore many is monotone increasing with respect to A.

Few is not monotone increasing with respect to A: (40a) does not entail (40b); there
may be many French researchers but few French linguists.
1.3.5.2 Monotone Increasing with respect to B
Det is monotone increasing with respect to B if, for all B¢ greater than B (e.g.,
B = French and B¢ = European), Det A B¢ is true if that Det A B is true:
(39) c. Many linguists are European.

(40) c. Few linguists are European.
(39a) entails (39c), therefore many is monotone increasing with respect to B.

Few is not monotone increasing with respect to B: (40a) does not entail (40c); there
may be few French linguists but many European linguists.
1.3.5.3 Monotone Decreasing with respect to A
Det is monotone decreasing with respect to A if, for all A¢ smaller than A (e.g.,
A = researcher and A¢ = linguist), Det A¢ B is true if that Det A B is true:
(41) a. Many researchers are French.

b. Many linguists are French.
(42) a. Few researchers are French.
b. Few linguists are French.
(41a) does not entail (41b), therefore many is not monotone decreasing with
respect to A. Likewise, few is not monotone decreasing with respect to A: (42a)
does not entail (42b); there may be few French researchers but many French lin-
guists, relative to linguists in general.
1.3.5.4 Monotone Decreasing with Respect to B
Det is monotone decreasing with respect to B if, for all B¢ smaller than B (e.g.,
B = European and B¢ = French), Det A B¢ is true if that Det A B is true:
(43) a. Few researchers are European.

b. Few researchers are French.
(43a) entails (43b), therefore few is monotone decreasing with respect to B.
1.3.6 The Semantic Characterization of Indefinites
The semantic properties of determiners presented above make it possible to identify

indefinite DPs by distinguishing them from both quantificational DPs and referen-
tial DPs. Indefinite determiners may be defined as the subclass of determiners that
are both symmetric and intersective (cf. Keenan 1987).
All cardinals (two, three, etc.) and modified cardinals (at least two, at most three)
are symmetric. They are also conservative and therefore, because of the relationship
between conservativity, intersectivity and symmetry mentioned in Sect. 1.3.3., they
are intersective as well.
DPs headed by intersective determiners are semantically flexible: depending on
the syntactic context and the lexical content of the determiner, they can be analyzed
as existential generalized quantifiers over amounts, as Skolem terms, or as
quantificational DPs (i.e., existential generalized quantifiers over individuals).
1.4 The Interpretation of Indefinites
1.4.1 The Interpretation of Indefinites and Presupposition
In order to describe the interpretation of indefinite DPs, we need to distinguish

between assertion and presupposition.
1.4.1.1 Assertion and Presupposition
Ever since the work of Frege, Russell and Strawson on the meaning and denotation
of linguistic expressions, it is common to distinguish between assertion and presup-
position: that which is asserted is subject to debate, while that which is presupposed
is part of the background information and as such is not included in that which can
be questioned. This distinction has an effect upon the denotation of a sentence.
1.4 The Interpretation of Indefinites 25
A false presupposition triggers a presupposition failure, which is usually analyzed

in terms of truth value gap: the sentence has no truth value, it is neither true nor
false. To illustrate this distinction, let us consider the following examples:
(44) a. Mary is pregnant and John doesn’t know it.

b. John doesn’t know that Mary is pregnant.
In (44a), it is asserted that Mary is pregnant, whereas in (44b) Mary’s being

pregnant is presupposed. So, if Mary is not pregnant, (44a) is false, while (44b) is
neither true nor false.
1.4.1.2 Presupposition of Existence and Assertion of Existence
Certain sentences presuppose the existence of individuals corresponding to the DPs

that they contain, while other sentences assert their existence. We will respectively
talk about presuppositional and existential (non-presuppositional) readings. Definite
articles are unanimously assumed to presuppose existence (see (45a), which presup-
poses the existence of a king of France), whereas indefinite articles and cardinals are
often existential and non presuppositional (see (45b)).
(45) a. John says that the king of France is bald.

b. John tells that a king of France is bald.
Based on the examples in (46) and (46¢), Reinhart (1995) points out that speakers
do not hesitate to say that (46¢a) and (46¢c) express false judgments and that (46¢b)
is true. On the other hand, she claims that judgments waver for examples (46a–c).
(46) a. An American king lived in New York.

b. No American king lived in New York.
c. Two American kings lived in New York.
(46¢) a. There was an American king in New York.
b. There were no American kings in New York.
c. There were two American kings in New York.
The examples in (46¢) are there-sentences, which do not presuppose but rather
assert the existence (or non existence) of one or more individuals, one or more
American kings. Thus they are either true (see (46¢b)) or false (see (46¢a) and
(46¢c)).
The examples in (46), on the other hand, are subject-predicate configurations in
which indefinites occupy the preverbal subject position. In this position, indefinites
can have two different interpretations: the first one is identical to the interpretation
found in (46¢); the second interpretation is partitive: “one of the American kings”,
etc. In the first case (assertion of existence), the judgments of acceptability for
(46a–c) are identical to those for (46¢a–c). But in the second case, the indefinites of
(46a–c) presuppose that the set of American kings is not empty and since the pre-
supposition contradicts their knowledge of the world, speakers cannot assign truth
values to (46a–c). The observed fluctuation in judgments is probably due to the fact
that speakers do not analyze these sentences in the same way. Some speakers ana-
lyze them as existential sentences and thus judge them to be true or false, the same
as they do for the examples in (46¢). Others analyze them as presuppositional.
1.4.1.3 Presupposition and Partitivity
Some indefinites, e.g., those introduced by certain, have a partitive interpretation.

Attal (1976) points out the contrast in French between certains, which gives rise to
a quantificational reading, and des, which cannot do so (except marginally, usually
in a contrastive context):
(47) a. Certains enfants étaient tristes.

‘Certain children were sad.’
b. ?? Des enfants étaient tristes.
‘Sm (unaccented some) children were sad.’
According to Milsark (1977), there are other indefinites, in particular cardinal

indefinites, which are compatible with a partitive reading. In (48), two girls has an inter-
pretation comparable to that of an explicitly partitive indefinite (i.e., two of the girls):
(48) Two girls were blond, all the others had dark hair.
Be they implicit or explicit, partitive indefinites presuppose the existence of a

contextually determined set of individuals (whence the term “D(iscourse)-linked”
due to Pesetsky 1987) and can be represented as tripartite structures, in the same
way as quantificational DPs:
(49) Two x (girl(x)) [x was blond]
1.4.2 Distributive and Collective Readings
Truly quantificational DPs (such as DPs with each in English and chaque in French)
are necessarily distributive, whereas definite and indefinite DPs are ambiguous
between a collective and a distributive reading.
1.4 The Interpretation of Indefinites 27
(50) a. My children sang songs.

b. Three children sang a song.
These sentences can mean that the children sang together, or that they sang
individually.
Theoreticians fail to agree on how to represent the distributive reading: should a
distributive operator be introduced in Logical Form, or is the difference between the
two readings not a true ambiguity but rather a matter of underdetermination, all
pluralities being open to different partitionings (cf. the “covers” used by Gillon
1996; Schwarzschild 1992)? A set made up of three individuals {a,b,c} could be
partitioned in several different ways: (i) as a group (a collective individual) made up
of three atomic individuals; (ii) as a set made up of three distinct individuals; (iii) as
a set made up of two groups, with three possibilities: {(a,b),c}, {(a,c),b}, {(b,c),a}.
Which of these structures is chosen can remain undetermined: besides the collective
and strictly distributive readings of (50a–b), corresponding to (i) and (ii), these
examples can be used to describe others situations or events, corresponding to the
possibilities enumerated in (iii), e.g., two children sang together and one sang indi-
vidually. That being said, there are syntactic contexts (syntactic position or lexical
properties of the predicate) that can impose one or another of the structures
(“covers”) that are compatible with a plural DP. For example, a predicate that can
only apply to atomic individuals forces a strictly distributive reading:
(51) Three students are strangers.
1.4.3 Scope Ambiguities
Since Fodor and Sag (1982), we know that the scopal properties of indefinite DPs
differ from those of quantificational DPs. The latter cannot scope out of the minimal
clause to which it belongs, whereas certain indefinites can do so:
(52) Two professors did not know whether every student had read an eighteenth
century writer.
In this example, the speaker may have in mind a particular eighteenth century
writer. In terms of scope, an eighteenth century writer can be said to have wider
scope than two professors, even though the latter is not contained in its minimal
clause.
For Fodor and Sag, the wide scope interpretation of indefinites is to be explained
as being due to a referential reading, comparable to that of demonstratives; the nar-
row scope interpretation would be due to a “quantificational” reading.
Fodor and Sag’s analysis cannot account for the so-called ‘intermediate
readings’ observed by Farkas (1981) and rediscovered by Abusch (1994). Thus, a
sentence such as (53) allows not only the interpretation paraphrased in (54a), which
is the only one predicted by Fodor and Sag, but also the ‘intermediate’ reading para-
phrased in (54b), according to which the indefinite scopes out of the embedded
sentence in which it occurs, but it does not take scope over the quantified DP that
occurs in the main sentence:
(53) Every professor rewarded every student who read a novel.

(54) a. There is a novel, such that every professor rewarded every student
who read it.
b. Every professor chose a specific novel and rewarded every student
who read that novel.
Moreover, certain indefinites, in particular bare mass nouns and bare plurals,
take “extra-narrow” scope (see Carlson‘s 1977a observations for English, which
extend to Romance languages such as Spanish, Italian and Romanian), which is
never found with canonical quantifiers. Yet another scope problem concerns
modified cardinals, which, unlike quantificational phrases (cf. (55a) vs (55b)), can-
not take scope over the QP in subject position (Liu 1990; Beghelli and Stowell
1997) when they occur in object positions:
(55) a. Every professor will examine at least two students.

b. Every professor will examine each student.
These various problems will be analyzed in Chaps. 2 and 6.
1.4.4 Specific/Non-specific/Generic Readings
Examples (56)–(58) respectively illustrate specific, non-specific and generic readings

of singular indefinites:
(56) John saw a good movie yesterday.

(57) John is looking for a secretary that can speak Russian.
(58) A bird can fly.
These readings have been relatively well described in the literature. One of our
goals is to correlate these interpretive differences with the different denotation types
of indefinites.
1.5 Conclusion 29
1.5 Conclusion
The aim of this book is to circumscribe the class of indefinite DPs and to propose a
unified analysis that may cover the diversity of their context-driven interpretations.
Each of the chapters is devoted to a particular empirical domain (bare NPs, existen-
tial sentences, ambiguities, scope, dependency relations, genericity) and presents
empirical arguments that may help us choose among the various analyses of
indefinites to be found in the literature, which we have reviewed above in Sect. 1.2.
We will also outline the theoretical challenges associated with treating indefinites as
referential, quantificational or property denoting. The latter hypothesis will be care-
fully examined, as it has played an important role in the recent literature starting
with McNally (1995a) and van Geenhoven (1996). Although we had ourselves
adopted the property analysis of indefinites in Dobrovie-Sorin and Beyssade (2004),
the empirical and theoretical limits of this account have led us to abandon it in this
book. We put forward a new hypothesis, based on distinguishing two types of enti-
ties: individualized entities and amounts, respectively notated ei and ea. Given this
distinction, we treat weak indefinites as existential generalized quantifiers over
amounts (type <<ea,t>,t>). Strong indefinites will be analyzed as either Skolem
terms (Steedman (2003, 2006)) or as quantificational. The choice among these pos-
sible analyses of indefinites will be shown to depend on the lexical content of verbal
and nominal predicates, the syntactic context and information structure (topicalization,
focalization).
Chapter 2
Bare Noun Phrases
Bare Noun Phrases (bare NPs henceforth) are nominal constituents that are not
preceded by an overt determiner. The analysis of bare NPs is challenging insofar as
we need to (i) account for the crosslinguistic variation regarding the use of bare
NPs, (ii) propose different analyses for bare plurals and count bare singulars, as they
exhibit very different distributions in some languages and (iii) account for the
similarities and differences between argumental bare NPs (i.e., bare NPs that occupy
argument positions) and predicative bare NPs (i.e., bare NPs occurring as predicates
of copular sentences). Moreover, we need to account for the differences and similari-
ties between bare NPs and indefinites headed by overt Determiners, in particular
the French du/de la/des indefinites.
2.1 Bare Noun Phrases across Languages
Given the standard view, according to which it is the determiner that is responsible
for referential or quantificational uses of nominal expressions, bare NPs should be
assumed to denote properties or, in extensional terms, sets. Qua property-denoting
expressions, bare NPs are expected to appear in predicate positions and ruled out
from argument positions. We will show, however, that there is no clear correlation
between absence vs presence of article and predicate vs argument positions, which
means that the possibility for some bare NPs to appear in predicate positions should
not be taken as evidence in favor of the property analysis of bare NPs occupying
argument positions.

32 2 Bare Noun Phrases
2.1.1 An Overview of Crosslinguistic Variation
All Romance languages, including French, allow bare NPs to occur in predicate
positions. Note however that bare NPs are allowed in predicate positions only with
nouns describing professions, roles, etc. (see Sect. 2.7 below):
(1) a. Jean est médecin.
Jean is doctor
‘Jean is a doctor.’
b. Jean et Marie sont médecins.
Jean and Marie are doctors
‘Jean and Marie are doctors.’
In examples such as (1a–b), the bare NPs can safely be assumed to denote
properties.
In all Romance languages other than French (Spanish, Catalan, Romanian,
Italian, as well as continental and Brazilian Portuguese), bare plurals and bare mass
nouns can also appear in most of the argument positions:
(2) a. Juan invitó estudiantes. Spanish
Juan invited students.
b. Ion a băut lapte. Romanian
Ion has drunk milk
‘Ion drank milk.’
In the same contexts, French disallows bare NPs (see (3a–b)) and uses indefinites
headed by de la/du/des instead (see (3a¢–b¢)):
(3) a. *Jean a invité étudiants.
Jean has invited students
a¢. Jean a invité des étudiants.
Jean has invited des students
‘Jean invited students.’
b. *Dans la rue jouaient étudiants.
in the street play.impf students
‘In the street were playing students.’
b¢. Dans la rue jouaient des étudiants.
in the street play.impf des students
‘Students were playing in the street.’
Turning now to bare count singulars (BSs henceforth),1 Romance languages (other
than French) differ among each other regarding their use in argument positions:
1
The label ‘bare singulars’ used in some of the current literature (Schmitt and Munn 1999; Munn
and Schmitt 2005; Farkas and de Swart 2003) and taken up here is somewhat misleading because:
(i) there is no singular marking on bare count nouns and therefore they can be viewed as
2.1 Bare Noun Phrases across Languages 33
(4) a. Joao invitou aluno. Brazilian Portuguese

b. *Juan invitó estudiante. Spanish
c. *Ion a invitat student. Romanian
d. *Gianni ha invitato student. Italian
John has invited student.
These examples show that in Brazilian Portuguese, BSs are freely used in postver-
bal positions (objects or postverbal subjects), whereas in Spanish, Catalan or Romanian
they are severely restricted, being able to appear only in the object position of certain
predicates. In Italian, BSs are ruled out even in these restricted contexts:
(5) a. Juan compró apartamento. Spanish
Juan bought apartment
‘Juan bought an apartment.’
b. Ion şi-a cumpărat apartament. Romanian
Ion se.dat-has bought apartment
‘Ion bought an apartment (for himself).’
c. * Gianni ha comprato apartamento. Italian
John has bought apartment.
English behaves like Romance languages other than French insofar as it freely
allows BPs and bare mass nouns in argument positions but differs from these lan-
guages insofar as it rules out BSs not only from all argument positions (the English
counterpart of (5)a–c is ruled out) but also from predicate positions.
(6) a. John invited students.
b. John drank wine.
c. *John bought apartment.
d. *John is professor.
The table below summarizes the generalizations illustrated above. The indica-
tions ‘yes’ and ‘no’ are straightforward: they mean that bare NPs are allowed and
respectively disallowed; the indication ‘yes/no’ means that bare NPs of the relevant
type are severely constrained (Table 2.1).
In this chapter we will attempt to explain the quite complex crosslinguistic
generalizations summarized in this table. We will leave aside Brazilian Portuguese,
for which the reader is referred to Schmitt and Munn (1999), Müller (2002), Munn
and Schmitt (2005), Dobrovie-Sorin and Pires de Oliveira (2008, 2010). We will
concentrate on Spanish, Catalan and Romanian, because these languages exhibit an
interesting contrast between BPs and BMNs, which are quite freely allowed, and BSs,
which are severely restricted in argument positions. The data in this group of
languages will be compared to those in French on the one hand and English on the
‘non plural’ rather than as ‘singular’; (ii) they are not semantically singular but rather number-
neutral. Nevertheless we have decided to adopt this label, in order to bring out the contrast between
BPs and these other unmarked count nouns. What we call ‘bare singular’ is called ‘count bare
noun’ by some authors (Müller 2002; Dobrovie-Sorin 2010; Espinal and McNally 2011.
Table 2.1 The use of bare NPs across languages

Predicate positions Postverbal argument positions
Bare NPs BPs/BMNs BSs
French Yes No No
Italian Yes Yes No
Spanish Yes Yes Yes/no
Catalan
Romanian
Brazilian Yes Yes Yes
English No Yes No
other. French disallows any kind of argumental bare NPs (including BPs and
BMNs) but allows bare NPs in predicate positions (more or less in the same contexts
as all the other Romance languages), whereas English allows argumental BPs and
BMNs but disallows BSs in all positions, including predicate positions.
The contrast between BSs on the one hand and BPs and BMNs on the other can
be restated as a generalization regarding indefinite articles. If we distinguish
indefinite articles (e.g., un and des in French, a in English, un(a) in Spanish, etc.)
from other indefinite determiners such as quelque(s), plusieurs in French or some,
certain, several in English, we can say that English, along with Romance languages
other than French, has indefinite articles for count singular nouns but lacks them for
plurals or mass nouns.
The brief presentation sketched above is sufficient for the reader to become aware
of the many questions that arise regarding the analysis of bare NPs: (a) What is the
syntactic analysis of bare NPs? Are they full DPs headed by covert Det’s or are they
truly bare? (b) Is the syntactic analysis of bare NPs different depending on whether
they appear in predicate or in argument positions? (c) Why is it that BPs and bare mass
nouns have a larger distribution than count BSs? Is the syntactic analysis of BPs and
bare mass nouns different from the syntactic analysis of BSs? (d) What is the semantic
analysis of bare NPs? Are they names of kinds (Carlson 1977a, c; Chierchia 1998),
property-denoting expressions (McNally 1995a, b, 1998; van Geenhoven 1996;
Dobrovie-Sorin 1995, 1997a, b; Dobrovie-Sorin and Laca 1996, 1999; Laca 1996;
Dobrovie-Sorin and Beyssade 2004) or do they introduce free variables that get bound
by existential closure, like indefinites (Diesing 1992; Kratzer 1988, 1995)?
2.1.2 The Distribution of Bare NPs in Romanian,

Spanish and Catalan
In this section we show that in Romanian, Spanish and Catalan, the distribution of
BSs is much more restricted than the distribution of BPs and bare mass nouns,
which indicates that BSs and BPs must be given different analyses (Dobrovie-Sorin
et al. 2005, 2006; Espinal and McNally 2011). The generalizations that hold in this
group of Romance languages will be illustrated with Romanian, which is less studied
than the other Romance languages and therefore, Romanian data are more informative
for the reader.
The distribution of bare NPs is more restricted than that of full DPs (nominals
with an overt determiner). Thus, bare NPs cannot appear in the preverbal subject
position2 nor as the object of experiencer-subject psych-verbs such as love, hate,
detest, respect, etc.
(7) a. ??Noroi curgea pe stradă.
mud flow.impf on street
‘Mud was flowing in the street.’
b. *Ion respectă profesori.
Ion respects professors
However, if we leave aside these contexts, bare plurals and bare mass nouns
can freely appear as direct objects of most verbs, as well as in the post-verbal
subject position3:
(8) a. Am desenat copaci.
have.1sg drawn trees
‘I drew trees.’
b. Ion mănîncă numai brânză.
Ion eats only cheese
‘Ion eats only cheese.’
(9) a. Aici se spală rufe
here se wash clothes
‘One does the laundry here.’
b. Pe stradă curgea noroi.
on street flow.impf mud
‘Mud was flowing in the street.’
Count bare singulars clearly differ from BPs and bare mass nouns insofar as
their distribution is much more restricted. Thus, BSs are generally ruled out in the
contexts illustrated above:
(10) *Am desenat copac.
have.1sg drawn tree
The example in (10) becomes grammatical if we add the singular indefinite article:
(11) Am desenat un copac.
have.1sg drawn a tree
Nevertheless, bare singulars can appear quite productively in the object position
of a reduced number of verbs, e.g., have and acquisition verbs, some intensional
verbs such as look for, some other verbs such as wear or use, some light verbs, and
also in some idiomatic expressions.
2
BMNs and BPs can appear in the preverbal subject position provided that they are modified (by an
adjective, PP or relative clause), coordinated or contrastively stressed (Longobardi 1994 among others).
3
We will leave aside the use of bare NPs after prepositions.
(12) a. Ion are casă.

Ion has house
‘Ion has a house.’
b. Casa asta are lift.
house-the this has lift
‘This house has a lift.’
c. Ne-au pus în sfârşit lift.
us.dat-have.3pl put in end lift
‘At last they put in a lift (for us).’
d. Ion şi-a cumpărat casă.
Ion se.dat-has bought house
‘Ion has bought a house (for himself).’
The data reviewed here show the necessity of a three-way distinction: (i) nominal
expressions with overt determiners; (ii) bare plurals and bare mass nouns (iii) count
bare singulars.
2.1.3 The Syntactic Structure of Bare NPs
The three-way distinction in the distribution of nominal constituents observed above

can be correlated with a three-way distinction in syntactic structure, as shown in
(13a–c); as an alternative to (13c) we may assume that the indefinite article a is
generated under Num and moved from there to Det:
(13) a. Bare singular

NP
copil
child
b. Bare plural
NumP
Num° NP
Pl copii
child.PL
c. Full DP
DP
Det NumP
a Num° NP
copil
Sg child.SG
Unlike full-fledged DPs, which are headed by a Det(erminer), the two types of
bare nominals are similar in that they both lack the functional category Det but
nonetheless different: bare plurals are NumPs, projections of the functional category
of Number, whereas bare (count) singulars are purely lexical projections of
the noun, NPs. An NP that lacks Number has in its extension both atomic entities
and sums thereof; a NumP constituent headed by Number has only atomic enti-
ties or only sums in its extension, depending on whether Number is valued as
singular or as plural.4 The role of Det is to ensure reference to entities or generalized
quantifiers.
Problems arise, however, as soon as we take into account the distribution of bare
NPs in predicate positions (see Sect. 2.8 below). It will turn out that we cannot treat
bare NPs in argument positions on a par with bare NPs in predicate positions and we
will therefore be led to distinguish between the two by assuming that bare NPs in
predicate positions are genuine bare NPs, whereas bare NPs in argument positions
are full DPs headed by a null Det.5 This hypothesis is welcome insofar as it is in line
with a widely assumed correlation: argument positions are occupied by full DPs,
which are entity-denoting constituents or generalized quantifiers; predicate positions
are occupied by bare NPs, which denote properties.
The hypothesis that bare NPs in argument positions are full DPs rather than mere
NPs is supported by a language such as French, which allows bare NPs in predicate
positions but not in argument positions, where des-phrases are used instead:
(14) a. Jean est professeur.
Jean is professor
‘Jean is a professor.’
b. Jean et Marie sont professeurs.
Jean and Marie are professors
‘Jean and Marie are professors.’
(15) a. Jean a une voiture/*voiture.
Jean has a car/*car
‘Jean has a car.’
b. Jean lisait des romans/*romans.
Jean read.impf des novels/*novels
‘Jean was reading novels.’
The French data can be captured by assuming that (i) argument positions must
host full DPs (universal constraint) and (ii) French does not have null Det’s. No
4
Note that this characterization of Number is compatible with the view that Number is not a func-
tional category but rather a feature that attaches to another (functional or lexical) category
(Bouchard 1998; Dobrovie-Sorin (to appear b)). Note also that Det does not necessarily subcatego-
rize for Number (see Munn and Schmitt (1999) on BSs in Brazilian Portuguese).
5
Alternatively, one may assume a type-shifting operation for bare NPs in argument positions. Note
however that such a type-shifting operation cannot be assumed to be general across languages.
easy account can be proposed if we assume the alternative view, according to

which bare NPs are allowed in argument positions: since French allows bare NPs
in predicate positions, we cannot understand why they should not be allowed in
argument positions.
2.2 Bare Plurals Are not the Plural Counterparts

of Singular Indefinites
Carlson (1977a, c) observed several important contrasts between the existential

reading of English BPs and the existential reading of singular indefinites and con-
cluded that BPs cannot be analyzed as the plural counterparts of singular indefinites.
Dobrovie-Sorin and Laca (2003) showed that Carlson’s observations for English
extend to Romance languages such as Spanish, Italian or Romanian.
2.2.1 Opacity
Carlson (1977a, c) noticed that in propositional attitude contexts, English BPs can
only take opaque readings, whereas singular indefinites are ambiguous, also allow-
ing transparent readings:
(16) a. Minnie wishes to meet a young psychiatrist.
b. Minnie wishes to meet young psychiatrists.
The opaque reading is possible for both examples but (16b) cannot be interpreted
as meaning ‘there is one or more young psychiatrists that Minnie wishes to meet’.
This interpretation is possible if we use an indefinite headed by some instead of the
BP:
(17) Minnie wishes to meet some young psychiatrists.
The same observations hold for the Romanian counterparts of these examples,
given in (18)–(19):
(18) a. Minnie vrea să consulte un tânăr psihiatru.
Minnie wants subj consult a young psychiatrist
‘Minnie wants to consult a young psychiatrist.’
b. Minnie vrea să consulte tineri psihiatri.
Minnie wants subj consult young psychiatrists
‘Minnie wants to consult young psychiatrists.’
(19) Minnie vrea să consulte nişte/câţiva tineri psihiatri.
Minnie wants subj consult some/a few young psychiatrists
‘Minnie wants to consult some/a few young psychiatrists.’
2.2 Bare Plurals Are not the Plural Counterparts of Singular Indefinites 39
2.2.2 Scope
Singular indefinites and BPs also differ regarding their scope interpretations:
(20) a. Everyone read a book on caterpillars.
b. Everyone read books on caterpillars.
(20b) only allows a reading on which books takes narrow scope with respect to
everyone whereas (20a), with a singular indefinite, is scopally ambiguous.
Again, the same observation can be made for corresponding examples in
Romance languages, e.g., Romanian:
(21) a. Toată lumea a citit o carte despre omizi.
everyone has read a book about caterpillars
‘Everyone read a book on caterpillars.’
b. Toată lumea a citit cărţi despre omizi.
everyone has read books about caterpillars
‘Everyone read books on caterpillars.’
BPs also take obligatory narrow scope with respect to negation, an issue that is
particularly important for BSs and as such will be discussed in Sect. 2.3.3 below.
2.2.3 Aspect
Dowty (1979) and Carlson (1977a, c) have noticed some other interesting properties
of bare nouns, concerning their aspectual effects on verbs. Unlike indefinites headed
by overt determiners, BPs cannot supply the boundedness that is required by telic-
ity. Therefore, those predicates that qualify as achievements or accomplishments
when their object is headed by an indefinite determiner6 yield activities when fol-
lowed by a BNP object. Examples (22a–b), from English and Spanish, thus show
that the alternation between culminative adverbs indicating duration (in X time) and
6
Note that French and Italian indefinite plurals headed by des and dei, respectively, are exceptional
in this respect, i.e., they resemble BPs insofar as they can induce atelicity effects. Cf. Dobrovie-
Sorin and Laca (2003):
(i) Jean a écrit des articles pendant trois ans/*en trois ans.
Jean has written des articles during three years/ *in three years.
‘Jean wrote articles during three years/*in three years.’
non-culminative adverbs (for X time) correlates with the alternation between plural
indefinites and BPs in object position.
(22) a. Max discovered two rabbits in his yard in two hours/?? for two hours.
a¢. Escribió dos artículos en tres meses/*durante tres meses.
wrote.3sg two articles in three months/*during three months
‘He wrote two articles in three months/*for three months’
b. Max discovered rabbits in his yard ??in two hours/for two hours.
b¢. Escribió artículos *en tres meses/durante tres meses.
wrote.3sg articles in three months/during three months
‘He wrote articles *in three months / for three months.’
Similarly, the use of BPs entails a durative interpretation of inherently
non-durative predicates such as kill or discover. Hence the possibility of embedding
these kinds of predicates under the aspectual verb continue:
(23) a. ?? Harvey continued to kill a rabbit. English
a¢. * El zorro siguió matando unas gallinas. Spanish
the fox continued killing some hens
‘The fox continued to kill some hens’.
b. Harvey continued to kill rabbits. English
b¢. El zorro siguió matando gallinas. Spanish
the fox continued killing hens
‘The fox continued to kill hens.’
The same remarks apply to postverbal subjects in Romance languages, in par-
ticular when they appear with intransitive verbs denoting a change of state:
(24) Seguían llegando invitados/*unos invitados. Spanish
continued arriving guests/ some guests
‘There continued to arrive guests /*some guests.’
2.2.4 Anaphoric Relations
BPs cannot function as antecedents of pronouns such as alţii ‘others’ or ceilalţi ‘the
others’. The unacceptability judgments illustrated in (25b) and (26b) can be attrib-
uted to the non-individualizable type of reference that characterizes BPs. The
Romanian examples in (25a) and (26a) show that plural indefinites headed by overt
determiners behave on a par with singular indefinites insofar as they are legitimate
as antecedents of alternative pronouns.
2.3 Count Bare Singulars Are not the Singular Counterparts of Bare Plurals 41
(25) a. Ion a recomandat un/trei/nişte romane, iar Maria a

Ion has recommended a/three/some novels and Maria has
recomandat altele.
recommended others
‘Ion recommended a/three/some novels and Maria recommended others.
b. # Ion a recomandat romane, iar Maria a recomandat altele.
Ion has recommended novels and Maria has recommended others
(26) a. Azi a/au venit un/nişte/trei studenţi străini.
today has/have arrived a/some/three students foreign
Mâine vor ajunge ceilalţi.
tomorrow will arrive the-others
‘A/some/three foreign student(s) arrived today. The others will arrive
tomorrow.’
b. # Azi au venit studenţi străini. Mâine vor ajunge ceilalţi.
today have arrived students foreign tomorrow will arrive the others.
Note that restrictions of this type do not indicate that BPs are non-referential.
Indeed, they allow appositive relatives, which are typical of referential DPs:
(27) Maria aducea de la bibliotecă cărţi, pe care Ion le citea.
Mary brought. impf from library books pe which Ion cl.acc read.impf
‘maria used to bring books from the library, which Ion read.’
2.3 Count Bare Singulars Are not the Singular

Counterparts of Bare Plurals
Recent work has shown that count bare singulars7 are allowed to occur in argument
positions not only in languages without articles (e.g., Russian or Hindi) but also in
languages with articles (e.g., Albanian, Norwegian, Hungarian and Romance lan-
guages). We will leave aside Brazilian Portuguese, a language in which BSs have a
distribution that is comparable to the distribution of BPs in the other Romance lan-
guages (see Sect. 2.1.1 above). The presentation below will concentrate on Romanian
but the core generalizations also hold for Spanish and Catalan, as described in
Bosque (1996), Laca (1999), Dobrovie-Sorin et al. (2005, 2006), Espinal and
7
Count bare singulars, which are not allowed in English (see Sect. 2.3.2 below), were neglected in
the post-carlsonian literature (with the notable exception of Kallulli’s (1999) analysis of Albanian)
and came to the foreground only recently, since 2003, due to work on Hungarian (Farkas and de
Swart 2003), Hindi (Dayal 2003, 2004), Norwegian (Borthen 2003) and Romance languages
(Dobrovie-Sorin et al. 2005, 2006).
McNally (2011). All these authors agree that the distribution of count bare singulars
is much more constrained than that of bare plurals. Italian BSs, which will not be
examined here, show an even more restricted distribution than the BSs of Romanian,
Spanish and Catalan.
2.3.1 Distribution
With most classes of verbs, count bare singulars are disallowed in object positions,
in clear contrast with bare plurals:
(28) a. Am văzut două/nişte/câteva păsări
have.1sg seen two some several birds
‘I saw two/some/several birds’
b. Am văzut *(o) pasăre
have.1sg seen a bird
‘I saw a bird’
Certain predicates do, however, allow BSs in their object positions. Such predi-
cates can be grouped into three lexical classes: (i) a avea ‘to have’ and acquisition
verbs; (ii) verbs such as a purta or a folosi ‘to wear, to use’, which can be viewed as
implying the verb ‘have’ itself; (iii) intensional verbs such as a căuta ‘look for’,
which can also be viewed as implying some kind of prospective possession:
(i) a avea ‘to have’ and acquisition verbs
(29) a. Ion are casă/ maşină/ copil /carte de credit/ paşaport/ bucătăreasă.
Ion has house car child card of credit passport cook
‘Ion has a house/car/child/credit card/passport/cook’
b. Casa asta are lift/ scară de serviciu/grădină.
house-the this has lift staircase of service garden
‘This house has a lift/staircase/garden.’
c. Ne-au pus în sfârşit lift.
us.dat-have put finally lift
‘At last they put in a lift.’
d. Ion şi-a cumpărat casă.
Ion se.dat-has bought house
‘Ion bought a house (for himself).’
(ii) a purta ‘to wear’, a folosi ‘to use’, a conduce ‘to drive’
(30) a. Maria poartă pantalon/ pălărie/ uniformă/ poşetă/ cravată/
Maria wears pant hat uniform purse tie
cămaşă/ rochie scurtă.
shirt dress short
‘Maria wears pants/hats/a uniform/a purse/ties/shirts/short dresses’
b. Ion foloseşte stilou/creion.
Ion uses pen pencil
‘Ion uses a pen/pencil’
c. Ion conduce camion.
Ion drives truck
‘Ion drives a truck’
(iii) prospective possession: a căuta’to seek’, a găsi ‘to find’, a vrea ‘to want’, a
dori ‘to wish’
(31) a. Ion caută secretară/nevastă/femeie/profesor/bucătar.
Ion seeks secretary wife woman professor/cook
‘Ion is looking for a secretary/wife/woman/teacher/cook’
b. Ion doreşte nevastă tânără.
Ion wishes wife young
‘John wants a young wife.’
The examples in (32a–b) show that the choice of the noun itself is constrained.
What seems to matter is that the verb together with the BS refers to a conventional-
ized type of possession:
(32) a. Ion are / a cumpărat (un) apartament/(o) casă /(o) maşină /
Ion has has bought an apartment a house a car
(un) calculator.
a computer
‘Ion has/bought an apartment/a house/a car/a computer’
b. *Ion a cumpărat castel/bloc /fabrică.
Ion has bought castle block factory
As the examples under (i)–(iii) show, the verbs that allow BSs describe actual or
prospective possession. However, no clear explanation can be found in the current
literature for the correlation between possession and the possibility of BSs in object
positions.
In the impersonal passive SE constructions illustrated in (33), the postver-
bal BSs occupy the object position of exactly the same verbs as those listed in
(i)–(iii) above but they qualify as subjects insofar as they agree with the main
verb:
(33) a. Ni s- a pus în sfârşit lift.
us.dat se-has put finally lift
‘At last they put in a lift (for us).’
b. Anul ăsta s-a cumpărat cuptor, anul viitor se va
year-the this se-has bought oven year-the next se will
cumpăra masină de spălat
buy machine of washing
‘An oven was bought this year, a washing machine will be bought
next year.’
c. De anul trecut se poartă pantalon/pălărie/uniformă/cravată
since year-the last se wears pant hat uniform tie
cămaşă rochie scurtă.
shirt dress short
‘Since last year it is fashionable to wear pants/hats/a uniform/a tie /shirts/
short dresses.’
d. Se caută secretară/ profesor de matematică/ bucătar.
se seeks secretary professor of mathematics cook
‘A secretary/mathematics professor/cook is needed.’
The only other predicate that allows postverbal subject BSs is the verb a fi ‘to be’.
Note that the examples below are paraphrasable by sentences with a avea ‘to have’:
(34) a. La facultate e secretară.
at faculty is secretary
‘The faculty has a secretary.’
b. În bloc nu e portar.
in building not is doorman
‘The building has no doorman.’
There are two other contexts in which bare singulars are allowed, namely with
light verbs and with idioms:
(35) a. a lua loc /parte/ fiinţă/ notă de
to take place part being note of
‘to take place/ to take part/ to come into being/ to take note of’
b. a da loc /exemplu /dovadă
to give place example proof
‘to give rise to/to give an example/to show’
c. a ţine loc de
to keep place of
‘to be a substitute for’
Idiomatic structures are characterized by the repetition of the same BS, occurring
as a subject and as the argument of a preposition:
(36) a. Cui pe cui se scoate.
nail pe nail se pulls-out
‘One nail drives out another’
b. Ban la ban trage.
money to money draws
‘Money attracts money’
c. Deal cu deal se întâlneşte, dar om cu om?
hill with hill se meets but man with man
‘Even hills meet, let alone men’
We may wonder whether the BSs in the latter two contexts should be analyzed in
the same way as those appearing in the previous contexts. A differentiating analysis
is suggested by the fact that in certain languages, e.g., French, bare singulars (as well
as bare plurals) are allowed with certain light verbs and in certain idiomatic expres-
sions but consistently banned with the verbs listed in (i)–(iii), including avoir ‘have’.
Summarizing, BSs can only be internal arguments, not external ones, and are
allowed only with a limited number of verbs. BPs have a larger distribution, which
suggests that the two types of bare NPs should not be analyzed in the same way.
2.3.2 Crosslinguistic Variation
An important argument in favor of treating bare singulars separately from bare plu-
rals is related to crosslinguistic variation: English is like Romance languages (other
than French) in allowing existential bare plurals but differs from them insofar as it
does not allow bare singulars, except in some idioms, as illustrated with (37).8 We
will not address here the use of BSs in idioms9:
(37) a. call (the) roll
b. take heart
c. suck face
8
There is a use of count bare singulars where they are interpreted as masses, in which case they are
allowed freely, as are other mass nouns (John was eating apple/meat/goulash).
9
It should also be noted that English allows bare singulars to appear productively inside N-V
compounds (or synthetic compounds of the form N-V + er).
(i) a. window washing, truck driving, deer hunting, pie baking
b. truck driver, deer hunter, pie baker
Most of the examples given in (29)–(31) can only be translated in English with
an indefinite singular DP10:
(38) a. *John has house/car/child/credit card/passport/cook.
a¢. John has a house.
b. *John bought house.
b¢. John bought a house.
(39) a. *Mary wears hat/uniform/handbag/tie/shirt/dress.
a¢. Mary wears a hat.
b. *John put on tie yesterday.
b¢. John put on a tie yesterday.
(40) a. *John is looking for secretary wife/woman/teacher/cook.
a¢. John is looking for a secretary.
b. *John wants young wife.
b¢. John wants a young wife.
The crosslinguistic variation regarding the (im)possibility of bare singulars in
object positions can be correlated with the variation regarding their (im)possibility
in predicate positions (Munn and Schmitt 2005).11
(41) a. *John is professor/student. English
b. Ion e profesor/student. Romanian
On the basis of these observations, Dobrovie-Sorin et al. (2005, 2006) suggest
the following generalization:
(42) If a language allows bare singulars in argument (more precisely in object)
position, then it also allows bare singulars in predicate position (Spanish and
Romanian vs. English).
In this case, bare singulars allow for modification by adjectives, similar to the Romance bare sin-
gulars in object position (e.g., Spanish (ii) or English (iii)):
(ii) Necesita cocinero mexicano
needs cook Mexican
‘(S)he needs a Mexican cook.’
(iii) a. I went big-game-hunting.
b. I’m going hot-babe-watching.
10
See Munn and Schmitt (2005) for further data and discussion. In addition to argument and predi-
cate positions, Munn and Schmitt examine BSs that occur as modifiers.
11
In English, the predicative construction typically involves a nominal preceded by an indefinite
article, whereas in Spanish and Romanian, among other Romance languages, there is a syntactic
and a semantic distinction between (i) and (ii). See also De Swart et al. (2005) for an analysis of
Note that the reverse does not hold: as already pointed out above, French allows
bare singulars (and bare plurals) in predicate positions but not in argument
positions.
2.3.3 Interpretation: Narrow Scope with respect to Negation
The interpretive restrictions regarding opacity, scope, aspect and anaphoric rela-
tions, described for BPs in Sect. 2.2 above, also characterize BSs. We will however
not illustrate all of these constraints, since they do not bring any new insight into the
phenomenon. Rather, we will concentrate on the obligatory narrow scope wrt nega-
tion, which can be observed in (43a–c) for BPs, bare mass nouns and BSs,
respectively:
(43) a. N-am citit romane.
neg-have.1sg read novels
‘I haven’t read novels’
b. De doi ani n-am băut vin.
of two years neg-have.1sg drunk wine
‘I haven’t drunk wine for two years (now)’
c. Ion nu are/nu şi-a cumpărat apartament/ maşină/calculator.
Ion neg has/neg se.dat-has bought apartment car computer
‘John didn’t buy himself an apartment/car/computer.’
Note that in examples such as (44a–b), built with run-of-the-mill verbs, narrow
scope with respect to negation cannot be indicated by using a BS:
(44) a. *N-am citit roman.
neg-have.1sg read novel
b. *De doi ani n-am întâlnit prieten.
of two years neg-have.1sg met friend
This ungrammaticality is expected, given that BSs can only be used with certain
verbs, call them ‘incorporating’ verbs. For non-incorporating verbs, there are two
bare predicate nominals in Dutch. Sentence (i), with a bare predicate nominal, is interpreted
intensionally as saying that John has the property associated with the profession ‘doctor’, whereas
sentence (ii), with the indefinite article, is interpreted extensionally as saying that John belongs to
the set of individuals that are doctors.
(i) Juan es médico.
Juan is doctor
(ii) Juan es un médico.
‘Juan is a doctor.’
ways of indicating narrow scope wrt negation. One possibility is to use a negative
word 12:
(45) a. N-am citit niciun roman.
neg-have.1sg read neither.one novel
‘I haven’t read any novel/I’ve read no novel.’
b. De doi ani n-am întîlnit niciun prieten drag.
of two years neg-have.1sg met neither.one friend dear
‘I haven’t met any dear friend for two years (now).’
The second possibility is to use bare plurals, which are allowed with a much
larger class of predicates than BSs:
(46) a. N-am citit romane.
neg-have.1sg read novels
‘I haven’t read novels.’
b. De doi ani n-am întâlnit prieteni.
of two years neg-have.1sg met friends
‘I haven’t met friends for two years (now)’
These examples indicate that the difference between incorporating and non-
incorporating verbs is maintained under negation.
Note now that in examples uttered without any special stress,13 DPs headed by
overt Determiners and in particular singular count indefinites, show a clear prefer-
ence for the wide scope interpretation:
(47) a. N-am citit un roman.
neg-have.1sg read a novel
Only possible reading: ‘There is a novel which I haven’t read.’
12
The use of niciun ‘neither.one.masc’ with verbs allowing BSs is marked:
(i) N-am nicio maşină.
neg-have.1sg neither.one car.
‘I don’t have any car.’
Nu caut nicio secretară.
neg seek.1sg neither.one secretary
‘I don’t look for any secretary.’
Examples like (i)–(ii) are used when the speaker negates an explicit or implicit assertion, e.g., ‘You
have a car/you look for a secretary’.
13
Parts of the indefinite can be interpreted in the scope of the negation provided that they are con-
trastively stressed. The alternative must usually be explicitly asserted. When it is the indefinite
article that is contrastively stressed, it is interpreted as the numeral ‘one’.
(i) a. N-am citit UN roman de Tolstoi, (ci două).
‘I haven’t read ONE novel by Tolstoy (but two).’
b. N-am citit un ROMAN de Tolstoi, (ci o nuvelă).
‘I haven’t read a NOVEL by Tolstoy (but a short story).’
‘Not-have.1.sg. read a novel by Tolstoy, (but a short story).’
b. N-am citit un ROMAN 14

neg-have.1sg read a NOVEL
‘I haven’t read a single novel.’
Turning now to incorporating verbs, the use of BSs becomes quasi-obligatory in
a negative context.
(48) a. ?? Ion nu şi-a cumpărat un apartament.
Ion neg se.dat- has bought a apartment
a¢. Ion nu şi- a cumpărat apartament.
Ion neg se.dat- has bought apartment
‘Ion didn’t buy an apartment.’
b. * Azi Ion nu poartă un pantalon
today Ion neg wears a pant
b¢. Azi Ion nu poartă pantalon.
today Ion neg wears pant
c. * Ion nu caută o secretară.
Ion neg seeks a secretary
c¢. Ion nu caută secretară.
Ion neg seeks secretary
Whenever a specific or a partitive reading is intended, the indefinite article must
be used. Thus, for houses that are closed with more than one key, one may say:
(49) Nu pot intra fiindcă nu am o cheie.
neg can.1sg enter because neg have.1sg a key
‘I cannot enter because I do not have a key’
2.3.4 Conclusions
In sum, using a bare noun, either singular or plural, is a way of indicating the nar-
rowest possible scope, in particular narrow scope with respect to negation. As we
will see in Chaps. 6 and 8 the obligatory narrow scope interpretation of bare NPs is
due to the fact that they are necessarily ‘weak’ (in the sense of Milsark 1977).
The difference in distribution between BSs and BPs indicates that BSs cannot be
viewed as the singular counterparts of BPs, a conclusion that is supported by the
observation that BSs have a number-neutral interpretation, i.e., they do not neces-
sarily refer to atomic entities but allow both atoms and sums in their denotation
(in other words, BSs are true of either atomic or plural entities, depending on the context).
14
Capitals indicate contrastive stress.
The number-neutral interpretation of BSs can be assumed to derive from the fact
that their syntactic representation lacks Number (Farkas and de Swart 2003;
Dobrovie-Sorin et al. 2005, 2006).
2.4 The Semantics of Bare Plurals
Recognizing that a distinction in distribution calls for a distinction in semantic

type and/or LF representation, we need to assign distinct analyses to singular
indefinites, existential15 bare plurals (BPs) and count bare singulars (BSs).
Granting that singular indefinites are entity-referring expressions (type e), we
need to propose a differentiating analysis for the two types of bare NPs. Since BPs
have been investigated more than BSs, we will concentrate on BPs. Although
most of the current analyses were intended to account only for BPs, we will check
whether they can also account for BSs. We will first show that Carlson’s (1977a,
c) analysis based on the hypothesis that BPs are names of kinds cannot be extended
to Romance languages, because in these languages bare NPs, regardless of their
number-marking, cannot have generic readings (Dobrovie-Sorin and Laca 1996,
2003). Turning next to the hypothesis that bare NPs denote properties,16 we show
that property-denotation is a plausible analysis for BSs but not for BPs. We will
therefore be led to revive the more traditional view that BPs are the plural coun-
terparts of non specific singular indefinites. Both types of DPs (i.e., singular
indefinites and BPs viewed as plural indefinites) can be analyzed as entity-
referring expressions, which nevertheless differ as to whether they refer to indi-
viduated entities, i.e., entities that are distinct from each other or to amounts/
quantities, i.e., non-individuated entities that are ordered by the part-whole
relation. Building on this denotational difference (Dobrovie-Sorin 2007), our pro-
posal will be that BPs should be analyzed as generalized quantifiers over amounts.
In sum, we assume a distinction between two types of entities (i.e., individuals
and amounts) and correlated to it a difference between two types of generalized
quantifiers, over individuals and over amounts.
15
We are not interested here in kind-referring BPs, which are found in English but not in Romance
languages (other than Brazilian Portuguese). As argued in Sect. 2.4.1, the existential readings of
Romance BPs cannot be derived from kind-reference.
16
This hypothesis is adopted by most of the theoreticians that have examined both types of bare
nouns (see in particular van Geenhoven 1996; Kallulli 1999; Carlson 2003; Chung and Ladusaw
2003; Dobrovie-Sorin et al. 2005, 2006). Dayal (2003) is one of the few authors who assumes a
type distinction between bare singulars and bare plurals: existential bare singulars denote proper-
ties, whereas bare plurals (both existential and generic) rely on kind-reference (as in Carlson
1977a, c; Chierchia 1998). This analysis cannot be adopted here, given that bare plurals in Romance
languages cannot denote kinds.
2.4 The Semantics of Bare Plurals 51
2.4.1 Bare Plurals and Reference to Kinds
2.4.1.1 The Carlsonian Analysis
In addition to existential readings, illustrated in (50), English BPs can also have
generic readings, illustrated in (51).
(50) a. Students are dancing in the street.
b. John is eating apples.
(51) a. Gorillas are on the verge of extinction.
b. Cats are intelligent.
c. John loves cats.
Carlson (1977a, c) proposed a unified analysis of English BPs, according to
which they are to be analyzed as names of kinds. On this view, the generic and exis-
tential readings of BPs are in complementary distribution, the choice between one
or the other reading depending on the context: generic readings arise when BPs
combine with i(ndividual)-level predicates, whereas existential readings are trig-
gered by s(tage)-level predicates.
The generic readings of BPs are directly explained by Carlson’s hypothesis
according to which BPs denote kinds, i.e., entities (type e expressions) whose ref-
erence is not fixed with respect to a time and a place.
In contrast, the existential reading of BPs depends on stages of kinds, resulting
from the use of kind denoting expressions in contexts that make reference to
particular events. According to Chierchia’s (1998) revised version of Carlson’s
analysis, the existential readings of BPs result from the application of a type-shifting
operation dubbed ‘Derived Kind Predication’ defined as in (52):
(52) Derived Kind Predication (DKP): (Chierchia 1998)
If P applies to objects and k denotes a kind, then
P(k) = $ x [ x £ È k Ù P (x) ]
As stated in (52), the DKP applies when bare plurals (which by assumption
denote kinds) appear as arguments of s-level predicates that apply to ordinary
objects, rather than kinds. The DKP introduces an existential quantifier over
instances of the kind. The DKP applies in an example such as (53), since barking is
a property of ordinary objects, not kinds. In (53), Çdogs notates ‘the kind dogs’,
obtained by applying the Down operator to the property dogs.:
(53) Dogs are barking.
barking (Ç dogs) Û (via DKP) $ x [ x £ È Ç dogs Ù barking (x) ]
A Carlsonian analysis adequately captures the generic reading of BPs (cf. Chap. 7)
but is problematic for the existential reading of BPs. This is particularly clear if we
try to give a unified analysis of the existential reading of bare nouns in English and
Romance languages.
2.4.1.2 Bare Plurals in Romance Languages Are Not Kind-Referring
Carlson’s (1977a, c) analysis cannot extend to Romance languages. Indeed, BPs in

Spanish, Italian and Romanian are not compatible with kind predicates such as be
on the verge of extinction17:
(54) a. În România sunt pe cale de dispariţie urşii /*urşi.
in Romania are on way to disappearance bears-the bears
‘In Romania bears are becoming extinct.’
b. O lege din 1950 protejează urşii /*urşi.
A law from 1950 protects bears-the bears
‘A law from 1950 protects bears.’
More generally, BPs do not yield generic readings when they combine with indi-
vidual-level predicates18:
(55) Lui Ion îi plac prăjiturile /*prăjituri
dat Ion him.dat appeal cakes-the cakes
‘Ion likes cakes.’
(56) Ion respectă profesorii /* profesori
Ion respects professors-the professors
‘Ion respects professors.’
In Spanish or Romanian, kind reference cannot be assumed for BSs either. Let
us indeed recall that BSs are allowed with a very restricted number of predicates
(see Sect. 2.3.1 above). Since kind-predicates and i-level predicates are not among
those predicates that can combine with BSs, kind-reference cannot be assumed for
this type of bare NP either.
We are thus led to conclude that in Romance languages, the existential readings
of bare NPs cannot be assumed to be derived from kind-denoting bare NPs.
2.4.2 Bare Plurals and Property Denotation
Let us now consider an alternative analysis, according to which existential bare

nominals denote properties (type < e,t>) even when they occur in argument
positions (McNally 1995a, b, 1998; van Geenhoven 1996; Dobrovie-Sorin 1997a, b;
Dobrovie-Sorin and Laca 1999; Dobrovie-Sorin and Beyssade 2004). This hypothesis
17
Examples with postverbal subjects (see (54)–(55)) or postverbal objects (see (56)) are used here
in order to avoid interference with another constraint, according to which unmodified bare nouns
cannot appear in preverbal position in Spanish, Italian and Romanian.
18
Chierchia (1998) observed that the generic reading of modified bare NPs is possible in a restricted
class of examples and he concluded that bare NPs denote kinds in Italian. This assumption was
criticized by Longobardi (2002) and Delfitto (2002).
can be implemented by assuming that verbal predicates can be represented not

only in the canonical way, as unsaturated expressions waiting for entity-denoting
constituents, but also as expressions that can be saturated by property-denoting
constituents.
2.4.2.1 Existential Predicates
Let us first illustrate the canonical type of semantic composition, by considering

examples such as (57), represented as shown in (57¢):
(57) a. John is handsome.
b. John admires Mary.
(57¢) a. lx.x is handsome (John)
b. lx.ly.x admire y (Mary) (John)
This type of predicate cannot apply to properties: the rules of semantic composi-
tion do not allow us to derive a truth value by combining two predicates. If we
assume that BPs in Romance languages denote properties, the ungrammaticality of
examples of the type in (58) is explained as being due to the fact that a property-
denoting expression cannot saturate predicates such as admire, which is represented
as in (57¢).
(58) * Ion admiră profesori.
Ion admires professors
According to van Geenhoven (1996), Dobrovie-Sorin (1997a, b) and McNally
(1998), certain predicates, call them ‘existential’, can be represented not only as
in (57¢) but also as shown in (59), where an existential operator binds some of the
verb’s argument variables. This kind of predicate is saturated in the lexicon by
existential closure and consequently cannot apply to entity-denoting expressions.
Variables bound by existential closure can range over domains that are either
non-restricted (cf. the implicit arguments of transitive verbs such as wash, eat,
etc.) or restricted by predicate variables notated P or Q in (59) below. Existential
predicates are thus represented by means of lambda abstraction over predicate
variables:
(59) a. lP lQ $x $y [x wash y Ù Q(x) Ù P(y)]
b. lP $x [x sleep Ù P(x)]
c. lP $x [x is available Ù P(x)]
According to the representations in (59), the predicates wash or sleep remain
unsaturated: they need to combine with properties that will saturate the predicate
positions represented as P and Q, which restrict the domain of argument
variables.
Summarizing, predicates can be represented in two different ways, either as

‘entity predicates’, i.e., lambda-abstracts over entities, or as existential predicates,
which have an existential quantifier in their lexical representation and function as
lambda-abstracts over the properties that restrict the range of their argument
variables. All predicates can be represented as entity predicates but only certain
predicates, e.g., wash, dance, sing, etc. but not intelligent or sad, can be represented
as existential predicates. The empirical criteria that allow us to distinguish between
those predicates that necessarily function as entity predicates and those that allow
both analyses will be examined in Sect. 2.5.
Given existential predicates of the type in (59), the examples in (60) can be
represented as in (60¢), which yield the representations in (60″) via lambda-
conversion19:
(60) a. Women were washing shirts in the garden.
b. Children were sleeping on the bed.
c. Books are available.
(60¢) a. lP lQ $x $y [x wash y Ù Q(x) Ù P(y)] (shirts) (women)
b. lP $x [x sleep Ù P(x)] (children)
c. lP $x [x are available Ù P(x)] (books)
(60″) a. $x $y [x wash y Ù women (x) Ù shirts (y)]
b. $x [x sleep Ù children (x)]
c. $x [x are available Ù books (x)]
The BPs women, children and books in (60¢) and (60″) have not been repre-
sented as entities (of type e) but as properties (of type < e,t>). The sentence has
an existential interpretation due to the presence of an existential quantifier in the
representation of the predicate.
2.4.2.2 Accounting for Carlson’s Observations Regarding Scope
Let us now briefly show that the property analysis of bare NPs can account for the
contrasts observed by Carlson between indefinite DPs and BPs regarding scope
19
On a classical analysis, we distinguish three types of conversion for terms of the lambda
calculus.
a-conversion : replace lx A with ly A[y/x] if y does not appear in A.
b-conversion : replace (lx A) t with A[t/x]
h-conversion : replace A with (lx Ax) if A is of type a → b and x is of type a and if there is no
instance of x in A.
These rules are all called lambda-conversion rules and by definition, two terms t and t¢ are said to
be lambda-equivalents if there is a sequence of lambda-conversions that allows reducing t to t¢ or
vice versa.
(see Sect. 2.2). Let us consider examples of the type in (14), repeated here as (61),
where meet is embedded under a propositional attitude verb:
(61) a. Minnie wishes to meet a young psychiatrist.
b. Minnie wishes to meet young psychiatrists.
Singular indefinites are analyzable as individual variables20 bound by an existen-
tial quantifier introduced by existential closure. Since existential closure can apply
at different levels (Heim (1982) and subsequent literature), (61a) is ambiguous,
allowing both an opaque (the existential operator is inserted below wish) and a
transparent reading of the indefinite (the existential operator is inserted above wish).
BPs, on the other hand, can only be legitimated by an existential quantifier present
in the lexical representation of the predicate with which they combine: in (61b),
the existential quantifier is introduced by the verb meet (which is analyzed as an
existential predicate, on a par with wash in (60’)) and therefore it has narrow scope
with respect to the verb wish, thus yielding an opaque reading.
This explanation for the obligatory narrow scope of BPs extends to other scopal
elements, e.g., quantified DPs or temporal adverbs, which appear above the main
verb that supplies the existential quantifier that legitimates the BP.
As we can see, in examples such as (61a–b) the same predicate (meet) receives
two different analyses: in order to account for the transparent reading of (61a), we
must assume that meet translates as a lambda-abstract over entities and in order to
account for the opaque reading of (61b), meet must be represented as an existential
predicate, which combines with an argument of type < e,t > (i.e., the bare plural
young psychiatrists).21
2.4.2.3 Problems
The property analysis of BPs is confronted with several problems. The first one is
related to the observation that those expressions that are clearly property-denoting,
e.g., adjectives, cannot occur in argument positions:
(62) a. *Ion mănîncă crud.
Ion eats raw
b. *Maria pictează roşu.
Maria paints red
20
Carlson (1977a, c) translates the singular indefinite article as an existential quantifier. We have
reformulated his analysis in the DRT framework because this facilitates the comparison with more
recent analyses of the same data.
21
The narrow scope reading of the indefinite DP in (61a) can be analyzed either as (i) relying on
the same representation as that of the BP in (61b), i.e., meet translates as an existential predicate
and the indefinite DP denotes a property or (ii) the indefinite DP translates as a variable over enti-
ties and an existential operator is supplied below wish. The choice between these options is not
relevant for the present discussion.
To answer this objection one may invoke a subcategorization constraint (e.g., the
syntactic category of adjectives is disallowed in argument positions) correlated with
a semantic distinction between two types of properties, those that are allowed in
argument positions and those that are not.
The second argument against the hypothesis that BPs denote properties comes
from French, which allows bare plurals in predicate positions (see (63) below) but
not in argument positions, where des-phrases must be used instead (see (64)):
(63) Jean et Jeanne sont professeurs.
Jean and Jeanne are professors
‘Jean and Jeanne are professors.’
In examples of the type in (63), the BP denotes a property. However, French does
not allow BPs in argument positions, as shown in (64a). Examples of this type
become grammatical if we replace the BPs by des-indefinites, which are the closest
French counterparts of BPs in the other Romance languages:
(64) a. *J’ai rencontré professeurs.
I have met professors
b. J’ai rencontré des professeurs.
I have met des professors
‘I have met professors’
The contrast between predicate and argument positions as to the possibility of
BPs in French suggests that property-denoting expressions cannot occur in argu-
ment positions.
2.4.2.4 The Property Analysis of Count Bare Singulars
The third argument against analyzing BPs as property-denoting expressions is

related to the observation made in Sect. 2.3 above, that in languages such as Spanish
or Romanian, count BSs occurring in argument positions are much more constrained
than BPs. Such a contrasting distribution calls for two distinct semantic analyses.
Property-denotation seems appropriate for count BSs in argument positions, insofar
as it might explain their highly constrained distribution: precisely because they are
property-denoting, they cannot occur in run-of-the-mill argument positions; even in
those contexts in which they do appear, they are not genuine arguments but instead
function – at some level of representation – as predicate-modifiers that combine
with the verbal predicate and yield complex predicates.
Dayal’s (2003) rule of Pseudo-incorporation, shown in (65), is a rule of predicate
modification22 based on the idea that certain transitive verbs can be represented as
‘incorporating predicates’:
22
Farkas and de Swart’s (2003) rule of Unification of thematic arguments and Chung and Ladusaw’s
(2004) rule of Restrict are different implementations of the same type of analysis.
(65) a. lx ly le [V(e) Ù Ag (e) = y Ù Th(e) = x]

b. lPly le [P-V (e) Ù Ag (e) = y Ù Appropriately Classificatory (e)]
(65a) represents a transitive verb and (65b) represents the incorporating version
of the same verb, which is obtained by replacing the Theme argument with a place-
holder for a predicate-modifier notated P. (65a) denotes a relation between individu-
als, whereas (65b) denotes a relation between individuals and properties, since
according to Dayal “the property argument does not correspond to a Theme but is
instead interpreted as a modification of the verb”. Given (65b), the object position
can be filled by property-denoting nominals and in particular by a bare singular. The
restriction to ‘appropriately classificatory’ events (cf. Dowty 1979) is meant to
account for the fact that V + bare singular sequences must refer to types of events
that are culturally stable.23
What (65b) does not make clear is the sort of relationship held between the ver-
bal predicate V and the property P. Because of this, we adopt the slightly different
predicate modification rule given in (65c), proposed in Dobrovie-Sorin et al. (2006),
which differs from Dayal’s rule in that the property denoted by the bare nominal is
predicated of the Theme of the verb24:
(65) c. lPly le [V(e) Ù Ag(e) = y Ù Th(e) has P Ù Approp. Classificatory (e)]
We assume that the rule of predicate modification applies only when two lexi-
cal constituents, with no functional category intervening between them: the verb,
which is a lexical head (V°), combines with a bare singular, which can be either a
lexical head (N°) or a phrasal constituent (NP). Compare bare plurals, which have
(at least) the functional category of Number and as such – given the constraint
suggested here – cannot combine with the predicate by predicate modification. In
sum, property denotation seems appropriate for BSs in argument positions. But
then, if we want to keep BPs distinct from BSs, we need another semantic analysis
for BPs.
The contrast described here between BPs and BSs in Romance languages is para-
digmatic of similar contrasts found in other languages (see the contrast described by
Dayal (2003) between pseudo-incorporated and non incorporated bare NPs in
Hindi) or crosslinguistically. Thus, according to Chung and Ladusaw’s (2003)
description, the Maori indefinites headed by te are full DPs,25 whereas the Chamorro
23
See also Carlson (2003), according to whom the semantic composition of weak bare nominals
deals with types of things and types of events (rather than tokens): ‘there are no times, no possible
worlds, no truth, only types’.
24
For a more precise definition of predicate-modification the reader is referred to Espinal and
McNally (2011), where a refined characterization of the predicates that allow it can also be
found.
25
Although they occupy different argument positions, both te-indefinites in Maori and incorpo-
rated objects in Chamorro are weak (in the sense of Milsark (1977)), in particular they take obliga-
tory narrow scope with respect to negation.
incorporated objects are bare NPs that can only appear with have and with existential
predicates. Because the Chamorro incorporated objects can be doubled by an inde-
pendent nominal expression (Gäi-ga’ yu’ kätu ‘have-pet I cat’ “I have a cat”) they
cannot be assumed to combine with a main predicate with a built-in existential
quantifier (as in Carlson (1977a, b) or van Geenhoven (1996)). Chung and Ladusaw
(2003) therefore propose a semantic analysis in which the rule of predicate-
modification (labelled Restrict) is factored out from the rule of existential closure,
which applies after Restrict and is parametrized: whereas the rule of existential
closure is optional in Chamorro, it is obligatory in Maori, and therefore the argu-
ment position is saturated, thus rendering impossible the doubling of te-indefinites.
Although Romance BSs cannot be doubled, they resemble Chamorro incorpo-
rated objects insofar as they are genuinely bare NPs with a highly restricted distribu-
tion, which led us to assume that they are property-denoting expressions that
compose with the main predicate via Dayal’s (2003) rule of Pseudo-incorporation,
which is a rule of predicate-modification comparable to Restrict. We can use the
parametrization of existential closure proposed by Chung and Ladusaw (2003) in
order to explain why, unlike Chamorro incorporated objects, BSs in Romance lan-
guages cannot be doubled by a DP with the same theta-role.
Turning now to Maori te-indefinites, we would like to suggest that their semantic
composition is radically different from that of Chamorro incorporated objects: they
are not property-denoting expressions (recall that they are DPs rather than genuine
bare NPs) and correlatively they cannot compose with the main verb via predicate
modification. We may instead assume that the semantic composition of Maori te-
indefinites is comparable to the semantic composition of Romance BPs, which we
are going to clarify in the following sections.
2.4.3 Bare Plurals and VP-level Existential Closure
2.4.3.1 VP-Level Existential Closure and Scope
Diesing’s (1992) analysis of BPs in terms of VP-level existential closure preserves

Carlson’s (1977a, c) insight that the existential readings of BPs crucially depend on
an existential quantifier that is supplied by their context. But instead of assuming
that the existential quantifier appears in the lexical representation of the predicate
(as proposed by Carlson (1977a) and later by van Geenhoven (1998)), Diesing pro-
poses a syntactic implementation based on a rule of VP-level existential closure:
(66) Attach an existential quantifier at the left edge of the VP-domain.26
26
Diesing (1992) and Kratzer (1995) correlate VP-level existential closure to the closure of the
nuclear scope proposed by Heim (1982).
The variables supplied by BPs can only be legitimated by VP level existential

closure, which explains their obligatory narrow scope: since the variables supplied
by BPs are bound by VP-level existential closure and since at LF any quantifier
takes scope over the VP (as a result of Quantifier Raising or of Quantifying in), BPs
will take narrow scope relative to any co-occurring quantifier.
Note also that the Carlsonian analysis cannot easily explain why, in those
languages that allow postverbal subjects (e.g., all Romance languages other than
French), the existential readings of BPs are blocked in the preverbal subject posi-
tion. Under Diesing’s proposal this generalization is accounted for: preverbal
subjects are above the VP and as such they are out of the reach of existential
closure.
2.4.3.2 VP-Level Existential Closure and Aspect
Another important observation made by Carlson (1977a, c) was that the existential
reading of bare plurals is allowed – even forced – by certain predicates and blocked
by others:
(67) a. Children were dancing in the street. (OK existential, * generic)
b. Doctors are intelligent. (* existential, OK generic)
Within Carlson’s own analysis, the grammaticality or the ungrammaticality of
(67) is explained as depending on the possibility or impossibility of introducing an
existential quantifier in the lexical representations of the predicates appearing in
these examples.
This type of account is not sufficient, because the existential readings of BPs
cannot be fully explained on the basis of the lexical properties27 of the predicates
with which they combine. Thus, most of – maybe all – those predicates that allow
existential readings for BPs also allow generic readings, depending on the tense and
aspect of the predicate:
(68) a. Students were dancing in the street.
b. Students dance rock’n roll.
Given pairs of this type, the existential quantifier cannot be introduced in the
lexical representation of the predicate but rather at some later stage in the derivation
of a given sentence, after Tense and/or Aspect have attached to the predicate. Under
Diesing’s analysis, one can assume that VP-level existential closure is allowed or
blocked by certain choices of Tense or Aspect.
27
Note that the relevant lexical classes of predicates are difficult to characterize (see Sect. 2.5
below).
2.4.3.3 Problems with Generic Objects
Generic objects, illustrated in (69), are problematic for Diesing:

(69) a. Maria loves coffee.
b. Juan hates lawyers.
In these examples, the BPs take generic reading despite the fact that they occur
in object positions and as such they should be caught by VP-level existential
closure. Diesing’s proposal is that emotional predicates such as hate, love, etc. force
their objects to scramble out of the VP at Logical Form, which puts them outside the
reach of existential closure.
It should be clear that Diesing’s analysis in terms of scrambling may be an ade-
quate explanation for why the existential readings of the objects of love/hate verbs
are blocked but it does not explain why their generic readings are allowed. Indeed,
Diesing’s analysis incorrectly predicts that bare mass NPs and BPs take generic
readings in Romance languages other than French:
(70) a. *María adora café
Maria loves coffee
b. *Juan detesta manzanas
Juan hates apples
The contrast between (69a–b) and (70a–b) can be explained as being due to the
fact that bare NPs in Romance languages (other than Brazilian Portuguese) are nec-
essarily ‘weak indefinites’, which as such must be legitimated by VP-level existen-
tial closure, whereas English bare NPs may also function as names of kinds.
To conclude, Diesing’s account is confronted with various problems, the most
important in the present context being that the rule of VP-level existential closure is
not compositional: the existential quantifier is neither introduced at the lexical level of
representation, nor is it strictly correlated with the application of a particular syntactic
rule. Also, in order for the rule of VP-level existential closure to do the work it is
needed for, Diesing must assume lowering rules, which we would like to avoid.
2.4.4 Bare Plurals as Amount-Referring Expressions
In this section we will present our own view regarding the semantic analysis of BPs.
Building on the denotational difference between count singular predicates on the
one hand and plural and mass predicates on the other hand, we will propose that
bare BPs (and bare mass NPs) can be analyzed neither as e-type expressions nor as
properties but rather as generalized quantifiers over amounts. This proposal pre-
serves Carlson’s intuition that BPs are legitimate only if an existential quantifier can
be supplied by the context.
2.4.4.1 Individuals vs. Amounts
The difference between count singular nouns (cat) on the one hand and plural nouns
(cats) and mass nouns (coffee) on the other hand can be characterized as follows:
count singular nouns describe individualized objects, viewed as ‘integrated wholes’
(Simons 1987; Moltmann 1997, 1998) whereas count plural predicates refer to
amounts/quantities of objects, on a par with mass nouns, which refer to amounts/
quantities of substance. Predicates that denote sets of integrated wholes are neither
cumulative nor divisive whereas predicates that denote sets of amounts are cumula-
tive and divisive.
In somewhat more technical terms, the denotation of a count singular noun is a
set of individuals, i.e., objects that cannot be ordered by the part-whole relation. All
the elements of this set are distinct from each other and do not overlap. The denotation
of plural and mass nouns, on the other hand, is a set of objects that has the algebraic
structure of a join semi-lattice: the objects of this set are amounts, which overlap
and are ordered by the part-whole relation.
Summarizing, we distinguish between two types of entities: individualized enti-
ties (for which we will use the label ‘individuals’ henceforth) and amounts.
The domain of denotation of amounts (of substance or of objects) is a set that
is structured by the part-whole relation, which explains why mass predicates, as
well as plural predicates, are both cumulative (if A is coffee and B is coffee,
then the sum of A and B is coffee; if A is books and B is books, then the sum
of A and B is books) and divisive (if A is part of an amount of coffee B, then A
is an amount of coffee; if A is part of an amount B of dogs, then A is an amount
of dogs).
It should be clear that some indefinites, e.g., singular indefinites such as a boy,
or cardinal indefinites, e.g., two students, are ambiguous: they can refer either to an
(singular or plural) individual or to an amount of objects (an amount of cardinality 1
for a boy and an amount of cardinality 2 for two boys). When they refer to individuals,
such indefinites are interpreted as specific indefinites.
2.4.4.2 Bare Plurals as Existential Generalized Quantifiers over Amounts
Following Carlson (1977a, c) and all subsequent proposals, we will assume that the
existential readings of BPs depend on an existential predicate. According to us, an
existential predicate is necessary because existential BPs (as well as bare mass NPs)
refer to non specific amounts and the existence of non specific amounts cannot be
presupposed; their existence must be asserted, hence the necessity to combine with
an existential predicate.
We depart from previous proposals regarding the denotation of existential BPs:
they are neither names of kinds (as proposed by Carlson (1977a, c)) nor properties
(as proposed by van Geenhoven (1996) among many others) but rather generalized
quantifiers over amounts. To illustrate our analysis, let us consider the following
examples:
(71) a. Children were sleeping.
b. Smoke is coming out of the chimney.
The representation in (72g) resembles previous proposals insofar as it assumes
an existential quantifier in the representation of main predicates (see the representa-
tion of sleep in (72b)) but is nevertheless different insofar as the BP is represented
as a generalized quantifier of the type shown in (72a), which itself contains an exis-
tential quantifier:
(72) a. children = lP $ xa (children(xa) Ù P(xa))
b. be sleeping = lxa $ xi (be-sleeping(xi) Ù Ind(xi, xa))
c. Children are sleeping = children (be-sleeping)
d. lP $ xa (children(xa) Ù P(xa)) • lxa $ xi (be-sleeping(xi) Ù Ind(xi, xa))
e. $ xa (children(xa) Ù [lxa $ xi (be-sleeping(xi) Ù Ind(xi, xa))] (xa))
f. $ xa (children(xa) Ù $ xi (be-sleeping (xi) Ù Ind(xi, xa)))
g. $ xa $ xi (children(xa) Ù be-sleeping(xi) Ù Ind(xi, xa))
In (72c), the two constituents are semantically composed by applying children
(viewed as denoting the generalized quantifier in (72a)) to be-sleeping. In (72d) we
have replaced the two constituents by their representations in (72a) and (72b). In
(72e) and (72f) we have applied lambda-conversion twice. In (72g) we have pre-
posed the existential quantifier that binds the individual variable.
Let us stress that according to this proposal, examples built with existential BPs
rely on LFs that contain two (rather than just one) existential quantifiers, which
respectively appear in the representation of the BP and of the main predicate: the
existential quantifier inside the generalized quantifier asserts the existence of an
amount that instantiates the property denoted by the noun (children), whereas the
existential quantifier in the representation of the main predicate asserts the existence
of a participant to the event (of sleeping); the variable over amounts of children and
the variable over participants to the event of sleeping are related by a relation
of individuation notated Ind, by virtue of which amounts of objects (in this case
children) are specified as being identified28 with the individuals that are the Theme
participants to the sleeping event.
Note that the contrasts between bare plurals and singular indefinites observed by
Carlson (1977a, c) and reviewed in Sect. 2.2 above can be explained as a conse-
quence of the fact that BPs must combine with existential predicates. And since
existential predicates are ingredients of all of the accounts of BPs reviewed
here, these accounts cannot be differentiated on the basis of the contrast between
28
Let us stress that Ind is not an identity relation (in which case the two terms of the identity should
be interchangeable) but rather an oriented identification relation, which allows an amount, i.e., a
not yet individualized entity, to be identified with an individual.
2.5 Existential Predicates and Entity Predicates 63
BPs and singular indefinites. The competing analyses of BPs can nevertheless
be evaluated based on other considerations. Carlson’s account in terms of kind-
referring BPs cannot be extended to Romance languages. Diesing’s account is not
compositional. Finally, the property analysis of BPs is both conceptually proble-
matic insofar as it runs against a fundamental homomorphism principle, according
to which argumental DPs denote either entities (type e) or generalized quantifiers
(type < <e,t>,t>) and empirically problematic insofar as it does not explain why
adjectives, which are property-denoting expressions, cannot appear in argument
positions. The generalized quantifier analysis of BPs proposed here obeys the homo-
morpism principle and is moreover able to capture the constraint against adjectives:
qua property denoting expressions, adjectives cannot occur in argument positions;
bare NPs occurring in argument positions do not denote properties but rather
generalized quantifiers over amounts, i.e., quantifiers that assert the existence of an
instantiation of a property.
2.5 Existential Predicates and Entity Predicates
We have previously argued (see Sects. 2.4.2 and 2.4.4) that the existential reading
of BPs is allowed only with certain predicates, dubbed ‘existential’, which contain
an existential quantifier in their representation. Let us now try to define the empiri-
cal criterion that allows us to distinguish between existential predicates and entity
predicates, which can only translate as lambda-abstracts over entities. In what fol-
lows, we will first show that the relevant distinction cannot be the one between
stage-level and individual-level predicates (Carlson 1977a, c; Kratzer 1988). We
will then suggest a different criterion, which relies on spatial localization.
2.5.1 Individual-Level and Stage-Level Predicates
On Carlson’s (1977a, c) approach, English bare nouns are analyzable as names

of kinds and the predicates responsible for their existential interpretation play a
double role: on the one hand, they introduce an existential quantifier and on the
other hand, they allow the denotation of the BNP to be shifted from a kind to the
stages that instantiate the kind. Given this analysis, the relevant predicates
have to be both “existential” and “transitory”, hence the hypothesis that the
existential quantifier is related to the transitory property of the predicate. This
hypothesis is crucial for Carlson but not for us: on the analysis adopted here, the
existential interpretation of bare nouns does not depend on “stages of individuals”
or “stages of kinds”. The only property of the predicate that is crucial for our
analysis is the fact that it allows one of its argument positions to be bound by an
existential quantifier.
The correlation between transitory predication and existential quantification is

therefore not necessary and hence we do not expect to find it in all contexts. Some
authors (Kiss 1994; Dobrovie-Sorin 1995) have observed that English has transitory
predicates, e.g., tired or sad, which do not allow existential readings of bare nouns
in argument position (cf (73)). Example (74), from Kiss (1994), shows that line is a
non transitory predicate that nevertheless allows an existential interpretation of a
bare NP in subject position.
(73) * Students were tired.
(74) Volcanoes line both sides of the river.
The same observations hold in Spanish, Italian and Romanian. Bare nouns,
which in these languages only allow an existential reading, are not always possible
as arguments of transitory predicates, nor are they always excluded as arguments of
permanent predicates. To give an example, the adjectives built with the copula estar
in Spanish express transitory predications but most of these predicates cannot com-
bine with bare nouns (cf. (75)).
(75) ??Estaban tristes/ inquietos/ enfermos medicos. Spanish
were sad worried sick doctors
‘Doctors were sad/worried/ sick.’
Only a subset of the adjectives that require estar, namely those that do not express
qualities of the subject but rather existence at a certain location, allow bare nouns as
arguments:
(76) Estaban presentes/ disponibles médicos. Spanish
were present available doctors
‘Doctors were present/ available.’
Conversely, even if the predicates in examples (77a–b) express stable situations,
they are compatible with bare nouns in argument position:
(77) a. Bordeaban cipreses el camino. Spanish
line.impf cypresses the road
‘Cypresses were lining the road.’
b. Este manuscrito contiene errores.
this manuscript contains errors
‘This manuscript contains errors.’
The correlation between transitory predication and existential readings is not
observed in French either. Kleiber (2001:55) argues that the temporal criterion can-
not explain data of the type in (78):
(78) a. ? Un avion est gris.
‘A plane is grey.’
b. ? Un médecin est disponible/triste.
‘A doctor is available/sad.’
These examples exhibit the same grammaticality judgments, despite the fact that
their main predicates differ regarding the temporal criterion: example (78a) is built
with an individual-level predicate (which expresses a permanent or stable property)
and (78b) is built with an s-level predicate (which expresses a transitory property).
In sum, there is no perfect correlation between transitory predicates and the so-
called existential predicates, which legitimate the existential reading of BNPs (as
well as the weak existential reading of indefinite DPs). We therefore need another
empirical criterion to identify this class of predicates.
2.5.2 Space Localization
According to Dobrovie-Sorin (1997a, b), what makes a predicate existential is the

possibility of localizing its arguments in space.29 It can indeed be observed that
those predicates that allow the existential reading of bare nouns (see (76)) also allow
space adverbials (see (79b)), whereas those that exclude existential bare nouns (see
(75a–b)) also exclude space adverbials (see (79a)):
(79) a. ??¿ Dónde estaba triste/ inquieto/ enfermo Juan? Spanish
where was sad worried sick Juan
b. ¿ Dónde estaba presente Juan?
where was present Juan
‘Where was Juan present?’
The same kind of correlation can be observed in French:
(80) a. * Du beurre était rance.
du butter was rancid
b. Du beurre était en train de fondre dans une assiette.
du butter was prog to melt in a plate
‘There was butter melting on a plate.’
According to the space location criterion, most i-level predicates qualify as non-
existential (as in (81)), because these predicates disallow not only the temporal
but also the spatial localization of their argument. For these predicates, the spatial
location criterion does just as well as an analysis in terms of a contrast between
permanent and episodic properties.
29
The use of spatial location as a criterion of identifying predicates that allow existential or weak
readings of indefinites was independently proposed by McNally (1995a, b, 1998), Glasbey (1998)
and Kleiber (2001).
(81) *Jean est intelligent/grand/beau dans sa chambre.

John is intelligent tall handsome in his room.
But the spatial location criterion can also account for those cases that
are problematic for the criterion based on the i-level versus s-level distinction.
We may thus suggest that permanent predicates such as line in (74) are compatible
with an existential reading because the direct object acts as the spatial localizer
of the subject. Being located in space, the subject can have an existential
reading.
The spatial location hypothesis also accounts for those predicates that do not
allow existential readings of BPs, although expressing a transitory property:
(82) a. ??Contrabassoonists were cheerful.
b. ??Peasants were angry.
c. ??During Chomsky’s lecture, students were asleep/drunk/hungry/tired.
d. ??When we arrived, students were asleep/drunk/hungry.
e. ??Look! students are drunk/ hungry in the street.
(83) *Du beurre était frais/ liquide/ mou
du butter was fresh/ liquid/ soft
The examples in (82)30 and (83) contain s-level predicates and yet they do not
allow existential readings. Applying the spatial location criterion, we again obtain
the correct result: the predicates in (82) and (83) do not locate their subject argu-
ment in space.
Finally, Dobrovie-Sorin (1997a, b) proposed that the spatial location criterion
explains the ungrammaticality of examples such as (84) pointed out by Laca (1990),
which show that bare nouns in Spanish, Italian and Romanian cannot appear in the
object position of verbs such as love, hate, etc. As expected, French de NPs behave
on a par with bare nouns in the other Romance languages:
(84) a. *María adora café. Spanish
a¢. *Marie adore du café French
Marie adores du coffee
b. *Juan detesta manzanas. Spanish
b¢. *Jean déteste des pommes French
Jean hates des apples
30
Some of these examples are acceptable for certain speakers but only with a “quasi-universal”
reading (or “functional” reading in Condoravdi’s (1992, 1994) terminology).
The ungrammaticality of examples of the type in (84a–b) correlates with the

ungrammaticality of examples such as (84c–d), which show that their predicates do
not allow for space adverbials:
(84) c. * María adora Juan en la cafetería. Spanish
c¢. * Marie adore Juan dans la cafétéria French
Marie adores Juan in the cafeteria
d. *Juan detesta Maria en la cocina. Spanish
d¢. *Jean déteste Maria dans la cuisine. French
Jean hates Maria in the kitchen
2.5.3 Some Apparent Problems
The space localization criterion must be refined in view of three types of data that
seem to be problematic.
Let us first observe that space adverbs are possible with any kind of predicate
when the subject is kind-referring. Thus, predicates such as expensive or sacred,
which disallow space adverbs when the subject DP is a proper name, allow them
when the subject is a name of kind:
(85) a. Les livres sont chers en France.
the books are expensive in France.
a¢. Books are expensive in France.
b. Les vaches sont sacrées en Inde.
the cows are sacred in India
b¢. Cows are sacred in India.
In examples of this type, space adverbials do not function as modifiers of the
main predicates but rather as modifiers of the subject DP, as suggested by glosses
such as ‘The books sold in France are expensive’ or ‘Cows in India are sacred’.
Kind-referring DPs can be easily modified by space localizers, because kinds refer
to entities that are spatially scattered; the role of the modifier is to restrict reference
to the realization of the kind at a certain location. Since individuals occupy one
location at a time, an individual-referring argument cannot be modified by a space
localizer. With those predicates that cannot be modified by space adverbs, examples
such as (86) are either unacceptable or else the space localizer is interpreted as a
time adverbial:
(86) a. ?? Jean est fatigué dans sa chambre.
Jean is tired in his room
b. Jean est heureux dans sa chambre.
Jean is happy in his room
(86b) is acceptable because it is easily interpretable as ‘whenever John is in his

room, he is happy’; (86b) is less acceptable because this type of interpretation is
more difficult with a predicate such as tired.
The sentences in (87) seem to provide another type of problem for the spatial
location hypothesis:
(87) a. * De la fumée était dans la pièce/ dans le four.
de la smoke was in the room in the oven
b. * Du beurre était sur la table.
du butter was on the table
* Du linge était dans la salle de bain.
du laundry was in the bathroom
Despite the presence of a space adverbial, these examples are unacceptable. But
note that in these examples the adverbial is not a modifier of the predicate but rather
it functions as the main predicate itself, the copula being semantically empty. If we
replace the copula with a verb, these sentences become acceptable.
(88) a. De la fumée s’élevait du four.
de la smoke refl-rise.impf from.the oven
‘Smoke was rising up from the oven.’
b. Du beurre traînait sur la table.
du butter lie.impf on the table
‘There was butter lying on the table.’
c. Du linge séchait dans la salle de bain.
du laundry dry.impf in the bathroom
‘There was laundry drying in the bathroom.’
To take care of this type of example, we may say that a predicate can be repre-
sented as existential if it allows space modifiers. Space predicates themselves can-
not function as existential predicates.
To sum up, we can distinguish two classes of predicates that allow for the exis-
tential reading of one of their arguments. The first class includes non-stative predi-
cates, all of which allow space modifiers, as in (89):
(89) a. Des enfants dormaient au coin du feu.
des children sleep.impf at-the corner of-the fire
‘There were children sleeping by the fire.’
b. Des enfants étaient en train d’écrire sur la table du séjour.
des children were prog to write on the table of the dining room
‘There were children writing on the table of the dining room.’
2.6 French Indefinites Headed by du/de la/des 69
Non-stative predicates describe events, which have to be localized in space and

time. The localization of an event indirectly locates the participants in the event, i.e.,
the arguments of the predicate, hence the existential reading of the subject in (89).
See also examples (88).
Stative predicates do not refer to events and therefore they do no allow space
adverbials. However, certain stative verbs are such that one of their arguments is
located with respect to another argument. One such predicate is surround, whose
subject is located with respect to the object. Although these predicates do not assert
the existence of an event, they allow an existential reading (cf. (77)) because they
locate their arguments with respect to each other.
The following generalizations summarize the previous discussion: (i) the argu-
ments of statives can only be located with respect to another argument of the same
predicate ; (ii) the arguments of non-statives can be located by means of the locative
expression that anchors the event in space (and time) and thus indirectly anchors the
participants in the event.
2.6 French Indefinites Headed by du/de la/des
In this section we show that French indefinites headed by du/de la/des are the clos-
est French counterparts of bare NPs. We start by illustrating the parallel distribution
and interpretive properties of BPs and bare mass NPs, which are echoed by the par-
allelism between plural des indefinites and mass indefinites headed by de la or du.
Nevertheless, the latter are more constrained than the former, which is a particular
case of a general crosslinguistic difference between mass and plural indefinites.
2.6.1 Bare Plurals and Bare Mass NPs
We have pointed out at several points in this chapter that mass nouns pattern with
plural nouns in English as well as in Romance languages other than French: both
types of nouns are allowed to occur without an article in argument positions (see
(90)–(91)), in clear contrast with singular count nouns, which need an indefinite
article in most contexts.
(90) a. John stole books from the library English
b. Juan robó libros de la biblioteca Spanish
c. Ion a furat cărţi de la bibliotecă Romanian
(91) a. John stole wood. English
b. Juan robó leña Spanish
c. Ion a furat lemn. Romanian
In Romance languages other than French, bare mass nouns also resemble BPs in
that they cannot combine with entity predicates, as illustrated in (92)–(93)31:
(92) a. *Juan adora café. Spanish
John loves coffee.
b. * Maria adoră cafea. Romanian
(93) a. *Juan respecta profesores Spanish
b. *Ion respectă profesori. Romanian
Moreover, bare mass nouns exhibit the interpretive effects that we have discussed
for BPs: they always take narrow scope (94) and induce atelicity effects. Examples
of the type in (94) are ungrammatical because the atelicity triggered by the BN fruta
is incompatible with the telicity induced by se comió:
(94) *Maria se comió fruta Spanish
Mary se ate fruit
2.6.2 Parallelisms between du/de la/des French Indefinites

and Bare NPs in the Other Romance Languages
In French, NPs are not allowed to appear bare in argument positions. But Dobrovie-
Sorin and Laca (2003) have shown that French DPs of the form de + definite arti-
cle + N32 represent the closest counterparts of existential bare NPs in other Romance
languages.
Like bare NPs in the other Romance languages, de + definite article NPs can be
used with plural and mass nouns but not with count singulars33:
31
The corresponding English examples are grammatical because English BPs and BMNs can refer
to kinds. The Spanish and Romanian examples become grammatical if we use a definite mass Ns
instead of bare mass nouns:
(i) Juan adora el café. Spanish

Juan loves the coffee
(ii) Maria adoră cafeaua. Romanian
32
De-expressions followed by a definite article (du/de la/des) are different from de-expressions
directly followed by a noun, which are required in negative contexts or when related to certain
adverbs such as beaucoup ‘many’ or peu ‘few’ (je n’ai pas vu de fille(s) ‘I haven’t seen de girl(s)’,
j’ai vu beaucoup / peu de filles ‘I have seen many / few de girls’). Here, we will leave aside this
other kind of expression.
33
In French, certain nouns, i.e., tomate ‘tomato’ can be interpreted, depending on the context,
either as a mass or as a countable expression. We can thus have des tomates (countable plural) as
well as de la tomate (mass, e.g., J’ai mangé de la tomate, or number neutral, e.g., Il y avait de la
tomate sur le marché).
(95) des (= de les) enfants

des (=de the.pl) children
*de l’ enfant
de le (=de the.msg) child
de la farine
de la (=de the.fsg) flour
du (= de le) beurre
du (=de the.msg) butter
Furthermore, du/de la/des NPs can combine with existential but not with entity
predicates:
(96) a. De la fumée s’élevait du four.
de la smoke se rise.impf from-the oven
‘Smoke was rising up from the oven.’
b. 300g de beurre étaient en train de fondre sur la table.
300 g of butter were prog to-melt on the table
‘300 g of butter were melting on the table.’
c. *Du beurre était savoureux.
du butter was delicious
d. *Du beurre était mou.
du butter was soft
De + definite article NPs are incompatible with a generic reading, just like
Spanish or Romanian bare plurals.
Given these similarities with existential BNs, French de + def art NPs can be
analyzed as weak indefinites, which in our implementation denote generalized
quantifiers over amounts.
2.6.3 On the Strong Reading of des Indefinites
It should nevertheless be mentioned that in addition to the unmarked use described

above, French plural des NPs may take strong readings in certain environments (Attal
1976, Chap. 4 below).34 Under their strong reading, des indefinites can combine with
34
Attal’s distinction is based on the observation that these plural des NPs can escape the scope of
negation, have specific readings in intensional contexts, or occupy argument positions of certain
verbs. Attal (1976: 142) leaves open the question of whether there is “one or two des”. More recent
studies (for an overview, see Bosveld-de Smet 1997: 5–54) show that there are good reasons to
assume the existence of the marked use of des.
entity predicates (96a), can escape the scope of negation (97b) and can appear in
contrastive contexts, where they supply an antecedent for the anaphoric expression
d’autres (97c):
(97) a. Des élèves étaient malades.
des students were sick
‘Sm students were sick.’
b. Des élèves ne m’ont pas rendu leurs devoirs.
des students neg me-have neg returned their assignments
‘Sm students didn’t hand in their assignments.’
c. Des enfants tambourinaient sur leurs tables, tandis que d’autres
des children drum.impf on their tables while others
criaient à tue-tête.
scream.impf out loud
‘Sm children were drumming on their tables, while others were
screaming out loud.’
The possibility of strong readings distinguishes French des indefinites from the
BPs of other Romance languages:
(98) a. *Ieri erau bolnavi elevi.
yesterday were sick students
b. *Nu mi-au adus lecţiile elevi.
neg me-have brought lessons-the students
This indicates that the presence of an overt determiner (e.g., a numeral or des in
French) is mandatory for a strong reading (leaving aside the generic reading). The
possibility of strong readings is indeed a general crosslinguistical property of plural
indefinites headed by overt determiners (see Chaps. 4 and 5):
(99) a. Ieri erau bolnavi nişte elevi.
yesterday were sick some students
‘Some students were sick yesterday.’
b. Nu mi-au adus lecţiile trei elevi.
neg me-have brought lessons-the three students
‘Three students didn’t bring me their lessons.’
In Sect. 2.4.4 above we have proposed that plural nouns denote sets of
amounts. Since both BPs and des indefinites refer to a random element in a set
of amounts, the default interpretation of both types of expressions is the weak
reading. The possibility of strong readings of des indefinites can be attributed to
the fact that the presence of the determiner signals the choice of a particular
plurality, which is isolated from the other pluralities in the domain and as such
it acquires the status of a plural individual, i.e., a plural entity that is distinct
from any other plural entity.
2.6.4 Mass Nouns and the Impossibility of Individuation
The examples in (100) below show that mass indefinites headed by du/de la cannot
take strong readings:
(100) a. *De la fumée était épaisse.
de la smoke was thick
b. *De l’eau n’était pas sale.
de l water neg was neg dirty
The contrast between the acceptability of plural des indefinites in (97) and
the unacceptability of mass de la/du indefinites in (100) is a particular case of a
general contrast between plural and mass indefinite DPs. Indeed, the examples
in (100) contrast with those in (101) below, in which the mass DP is built with
an overt determiner:
(101) a. *Ieri era limpede nişte apă.
yesterday was clear some water.
b. *200g de apă nu era murdară.
200g of water neg was dirty
Moreover, the examples below show that whereas plural indefinite DPs can be
used as arguments of love or hate, mass indefinites cannot appear in this context
even when preceded by a determiner:
(102) a. Maria urăşte doi băieţi.
Maria hates two boys
‘Maria hates two boys.’
b. *Maria adoră 200g de cafea.
Maria adores 200g of coffee
In sum, mass indefinites headed by overt determiners, in particular by du / de la,
differ from plural indefinites headed by overt determiners, in particular by des, inso-
far as they cannot take strong readings. We are thus led to conclude that the use of
an overt indefinite determiner is not sufficient for isolating a particular amount of
substance from the other amounts that satisfy the same property. This difference
between mass and plural indefinites can be attributed to the properties of their
respective domains of denotation: the plural domain is generated from a set of
individuals (i.e., entities that are distinct, non overlapping), whereas the mass
domain is generated from a set of amounts (i.e., overlapping entities). Reference to
individualized amounts is possible for demonstratives (this water, these 200 g of
butter) and for definites (the water in the tub, the butter on the table) because these
determiners correspond to operations (deixis and maximalization, respectively) that
yield individualized entities.
2.7 Bare NPs in Predicate Positions
Let us now examine bare NPs in predicate positions. French shows that the use
of bare NPs in predicate positions is disconnected from the use of bare NPs in
argument positions. More precisely, French allows BNPs in predicate positions
(with certain nouns), although it does not allow them in argument positions:
(103) a. *Jean était en train de lire livre.
Jean was prog to read book
b. Jean et Marie sont étudiants.
Jean and Marie are students
c. Jean est étudiant.
Jean is student
The absence of correlation between the use of bare NPs in argument and predicate
positions is also visible in the other Romance languages, where the use of BSs in
argument positions is much more restricted than that of BPs (see Sect. 2.1.2 above),
whereas in predicate positions, BSs can be used in all the contexts in which a BP
can be used.
Given that predicate positions presumably require property-denotation, we
might expect bare nouns in predicate positions to be possible (at least for a certain
lexical class of nouns) in all languages. English is a counterexample to this
generalization35:
(104) a. *John is student.
b. John is a student.
c. John and Mary are students.
The impossibility of using BSs in predicate (as well as in argument) positions is
to be analyzed as being due to some syntactic constraint that operates in English as
opposed to all the other Romance languages, including French.36
In the rest of this chapter we will concentrate on the contrast between BSs and
indefinite singulars (ISs henceforth) in predicate positions, which is exhibited by all
35
English does allow BSs in predicate positions but only if they have unique reference, e.g., with
some functional nouns:
(i) Mary is chairwoman.
36
According to Dobrovie-Sorin (to appear b), Number is a feature that attaches to little n in
English and as such it is necessarily realized (as –s in the plural and as the indefinite article a in
the singular) on nouns, even in predicate positions. In Romance languages, on the other hand,
Number would attach to Det, which would allow nouns to show up without any Number
marking.
2.7 Bare NPs in Predicate Positions 75
Romance languages. In French, this contrast is paralleled by the contrast between

BPs and des indefinites:
(105) a. Minou est un chat.
Minou is a cat
a¢. Minou et Minette sont des chats.
Minou and Minette are des cats
‘Minou and Minette are cats.’
b. Jean est étudiant.
Jean is student
b¢. Jean et Marie sont étudiants.
Jean and Marie are students
Due to lack of space, the contrast between BPs and des indefinites will not be
illustrated with examples in the rest of this chapter but readers may check on their
own that modulo the singular/plural difference, the proposed analyses cover
the corresponding contrasts between BPs and plural des indefinites. Similar con-
trasts between BPs and plural indefinites can be found in the other Romance
languages.
Our generalizations will be illustrated with French examples. This choice is
motivated by several reasons: (i) as explained above, this language shows clearly
that the use of bare NPs, in particular BSs, in predicate positions is a phenomenon
that is distinct from the use of bare NPs in argument positions; (ii) the contrast
between BSs and ISs is paralleled by the contrast between BPs and des indefinites;
(iii) French has constructions of the type Marie, c’est une belle femme ‘Mary, this is
a good-looking woman’, which are crucial for the analysis of the contrast between
BSs and ISs.
Both the generalizations and the proposed analyses are assumed to cover the
other Romance languages but not English, as in English copular sentences ISs do
not alternate with BSs and thus what we say about ISs in Romance languages is not
expected to hold for English ISs.
In order to compare ISs and BSs in predicate positions in French, we will
consider only unmodified nouns, for which the contrast between ISs and BSs is
clear cut.
2.7.1 A Subclass of Nouns
A restricted class of common nouns may appear bare in the predicate positions of
copular sentences. Such predicates refer to professions (professeur ‘professor’,
avocat ‘lawyer’), titles (prince), hobbies (alpiniste ‘climber’), functions (secrétaire
‘secretary’), occupations (étudiant ‘student’) or social status (chomeur ‘unem-
ployed’) (cf Laca and Tasmowski 1994; Matushansky and Spector 2004; de Swart
et al. 2007; Roy 2006).
Several authors have tried to characterize these nouns. According to Roy (2001,
2006), those nouns that can be used bare in the postcopular position are event nouns,
i.e., nouns that have an event-variable in their argument structure. Thus, profession
nouns such as teacher, dancer or professor would refer to activities when used
bare in predicate positions: Jean est enseignant means that John teaches, which
refers to an activity. According to Roy, all nouns can be used bare when they are
predicated of human individuals, provided that they allow an eventive reading.
Roy’s characterization seems problematic for nouns such as père ‘father’ or fille
unique ‘only daughter’:
(106) a. Jean est père (de trois enfants)
Jean is father (of three children)
‘Jean is the father of three children.’
b. Marie est fille unique.
Mary is daughter unique
‘Marie is an only daughter.’
Matushansky and Spector (2004) claimed that nouns that can appear bare in
predicate positions share the property of being non-scalar. These authors distinguish
nouns like génie ‘genius’, which can never be bare, from nouns like professeur
‘professor’, by pointing out that only the former are compatible with degree
modification: one can be a real or an absolute genius but cannot be more or less of
a professor. However, this description is not entirely adequate, given the uncertain
boundary between scalar and non scalar predicates and the fact that non scalar
predicates can be easily coerced into acquiring scalar readings (cf Kennedy 1999).
A non-scalar adjective like français ‘French’ can be used in the comparative (Jean
n’est pas plus français que Pierre ‘John is not more French that Peter’), thus
acquiring a scalar interpretation.
Beyssade and Dobrovie-Sorin (2005) argue that nouns that can be bare in
predicate positions are non sortal (cf Gupta 1980), i.e., nouns that are not inherently
associated with an individuation principle. Sortal and non-sortal nouns can be
distinguished on the basis of the counting test. Imagine a situation where I face a
group formed by two men (a math teacher and a physics teacher) and a woman
teaching both literature and English. If I am asked how many women there are in the
group in question, I can easily answer ‘only one’. However, if I am asked how many
teachers there are, the answer is less straightforward, depending on whether I am
supposed to count the woman once or twice (once as a literature teacher and once as
an English teacher). Professeur ‘teacher’ is a non-sortal noun, without any inherent
individuation principle: when counting teachers, we can either count the actual
individuals or the disciplines they teach. The situation is even clearer for nouns
like passager ‘passenger’, which refers to individuals who take part in an event.
In order to count passengers we count either individuals (and if one individual
travels twice, he counts as one passenger), or trips (and therefore if one individual
travels more than once, he gets counted several times). Beyssade and Dobrovie-
Sorin claim that in French, bare nouns in predicate positions are non sortal, a notion
that covers nouns referring to professions, titles, hobbies, as well as relational nouns,
functional nouns and event nouns:
(107) a. Jean est père.
Jean is father
‘Jean is a father.’
b. Jean est ami avec Marie.
Jean is friend with Marie
‘Jean is friends with Marie.’
c. Jean est passager sur le vol n° 345.
Jean is passenger on the flight n° 345
‘Jean is a passenger on flight nº 345.’
2.7.2 Distributional Differences between Singular Indefinites

and Bare Singulars
There are several syntactic and semantic differences between ISs and BSs in predi-
cate position in Romance languages. Here we will present two of these differences.
First, singular indefinites cannot appear as predicates of small clauses, in clear
contrast with bare singulars.
(108) a. *Marie imagine Paul un ministre.
Marie imagines Paul a minister
b. Marie imagine Paul ministre
Marie imagines Paul minister
With respect to this context, BSs behave on a par with adjectives and PPs, which
can also appear in small clauses, and differ from DPs headed by determiners, which
cannot appear in small clauses, regardless of whether they are indefinite (as in
(109c) or definite as in (109d)).
(109) a. Marie croit Paul coupable.
Marie believes Paul guilty
‘Marie believes Paul guilty.’
b. Marie croyait Jean en difficulté.
Marie believe.impf Jean in trouble
‘Marie believed John was in trouble.’
c. *Marie croyait Jean un avocat.
Marie believe.impf Jean a lawyer
d. *Marie croyait Jean le meilleur avocat.
Marie believe.impf Jean the best lawyer
Secondly, an interesting correlation holds in French copular sentences (a.o.,

Kupferman 1979) between the type of subject pronoun, namely il/elle vs. ce ‘this,
that’ and the type of predicate, namely ISs vs. BSs:
(110) a. Marie, elle est professeur.
Marie she is professor
b. ?Marie, elle est une belle femme.
Marie she is a beautiful woman
(111) a. *Marie, c’est professeur.
Marie that is professor
b¢. Marie, c’est une belle femme.
Marie that is a beautiful woman
2.7.3 Higgins’ Typology Revisited
According to Higgins (1979), four types of copular sentences need to be

distinguished:
(112) a. John is a doctor. predicational
b. Clark Kent is Superman. equative
c. That’s John. identificational
d. The problem is John. specificational
Roy (2006) and Beyssade and Dobrovie-Sorin (2005) have shown that the con-
trasts between BSs and ISs cannot be explained on the basis of Higgins’s typology.
Roy postulates a three-way distinction among predicational copular sentences37 and
Beyssade and Dobrovie-Sorin assume two distinct rules of predication, called
attributive predication and classifying predication. According to both proposals,
Higgins’ class of predicational sentences is split into two types, corresponding to
ISs and BSs, respectively.
Both of these analyses are problematic insofar as they do not take into account
important similarities between copular constructions built with ISs and the three
types of non-predicational sentences defined by Higgins. Indeed, copular sentences
built with ISs pattern with the non-predicational copular sentences wrt to the
two tests used above, i.e., the compatibility with small-clauses (see (113)) and the
alternation between ce / il pronouns (see (114)).
37
The notions of ‘defining’, ‘characterizing’ and ‘situation-descriptive’ predicates, which are
defined in semantic terms, are assumed to entertain a one-to-one relation with distinct grammatical
categories:
(i) Defining predicates are expressed by ISs.
(ii) Characterizing predicates are expressed by BSs.
(iii) Situation-descriptive predicates are expressed by adjectives.
(113) a. *Mary croit John un médecin. predicational

Mary believes John a doctor
b. *Mary croit Clark Kent Superman. equative
Mary believes Clark Kent Superman
c. *Mary croit ce John. identificational
Mary believes that John
d. *Mary croit le problème John. specificational
Mary believes the problem John
(114) a. Jean, c’ est un médecin. predicational
Jean that is a doctor
b. Clark Kent, c’est Superman. equative
Clark Kent that is Superman
c. Ca, c’est John. identificational
that that is John
d. Le problème, c’est John. specificational
the problem that is John
The parallelisms between copular sentences built with ISs and the non predica-
tional types of copular sentences led Beyssade and Dobrovie-Sorin (2009) to pro-
pose that copular sentences built with ISs are not predicational but instead should be
grouped together with the three classes of non-predicational sentences. Higgins’
four-way typology can thus be reduced to a binary distinction between predicational
sentences and identity sentences (equative, identificational and specificational sen-
tences can be viewed as subtypes of identity sentences).38 As we will see in
Sect. 2.7.4 below, predicational sentences are defined in a more constrained way
that allows us to split Higgins’s predicational sentences into predicational (in this
more restricted sense) and identity sentences. This binary distinction correlates with
a difference in syntactic categories:
(115) a. Identity sentences require full DPs in the post-copular position.
b. Bare singulars can only appear in predicational sentences, on a par with
adjectives.
2.7.4 Semantic Composition
According to Beyssade and Dobrovie-Sorin (2009), predicational copular sen-

tences rely on attributive predication, whereas identity copular sentences rely on
38
Beyssade and Dobrovie-Sorin’s hypothesis that equatives, identificationals and specificationals
are all identity sentences builds on Heycock and Kroch’s (1999) view that specificationals are
base-generated as equatives rather than derived via Predicate Inversion from a predicational
configuration (Moro 1997; den Dikken 2006).
identity predication. Below we present the definitions of these two types of

predications and we make explicit the semantic composition for each of them.
(116) Attributive predication
[[DP is Adj/BS]] = 1 iff [[DP]] has PAdj / PBS
or in other terms
iff PAdj / PBS ∈ [[DP]]
In words, a sentence of the form DP is Adj/BS is true iff the property denoted by
the Adj or BS belongs to the set of properties denoted by the DP.
Crucial for our present purposes are the semantic types assumed for the two
expressions around the copula: the subject must be analyzed as denoting a set of
properties (i.e., as a generalized quantifier) and the predicate as a property-denoting
expression. The copula plays no role, it is empty and can be represented by the
identity function lPP. The compositional analysis of a sentence based on attributive
predication is given in (117); j is a constant that refers to John and H is a unary
predicate corresponding to handsome.
(117) a. John is handsome
b. S
PP(j). xH(x) which reduces first to xH(x)(j)
and then to H(j)
John is handsome
PP(j) PP. xH(x) which reduces to xH(x)
is handsome
PP xH(x)
Turning now to identity sentences, Beyssade and Dobrovie-Sorin (2009) analyze

them as relying on an identity predicate, which relates two arguments of the same
kind (two entities, two properties, etc.). In these sentences, the copula is interpreted
as an identity predicate. This does not mean that we assume identity sentences to be
symmetric. Indeed, even equatives can be shown to be asymmetric, as observed by
Groenendijk et al. (1996). The asymmetry that characterizes identity sentences can
be captured by assuming that they do not rely on the symmetric identity relation, as
in (118a) but rather on a one-place predicate derived from an identity relation by
saturating the postcopular argument. The asymmetry between the two arguments is
due to information structure.
(118) a. identity relation: lx ly (x = y)
b. identificational predicate, e.g., BE a lx (x = a)
Equative sentences can thus be represented as in (119c), which is obtained by
applying an identificational predicate of the form shown in (118b) to the individual
denoted by the subject DP, as in (119b), which reduces to (119c) by lambda-conversion.
The two constants ck and s correspond to the individuals respectively denoted by
Clark Kent and Superman.
(119) a. Clark Kent est Superman

b. lx (x = s) ck
c. ck = s
Higgins’ three classes of non-predicational sentences illustrated in (112)b–d can
all be viewed as relying on a predicate of the type in (118b) and as such all of them
can be called ‘identity sentences’. They differ from each other by the type of the
subject: (i) equative sentences are those identity sentences built with proper names
or definite descriptions; (ii) identificational sentences are identity sentences built
with presentational pronouns; (iii) specificational sentences are identity sentences
built with attributive subjects:
(120) Types of identity sentences
a. Clark Kent is Superman.
Equative: the subject DP = proper name
b. That’s John.
Identificational: the subject DP = presentational pronoun
c. The problem is John.
Specificational: the subject DP = attributive use
Beyond their differences, all of the three types of subjects enumerated here are
entity-denoting expressions. Specificational sentences are special insofar as they are
known to be built with DPs used attributively.39 Note however that the subjects of
equatives themselves have been argued to be used attributively (Groenendijk et al.
1996), in which case they would mean ‘the individual referred to by the name of
Clark Kent’ rather than ‘the individual Clark Kent’. Finally, the presentational pro-
nouns that appear in identificationals are also entity-denoting expressions that might
be viewed as involving a particular type of attributive use: ‘the individual referred
to by using ce ‘this/that’ (whom you don’t know)’. These observations suggest that
the subjects of our three types of identity sentences are much more similar than one
would believe: all of them rely on e-type expressions that are used attributively.
2.7.5 Copular Sentences Built with Singular

Indefinites as Equatives
Turning now to the postcopular expressions, identificational and specificational sen-

tences allow not only proper names but also indefinite DPs, as illustrated below:
(121) a. That is a woman. Identificational
b. What John saw is a woman. Specificational
39
Comorovski (2007) proposes to represent the attributive use of specificational subjects as inten-
sional expressions of type < s,e > .
Let us now assume that the third type of identity sentence, equatives, may also be
built with a postcopular indefinite. Such a hypothetical equative would be a copular
sentence of the form in (122b), which is obtained from the canonical equative given
in (122a) by replacing the proper name in the postcopular position with an indefinite
DP. Analyzed in this way, the sentence in (122b) says that there is an actor to which
Clark Kent is identified:
(122) a. Clark Kent is Superman.
b. Clark Kent is an actor.
This analysis of copular sentences of the type in (122b) fills a gap in the paradigm
of identity sentences, insofar as postcopular indefinite DPs appear not only in
specificational sentences and identificational sentences but also in equatives. We are
thus led to conclude that copular sentences built with postcopular singular indefinites
are not to be analyzed as predicational sentences – as currently assumed – but rather as
equatives: a singular indefinite in the postcopular position denotes an individual to
which the entity denoted by the subject is identified. Under our proposal, a strong
dividing line separates copular sentences built with bare singulars from those built with
singular indefinites: only the former are predicational sentences (and as such group
with copular sentences built with adjectives), whereas the latter constitute a particular
type of identity sentence, namely equatives built with ISs in the predicate position.
Predicational sentences rely on a rule of attributive predication, which estab-
lishes a relation between a property-denoting predicate and a subject DP that denotes
a generalized quantifier (i.e., a set of properties). Identity sentences, on the other
hand, rely on an identity predicate, which establishes a relation between a subject
DP that denotes an entity (since it is used attributively, the subject DP cannot be
shifted to denote a generalized quantifier) and another entity. We thus postulate the
existence of two distinct copulas. In predicational sentences, the copula is semanti-
cally empty and as such can be represented as the identity function (see (123a)),
which applies to a formula and yields the same formula. In identity sentences, on
the other hand, the copula denotes the identity relation (see (123b))40:
40
The distinction between the identity and the predicational copula is a standard one. Less standard
is the view that specificational, identificational as well as copular sentences built with ISs all rely
on an identity copula. Although den Dikken’s (2006) account of copular sentences differs from
ours insofar as it crucially relies on Predicate Inversion for the analysis of specificationals and
refutes the existence of base-generated equatives, it is nevertheless similar to ours insofar as it
recognizes only two types of copular sentences, Moro’s (1997) canonical and inverse predication.
Correspondingly, den Dikken’s copulas are either ‘relators’ or ‘linkers’. Our approach in terms of
identity predication (rather than Inverted Predication) is well motivated on semantic grounds (an
identity predicate is needed for the semantic composition), whereas under den Dikken’s account,
the ‘linker’-type of copula has nothing to do with identity, being syntactically triggered by the
movement underlying Predicate Inversion. According to den Dikken, the Predicate Inversion anal-
ysis of specificationals and equatives has the advantage of explaining the ban on A’-extraction of
the postcopular expression, e.g., *whose opinion of Smith do you think your opinion of Jones is?
We leave it open for further research the objective of showing that this constraint can be explained
as an effect of identity predication (as opposed to attributive prédication).
(123) a. lP lx P(x) copula in predicational sentences

b. lx ly (x = y) copula in identity sentences
Beyssade and Dobrovie-Sorin’s (2009) analysis presented above can be summa-
rized as follows:
(124) a. Predicational copular sentences are built with a predicate that denotes a
property and a subject that denotes a generalized quantifier. The copula is
semantically empty.
b. Identity copular sentences are built with an identity predicate that estab-
lishes an identification relation between the entity denoted by the subject
and the entity denoted by the postcopular DP. The copula denotes the
identity relation.
2.7.6 Explaining the Contrasts between Bare Singulars

and Indefinite Singulars
In what follows we will show that Beyssade and Dobrovie-Sorin’s (2009) pro-
posal is able to explain the contrasts between ISs and BSs in postcopular posi-
tions. We begin with the two contrasts presented in Sect. 2.7.2 above and continue
with some others.
2.7.6.1 Small Clauses and Secondary Predication
Small clauses are characterized by the absence of the copula. This absence is legiti-
mate with BSs because BSs are property-denoting expressions and correspondingly
the copula preceding them is semantically empty, which explains why it can be
dropped. ISs, on the other hand, are entity-denoting expressions that must combine
with a copula that denotes the identity relation. Being semantically non empty, the
copula cannot be dropped, which explains why ISs cannot appear as predicates of
small clauses.
The hypothesis that BSs can denote properties, whereas ISs cannot do so, also
explains why BSs can function as secondary predicates, whereas ISs cannot do so:
(125) Jean est né (*un) prince et est mort (*un) mendiant.
‘John was born a prince and died a beggar.’
In sum, both small clauses and secondary predications are instances of attributive
predication (as opposed to identity predication) and therefore ISs are ruled out in
these environments.
2.7.6.2 Alternation Between ce ‘that’ and il/elle ‘he/she’
Turning now to the alternation between ce ‘that’ and il ‘he’ (see (110–111)), these
pronouns differ with respect to their denotation: only the former can be analyzed as
a generalized quantifier (type < <e,t>,t>).41 Since the pronoun ce necessarily refers
to an entity of type < e>, it is ruled out from predicational sentences.42
The examples in (126) show that only ISs can provide an answer to questions of
the type qui est…? ‘who is’, whereas only BSs can be the antecedent of an anaphor
like ce que ‘which’ or l’est aussi ‘also’.
(126) a. Qui est Jean? Jean est *( un) chanteur
who is Jean Jean is a singer
‘Who is Jean? Jean is a singer’
b. Jean est *( un) docteur, ce que je ne serai jamais.
Jean is a doctor that which I neg be.fut never
‘John is a doctor, (that) which I’ll never be.’
c. Marie est * (une) institutrice, et sa sœur l’est aussi.
Marie is a primary school teacher and her sister that is also
‘Marie is a primary school teacher, and her sister is that too.’
This behavior is expected, given that qui ‘who’ refers to entities of type < e>,
whereas the antecedent of ce que ‘which’ and l’est aussi ‘also’ is property-denoting.
2.7.6.3 Modifying PPs for Names of Role
It is also expected that only bare nouns are compatible with expressions like de
profession ‘by profession’, de nationalité ‘by nationality’ and with qua phrases (en
tant que ‘as’), which select a property-denoting argument:
41
Partee (1987) showed that a DP could be associated to several semantic types, which are related
to each other by type-shifting rules. According to Beyssade and Dobrovie-Sorin, ce ‘that’ is special
in that it is necessarily of type e, i.e., it can only denote an entity. Ad-hoc as it may seem, this
characterization of ce seems natural: languages may resort to special lexical items in order to make
reference to objects that are viewed as having no property at all, which means that they cannot be
viewed as denoting sets of properties.
42
Note however that when it refers to either a kind or a proposition (finite or not), ce can be the
subject of attributive predication:
(i) La soupe, c’est bon.
The soup that’s good,
‘Soup, that’s good.’
(ii) Que Jean fasse du jogging, c’est surprenant.
That Jean do of jogging, that’s surprising,
‘That Jean jogg is surprising.’
(iii) Faire du jogging, c’est sain.
doing of jogging, that’s healthy,
‘Jogging is healthy.’
(127) a. Pierre est (* un) avocat de profession.

Pierre is a lawyer of profession
‘Pierre is a lawyer by profession’
b. Pierre est(* un) chrétien de religion.
Pierre is a Christian of religion
‘Pierre is a Christian by religion’
c. Pierre est (* un) français de nationalité.
Pierre is a Frenchman of nationality
‘Pierre is French by nationality’
(128) En tant que (*un) médecin, Pierre n’a pas voulu prendre
as such as a doctor Pierre neg has neg wanted take
position sur ce sujet.
position on this subject
‘As a doctor, Pierre didn’t want to take a position on this matter.’
2.7.6.4 Quantified Subjects
Bare nouns can appear in the predicate position of copular sentences built with a
quantified subject, whereas singular indefinites cannot do so:
(129) a. Parmi ses amis, personne n’est (* un) acteur.
among his friends nobody neg is a actor
‘Among his friends, nobody is an actor.’
b. Dans sa famille, tout le monde est (* un) médecin de père en fils.
in his family everybody is a doctor from father to son
‘In his family, everybody is a doctor from father to son.’
Sentences like (129a–b) have the Logical Form in (130), in which the quantifier
denotes a relation between two sets:
(130) a. NOx (person (x)) actor (x)
b. "x (person (x)) doctor (x)
The ungrammaticality of (129a–b) is expected under the hypothesis that ISs can-
not denote properties and as such they cannot supply one of the two sets related by
the quantifier. Combined with the copula, ISs necessarily form identity predicates
and therefore can only take entity-denoting subjects.
2.7.6.5 Lifetime Effects
As observed by Roy (2006) and Matushansky and Spector (2004), a.o., ISs and
BSs contrast with respect to the so-called ‘lifetime effects’, which can be analyzed
in terms of implicature. Thus, the sentence in (131a), built with an indefinite, is

both true and appropriate, although Balzac is actually dead, whereas the sentence
in (131b), with a bare noun, is odd, because it carries the implicature that Balzac
is alive:
(131) a. Balzac est un écrivain.
Balzac is a writer
‘Balzac is a writer’
b. ?Balzac est écrivain.
Balzac is writer
The contrast between ISs and BSs regarding lifetime effects can be explained in
terms of the distinction between predicational and identity sentences if we assume
that identity is an existent-independent predicate (the notion of existence-indepen-
dent predicate is borrowed from Musan (1997)). Identity expresses an atemporal
relation: an identity sentence imposes no requirement concerning the times of exis-
tence of the referents of DP1 and DP2. Indeed, as shown in (132), it is possible to
establish an identity relation between two DPs that are associated to two disjoint
time intervals corresponding to the past and future times with which the referents of
Jean and Marie are associated:
(132) Ce que Jean a ressenti hier est exactement ce que
that what Jean has felt yesterday is exactly that what
Marie ressentira demain.
Marie feel.fut tomorrow
‘What John felt yesterday is exactly what Mary will feel tomorrow.’
Attributive predication, on the other hand, requires that the time of existence te of
the subject must cover the situation time ts:
(133) In predicational sentences, te É ts
The present tense on the copular verb conveys the same meaning as an explicit
adverb such as today. Therefore, because ts = today, the requirement in (133) amounts
to the requirement te É today, which implies that the referent of DP1 is alive. This
explains why copular sentences built with the present tense and BSs, which are to
be interpreted as relying on attributive predication, exhibit lifetime effects. No such
effect arises with ISs, because they rely on identity predication.
According to Roy, among others, the lifetime effect also characterizes copu-
lar sentences such as (134a), built with a past tense and a bare singular in predi-
cate position. Note however that in the past sentence, the implicature ‘no longer
alive’ is easily cancelable by adding a time adverbial, as in (134b). The only
inference of (134b) is that Paul is no longer a doctor, hence (134b) is true if Paul
has retired or changed his job. This inference is itself easily cancelable, as
shown in (134c):
(134) a. Paul était médecin.

Paul was doctor
‘Paul was a doctor’
b. Paul était médecin avant d’émigrer aux Etats-Unis.
Paul was doctor before to emigrate to the US
‘Paul was a doctor before he emigrated to the US’
c. Paul était médecin avant d’émigrer aux Etats-Unis,
Paul was doctor before to emigrate to the US
et il l’est resté.
and he that is remained
‘Paul was a doctor before he emigrated to the US and he remained a
doctor.’
2.7.6.6 Spatial and Temporal Modification
Let us finally consider the examples in (135), which show that spatial and tem-
poral modifiers are allowed in predicational sentences but not in identity
sentences:
(135) a. Paul est (* un) médecin à Paris.
Paul is a doctor in Paris
‘Paul is a doctor in Paris’
b. Paul a été (*un) professeur à trois occasions dans sa vie.
Paul has been a professor at three times in his life
‘Paul has been a professor three times in his life’
c. Marie est de temps en temps (* une) ouvreuse à l’Odéon.
Marie is from time to time a usher to the Odéon
‘Marie is occasionally an usher at the Odéon’
d. Jean a été (* un) étudiant de 1995 à 1998.
Jean has been a student from 1995 to 1998
‘John was a student from 1995 to 1998.’
In the non starred versions of the examples above, the locative or temporal phrase
is a sentential complement, whose semantic role is that of a property modifier: it
turns a given property into a new one. If be doctor is a property, be doctor in Paris,
or be doctor during the day are two other properties. The ungrammaticality of the
starred versions is due to the fact that in identity sentences, the postcopular expres-
sion does not denote a property and as such it cannot be modified. DP–internal
modifiers are allowed, as in (136b):
(136) a. Paul est un médecin de jour/ * le jour.

Paul is a doctor of day the day
‘Paul is a doctor during the day.’
b. Jean est un médecin de la faculté de Paris.
Jean is a doctor of the faulty of Paris
‘Jean is a doctor from the university of Paris.’
2.7.6.7 Attributive Uses of Indefinite Singulars
Note finally that copular sentences convey different meanings depending on whether
they are built with or without an indefinite article, as illustrated by (137):
(137) a. Jean est clown.
Jean is clown
‘Jean is a clown’
b. Jean est un clown.
Jean is a clown
‘Jean is a clown’
According to Laca and Tasmowski (1994), the IS has a metaphoric meaning.
This is an appropriate description of the contrast given in (137) since être clown ‘(to
be [a] clown)’ in (137a) means to be a professional clown, whereas être un clown ‘(to
be a clown)’ in (137b) is less precise: the noun clown in (137b) does not necessarily
refer to John’s profession but rather says something about John’s character, e.g., he
likes to fool around or he is a funny guy.
Although they are built with ISs, copular sentences of the type in (137b) are
predicational rather than identity sentences. Indeed, in these examples, the subject
DP does not have an attributive use, as in identity sentences but rather a referential
use, as in predicational configurations. Correspondingly, the postcopular IS is not
interpreted as referring to an individual belonging to the class of (professional)
clowns but rather as an attributive predicate roughly paraphrasable by ‘have the
qualities of a clown’ or ‘behave as a clown’.
The predicational use of ISs can also explain a number of examples that are
problematic for Beyssade and Dobrovie-Sorin’s (2009), as correctly pointed out by
a reviewer of this book:
(138) a. Chaque langue est un art.
every language is an art
‘Every language is a work of art.’
b. Chaque maison est un royaume.
every house is a kingdom
‘Every house is a kingdom.’
c. Chaque homme est un orateur.
every man is an orator
‘Every man is an orator.’
In all of these examples, the postcopular IS is not interpreted as denoting an

individual but rather as denoting properties associated to prototypical entities such
as a work of art, a kingdom or an orator: ‘every language has the status/ (some of)
the qualities of a work of art’, ‘each house has the status/ (some of) the qualities
of a kingdom’, ‘every house has (some of) the status/ (some of) the qualities of a
kingdom’, ‘every man has (some of) the qualities/status of an orator’. The example
in (138d) can be explained along the same lines:
(138) d. Chaque ouvrage est un tout.
every work is a whole
‘Every piece of work is a whole.’
This example is to be analyzed as relying on attributive predication rather than
on an identity relation because be a whole means ‘form a whole’.
The explanation proposed above seems correct but we should acknowledge that
it weakens Beyssade and Dobrovie-Sorin’s (2009) assumption that unmodified ISs
necessarily form an identity predicate. They do so in those contexts in which ISs
denote an individual belonging to the class denoted by the noun. But there are con-
texts in which ISs seem to be able to function as attributive predicates.
2.7.7 Modified Bare Nouns
Unlike singular indefinites, which are compatible with any kind of modifier, bare
noun modification is much more constrained:
(139) a. Jean est (un) médecin généraliste.
Jean is a doctor generalist
‘Jean is a general practitioner.’
b. Jean est *( un) médecin honnête.
Jean is a doctor honest
‘Jean is a honest doctor’.
Modified bare nouns denote complex predicates, built from the simple properties
denoted by the unmodified bare nouns. There are at least three different ways of
building complex properties by modification.
First, a noun of profession N may be modified by a postnominal adjective or by
another noun preceded by a functional preposition, yielding a hyponym of N: méde-
cin généraliste ‘general practitioner’ in (140a) designates a type of doctor and chan-
teur de jazz ‘ jazz singer’ in (140b) a type of singer. The only constraint on these
constructions is a purely lexical one: the resulting NP must denote a subtype of the
unmodified N.
(140) a. Jean est médecin généraliste.

Jean is doctor generalist
‘Jean is a general practitioner.’
b. Jean est chanteur de jazz.
Jean is singer of jazz
‘Jean is a jazz singer.’
Let us now consider BSs preceded by adjectives of the type bon ‘good’ or mau-
vais ‘bad’:
(141) a. Jean est bon / mauvais médecin.
Jean is good bad doctor
‘Jean is a good/bad doctor’
In cases of this type, we may assume that the head of the postcopular phrase is
the adjective itself rather than the noun. A complex property is built by modifying
an adjective by a name of role. Therefore, names of roles can be taken to denote,
depending on the context, either a property or a property modifier. In (141), the
name of role médecin is analyzed as a property modifier rather than as a property:
the property attributed to Jean is not the property of being a doctor but the property
of being bad or good, as a doctor.
This analysis presents two advantages compared to some other recent proposals
(Matushansky and Spector 2004; de Swart et al. 2007). First, we can explain why
danseur in (142) does not have the restricted meaning of professional dancer (com-
pare unmodififed BSs, which have only the restricted meaning). (142) can be under-
stood as meaning ‘John is bad when he dances’ and not necessarily as ‘John is a
professional dancer who dances badly’.
(142) Jean est bon danseur.
Jean is good dancer
‘Jean is a good dancer.’
The examples in (143) corroborate our analysis:
(143) a. *Pierre trouve Jean danseur /génie.
Pierre finds Jean dancer /genius
b. Pierre trouve Jean bon danseur.
Pierre finds Jean good dancer
‘Pierre finds Jean [to be a] good dancer.’
c. Pierre trouve Jean bon (comme danseur).
Pierre finds Jean good as dancer
‘Pierre finds Jean [to be] good (as a dancer).’
The acceptability of (143b) is unexpected if bon danseur ‘good dancer’ is ana-
lyzed as a BS, i.e., a bare noun modified by an adjective. This acceptability can be
explained under the proposal made above: bon danseur ‘good dancer’ is an AdjP
headed by bon ‘good’, which is modified by danseur ‘dancer’. Under this analysis,
(143b) is acceptable on a par with (143c), because adjectives (unlike nouns) satisfy
the selectional restrictions of trouver ‘find, consider’.
Our proposal can be extended to cover examples of the type shown in (144),
which are usually analyzed as lexicalizations or idioms. Within our account, they
can instead be analyzed in terms of property modification:
(144) a. Jean est beau/ gentil garçon.
Jean is handsome / kind boy
‘Jean is a handsome/kind boy.’
b. Marie est vieille fille/ jeune grand-mère.
Marie is old girl young grand-mother
‘Mary is a spinster/a young grand-mother’
c. Jean est simple soldat / petit commerçant.
Jean is simple soldier / small shopkeeper
‘Jean is a simple soldier/small shopkeeper.’
2.7.8 The Argument Structure of Relational Nouns
Turning now to relational nouns such as ami, the presence or the absence of the
indefinite article depends in this case on the type of preposition that heads the com-
plement of the relational noun. With PPs such as avec DP ‘with DP’ or à DP ‘to
DP’, which are headed by lexical prepositions, the indefinite article is ungrammati-
cal (as shown in (145a–b)). With PPs such as de DP, built with a semantically empty
preposition, functioning as a marker of Genitive Case, the indefinite article is
obligatory:
(145) a. Jean est (* un) ami avec Pierre.
Jean is a friend with Pierre
‘Jean is friends with Pierre.’
b. * Jean est *( un) ami de Pierre.
Jean is a friend of Pierre
‘Jean is friends with Pierre.’
Kinship nouns are a sub-class of relational nouns, which are special in that their
complement can itself be a bare noun:
(146) a. Jean est ( un) fils d’avocat
Jean is a son of lawyer
‘Jean is a son of lawyer.’
b. Jean est *( le) fils d’un avocat.
Jean is the son of a lawyer
‘Jean is the son of a lawyer.’
Granting that bare nouns denote properties, examples of the type in (146a)
indicate that contrary to current assumptions, the second argument of kinship nouns
need not be saturated by an individual but may also be saturated by a property. In
other words, kinship nouns may denote a relation between an individual and a
property referring to a role. In (142b), fils de diplomate denotes a complex property,
obtained by applying a function (fils ‘son’) to a property (diplomate ‘diplomat’):
(147) a. lx lP son (x, P) (lawyer) j
b. lx son (x, lawyer) j
(147) shows how the complex predicate être fils d’avocat ‘be son of lawyer’ is
built. The kinship noun son is analyzed as a binary predicate, which denotes a rela-
tion between an individual and a property (rather than a relation between two indi-
viduals, as currently assumed). This is indicated by the two lambda abstractions lx
lP in (147a). This binary predicate is applied to the property denoted by lawyer and
to the constant ‘j’ referring to John. (147b) is obtained by reduction from (147a), the
name of role lawyer instanciating the value of P. In (147b) the complex property être
fils d’avocat represented by ‘lx son (x, lawyer)’ is applied to j.
2.7.9 Comparison with Other Approaches
Let us now compare the analysis adopted above with the proposal in de Swart et al.
(2007). Both approaches share the assumption that bare nouns and singular indefinites
have different denotations. However, the distinctions they make are not the same.
On our proposal, bare nouns in predicate position are property-denoting expres-
sions, just like adjectives, which can be directly attributed to individuals, whereas
singular indefinites refer to entities. Accordingly, we posit the existence of two types
of copular sentences, predicational and identity sentences, with two distinct copulas.
In contrast to this, de Swart et al. make no distinction concerning the copulas and
assume the existence of two operators (REL and QUAL) that map kinds to indi-
vidual entities. Following Carlson, they assume that nouns lexically denote kinds of
type k and they use a realization operator REL in order to map a kind to the set of
entities realizing it. Copular sentences built with a singular indefinite would estab-
lish a relation between the entity denoted by John and the kind denoted by teacher
via the use of the realization operator, which mandatorily applies at the NumP level.
In the case of copular sentences built with bare nouns, they assume the existence of
capacity qualifiers, of the type by profession, by nationality, which may be overt or
covert. Capacity qualifiers are analyzed as operators that map the kind k denoted by
the noun to the set of entities realizing the kind as a particular role in society, often
associated with its typical activities. Capacity qualifiers, noted QUAL, are more
restrictive than REL: only entities that realize the kind k in the way that is expressed
by the qualifier are in the set QUAL(k).
The two accounts cover different empirical domains. De Swart et al. restrict their
attention to role names and therefore nothing is said regarding other types of nouns
2.8 Conclusions 93
that can appear bare in predicate positions, e.g., relational or kinship nouns. Note
furthermore that their account, which builds on the lexical properties of role names,
does not seem able to explain the use of role names modified by adjectives such as
bon. Claiming that mauvais médecin ‘bad doctor’ is a role name like médecin ‘doctor’
and that this information is lexically encoded would be an ad hoc and counter-intuitive
solution. The same observation holds for kinship nouns like fils de diplomate ‘son of
diplomat, diplomat’s son’, fils d’avocat ‘son of lawyer, lawyer’s son’, which are unlike
to be stored as such in the lexicon, a proposal that would obscure their compositional
and productive character.
2.8 Conclusions
In this chapter we were led to challenge several widely assumed generalizations and
hypotheses regarding the analysis of bare NPs. The most important observation is
that the use of bare NPs in argument positions should not be viewed as being related
to the use of bare NPs in predicate positions. These are two separate phenomena, for
which separate analyses are needed. This observation strongly suggests that the
property analysis of argumental BPs is misguided. We remain open to the possibil-
ity of analyzing count BSs in Spanish, Romanian or Catalan as property-denoting.
Carlson’s (1977a, c) observations regarding the differences between BPs and singular
indefinites can be attributed to the fact that, due to the absence of an overt Det, BPs
are necessarily weak (in the sense of Milsark 1977), whereas singular indefinites are
both weak and strong. In Sect. 2.4.4 we have suggested that the lack of indefinite
article is related to the denotation: mass and plural nouns refer to amounts (of sub-
stance or of objects), i.e., entities that are ordered by the part-whole relation, whereas
singular count nouns refer to individuals, i.e., entities that are not ordered by the
part-whole relation.
The main theoretical proposal of this chapter is that argumental BPs (as well as
bare mass NPs) should be analyzed as generalized existential quantifiers over
amounts, which are defined in such a way that they need to combine with existential
predicates. In Chaps. 3, 4 and 5 we will show that this analysis extends to non-bare
weak indefinites.
Turning now to bare NPs in predicate positions, let us insist that the interesting
phenomenon concerns only a limited class of nouns: names of roles, professions
and relational nouns. All other nouns occurring in postcopular positions require the
indefinite article. The differences between copular sentences built with BSs as
opposed to ISs point to a necessary distinction between two types of copular sen-
tences (and two types of copulas), which we have labeled attributive and identity
sentences. In identity sentences, the postcopular indefinite denotes an individual,
just as it does in argumental positions. In attributive sentences, the postcopular noun
denotes a property, which explains why ISs are disallowed and BSs allowed, on a
par with adjectives.
Chapter 3
Existential Sentences
It is usually assumed that existential sentences are subject to an “indefiniteness

restriction”: only indefinite DPs can follow sequences such as There is in English or
Il y a in French1; definite and quantificational DPs (cf (2)) are excluded from such
contexts. In this chapter, those French examples that are not glossed are the counter-
parts of the immediately preceding English examples:
(1) a. There was a man in front of the door.
a¢. Il y avait un homme devant la porte.
b. There is a knife on the table.
b¢ Il y a un couteau sur la table.
(2) a. ?There is Fred outside.
a¢. ?Il y a Fred dehors.
b. ?There was the table in the garden.
b¢. ? Il y avait la table dans le jardin.
c. *There was each table in the garden.
c¢. * Il y avait chaque table dans le jardin.
1
In this chapter we will present recent analyses of English existential sentences and see how they
can be refined in order to account for the French data. Beyond the similarities between there be
sentences in English and il y a sentences in French, these constructions also differ in interesting
ways. In addition to the difference in verbs (avoir ‘to have’ in French and to be in English) and the
presence of two pronouns il and y in French, the two constructions differ with regard to agreement:
in French, the verb avoir is always found in the third person singular, regardless of the number
(singular or plural) of the DP argument, whereas in English, BE agrees with the DP argument. We
will not attempt to explain these syntactic differences here. Nor are we going to examine Landman’s
(2003) observation that a definite DP is more easily accepted following il y a in French than fol-
lowing there be in English.

96 3 Existential Sentences
McNally (1998) restates the indefiniteness constraint as a selectional restriction

according to which existential sentences select property-denoting arguments. In this
chapter we will show that McNally’s proposal is problematic and we will adopt
Heim’s (1987) negative constraint: a variable that ranges over individuals cannot
occur in the LF representation of existential constructions. Basing our analysis on
French, we will distinguish three types of existential sentences, a locative existential
construction, an eventive construction and an enumerative construction, which can
be distinguished by their respective codas. Locative existentials are built with a
locative PrepP (note that the examples in (1a–c) are all of the locative type), in even-
tive existentials the coda is an eventive relative clause in French and an eventive
gerund in English and finally, in enumerative existentials the coda is absent or
implicit. These three constructions all obey Heim’s constraint but they nevertheless
differ from each other regarding the type of argument that they allow. The most
restrictive constraint holds for locative existentials, which can only take non specific
amounts as arguments. The argument of eventive existentials is an event participant,
which can be an individual or an amount. Finally, the argument of enumerative
existentials is a list. At the end of the chapter we will examine two types of existen-
tial sentences that seem to constitute counterexamples to Heim’s generalization:
existential sentences built with quantificational arguments and existential sentences
that allow for the relativization of their argument. We will explain why and how
such sentences obey Heim’s constraint.
3.1 Constraints on Existential Sentences
3.1.1 Existential Sentences Have Property-Denoting Arguments
3.1.1.1 The Semantic Composition of Existential Sentences according

to McNally (1998)
McNally’s (1998) analysis builds on a remark due to Strawson (1959: 241), who
says that existential sentences can be construed as propositions of the form subject-
predicate “in which the subject is a property or concept and in which the predicate
declares, or denies, its instantiation.” Accordingly, McNally proposes that there-
sentences are built with a predicate of instantiation that applies to a property. Thus
the sentence in (3) is true if and only if the property snow, represented in (4) as ly
snow(y), is instantiated.
(3) There is snow.
(4) to be instantiated (ly snow(y))
As McNally points out, a property is instantiated if and only if there is a particu-
lar of which it is true. To be instantiated is reformulated in (5a) so as to show that it
3.1 Constraints on Existential Sentences 97
is this predicate that introduces existential quantification: to be instantiated

translates as lP [$x P(x)], i.e., the set of properties P such that there is a particular
x for which P is true. We go from (5a) to (5b) and then to (5c) via two applications
of the rule of l-reduction. In (5a), ly snow(y) represents the set of amounts of mat-
ter that are snow. (5b) is obtained by substituting the value ly snow(y) for P in $x
P(x) in (5a). And finally (5c) is obtained from (5b) after reduction of ly snow(y) (x),
which yields snow(x).
(5) a. lP $x P(x) (ly snow(y))
b. $x (ly snow(y)) (x)
c. $x snow(x)
According to this analysis, there is snow is true if and only if a particular, namely
a certain quantity of snow, exists. It is the logical counterpart of there is that is
responsible for introducing existential quantification, not the postverbal DP snow.
This representation of there sentences should be distinguished from the tradi-
tional formalization (6b), which is still often used, according to which (3) contains
a predicate of existence (there is) and an existential quantification over particulars
(corresponding to snow). In (6a), the unary predicate lx exist(x) is the logical coun-
terpart of there is and the generalized quantifier lP $y [snow(y) Ù P(y)] translates
the expression snow. The existential quantifier $, which quantifies over a particular,
is part of the representation of snow. This is particularly clear in (6b), which is the
reduction of (6a).2 Note that (6b) is somewhat redundant, as the predicate exist
duplicates information given by the existential quantifier $.
(6) a. lP [$y [snow(y) Ù P(y)]] (lx exist(x))

b. $y [snow(y) Ù exist(y)]
Going back to McNally’s (1998) analysis, it explains the restrictions on the dis-
tribution of postverbal DPs in existential constructions by means of a constraint on
the semantic type of the DP3:
2
To reduce (6a), P in $y [snow(y) Ù P(y)] is replaced by the predicate lx exist(x). We thus obtain
(i). y is then substituted for x in [exist(x)], yielding (6b).
(i) [$y [snow(y) Ù (lx exist(x)) (y)]
3
McNally views the constraint in (7) as a selectional restriction comparable to the requirement of
plural referents imposed by predicates such as gather, which can be satisfied by morphological
plurals (ii) but also by singular DPs that denote groups (iii). Example (i) is excluded, not because
the subject is singular but rather because it denotes an atomic individual:
(i) *A girl is gathering around the table.
(ii) The girls are gathering around the table.
(iii) The dense crowd is gathering around the building.
(7) McNally’s (1998) constraint

Sentences of the form There be DP W are well-formed only if DP denotes a
property (in other words, DP is of type < e,t>).
It follows from the formulation in (7) that There be DP W and Il y a DP W are
ungrammatical when DP denotes an individual or a generalized quantifier.4
Summarizing, according to McNally (1998), the DP in an existential construc-
tion does not introduce a discourse referent but rather a property. It is the predicate
there be / il y a that introduces existential quantification, as well as an individual
variable. To assert existence is to predicate a property of another property, i.e., to
assert that a property is instantiated. Such an assertion is about a property and only
indirectly about the individuals of which that property is true.
3.1.1.2 Negative Existential Sentences
We have already noticed that McNally’s formalization of There is / Il y a sentences

should be distinguished from the traditional formalization of existential sentences,
given in (6b) and repeated here in (8), which involves a predicate of existence (there
is / il y a) and existential quantification over particulars (corresponding to snow / de
la neige).
(8) $y [snow(y) Ù exist(y)]
This traditional analysis is problematic for negative sentences: the DP snow

introduces existential quantification in the logical form, which contradicts the nega-
tion of the predicate exist, which translates there is no. In sum, the Logical Form
(9b) conveys a contradiction:
(9) a. There is no snow.

b. $y [snow(y) Ù ¬ exist(y)]5
McNally’s analysis avoids the paradox of negative existential sentences because

in both positive and negative existentials, snow is analyzed as denoting a property
and does not introduce any existential quantification.
4
Apparent exceptions will be examined later in this chapter: the case of definite descriptions in
Sect. 3.3 and the case of quantified expressions in Sect. 3.4.
5
If negation took scope over the entire sentence, we would have the following representation:
(i) ¬ $y [snow(y) Ù exist(y)]
This representation is also somewhat paradoxical due to the simultaneous presence of the negated
existential quantifier and the predicate exist. It could be paraphrased as: it is false that there exists
snow that exists.
(10) a. There is snow.

b. lP $x P(x) (ly snow(y))
(11) a. There is no snow.
b. lP ¬$x P(x) (ly snow(y))
3.1.2 Problems with the Property Analysis
3.1.2.1 Definite DPs
On McNally’s account, definite DPs are expected to be ruled out of existential

constructions as a consequence of the fact that they are of type e, rather
than < e,t>. However, McNally acknowledges the occurrence of definite DPs in
existential sentences. This can be seen in the English example in (12), due to
Lambrecht (2002), and its French equivalent in (13). Furthermore, she notes
that in languages like Catalan, definite DPs are always acceptable in existential
constructions (cf. (14))
(12) There is the mailman coming.
(13) Il y a le facteur qui arrive.
there has the mailman who arrives
‘There is the mailman coming.’
(14) Hi havia la Joana a la festa.
there has the Joana at the party
‘Joan was at the party.’
To account for these examples, McNally resorts to type-shifting. Ever since
Montague, it is well established that a proper name can denote either something of
type e (a constant), or a set of properties, of type < <e,t>,t>. Jean translates either as
the constant j in (15a), or as the set of properties that Jean has in (15b). Partee and
Rooth (1983) and Partee (1987) have furthermore shown that it is possible to multi-
ply type-shifting rules and thus go from a type e denotation to one of type <e,t>.
Jean may be represented as the set of individuals that are identical to Jean, as shown
in (15c).
(15) a. j
b. lP (P(j))
c. lx (x = j)
Accordingly, for McNally, any referential expression of type e can type-shift to
an expression of type <e,t>. This would account for the occurrence of definite
descriptions in existential sentences, by simply assuming that they are subject to the
above-mentioned type-shifting rule.
According to McNally, the marginal acceptability, viz. the unacceptability, of

definite DPs and proper names is not due to the selectional constraint (which allows
them, due to type-shifting) but rather to a pragmatic novelty condition on the refer-
ent denoted by the postverbal argument: There be can only be used in a context C
if the DP serving as its argument introduces a novel discourse referent. This prag-
matic constraint adds up to the semantic selectional constraint according to which the
DP argument of existential sentences denotes a property. The pragmatic condition
accounts for the contrast between (16) and (17):
(16) a. *There was the table in the garden.
a¢. *Il y avait la table dans le jardin.
(17) a. There was a table in the garden.
a¢. Il y avait une table dans le jardin.
b. There were several tables in the garden.
b¢. Il y avait plusieurs tables dans le jardin.
Indefinite DPs are frequent in existential sentences because, as shown by Heim
(1982), they always introduce a new discourse referent. Definite DPs, on the other
hand, do not usually meet the novelty condition. Instead, they refer back to a dis-
course referent that has already been introduced or is part of the common
ground.
McNally claims that definite DPs are excluded of existential sentences because
they never introduce new entities. But this is not exact. The difference between
definiteness and indefiniteness does not perfectly correlate with that between
familiarity and novelty. Indeed, there are instances of definite DPs that introduce
new discourse referents, in particular, what Prince (1981: 235) calls an ‘unused’
new entity.6 An unused entity is a referent that the addressee is assumed to be
acquainted with (e.g., a contextually salient referent) but which she has not used
yet. This is illustrated with the definite descriptions le chat / the cat and the proper
name John in (18) and (19), which can be uttered out of the blue:
(18) a. Have you seen the cat in the yard?
a¢. As-tu vu le chat dans la cour ?
(19) a. Peter went to Paris. He saw John over there.
a¢. Peter est allé à Paris. Il a vu John là-bas.
6
Prince distinguishes between two kinds of new discourse referents: one kind is “brand new” dis-
course referents, corresponding to cases where the addressee must create a new entity and intro-
duce it into the universe of discourse; the other kind is ‘unused’ discourse referents, corresponding
to cases where the addressee already has a discourse referent in his/her own model and only needs
to introduce or copy it into the discourse-model. The referent does not need to be created but
merely activated, as it were.
The examples in (20)–(21) show that despite their being definite, DPs that refer
to ‘unused new’ entities may be used in existential sentences:
(20) a. There is the cat in the yard.
a¢. Il y a le chat dans la cour.
(21) a. There is John over there.
a¢. Il y a John là-bas.
Another case in point, observed by Comorovski (1995), involves examples rely-
ing on the mechanism known as ‘cross-reference’. This can be observed with cer-
tain definite DPs, e.g., its author / son auteur in (22), or the son of the neighbor / le
fils de la voisine in (23):
(22) a. John read [a book on Schubert]i and wrote to [itsi author]j.
a¢. John a lu [un livre sur Schubert]i et a écrit à [soni auteur]j.
(23) a. Jean has [a neighbor who plays the piano]i. As for [the neighbor’si son]j,
he plays the violin.
a¢. Jean a [une voisine qui joue du piano]i. Quant au [fils de la voisinei]j, il
joue du violon.
In (22), its author/son auteur introduces a novel discourse referent but this refer-
ent is dependent on the DP a book on Schubert/un livre sur Schubert. Similarly in
(23), the son of the neighbor/le fils de la voisine introduces a discourse referent not
yet present in the context but the referent depends on another referent that has
already been mentioned (i.e., the neighbor).
Cross-reference explains why definite descriptions with an indefinite comple-
ment, like the French DP le fils d’une voisine or the English the son of a neighbor,
are allowed to appear in existential sentences (see Jackendoff (1974) for English
and Milner (1982) for French).
(24) a. There was the son of a neighbor at the party.
a¢. Il y avait le fils d’une voisine à la fête.
In sum, the study of existential sentences calls attention to definite DPs that refer
to new entities: although they refer to new discourse-referents, these DPs are definite
either because their referents are salient or ‘active’ in the extralinguistic context, or
because they are introduced by means of cross-reference (see Chap. 8 on dependent
DPs).
McNally’s account relies on two constraints: a semantic one, on the (semantic)
type of the DP argument in existential sentences, and a pragmatic one, on the nov-
elty of the discourse referent associated with the DP. The two constraints are of a
very different nature: the former rules out quantified expressions from existential
sentences, the latter explains why definite DPs are often less acceptable than
indefinite DPs. We will not pursue this analysis, which makes use of type-shifting
rules without restricting the range of application of these rules. Rather, we seek to
avoid type-shifting rules as much as possible. We consider it is always possible to
have a type-lifting operation that applies to a constant, i.e., an expression referring
to an individual and returns an expression of type < e,t>. To put it differently, in our
view, a constant can either refer to an individual, or to a set of properties. But we do
not allow for any other type-shifting operations and we think that type-shifting can
only result from the application of a covert semantic operator.
Let us now observe that the presence of definite DPs in existential sentences is
problematic not only for McNally’s analysis but also for Milsark’s (1977) view
according to which the arguments of there-sentences are necessarily weak indefinite
DPs (see Chaps. 4 and 5 below). This problem can be avoided by extending the
typology of French existential sentences to English. As shown in Sect. 3.2.2 below,
definite DPs can freely occur in eventive existentials (i.e., those existentials that
have an eventive relative (in French) or gerund (in English) as a coda) but not in
locative existentials (i.e., those existentials that have a locative coda).
3.1.2.2 Adjectives
A simple and yet compelling argument against the property analysis of the argu-
ment of existential sentences (and more precisely locative existentials) comes from
the fact that adjectives, which are typically property-denoting expressions and
which are always felicitous in predicate positions, are excluded from existential
sentences. This generalization holds cross-linguistically and in particular in both
English and French:
(25) a. * There is {happy/red}.
a¢. * Il y a {heureux/rouge}.
(26) a. She was {happy/red}.
a¢. Elle était {heureuse/rouge}.
The clear contrast illustrated in (25)–(26) is unexpected under McNally’s property
analysis. In fact, McNally’s analysis of existential sentences is a particular case of a
more general hypothesis according to which weak DPs are property-denoting
expressions (McNally (1995a, b, 1998), van Geenhoven (1996), McNally and van
Geenhoven (1998), Dobrovie-Sorin (1995, 1997a, b), Ladusaw (1994), a.o.), which
we have shown to be problematic in section 4.2 of Chap. 2 above. Note that McNally
might have argued that the arguments of existential sentences are weak DPs of a
special, more constrained sort. However, the hypothesis of property-denotation does
not seem to be adequate and the simplest view is to go back to Milsark’s (1977)
observation that the arguments of existentials must be weak DPs (see Chaps. 4 and
5 of this book).
In order to further understand Milsark’s generalization we need to define weak
DPs and explain why only weak DPs are allowed as arguments of existential sen-
tences (as we will see below, this requirement holds for locative existentials but not
for the other two types of existentials discussed in Sect. 3.2 below).
3.2 Existential Sentences in French 103
3.1.3 Our Proposal in a Nutshell
The conclusions in (27a) and (27b) can be respectively drawn from our examination
of existential sentences built with definite DPs and adjectives:
(27) a. The property analysis cannot be a constraint on all types of existen-
tial sentences.
b. The property analysis cannot be a constraint on locative existential
sentences.
In the remainder of this chapter we will attempt to show that existential sentences
are of different types and that there is a correlation between the type of DP and the
type of existential sentence:
(28) a. Locative existentials have weak DPs as arguments.
b. Definite DPs are allowed in eventive and enumerative existentials.
We will also make explicit the semantic composition of locative existentials
based on the hypothesis stated in (29), proposed in Chap. 2 and further motivated in
Chaps. 4 and 5:
(29) Weak DPs refer to non specific amounts.
3.2 Existential Sentences in French
In this section, we will concentrate on French il y a existentials and propose,

following Giry-Schneider (1988), Comorovski (1995) and Lambrecht (2002), that
three types of existential sentences must be distinguished: a locative existential con-
struction, an eventive construction and an enumerative construction,7 which can be
distinguished by their respective codas. Locative existentials are built with a locative
7
We do not adopt Comorovski’s (1995) distinction between the existential and the presentational
readings of there is sentences. For Comorovski, the existential reading corresponds to an asser-
tion of existence, as in (i), and the presentational reading, to the introduction of a new referent, as
in (ii):
(i) Il y a peu de neige dehors.

‘There is little snow outside.’
(ii) Il y a beaucoup de gens qui meurent de froid dehors.
‘There are many people freezing to death outside.’
According to Comorovski, these two readings correspond to different ways of calculating the
semantic value of the sentence. Under the existential reading, the denotation of the predicate struc-
ture is calculated and combined with the meaning of there be. Under the presentational reading, the
denotations of be and the coda XP combine with each other before combining with the denotation
of the DP. The expletive “there” contributes nothing to the meaning of the sentence.
coda. Eventive existentials are built with an eventive coda, more precisely an even-
tive relative clause or an eventive present participle in French (the English counter-
parts of this type of existential are generally built with a gerund). Finally, enumerative
existentials do not need a coda and when a coda is present, it does not introduce new
information; enumerative constructions are characterized by the fact that their argu-
ment is an enumeration.
3.2.1 The Locative Existential Construction
Locative existential constructions are of the type shown in (30):

(30) a. Il y avait autrefois une belle princesse qui vivait dans un
there has.impf once a beautiful princess who live.impf in a
grand château
big castle
‘Once there was a beautiful princess who lived in a great castle.’
b. Il y a de la neige sur les toits.
there has de la snow on the roofs
‘There is snow on the roofs.’
Here, the existential construction is used to introduce a new discourse referent
that will be the theme of the discourse that follows. Often, the discourse referent is
linked to worlds and times that are also unidentifiable (an imaginary world, an unan-
chored time like un jour ‘one day’ or autrefois ‘once’). This type of sentence is
often found in narratives.
Locative existential constructions are incompatible with definite DPs, as can be
seen by comparing the judgments in (30) with those in (31):
(31) a. * Il y avait autrefois la belle princesse qui vivait dans
there has.impf once the beautiful princess who live.impf in
un grand château
a big castle
‘Once there was the beautiful princess who lived in a great castle.’
b. * Il y a la neige sur les toits.
there has the snow on the roofs
‘There is the snow on the roofs’
The sentence asserts the existence of an entity at a certain location. When the
coda is empty, as in (32), an implicit locative may be reconstructed: ici ‘here’ in
(32a) or sur la terre ‘on earth’ in (32b–c). This can be explained if we assume that
the instantiation predicate, there be or il y a, is an episodic predicate that contains a
situation or a location in its argument structure (Kratzer 1988). If this argument is
not expressed, then it must be recoverable from the context.
(32) a. Il y a un problème.
there has a problem
‘There is a problem.’
b. Il y a un Dieu.
there has a God
‘There is a God.’
c. Il n’y a pas de justice.
there neg has neg de justice
‘There is no justice.’
On our analysis, the existential predicate is a localization predicate, which can
only combine with an amount-referring expression. Such a predicate asserts the
existence of an entity at a certain location; the localized entity individualizes the
amount-referring DP with which the existential predicate combines.
Our proposal for the semantic composition of existential sentences is given
below. Consider sentence (33). It will be associated with the Logical Form (34),
which is obtained by combining the locative existential predicate ‘il y a x sur la
table’ with the weak DP ‘de l’eau’. The detail of the semantic composition is given
in (35): as indicated in (35c), the generalized quantifier denoted by ‘de l’eau’ (see
(35a)) applies to the property denoted by il y a x sur la table (see (35b)). In (34) and
(35), Ind is an individuating relation by which an amount (which instantiates a cer-
tain property) is individuated by identification to a localized entity. (34) is equiva-
lent to (35c) and is obtained by applying to (35c) two lambda-conversions, which
are detailed in (35d) and (35e).
(33) Il y a de l’eau sur la table.
there has de water on the table
‘There is water on the table.’
(34) $ xi $ xa (Water (xa) Ù Ind (xi,xa) Ù (xi is at Loc))
(35) a. de l’eau = lP $ xa (water (xa) Ù P(xa))
b. il y a x sur la table = lxa $ xi (Ind (xi,xa) Ù (xi est sur la table) )
c. lP $ xa (water (xa) Ù P(xa)) • lxa $ xi (Ind (xi,xa) Ù (xi est sur la table))
d. $ xa (water (xa) Ù lxa $ xi (Ind (xi,xa) Ù (xi est sur la table)) (xa))
e. $ xa (water (xa) Ù $ xi (Ind (xi,xa) Ù (xi est sur la table)))
According to this analysis, existential sentences are not built by saturating a

predicate with an argument. Instead, the argument of an existential sentence is
necessarily treated as a generalized quantifier that applies to an existential
predicate.
3.2.2 The Eventive Construction
In addition to locative existentials, we can identify a construction that we will refer

to as an eventive existential:
(36) a. Il y a le téléphone qui sonne.
there has the phone that rings
‘There is the phone ringing.’
b. Il y a le chat qui meurt de froid dehors.
there has the cat that dies of cold outside
‘There is the cat freezing to death outside.’
c. Il y a la fille de la voisine qui cueille des fleurs dehors.
there has the daughter of the neighbor that picks des flowers outside
‘There is the neighbor’s daughter out picking flowers.’
This construction does not assert the existence of a new entity but rather of a
new event; a new entity may nevertheless be introduced in an argument position
as a participant to the event. In French, this kind of existential sentence may
constitute an answer to the question Qu’est-ce qu’il y a ? (‘what is the matter?’)8
and il y a can sometimes be paraphrased by il y a que, as in (37), which corre-
sponds to (36)a:
(37) Il y a que le téléphone sonne.
There has that the phone rings
‘It’s that the phone is ringing.’
As noted by Lambrecht, the utterances in (36) are not used to inform the addressee
about the existence of a telephone, a cat or a neighbor’s daughter, nor do they
call attention to these referents but instead they are meant to inform the addressee
8
For this reason, it can be said that it is the proposition associated with DP W that is the focus in
the eventive construction, as opposed to the proper existential construction, where the DP is the
focus. Lambrecht (2002) claims that in the locative existential construction, W counts as a second-
ary predicate.
of an event.9 And it is only insofar as they are participants in the event that the tele-
phone, cat and neighbor’s daughter are mentioned. No semantic constraint bears on
these event participants; they may refer to individuals or amounts and they may be
new or familiar.
In eventive constructions, the coda is obligatorily overtly expressed. This is easy
to understand, since it is the coda itself that supplies an event variable and a restric-
tion on its range:
(38) a. ? Il y a le téléphone.
there has the phone
‘There is the phone.’
b. ? Il y a le chat.
there has the cat
‘There is the cat.’
c. ? Il y a la fille de la voisine.
there has the daughter of the neighbor
‘There is the neighbor’s daughter.’
According to Lambrecht, the coda in this type of existential sentence is a “pre-
sentational relative construction”.10
Note furthermore that the coda predicate must denote a transitory, or stage-level,
property.
9
It may be noted that the eventive reading can also be obtained by replacing the expletive il and the
clitic y with a personal pronoun:
(i) J’ai eu mon beau-frère qui a fait Paris-Nice.

I have had my brother-in-law who has done Paris-Nice
‘I had my brother-in-law who competed in Paris-Nice’
(ii) Il y a eu mon beau-frère qui a fait Paris-Nice
There has had my brother-in-law who has done Paris-Nice
‘There was my brother-in-law competing in Paris-Nice.’
10
Presentational relative constructions are not restricted to il y a sentences but are also present in
example (i) given in the preceding footnote and in (i)–(ii) below. All these examples are from
Lambrecht (2002):
(i) Je vois le facteur qui arrive

I see the mailman who arrives
‘I see the mailman coming.’
(ii) Voilà le facteur qui arrive.
Look the mailman who arrives
‘Here comes the mailman.’
(39) a. Il y a les enfants de la voisine qui jouent dehors.

there has the children of the neighbor that play outside
‘There are the neighbor’s children playing outside.’
b. ?? Il y a les enfants de la voisine qui sont intelligents.
there has the children of the neighbor that are intelligent
‘There are the neighbor’s children intelligent.’
Eventive existentials clearly differ from locative existentials in that their argu-
ment need not be a weak DP; individual-referring expressions (type e) such as
proper names, as well as definite or demonstrative DPs are also allowed:
(40) a. Il y a {un/le} téléphone qui sonne.
There has a/the phone that rings
‘There is {a/the} telephone ringing.’
b. Il y a Jean qui sonne.
There has Jean that rings
‘There is Jean ringing.’
The possibility of eventive existentials taking both weak and strong DPs in their
argument position is due to the fact that their coda contains an eventive predicate,
which is ambiguous, translating either as an existential predicate or as an entity-
predicate and correlatively being able to combine with both weak and strong DPs.
To illustrate this point, let us consider the two eventive existential sentences below
and their Logical Forms:
(41) a. Il y a Jean qui dort.
‘There is Jean sleeping’
b. $ e (sleep (e) Ù Subject (e, j))
(42) a. Il y a des enfants qui dorment.
‘There are children sleeping’
b. $ e $ xi (sleep (e) Ù Subject (e, xi) Ù Ind (xi,xa) Ù des enfants (xa))
The eventive existential predicate il y a is used to assert the existence of the
event described in the proposition to which it applies. In the case of (41), the predi-
cate sleep is analyzed as an entity-predicate that combines with the proper name
Jean, whereas in (42), the predicate sleep is analyzed as an existential predicate
that introduces an existential quantifier on one of the participants in the event, as
shown in (42b).
It is interesting to compare existential sentences and thetic sentences with pre-
verbal subjects. Ulrich (1985) and Sasse (1987) propose that there are two types of
thetic sentences, labeled entity-central and event-central, which are illustrated in
(43a) and (43b) respectively:
(43) a. Un gros nuage apparaît à l’horizon.

a large cloud appears at the horizon
‘A large cloud is appearing on the horizon.’
b. Il fait froid.
it makes cold
‘It is cold.’
In the thetic sentence in (43a), a novel entity is introduced as a participant to the
event denoted by the main predicate: the verb apparaître ‘to appear’, which is an
eventive verb, cannot be replaced by a stative verb, such as être ‘to be’ or se trouver
‘to be situated’, as illustrated by (44a–b):
(44) a. * Un gros nuage se trouve à l’horizon.
a large cloud refl situate at the horizon
‘A large cloud is situated on the horizon.’
b. * Un gros nuage est à l’horizon.
a large cloud is at the horizon
‘A large cloud is on the horizon.’
Existential sentences are of two types, eventive and locative. Eventive existential
sentences correspond to the acceptable entity-central thetic sentences (see (45a) and
(43a)), whereas locative existential sentences correspond to the unacceptable ones
(see (45b) and (44a–b)):
(45) a. Il y a un gros nuage qui apparaît à l’horizon.
There has a large cloud that appears at the horizon
‘There is a large cloud appearing on the horizon.’
b. Il y a un gros nuage à l’horizon.
There has a large cloud at the horizon
‘There is a large cloud on the horizon.’
(45a) can be said to be eventive, whereas (45b) is a locative existential sentence
In sum, at this point, we can distinguish between two types of existential con-
structions in French, locative existentials and eventive constructions. In the follow-
ing table, we present their respective properties (Table 3.1):
Table 3.1
Locative existential sentences Eventive sentences
Coda Can be implicit Necessarily explicit
Locative Eventive
DP Weak DP No restriction
Novel Type e
Unlike McNally, who analyzes existentials with definite DPs as cases of locative
existential constructions with type-raising of the definite DP (from e to <e,t>), we
believe that such sentences are eventive existentials, a type of construction that does
not constrain the postverbal DP, which can be not only indefinite but also definite
(see Sect. 3.3.2 on the use of certain quantified DPs).
Giry-Schneider (1988) and Comorovski (1995) both maintain that existential
sentences built with locative codas are ambiguous, invoking, respectively, the fol-
lowing examples:
(46) a. Il y a un lion sur le canapé.

there has a lion on the couch
‘There is a lion on the couch.’
b. There is a woman in the house.
(46a) is ambiguous between an eventive reading, e.g., as an answer to the ques-
tion Qu’est-ce qu’il y a ? ‘What is the matter?’ and a purely existential reading;
correlatively, the locative PP serves to localize an event or an entity. Comorovski
claims that (46b) is ambiguous in the same way. There seem to be two different
stress patterns, noted by Rando and Napoli (1978), which correspond to the two
different readings: when only woman is stressed, the sentence has an existential
reading; when stress is placed on both woman and house, it has an eventive read-
ing. A study of French prosody might give similar results. We believe that this kind
of ambiguity is in fact quite rare. When the coda is a locative, an eventive reading
(i.e., as an answer to the question ‘What is happening?’) is only possible if it is
truly exceptional for the referent denoted by the DP to be in the place described in
the coda.
3.2.3 The Enumerative Construction
The enumerative construction is found in both French and English but it is more
frequent in French:
(47) a. Il y a Marie qui viendra, il y a Jean aussi, un copain de Jean et sans doute
toi aussi.
a¢. Marie is going to come, and there’s also Jean, a friend of John’s and prob-
ably you.
b. Je crois qu’on a appelé tout le monde. Non, il y a encore Marie et Jean.
b¢. I guess we’ve called everybody. No, there is still Mary and John.
Enumerative constructions are usually used as answers to questions and their
coda takes up the question itself, e.g., (47a) can be an answer to Qui viendra ‘Who
will come?’. Generally speaking, in enumerative constructions, the coda does not
convey new information: it is not included in the focus, its content corresponds to
the discourse topic. Consequently, even if enumerative sentences without a coda
3.3 Existential Sentences Cannot Have Individual Variables as Arguments 111
can be found (see (47b)), in fact the coda is not empty but elliptic: it can easily be
recovered from the context. Thus, in (47b), an elliptic coda takes up the discourse
topic, which in this context can be ‘Whom do we have to call?’ The argument of
enumerative existentials is a list that can contain any type of DP, in particular proper
names or definite DPs.
3.2.4 Conclusions
To sum up, we have distinguished three types of existential constructions in French,

which can be characterized by their respective codas: locative existentials have loc-
ative codas, eventive existentials have eventive codas and enumeratives have a coda
that corresponds to a discourse topic. Concerning the constraints on the DPs occur-
ring in these constructions, Milsark’s generalization seems adequate for locative
existentials but not for eventives or enumeratives. Unlike locative existentials, which
can only combine with weak DPs, eventives and enumeratives are compatible with
entity-denoting DPs, e.g., definite DPs, as well as proper names.
3.3 Existential Sentences Cannot Have Individual

Variables as Arguments
3.3.1 Heim’s (1987) Constraint
Heim (1987) observed that personal pronouns cannot appear in English existential
constructions. This generalization extends to il y a-sentences in French. The exam-
ples in (49) are the French counterparts of the immediately preceding English
examples:
(48) a. Few people admitted that they had been at the party.
b. *Few people admitted that there had been them at the party.
(49) a. Peu de gens ont reconnu qu’ils étaient venus à la fête.
b. *Peu de gens ont reconnu qu’il y avait eu eux à la fête.
Based on examples of the type in (48b) and (49b) and assuming that personal
pronouns typically translate as individual variables, Heim (1987:23) proposed the
constraint in (50):
(50) Heim’s constraint
There be x is ungrammatical where x is an individual variable.
In what follows we will provide further evidence for this constraint by examining
existential sentences built with quantificational DPs on the one hand and existential
sentences inside relatives on the other hand.
3.3.2 Quantified DPs in Existential Sentences
3.3.2.1 Quantification over Types Rather Than Tokens
Given Heim’s constraint, quantified DPs are predicted to be excluded from existential
sentences: at LF, quantified DPs are moved (via Quantifier Raising) out of their
argument position, leaving behind an individual variable, which yields a violation
of Heim’s constraint. However, existential sentences built with quantified DPs can
be found, as observed by Lumsden (1988), who gave the following examples:
(51) a. There was every kind of doctor at the convention.
b. There were most sorts of books in his library.
c. There were both varieties of wine for sale.
d. There was each kind of question on the exam.
In all of these examples, the postverbal DP contains a noun like kind, sort or
variety.11 The examples become ungrammatical when the terms kind, variety, type…
are removed:
(52) a. *There was every doctor at the convention.
b. *There were most books in his library.
c. *There was each question on the exam.
These data show that quantified DPs are not ruled out as such from appearing in
the argument position of an existential construction. Rather, their acceptability
depends on the descriptive content of the noun, which must refer to subkinds. DPs
such as a kind of doctor, a sort of book and a kind of question do not denote indi-
viduals but classes of individuals. In (51a), quantification is not over the set of
individuals that are doctors but over the set of medical specializations (general prac-
titioners, surgeons, dentists…); in (51b), the domain of quantification is not the set
of books but rather the set of literary genres; and in (51c), the domain of quantification
is not the set of actual questions, but rather the set of types of questions.
Similar observations hold in French. It is possible to construct minimal pairs
similar to Lumsden’s by replacing the postverbal N with toutes sortes de N
(‘every sort of N’) or toute espèce de N (‘every kind of N’) or tout type de N
11
For a more detailed analysis of the word kind and for more details on the differences between
every kind of N and an N of every kind, cf. Wilkinson (1991, 1995). The differences seem to disap-
pear in existential sentences. This seems to be true as well for the expressions Det genre de N and
(un/des) N de Det genre in French:
(i) Il y avait ce genre de médecins à la réunion.

‘There was that sort of doctor at the convention.’
(ii) Il y avait des médecins de ce genre à la réunion.
‘There were doctors of that sort at the convention.’
(‘every type of N’).12 Note that these expressions convey indefiniteness. Toutes
sortes de médecins (‘every sort of doctor’) is equivalent to des médecins de toutes
sortes (‘doctors of every sort’), tous les types de questions (‘every type of ques-
tion’) can be replaced by des questions de tous les types (‘questions of every
type’) and toute espèce de gens (‘all kinds of people’) means the same as des
gens de toutes espèces (‘people of all kinds’).
(53) a. Il y avait toutes sortes de médecins à la réunion.
There have.impf all sorts of doctors at the convention
‘There was every sort of doctor at the convention.’
b. Il y avait des médecins de toutes sortes à la réunion.
There have.impf des doctors of all sorts at the convention
‘There were doctors of every sort at the convention.’
c. ? Il y avait tous les médecins à la réunion.
There have.impf all the doctors at the convention
‘There were all the doctors at the convention.’
(54) a. Il peut y avoir tous les types de questions à l’examen.
There may have all the types of questions at the exam
‘There may be any type of question on the exam.’
b. Il peut y avoir des questions de tous les types à l’examen.
There may have des questions of all the types at the exam
‘There may be questions of any type on the exam.’
c. ? Il peut y avoir toutes les questions à l’examen.
There may have all the questions at the exam
‘There may be all the questions on the exam.’
12
According to Vaugelas, with toute sorte de ‘all sorts of’, followed by a complement, the
singular is used if the noun is singular and the plural is used if the noun is plural:
(i) Je vous souhaite toute sorte de bonheur.

I you wish every sort of happiness
‘I wish you all sorts of happiness.’
(ii) Dieu vous préserve de toutes sortes de maux.
God you protect of every.pl sorts of evils
‘May God protect you from all sorts of evil.’
But this is not always the case: thus, for example, we may find toutes sortes de gibier ‘all sorts of
game’ in the writings of R. Bazin, or toute sorte de charges ‘every sort of charges’ in the writings
of A. Suarès.
(55) a. On rencontre toute espèce de gens dans ces pays. (H. Bosco)
3sg meets every kind of people in these countries
‘One meets all kinds of people in these countries.’
b. On rencontre des gens de toutes espèces dans ce pays.
3sg meets des people of all kinds in this country
‘One meets people of all kinds in this country.’
c. ? On rencontre tous les gens dans ce pays.
3sg meets all the people in this country
‘One meets all people in this country.’
More interestingly, French exhibits grammatical examples such as (56a), where
the postverbal argument is quantified and nouns such as type, sort or kind do not
appear.13
(56) a. Il y avait {la plupart des /tous les} livres de Beckett dans
there have.impf the majority of the/all the books of Beckett in
cette librairie.
this bookstore
‘There were {most /all of} Beckett’s books in that bookstore.’
b. * Il y avait {la plupart des/ tous les} livres dans
there have.impf the majority of the/ all the books in
cette librairie.
this bookstore
‘There were {most /all of} books in that bookstore.’
c. * Il y avait {la plupart des/ tous les} livres reliés dans
there have.impf the majority of the/all the books bound in
cette librairie.
this bookstore
‘There were {most/all of} bound books in that bookstore.’
The contrast between (56a) vs. (56b–c) deserves an explanation. In (56a), the
nominal predicate livre de Beckett ‘book by Beckett’ applies to book titles and not
to copies of books. (56a), with la plupart / most, does not mean that most copies of
Beckett’s books were in the bookstore but rather that for most books by Beckett, at
least one copy can be found in the bookstore. Livre de Beckett does not refer to an
13
Note however an interesting contrast between tous les (‘all the’) and tout (‘every’) or chaque
(‘each’) in French. Tout and chaque cannot easily appear in existential constructions, even when
they are followed by a noun that denotes a kind instead of a token.
(i) * Il y avait tout livre de Beckett dans cette librairie.

there have.impf every book of Beckett in this bookstore
‘There was every book by Beckett in that bookstore.’
(ii) ?? Il y avait chaque livre de Beckett dans cette librairie.
there have.impf each book of Beckett in this bookstore
‘There was each book by Beckett in that bookstore.’
object, a token but to a title, a type. Livre de Beckett is used in this sentence as a
predicate of kinds, not as a predicate of objects. Beckett’s books refers to a kind,
divided into subkinds corresponding to each title. Livre de Beckett refers to a set of
subkinds: the set of titles of texts written by Beckett, with each title corresponding
to a subkind, whose members are copies of that title. (56a) means that for most titles
of Beckett, there is at least one copy of this title (i.e., one instantiation of the sub-
kind corresponding to the title) in the bookstore.
The contrast between (56a) on the one hand and (56b) and (56c) on the other
hand shows that reference to book-types rather than to book-tokens requires a
restriction on the noun livre and not just any restriction will do.
In sum, quantified DPs are not entirely excluded from existential constructions.
They are acceptable, provided that they do not quantify over individuals but over
subkinds, types or sorts.
3.3.2.2 Representation
To account for the grammaticality of examples such as (51), McNally proposes to

identify kinds with property-denoting DPs. According to her, the contrast in (57) is
predictable, if one considers the Logical Forms (58a) and (58b), respectively associ-
ated with (57a) and (57b). Tripartite representations – of the form Quantifier (restric-
tion) [scope] – allow us to call attention to the type of the variable bound by the
quantifier, which occurs both in the restriction and in the scope. Individual variables
are represented by lowercase letters (x, y, z…) and uppercase letters (P, Q…) repre-
sent predicate variables.
(57) a. There was every kind of doctor at the convention.
b. *There was every doctor at the convention.
(58) a. ∀P (kind-of-doctor (P)) [there was P at the convention]
b. ∀x (doctor (x)) [there was x at the convention]
According to McNally, the Logical Form in (58b) is ill-formed because it
violates the constraint on the semantic type of the argument of there be, which has
to be a property (type <e,t>) and not an individual of type e. The Logical Form in
(58a) on the other hand is well-formed, because P ranges over properties. The
restrictor ‘(kind-of-doctor (P))’ means that P is a medical specialization.
McNally observes that the contrast between ‘every N’ and ‘every kind of N’ is
also found in copular constructions:
(59) a. *Martha has been every doctor.
b. Martha has been every kind of doctor.
(59a) is ungrammatical because a predicative position, which needs to be filled by a
property-denoting DP, is occupied by an essentially quantified expression, every doctor,
which is of type <<e,t>,t> and cannot be type-shifted to <e,t>. Every doctor cannot
denote a property. (60a), the tripartite representation associated with (59a), is ill-formed
since the variable x ranges over atomic individuals and not over properties.
(60) a. ∀x (doctor (x)) [Martha has been x]

In contrast, (59b) is grammatical and associated with the Logical Form (60b).
The quantifier binds the variable P, which does not refer to an individual but to a
property. P takes its value in the set of medical specializations. Consequently the
Logical Form (60b) is well-formed.
(60) b. ∀P (kind-of-doctor (P)) [Martha has been P]
McNally’s analysis crucially relies on the idea that kind-reference is identical to
properties, which we will not endorse. Thus, McNally’s attempt to account for the
grammaticality of examples of the type in (51) is problematic, which sheds further
doubt on McNally’s hypothesis that the argument of existential sentences is prop-
erty-denoting, which we have already rejected in Sect. 3.2 above, on the basis of
independent evidence.
Let us repeat that McNally’s hypothesis is not needed in order to explain the
ungrammaticality of quantificational DPs in existential sentences: Heim’s constraint
is sufficient. But we still need to understand why the Lumsden-type of example is
allowed. We will assume, following Carlson (1977a), that kinds are entities of type
e, which are nevertheless distinct from individual entities: “kinds are a little differ-
ent from more normal individuals in that kinds can be here and there, whereas nor-
mal individuals are generally confined to one location, at a given time” (Carlson
1977a: 442). To capture this difference between the two types of entities we will use
indices o and k to notate variables that range over objects and kinds respectively.
The Logical Forms associated with (57) are given below:
(61) a. ∀xk (kind-of-doctor (xk)) [there was xk at the convention]
b. ∀xo (doctor (xo)) [there was xo at the convention]
Because (61a) contains a variable over kinds rather than a variable over individu-
als, this LF does not violate Heim’s constraint.
The analysis we have proposed for the existential predicate (il y a / there is) in
Sect. 3.2.1 predicts the paraphrase relation between (62a) and (62b).
(62) a. Il y a toutes les sortes de docteurs à la réunion.
There have all the sorts of doctors at the convention
‘There is every sort of doctor at the convention.’
b. Il y a des docteurs de toutes les sortes à la réunion.
There have des doctors of all the sorts at the convention
‘There are doctors of every sort at the convention.’
Thus, the LF associated with the nuclear scope of (61a) is developed in (63a)
below. The tripartite formula in (61a), repeated in (63b), can be translated into the
logical formula in (63c), which conveys that for every subkind of doctor, there is at
least one individual entity that belongs to this subkind and is at the convention.
Accordingly, the logical form in (63c) can represent the semantic content of both
(62a) and (62b). This follows from the fact that the instantiation of a subkind amounts
to the consideration of instances of the subkind in question (see (63d)). xa is used to
represent a variable over amounts, M(x) means that x is a doctor and D(x) that x is a
dentist. If we consider a specific subkind of doctors (dentists for example) and we
associate it with a constant D, then the DPs ‘the dentists’ and ‘the doctors of subkind
D’ have the same denotation and correspond to the existential generalized quantifiers
given in (63e). (63f) is equivalent to (63e), since D is a subkind of doctors.
(63) a. there is xk at the convention:
lxk ∃ xi (Ind (xi,xk) ∧ (xi is at the convention))
b. ∀xk (kind-of-doctor (xk)) [there was xk at the convention]
c. ∀xk (kind-of-doctor (xk)) → ∃ xi (Ind (xi,xk) ∧ (xi is at the convention))
d. Ind (xi,xk) ⇔ ∃xa (Real(xa, xk) ∧ Ind (xi,xa))
e. Des dentistes ‘dentists’: lP ∃xa [D(xa) ∧ P(xa)]
f. Des médecins de la classe des dentistes ‘doctors from the kind of dentists’:
lP ∃xa (M (xa) ∧ Real (xa, D) ∧ P(xa))
lP ∃xa (D(xa) ∧ P(xa))
g. (M (x) ∧ Real (x, D)) ⇔ D(x)
Returning to the French example in (56a), repeated in (64), it will have the
Logical Form in (65):
(64) Il y avait la plupart des livres de Beckett dans cette librairie.
there have the majority of the books of Beckett in this bookstore
‘There were all of Beckett’s books in that bookstore.’
(65) MOSTxk (book-by-Beckett (xk)) [there is xk in that bookstore]
Livre-de-Beckett in (64) is analyzed as a property of (sub) kinds. Thus, the
quantifier does not bind an individual variable but a kind variable.
To sum up, in all these examples, the quantifier raises out of the scope of the
predicate of existence, leaving a variable behind. For the sentence to be grammati-
cal, the variable has to obey Heim’s constraint, according to which, in existential
sentences, the argument cannot be an individual variable. The presence of words
like kind, sort, type is crucial for the examples in (51) and (53)–(55), because they
guarantee that the variable ranges over subkinds and not over individual entities.
3.3.3 Existential Sentences Inside Relative Clauses
This section focuses on existential sentences inside relative clauses. Carlson (1997b)
identified a subtype of relative clause, labeled ‘amount relatives’, and he observed
that these relatives are the only type of relative clause that admits existential
constructions:
(66) a. Every man there was on the life-raft died.
b. *Some man there was on the life-raft died.
In what follows we will show that this type of relative clause obeys Heim’s con-
straint. The presentation below is based on French and English examples. Most
English examples come from either Carlson (1977b) or Heim (1987). Most of the
French examples come from Frantext, a corpus of French novels.
3.3.3.1 Amount Relatives
Carlson (1977b) showed that relative clauses of the type shown in (67c) are neither
non restrictive (like (67a)) nor restrictive (see (67b)).
(67) a. Les étudiants, qui ont le droit à une réduction, passeront après
the students who have the right to a discount pass.fut after
tout le monde.
everybody
‘Students, who get a discount, will go after everyone else has.’
b. Les étudiants qui ont réussi l’ examen peuvent s’inscrire pour
the students who have passed the exam may refl register for
le second semestre.
the second semester
‘Students who passed the exam may register for second semester.’
c. Every man there was on the life-raft died.
Although amount relatives are superficially very much like restrictive relatives,
they differ from canonical restrictive relatives by certain syntactic and semantic
properties: (i) while restrictive relatives can be introduced by wh- pronouns, amount
relatives are only compatible with that and the null complementizer; (ii) unlike the
other types of relative clauses, amount relatives allow the constituent that follows
there be to be relativized. The latter observation is important to our discussion. It
corroborates the idea that there is a selectional restriction on the semantic type of
the argument of existential constructions.
To properly understand the data, let us take a closer look at Carlson’s observa-
tions. Drawing a parallel between amount relative clauses and comparative clauses,
Carlson observes that in a comparative clause, the compared DP may appear inside
an existential sentence:
(68) a. There are more women in high school than there are in college.
b. There aren’t as many women in college as there are in high school.
These comparative clauses are assumed to be formed by a deletion transforma-
tion applied to underlying configurations of the type shown in (69), corresponding
to (68a): in (69); the deleted element is X women, where X notates a quantity phrase
that may be paraphrased by AN AMOUNT X.
(69) There are more women in high school than there are X women in college.
Two numbers or quantities are compared: the quantity or number of women in

high school and the quantity or number of women in college. The variable X in (69)
does not refer to a variable over individuals but rather to a variable over quantities.
The nature of the comparison is clear when paraphrased as in (70):
(70) The amount of women in high school is greater than the amount of women
in college.
Carlson suggests that amount relatives resemble comparatives. He considers the

two following sentences, which respectively involve an amount relative clause and
a restrictive relative clause.
(71) a. Marv put everything (Ø / that /*which) he could in his pocket.
b. Huet put everything which was red in his crib.
(71b) can be represented as in (72b). But (71a) and (72a) do not have the same
truth conditions.
(72) a ∀x (Marv can put x in his pocket) → (Marv put x in his pocket)
b. ∀x (x is red) → (Huet put x in his crib)
(71a) is understood to mean that once Marv’s pocket is full, no further object
may be placed in it, even an object that would otherwise easily fit in the pocket.
According to (72a), Marv would have to put in his pocket any object that would
normally fit and continue to add such objects even beyond the point when his pocket
becomes full. So, (72a) is not a proper formalization of (71a). Instead, it is interest-
ing to compare (71a) to a comparative of equality such as (73):
(73) Marv put as many things as he could in his pocket.
Carlson proposes that amount relatives differ from restrictive relatives in that the
relativized element is filled by an amount expression rather than by an individual
denoting expression. Consequently, the underlying structure associated with (71a),
repeated in (74a), is (74b):
(74) a. Marv put everything that he could in his pocket.
b. Marv put everything [that he could put [NP that amount] of things in his
pocket]
In French, where the data are somewhat different, this analysis in terms of amount
is even more obvious. In the French sentences corresponding to (74a), either there
is ellipsis of the entire VP (Il y a plus de femmes au lycée qu’à l’université ‘There
are more women in high-school than at the university’), or the existential construc-
tion is retained and the pronoun en (corresponding to the phrase de femmes) is
obligatory (il y a plus de femmes au lycée qu’il n'y en a à l’université ‘there are more
women in high-school than there are at the university’). In the latter case, it is clear
that only the amount element in the DP is relativized, because the nominal element
remains in the relative clause in the form of the pronoun en.
Let us now return to the example in (67c), repeated here as (75a) and contrast it
with (75b).
b. *Some man there was on the life-raft died.
The difference in acceptability between the two examples can be attributed to
their not having the same structure. (75a) is an amount relative and therefore the
relativization of the position following there was is possible. (75b), on the other
hand, is a restrictive relative and the relativized element denotes an individual. The
wh-trace is therefore a variable that ranges over individuals, which is ruled out by
Heim’s constraint, hence the ungrammaticality of (75b). According to Carlson, the
structures associated with these two sentences are as follows:
(76) a. Every man [there was [that amount] of men on the life-raft] died
b. Some man [there was [that man] on the life-raft] died
Heim (1987) proposes a slightly different notation for amount relatives. She
replaces the trace of the relativized element with a phrase having the structure
x-many N or x-much N, which expresses the idea of quantity. If the relativized
expression is of the individual-type, then its trace is an individual-type variable x.
(75a,b) are accordingly represented as (77a, b):
(77) a. Every man [there was [x-many] men on the life-raft] died
b. *Some man [there was [x] on the life-raft] died
The advantage of Heim’s notation in (77) is that it allows the same analysis to be
used to account for both the contrast between (75a) and (75b) and the contrast
between (78a) and (78b).
(78) a. There were that many horses in the pasture.
b. *There was that horse in the pasture.
We still need to explain why every man is compatible with amount relatives,
whereas some man is not. Carlson (1977b) observes that it is only those determiners
that can precede an amount expression that can head an amount relative. Determiners
meeting this condition appear in the first list in (79) and those that do not appear in
the second:
(79) THE 40 men *TEN many people
THESE few insects *FEW several incidents
EVERY ten minutes *LOTS of many boys
ANY five cigars *MANY twelve pounds
ALL fifty Vikings *A several clods
WHAT few remarks *SOME eight mammals
THESE two answers *SEVERAL many ladies
THESE five criminals *MOST nine squids
MY many dreams *EACH fifty minutes
Since the quantifier of the relativized DP of an amount relative is an amount

expression, one would not expect that singular nouns could head amount relatives.
This is indeed true for count nouns, which accounts for the ungrammaticality of
(80b).
(80) a. The men that there were in Austria like Bob.
b. *The man that there was in Austria likes Bob.
Since mass nouns refer to amounts of substance, we expect them to be allowed
to head amount relatives:
(81) The meat there was soon eaten by the cougar.
The acceptability of (81) is explained by analyzing the mass DP as having a
determiner of the type that much in the underlying structure.
Carlson’s analysis of amount relatives accounts for most of the data. For instance,
it accounts for the constraint on complementizers that may introduce amount rela-
tives. Carlson (1977b), Safir (1982) and Heim (1987), among others, observe that
amount relatives are never introduced by wh- relative pronouns (cf. (67)). The dif-
ference between a wh- element and the complementizer that can be analyzed in
terms of type. A wh- pronoun is analyzed as a variable that ranges over individuals,
whereas the relativized element of an amount relative denotes an amount and not an
individual. The complementizer that is compatible with the relativization of
amounts:
(82) a. *The few men who there were at the party were awfully boring.
b. The few men that there were at the party were awfully boring.
Under the assumption that there be selects an amount-denoting argument, it is
clear why existential constructions can appear in amount relatives but never in rela-
tives introduced by a wh- word:
b. *Every man who there was on the life-raft died.
Note finally that Carlson’s observations concerning whether a determiner is or is
not compatible with amount expressions are able to account for the contrasts illus-
trated in (84) and (85):
(84) a. The people there were at that time only lived a few decades.
b. {What/That} light there is in this painting is quite diffuse.
c. Any beer there may be left in that cooler is mine.
(85) a. {* Five/*Most/*Several/*Many/} men there were here disagreed.
b. {* Some/*Each/* A} man there was disagreed.
So, although it may often be impossible to see a difference in interpretation
between a restrictive relative and an amount relative, example (71a) clearly demon-
strates that in certain cases the notion of quantity is necessary.
3.3.3.2 Existential Sentences in French Relative Clauses
In French, observations regarding amount relatives are not as easy to make as in

English, because in French there is no equivalent of the who/that alternation in
relative clauses. However, an interesting difference exists between que ‘that’ and
ce que in non restrictive relative clauses. The examples below show that ce que
‘that which’ must be used when the relativized DP is not a referential DP but an
attributive one.14
(86) a. Elles s’habillaient comme des femmes excentriques, {ce qu’/

they refl dress.impf like des women eccentric dem that
*qu’} elles étaient.
that they were
‘They dressed like eccentric women, which is what they were.’
b. Je doute que Terry soit un génie, {ce qu’/*qu’}ils
I doubt that Terry be.subj a geniuswwww dem that/that they
considèrent tous qu’ elle est.
consider all that she is
‘I doubt that Terry is a genius, which is what they all consider her to be.’
Nevertheless, it is easy to observe that existential constructions do not appear

frequently in French relative clauses. A search of the Frantext database only turned
up some 30-odd hits. Let us consider some of them.
14
These examples correspond to the French translation of English examples borrowed from
McNally (1998), who uses them in order to illustrate the difference between individual- and
property-denoting expressions:
(i) a. *They dressed like the eccentric women who they were.
a¢. They dressed like the eccentric women that they were.
b. *I doubt that Terry is the genius who they consider her to be.
b¢. I doubt that Terry is the genius that they consider her to be.
For reasons not directly related to the topic under discussion here, there are no minimal pairs with
definite DPs in French: both (iia) and (iib) are acceptable.
(i) a. Elles s’habillaient comme les femmes excentriques, ce qu’elles étaient.
b. Elles s’habillaient comme les femmes excentriques qu’elles étaient.
(87) a. Il faisait chaud, ça sentait bon, et cela venait des odeurs qu’il y avait sur
sa table de toilette.
‘It was warm, there was a pleasant smell, and this was coming from the
fragrances that were on her bathroom table.’
b. Il régnait à la façon des tyrans qu’il y avait alors dans la plupart des villes
grecques.
‘He ruled in the manner of the tyrants that were in most Greek cities at
the time.’
c. Un jour, l’insuffisance cardiaque, alliée à la surcharge de poids qu’elle
prenait parce qu’elle mangeait pour toutes les femmes qu’il y avait en
elle, l’attendit au coin d’un couloir de l’hôpital.
‘One day, her heart condition, combined with the weight she had put on
because she ate for all the women that there were inside her, caught up
with her in the corner of a hospital corridor.’
d. Je ne l’aime pas comme un individu, mais comme le fond religieux de
ma race, comme quelque chose qu’il y a chez tous et chez moi.
‘I do not love her as an individual, but as the religious foundation of my
race, as something there is in everyone and in me.’
In (87a) and (87b), relative clauses are headed by indefinites DPs (des odeurs
‘[des] fragrances’, des tyrans ‘[des] tyrants’) that do not refer to specific enti-
ties. These DPs translate at Logical Form as existential generalized quantifiers
that combine with the existential predicate ‘il y avait’ exactly as in (35),
Sect. 3.2.1. In (87c), toutes les femmes qu’il y avait en elle (‘all the women that
there were inside her’) does not denote several particular women, unlike the
analogous complex DP in Toutes les femmes qui sont parties ont eu raison (‘All
the women that left were right’). The common noun femme (‘woman’) does not
refer to a property of objects (a woman as an individual) but rather to a property
of kinds (a woman as a type of woman). Consequently the universal quantifier
binds a kind variable, not an individual variable. In a more explicit paraphrase
of the example, the different types of women could be listed: elle mangeait pour
toutes les femmes qu’il y avait en elle : la nerveuse, la travailleuse, la
paresseuse… (‘she ate for all the women that there were inside her: the nervous
one, the hard-working one, the lazy one…’). So Heim’s constraint is verified: in
none of the acceptable examples does the variable range over individuals.
Finally, (87d) illustrates a relatively frequent type of example involving the
expression quelque chose qu’il y a… (‘something there is…’). The use of this
indefinite corroborates the idea that the relativized DP cannot refer to an individual
variable.
To the database examples above, we can add similar cases found on the Internet
(see (88)) or constructed examples, as in (89).
(88) a. Que faire de tout ce qu’ il y a dans mon placard ?

what do of all dem that there has in my wardrobe
‘What to do with all that there is my wardrobe?’
b. C’est tout ce qu’ il y a de plus réussi/ de sérieux.
It is all dem that there has of more successful of serious
‘There couldn’t be anything more successful/serious.’
c. Je vous souhaite tout ce qu’ il y a de meilleur pour 2011.
I you wish all dem that there has of better for 2011
‘I wish you all the best for 2011.’
d. On ne voit vraiment rien de ce qu’ il y a autour.
3sg neg sees really nothing of dem that there has around
‘One can’t see anything of what is around’
(89) a. Tous les livres qu’ il y avait dans la bibliothèque de ma
all the books that there have.impf in the library of my
grand-mère ont brûlé.
grandmother have burnt
‘All the books that there were in my grandmother’s library have burnt.’
b. Les livres qu’il y avait dans la bibliothèque de ma grand-mère
the books that there were in the library of my grandmother
ont brûlé.
have burnt
‘The books that there were in my grandmother’s library have burnt.’
All of these examples involve amount relatives. Note that the sentences in (88)
involve ce que ‘that which’, only possible after tout ‘all’ or rien ‘nothing’. The
examples in (89) are similar to those provided by Carlson, repeated in (90). Carlson
has pointed out that singular nouns cannot be modified by amount relatives. The
same restriction holds in French, as attested by (92).

b. The men that there were in Austria like Bob.
(91) *The man that there was in Austria likes Bob
(92) *Le livre qu’ il y avait dans la bibliothèque de ma grand-mère
the book that there was in the library of my grandmother
a brûlé.
has burnt
‘The book that there was in my grandmother’s library has burnt.’
Summarizing, Carlson’s amount relatives obey Heim’s constraint, because in this

type of relative, the variable does not range over individuals but rather over
quantities.
3.4 Conclusion 125
It should be observed that in the French examples examined above, the existen-
tial sentences inside the relatives are all of the locative type. Turning now to the
other types of existentials, it seems very difficult, if not impossible, to relativize the
enumerative constructions (cf. (93)). This is even clearer for eventive existential
sentences, which cannot appear in relative clauses, as illustrated by (94):
(93) a. Il ya Jean, Marie et un ami.

There have Jean Marie and a friend
‘There are Jean, Marie and a friend.’
b. *Jean, Marie et un ami qu’ il y a....
Jean Marie and a friend that there has
‘Jean, Marie and a friend who there are…’
(94) a. Il y a le téléphone qui sonne dans la pièce du fond.
There has the phone that rings in the room in the back
‘There’s the phone ringing in the back room.’
b. *Le téléphone qu’ il y a qui sonne dans la pièce du fond…
the phone that there has that rings in the room in the back
‘The telephone that there was ringing in the back room…’
3.4 Conclusion
Contrary to what is usually assumed, existentials are not subject to Milsark’s con-
straint, according to which only weak indefinite DPs may be arguments of there-
sentences, definite and quantified DPs being excluded from this position. No
determiner is completely excluded from these constructions: provided they quantify
over non individual entities, quantified DPs are allowed in existential sentences; as
for definite DPs or proper names, they are licensed in eventives (as event partici-
pants) and in enumeratives (as members of a list). We have shown that Heim’s
constraint is relevant for all the data we have examined, in English as well as in
French: existential sentences are incompatible with individual variables in their
argument position.
We have distinguished three types of existential constructions (locatives, even-
tives and enumeratives), which can be characterized by their respective codas.
Locative existential sentences are built with weak DPs (analyzed as generalized
existential quantifiers over amounts), which combine with the instantiation predi-
cate il y a. They are also compatible with quantified DPs, when the quantification is
not over individuals but over other types of entities such as kinds, types, sorts as
well as amounts. The analysis of amount relatives provides an argument in favor of
Heim’s constraint.
Eventive existential sentences are built by an application of the instantiation

predicate to an event, described by a constituent made up of the DP and the coda.
Enumerative existential sentences take a list in their argument position.
In all of these cases, the instantiation predicate il y a/there be can be seen as a
particular type of existential predicate. Its semantic contribution reduces to assert-
ing the instantiation of its argument.
In terms of formalization, our contribution is the representation of weak DPs as
generalized existential quantifiers. These generalized quantifiers differ from other
generalized quantifiers on two points: (i) they quantify over amounts and not over
individuals; (ii) they can only apply to existential predicates.
Chapter 4
The Ambiguity of Indefinites: Towards
a Denotational Definition of the Weak/Strong
Distinction
According to Milsark (1977), certain indefinites have two readings, a “weak” and a
“strong” reading. In this chapter we will show that this two-way distinction should
be replaced by a three-way distinction: in addition to the weak reading, we will
distinguish two types of strong readings, a quantificational and a non-quantificational
one. Three distinct representations will be proposed for each of these three readings
and correlations between the denotations of indefinites and presuppositionality or
partitivity will be highlighted.
4.1 Weak and Strong Indefinites
Although it plays a central role in the study of indefinites as well as the study of the
semantics of DPs in general, the definition of the weak/strong distinction remains
problematic,1 despite efforts to formalize it within the generalized quantifier frame-
work (see especially Barwise and Cooper 1981). In what follows, we present those
contexts in which the difference between weak and strong indefinites can be dis-
cerned (Sect. 4.1).
Milsark (1977) observed that only indefinite DPs can appear in existential sen-
tences introduced by il y a / there is; quantificational or definite DPs are
unacceptable2:
1
For a clear presentation of the problems raised by current analyses of the weak/strong distinction,
the reader is referred to McNally and van Geenhoven (1998).
2
See Chap. 3 above for refinements of this generalization.

128 4 The Ambiguity of Indefinites: Towards a Denotational Definition…
(1) a. Il y avait deux enfants devant la porte.

There have.impf two children in front the door
‘There were two children in front of the door.’
b. *Il y avait chaque enfant devant la porte.
There have.impf every child in front the door
‘There was each child in front of the door.’
c. ??Il y avait l’ enfant devant la porte.
There have.impf the child in front the door
‘There was the child in front of the door.’
Milsark’s distinction between weak and strong DPs was meant to capture this
differential behavior: weak DPs can occur in existential sentences, whereas strong
DPs cannot.
Milsark further observed that indefinite DPs are not always weak. They may also
have a strong construal, which corresponds to a partitive meaning. The crucial point
is that indefinites such as two students, which are not overtly partitive (they do not
have a partitive complement), are necessarily interpreted as partitive in certain
contexts. Consider the examples in (2) and (3):
(2) a. Deux filles étaient en train de jouer dans la cour.
Two girls were prog to play in the courtyard
‘Two girls were playing in the courtyard.’
b. Deux filles étaient blondes, toutes les autres avaient des
two girls were blond all the others had des
cheveux foncés.
hair dark
‘Two girls were blond, all the others had dark hair.’
(3) a. Two girls were playing in the courtyard.
a¢. Sm students were reading a poem. (sm = unaccented form of some)
b. Two students were blond.
b¢. (Some/*Sm) students are intelligent / handsome / blond.
In (2a), deux filles ‘two girls’ is ambiguous allowing two interpretations: “There
were two girls who were playing in the courtyard” and “Two of the girls were play-
ing in the courtyard”. Note however that the weak reading (i.e., “there were two
girls …’) is more salient. In (2b), on the other hand, the partitive reading (i.e., “two
of the girls”) is the only possible one; the group denoted by deux filles is construed
as belonging to a larger group of girls in the common ground (hence Pesetsky’s
(1987) label “D(iscourse)-linked”). The same observations hold for English, as illus-
trated in (3). Only the partitive reading is available in (3b–b¢), as in (2b), due to the
use of predicates like intelligent, handsome, blond (see Sect. 4.2.2).
Let us warn the reader that the correlations between strong vs. weak readings and
partitivity or non partitivity should be carefully formulated. Milsark does not say
anything about overtly partitive indefinites, e.g., two of your students, and it can be
4.2 Weak Indefinites 129
quite easily shown that such indefinites are not necessarily strong, as indicated by
examples such as there were two of your students in the street yesterday (see section 1.2
of Chap. 5). To repeat the main point, Milsark’s observation regarding partitivity is that
in certain contexts, namely in the subject position of certain predicates, indefinites
necessarily take a partitive reading, although they are not overtly partitive.
Milsark’s generalizations regarding weak and strong interpretations can be sum-
marized as in (4):
(4) Milsark’s (1977) generalizations

a. Indefinites occurring in existential sentences are necessarily weak.
b. Indefinites occurring in the subject position of a certain class of predicates
(individual-level) are necessarily strong.
c. Strong indefinites have partitive interpretations.
In what follows we will provide further evidence in favor of (4a–b) but we will
depart from (4c), by showing that strong indefinites are not always partitive/
(“D-linked”) (see Sect. 4.3.2), which will lead us to distinguish two types of strong
indefinites.
4.2 Weak Indefinites
In this section we point out some disadvantages of the current analyses of weak
indefinites and we make a new proposal, namely the one already introduced in
Chap. 3 above, according to which weak indefinites refer to amounts and should be
represented as generalized existential quantifiers over amounts.
4.2.1 Weak Indefinites as Individual Variables

Bound by Existential Closure
Following Kamp (1981) and Heim (1982), most theoreticians assume that indefinites
are to be represented as individual variables (see especially Kratzer 1988, 1995;
Diesing 1992, a.o.).
In the framework of Heim (1982), individual variables are bound by existential
closure, which may apply in two contexts:
(i) in the nuclear scope of every quantifier, or
(ii) at the highest level, the level of text.3
3
Heim (1981:139) explains that in order to take care of indefinites occurring in unembedded sen-
tences, we have to ‘make reference to expressions larger than sentences, namely texts, which are
sequences of sentences of unlimited length’. Heim assumes a rule of construal called Text
Formation, which attaches a sequence of sentences under a T(ext)-node. Text-level existential
closure consists in adjoining the existential quantifier to T.
These two possibilities are illustrated in (5a) and (5b), respectively:
(5) a. Chaque élève écrira un poème.

Each student write.fut a poem
‘Each student will write a poem.’
b. Un élève m’a annoncé une mauvaise nouvelle.
A student me-has announced a bad news
‘A student gave me a bad piece of news.’
(5¢) a. ∀x ∃y (student (x) ∧ poem (y)) [x will write y]
b. ∃x ∃y (student (x) ∧ bad piece of news(y)) [x told me y]
Diesing (1992) observed that Heim’s rules cannot account for the weak readings
of indefinites and proposed instead the following rule:
(6) Existential closure applies at the VP level.

Besides this definition of existential closure, Diesing’s analysis relies on certain
other hypotheses:
(7) a. Subjects of individual-level (permanent or stable) predicates must occupy

the (Spec, IP) position (which is external to the VP constituent).
b. Subjects of stage-level (transitory) predicates can be either in (Spec, IP) or
in a VP-internal subject position.
c. Subjects of stage-level predicates may be affected by a lowering rule,
which moves them from (Spec, IP) back to their initial VP-internal subject
position.
Taken together, these hypotheses account for the data in (2). The predicates in
(2b) are individual-level and therefore their subjects must occupy a preverbal posi-
tion (technically referred to as (Spec, IP)), which is above VP and thus outside the
scope of existential closure (see (6)). As a result, a weak (or existential) reading of
(2b) is blocked; only the strong reading (the presuppositional reading, in Diesing’s
terms) is available. Sentence (2a) is built with a stage-level predicate and therefore
the subject can be lowered to a position inside VP, where it is in the scope of exis-
tential closure, as defined in (6). Thus, a weak reading is available in (2a).
Although one may disagree with some parts of Diesing’s analysis (in particular
the lowering of the subject from Spec, IP back into VP), it seems necessary – granting
that indefinites translate as variables – to constrain their weak readings by using
rules of existential closure that are stricter than those of Heim (1982).
The hypothesis that an existential quantifier is present in the lexical representa-
tion of certain predicates goes back to Carlson (1977a, c), according to whom the
relevant class can be identified with s(tage)-level predicates, as opposed to
i(ndividual)-level predicates. In Sect. 2.5 of Chap. 2 we have proposed a different
distinction, between entity-predicates and existential predicates, which cuts across
the i-level/s-level predicate distinction. Furthermore, the lexical nature of the predicate
is not a sufficient condition for an indefinite to be able to take a weak reading: the pos-
sibility of weak readings of indefinites depends not only on the lexical properties of
the predicates with which indefinites combine but also on the position occupied by the
indefinite. Thus, the pair in (8a) vs. (8b) shows that in Romanian, bare plurals are
allowed in the postverbal subject position with a verb such as a veni ‘to come’ but
disallowed in the preverbal subject position of the same verb:
(8) a. Ieri au venit studenţi să-mi spună..

yesterday have come students subj-me say
‘Yesterday there came students to tell me …’
b. *Studenţi au venit ieri să-mi spună
students have come yesterday subj-me say
‘Students came yesterday to tell me …’
Diesing’s (1992) rule of VP-level existential closure is an attempt to capture this
additional constraint on weak indefinites.4 According to Diesing, the contrast in (8a)
vs (8b) can be explained as follows: (i) existential BPs are weak indefinites; (ii)
weak indefinites must be bound by existential closure; (iii) existential closure
applies to the VP. The example in (8a) is grammatical because the subject is
VP-internal and as such the BP can be legitimated by existential closure; the exam-
ple in (8b) is ungrammatical because the subject is VP-external and as such the BP
cannot be legitimated by existential closure; and since existential BPs can only be
weak indefinites, (8b) is ungrammatical.
Note that this explanation relies on the assumption that in Romance languages
other than French (as well as in German) subjects sitting in Spec, IP cannot be lowered
to a VP-internal position. Compare English and French, where this type of lowering
rule must be assumed in order to account for the acceptability of (2a) and (3a, a¢).
Lowering seems to be blocked by the Theme/Topic status of preverbal subjects (see
Sect. 5.2 of Chap. 5).
Let us finally observe that in order to account for both weak and strong readings
of indefinites we need to assume both Heim’s (1982) rules of existential closure (for
strong and dependent readings) and Diesing’s rule of VP-level existential closure
(for weak readings). Because this is not a desirable set up, Kratzer (1988, 1995) and
Diesing (1992) suggest that it is possible to dispense with Heim’s rules.
4.2.2 Weak Indefinites as Property-Denoting Expressions
Another possibility is to analyze weak indefinites as property-denoting expressions

(Heim 1987; Ladusaw 1994; Dobrovie-Sorin 1997a, b; McNally 1998; McNally
4
Diesing’s own data concern the postverbal and preverbal positions of subject bare plurals in
German.
and van Geenhoven 1998). This analysis relies on the hypothesis that certain predicates
are not necessarily represented as in (9a) but may also be represented as in (9b):
(9) a. lx P(x)
b. lQ ∃x (P(x) ∧ Q(x))
(9a) is the canonical representation of one-place predicates: they denote sets of
individuals (type <e,t>) and as such can only apply to individual-denoting DPs (type
e). The representation in (9b) is designed in such a way that certain predicates, e.g.,
dormir ‘sleep’, laver ‘wash’ and disponible ‘available’ have argument positions that
are already saturated by existential closure in the lexicon. But they are not fully
saturated, as they still require a property to specify the domain of variation of their
argument variable. In (9b), this is represented by the property variable Q restricting
the argument variable.5
Granting (i) that sentential predicates can be represented as in (9b) and (ii) that
weak indefinites denote properties, we can analyze sentences such as (10) as shown
in (10¢):6
(10) a. Des enfants sont en train de dormir.

des children are prog to sleep
‘Children are sleeping.’
b. De la fumée s’échappe de la cheminée.
de la smoke escapes of the chimney
‘Smoke is coming out of the chimney.’
The representation in (10¢a) contains a predicate analyzed as in (9b), with dormir
replacing P and applying to the property denoted by des enfants ‘children’. (10¢a¢)
is obtained by l-conversion of (10¢a):
(10¢) a. (lQ ∃x (dormir(x) ∧ Q(x))) (des enfants)

a¢. ∃x (dormir(x) ∧ des enfants(x))
5
Sometimes, the domain of variation of the existentially bound variable is not restricted. Such is
the case of implicit arguments, which appear in the thematic structure of a predicate but are not
lexically projected, as in sentences like Jean a mangé ‘Jean ate’ or Marie lit beaucoup ‘Marie reads
a lot’.
6
The weak reading of indefinites is sometimes analyzed as predicate modification (Farkas 2001).
And yet, rules of modification of the sort proposed in Chung and Ladusaw (2003), for example, are
technically different from the type of representation used here. At this point in the investigation of
weak indefinites, there are two possibilities. The first one is that weak indefinite readings are all of
a kind, in which case the different representations proposed in the recent literature are notational
variants of the same phenomenon. Alternatively, it may be that there are several distinct types of
weak indefinites (for example, bare plurals and bare count singulars in Spanish or Romanian would
belong to distinct types), licensed by different syntactic contexts and assigned different semantic
representations.
All predicates can be represented as in (9a) but only a subclass of predicates

(albeit a large one), call them existential,7 can be represented as in (9b). This amounts
to saying that in the lexicon existential predicates are systematically ambiguous
between the two kinds of representation, (9a) and (9b).
The hypothesis that weak indefinites are not represented as individual vari-
ables correlates with their not introducing discourse referents. It is nevertheless
possible for a weak indefinite to be the antecedent of an anaphor, because a
discourse referent is introduced indirectly as a participant to the event denoted
by the VP.
The property analysis of weak indefinites is problematic insofar as it cannot
explain why prototypical property-denoting expressions such as adjectives may
occur in those argument positions that allow weak indefinites:
(11) a. John reads novels.

b. *John reads romantic.
4.2.3 Weak Indefinites as Amount-Referring Expressions
We have so far discarded two analyses of weak indefinites, one in terms of individual
variables bound by VP-level existential closure, the other in terms of properties. We
will nevertheless preserve the common ingredient of these analyses, namely the
hypothesis that weak indefinites are legitimate only if an existential quantifier can
be supplied by the predicate with which the indefinite combines.
In Chap. 3 above we have already proposed that weak indefinites occurring in
existential sentences are to be analyzed as amount variables bound by the existential
quantifier introduced by il y a. We repeat below our proposal for the semantic
composition of existential sentences:
(12) de l’eau = lP ∃ xa (water (xa) ∧ P(xa))

(13) il y a = lxa ∃ xi (Ind (xi,xa))
(14) a. il y a de l’eau = de l’eau (il y a)
b. lP ∃ xa (water (xa) ∧ P(xa)) • lxa ∃ xi Ind (xi,xa)
c. ∃ xa (water (xa) ∧ lxa ∃ xi Ind (xi,xa) (xa))
d. ∃ xa (water (xa) ∧ ∃ xi Ind (xi,xa)
e. ∃ xi∃ xa (water (xa) ∧ Ind (xi, xa ))
7
“Existential predicate” is an ambiguous term, in that it can refer either to lexical classes or to
representations of verbs, in a given context. Under the latter meaning, “existential predicate” refers
to (9b).
In (14a), the two constituents are semantically composed by applying the generalized
quantifier denoted by de l’eau to the property denoted by il y a. In (14b) we have
replaced the two constituents by their representations in (12) and (13). In (14c) and
(14d) we have applied lambda-conversion twice. In (14e) we have preposed the
existential quantifier that binds the individual variable.
Let us now turn to the representation of examples of the type in (10a–b), repeated
under (15a–b). Both examples can be analyzed in the same way, so that it is sufficient
to give the semantic composition of one of the two examples:
(15) a. Des enfants sont en train de dormir.

des children are prog to sleep
‘Children are sleeping.’
b. De la fumée s’échappe de la cheminée.
de la smoke escapes of the chimney
‘Smoke is coming out of the chimney.’
Granting that des enfants ‘[des] children’ in (15a) is a weak indefinite and grant-
ing that weak indefinites are to be represented as generalized quantifiers of the type
shown in (12) above, des enfants is to be represented as in (16a). Following Carlson,
we will assume an existential quantifier in the lexical representation of predicates
such as sleep:
(16) a. des enfants = lP ∃ xa (children (xa) ∧ P(xa))

b. sont en train de dormir = lxa ∃ xi (sleeps (xi) ∧ Ind (xi, xa))
c. Des enfants sont en train de dormir = children (be sleeping)
d. lP ∃ xa (children (xa) ∧ P(xa)) • lxa ∃ xi (sleeps (xi) ∧ Ind (xi, xa))
e. ∃ xa (children (xa) ∧ [lxa ∃ xi (sleeps (xi) ∧ Ind (xi, xa))] (xa))
f. ∃ xa (children (xa) ∧ ∃ xi (sleeps (xi) ∧ Ind (xi, xa)))
g. ∃ xa ∃ xi (children (xa) ∧ sleeps (xi) ∧ Ind (xi, xa))
4.3 Strong Indefinites
In this section, we will demonstrate the existence of two distinct types of strong
readings, a quantificational and a non quantificational one.
4.3.1 Quantificational Strong Indefinites
According to Milsark, strong indefinites have partitive interpretations (note how-

ever that partitive indefinites, in particular overt partitives, e.g., two of my students,
are not necessarily strong (see Sect. 4.4 for further discussion)). In this subsection
4.3 Strong Indefinites 135
we will propose that the partitive interpretations of those indefinites that are not
overtly partitive are an effect of their quantificational status, which is triggered by
the context. In the next subsection we will argue that strong indefinites are not nec-
essarily partitive and correlatively that they are not necessarily to be represented as
quantificational.
In order to clarify as much as possible the difference between the weak and
strong readings of indefinites, let us consider the example in (17), built with a car-
dinal and let us further assume that the Logical Forms associated with each reading
are as indicated in (18) and (19):
(17) Deux enfants sont en train de jouer dehors.

two children are prog to play outside
‘Two children are playing outside.’
(18) Weak reading (xa is a variable over amounts)
∃ xa (child (xa) ∧ two(xa) ∧ playing (xa))]8
(19) Strong reading (xi is a variable over individuals)
Two xi (child (xi)) [xi is playing]
As explained above in Sect. 4.2.3, the weak readings of indefinites are associated
with logical forms without quantification over individuals, in which an amount vari-
able is bound by an existential quantifier. Indeed, the variable in (18) does not range
over individual children but rather over amounts of children of cardinality two,
whereas the variable in (19) ranges over individual children. Correlatively two func-
tions as a cardinality predicate (which is a particular type of measure phrase) in (18)
but as a quantifier in (19). For the sentence in (17) to be true under the weak reading
of two children, we only need to find an amount of two elements for which both predi-
cates are true. To say that a strong indefinite is partitive means that it presupposes the
existence of a contextually determined set of individuals. We assume that this presup-
posed set corresponds to the restriction of a tripartite structure. The tripartite represen-
tation in (19) is built on the model of sentences containing quantificational DPs:
(20) a. Each child is blond.

b. ∀ x (child (x)) [x is blond]
It should be stressed that in the case of quantificational determiners such as tous
(‘all’), la plupart (‘most’) or 3%, tripartite representations are due to their lexical
properties (proportionality, non-intersectivity). In the case of indefinite DPs, on the
other hand, tripartite representations are not due to the lexical properties of indefinite
determiners (which are intersective, symmetric and non proportional) but rather to
the syntactic context (in particular contexts in which a weak reading is blocked) or
to a partitive / D-linked reading that is intended by the speaker.
8
The LF representations of examples of this type might be more adequately represented in terms
of existential quantification over events. However, this refinement can be disregarded here.
4.3.2 Non-quantificational Strong Indefinites
In this section we show that strong indefinites are not necessarily partitive (which
goes against Milsark’s generalization given in (4c)). We propose that strong non
partitive indefinites are not to be represented as quantificational DPs but rather as
individual variables or more precisely as Skolem terms (cf. Steedman (2003)).9
4.3.2.1 Non-partitive Strong Readings
Let us consider the following examples:

(21) a. Une voiture est dans le garage.
a car is in the garage
‘A car is in the garage.’
b. Un homme était à côté de moi.10
a man was next of me
‘A man was next to me.’
c. On a dû s’arrêter parce que trois enfants étaient malades/fatigués.
3sg has had refl-stop because three children were sick tired
‘We had to stop because three children were sick.’
The indefinites appearing in (21a–c) are not necessarily interpreted as partitive:
these examples do not necessarily presuppose the existence of contextually deter-
mined sets of cars, men or children. Indeed, in (21a), a continuation like les autres
sont à tel endroit ‘the others are in such-and-such a place’ is neither needed nor
expected. There are contexts in which (21a) may be uttered felicitously without
causing a potential interlocutor to ask “où sont les autres?” (‘where are the oth-
ers?’). In short, (21a–c) do not require a partitive reading.
Note now that despite the intuition that the examples in (21) can be paraphrased
by existential sentences, the indefinites in (21) cannot be assumed to have weak
readings, because these examples are built with entity predicates (see Sect. 2.5 of
Chap. 2), which cannot apply to DPs that are necessarily weak, e.g., mass indefinites
introduced by du/de la in French:
9
Within DRT, non-quantificational strong indefinites correspond to free variables bound by text-
level existential closure.
10
For some speakers, examples of the type in (21a–b), borrowed from Kleiber (1981a, b), are
highly marginal.
4.3 Strong Indefinites 137
(22) a. *De la fumée est dans le garage.

de la smoke is in the garage
‘Smoke is in the garage.’
b. *Du beurre était à côté de la farine.
du butter was next of the flour
‘Butter was next to the flour.’
These examples are ungrammatical due to two conflicting requirements: mass
indefinites introduced by du/de la are necessarily weak, i.e., they supply an amount
variable and as such they must be legitimated by existential closure, but copular
sentences with a localizing PP predicate block this reading.
We are thus led to conclude that, although they do not have partitive meanings,
the indefinites that appear in (21) have strong readings. This type of reading is
difficult to obtain; judgments vary and are context-dependent, as shown in (23),
built with joyeux ‘joyous’, a stage-level predicate that qualifies as an entity predi-
cate, i.e., a predicate that cannot supply an existential quantifier.
(23) a. ??Deux filles étaient joyeuses.

two girls were joyous
‘Two girls were joyous.’
b. Deux filles étaient si joyeuses que tout le monde était de bonne humeur.
two girls were so joyous that everyone was of good mood
‘Two girls were so joyous that everyone was in a good mood.’
Non-partitive strong readings are easy to find with indefinites that are specific,
e.g., speaker-specific:11
(24) a. Une femme que je connais est chauve.12

a woman that I know is bald
‘A woman I know is bald.’
b. Un collègue à moi est daltonien.
a colleague of me is color-blind
‘A colleague of mine is color-blind.’
Certain indefinite determiners are specialized for non-partitive strong readings.
A clear example of this type of determiner is unos ‘some’ in Spanish (see Villalta
1995; Tasmowski and Laca 2000; Lopez Palma 2007). In the following examples,
due to Helena Lopez Palma (p.c.), we see that unos N can be represented neither as
a quantificational indefinite, (25a–b), nor as a weak indefinite, (25c).
11
See Farkas (1994), who distinguished among speaker-specificity, epistemic specificity and scope
specificity, by virtue of their descriptive content.
12
Example from Galmiche (1986).
(25) a. Unos estudiantes le han regalado un libro a Carmen.

some students her.dat have given a book to Carmen
‘Some students gave a book to Carmen.’
b. *Unos estudiantes le han regalado sendos libros a Carmen
some students her.dat have given sendos books to Carmen
‘Some students gave sendos books to Carmen.’
c. Los estudiantes han leido unos libros.
the students have read some books
‘The students read some books.’
Unlike quantificational DPs, unos N cannot have a distributive interpretation:
(25a) describes a situation in which a group of students gave a book together; (25b)
is ungrammatical because sendos forces a distributive interpretation (‘some stu-
dents each gave a book to Carmen’) and is therefore incompatible with unos.13
Unlike bare plurals in Spanish, unos N cannot be interpreted as a dependent plural:
(25c) cannot mean that the students each read a book (whereas this meaning is pos-
sible for Spanish bare plurals, as in Los estudiantes han leido libros).14 We can
therefore conclude that indefinite DPs headed by unos are strong but not
quantificational.
This third type of reading, although crucial for the theory of indefinites,15 has escaped
notice because – unless it is lexically marked, as is the case for unos – it is only margin-
ally accepted and requires special contexts. This kind of interpretation is facilitated by
indications allowing the discourse referent to be introduced by way of a separate predi-
cation, such as “that I/you (don’t) know)” or “that I met last year”, as in (24).
4.3.2.2 Non-quantificational Strong Indefinites and Wide Scope Effects
Since Fodor and Sag (1982), indefinites are known to allow wide scope interpreta-
tions that cannot be analyzed as relying on mechanisms of quantifier scope:
(26) Each professor knows that a student I’m advising failed the test.
13
Helena Palma Lopez points out that (25b) becomes acceptable if unos estudiantes is replaced
with unos estudiantes mios ‘some students of mine’:
(i) Unos estudiantes mios le han regalado sendos libros a Carmen.

some students mine her.dat have given sendos books to Carmen
Distributivity here owes to the specificity of the DP (the referent of unos estudiantes mios is acces-
sible to the speaker), which in turn owes to the presence of mios. For more on how the specificity
of indefinites and distributivity are linked, see Sect. 5.1.5 of Chap. 5.
14
This does not mean that, when in object position, unos N will always take wider scope than the
subject. In (25c), for example, the relative scope of the two DPs is ambiguous: each student may
have read the same group of books, or each may have read a different group.
15
Acknowledging the existence of non-quantificational strong indefinites may help resolving a
contradiction between the analyses of Heim (1982) and Heim (1987) (remarked upon but left
unresolved in note 1 of the latter): the analysis of Heim (1987) pertains to weak indefinites, while
the analysis of Heim (1982) is adequate for non-quantificational strong indefinites.
4.4 The Weak/Strong Distinction and Presuppositionality 139
Example (26) allows a specific reading according to which a student that I’m advis-
ing is such that each professor knows that he failed the test. This reading cannot be
analyzed in terms of quantifier scope because canonical quantifiers cannot scope out
of the minimal clause to which they belong, which led Fodor and Sag to propose that
the wide-scope interpretation of indefinites derives from a referential interpretation of
the indefinite. Within the three-way analysis of indefinites proposed here, wide-scope
indefinites can be represented neither as weak indefinites (such indefinites are neces-
sarily narrow-scoped) nor as quantificational strong indefinites (the scope of quantified
DPs cannot exceed their minimal clause). This leaves us with the third reading that we
have identified here: non-quantificational strong indefinites.
Following Steedman (2003), we propose to represent this type of indefinite,
which exhibits wide-scope effects, as a Skolem term (see Chap. 6 below).
4.3.3 The Two Strong Readings of Indefinites

and the Denotation of DPs
The two strong readings of indefinites identified above are parallel to two different
types of strong DPs, as shown in (27):
(27) Strong DPs Indefinites Other than Indefinites

Type <<e,t>,t> quantificational quantificational
Type e Referential proper names, definites, demonstratives
We have so far suggested that partitive strong indefinites can be analyzed as
quantificational DPs, which can be represented by means of tripartite structures. Non-
partitive strong indefinites on the other hand are referential DPs, which denote indi-
viduals. The similarity between referential DPs and non partitive strong indefinites
appears clearly if the latter are analyzed as Skolem terms (see Steedman 2003 and
Chaps. 6 and 8). This analysis has not yet become predominent. In the current literature
non-partitive strong indefinites are treated as specific. For a general overview of differ-
ent semantic approaches to specific indefinites see von Heusinger (2011).
4.4 The Weak/Strong Distinction and Presuppositionality
Following Diesing (1992), we will assume the “presuppositionality hypothesis”

(cf. Heim and Kratzer (1998: 163)) according to which quantified DPs are presup-
positional. Sentences built with quantified DPs are represented as tripartite struc-
tures, in which the restriction may be identified with the presuppositional part of the
sentence. Since referential DPs are also presuppositional (cf. Strawson), we can
assume the generalization in (28), which establishes a correlation between the
denotation of DPs and presuppositionality:
(28) Strong DPs (referential DPs and generalized quantifiers) are presuppositional.
Two types of presuppositionality have to be distinguished, correlating with the

two types of strong DPs. Referential DPs, e.g., proper names, presuppose the exis-
tence of their referent, whereas quantificational DPs presuppose that the set denoted
by the head noun N (or by the noun together with its modifiers) is not empty. In
other terms, quantificational DPs presuppose the existence of the non-empty set
denoted by their NP constituent.
Since Diesing (1992), the term “presuppositional” is commonly used as a syn-
onym for “partitive”, “D-linked” or “quantificational”. We do not adopt this use: for
us, presuppositional is equivalent to “strong” but different from “partitive” and
“quantificational”. Indeed, there are both partitive and non partitive presuppositional
indefinites. The distinction between partitive and non-partitive presuppositional
indefinites corresponds to the distinction that we have established between the two
strong readings of indefinites: quantificational strong indefinites are presuppositional
and partitive, whereas non-quantificational strong indefinites are presuppositional
but not partitive. Non-quantificational strong indefinites resemble definite and
demonstrative DPs by their denotational type (type e) and by their non partitivity.
“Wide scope” indefinites allow us to illustrate the case of non partitive presup-
positional indefinites:
(29) Each professor knows that a student failed the test.

The so-called “wide scope” interpretation of a student is presuppositional: if the
set of students is empty, then the sentence is uninterpretable (neither true nor false).
However, the wide-scoped indefinite is not necessarily partitive: there need not be a
contextually relevant proper subset of students (except for the singleton containing
the student in question).
On the other hand, overtly partitive indefinites are not always presuppositional,
as indicated by examples such as (30), which show that indefinites like one/two/
three/several of your students/children/friends are acceptable in existential
sentences:
(30) Dans la salle il y avait deux de tes étudiants.

in the room there have.impf two of your students
‘In the room there were two of your students.’
Examples of this type, in which overtly partitive indefinites occur in existential
sentences, can be represented as in (31), where the partitive constituent introduced
by de ‘of’ does not map to the restriction of a tripartite configuration but instead is
analyzed as a property that restricts the domain of variation of an amount variable:
(31) $ xa [xa belongs to your students Ù |xa| = two]
Represented as in (31), the example in (30) asserts the existence of an amount

that ranges over students of yours and the measure of which is two. Here and else-
where, the cardinals that modify weak indefinites are to be analyzed as the values
obtained by applying a measure function to an amount of objects.
4.5 Conclusions 141
Summarizing, the denotational distinction between weak and strong indefinites

correlates with the distinction between non-presuppositional and presuppositional
indefinites:
(32) a. Strong indefinites (which denote either individuals or generalized

quantifiers) are presuppositional.
b. Weak indefinites are non-presuppositional.
As indicated in (32a), presuppositional indefinites are not necessarily
quantificational. They may also be referential (i.e., individual referring).
4.5 Conclusions
Most theoreticians distinguish only two readings of indefinites, weak and strong.
We have shown that indefinites are in fact compatible with three distinct readings
(weak, quantificational strong and non-quantificational strong) and we have pro-
posed three possible analyses of indefinites, which can be represented as follows:
a. Weak indefinites refer to non specific amounts and must combine with an exis-
tential predicate. They denote generalized existential quantifiers over amounts.
b. Non-quantificational strong indefinites are referential expressions that are repre-
sented as Skolem terms.
c. Quantificational indefinites are generalized quantifiers, which we have repre-
sented in terms of tripartite configurations.
Chapter 5
Disambiguating Indefinites
In the preceding chapter, we established that indefinites are potentially ambiguous,

being compatible with three different types of representation. The evidence in favor
of this tripartition was mainly based on two contexts that restrict the interpretations
of indefinites: il y a sentences and a lexical distinction between two classes of
predicates. In what follows, we will examine DP-internal properties that help dis-
ambiguate indefinites: (a) the lexical properties of determiners; (b) the internal
structure of indefinite DPs. We will then consider contextual factors, such as syntac-
tic position and information structure. The main generalizations will be based on
French examples, which do not seem to differ from the corresponding English
examples, as can be inferred from the glosses. Spanish and Romanian examples will
be found in Sect. 5.1.4, where they illustrate the phenomenon known as ‘preposi-
tional accusatives,’ a type of construction that does not exist in French or English.
5.1 Disambiguating Indefinites: DP-Internal Factors
5.1.1 Lexical Specification of Plural Indefinites: Partitivity,

Contrastivity and Distributivity
Many languages have more than one plural indefinite determiner. French, for exam-
ple, has certains, quelques, des and plusieurs and English has certain, some, sm
(deaccented some) and several. They are all indefinite in that they are intersective
and symmetrical but they differ with respect to other semantic properties, such as
partitivity, contrastivity and distributivity. These features restrict the interpretative
possibilities of indefinite DPs.

144 5 Disambiguating Indefinites
Attal (1976) notes that there is an important contrast between certains ‘somepl’
and des in French: certains gives rise to quantificational sentences with distributive
interpretations, while des does not, except in contrastive contexts:
(1) a. Certains enfants étaient tristes/ malades / indisciplinés.
certain children were sad/ sick / undisciplined
b. ?? Des enfants étaient tristes/ malades/ indisciplinés.
des/sm children were sad/ sick / undisciplined
‘Some children were sad/sick/undisciplined.’
In unmarked cases, DPs introduced by certains are to be analyzed as
quantificational, whereas DPs introduced by des refer to non specific amounts. The
other plural indefinite determiners (quelques, plusieurs) are compatible with all
three representations introduced in the previous chapter; the choice of a particular
representation is contextually determined.
In what follows, we would like to suggest that the quantificational properties of
certains derive from its lexical meaning. More precisely, certains is intrinsically
contrastive and as such it gives rise to the implicature “not all”. The speaker uses
certains rather than des or quelques when the individuals characterized by the prop-
erty denoted by N may be divided into two subsets, the set of those who satisfy the
predicate of the sentence and the set of those who do not. The contrastitivity that
characterizes certains thus induces partitivity. In the words of Corblin (2001:105),
“certains N laisse entendre une partie de N mais pas tous les N”.1
(2) Certains étudiants ont rendu leur copie.
certain students have returned their paper
‘Certain students turned their paper in.’
Determiners other than certains (such as des, quelques, plusieurs, etc.) are lexi-
cally non-contrastive. If, from quelques N VP, we may conclude that not every N
satisfies VP, it is not because of the lexical meaning of quelques but rather because
of an application of Grice’s conversational maxim of quantity.
The contrastive partitivity of certains must be distinguished from the partitivity
of explicitly partitive indefinites, as in (3):
(3) Deux de tes étudiants ont rendu leur copie.
two of your students have returned their paper
‘Two of your students handed their paper in.’
1
‘Certains N suggests some of the Ns but not all of the Ns.’
This is not a logical implication but only an implicature, because it is possible to cancel it by using
même ‘even’ and say:
(i) Certains étudiants sont indisciplinés. Tous même, me semble-t-il.
certain students are undisciplined all even me seems- it
‘Certain students are undisciplined. All of them, even, it seems to me.’
5.1 Disambiguating Indefinites: DP-Internal Factors 145
For (3) to be true, it is enough that two students of yours handed their paper in.
Whether your other students handed theirs in or not does not affect the truth value
assigned to the sentence.
Let us now return to Attal’s observation that indefinites introduced by certains
favor a distributive reading. This interpretation distinguishes certains not only from
des (see (1a–b)) but also from other plural indefinites, which favor a collective
reading:
(4) a. Certains étudiants sont passés me dire que…
certain students are come me tell that
‘Certain students came by to tell me that…’
b. Plusieurs/ des/ quelques étudiants sont passés me dire que…
several des some students are come me tell that
‘Several / sm / a few students came by to tell me that…’
It can be shown that distributivity per se is not part of the lexical meaning of
certains. Indeed, the distributivity of certains N can be blocked by predicates that
select pluralities:
(5) Certaines filles se ressemblait.
certain girls refl resemble.impf
‘Certain girls looked alike.’
In examples of this kind, certains remains contrastive-partitive, which gives rise
to the implicature that there are girls that do not look alike. Quantificational DPs,
which are necessarily distributive (see Sect. 6.3 of Chap. 6), are unacceptable in the
same context:
(6) a. *Chaque fille se ressemblait.
each girl refl resemble.impf
‘Each girl looked alike.’
b. *Toute fille se ressemblait.
every girl refl resemble.impf
‘Every girl looked alike.’
We may thus conclude that certains is not inherently quantificational.
However, some correlation seems to exist between the distributivity and the con-
trastive meaning of certains: certains is contrastive and by default distributive
(unless the sentential predicate is collective), whereas des, plusieurs and quelques
are non-contrastive and – in unmarked cases – non-distributive.
It seems possible to account for this correlation as follows. The contrastive meaning
(‘not all’) that characterizes certains must take into account not only the intersection of
the nominal predicate with the predicate of the sentence but also other elements belong-
ing to the set described by the nominal predicate. However, certains is not a proportional
determiner, because we need not check every element in the set denoted by the nominal
predicate in order to assign a truth value to a sentence containing certains. It is sufficient
to find one element in this set that does not have the property denoted by the verb.
The proposal we have just outlined regarding certains can be summarized by

saying that its (contrastive-partitive) lexical meaning gives rise to a tripartite
Logical Form and thus to distributivity. In unmarked cases, distributivity will be
“strict”: the quantification domain will contain atomic individuals (see (4¢)). If the
predicate of the sentence blocks this reading (as with ressembler ‘look alike’,
which can only apply to groups), the quantification domain will contain groups of
individuals (see (5¢)):
(4¢) Certains x (student (x)) [x came by to tell me that …]
(5¢) Certains x (x is a group of girls) [x look alike]
This representation says that, given a contextually identifiable group of girls,
certain subgroups (but not all) belonging to this group contain girls who look
alike.
Indefinites that are neither explicitly partitive nor contrastive are necessarily
interpreted as partitive when they appear in a contrastive context or as subjects of
negative predicates:
(7) a. Plusieurs enfants ne m’ont pas écrit.
several children neg me have neg written
‘Several children didn’t write to me.’
b. Quelques enfants sont arrivés à temps, d’autres ne sont venus que
some children are arrived at time others neg are come only
deux heures plus tard.
2 hours more late
‘Some students arrived on time, others did not come until 2 hours later.’
In this case as well, contrast triggers the distributive reading. This correlation
between contrast and distributivity suggests again that contrast gives rise to a
quantificational structure, which correlates with distributivity:
(7¢) a. Plusieurs x (children (x)) [x did not write to me]
Let us now see how plusieurs differs from quelques.
(8) a. Plusieurs enfants étaient intelligents/ tristes/ blonds/ fatigués.
several children were intelligent/ sad/ blond/ tired
‘Several children were intelligent / sad / blond / tired.’
b. ?? Quelques enfants étaient intelligents.
some children were intelligent
‘Some children were intelligent.’
c. ? Quelques enfants étaient tristes / blonds / fatigués.
some children were sad/ blond/ tired
‘Some children were sad / blond / tired.’
d. Plusieurs/ quelques enfants se ressemblaient.
several/ some children refl resemble
‘Several / some children looked alike.’
Examples (8a–c) have predicates that must apply to atomic individuals.

Example (8a) is acceptable because plusieurs can be distributive (though it does
not have to be, as indicated by (8d). The marginality of (8b–c) is due to quelques
being marked as collective2; it may be distributive only in the presence of a con-
trastive predicate (as in Quelques enfants étaient fatigués, les autres étaient en
train de jouer dans la cour ‘Some children were tired, the others were playing in
the yard’).
Turning now to des-indefinites, on their unmarked use they have a weak read-
ing (see Chap. 2, where parallelisms are drawn between bare plurals in Spanish
and Romanian and indefinites with des in French), which is incompatible with
the quantificational configuration induced by contrast or explicit negation. This
accounts for the marginality of (7c), which contrasts with the acceptability of
(7a–b).3
(7) c. ??Des enfants ne m’ont pas écrit.
des children neg me have neg written
‘Sm children didn’t write to me.’
Let us now consider the examples in (9), due to Galmiche (1986):
(9) a. Des verres sont vides.
des glasses are empty
‘Sm glasses are empty.’
b. Des fauteuils sont bancals.
des armchairs are wobbly
‘Sm chairs are wobbly.’
c. Des verres sont ébréchés.
des glasses are chipped
‘Sm glasses are chipped.’
d. Des fourchettes sont sales.
des forks are dirty
‘Sm forks are dirty.’
e. Des fleurs sont fanées.
des flowers are wilted
‘Sm flowers are wilted.’
2
The collective meaning is even clearer in the case of the Spanish determiner unos (Villalta
1995).
3
There is however a partitive des, which is interpreted like certains (cf. Bosveld de Smet 1997).
(i) Des élèves étaient absents hier.
des students were absent yesterday
‘Some students were absent yesterday.’
f. Des ampoules sont usagées.

des bulbs are worn out
‘Sm bulbs are worn out.’
g. Des cendriers sont pleins.
des ashtrays are full
‘Sm ashtrays are full.’
The acceptability of these examples is surprising: on their unmarked interpreta-
tion, des indefinites are weak, i.e., amount referring, but the examples in (9) are
constructed with entity predicates (see Sect. 2.5 of Chap. 2), which cannot com-
bine with amount-denoting arguments. Note however that the interpretations of the
examples in (9) are not partitive: to paraphrase them, one would not use certains
but rather quelques. The lack of a partitive interpretation is expected under the
assumption that on their unmarked reading, des-indefinites cannot be quantificational
(cannot be represented by tripartite structures). We are therefore led to assume that
in examples such as (9), des-indefinites function as referential terms that refer to
plural individuals. From the perspective proposed here, plural individuals differ
from amounts insofar as they carry a referential index that distinguishes them from
other plural individuals that satisfy the same description. Examples of the type in
(9) can be represented as predicational configurations in which the main predicates
apply to referential terms (see (9¢a)). Note that in (9¢a), the numerical index on
glasses does not indicate the number of glasses but instead is a referential index.
(9¢) a. empty (glasses33)
The referential use of des indefinites illustrated in (9) is subject to severe con-
straints: as pointed out by Galmiche (1986), examples of this kind are acceptable
only when both the subject and the predicate of the sentence express directly per-
ceptible objects and properties, respectively. This indicates that direct perception
allows referential interpretations even for those indefinites that are not lexically
marked as specific. Moreover, examples that trigger a specific reading of des
indefinites are subject to rules of discourse relevance for which the interaction
between the meaning of the verb and the meaning of the noun is a decisive factor.
Thus, examples (9a–g) become ungrammatical if we replace the matrix predicates
by their antonyms. Note that the type of predicate in (9) and (10) does not differ;
entity-predicates are used in both cases:
(10) a. ?? Des verres sont pleins.4
des glasses are full
‘Sm glasses are full.’
4
These examples are acceptable with the partitive meaning equivalent to certains. This is expected
because the partitive des is quantificational and entity-predicates allow quantificational readings.
b. ?? Des fauteuils sont confortables.

des armchairs are comfortable
‘Sm chairs are comfortable.’
c. ?? Des verres sont entiers.
des glasses are whole
‘Sm glasses are whole.’
d. ?? Des fourchettes sont propres.
des forks are clean
‘Sm forks are clean.’
e. ?? Des fleurs sont fraîches.
des flowers are fresh
‘Sm flowers are fresh.’
f. ?? Des ampoules sont neuves.
des bulbs are new
‘Sm bulbs are new.’
g. ?? Des cendriers sont vides.
des ashtrays are empty
‘Sm ashtrays are empty.’
According to Galmiche, examples (10a–g) are unacceptable because they are not
‘relevant’, which means that their content is not sufficiently informative. At a feast,
glasses are supposed to be full. In this context, an utterance is relevant if it describes
the unexpected situation of glasses being empty (see (9a)) and therefore needing to
be refilled.
Evidence in favor of the idea that the predicates in (9) above are entity-predicates
is provided by examples with mass nouns. Mass indefinites are necessarily weak
(compare plural indefinites which may take an entity type denotation as explained
in (9¢a)) and as such they cannot combine with entity-predicates.
(11) a. * De la fumée était noire.
de la smoke was black
‘Sm smoke was black.’
b. ?? Du mobilier était abîmé / en mauvais état.
du furniture was damaged/ in bad condition
‘Some furniture was damaged/ in bad condition.’
To summarize, plural indefinite determiners in French present notable interpre-
tative differences, corresponding to different choices from among the three types
of representation proposed in Chap. 4. In unmarked uses, des-indefinites refer to
amounts, indefinites headed by quelques have a collective interpretation (they
supply variables that range over plural individuals) and certains has a contrastive
meaning (and is correlatively to be represented as a quantificational DP). Plusieurs
is interpreted as a cardinality predicate meaning ‘more than one, several’; it is
compatible with the three types of readings just described. Although they are not
designated as partitives in the lexicon, quelques and plusieurs can acquire a partitive
meaning in the presence of a contrastive predicate. Whether it is due to the lexical
specification of the determiner (as is the case for certains) or induced by the syn-
tactic context, the contrastive interpretation gives rise to a quantificational Logical
Form (tripartite structure) and this representation correlates with a distributive
reading.
5.1.2 Partitive Indefinite DPs
Partitive indefinites such as un/deux/plusieurs des N ‘one/two/several of the N’ are

not necessarily interpreted as distributive, which indicates that they are not neces-
sarily quantificational:
(12) Deux de tes étudiantes se ressemblaient incroyablement.
Two of your students refl resemble.impf incredibly
‘Two of your students looked incredibly alike.’
This example can be represented as in (12¢), where the indefinite is referential
(more precisely it is to be analyzed as a Skolem term, see Chap. 6), which is indi-
cated by the referential index 35:
(12¢) looked incredibly alike (two-of-your-students35)
Overt partitives can even appear in existential sentences:
(13) Dans la salle, il y avait deux de tes étudiants.
in the room there have.impf two of your students
‘In the room were two of your students.’
We know that existential constructions require amount-denoting DPs and are
therefore incompatible with quantificational DPs that quantify over individuals (see
Chap. 3). An example like (13) thus constitutes an argument in favor of the idea that
explicitly partitive indefinites are not necessarily quantificational.5 Formally, this
means that the de DP ‘of DP’ constituent need not always be analyzed as belonging
to the restriction of a tripartite structure; it can also be analyzed as a property restrict-
ing the range of the variable introduced by the weak indefinite, translated in the
Logical Form as an existential generalized quantifier over amounts that combines
with the instantiation predicate il y a. By applying two successive lambda-
conversions to (13¢a), we obtain (13¢b) and finally (13¢c).
5
There are notable differences between languages regarding this point. For example, explicit
partitives in Spanish cannot appear in hay ‘there is’ constructions: (i) *habia dos de tus estudi-
antes en la sala ‘there were two of your students in the room’ vs. (ii) habia dos estudiantes tuyas
en la sala ‘there were two students of yours in the room’ (judgments provided by Helena Lopez
de Palma, p.c.).
(13¢) a. lP $ xa (xa belongs to your students Ù |xa| = 2 Ù P(xa)) • lxa $ xi

(Ind (xi,xa) Ù (xi is in the room))
b. $ xa (xa belongs to your students Ù |xa| = 2 Ù lxa $ xi (Ind (xi,xa)
Ù (xi is in the room)) (xa))
c. $xi$xa (Ind(xi, xa)) Ù xa belongs to your students Ù
xa is in the room Ù |xa| = 2)
The dissociation between partitivity and quantificational status is worth stressing, for
it goes against most of the current analyses of indefinites. It is thus commonly assumed
that an indefinite that is partitive (or D-linked, in the sense of Pesetsky (1987)) is also
quantificational. We have just observed that indefinites that are explicitly partitive are
not necessarily quantificational. On the other hand, it seems plausible that the partitive
meaning of those indefinites that are not explicitly partitive is due to a quantificational
structure: the quantification domain (corresponding to the restrictive term of a tripartite
structure) is interpreted as a contextually determined subset of the set described by
the nominal predicate. The quantificational structure itself is not brought about by the
inherent features of the indefinite DP but rather by properties of the context: the type
of predicate (see Chap. 4, Sect. 4.3.2), the presence of a contrastive context (see
the examples in (7), above), or yet other factors to which we will return below.
In sum, overtly partitive indefinite DPs are ambiguous: they may be amount-
referring (more precisely and in our implementation they denote a generalized
existential quantifier over amounts) or quantificational (denote a generalized
quantifier over individual entities). Contrast induces a tripartite representation,
interpreted as a quantificational configuration. This kind of representation is possi-
ble but not necessary for partitive DPs.
5.1.3 Modified Cardinals
Westerstähl (1989) and Liu (1990) observed that modified cardinals (au moins n ‘at
least n’, exactement n ‘exactly n’, n au plus ‘at most n’) have interesting scopal
properties: unlike unmodified cardinals, which allow several readings (wide scope
(14a), intermediate scope (14b) and narrow scope (14c)), modified cardinals favor a
narrow scope reading (cf. (15)):
(14) Chaque professeur a récompensé chaque élève qui a lu
each professor has rewarded each student who has read
deux romans.
two novels
‘Each professor rewarded each student who read two novels.’
a. There are two novels such that each professor rewarded each student
who read them.
b. Each professor chose two novels and rewarded each student who read
them.
c. Each professor rewarded each student who read any two novels.
(15) Chaque professeur a récompensé chaque élève qui a lu

each professor has rewarded each student who has read
au moins deux romans.
at least two novels
‘Each professor rewarded each student who read at least two novels.’
Kamp and Reyle (1993) showed that modified cardinals behave both like indefinites
and like quantifiers. With respect to scope, they behave like quantifiers (in that they
cannot scope out of the minimal sentential domain to which they belong) but their
anaphoric potential makes them comparable to indefinites. As illustrated by (16b), they
introduce a discourse referent that is accessible in “donkey sentences”, on a par with
the singular indefinite in (16a) and in contrast with the quantified DP in (16c).
(16) a. Tout fermier qui possède un âne le bat.
every farmer who owns a donkey it beats
‘Every farmer who owns a donkey beats it.’
b. Tout fermier qui possède au moins deux ânes les bat.
every farmer who owns at least two donkeys them beats
‘Every farmer who owns at least two donkeys beats them.’
c. * Tout fermier qui possède chaque âne le/ les bat.
every farmer who owns every donkey it/ them beats
‘Every farmer who owns each donkey beats it/them.’
Corblin (2001) points out additional differences between modified and unmodified
cardinals: modified cardinals favor distributive over collective readings and they do
not allow generic readings. These two points are illustrated in (17).
(17) a. Cinq étudiants m’ont offert un livre. (distributive or collective)
five students me have offered a book
‘Five students offered me a book.’
b. Au moins cinq étudiants m’ont offert un livre. (distributive)
at least five students me have offered a book
‘At least five students offered me a book.’
c. Deux jumeaux se ressemblent. (generic reading possible)
two twins refl resemble
‘Two twins look alike.’
d. Au moins deux jumeaux se ressemblent. (generic reading impossible)
at least two twins refl resemble
‘At least two twins look alike.’
These observations led Kamp and Reyle (1993) and Corblin (2001) to postulate
the existence of an intermediate class between cardinals and quantifiers, labeled
‘cardinal quantifiers’. As the name suggests, these expressions would be a sub-type
of quantifiers rather than a sub-type of indefinites.
We will not adopt this analysis: modified cardinals may but need not be
quantificational. DPs introduced by modified cardinals acquire quantificational status
under the same contextual conditions as other indefinites, in particular in the subject
position of entity-predicates (cf. (18a)) or in contrastive contexts (see (18b)):
(18) a. Au moins trois filles étaient blondes.
at least three girls were blond
‘At least three girls were blond.’
b. Même si la plupart des étudiants n’ont pas réussi,
even if the majority of the students neg have neg succeeded
au moins trois étudiants ont eu 18/20.
at least three students have had 18/20
‘Even if most students did not pass, at least three students received 18 out
of 20.’
Like indefinites and unlike quantificational DPs, they may appear in il y a
constructions:
(19) a. Dans la salle, il y avait au moins deux étudiants.
in the room there have.impf at least two students
‘In the room were at least two students.’
b. *Dans la salle, il y avait chaque étudiant.
in the room there have.impf each student
‘In the room was each student.’
c. Il y a au moins trois étudiants qui ont triché.
there has at least three students who have cheated
‘There are at least three students who cheated.’
d. * Il y a chaque étudiant qui a triché.
there has each student who has cheated
‘There is each student who cheated.’
In this context, modified cardinals cannot be analyzed as quantificational but
must be analyzed as weak, i.e., as referring to amounts.
Our proposal is that modified cardinals differ from run-of-the-mill indefinites in
that they do not allow a non-quantificational strong reading, i.e., a specific referen-
tial interpretation; in other words, they cannot be represented as Skolem terms (see
the difficulty of a collective reading in the subject position in (17b)6). The other
6
Note that a collective interpretation of modified cardinals is related to an eventive interpretation
of the sentence, in which the modified cardinal is a weak indefinite:
(i) Au moins cinq étudiants se sont réunis hier.
at least five students refl are gathered yesterday
‘At least five students have gathered yesterday.’
effects observed by Kamp and Reyle (1993) and Corblin (2001) can be shown to
follow from this hypothesis. As we will see in Chaps. 6 and 7, wide and intermediate
scope effects (see (14a–b)) on the one hand and generic interpretation (see (17c) vs.
(17d)) on the other hand, are possible only if a given indefinite can translate as an
individual variable.
In short, modified cardinals are indefinites that cannot be represented as indi-
vidual variables. We may now wonder whether this impossibility is linked to the
form of the DP (i.e., to the modification of the cardinal). A possible hypothesis
relies on postulating a homogeneity condition on quantification domains: a plural
indefinite can map onto a variable that ranges over pluralities only if the domain of
variation is homogeneous. A modified cardinal like au moins deux N ‘at least two
N’ denotes a set of amounts of varying cardinality: amounts of two N, of three N,
etc. However, this constraint could not explain why exactement un N ‘exactly one
N’ (or more generally exactement n N ‘exactly n N’), which defines a homogeneous
quantification domain, behaves in the same way as au moins n N ‘at least n N’ or au
plus n N ‘at most n N’.
Another line of explanation was proposed by Krifka (1999), who assumes that
modified cardinals are not determiners but rather complex expressions made up of
an adverb and a cardinal. Au moins, au plus and exactement have the same seman-
tics when applied to a numeral like quatre ‘four’ or when applied to some other
linguistic unit, as illustrated in (20). These adverbs affect the truth conditions of
the sentence, not the interpretation of the determiners. In others terms, even when
they are modified, cardinals are analyzed as cardinals, i.e., as indefinites, which
are ambiguous:
(20) a. Jean a vu au moins Marie.
Jean has seen at least Marie
‘Jean saw Mary at least.’
b. Pierre était exactement à l’heure.
Pierre was exactly at the hour
‘Pierre was exactly on time.’
c. Jean demandera au plus une petite somme d’ argent.
Jean ask.fut at most a small sum of money
‘Jean will ask for at least a small amount of money.’
Further reasons can be found for not considering modified cardinals to be deter-
miners. In the theory of generalized quantifiers, it was shown that conservativity is
a characteristic property of natural language determiners. Accordingly, elements
such as only in English, which can appear in a prenominal position but are not
conservative, are excluded from the set of English determiners. Certain modified
cardinals are, at least in certain uses, comparable to only in that they are not con-
servative. Take the case of exactement trois ‘exactly three’ and au plus trois ‘at
most three’.
(21) a. Exactement trois filles ont menti.

exactly three girls have lied
‘Exactly three girls lied.’
b. Au plus trois filles ont menti.
at most three girls have lied
‘At most three girls lied.’
In these examples, the modified cardinal is not conservative: in order to deter-
mine the truth value of these sentences, one must know not only the set of all girls
and the set of girls who lied but also the set of all liars. So for example, (21a) and
(21b) can be said to be false in contexts where the set of liars contains boys as well
as girls. Assigning a truth value requires examining the set of liars and checking that
it is composed of exactly three or at most three girls, respectively. One must make
sure that there are no liars that are boys.
A complete analysis of modified cardinals goes far beyond the aims of this book.
For our present concerns it is sufficient to assume a negative characterization,
according to which DPs headed by modified cardinals cannot be represented as non
quantificational strong indefinites. The scopal phenomena examined in Chap. 6 will
give further evidence in favor of this hypothesis.
5.1.4 Prepositional Accusatives: Denotation Type and Specificity
Certain languages have two distinct forms for direct objects, one unmarked, the
other marked with a morphological case – as in Turkish, cf. Enç (1991) – or with a
preposition – like a in Spanish (Jaeggli 1982; Bleam 1999) and pe in Romanian
(Dobrovie-Sorin 1990, 1993; Cornilescu 2000). In this section we will examine
Romanian and Spanish prepositional Accusatives and we will show that the use of
a marked accusative blocks the weak reading; in other words, marked accusatives7
cannot be analyzed as amount-referring expressions. This constraint leaves room
for crosslinguistic variation.8 Thus, Romanian prepositional Accusatives necessar-
ily take specific and wide-scoped interpretations (Farkas 1978; Dobrovie-Sorin
1987, 1990, 1993, 1995), whereas Spanish prepositional Accusatives may also be
interpreted as non specific and narrow scoped indefinites.
5.1.4.1 Romanian
In Romanian, the prepositional marking of the direct object correlates with clitic dou-
bling: in (22b) and in (24b), the direct object is doubled by the clitic l- (accusative
7
The notion of ‘strong Case’ was used by de Hoop (1992) to refer to prepositional or morphologically
marked Accusatives, as opposed to unmarked objects, assumed to be marked with a ‘weak Case’.
8
For a crosslinguistic analysis of specific indefinites, and in particular of prepositional Accusatives,
see von Heusinger (2002).
masculine third person singular) and in (23b), the direct object is doubled by the clitic
o (accusative feminine third person singular). The clitic doubling of the direct object
was optional in earlier stages of Romanian but it has become obligatory in the modern
language. Examples (22a) and (23a) show that non-prepositional accusatives cannot
be doubled by clitics (this impossibility characterizes all stages of Romanian):
(22) a. *L-am întîlnit un prieten.
him.acc-have.1sg met a friend
b. L-am întîlnit pe un prieten.
him.acc-have.1sg met pe a friend
‘I have met a friend.’
Turning now to their interpretation, prepositional accusative indefinites receive a
specific interpretation and they take obligatory wide scope.9
The specific interpretation of prepositional objects is indicated by the fact that
they do not give rise to “weak cross-over” (or “WCO”) effects, whereas unmarked
objects yield WCO effects10:
(23) a. ?* Mama eii va ajuta [una din studentele tale]i.
mother her.gen will help one of students yours
b. Mama eii oi va ajuta pe [una din studentele tale]i.
mother her.gen her.acc will help pe one of students yours
‘Heri mother will help [one of your students]i.’
Note also that a donkey sentence type of reading is blocked by prepositional
accusatives:
(24) a. Toţi profesorii care au cunoscut un student excepţional
all professors-the who have met a student exceptional
îşi amintesc de el.
refl remember of him
‘All the professors who knew an exceptional student remember him.’
b. ?* Toţi profesorii care l-au cunoscut pe un student
all professors-the who him-have met pe an student
excepţional îşi amintesc de el.
exceptional refl remember of him.
9
Here, we will not be looking at the distribution of prepositional accusatives with pronouns like
nimeni ‘nobody’, cineva ‘someone’, etc.
10
The term cross-over refers to the following generalization (cf. Postal 1971; Wasow 1972, 1979):
a relative or interrogative pronoun cannot cross (or move past) a pronoun with which it is coin-
dexed. Weak cross-over violations arise in structures where the coindexed pronoun does not
c-command the trace left by the movement of the relative pronoun. For example, in *Whoi does
hisi mother love ei, the pronoun his is embedded in the DP his mother and thus does not c-com-
mand the trace of who in the object position. Weak cross-over effects are also observed where a
pronoun is coindexed with a quantified DP to its right (*hisi mother loves nobodyi). This kind of
example is parallel to examples with wh-movement (in relatives or interrogatives) if it is assumed
that quantified DPs move at the level of Logical Form.
These properties show that prepositional accusative indefinites behave like

referential DPs, e.g., definite DPs or proper names. (25) shows that proper names do
not give rise to weak cross-over violations and that they can be doubled by clitics
(in fact they must be clitic doubled in contemporary Romanian):
(25) Mama luii îli va ajuta pe Ioni.
mother his him.acc will help pe John
‘Hisi mother will help Johni.’
Finally, Romanian prepositional-marked indefinite objects cannot take narrow
scope:
(26) a. i. Fiecare profesor va examina zece elevi.
each professor will examine ten students.
‘Each professor will examine ten students.’
ii. Toţi judecătorii cunosc doi avocaţi.
all judges-the know two lawyers
‘All the judges know two lawyers.’
iii. Toţi copiii au văzut un actor celebru.
all children-the have seen an actor famous
‘All the children saw a famous actor.’
b. i. Fiecare profesor îi va examina pe zece elevi.
each professor them will examine pe ten students
ii. Toţi judecătorii îi cunosc pe doi avocaţi.
all judges-the them know pe two lawyers
iii. Toţi copiii l-au văzut pe un actor celebru.
all children-the him-have seen pe a actor famous
(27) a. Ieri n-am examinat mulţi elevi.

yesterday neg-have.1sg examined many students
b. Ieri nu i-am examinat pe mulţi elevi.
yesterday neg them-have.1sg examined pe many students
‘I didn’t examine many students yesterday.’
In the examples in (26a), each professor, judge or child can examine, know
or see different students, lawyers or actors, respectively. The examples in (26b),
on the other hand, are concerned with specific groups of ten students, two law-
yers or one actor, with which the speaker is acquainted. In other words, the
prepositional marked Accusatives in (26b) show speaker-specific readings
(Farkas 1978). In (27a), mulţi ‘many’ has to be interpreted in the scope of nega-
tion and the sentence therefore means that I examined few children, as opposed
to (27b), which must be interpreted as ‘many children are such that I did not
examine them’.
A specific indefinite is necessarily presuppositional but not necessarily partitive.11

We can see this in (26b)(i), which presupposes the existence of a group of ten stu-
dents, without necessarily presupposing the existence of another set of students,
contextually determined, to which the group of ten belongs. Note that definite DPs,
as well as proper names are also presuppositional but non partitive.
The correlation between specificity and wide-scope observed here for Romanian
prepositional-marked indefinite objects constitutes evidence in favor of the analy-
sis of scope adopted in Chap. 6, where we will see that the so-called ‘wide scope’
interpretation of indefinites is in fact not due to a genuine scope mechanism, e.g.,
Quantifier Raising, but rather to the specific interpretation of indefinites. A specific
indefinite refers to a particular individual, which belongs to the world of the speaker
(therefore, we can speak about speaker specificity). Their specificity prevents
Romanian prepositional indefinite objects from taking narrow scope with respect
to other quantifiers.
In sum, Romanian prepositional indefinite objects qualify as strong non
quantificational indefinites: they are necessarily specific (i.e., they carry a referen-
tial index that is fixed with respect to a given discourse context or with respect to the
world of the speaker) and as such they cannot take narrow scope. However, they do
not qualify as quantificational, because they are not necessarily partitive.
5.1.4.2 Spanish
Romanian and Spanish differ regarding the use and interpretation of prepositional
accusatives. Thus, there are contexts in which the prepositional accusative is obliga-
tory in Spanish, even if the indefinite in question does not receive a specific
interpretation. In (28b) the girl that John loves need not be known by the speaker or
assumed to refer to a particular girl in the discourse context:
(28) a. *Juan ama una chica.
Juan loves a girl
‘Juan loves a girl.’
b. Juan ama a una chica.
Juan loves a a girl
‘Juan loves a girl.’
Examples (28a–b) are constructed with entity-predicates, which block the weak
reading of the DP in the object position, while allowing both a specific and a non-
specific reading. The examples in (28a–b) show that in Spanish, prepositional accu-
satives are obligatory for non-weak indefinite objects, regardless of whether they
11
Enç (1991) analyzes morphological accusatives in Turkish as being necessarily partitive. In
Romanian as well, prepositional accusatives are partitive in most contexts. However there are
examples of Romanian prepositional accusatives that are specific without being partitive.
are specific or not. An indefinite that is not a-marked is necessarily interpreted as

weak.12
The Romanian examples in (29a–b), which correspond with (28a–b), are both
grammatical but differ with respect to speaker-specificity:
(29) a. Ion iubeşte o studentă.
Ion loves a student
b. Ion o iubeşte pe o studentă.
Ion her loves pe o student
‘Ion loves a student.’
In (29a), the object is interpreted as non-specific, more precisely, the student
loved by Ion is not known by the speaker and non-partitive, whereas in (29b) it can
have either a partitive reading (presupposing a contextually determined set of stu-
dents) or a specific reading (the student in question can be identified by the
speaker).
In the brief presentation of the Romanian data, we showed that the specificity of
the prepositional indefinite objects in that language correlates with the impossibility
of taking narrow scope. Given that Spanish prepositional accusatives are not neces-
sarily specific, we expect them to allow a narrow scope interpretation.
Examples (30a–c), taken from Bleam (1999:176), show this prediction to be
correct:
(30) a. Juan y Maria buscan (a) una chica que sepa español.
John and Mary seek (a) a girl who know.subj Spanish
b. Juan y Maria buscan a una chica que sabe español.
John and Mary seek a a girl who knows Spanish
c. *Juan y Maria buscan una chica que sabe español.
John and Mary seek a girl who knows Spanish
It is known that subjunctive relative clauses force a de dicto reading (i.e., a read-
ing in which the indefinite is interpreted as narrow-scoped with respect to an opacity
operator, e.g., the verb seek), while indicative relatives force a de re reading (i.e.,
readings in which the indefinite is interpreted out of the scope of an opacity opera-
tor, e.g., the verb seek). Thus, (30a), built with a subjunctive, can only have a de
dicto interpretation. The fact that the prepositional accusative is optional in (30a)
indicates that it does not block the de dicto reading, i.e., a narrow scoped reading
with respect to seek.
The indicative, on the other hand, forces a de re reading, i.e., a reading in which
the indefinite is not interpreted in the scope of seek. Such a reading is allowed in
(30b), where the object is marked by a, and impossible in (30c), where the object
12
In the words of Bleam (1999:180), who identifies “weak” and “property denoting”: “non-a-
marked animate DPs [in Spanish] are interpreted as properties”.
is not marked. The example in (30c) is ungrammatical due to the contradictory

requirements imposed on the indefinite: the presence of an indicative relative
clause forces the indefinite to take a de re reading but the absence of the preposi-
tional marking forces the indefinite to take a weak reading and therefore a de
dicto reading.
In sum, an unmarked indefinite object is necessarily interpreted as taking narrow
scope in Spanish, whereas a prepositional accusative can take both narrow and wide
scope with respect to an opacity creating verb. Granting that the wide-scope (or
rather the non-narrow-scoped) interpretation of indefinites is in fact due to a specific
reading, prepositional Accusatives in Spanish must be assumed to allow for specific
readings. However, the fact that in this language prepositional Accusatives can also
take narrow scope indicates that they are not necessarily specific.
According to some theoreticians (see in particular McNally and van Geenhoven
(1998)) the narrow scope interpretation of indefinites correlates with a weak reading
of the indefinite. The prepositional accusative version of (30a) shows that this cor-
relation cannot be maintained: indeed, the object takes narrow scope, although it is
marked by a, which blocks the weak reading. In Chap. 6 it will be proposed that
certain narrow-scoped indefinites are to be analyzed as dependent indefinites (rather
than as weak indefinites) and that dependent indefinites are to be represented as
dependent Skolem terms. This means that narrow-scoped prepositional Accusatives
in Spanish can be analyzed as dependent Skolem terms, whereas narrow-scoped
unmarked Accusatives are weak indefinites.
Going back to Romanian, unmarked accusatives are ambiguous, allowing both
specific and non specific readings and correlatively allowing both wide and narrow
scope (see (31b–c)), whereas the prepositional accusative, which forces specificity,
rules out the possibility of a de dicto reading (see (31a)).
(31) a. *Ion şi Maria caută pe o fată care să vorbească spaniola.
John and Mary seek pe a girl who speak.subj Spanish
b. Ion şi Maria caută o fată care să vorbească spaniola.
John and Mary seek a girl who speak.subj Spanish
c. Ion şi Maria o caută (pe) o fată care să vorbească spaniola.
John and Mary her seek pe a girl who speak.subj Spanish
5.1.4.3 Conclusions
The generalizations presented above regarding prepositional Accusatives and

unmarked objects in Spanish and Romanian can be summarized as follows:
Weak Non- Narrow Wide
reading specific scope Specific scope
Romanian PrepAccus No No No Yes Yes
Spanish PrepAccus No Yes Yes Yes Yes
Romanian Unmarked Obj Yes Yes Yes Yes Yes
Spanish Unmarked Obj Yes Yes Yes No No
5.2 Information Structure and the Disambiguation of Indefinites 161
In both Romanian and Spanish, the use of the prepositional Accusative trig-
gers a strong non quantificational reading.13 But Romanian and Spanish preposi-
tional Accusatives differ regarding specificity and correlatively regarding their
scopal interpretations: Romanian prepositional Accusatives are necessarily
specific and wide-scoped, whereas Spanish prepositional Accusatives can be
both specific and non-specific and correlatively they can allow both wide and
narrow scope. It seems reasonable to suggest that the obligatory specificity of
Romanian prepositional Accusatives is due to the fact that in this language, prep-
ositional Accusatives are clitic-doubled: it is the presence of the doubling clitic
that would trigger the referential indexing of prepositional Accusatives. Since
a-marked Accusatives are not clitic-doubled in Spanish, they are not necessarily
referentially indexed.14
5.2 Information Structure and the Disambiguation

of Indefinites
Insofar as the interpretation of indefinites is contextually determined, the best

theoretical option would be to assign a unique representation to the indefinite
itself and analyze interpretative differences as being induced by a difference in
context. Such is the theoretical position adopted by Reinhart (1995), who claims
that the different readings of indefinites can be characterized in terms of (the pres-
ence or absence of) presuppositionality and depend on the information structure
of the sentence. We will follow Reinhart’s view that indefinites in the Topic posi-
tion are necessarily presuppositional but we will depart from Reinhart by showing
that there are indefinites that are presuppositional without occupying the Topic
position.
5.2.1 Indefinites in the Topic Position

Are Presuppositional
Syntactic structure is underdetermined with respect to Topic-Comment structure,

the same syntactic structure being compatible with more than one information
structure. Accordingly, a sentence of the form Subject-Verb-Object can be assigned
at least two different structures: (i) no Topic, or (ii) the subject is Topic. The Topic
may be defined as being the constituent about which something is asserted in a
13
This generalization holds for many other languages, in particular for Turkish (Enç 1991) and
Hebrew (Danon 2002).
14
In Chap. 6 it will be proposed that narrow scoped indefinites that are not weak are to be ana-
lyzed as dependent indefinites.
categorical statement. As an answer to a question like ‘What happened?’, an SVO

sentence is analyzed as having a thetic structure (i.e., as having no Topic, its whole
content belonging to the Comment). In an answer to a wh-question like ‘What did
Jean do ?’, the element corresponding to the wh-phrase is mapped onto the Focus
and everything else, in particular the subject, goes to the Ground. Since in the gen-
eral case the Topic belongs to the Ground, the subject in the type of sentence just
described is interpreted as a Topic. The choice between options (i) and (ii) is
context-dependent and can sometimes be marked into nationally. In what follows
we will be concerned with how the Topic-Comment partition bears on the interpre-
tation of indefinites.
Let us examine the difference between existential il y a sentences and sentences
with preverbal subjects:
(32) a. Il y avait un roi américain à New York.
there have.impf a king American in New York
‘There was an American king in New York.’
b. Il n’y avait pas de roi américain à New York.
there neg have.impf neg de king American in New York
‘There were no American kings in New York.’
c. Il y avait deux rois américains à New York.
there have.impf two kings American in New York
‘There were two American kings in New York.’
(33) a. Un roi américain vivait à New York.
a king American live.impf in New York
‘An American king lived in New York.’
b. Aucun roi américain ne vivait à New York.
no king American neg live.impf in New York
‘No American king lived in New York.’
c. Deux rois américains vivaient à New York.
two kings American live.impf in New York
‘Two American kings lived in New York.’
Reinhart (1995) points out that speakers do not hesitate in judging (32a) and
(32c) to be false and (32b) to be true. These judgments are consistent with the fact
that existential il y a sentences are structurally unambiguous: the entire sentence
belongs to the Focus and correlatively, the indefinite DP receives a weak reading.
In contrast, speakers are hesitant when judging examples like (33a–c), which are
compatible with two distinct information structures, all-Comment and Topic-
Comment. Correlated with these two configurations are two different interpreta-
tions of indefinites occupying the preverbal subject position: weak (i.e., non-
presuppositional) and presuppositional, respectively. In the first case, the judgments
of acceptability for (33a–c) are identical to those for (32a–c). But in the second
case, the indefinites in (33a–c) presuppose that the set of American kings is not
empty. As this presupposition conflicts with the speakers’ knowledge, they are
unable to assign a truth value to (33a–c). The fact that speaker judgments vary for
(33) can be understood as being due to the uncertainty regarding the analysis of
these examples: some speakers may analyze them as existential sentences (i.e., as
conveying the same meaning as (32)) and thus judge them true or false; other speakers
may analyze them as Topic-Comment configurations and thus judge them to be
indeterminate, i.e., neither true nor false, because of presupposition failure.
While we agree with this analysis, we do not agree with Reinhart’s hypothesis
that the (non-)presuppositionality of indefinites is a purely pragmatic effect due to
the process by which speakers assign a truth value to a Topic-Comment structure.
Contrary to Reinhart, we believe that the differences between presuppositional and
non-presuppositional interpretations are not simply pragmatic but that they corre-
late with differences in denotation type: presuppositional DPs denote either indi-
viduals (type e) or generalized quantifiers over individuals (type < <e,t>,t>), whereas
non-presuppositional DPs are weak DPs (which denote generalized existential
quantifiers over amounts). Information structure partly determines which of these
denotations are possible in a given context.
5.2.2 Non-topical Presuppositional Indefinites
Let us consider the following examples:

(34) Un roi américain était ivre/ malade/ affamé.
a king American was drunk/ sick/ starving
‘An American king was drunk/ sick/ starving.’
These sentences seem to be indeterminate rather than false. In line with Reinhart’s
attempt to derive the presuppositionality of indefinites from their topical status, one
might assume (35):
(35) The subject of an entity-predicate is necessarily a Topic.
But this generalization is incorrect, as indicated by the fact that in the context of
questions (36) and (37), the subject DPs of the answers given in (36¢) and (37¢) are
not Topics. Indeed, as answers to (36) and (37), the sentences in (36¢) and (37¢) are
all-Focus, and since all-Focus sentences are thetic, they have no Topic and therefore
their subject cannot be assumed to be a Topic.
(36) Pourquoi es-tu sous le choc?
why are you under the shock
‘Why are you in shock?’
(36¢) Pendant mon cours, un roi Américain était ivre/ malade
during my class a king American was drunk / sick
‘During my class, an American king was drunk / sick.’
(37) Pourquoi veux-tu t’arrêter?

why want you refl stop
‘Why do you want to stop?’
(37’) (Parce que) plusieurs élèves sont malades.
(because) several students are sick
‘(Because) several students are sick.’
And yet, these examples display presuppositional effects comparable to those in
(34). These effects are unexpected under Reinhart’s analysis, for which the presup-
positionality of indefinites is entirely due to information structure.
These facts are accounted for under the alternative analysis proposed here,
according to which presuppositionality is due to denotation type: entity-predicates
are incompatible with weak (i.e., amount-denoting) DPs, allowing only individual-
type or generalized quantifier denotations, which are necessarily associated with
presuppositionality.
An even clearer case is provided by examples built with verbs like aimer ‘to
love’, haïr ‘to hate’, admirer ‘to admire’, as in (38), which contrast with (39)15:
(38) Jean aime la musique / les femmes / *de la musique / ?? des femmes.
Jean likes the music the women de la music des women
‘Jean likes music / women.’
(39) Jean écrit de la poésie/ des poèmes/* la poésie /* les poèmes.
Jean writes de la poetry des poems the poetry the poems
‘Jean writes poetry / poems.’
The ungrammaticality of de la / des indefinites in the object position of (38) can
be attributed to the fact that predicates such as aimer ‘love’ are entity-predicates,
which block the weak reading of their object. This analysis correctly predicts that
these predicates give rise to presuppositionality effects regarding the object. Note
now that in these examples, the object DPs are not Topics: the sentences in (38) say
something about Jean, not about music or women.
The important point here, which goes against Reinhart’s proposal, is that entity-
predicates give rise to strong readings (and thus to presuppositionality) of DPs that
are not Topics.
5.2.3 Only Indefinites in Topic Positions Are Quantificational
Following Diesing (1992), most theoreticians assume a “semantic partition” mecha-

nism by virtue of which:
(40) a. An indefinite in Spec, IP is quantificational.
b. An indefinite belonging to the VP is existential.
15
For further observations regarding this phenomenon see Kanouse (1972), Lawler (1972),
Declerck (1987) and Laca (1990).
As an illustration, consider the example in (41), which is ambiguous, as indicated

in (42)a–b :
(41) Deux professeurs ont acheté un livre.
two professors have bought a book
‘Two professors bought a book.’
(42) a. ‘Two of the professors bought a book’.
b. ‘There were two professors who bought a book’
Diesing’s hypothesis can account for the ambiguity of (41) by assuming that
the subject DP can either remain in the position it occupies at S-Structure
(namely, (Spec, IP)), or it can lower into the VP-internal subject position (namely,
(Spec,VP)):
(43) a. [Spec,IP Deux professeurs] [I’[I ont] [VP acheté un livre]
b. [Spec,IP tDeux professeurs] [I’[I ont] [VP Deux professeurs acheté un livre]
Given these two representations, the two interpretations of (41) follow from
hypotheses (40a–b). There is however no independent motivation for postulating
the representations in (43a–b) and lowering is a problematic syntactic operation.
We therefore need to revise Diesing’s analysis, by taking out the syntactic
implementation, which obscures that which truly disambiguates indefinites, namely
information structure. As we already saw in Sect. 5.2.1, in languages such as
English or French, a preverbal subject can be a Topic but does not have to be one
(see examples (33) above). So (41) can have either of two information structures,
which correlate with different representations of the indefinite itself. We will
assume the principle stated in (44), according to which Topic status is a necessary
condition for an indefinite to be quantificational:
(44) An indefinite DP may be quantificational only if it is a Topic.16
The conceptual motivation for this correlation is that information structure and
quantification are both asymmetrical. Absence of symmetry is lexical for
quantificational determiners such as tout ‘every’ and chaque ‘each’, whereas for
determiners that are intersective and symmetric, a quantificational representation is
only possible if the asymmetry that is characteristic of quantification is induced by
the syntactic configuration (e.g., the partition between VP-internal and VP-external
material) or a Topic-Comment partition.
5.2.4 Indefinites at the Left Periphery
Above, we have shown that indefinites may appear in the Topic position, with
notable semantic effects. Yet, this generalization seems problematic in view of the
16
This is an extension of the generalization proposed by Tasmowski and Laca (2000) for a specific
case, that of unos (‘some’, ‘a few’) in Spanish.
following examples, in which an indefinite cannot be left dislocated, although it is

currently assumed that dislocated constituents are Topics:
(45) a. *Un étudiant, il m’a aidé à porter la table.
a student he me has helped to carry the table
‘A student, he helped me with carrying the table.’
b. *Un étudiant, Jean l’a examiné deux fois.
a student Jean him has examined two times
‘A student, Jean examined him twice.’
The examples in (45) differ from those examined in previous sections in that
the indefinite DP is a left dislocated constituent, separated from the rest of the
sentence by an intonational pause (indicated by a comma) and resumed by a per-
sonal pronoun. Such left-dislocated elements are adjoined to a sentence without
being a part of it: they are neither constituents belonging to the syntactic structure
of the minimal sentence, nor are they elements of the semantic composition of the
sentence. Their only link to their host sentence is anaphoric, usually indicated by
means of an anaphoric pronoun. In sum, the Topics examined in previous sections
are sentence-internal, whereas the Topics in (45a–b) are not part of the minimal
sentence to which they attach. Although we cannot pursue this hypothesis here, it
seems natural to assume that the structural difference between sentence-internal and
sentence-external Topics17 correlates with the pragmatic difference between senten-
tial and discourse Topics (for the notion of ‘discourse Topic’ see Büring (1997)).
As far as we know, the ungrammaticality of (45) has not yet received a satisfac-
tory explanation. Our tentative suggestion is that a pragmatic constraint on the use
of sentence-external Topics may be at work. The ‘common ground’ (to be kept dis-
tinct from the notion of Ground used in Sect. 5.2.1) refers to the shared knowledge
of the speaker and the hearer:
(46) A sentence-external Topic must belong to the common ground (of the speaker
and the hearer).
Because proper names and definite DPs that have unique referents are part of the
common ground, such DPs are legitimate sentence external Topics:
(47) a. Jean, il m’a aidé à porter la table.
Jean he me has helped to carry the table
‘John, he helped me with carrying the table.’
b. Mon frère, Jean l’a examiné deux fois.
my brother Jean him has examined two times
‘My brother, Jean examined him twice.’
17
Regarding the various left peripheral constructions in French see Fradin (1988).
Going back to indefinites, it is important to observe that specificity is not sufficient

for an indefinite to belong to the common ground. Indeed, speaker-specific indefinites
(which belong to the world of the speaker) are hard to interpret as belonging to the
world of the hearer. Therefore the indefinites in (45) do not belong to the common
ground and as such they are not legitimate sentence-external Topics. Hence the
unacceptability of (45).18
Let us now turn to examples of the type in (48), in which the dislocated indefinites
refer to kinds, as clearly indicated by the use of the special pronoun ça19:
(48) a. Un étudiant, ça sait lire et écrire.
a student this knows read and write
‘A student, it knows how to read and write.’
b. Un chat, ça ronronne.
a cat this purrs
‘A cat, it purrs.’
18
Jean-Marie Marandin (2011, personal communication) observes that specificity does play a role
in the acceptability of certain examples:
(i) Un étudiant que je n’avais pas vu depuis 30 ans, je l’ai rencontré hier à l’opéra.
‘A student that I had not seen for 30 years, I met him yesterday at the opera.’
(ii) Un étudiant à toi, il m’a aidé à porter la table.
‘A student of yours, he helped me with carrying the table.’
(iii) Un de tes étudiants, je l’ai examiné deux fois.
‘One of your students, I examined him twice.’
According to us, these examples are not fully acceptable (they are clearly degraded compared
to (47)). We furthermore believe that the structure of these examples is different from those in
(45) and (47). The acceptability of certain examples is ameliorated if the indefinite is modified by
a restriction that indicates some relation to the hearer (see (ii) and (iii)). However, these examples
are not perfectly acceptable, because the referent of the indefinite itself is not part of the knowl-
edge of the hearer.
19
Besides kinds, ça can refer to propositions, events, or not yet individualized entities but not to
individualized or specific entities:
(i) Je ne peux pas croire ça.
I NEG can NEG believe this
‘I cannot believe this.’
(ii) Je ne veux pas que ça m’arrive une nouvelle fois.
I NEG want NEG that this me-happens a new time
‘I don’t want this to happen to me once again’
(iii) *Marie, je ne veux pas que ça vienne me chercher.
Marie I NEG want NEG that this comes me look for
(iv) *Un étudiant à toi, je ne veux pas que ça vienne me chercher.
A student of yours I NEG want NEG that this comes me look for
The grammaticality of dislocated kind-referring indefinites is expected under the

proposal sketched here. Indeed, the constraint in (46) is satisfied, since kinds belong
to the common ground of the speaker and the hearer.
Let us further observe that amount-referring indefinites may also be viewed as
belonging to the common ground: the existence of unspecified amounts that instan-
tiate kinds is part of the assumed knowledge of speakers and hearers. We therefore
expect amount-referring indefinites to be acceptable in dislocated positions. This is
indeed true, as shown in examples (49),20 built with French indefinites headed by
du/ de la, which have been analyzed as referring to amounts of substance in Chaps.
2, 3 and 4:
(49) a. Du caféi, j’eni ai acheté ei hier. (e = empty category)
du coffee I en have bought yesterday
‘Sm coffee, I bought some yesterday.’
b. Du caféi, je sais qu’il eni a acheté ei hier.
du coffee I know that he en has bought yesterday
‘Sm coffee, I know that he bought some yesterday.’
c. De l’ori, je sais qu’il eni a été acheté ei.
de l gold I know that he en has been bought
‘Sm gold, I know some was bought.’
Let us finally consider the Romanian examples in (50), which allow direct objects
(marked by PE) to occur at the left edge of the sentence:
(50) a. Pe un student Ion îl va examina de două ori.
pe a student Ion him will examine of two times
‘There is a student that Ion will examine twice.’
Although these examples seem to be built in the same way as the French exam-
ples in (45), they are in fact different. Note first that there is no perceptible pause
between the preposed object and the rest of the sentence and correlatively no comma
is used in writing. Furthermore, the relation between the object pronoun and the
preposed object DP is not an anaphoric relation but rather the type of relation that
underlies clitic-doubling. Indeed, if the indefinite is speaker-specific, the clitic is
20
The dislocation of mass DPs proves ungrammatical in (i) and (ii) but this is due to syntactic
reasons: en cannot be linked to an empty category in the preverbal subject position. Example (iii)
is grammatical, for en is linked to the postverbal position of an impersonal il construction.
(i) *Du caféi ei s’eni est acheté hier. (e = empty category)
‘Sm coffee was bought yesterday.’
(ii) *Du caféi ei eni a été acheté hier.
‘Sm coffee was bought yesterday.’
(iii) *Du caféi, il s’eni est acheté ei hier.
‘Sm coffee, there was some bought yesterday.’
5.3 Conclusions 169
present regardless of whether the object DP is at the left edge of the sentence, as in
(50), or inside the sentence, as in (51):
(51) Ion îl va examina pe un student de două ori.
Ion him will examine pe a student of two times
‘Ion will examine a student twice.’
These observations indicate that preposed direct objects in Romanian exam-
ples of the type in (50) are not external to the minimal sentence: they may be
analyzed as occupying (Spec, IP), a position that in this language can host not
only subjects but also objects (cf. Dobrovie-Sorin 1987). Indefinites occupying
this position are not subject to the constraint in (46), which explains why the
Romanian examples in (50) are fully grammatical, in clear contrast with the
French examples in (45).
In sum, sentence external Topics should be distinguished from sentence internal
Topics. Indefinites can occupy the position of sentence internal Topics (in which
case they can be either specific or quantificational but not weak) but specific
indefinites cannot function as sentence external Topics, whereas weak indefinites
(amount-referring) as well as kind-referring indefinites can do so. Because they are
outside the minimal sentence, sentence external Topics are not part of the semantic
composition of the sentence to which they attach. The semantic composition takes
into account the pronoun that has the sentence external Topic as an antecedent.
Sentence internal Topics, on the other hand, are part of the semantic composition
and they impose a strict constraint on the DP in Topic position: it cannot be weak.
5.3 Conclusions
Even though indefinite determiners are unambiguous (they are semantically defined
as intersective), indefinite DPs allow three distinct representations: (i) weak
indefinites refer to amounts, i.e., to non individualized entities; (ii) entity-referring
indefinites (type e); (iii) quantificational indefinites. For certain determiners, cardi-
nals in particular, all three possible analyses can be observed. One or two of these
representations may be unavailable due to various factors, such as: (a) the lexical
properties of certain determiners (see the differences between certains ‘certain’,
quelques ‘some’, plusieurs ‘several’ and des ‘deaccented ‘some”); (b) the internal
structure of an indefinite DP (absence of a determiner, partitive structure); (c) the
lexical properties of the matrix predicate (entity-predicates, contrastivity); (d) use of
object markers; (e) syntactic position; (f) information structure. The first four factors,
which are lexical or concern the structure of the indefinite DP, bear on the choice
between the weak and strong readings of indefinites, whereas the last two factors,
which are contextual, make a quantificational representation possible. Let us stress
that according to us, the quantificational interpretation of indefinites is never lexically
specified (compare Diesing (1992), who assumes that indefinites are ambiguous
between a quantificational and a weak interpretation) but depends on the context:

only indefinites in Topic positions, as well as indefinites in the subject position of
contrastive predicates can be assigned a quantificational representation. In sum, we
have proposed that certain indefinites, in particular singular indefinites and cardinal
indefinites, are ambiguous between a weak and a strong reading. Other indefinites
can be only weak or only strong depending on the lexical properties of their deter-
miners and/or their internal structure. The difference between non quantificational
strong and quantificational strong indefinites, on the other hand, is not a matter of
ambiguity but rather a matter of indeterminacy, the choice between the two options
being determined by the context.
Chapter 6
The Scope of Indefinites
In contexts where indefinite DPs co-occur with operators of various sorts

(quantificational DPs, intensional predicates, negation, quantificational adverbs),
we can observe certain interpretive effects, which have been traditionally analyzed
in terms of the relative ‘scope’ of the indefinite DP with respect to these operators.1
This type of analysis is based on the logical tradition, according to which indefinites
are existentially quantified DPs. In more recent theoretical frameworks, mainly
following Heim (1982) and Kamp (1981), indefinites are analyzed as individual
variables. Scope interactions with other operators are captured by postulating free
insertion of an existential quantifier $, through existential closure. The main prob-
lem with this approach is that it predicts more interpretations than the ones attested,
in other words, it over-generates.
The analysis we will adopt in this chapter is close to the one in Farkas (1997a, b,
2001) and Steedman (2003), who analyzed ‘scope’ effects in terms of dependency
relations. Accordingly, indefinites do not have ‘wide’ or ‘narrow scope’ but are
dependent or independent with respect to another operator.2
After briefly reviewing existing analyses (Sect. 6.1), we show that interpretive
effects usually attributed to scope in fact depend on the denotation of indefinite
DPs: weak indefinites take obligatory narrow scope, whereas intermediate and
wide scope are associated with individual-denoting indefinites (Sect. 6.2). Since
indefinites lack inherent quantificational force, they are not distributive and as such
do not induce any referential dependency. We show that distributive construals,
which are possible in certain contexts, are due to distributive predication rather than
1
In this chapter we will only examine the scope of a DP with respect to other DPs.
2
The scope ambiguities triggered by the interaction between tense or modality on the one hand and
DPs on the other hand could also be analyzed in terms of dependencies. But in these cases, the DPs
would be dependent on times and worlds.

172 6 The Scope of Indefinites
to quantification (Sect. 6.3). In Sect. 6.4, we propose an analysis of the dependent

readings of indefinites, analyzed as Skolem terms.
6.1 Scope: Current Analyses
6.1.1 Scope and Quantifier Raising
The two interpretations of the sentence in (1) are generally assumed to correspond
to two distinct representations:
(1) Someone insulted everyone.
(1¢) a. (Someone x) (everyone y) (x insulted y)
b. (Everyone y) (someone x) (x insulted y)
(1″) a. $x "y (human(x) Ù human(y)) [x insulted y]
b. "y $x (human(x) Ù human(y)) [x insulted y]
The representations in (1¢a–b) can be derived by the rule of Quantifier Raising
(May 1977, 1985), which attaches quantified DPs to the first dominating sentential
node; at this level, the relative order of the two quantified DPs is not constrained,
thus yielding the two possibilities in (1¢a–b). The representations in (1″) are the
result of replacing raised DPs with the corresponding logical quantifiers and adding
the restriction on the domain of quantification (here, someone and everyone range
over the set of humans).
The rule of Quantifier Raising is motivated by the fact that the rules of semantic
composition cannot interpret quantificational DPs in argument positions. The rule
of Quantifier Raising solves this problem by pulling the quantified DP out of the
argument position:
(2) I examined every student.
(2¢) every x (student(x)) [I examined x]
In this representation, the object position is occupied by an individual variable
and the determiner every denotes a relation between the property ‘be a student’
(in extensional terms, the set of x such that x is a student) denoted by the restriction
and the property ‘be examined by me’ (in extensional terms, the set of x such that I
examined x) denoted by the nuclear scope.
Unlike May (1977, 1985), we do not think that Quantifier Raising is a movement
rule that exhibits properties typical of syntactic movement, in particular wh-movement,
which occurs in interrogatives and relatives. Wh-movement is subject to locality
constraints, as well as constraints related to the target position, but it can avoid
locality violations by successive movement:
(3) a. Whoi do you think [ei [that Mary forgot [ei [to call ei] ]]?
b. Whoi does this student believe [ei [that Mary admires ei]]?
6.1 Scope: Current Analyses 173
Quantifier Raising, on the other hand, is a strictly local rule, which cannot go
beyond the limits of the minimal clause.3 The sentence in (3c) cannot have an
interpretation where every professor scopes over a student; the sentence is not an
assertion about several students but about a single student, the same for every
professor:
(3) c. A student believes that Mary admires every professor.
We may thus conclude that Quantifier Raising cannot be assimilated to other
instances of syntactic movement and therefore it is conceptually indistinguishable
from Montague’s rule of ‘Quantifying-in’: both implementations rely on a semantic
composition where quantified DPs are attached to the immediately dominating
clausal node, their interpretation being independent of their surface position. We
will use Quantifier Raising for ease of exposition, as this implementation provides
a straightforward way of deriving Logical Forms on the basis of corresponding
syntactic structures. However, the analyses developed here do not hinge on this
specific formalism.
6.1.2 Scope Ambiguity or Ambiguous Indefinites?
According to a widely accepted line of thinking (see in particular Farkas (1985) and
Corblin (1997)), going back to the logical tradition, the ambiguity between specific
and non-specific indefinites is a scope phenomenon:
(4) All the students saw a counselor.
(4¢) a. "y $x (student(x) Ù counselor(y)) [x saw y]
b. $x "y (student(x) Ù counselor(y)) [x saw y]
According to Farkas (1985) and Corblin (1987), the non-specific reading of the
indefinite in (4) is the result of its narrow scope with respect to the subject DP (see
(4¢a)), whereas its specific reading results from wide scope (see (4¢b)).
However, this account cannot be extended to the example in (5a), due to Fodor
and Sag (1982):
(5) a. Every professor heard the rumor that a student of mine failed the test.
The sentence in (5a) is ambiguous: either every professor heard that there is a
(possibly different from one professor to another) student who failed, or, there is a
specific student x such that every professor heard the rumor that x failed the test.
3
Farkas and Giannakidou (1996) challenged the strict locality of Quantifier Raising by pointing out
some exceptions:
(i) A/some student made sure that each/every invited speaker has a seat.
For further discussion of scope inversions see Sect. 6.2.2.
This second reading is unexpected under the assumption that indefinite DPs are
quantificational, since as already mentioned, QPs cannot scope out of the minimal
clause to which they belong, which implies that they cannot scope out of syntactic
islands as shown in (5b):
(5) b. Some professors heard the rumor that every student of mine failed the test.
The sentence in (5b) can only have a reading where every student has narrow
scope: a group of professors heard the rumor that every student of mine failed.
We must therefore conclude, following Fodor and Sag (1982), that the second
interpretation of (5a) cannot be analyzed as relying on the wide scope of the
indefinite DP. In sum, the ambiguity of (5a) cannot be analyzed as resulting from the
relative scope of two quantifiers. Fodor and Sag (1982) derive the two possible read-
ings from the ambiguity of indefinites, between a ‘referential’ and a ‘quantificational’
reading.4 Referential indefinites have no scope5 but can receive an apparent wide
scope interpretation in virtue of the fact that they refer to a specific individual.
With this in mind, let us now go back to example (4). If some wide-scope
indefinites are quantificational, the ambiguity of (4) could be derived in terms of
scope, since the indefinite is in the same clause as the quantified subject DP. However,
this analysis is not necessary, as we could just as well assume that the apparent wide
scope reading results from the referential interpretation of the indefinite.
6.1.3 Intermediate Scope or Referential Dependency?
Fodor and Sag’s proposal cannot account for the intermediate scope reading of
indefinites, a fact noted in Farkas (1981) and rediscovered by Abusch (1994) and
Corblin (1997). Thus, in examples such as (6a), the indefinite scopes out of the
minimal clause containing it, without, however, acquiring widest scope.
(6) a. Every professor rewarded every student who read a novel.
The sentence in (6a) has three possible readings. On the first one, the indefinite
takes maximal scope: there is a (single) novel, such that every professor
rewarded every student who read it. Another option is that the indefinite has nar-
row scope: every professor rewarded every student who read a novel (whatsoever),
the result being that there could be as many novels as students. On the third
4
Let us warn the reader that we do not endorse Fodor and Sag’s view that the narrow scope reading
of indefinites depends on their quantificational status. As made clear below, our proposal is that
narrow scope indefinites are either weak or dependent indefinites.
5
Having no scope means getting wide scope interpretation without giving rise to any kind of
dependency.
6.1 Scope: Current Analyses 175
possible reading, the indefinite takes intermediate scope: every professor chose a
specific novel and rewarded every student who read that novel; there can be at most
as many different novels as professors but crucially, fewer novels than students.
This last reading cannot be easily obtained but becomes more accessible in the pres-
ence of a certain type of relative (see Sect. 6.4 below):
(6) b. Every professor rewarded every student who read a novel he had
recommended.
This reading is unexpected under Fodor and Sag’s account: a referential indefinite
should only be able to take widest scope, or, to put it differently, its reference should
not be dependent on that of other quantifiers.
In view of these facts, Fodor and Sag’s proposal needs to be revised, in order to
account for both maximal scope and intermediate readings. Using the DRT frame-
work, Abusch (1994) analyzes the scope ambiguities of indefinites (represented as
variables) as a consequence of existential closure, which can attach to different
syntactic constituents.
Thus, the three readings of the sentence in (6a) would correspond to the different
positions occupied by the existential quantifier at Logical Form. This yields the
wide scope representation in (7a) (existential closure at text-level), the intermediate
scope reading in (7b) (existential closure in-between the two universal quantifiers)
and the narrow scope reading in (7c) (in situ existential closure). It is important to
note that the condition novel (z) associated with the indefinite DP a novel occupies
different positions in (7)a–c. This is due to the fact that in the case of wide and
intermediate scope interpretation, the condition supplied by the indefinite DP is
inserted in the position in which the existential quantifier is itself inserted. Although
this analysis is not stated in terms of movement, it has the same effect as a move-
ment analysis:
(7) a. $z [novel(z) Ù"x [professor(x) →"y [(student(y)
Ù read(y,z)) → rewarded(x,y)]]]
b. "x [professor(x) → $z [novel(z) Ù "y [(student(y) Ù read
(y,z)) → rewarded(x,y)]]]
c. "x [professor(x) → "y [student(y) → $z [novel(z)
Ù (read(y,z) → rewarded(x,y))]]]
Reinhart (1995) puts forth a similar proposal, the difference being that the
indefinite DP is not translated as a free variable but as a choice-function ‘f(novel)’.
In this framework, the three readings of the sentence in (6a) are associated with the
representations in (8a–c):
(8) a. $f "x [professor(x) → "y [(student(y) Ù read(y, f(novel)))
→ rewarded(x,y)]]
b. "x [professor(x) → $f "y [(student(y) Ù read(y, f(novel)))
→ rewarded(x,y)]]
c. "x [professor(x) → "y (student(y) → $f [read(y, f(novel))
→ rewarded(x,y)])]
The representations in (8) do not rely on movement: in all three formulas in

(8a–c), a novel stays in situ and occupies the position of the second argument of the
predicate read, thus yielding ‘read(y, f(novel))’. The absence of movement consti-
tutes an important advantage of the choice function account compared to the one
that makes use of free variables.
One may wonder whether we can implement Reinhart’s idea of leaving the
indefinite DP in situ without using choice functions. One might thus postulate LFs
representations such as (7¢a-c), where the existential quantifier is separated from
the descriptive condition, which occupies the surface position of the indefinite.
(7¢) a. $z "x professor(x) → "y [(student(y) Ù novel (z)
Ù read(y,z)) → rewarded(x,y)]]
b. "x [professor(x) → $ z "y [(student(y) Ù novel (z) Ù read (y,z))
→ rewarded(x,y)]]
c. "x [professor(x) → "y (student(y) → $ z [(novel(z)
Ù read(y,z)) → rewarded(x,y)])]
Representations of the type in (7¢c) are confronted with a well-known problem in
situations where there are no novels. Indeed, if there is no novel, ‘novel(z)’ is false
and any formula of the type ‘$ z [novel(z) → f]’ is true, thus rendering the sentence
in (6a) true. This is in contradiction with the speakers’ intuition, especially in the
case of the wide scope interpretation of (6a).
The analysis in terms of choice functions avoids this problem: if there is no
novel, there is no choice function that ranges over novels and consequently, formu-
las (8a–c) are false. Therefore, the analysis in terms of choice functions offers a
technical solution to a specific problem but it stays conceptually close to DRT-based
approaches. Both accounts rely on the assumption that indefinites are not
quantificational expressions but introduce free variables (individual and function
variables, respectively). The different possible readings result from free insertion of
existential closure operators. It should be observed that although there is no move-
ment of the indefinite DP, the position where the existential quantifier is inserted
still plays an important part, which means that these analyses adhere to the tradi-
tional view that the ambiguity of indefinites is a scope ambiguity.
A different line of investigation has been pursued in Farkas (1997a, b, 2001),
Kratzer (1998) and Steedman (2003), who restate the so-called “scope phenomena”
in terms of dependency relations. Under this view, what is crucial is whether a given
indefinite DP has a stable denotation, or alternatively, its reference is dependent on
some other expression in the sentence with which the indefinite co-varies.
In this chapter we will continue to talk about wide, intermediate or narrow scope
to describe the different interpretations, even if we will adopt an analysis inspired
by Steedman (2003), which is not formulated in terms of scope. The so-called wide
scope reading is now treated as involving indefinites that take constant reference,
whereas so-called intermediate and narrow scope readings rely on a dependent
indefinite.
6.2 Scope and Type of Denotation 177
6.2 Scope and Type of Denotation
In the previous chapters, we made a three-way distinction regarding the denotation of

indefinite DPs, which can be individual-referring, amount-referring or quantificational.
In two of these cases (individual-referring and amount-referring), QR is not motivated,
because the nominal expression does not contribute a quantifier. Indeed, those indefinites
that translate as individual variables have the semantic type that is required for argu-
ment positions and consequently, there is no type-mismatch that could force move-
ment. As for weak indefinites, their legitimation depends on the existential quantification
introduced by the predicate with which the indefinite combines.
Having argued that individual-denoting and amount-denoting indefinites are not
subject to QR, we can now turn to the third type of reading previously identified,
namely the ‘strong quantificational’ one and ask whether an analysis relying on QR
is adequate in this case, as is typically assumed in analyses that treat indefinite
determiners as ambiguous between a quantificational and a non-quantificational
reading (see Diesing 1992). In such a framework, an indefinite DP with a non-
quantificational determiner is not visible for QR, whereas an indefinite DP with a
quantificational determiner is subject to obligatory QR. In Sect. 6.2.2, we show that
this analysis is not appropriate: the inverse scope of indefinites is not derived by QR
and should instead be analyzed as a non-dependent reading, which is one of the pos-
sible interpretations of individual-denoting indefinites (type e). Distributivity can be
derived from plural predication.
6.2.1 On the Obligatory Narrow Scope of Weak Indefinites
6.2.1.1 Bare Nouns
The example in (9) shows that bare nouns like students of mine cannot take wide
scope.
(9) Every professor heard the rumor that students of mine failed the test.
This scope restriction is a consequence of the analysis of bare NPs proposed in
Chap. 2: they are necessarily weak indefinites (amount-referring expressions),
which as such must combine with an existential predicate, which prevents them
from being interpreted out of the position in which they appear. The obligatory
narrow scope of bare NPs is thus a direct consequence of their denotational type.
6.2.1.2 Modified Cardinals in Object Position
Consider now the sentence in (10), built with a modified cardinal (e.g., at least two,
maximum three, exactly two):
(10) Every professor heard the rumor that at least a student of mine failed the test.
The example in (10) shows that the scope of modified cardinals is clause-bound.
The DP at least a student of mine can acquire neither a wide scope interpretation
(this meaning, which is missing, is in (11a)) nor an intermediate scope reading, on
which it would outscope the intensional operator rumor, but would take scope below
the quantified DP every professor (cf. (11b)):
(11) a. There is at least a student of mine such that every professor heard the
rumor that he failed the test.
b. Every professor heard, about at least a student of mine, the rumor that he
failed the test.
One way to account for these observations is to assume, following Kamp and
Reyle (1993), that modified cardinals are quantificational and as such, their scope is
clause-bound.
However, Kamp and Reyle’s analysis cannot account for the scope properties of
a modified cardinal within its minimal clause. The facts illustrated in (12a–b) were
noticed by Ben-Shalom (1993), Beghelli (1995), Beghelli and Stowell (1997),
among many others:
(12) a. Every critic panned at least two books by H. Miller. (S > O, *O > S)6
b. A psychologist examined at least 30 students. (S > O, ??O > S)
c. A boy invited at least three girls. (S > O, *O > S)
These examples show that modified cardinals cannot take ‘inverse’ scope7: the
object, which is structurally below the subject in surface structure, cannot scope
over the subject. The lack of inverse scope of modified cardinals, illustrated in
(12a–c), is unexpected under the hypothesis that these indefinites behave like
quantificational DPs. If modified cardinals were quantified DPs, they would be
subject to Quantifier Raising, which would predict the wide scope reading, just like
in the case of every book in (12d):
(12) d. At least two critics panned every book by H. Miller.
(S > O, O > S)
In sum, the fact that modified cardinals in object position cannot take scope over
the subject DP argues against the quantificational analysis of modified cardinals
advocated by Kamp and Reyle.
6
S > O and O > S indicate wide scope of the subject over the object and wide scope of the object
over the subject, respectively. *O > S says that the object cannot scope over the subject.
7
Inverse scope corresponds to a situation where the order of interpretation of two DPs is inverted
with respect to their order in surface structure.
6.2 Scope and Type of Denotation 179
6.2.2 Inverse Scope and Individual-Type Denotation
Let us now consider the following examples, in which an indefinite in object

position can take scope over the subject (inverse scope):
(13) a. Every student saw a counselor. (S > O, O > S)
b. Every woman will consult two doctors. (S > O, O > S)
If indefinite DPs were assumed to be ambiguous, allowing either a quantificational
or a non-quantificational construal, their inverse scope could be analyzed as being
due to Quantifier Raising (see in particular, Diesing (1992)). But this account can-
not be integrated in the analysis we pursued so far: since their determiners are both
intersective and symmetric, indefinites are not inherently quantificational and as
such they are invisible for QR.
The hypothesis that the inverse scope of indefinites is not derived by QR is
supported by the observation, that the inverse scope reading of an indefinite
(see (13a–b)) is more easily accessible than the inverse scope reading of quantified
DPs ((14a–b))8:
(14) a. A professor examined every student. (? O > S)
b. A student read every book. (?? O > S)
If QR was needed in both (13a–b) and (14a–b), we would expect the wide scope
reading of the object to be equally available in the two groups of examples. The
observed difference between (13a–b) and (14a–b) suggests that the so-called ‘inverse
scope’ reading of the object in (13) is not derived by QR (nor “Quantifying-in”) but
depends on an individual-type denotation (type e), which correlates with
specificity.
This conclusion is confirmed by the fact that the inverse scope reading of plural
indefinites in object position does not correlate with a dependent reading of the
subject:
(15) a. Three experts will review each paper.
b. Three experts will review two papers.
8
Although the inverse scope reading can be hard to obtain, this interpretation is available both in
English and French. Note that when the sentence contains indefinite pronouns like quelqu’un
‘somebody’ or personne ‘nobody’, the ambiguity is salient, as made explicit in (i′) and (ii″)
(i) Quelqu’un relira chaque papier soumis.
Somebody will read each submitted paper.
(i¢) One and the same person will read each of the submitted papers.
(i″) Each submitted paper will be reviewed by somebody, but not necessarily by the same
person.
Focusing on inverse scope readings only, note that in (15a), with a quantified DP
in object position, there is a correlation between wide scope of the object over the
subject and the dependent reading of the indefinite subject. The interpretation of
three experts varies with the interpretation of every paper: each paper may be sub-
mitted to a different group of three experts.
On the other hand, the inverse scope reading in (15b) is clearly different from the
one just described for (15a). Under the most salient interpretation, there is a specific
set of two papers, for example the best two papers, which will be reviewed by a
specific group of three experts: in other words, the total number of experts is not
higher than three. This type of reading (known as the cumulative reading, see a. o.,
Scha (1981), Krifka (1992a), Kratzer (2007)) is typically associated with referential
expressions, especially definite DPs, as shown in (15c):
(15) c. Three experts will review the best two papers.
On the basis of the following examples, Ruys (1992) and Reinhart (1995) point
out that the same generalization holds for indefinites which scope outside their min-
imal clause: an inverse scope indefinite does not give rise to a dependent reading of
another indefinite:
(16) a. If three aunts of mine die, I will inherit a house.
b. Exactly half the critics heard the rumor that two books by Henry Miller
had been panned.
The sentence in (16a) can be paraphrased as in (17) but, crucially, cannot have
the reading in (18):
(17) There are three aunts of mine such that, when they are all dead, I inherit a
house.
(18) There are three aunts of mine such that, for each of them, if she dies, I inherit
a house.
Similarly, in (16b), we can talk about two specific books, in which case two
books by Henry Miller has inverse scope. However, the DP exactly half the critics
does not acquire dependent scope: the sentence in (16b) refers to a specific group of
critics, the same for both books. The sentence cannot make reference to two differ-
ent groups of critics, one for each book by Henry Miller, with each group containing
exactly half of the critics.
We conclude that the so-called ‘inverse scope’ reading of inde fi nites is dif-
ferent from the inverse scope of quantified DPs. Whereas the inverse scope read-
ing of quantified DPs is necessarily associated with the dependent interpretation
of some other DP, no dependency relation can be observed for the inverse scope
reading of an indefinite. This generalization can be reformulated in terms of dis-
tributivity: a quantificational DP is distributive and thus induces a referentially
dependent reading of a DP within its scope; in contrast to this, an indefinite DP is
not inherently distributive and therefore a ‘wide-scoped’ object indefinite does
6.3 The Distributivity of Indefinites: Quantification or Distributive Predication? 181
not trigger any dependency effect on another indefinite. In sum, the inverse scope
interpretation of indefinites is merely a non-dependent reading: the indefinite is
referential or specific.
6.3 The Distributivity of Indefinites: Quantification

or Distributive Predication?
Due to their inherent distributivity, quantificational DPs may induce dependent

readings on indefinites occurring within their scope. Because indefinite DPs are not
inherently distributive, we expect that the so–called ‘inverse scope’ of object
indefinites does not induce dependent readings on subject indefinites. The examples
discussed in the previous section confirm this prediction. There are, however,
indefinites which seem to contradict this generalization. In the following section,
we show that the distributivity of specific indefinites is the result of distributive
predication. In Sect. 6.3.2, we show that the subject position can give rise to dis-
tributivity effects for indefinites, even in cases where they are non-specific. This is
expected under the analysis put forth in Chap. 5, according to which indefinite DPs
in subject positions can be represented as quantificational.
6.3.1 The Distributivity of Inverse Scope Specific Indefinites
The distributive reading of an indefinite in object position is more easily accessible

if the indefinite is interpreted as specific (cf. Ben-Shalom 1993; Dobrovie-Sorin
1995; Krifka 1992a). A clear set of empirical data arguing in favor of this general-
ization is provided by Romanian examples, such as (19)a–b, in which a preposi-
tional accusative co-occurs with an indefinite subject marked by câte, which triggers
a “distributed” reading.9
(19) a. ? Câte un băiat le-a invitat pe trei fete10.
cate a boy them-has invited pe three girls
‘A boy invited three girls’
9
The type of interpretation induced by câte is easier to grasp in the example in which câte attaches
to the object DP.
(i) Fetele vor citi câte o poezie.
the girls will read cate a poem
‘The girls will each read a poem’
10
In contemporary Romanian, object clitic-doubling is obligatory with objects marked by preposi-
tional accusative.
b. Pe trei fete le-a invitat câte un băiat.

pe three girls them has invited cate a boy
The difference in acceptability between (19)a–b and (20)a–b shows that the
prepositional accusative – which induces specificity (see Chap. 5, Sect. 5.1.4) facili-
tates the distributive reading of the object:
(20) a. *Câte un băiat a invitat cel puţin trei fete.
cate a boy has invited at least three girls
A boy invited at least three girls’
b. ?? Câte un băiat a invitat trei fete.
cate a boy has invited three girls
The ungrammaticality of the example in (20a) comes from the co-occurrence of
two diverging markers, the presence of câte on the subject and modified cardinals in
the object position. The presence of câte on the subject requires distributivity of the
object indefinite, which can only arise if the indefinite object is specific (see below).
But modified cardinals cannot have a specific interpretation (see Chap. 5, Sect.
5.1.3). This is illustrated by example (21).
(21) a. Ion i-a examinat pe doi elevi.
John them-has examined pe two students
‘John examined two students’
b. ?? Ion i-a examinat pe cel puţin doi elevi.
John them-has examined pe at least two students
‘John examined at least two students.’
The example in (20b) is slightly more acceptable than (20a) because a non-
modified cardinal can acquire a specific interpretation (cf (21a) vs. b). If the
specificity is overtly marked by the prepositional accusative, as in (19a), the accept-
ability improves. And finally, the example is perfectly grammatical when the rela-
tive order of the two indefinite DPs at S-structure matches their order of interpretation
(see the contrast between (19a) and (19b)).
It is well-known that distributive readings may come from two different sources,
namely quantification and plural predication, i.e., predication applying to a plural-
ity. Since indefinites are not inherently quantificational, a quantificational reading of
indefinite objects cannot be assumed. We may therefore analyze the distributive
readings of specific indefinites as resulting from plural predication, rather than from
a quantificational structure: (19b) is true if and only if the individual denoted by trei
fete ‘three girls’ belongs to a predicate obtained by the pluralization of the predicate
obtained by lambda-abstraction over the position it occupies; the pluralization
(notated by the so-called ‘star operator’) of the predicate is triggered by the fact that
6.3 The Distributivity of Indefinites: Quantification or Distributive Predication? 183
the argument of the predication, i.e., three girls, is plural.11 In other words, (19b) is
verified if and only if:
(22) [[three girls]] Î *invited-by-a-boy
(22) says that [[three girls]] belongs to the set of individuals and sums of indi-
viduals invited by a boy. Due to the presence of the preposition pe in (19b), the DP
three girls denotes a specific group of three girls, which can be represented as
g1 + g2 + g3. (22) is true if g1 + g2 + g3 belongs to the denotation of the pluralized predi-
cate, i.e., if g1 + g2 + g3 is one of the groups in the set denoted by the predicate. The
distributive interpretation is due to the generalization in (23):
(23) If a sum of individuals satisfies a pluralized predicate12 then each individual
in that sum satisfies the corresponding singular predicate (see Link 1983 and
Landman 1989a, b).13
Accordingly, the sentence is verified if each girl was individually invited by a
(potentially different) boy.
6.3.2 The Quantificational Status of Indefinites

in Subject Position
So far, we have argued that the distributive reading of object indefinites cannot be
due to a quantificational structure but rather to distributive predication, which
depends on specificity. In what follows, we show that subject indefinites can be
distributive even when non-specific. In view of these facts, we are led to propose
that indefinites can acquire a distributive interpretation due to their syntactic
position.
In order to illustrate this point let us consider again the behavior of modified
cardinals. We have already shown that (i) modified cardinals cannot take a type e
denotation (a restriction that applies to both object and subject positions, see the
discussion in Sect. 5.1.3 in Chap. 5) and (ii) they should be analyzed as amount-
referring in the object position (see the discussion of examples (10)–(12) in
11
See Landman (1989a, b), who extends Link’s (1983) analysis of plural nominal predication to
plural verbal predication. Generally speaking, a predicate that denotes a set of atoms is pluralized
if its subject DP is plural.
12
Pluralized predicates should be distinguished from collective predicates of the type meet.
13
This generalization follows from the definition of the pluralization operation itself: since the
predicate is derived by pluralization from the corresponding singular predicate, each of the indi-
viduals in any of the groups in the denotation of the pluralized predicate necessarily satisfies the
corresponding singular predicate.
Sect. 6.2.1.2). Note now that examples such as (24) indicate that modified cardinals
in the subject position favor a distributive reading:
(24) a. At least two students offered me a gift. (S > O)
b. Exactly two students offered me a gift. (S > O)
c. At least three boys invited a girl. (S > O)
Modified cardinals can thus acquire clearly different interpretations according
to whether they occupy an object or a subject position: in the object position, they
can only be amount-referring, whereas in the subject position, they can have a
distributive reading, which can be analyzed as being due to a quantificational
representation.
We may thus conclude that indefinites are quantificational only when they occur
in the preverbal subject position. A natural assumption to make is that intersective
and symmetric determiners can function as quantificational determiners only if the
asymmetry that is characteristic of quantificational structures is a property of the
syntactic context in which they occur. We may thus suggest that an intersective deter-
miner can function as a quantifier only if the partition between restriction and scope
corresponds to overt syntax (preverbal subjects are mapped onto the restriction)
or to information structure (Themes are mapped onto the restriction).
6.4 Referential Dependencies and Skolem Functions
In the previous sections, we argued that the scope effects exhibited by indefinites
should in fact be analyzed in terms of dependency relation. Let us go back to the
different readings of a sentence like (6a), repeated below as (25). According to our
analysis, the wide scope interpretation paraphrased in (26a) corresponds to a non-
dependent reading of the indefinite, whereas the intermediate (26b) and narrow
scope (26c) readings are the result of dependency relations of the indefinite DP with
respect to other DPs in the sentence:
(26) a. Il existe un roman et un seul, tel que chaque professeur a récompensé
chaque étudiant qui l’a lu.
‘There is a unique novel, such that every professor rewarded every
student who read it’
b. Chaque professeur a choisi un roman particulier et a récompensé tous les
étudiants qui l’ont lu.
‘Every professor chose a specific novel and rewarded every student who
read that novel’
c. Chaque professeur a récompensé chaque étudiant qui a lu un roman
quelconque.
‘Every professor rewarded every student who read some novel.’
6.4 Referential Dependencies and Skolem Functions 185
In (26b), the indefinite DP a novel is dependent on the quantified DP every

professor: there are at most as many different novels as (there are) professors. In
(26c), the indefinite DP a novel is dependent on the quantified DPs every professor
and every student: there can be as many different novels as there are (professor,
student) pairs. Let us illustrate these various readings with an example: assume a
world in which there are three professors and two students. The same student can
attend several courses and thus have several professors. In the case of the so-called
intermediate scope reading in (26b), each professor picks some novel and rewards
all the students who read that novel. Since there are three professors, there can be at
most three novels. In the narrow scope reading, each student can read some novel,
so for any given course, each student can read a different novel: there can be at most
two different novels in each course; since there are three professors and thus three
courses, the number of novels is multiplied and can be six (three times two).
In order to represent this dependency, we can use Skolem functions.14 A Skolem
function is a function that maps an element x onto an element y, such that y = f(x).
By writing f(x) instead of y, we make it clear that y depends on x. Hintikka (1974,
1986), Groenendijk and Stokhof (1984) and Engdahl (1986) were among the first to
use Skolem functions to account for the type of data illustrated in (27):
(27) Every man hates a (certain) woman, his mother-in-law.
In (27), the indefinite DP a (certain) woman does not refer to a unique individual,
the same woman for every man, but instead it varies depending on the man. This
dependency relation can be indicated by translating a certain woman by the func-
tional term f(x), where x ranges over men; mother-in-law can be taken as the name
of the function f that relates every man x to the woman y that he hates. The sentence
in (27) can then be represented by the tripartite formula in (27¢). In (27¢), every man
translates as the universal quantifier "x followed by the restrictive term man(x). The
whole of the sentence with x in the position of the quantified DP every man and f(x)
translating a woman constitutes the nuclear scope (indicated by square brackets) of
the quantifier.
(27¢) $f "x (man(x)) [hate(x,f(x)) Ù woman(f(x)) Ù mother-in-law(f(x),x)]
14
Skolem functions were initially used in logic in order to eliminate existential quantifiers in the
scope of a universal quantifier, the existential quantifier that binds the Skolem function itself
always takes largest scope (over the universal quantifier). Thus, the formulas in (i) and in (ii) are
truth-conditionally equivalent:
(i) "x $y (f(x,y))
(ii) $f "x (f(x,f(x)))
In (i), y depends on x because the existential quantifier $ binding y is within the scope of the uni-
versal quantifier " that binds x. In (ii), this dependency is expressed in functional terms: y is
replaced by f(x).
The tripartite representation in (27¢) is equivalent to the logical formula in (27″):

(27″) $f ["x [man(x) → (hate(x,f(x)) Ù woman(f(x))] Ù mother-in-law(f(x),x)]
The first conjunct ‘∀x [man(x) → (hate(x,f(x)) Ù woman(f(x))]’ corresponds to
‘every man hates a woman’; the second conjunct ‘mother-in-law (f(x),x)’ indicates
the name of f. The woman in question is every man’s mother-in-law.
In order to replace a variable y with a functional term, marking the dependency
of y with respect to x (i.e. y = f(x)), the function f has to be properly defined, namely
to associate to every x a single element y, such that y = f(x). In (27), these conditions
are met: mother-in-law is a function whose domain is the set of married men (the
overtly expressed restrictor man can be easily accommodated to refer to married
men rather than to the overall domain of men), its range is the set of women, this
function associates with each married man a unique woman, his mother-in-law. The
degree of acceptability of a sentence decreases when these requirements are not
observed or hard to accommodate.
(28) a. ? Every man hates someone. His elder brother.
b. ?? Every man hates an animal. His snake.
In (28a), we have to accommodate the domain of f by restricting it to men who
have an elder brother. The decreased acceptability of (28b) can be attributed to the
fact that its interpretation requires two accommodations: not only do we have to
restrict the domain to men owning snakes but we must further restrict it to men who
own only one snake. This second readjustment is needed because a function can be
defined if and only if each element in the domain is mapped to one and only one
element of the co-domain. Therefore, in (28b) his snake can be represented as a
Skolem function f(x) if and only if the domain is restricted to men owning one and
only one snake. If a man owns more than one snake, it is impossible to define a func-
tion f such that f(x) is ‘the snake of x’.
Going back to example (25), repeated below, we can make use of Skolem func-
tions to account for intermediate scope (26b) and narrow scope (26c) readings, rep-
resented by the tripartite configurations in (29) and (30):
(29) $f"x"z (professor(x) Ù student(z) Ù novel(f(x)) Ù read(z,f(x)))
[rewarded(x,z)]
(30) $f"x"z (professor(x) Ù student(z) Ù novel(f(x,z)) Ù read(z, f(x,z)))
[rewarded(x,z)]
In these representations, the two quantified DPs (every professor and every student
who read a novel) are mapped onto the two first parts, the quantificational part ("x"z)
and the restriction (professor(x) Ù student(z) Ù novel(f(x)) Ù read(z, f(x))) in (29) or
(professor(x) Ù student(z) Ù novel(f(x,z)) Ù read(z, f(x,z))) in (30). The nuclear scope,
[rewarded(x,z)], corresponds to the whole of the sentence, in which the two quantified
DPs have been respectively replaced by the variables that they bind, x and z.
In (29), which corresponds to the “intermediate scope” reading, the procedure of
skolemization applies once, f(x) refers to the novel recommended by the professor
6.4 Referential Dependencies and Skolem Functions 187
x to all of his students. The choice of novels varies with professors. In (30), which
translates the “narrow scope” reading, the procedure of skolemization applies twice.
The f function depends both on x and on z and attributes a novel to pairs of elements
(professor, student).
The tripartite representations in (29) and (30) are equivalent to the logical formu-
las in (29¢) and (30¢), respectively:
(29¢) $f"x"z [(professor(x) Ù student(z) Ù novel(f(x)) Ù
read(z,f(x))) → rewarded(x,z)]
(30¢) $f"x"z [(professor(x) Ù student(z) Ù novel(f(x,z)) Ù read(z,
f(x,z))) → rewarded(x,z)]
These two formulas are, however, problematic, because there is no constraint on
the existentially-quantified function, meaning that the formulas in (29¢) and (30¢)
are always true: any choice of function that falsifies the antecedents in (29¢) and
(30¢) makes the formulas true. Thus, if we choose a function that does not associate
a novel to f(x), then the condition novel(f(x)) is false and the result is always true.
The problem was noticed by Winter (1997) and Dekker (2004), among others. This
problem can be solved by introducing constraints forcing the function to be an
appropriate Skolem function, by specifying its domain and range.15 But we will
choose an alternative solution by adopting Steedman’s (2003) implementation in
terms of Skolem terms. Using Skolem terms, we can associate the LFs in (29″) and
(30″) with the “intermediate scope” and the “narrow scope” readings of (25):
(29″) "x"z (professor(x) Ù student(z) Ù novel(f(x)) Ù
read(z,f(x))) → rewarded(x,z)
where f is a function from the set of professors to the set of novels
(30″) "x"z (professor(x) Ù student(z) Ù novel(f(x,z)) Ù read(z,
f(x,z))) → rewarded(x,z)
where f is a function from the set of (professor, student) pairs to the set of
novels
Our account is similar to the one advocated in Steedman (2003): the dependent
indefinite DP is not represented as an existentially-quantified function variable but
rather as a Skolem term, i.e., the value of a Skolem function f; f is the name of a
function; f does not vary, it is only the value of f(x) or f(x,z) that varies. The
specification of the range and the domain of f are not indicated inside the formulas
but are stated as a constraint on f.
The constraints on the Skolem function that hold for the intermediate and narrow
scope readings are given by the linguistic form of the sentence. The domain of f is
the extension of the noun generating the dependency (professor in the “intermediate
15
In the following formula, which refines (29¢), the range and the domain of the function are
specified in the first conjunct:
(i) $f ["x"y [(professor(x) Ù f(x) = y) → novel(y))] Ù "x"z [(professor(x) Ù student(z)
read(z,f(x))) → rewarded(x,z)])
reading” of (25) for example), the range of f corresponds to the extension of the
dependent noun (novel in both readings of (25)) and the function itself can also be
explicit, as in the case of mother-in-law in (27).
As observed in Kratzer (1995), intermediate scope readings can be hard to obtain
but become more accessible when the dependent noun is modified by a relative that
contains a pronoun that is anaphorically related to the (dependency-inducing)
quantified expression. Consider the following examples, adapted from Kratzer
(1995):
(31) a. Every professor rewarded every student who read a/some book I had
recommended.
b. Every professor rewarded every student who read a/some book he had
recommended.
c. Every professor rewarded every student who read a/some book he had
recommended to him.
In (31a), a/some book that I had recommended can acquire a wide scope inter-
pretation, where it denotes a specific book. Another possibility is the narrow scope
reading, according to which the indefinite DP depends both on professors and
students. The intermediate scope reading, where books vary with professors but not
with students is possible but rather difficult. In (31b), the occurrence of the pronoun
he (instead of I), which gets bound by the quantified expression every professor,
favors the intermediate reading, in which a/some book is a dependent indefinite.
Due to the presence of he, the value of the f function is almost explicit in this sen-
tence: f is a function that has in its domain the set of professors, in its range the set
of books and that associates every professor to the book he recommended. If a pro-
fessor recommended several books, one book is chosen from the set of possible
ones. In (31c), the presence of two bound variable pronouns (he and to him) rules
out both the wide scope and the intermediate scope readings: the indefinite DP is in
the dependency of both quantified expressions, the books vary both with professors
and with students.
Turning finally to the wide scope reading, it need not be represented in terms of
Skolem functions, given that the so-called ‘wide scope’ indefinites are not depen-
dent on any other quantified DP but only on the situation. These indefinites can be
represented as referential terms, as in (32a) or (32b):
(32) a. "x"y ((professor(x) Ù student(y) Ù read(y,n) Ù
novel(n)) → rewarded(x,y))
b. "x"y ((professor(x) Ù student(y) Ù read(y,f(novel))) → rewarded(x,y))
where f is a function defined over the domain of novels and distinguishes
one novel among all novels.
In (32a) the referential indefinite a novel is represented by a constant and in
(32b), which is equivalent to (32a), it is represented by f(novel), where f is the name
of a choice function applied to the domain of novels. Choice functions can be viewed
as a special case of Skolem functions, where there is no dependency on other
quantified DPs.
6.5 Conclusions 189
6.5 Conclusions
Summarizing, we have shown that scope effects can be analyzed as dependency

relations. Among the different semantic types of indefinites identified in previous
chapters, only referential indefinites, i.e., indefinites of type e, can be dependent.
Dependent indefinites can be represented as Skolem terms, the reference of which
co-varies with the quantificational DPs on which they depend.
When an indefinite DP of type e is not dependent, it is specific and functions as
a constant. These indefinites are improperly called wide scope indefinites.
Chapter 7
Genericity, (In)Definiteness and Bare Nouns
In line with most theoreticians of genericity, we assume that there are two types of
genericity: the so-called ‘nominal’ genericity associated with DPs denoting kinds and
the so-called ‘sentential’ genericity associated with generic quantification. We will also
follow the current view that the generic reading of indefinites is not related to kind
denotation1 but instead depends on sentential genericity: the indefinite contributes an
individual variable that gets bound by an overt adverb of quantification or by a covert
generic operator GEN (which can be assimilated to an adverb of quantification).
Current theories differ, however, regarding the analysis of adverbs of
quantification, which leads to different analyses of generic indefinites and bare plu-
rals. According to Lewis (1975), Kamp (1981) and Heim (1982), adverbs of
quantification are unselective: they bind all free variables in their domain. This
“unselective binding” hypothesis is confronted with several empirical and concep-
tual problems, which led theorists to gradually abandon it. The two competing
hypotheses still under debate are:
(i) In all of their contexts, Q-adverbs quantify over events or situations (Rooth
1985, 1995; Schubert and Pelletier 1987, 1988; de Swart 1991, 1996; Krifka
et al. 1995; Krifka 1995, etc.).2 The configurations in which adverbs of
quantification seem to bind one or more individual variables (or a tuple of indi-
viduals) would constitute a side effect of quantification over events: quantifying
over events induces quantifying over the tuple of participants in the event.
According to this analysis of adverbs of quantification, a generically-interpreted
indefinite (“generic indefinite”, henceforth) is always indirectly bound by an
adverb quantifying over events.
1
In this chapter we set aside the taxonomic readings of indefinites, which do not raise any particu-
lar problem: with kind predicates, or in contexts that trigger a kind interpretation, indefinites take
a taxonomic reading (e.g., Two tigers are extinct, Some musk rats arrived in Europe during the
sixteenth century).
2
Although event-based and situation-based approaches are technically different, there are many
cases in which they have the same empirical coverage. Here, we will use an event-based notation.

192 7 Genericity, (In)Definiteness and Bare Nouns
(ii) Depending on the context, Q-adverbs quantify either over events or over
individuals (Chierchia 1998; Kratzer 1998).
In this chapter, we will assume the second hypothesis. The distinction between
these two types of quantification correlates with a distinction between two types of
indefinites occurring in generic contexts: those indefinites that are bound by the
Q-adverb have a genuine generic reading, whereas those indefinites that occur in the
context of a Q-adverb that quantifies over events are ‘pseudo-generic’. These dis-
tinctions will prove crucial for the analysis of habitual sentences (cf. Sect. 7.2) and
of the constraints on the generic readings of singular (cf. Sect. 7.3) and plural (cf.
Sect. 7.4) indefinites. We will show that those plural indefinites (e.g., French des-
indefinites) that are directly bound by an adverb of quantification can only express
generalizations over groups of individuals. The analysis of plural indefinites sheds
light on the analysis of bare plurals and constitutes an argument in favor of the
hypothesis that, on their generic reading, bare plurals should not be analyzed as
indefinite DPs but rather as kind-referring DPs (cf. Carlson 1977a, c).
7.1 Generic Indefinites: Quantification over Events

and over Individuals
7.1.1 Quantification over Events and Indirect Binding

of Indefinites
Consider example (1), which may be represented as in (1¢), where the adverb
toujours ‘always’ has been replaced by the universal quantifier:
(1) When John invites a friend, he always cooks dinner for her.
(1¢) "e,x (invite (e, John, x) Ù friend (x)) [$e’ (overlap (e, e’) Ù cook-dinner
(e’, John, x))]
Representations of the type in (1¢), where different variables corresponding to
different types (i.e., e for events and x for individuals) are bound by the same
quantifier, are used by different theorists to mean different things. Lewis (1975), for
example, uses this kind of notation to represent unselective binding (the adverbs of
quantification bind all free variables in their scope), whereas Chierchia (1995c) uses
it to represent multiple binding.3 Other authors assume that in if / when-clauses, the
adverb quantifies only over events, the individual variables being indirectly bound
3
Chierchia (1995c: 101–102) examines the asymmetric readings of sentences such as When a
painter lives in a village, it is usually pretty. This sentence is ambiguous between a generalization
over painters living in a village and a generalization over villages in which a painter lives. Chierchia
suggests that the difference between the two readings can be explained if we assume that a
quantifier may bind an event as well as an individual variable. For an alternative solution to the
problems that asymmetric readings of if / when-clauses pose, see Chap. 8.
7.1 Generic Indefinites: Quantification over Events and over Individuals 193
(Rooth 1985, 1995; Schubert and Pelletier 1987; 1988; de Swart 1991, 1996; Krifka
et al. 1995; Krifka 1995, etc.). For this third type of analysis, the notation in (1¢) is
not adequate. We may instead assume a representation of the type in (1″) inspired
by Steedman (2003), in which the indefinite a friend translates as an event-dependent
Skolem term (see Chaps. 6 and 8) notated f(e),4 where f(e) is interpreted as
Theme(e):
(1″) "e (invite (e, John, f(e)) Ù friend (f(e))) [$e’ ((overlap (e, e’) Ù cook-dinner
(e’, John, f(e)))]
The adverb of quantification binds the event variable, which ranges over events
in which John invites someone. The indefinite in (1″) is represented by means of a
Skolem term f(e) denoting an individual (type e) that is the Theme of the invitation
and whose referent varies depending on the inviting events.
7.1.2 Adverbial Quantification over Individuals
Characterizing sentences of the type in (2), built with individual-level predicates,

express generalizations that hold of classes of individuals:
(2) A dog is usually intelligent.
(2¢) GENx (dog (x)) [intelligent (x)]
Representations such as (2¢) assume the analysis of adverbs of quantification
proposed by Lewis (1975), Kamp (1981) and Heim (1982), according to which
(overt or covert) adverbs of quantification are unselective: they bind any kind of free
variable occurring in their scope and since i-level predicates such as intelligent sup-
ply an individual variable and no time variable, the adverb of quantification binds an
individual variable.
Other authors adopt a more constrained view, according to which adverbs of
quantification can quantify only over events or situations (Rooth 1985, 1995;
Schubert and Pelletier 1987, 1988; de Swart 1991). Note however that some of
these theorists were led to relax this assumption: de Swart and Farkas (2005) and
Farkas and de Swart (2007) propose that in examples such as (2), which express
generalizations over individuals, the Q-adverb binds both an event-variable and
the variable supplied by the indefinite, whereas in examples that express quantifica-
tion over events (see example (1) above), the Q-adverb binds only an event-variable.
Such a differentiated analysis is an implicit acknowledgment of the fact that a
4
The cooking event may itself be represented as a Skolem term depending on inviting events,
which allows us to dispense with the existential closure of the variable over cooking events: (e)
(i) "e (invite (e, John, f(e)) Ù friend (f(e))) [cook-dinner (g(e), John, f(e))].
uniform analysis of Q-adverbs as exclusively quantifying over events cannot be

maintained.5
In this chapter we will assume without further discussion that Q-adverbs can
quantify not only over events but also over individuals (Lewis 1975; Kamp
1981; Heim 1982; Kratzer (1988, 1995); Diesing 1992; Chierchia1995a; Krifka
et al. 1995). This does not, however, mean that Q-adverbs are unselective
quantifiers: in one and the same context they can quantify either over events or
over individuals; in other words, Q-adverbs can only bind one variable in a
given context.
7.1.3 Syntax-Semantics Mapping Rules
If we allow adverbs of quantification to quantify not only over events but also over
individuals, the question that arises is how to choose between these two possibili-
ties: when does an adverb quantify over events and when does it quantify over
individuals? We will assume that overt syntax (including information structure) is
mapped onto LF representations by rules such as (3)–(5):
(3) If /when-clauses go to the restriction of adverbs of quantification (Lewis 1975;
Heim 1982, a.o.).
(4) Subjects of generic predicates6 go to the restriction of adverbs of
quantification.7
(5) A sentence containing an adverb of quantification goes to the nuclear scope
and the restriction is filled by its “focus closure”, obtained by replacing the
focused constituent with a variable (Rooth 1985, 1995).
The choice among these different rules depends on the lexical properties of the
main predicate, on the syntactic configuration and on information structure. Quite
clearly, the rule in (3) can apply in (1) but not in (2), which does not contain a
5
Empirical arguments against representing characterizing sentences built with i-level predicates as
relying on quantification over events can be found in Dobrovie-Sorin (2003).
6
This label groups together individual-level predicates such as intelligent, handsome, blond, etc.
(Carlson 1977a, c; Kratzer 1988, 1995) and stage-level predicates on their habitual or dispositional
use (e.g., John reads novels, Paul smokes, Mary drinks wine). Leaving aside their differences, these
two types of predicates are alike in that their arguments are individuals rather than stages of
individuals.
7
This principle is reminiscent of Diesing’s (1992) hypothesis that DPs sitting in Spec IP (the pre-
verbal subject position) go to the restriction of adverbs of quantification. Yet, the mapping rule (4)
differs from Diesing’s principle in that it is not stated in terms of a particular syntactic position; the
notion of ‘subject of a generic predicate’, used in (4), depends on (i) a lexical distinction between
classes of predicates and (ii) information structure.
7.1 Generic Indefinites: Quantification over Events and over Individuals 195
subordinate if / when-clause. The rule in (4), on the other hand, cannot apply in
(1) but can apply in (2),8 yielding the representation in (2¢), where the adverb
quantifies over individuals.
Rule (5) is needed in order to analyze examples such as (6), in which the capital
letters indicate the focused constituents9:
(6) a. In St. Petersburg OFFICERS always escorted ballerinas.
b. In St. Petersburg officers always escorted BALLERINAS.
(6¢) a. "e,x,y (escort (e,x,y) Ù ballerinas (y))
[escort (e,x,y) Ù ballerinas (y) Ù officers (x)]
b. "e,x,y (escort (e,x,y) Ù officers (x))
[escort (e,x,y) Ù ballerinas (y)) Ù officers (x)]
These two representations capture the different interpretations associated with
the two sentences, which express different generalizations according to the place-
ment of focus: (6a) says that every time ballerinas were escorted, they were escorted
by officers, whereas (6b) says that every time officers escorted someone, they
escorted ballerinas.
In sum, quantification over events obtains due to an application of rule (3) or rule
(5), whereas quantification over individuals obtains due to an application of rule (4).
7.1.4 Two Types of Generic Readings for Indefinites
Given the difference between quantification over events and quantification over
individuals, it is possible to distinguish between two types of generic readings for
indefinites:
(7) a. An indefinite DP takes a ‘truly’ generic reading if and only if it is directly
bound by an adverb of quantification (in other words, the adverb quantifies
over the set of individuals supplied by the indefinite NP).
b. An indefinite DP takes a ‘pseudo-generic’ reading if and only if it is
indirectly bound by an adverb of quantification that quantifies over events.
The LFs given in (1″) and (2¢), repeated here with the corresponding examples,
allow us to distinguish between pseudo-generic readings and truly generic readings
of indefinites:
(1) When John invites a friend, he always cooks dinner for her.
(1″) "e (invite (e, John, f(e)) Ù friend (f(e))) [$e’ (overlap (e, e’) Ù cook-dinner
(e’, John, f(e)))]
8
This does not exclude the possibility of finding examples where the syntax allows for two differ-
ent mapping rules, which would give rise to an ambiguity.
9
These examples are due to Rooth (1985, 1996:272).
(2) A dog is usually intelligent.

(2¢) GENx (dog (x)) [intelligent (x)]
In the first case the indefinite translates as an event-dependent Skolem term and
in the second case as an individual variable bound by a generic operator. The Skolem
term f(e) is a function that applies to an event and returns an individual participating
to the event (see Parsons’ 1990 neo-Davidsonian theory). Since the value of f(e)
varies with the value of the event itself, the result is a dependent reading (of a friend
as well as of the anaphoric pronoun her).
In sum, we assume that Q-adverbs quantify, depending on the context, either
over events or over individuals. Consequently, the indefinites that occur in
generic contexts can be analyzed either as Skolem terms depending on an event
variable or as individual variables bound by a Q-adverb. In what follows, we
will show that the distinction between the two types of adverbial quantification
and the correlative difference between the two types of indefinites occurring in
generic contexts is crucial in explaining the constraints on the generic readings
of indefinites.
7.2 Characterizing Sentences with Habitual Predicates
Let us now examine examples of the type in (8):

(8) A bird flies.
It is commonly accepted that the representation of sentences such as (8) involves
a default adverb of quantification GEN, which is the ‘silent’ counterpart of overt
adverbs such as usually, habitually, etc. However, there is no consensus concerning
the domain of quantification of this GEN operator. The two analyses that can be
found in the current literature are given in (8¢a) and (8¢b):
(8¢) a. GENe,x (bird (x) Ù C(e,x)) [fly (e,x)]
b. GENx (bird (x)) [fly (x)]
According to the LF in (8¢a), GEN quantifies over event-individual pairs (see
Chierchia (1995a, b), Schubert and Pelletier (1988), Krifka et al. (1995), among
many others). The corresponding interpretation can be paraphrased as follows:
“Take any bird and any situation in any world maximally similar to ours where
the felicity conditions for flying (such as, e.g., presence of the right triggers) are
satisfied […]. Any bird will fly in such a situation” (Chierchia 1995b:196). C(e,x)
is a condition that restricts the domain of quantification to relevant situations; C
is a contextually-specified constant. In the case under consideration, only birds
that are in a situation in which they are likely to fly are taken into account. This
allows us to exclude other situations, like for instance, situations in which the
birds are sleeping.
7.2 Characterizing Sentences with Habitual Predicates 197
On the analysis adopted here, this kind of representation can be rewritten using
quantification over events as in (8¢c), where the indefinite is represented as a
Skolem term:
(8¢) c. GENe (bird (f(e)) Ù C(e,f(e))) [fly (e,f(e))]
As for the LF given in (8¢b), it relies on the hypothesis that GEN quantifies over
an individual variable supplied by the indefinite (Heim 1982; Diesing 1992).
We will show that LFs (8¢a–c) are problematic and that examples of the type in
(8) rely on generic quantification over individuals combined with a habitual predi-
cate. Quantification over individuals is allowed due to the fact that the s-level predi-
cate is turned into an i-level predicate via a HAB operator.
7.2.1 Adverbial Quantification over Events
The representations in (8¢a) and (8¢c) cannot explain why sentences such as those in
(9) cannot have a generic reading:
(9) a. A dog is tired.
b. A cow is infected.
c. A police officer is available.
These examples are similar to (8) insofar as they are built with an indefinite in the
subject position of an s-level predicate and therefore they should in principle be
analyzed in a similar way, i.e., as expressing generalizations over episodic events:
(9¢) a. GENe (dog (f(e)) Ù C(e,f(e))) [tired (e,f(e))]
Since the LFs in (8¢c) and (9¢a) are built in the same way, they are equally well-
formed. The unacceptability of the generic reading for sentences like (9) is therefore
surprising.
Another problem arises with ambiguous examples such as (10) below, which can
be paraphrased as (10a) and (10b), respectively:
(10) A student rarely smokes.
a. Few students smoke.
b. Generally, a student smokes rarely.
In (10a), the adverb of quantification rarely (translated as few in the gloss (10a))
quantifies over individuals, whereas in (10b), the same adverb is analyzed as a fre-
quency adverb, translated as rarely in (10b). This ambiguity is unexpected on the
analysis given above, on which (10) would have only one representation, that in
(10¢):
(10¢) FEWe (student (f(e)) Ù C(e,f(e))) [smoke (e,f(e))]
The interpretation associated with the LF in (10¢) can be paraphrased as follows:

few events in which smoking by students might take place are events in which
smoking by students really takes place. This analysis is problematic, since it does
not distinguish between the interpretations in (10a) and (10b).
7.2.2 Adverbial Quantification over Individuals
According to another widespread view, (8) can be analyzed as involving quantification

over individuals:
(8) A bird flies.
(8¢) b. GENx (bird (x)) [fly (x)]
The intuition behind this analysis is that (8) expresses a generalization over
individuals (birds) rather than over events or situations. The problem with the
representation in (8¢b) is that the main predicate in (8) is a stage-level predi-
cate, which projects a time variable (Kratzer 1988, 1995) but this has not been
represented in (8¢b). If we introduce a time variable and we allow the GEN
operator to bind both variables, t and x, we obtain GENx,t (bird (x)) [fly (x,t)],
which would correspond to the following interpretation: ‘in general, any bird
at any moment flies’. But this interpretation does not correspond to the intui-
tive reading. On the other hand, if we assume that GEN binds only individual
variables, we obtain the illegitimate representation in (8¢d), in which the time
variable remains free:
(8¢) d. # GENx (x is a bird) [x flies at t]
7.2.3 Proposal: Quantification over Individuals Combined

with Quantification over Times
Following Dobrovie-Sorin (2003) we will assume that generic characterizing sen-

tences (i.e., sentences that attribute a property to a class of individuals) built with a
habitual or dispositional predicate (i.e., a transitory predicate used habitually or
dispositionally) involve not one but two different adverbs:
(8¢) e. GENx (bird (x)) [HABt [fly (x,t)]]
These two operators differ in two respects. On the one hand, they quantify over
different domains (individuals and times, respectively) and, on the other, one of
them is two-place and the other is unary. GEN is a two-place operator: it denotes the
relation between two sets corresponding to the restriction and the scope. HAB, on
7.2 Characterizing Sentences with Habitual Predicates 199
the other hand, is comparable to a frequency adverb10: it is a unary operator that

applies to a transitory predicate, binding its time variable and returning the corre-
sponding habitual predicate.
The hypothesis that a covert frequency adverb HAB co-exists with the binary
operator GEN enables us to reconcile Carlson’s (1977a, c) proposal, according to
which a generic operator is unary, with later analyses, including Carlson (1989),
which introduce a binary operator. On the analysis proposed here, these two
accounts are not mutually exclusive; rather, they are both necessary. Following
Carlson (1977a, c), we assume that the operator responsible for the habitual use
of episodic predicates is unary. But unlike Carlson, we also assume a two-place
operator GEN.
The co-occurrence of a frequency adverb and an adverb of quantification is sup-
ported by examples such as (11), with two overt adverbs:
(11) In general, a student rarely reads novels.
(11¢) GENx (student(x)) [FEWt (reads novels (x, t)]]
The hypothesis assumed here, according to which a default frequency adverb
HAB co-occurs with a binary operator GEN is plausible given the systematic
ambiguity of overt adverbs of quantification. With possible few exceptions,
adverbs of quantification (in particular, rarely, sometimes, usually) may function,
depending on the context, both as quantifiers and as frequency adverbs (de Swart
1991).
Let us now show that the analysis of Q-adverbs proposed here can solve the
problems noted above. To account for the impossibility of the generic reading of
(9a–c), which are built with adjectival s-level predicates, we have to explain why
examples of this kind cannot be associated with LFs such as (8¢e), i.e., with LFs
that rely on HAB. In purely descriptive terms, the problem is clear: the main
predicates in (9) cannot have a habitual interpretation (tired, infected, or available
cannot be interpreted as meaning ‘usually be tired, infected, available’). The
unavailability of this reading may be explained by assuming that unlike verbs,11
adjectives cannot legitimate the insertion of a default HAB operator. And because
HAB is not inserted, the time variable remains unbound, yielding an ill-formed
representation:
(9¢) a. # GENx (dog (x)) [tired (x, t)]
10
This does not mean, however, that HAB, which is covert, has the same scopal interpretation as
an overt frequency adverb; HAB may instead be viewed as a pluractional operator, as proposed by
van Geenhoven (2004, 2005). See Cabredo-Hofherr (to appear) on differences between plurac-
tional operators and frequency adverbs.
11
It might be the case that certain adjectives take a habitual reading and conversely that certain
s-level verbal predicates do not. What is relevant here is the fact that certain s-level predicates
(be they adjectives or verbs) cannot take a habitual reading.
Note now that the examples in (9) become acceptable if we add an overt
Q-adverb:
(12) a. A dog is rarely tired.
b. A cow is rarely infected.
c. A fireman is rarely available.
Because the Q-adverb can bind the time variable introduced by tired, infected or
available, LFs such as (12¢a) are well-formed:
(12¢) a. GENx (dog (x)) FEWt [tired (x, t)]
We can also account for the ambiguity of (11) if we assume the LFs in (11¢a–b),
where a default adverb (HAB and GEN, respectively) co-exists with an overt adverb,
‘rarely’, translated as FEW, which can function either as a binary operator or as a
frequency adverb. The ambiguity of (11) can thus be analyzed as an ambiguity
between the relative scope of two operators:
(11¢) a. GENx (student (x)) [FEWt [smokes (x, t)]]
b. FEWx(student (x)) [HABt [smokes (x, t)]]
7.2.4 Conclusions
We have proposed that the LF representations of habitual characterizing sentences

like (8) involve not one but two adverbs of quantification, a binary and a unary one
(frequency adverb). According to this analysis, habitual characterizing sentences do
not express generalizations over events but rather generalizations over individuals
that are characterized by a habitual or a dispositional property.
7.3 The Genericity of Singular Indefinites
In this section we will examine two constraints on the generic readings of singular
indefinites.12 The so-called ‘nomicity constraint’ concerns the type of predicate
required in order for a subject indefinite to be interpreted as generic. The second
constraint explains why indefinites in object positions cannot receive a genuine
generic reading.
12
The genericity of plural indefinites raises further problems, which will be presented in Sect. 7.4
below.
7.3 The Genericity of Singular Indefinites 201
7.3.1 GEN and the Nomicity Constraint
Let us now consider the following examples, due to Lawler (1973) and Burton-
Roberts (1977):
(13) a. ◊ A madrigal is popular.13
b. ◊ A king is generous.
c. ◊ A room is square.
To account for the unacceptability of examples of this kind, Lawler and Burton-
Roberts have proposed a nomicity constraint: the generic reading of singular
indefinites is possible only if the generalization expresses a property that is nomic,
necessary, essential, inherent or analytic (for a recent discussion of this constraint
see Cohen (2001) and Greenberg (2007)). Note that the predicates in (13a–c) denote
contingent properties (of the subject), whereas those in (14) denote inherent proper-
ties (of the subject):
(14) a. A madrigal is polyphonic.
b. A square has four sides.
Note that (13a–c) can have a generic reading if we add an overt adverb of
quantification14:
(15) a. A madrigal is usually popular.
b. A king is rarely generous.
c. A room is usually square.
The contrast between (13a–c) and (15a–c) suggests that the unavailability of the
generic reading of (13a–c) is not due to a constraint on adverbial quantification in
general but rather to a constraint on the insertion of the default operator GEN. We
can thus suggest that the nomicity constraint mentioned earlier is a condition on the
insertion of GEN.
(16) In characterizing sentences, GEN can be introduced by default only if an
inherent relation holds between the property denoted by the indefinite and
the property denoted by the sentential predicate.
Bare plurals in English (as well as plural definite DPs in French) are not subject
to this constraint.15 The glosses of the primed examples in (17) are not indicated
13
The symbol ◊ means that the sentence cannot have a generic interpretation.
14
There are other factors that make the generic reading possible in these examples, such as the
presence of modality, which can be expressed by a modal verb or simply by intonation:
(i) A king must be generous.
15
The contrast between singular indefinites and bare plurals in English with respect to the nomicity
constraint was observed by Lawler (1972, 1973) and Burton-Roberts (1977). In Sect. 7.5 below we
will show that English generic bare plurals cannot be analyzed as indefinite-like expressions.
because they are identical to the corresponding English examples (modulo the
definite article):
(17) a. Madrigals are popular.
a¢. Les madrigaux sont populaires.
b. Kings are generous.
b¢. Les rois sont généreux.
c. Rooms are square.
c¢. Les chambres sont carrées.
The acceptability of these examples is expected on the hypothesis that the
genericity of English bare plurals and French plural definites should be analyzed in
terms of kind-reference (rather than in terms of an indefinite being bound by GEN).
The examples in (17) can thus be analyzed as involving generic predication: a
generic (i.e., stable) property is attributed to a kind-referring DP. Because GEN
need not be introduced, the nomicity constraint is irrelevant for generic sentences
built with kind-referring DPs (see Dobrovie-Sorin and Laca 1996): we can attribute
to a kind (i.e., to a maximal intensional group of individuals) a contingent property
as well as an inherent property.
Note that the constraint in (16) is also relevant for modified indefinites. Thus, the
example in (18a) is more acceptable than that in (18b), because a relation exists
between being good and being generous, whereas no natural (e.g., lexical or causal)
relation exists between being blond and being generous:
(18) a. A good king is generous.
a¢. Un bon roi est généreux.
b. ◊ A blond king is generous.
b¢. ◊ Un roi blond est généreux.
Note finally that the constraint in (16) does not concern generic sentences built
with if/when-clauses. In this type of sentence, a default GEN operator is supplied by
the connector if/when itself (Heim 1982; Farkas and Sugioka 1983):
(19) a. A madrigal is popular when it is short.
a¢. Un madrigal est populaire quand il est court.
b. A madrigal is popular if it is written by Orlando di Lasso.
b¢. Un madrigal est populaire s’il est écrit par Orlando di Lasso.
(19¢) a. GENx (madrigal (x) Ù short (x)) [popular (x)]
b. GENx (madrigal (x) Ù written-by (x, Orlando di Lasso)) [popular (x)]
These LFs can be rewritten as in (19″), where quantification over events is
explicit.
(19″) a. GENe (short (e, f(e)) Ù madrigal (f(e))) [popular (e, f(e))]
b. GENe (written by (e, f(e), Orlando di Lasso) Ù madrigal (f(e))) [popular
(e, f(e))]
7.3 The Genericity of Singular Indefinites 203
The observations made in this section, which bear on indefinites in subject

positions, can be summarized as follows:
(20) a. The generic readings of indefinites in subject positions are due to
sentential genericity: the presence of an overt or covert adverb of
quantification is necessary.
b. Depending on whether the adverb of quantification binds the indefinite
directly or indirectly, we may talk of ‘truly generic’ readings and
‘pseudo-generic’ readings, respectively.
c. A GEN operator can be supplied by default either due to the presence of
an if/when operator or due to a ‘nomic’ relation between the predicates
that are mapped onto the restriction and the scope.
7.3.2 Pseudo-Generic Indefinites in Object Positions
According to rule (4), quantification over individuals is obtained if: (i) the indefinite is
in a subject position (more precisely in the preverbal subject position) and (ii) the predi-
cate is generic (i.e., a predicate lexically specified as stable, or a transitory predicate in
its habitual use). The constraint on the predicate can be easily explained: if the predicate
denotes a transitory property, it projects a time variable that must be either saturated by
a time index (in this case we obtain a particular and not a generic statement) or bound by
the adverb of quantification, which therefore cannot bind the individual variable that the
indefinite supplies (recall that we do not assume that Q-adverbs are unselective binders:
depending on the linguistic context, they can bind either an event variable or an indi-
vidual variable but not both of them). As for the constraint regarding the position occu-
pied by the indefinite, we suggest that this constraint is due to the need for an asymmetric
configuration such as subject/VP or Theme/Rheme, which is mapped onto the restric-
tion / scope partition characteristic of quantificational structures.16
Rule (4) cannot apply if the indefinite occupies an object position, which explains
why a generic reading is not available for the indefinite object in (21), which can
only refer to a particular individual:
(21) John respects a professor.
The generic reading becomes available if (i) the verb is stressed or (ii) the indefinite
in the object position contains an adjective or some other type of adnominal modifier:
(22) a. In general, John RESPECTS a professor.
b. In general, John respects an intelligent professor.
c. In general, John respects a professor who takes care of his students.
d. Mary rarely likes a short sleeved-dress.
The generic reading of (22a) is made possible by contrastive stress, which
enables the application of rule (5). The whole sentence goes to the nuclear scope
and the restriction is filled by focus closure, which is obtained by replacing the
16
For a different account of the role that information structure plays in generic sentences and cor-
relatively in the choice of the generic interpretation of BPs, see Cohen & Erteshik-Shir (2002).
focused constituent (i.e., the predicate respect) with a context-dependent predicate

of events (compatible-with-respect):
(22¢) a. GENe (compatible-with-respect (e, John, f(e)) Ù professor (f(e)))
[respect (e, John, f(e))]
This representation, in which the GEN operator quantifies over events, can be
paraphrased as follows: most events compatible with respect in which John and a
professor participate are events in which John respects that professor.
Turning now to (22b–c), we can assume that it is the presence of an adnominal
modifier that allows the indefinite NP to occur in the restriction:
(23) An indefinite DP that contains an adnominal modifier (adjective, relative
clause, prepositional DP) goes to the restriction of the adverb of quantification
that takes scope over it.
(22¢) b. GENx (professor (x) Ù intelligent (x)) [respect (John, x)]
LFs such as (22¢b), which contain two conditions in the restriction, can be rewrit-
ten using quantification over events, by analyzing the condition supplied by the
modifier intelligent as a condition on events:
(22¢) c. GENe (intelligent (e, f(e)) Ù professor (f(e))) [respect (e, John, (f(e)))]
We can conclude that object indefinites can take pseudo-generic readings in
those contexts in which quantification over events is induced by contrastive stress or
nominal modification.
Since neither contrastive stress nor nominal modification are present in (24), the
BP cannot be analyzed as an indefinite bound by a Q-adverb.17
(24) John respects teachers.
In this example, the genericity of the bare NP can instead be analyzed as an
instance of nominal genericity, as English bare plurals can function as names of
kinds (Carlson 1977a, c).
The analysis sketched here explains why French definite plurals occurring in
object positions can take generic readings, in clear contrast with French des-
indefinites, which are not acceptable18:
(25) a. Jean respecte les professeurs. (generic/specific)
Jean respects the teachers
b. *Jean respecte des professeurs.
Jean respects des teachers
17
See Laca (1990), Dobrovie-Sorin and Laca (1996) and Dobrovie-Sorin (1997a, b) for further
evidence against this analysis of ‘generic objects’, advocated in particular by Diesing (1992) and
Kratzer (1988, 1995).
18
Plural indefinites headed by certains ‘certain’ or quelques ‘some’ are acceptable in this context,
but they can only take a specific reading:
(i) Jean respecte certains/quelques professeurs.
Jean respects certain some teachers
7.4 The Genericity of French des-Indefinites 205
7.4 The Genericity of French des-Indefinites
The generic reading of French plural indefinites is known to be subject to strict

constraints, depending on the type of plural determiner involved. We will focus on
the properties of des-indefinites, leaving aside other plural determiners such as
plusieurs ‘several’ or certains ’some’ etc. (for an overview of the basic data,
see Corblin (1987, 2001), from whom we borrow most of the examples below). We
distinguish ‘truly generic’ from ‘pseudo-generic’ indefinites and we show that while
the former occur in sentences that express generalizations over groups of individu-
als, the latter are related to generalizations over events.
7.4.1 The Individuation Constraint on Quantification
Corblin (1987:57–58) observed that generic cardinal indefinites cannot express

generalizations over atomic individuals: “Il n’existe pas d’interprétation
générique distributive stricte des indéfinis nombrés” [‘There is no strictly dis-
tributive generic reading for cardinal/numbered indefinites’]. Corblin’s explana-
tion relies on a pragmatic principle according to which examples like those in
(26b) are blocked (or, in Corblin’s terms, neutralized) because they can be more
readily expressed by using a singular indefinite (see (26a)) instead of a cardinal
indefinite (see (26b)):
(26) a. Une tortue vit longtemps.
A turtle lives long-time
‘A turtle has a long life-span.’
b. *Deux / Trois tortues vivent longtemps.
Two three turtles live long-time
‘Two/three turtles have a long life span.’
Coming back to our present concerns, the same principle would account for sim-
ilar restrictions on des-indefinites.19 Thus, examples like (27b) would be blocked by
the corresponding (27a), with a singular indefinite:
(27) a. Un carré a quatre côtés.
A square has four sides
‘A square has four sides.’
b. *Des carrés ont quatre côtés.
des squares have four sides
19
Corblin (1987) notes that ‘strictly distributive’ readings (i.e., generalizations over atomic indi-
viduals) are marginally possible with des-indefinites. We will come back to this issue in Sect. 7.4.3
but for now we focus only on examples that express generalizations over groups.
Below we give further examples of the same type, grouped according to whether
they contain an overt adverb of quantification or not:
(28) a. *Des enfants marchent rarement avant 10 mois.
des children walk rarely before 10 months
b. *Des Indiens meurent en général jeunes.
des Indians die usually young
(29) a. *Des chats sont intelligents.
des cats are intelligent
b. *Des Roumains parlent français.
des Romanians speak French
Dobrovie-Sorin (2004) shows that in order to understand the generic readings of
des-indefinites, we need to distinguish between examples of the type illustrated
above, which rely on generic quantification over (plural) individuals and examples
like (30), which rely on generic quantification over events20:
(30) a. Méfie-toi, des guêpes énervées sont un danger terrible.
watch out des wasps excited are a danger terrible.
‘Watch out, excited wasps are a terrible danger.’
b. Des éléphants blancs se promenant dans la rue ont parfois/
des elephants white strolling in the street have sometimes/
toujours suscité une très vive curiosité.
always aroused a very vivid curiosity
‘Always/sometimes if white elephants stroll in the street they arouse
curiosity.’
In this section we will only be interested in adverbial quantification over
individuals and therefore we will avoid overt Q-adverbs, modal operators or nomi-
nal modifiers, all of which may allow, in certain contexts, a quantification-
over-events interpretation.
Going back to the unacceptability of the generic reading of plural des-indefinites
illustrated in (27b) and (29), we will depart from Corblin’s pragmatic explanation
and pursue the line of investigation originating in Dobrovie-Sorin and Mari (2007a,
b) and further developed in Dobrovie-Sorin (to appear a), according to which the
examples under examination here violate a constraint on quantification. The label
and the definition in (31) are borrowed from Dobrovie-Sorin (to appear a):
(31) Individuation Constraint on Quantification (ICQ)
A variable that ranges over elements ordered by the part-whole relation cannot
be bound by a quantifier.
20
Example (30b) is adapted from an Italian example due to Longobardi (2002). Longobardi’s
account for generic indefinites is different from the one pursued here (for discussion, see Dobrovie-
Sorin and Laca (2003)).
Following Link (1983), plural (and conjoined) DPs (Determiner Phrases) are
viewed as denoting sums (of individuals), which are currently represented as
elements of a join semi-lattice (Scha 1981; Link 1983; Landman 1989a, b):
a+b+c
(32)
a+b a+c b+c
a b c
This lattice structure is generated by applying the join-operation to the set of

individuals that stands at its bottom, {a,b,c}. Thus, given a domain of reference that
contains three boys, a, b and c, the denotations of boy and boys are those in
(32¢a–b)21:
(32¢) a. [[boy]] = {a, b, c}
b. [[boys]] = { a + b, a + c, b + c, a + b + c}
Plural definite DPs such as these boys or the boys refer to a unique sum (relative
to a given context) in the domain, i.e., they refer to either a + b or to a + c, or to b + c
or to a + b + c. Indefinite plurals, e.g., des garçons in French or some boys in English,
as well as English bare plurals (on their existential readings) supply variables that
range over the sums of the semi-lattice in (32).
Given this analysis of plural indefinites, the LF representation in (27¢b) is ruled
out (# indicates an illegitimate configuration) by the ICQ, because the variable sup-
plied by des carrés ‘[DES] squares’ ranges over elements ordered by the part-whole
relation and as such it cannot be bound by the Q-adverb:
(27¢) b. # GEN X (X is a sum of squares) [X has four sides]
The ban against quantifying over elements ordered by the part-whole relation
has been frequently invoked in relation to quantification over situations; see in
particular Kratzer (1990, 1998) and Heim (1990). In the words of Kratzer (1995
:169): ‘Quite generally, any sort of quantification seems to require that the domain
of quantification is set up in such a way that its elements are truly distinct’, where
‘truly distinct’ means ‘not related to each other by part-whole relations.’ According
to Kratzer (1990, 1998), situations are particulars that are parts of worlds (worlds
are maximal situations, i.e., situations that are not a proper part of any other situa-
tion). Because of their inherent mereological part-whole structure, situations can be
quantified over only if we take into account situations ‘of the right size’, a quite
21
Theorists disagree as to whether the denotation of plural nouns contains the set of singular
individuals that generates the lattice or not. For simplicity we will adopt the latter view in this
chapter.
complex and highly debated issue not relevant for our present concerns, which relate
to quantification over the domain of individuals.
Although central to the investigation of quantification over situations, the
Individuation Constraint on Quantification has gone largely unnoticed (prior to
Dobrovie-Sorin and Mari (2007a, b)) in the literature dedicated to quantification
over individuals. The reason for this state of affairs is presumably the fact that
singular count nouns denote sets of singular individuals, which are not ordered by
the part-whole relation and as such trivially allow for quantification. Characterizing
sentences built with bare plurals and Q-adverbs were themselves treated as involving
quantification over singular individuals. Thus, examples of the type in (33) were
analyzed as relying on the LF representations in (33¢), in which the bare plural sup-
plies a variable over singular individuals (Diesing 1992)22:
(33) Squares have four sides.
(33¢) GENx (x is a square) [x has four sides]
Granting that plural marking cannot be ignored by the syntax-semantics map-
ping rules and assuming that the bare plural in (33) is comparable to a plural
indefinite (rather than a name of kind), the LF representation associated with
(33) should be (27¢b) rather than (33¢). Note also that the analysis according to
which (33) translates as (33¢) predicts that French des-indefinites, which are the
closest counterparts of indefinite-like English bare plurals, allow generic inter-
pretations. The unacceptability of examples of the type in (27b) shows that this
prediction is wrong.
The unacceptability of the generic reading of French plural indefinites sheds
serious doubt on the hypothesis that English generic bare plurals are indefinite-like
expressions. They can instead be analyzed as names of kinds, an issue that is not
relevant at the present stage of this presentation. We will return to this issue in
Sect. 7.5 below, where further arguments against representing (33) as relying on
(33¢) will be given.
The ICQ accounts not only for our initial examples, built with plural indefinites,
but also for the unacceptability of generic mass indefinites, which have not been
examined in previous literature:
(34) a. *De l’eau est liquide.
de l’water is liquid
b. *De l’eau gèle à 0° Celsius.
de l’water freezes at 0° Celsius
c. *Du vin est cher ici.
du wine is expensive here
22
Although theorists of genericity do not agree on whether bare plurals in examples like (33) are
indefinite-like (Diesing 1992; Kratzer 1988, 1995, a.o.) or kind-referring (Carlson 1977a, c), all of
them seem to assume LFs of the type (33¢).
The variables introduced by mass indefinites range over amounts/portions of

matter, which are ordered by the part-whole relation and as such cannot be bound
by a Q-adverb.23
It is easy to observe that des-indefinites can be represented as sum-variables
bound by existential closure:
(35) Dans la rue, des étudiants étaient en train de danser.
in the street des students were prog dancing
‘In the street, sm (unstressed some) students were dancing.’
(35¢) $X (X is a sum of students Ù X is dancing)
We must therefore acknowledge that the ICQ does not concern existential closure.
This is not surprising, since existential closure cannot be assimilated to two-place
quantification. Note indeed that in order to assign a truth-value to an LF of the type
in (35¢) we do not need to count the sums of students in the domain. All we need to
know is whether or not there is a sum that is students and dances.
Let us also observe that cardinal determiners, e.g., ten in Ten students were
dancing in the street, are not problematic either, since in the corresponding LF such
determiners do not function as genuine quantifiers (they do not denote relations
between sets) but rather as cardinality predicates: $X (X is a sum of students ∧ X is
dancing ∧ | X| = 10). To conclude, the ICQ does not prevent sums from being mea-
sured (with numbers or with other measure units).
7.4.2 The Generic Readings of Plural Indefinites Built

with Symmetric Nouns
Let us now consider the examples in (36)24:

(36) a. Deux/Des parrallèles se croisent à l’infini.
Two des parallels refl cross at the infinite
‘(Two) Parallels cross each other at the infinite.’
b. Deux/Des pays limitrophes ont souvent des rapports difficiles.
Two des countries neighboring have often des relations difficult
‘(Two) Neighboring countries frequently/often have difficult relations.’
23
For further empirical support in favor of the ICQ see Dobrovie-Sorin (to appear a), who shows
that this constraint is relevant not only for Q-adverbs but also for quantificational determiners such
as keine ‘noplur’ in German. The unacceptability of the generic readings of bare plurals and bare
mass nouns in Romance languages other than French can also be explained as consequences of the
ICQ (see Sect. 7.5).
24
The versions of (36a–c) built with deux are taken from Corblin (1987). We have added the
examples with des-indefinites. For reasons we will not address here, the generic readings of
des-indefinites are less acceptable than those of cardinal indefinites.
c. Deux/De vrais jumeaux se ressemblent dans les moindres détails.

Two de true twins refl resemble in the smallest details
‘(Two) True twins look alike down to the smallest details.’
d. Deux/Des amis intimes se critiquent toujours.
Two des friends intimate refl criticize always
‘(Two) Intimate friends always criticize each other.’
Note that these examples are built with plural symmetric nouns, e.g., friends,
twins, neighbors, etc., whereas the ungrammatical examples in (27b), (28)–(29) are
built with plural sortal nouns, e.g., squares, child, cats, etc. Building on Dobrovie-
Sorin and Mari (2007a, b), let us assume that plural symmetric nouns denote sets of
groups. Compare the denotation of friends given in (37a) to the denotation of boys
given in (32¢b) and repeated in (37b):
(37) a. [[friends]] = {P,Q,R}
(where P,Q and R are groups, the members of which entertain a reciprocal
friendship relation)
b. [[boys]] = {a + b, a + c, b + c, a + b + c}
(where a, b and c are individual boys)
Crucially, although friends is a plural form, it is not obtained by semantic plural-
ization from the singular noun friend (such singular nouns do not denote a set of
individuals but rather a symmetric relation between two individuals). Such plural
forms are ‘inherent plurals’, comparable to reciprocals (love/hate each other, meet,
etc.). The group interpretation of friends represented in (37a) should be distin-
guished from the sum interpretation of friends. Note indeed that examples such as
(38) are ambiguous between the two interpretations in (38a) and (38b)25:
(38) John and Mary are friends.
a. John and Mary entertain a friendship relation.
b. John and Mary are friends of mine
The (b) interpretation relies on the pluralization of the singular noun friend,
which has one of its argument positions saturated by an implicit argument (the
speaker in this case). In other words, the sum-referring friends corresponds to the
25
Note that in French this ambiguity exists for des-indefinites in predicate position (see (i)),
whereas the BP in (ii) necessarily takes the reciprocal interpretation:
(i) Jean et Marie sont des amis.
Jean and Mary are des friends
(ii) Jean et Marie sont amis.
Jean and Mary are friends
‘Jean and Mary are friends’
pluralization of friend (of me), whereas the group-referring friends is an inherent

plural, which denotes a set of groups, each group being made up of individuals that
entertain a friendship relation.
Since (under the relevant interpretation) plural symmetric nouns denote sets of
groups (i.e., pluralities that are not ordered by part-whole relations), the variables
supplied by plural symmetric nouns can be bound by a Q-adverb. In other words,
the LF in (36¢) is well-formed:
(36¢) a. GENx (x is a group of parallels) [x cross each other at the infinite]
Dobrovie-Sorin and Mari (2007a, b) make two observations in support of their
hypothesis that plural symmetric nouns may denote sets of groups. They point out
that group-denotation requires the plural noun to describe a symmetric relation that
holds among the elements of the group (Landman 1989b; Simons 1987; Moltmann
1997), which ensures that the elements of the plurality form an ‘integrated whole’
rather than a mere sum/aggregate of individuals that do not bear any relation to each
other. More precisely, the groups denoted by symmetric plural nouns are examples
of Simons’ (1987) ‘relation-based integrated wholes’. The second observation is
that plural symmetric nouns do not satisfy cumulativity, in clear contrast with plu-
ralized sortal nouns26:
(39) a. A are children
B are children
A and B are children
b. A are friends (with each other)
B are friends (with each other)
# A and B are friends (with each other)
The following examples, taken from Carlier (2000:184), provide further support
for our analysis:
(40) a. Des amis s’ entraident.
des friends refl mutual-help
‘Friends help each other.’
b. Des soeurs rivalisent souvent.
des sisters compete often
‘Sisters often compete with each other.’
26
Dobrovie-Sorin (to appear a) observes that plural symmetric nouns are divisive (if A is friends,
then any group that is part of A is friends), which is incompatible with Dobrovie-Sorin & Mari’s
(2007a, b) view that such plural nouns denote sets of groups that are not ordered by the part-whole
relation (since divisiveness depends on the part-whole relation). Dobrovie-Sorin (to appear a) was
therefore led to revise the analysis of plural symmetric nouns: plural symmetric nouns allow
generic readings because their domain contains a legitimate domain of quantification, namely the
set of the maximal groups that satisfy the relation denoted by the noun.
c. Des aimants s’ attirent.

des magnets refl attract
‘Magnets attract each other.’
d. Des jumeaux ont souvent des affinités.
des twins have often des affinities
‘Twins often have affinities.’
These examples are built with symmetric plural predicates and as such can be
represented as relying on quantification over groups of individuals:
(40¢) a. GEN X (friends (X)) [X help each other]
The interpretation conveyed by these LF representations corresponds to our intu-
itions: the sentences in (40) express generalizations over groups of individuals, who
entertain a stable relation, which is either habitual or dispositional.
Summarizing, adverbial quantification over plural individuals is severely con-
strained: only variables that range over plural individuals that are not ordered by
part-whole relations can be bound by a Q-adverb. This constraint explains why
plural symmetric nouns but not plural sortal nouns allow generic readings.
7.4.3 The Pseudo-Generic Reading of Plural Indefinites
In this paragraph we will briefly examine examples that arguably rely on adverbial
quantification over events (rather than over individuals).
Let us first consider the examples in (41):
(41) a. Trois/Des hommes forts peuvent soulever un piano.
Three des men strong can lift a piano
‘(Three) Strong men can lift a piano.’
b. Deux/des hommes grands attirent toujours l’attention.
Two des men tall draw always the attention
‘(Two) Tall men always draw attention.’
c. Deux/Des pipelettes ne se supportent pas longtemps.
Two des chatterboxes neg refl stand neg for long
‘(Two) Chatterboxes won’t stand each other for a long time.’
d. Deux/Des enfants en bas âge donnent toujours beaucoup de travail.
Two des children small give always a lot of work
‘(Two) Small children always involve a lot of work.’
e. Trois/des petites filles sont toujours en train de te préparer une
Three des small girls are always prog you prepare a
surprise.
surprise
‘(Three) Small girls are always preparing surprises.’
The sentence in (41a) does not say that any group of men have a certain habit or
typical behavior but rather that any events/situations in which three strong men try
to lift together a piano are such that they manage to lift it.
In order to account for this type of example, we must assume a new mapping rule:
(42) Plural indefinites can be analyzed as supplying a predicate of events
paraphrasable as ‘be together’, which is mapped into the restriction.
This rule yields the representations in (41a–e)27:
(41¢) a. GENe (be together (e, f(e)) Ù strong men (f(e)) Ù |f(e)| ³2) [can lift a
piano (e, f(e))]
b. GENe (be together (e, f(e)) Ù tall men (f(e)) Ù |f(e)| ³2) [draw attention
(e, f(e))]
c. GENe (be together (e, f(e)) Ù chatterboxes (f(e)) Ù |f(e)| ³ 2) [stand each
other for a long time (e, f(e))]
d. "e (be together (e, f(e)) Ù small children (f(e)) Ù |f(e)| ³ 2) [involve a lot
of work (e, f(e))]
e. "e (be together (e, f(e)) Ù small girls (f(e)) Ù |f(e)| ³ 3) [prepare surprise
(e, f(e))]
Let us now observe that the examples in (43) are not necessarily interpreted as
expressing generalizations about events that contain more than one lion. Therefore,
we cannot assume that the restriction of the corresponding LFs contains ‘be
together’:
(43) a. Des lions blessés sont toujours vulnérables.
des lions injured are always vulnerable
‘Injured lions are always vulnerable.’
b. Des enfants malades sont souvent grincheux.
des children sick are often grumpy
‘Sick children are often grumpy.’
As observed by Heyd (2002), these examples are built with des-indefinites
modified by adjectives that can function as sentential predicates. Such adjectives
can be analyzed as predicates of events and can thus provide the restriction of a
quantifier over events:
(43¢) a. GENe (injured(e,f(e)) Ù lions (f(e))) [vulnerable (e,f(e))]
b. GENe (sick (e,f(e)) Ù children (f(e))) [grumpy (e,f(e))]
27
The LF in (41¢), as well as most of the other LFs in this section, are simplified versions. Thus, in
(41¢a), the events of being together and being able to lift a piano are not the same event but rather
two overlapping events (see the LF in (1¢)). Furthermore, we have not given a full analysis of the
DPs themselves, since we have not separated the predicates supplied by the noun from the predi-
cates supplied by their modifiers.
Note finally that the (pseudo-)generic interpretation of plural indefinites is

facilitated by the use of deontic modal verbs (Carlier 1989). Although the
(b) example below does not contain any overt deontic operator, the negated verb
is interpreted as containing a negated deontic : ‘do not behave like that’ means
‘must not behave like that’ (because if they behaved like that they would not be
‘true’ policemen’):
(44) a. Des jeunes filles doivent se montrer discrètes.
des young girls must refl show discrete
‘Young ladies must behave discretely.’
b. Des agents de police ne se comportent pas ainsi dans une
des agents of police neg refl behave neg like this in a
situation d’ alarme
situation of alarm
‘Police agents do not behave like that in emergency situations.’
One may wonder why deontic modality facilitates the pseudo-generic reading of
a plural indefinite. The first answer that comes to mind is that the modal operator,
due to its introducing possible worlds into the computation, introduces quantification
over events. However, this cannot be a complete solution, since it is not the case that
all modal operators have the same effect. We may therefore conjecture that the role
of the deontic operator is to express the idea that there is a necessary relation between
the predicates in the restriction and in the scope of the Q-adverb.
If this suggestion is on the right track, it seems to indicate that some kind of
‘nomicity’ constraint holds not only for the default GEN operator that quantifies
over individuals but also for the default GEN operator that quantifies over events. In
other words, in the absence of an if/when configuration, a default GEN operator can
be supplied only if a nomic relation can be established between the restriction and
the scope. This suggestion is supported by Corblin’s (1987) observation that exam-
ples like (44) can be felicitously used only if the speaker talks to a group of police
agents or young ladies. We may speculate that Corblin’s pragmatic condition on the
use of examples like (44) can be explained as being due to the fact that the speaker
intends his utterance as advice for the hearer(s) and as such the interpretation is
facilitated if the subject of the generalization underlying the advice has the same
cardinality as the hearer. In sum, utterances such as (44a–b) are intended to mean:
‘If young ladies [such as you] want to be considered to be true young ladies, then
they must behave discreetly’; ‘If police agents [such as you] want to be considered
to be true police agents, then they must not behave like that in an alarm situation.’
Let us now observe that in examples like (43) above, the plural indefinites are
‘number-neutral’, in the sense that they express generalizations that hold not only
for plural individuals but also for atomic individuals.28 The examples in (30),
28
According to Corblin (1987), the relevant examples allow ‘strictly distributive readings’. This
generalization is correct only if we interpret ‘allow strictly distributive readings’ as meaning ‘num-
ber neutral readings’.
introduced in Sect. 7.4.1 above and repeated below, illustrate the same
phenomenon:
(30) a. Méfie-toi, des guêpes énervées sont un danger terrible.
watch out des wasps excited are a danger terrible.
‘Watch out, excited wasps are a terrible danger.’
b. Des éléphants blancs se promenant dans la rue ont parfois/
des elephants white strolling in the street have sometimes/
toujours suscité une très vive curiosité.
always aroused a very vivid curiosity
‘If white elephants stroll in the street they always/sometimes arouse
curiosity.’
The sentence in (30a) can be interpreted as a warning against either groups of
excited wasps or against single wasps and in (30b) the curiosity might have been
caused by groups of elephants or by a single elephant strolling in the street.
In what follows, we show that number neutralization can arise only in examples
that rely on quantification over events.29 Consider the LF representation of (30b):
(30¢) b. "e (stroll in the street (e,f(e)) Ù white elephants (f(e)))
[arouse curiosity (e,f(e))]
In these LF representations, the plural indefinites are not translated as variables
that range over (groups of) individuals but as Skolem terms, the values of which
co-vary with the events they depend on. The event participants may be atomic indi-
viduals or groups, depending on the lexical properties of the main predicate as well
as pragmatic factors. In sum, we propose that number neutralization characterizes
event-dependent indefinites, which are represented as Skolem terms:
(45) des-indefinites that translate as event-dependent Skolem terms are number-
neutral.
Number neutralization should be kept distinct from strictly distributive
readings. Strict distributivity, on which the main predicate holds of atomic indi-
viduals, is obligatorily associated with generic sentences built with singular
indefinites (Sect. 7.3 above) and can sometimes characterize generic sentences
built with plural definites (Sect. 7.5 below). Truly generic plural indefinites dis-
allow both number neutralization and strict distributivity: they supply a group
variable and as such, the sentences containing them necessarily express gener-
alizations over groups.
29
It goes without saying that number neutralization cannot arise in examples such as (41), which
according to the analysis presented above express generalizations over events of ‘being together’,
thus constraining the cardinality of the event participants to be at least two.
7.4.4 Conclusions
To conclude, French plural des-indefinites have two different kinds of generic

readings. On the one hand, plural indefinites can be indirectly bound: they are
pseudo-generic, rather than truly generic. The adverb of quantification (or the
default quantifier GEN) quantifies over events or situations and indirectly over
groups of individuals. On the other hand, plural indefinites can be directly bound by
the Q-adverb if they are built with symmetric relational nouns, which denote a
stable property of groups of individuals.
7.5 The Genericity of English Bare Nouns
English bare plurals (and bare mass nouns) differ from their counterparts in Romance
languages like Italian, Spanish or Romanian and from French des-indefinites as well
as from French mass indefinites headed by de la or du (contracted form of de and
le), insofar as they systematically allow generic readings.30 We assume that English
bare plurals (and bare mass nouns) with generic readings should be analyzed as
names of kinds (cf. Carlson 1977a), or more precisely as intensional maximal sums
(Chierchia 1998; Dobrovie-Sorin and Laca 1999, 2003).
7.5.1 The Genericity of French Plural and Mass Indefinites

and the Genericity of English Bare Nouns
As already mentioned in Sect. 7.4.1 in relation to examples (28), the restrictions on

the generic readings of French plural des-indefinites cast serious doubt on the pre-
dominant analysis of English, according to which bare plurals like those in (46)–(47)
are plural indefinites that are respectively bound by an overt adverb of quantification
or by the generic operator GEN:
(46) a. Children rarely walk before the age of 10 months.
b. Indians usually die young.
(47) a. Cats are intelligent.
b. Romanians speak French.
30
Generic readings of modified bare nouns are possible in Italian for a restricted set of examples,
which led Chierchia (1998) to analyze Italian bare nouns – and by extension bare nouns in
Romance languages other than French – as kind-denoting. However, Longobardi (2002) and
Delfitto (2002) have shown that modified bare nouns in Italian are ruled out in contexts that allow
names of kinds. Consequently, a Carlson-style analysis cannot account for the Italian data. We
agree with Longobardi and Delfitto that the generic reading of Italian bare plurals relies on
quantification over events.
7.5 The Genericity of English Bare Nouns 217
These examples contrast with the corresponding French sentences, built with
des-indefinites:
(48) a. *Des enfants marchent rarement avant 10 mois.
des children walk rarely before 10 months
b. *Des Indiens meurent en général jeunes.
des Indians die usually young
(49) a. *Des chats sont intelligents.
des cats are intelligent
b. *Des Roumains parlent français.
des Romanians speak French
The examples below illustrate the perfectly parallel contrast between English
bare mass nouns and French mass indefinites headed by de la/du:
(50) a. Water is liquid (most of the time).
b. Gold is (often) yellow.
(51) a. *De l’eau est liquide (la plupart du temps).
de l’water is liquid the majority of time
b. *De l’or est (souvent) jaune.
de l’gold is often yellow
Given the unacceptability of plural and mass indefinites in the examples in
(48)–(49) and (51),31 the English bare nouns in (46)–(47) and (50) cannot be
analyzed as indefinites. We are thus led to assume that English bare plurals (and
bare mass nouns) function as names of kinds (Carlson 1977a, b) or rather as
intensional maximal sums (Chierchia 1998) not only in sentences built with
predicates that select kinds, e.g., Cats are on the verge of extinction, but also in
generic characterizing sentences of the type in (46)–(47), built with Q-adverbs or
the GEN operator.
On this proposal, the French counterparts of (46)–(47) and (50) are not the exam-
ples in (48)–(49) and (51) but rather those in (52)–(53), with plural and mass
definites (see Dobrovie-Sorin and Laca 1996, 1999):
(52) a. Les enfants marchent rarement avant 10 mois.
the children walk rarely before 10 months
b. Les Indiens meurent en général jeunes.
the Indians die usually young
31
De Swart (1992, 1996) suggests that the restricted use of French des-indefinites is due to the fact
that des/du goes back to a partitivity marker, which would be incompatible with a quasi-universal,
generic interpretation. However, this hypothesis cannot be maintained: on the one hand, the exam-
ples analyzed here contain des-indefinites with no partitive meaning and on the other, indefinites
with partitive des can also have generic readings (see Heyd 2002).
(53) a. Les chats sont intelligents.

the cats are intelligent
b. Les Roumains parlent français.
the Romanians speak French
(54) a. L’ eau est liquide (la plupart du temps).
the water is liquid the majority of time
b. L’ or est (souvent) jaune.
The gold is often yellow
What remains to be explained is why examples of the type in (46)–(47) and
(50) or (52)–(53) and (54), built with kind-referring expressions, allow for adver-
bial quantification over atomic individuals, in clear contrast with examples built
with plural indefinites.
7.5.2 Adverbs of Quantification and Kind-Predication
According to the current view (Chierchia 1998, a.o.), generic sentences built with
names of kinds, e.g., (55a), repeated below, are to be represented as shown in (55¢):
(55) a. Les chats sont intelligents.
the cats are intelligent
‘Cats are intelligent.’
(55¢) a. GENx (x is an instantiation of Çcats) [x is intelligent]
In this representation, the name of kind (or more precisely the maximal inten-
sional sum of cats, notated Çcats, which is obtained by applying the Down operator
Ç
to the plural property cats) has been shifted to the set of atomic individuals that
realize the kind and the GEN operator binds a variable that ranges over those indi-
viduals. This type-shifting operation might be motivated by the need to adjust the
denotation of the subject to the denotation of the main predicate: the main predicate
expresses a property of individuals, whereas the subject refers to a name of kind.
Note however that this motivation is rather weak, given that according to Chierchia,
bare plurals are not really names of kinds (as they are for Carlson (1977a, c)) but
rather intensional maximal sums of individuals and therefore no sortal conflict arises
between the subject and the predicate. Moreover, this analysis cannot be extended to
examples such as (54), repeated in (56) below, which are built with kind-referring
mass DPs:
(56) a. L’eau est liquide la plupart du temps.
The water is liquid the majority of the time
‘Water is liquid most of the time’
b. L’or est souvent jaune.
the gold is often yellow
‘Gold is often yellow.’
7.5 The Genericity of English Bare Nouns 219
Given the analysis of kind-referring plural DPs sketched above, the LF

representations of (56)a–b would be of the type in (56¢):
(56¢) GENX (X is an instantiation of Çwater) [X is liquid]
In (56¢) Çwater notates the maximal intensional sum of water, obtained by apply-
ing the Down operator Ç to the mass property water.32 We use capitals in order to
notate variables that range over non-atomic entities, in this case, over amounts or
portions of water. GEN quantifies over the portions that instantiate the maximal
intensional amount of water.33
The problem is that LFs of the type in (56¢) are ruled out by the Individuation
Constraint on Quantification, since in (56¢), X ranges over elements ordered by the
part-whole relation.
We are thus led to conclude that the possibility of the generic readings of definite
mass nouns vs. the impossibility of the generic readings of indefinite/bare mass
nouns cannot be explained by the current analysis of generic sentences and of mass
quantification.
In order to solve this problem, we may adopt Higginbotham’s (1994) analysis of
mass quantifiers. According to Higginbotham, mass quantifiers denote relations
between objects (unlike count quantifiers, which denote relations between sets).
(57) Mass quantifiers denote relations between objects.
(Higginbotham (1994), following Moravicsk (1973), Roeper(1983), Lønning
(1987))
Higginbotham proposes to associate the FL in (58¢) to examples like (58):
(58) a. All gold is yellow.
b. Most water is liquid.
(58¢) a. ALLmass (sx. gold(x), sx. yellow(x))
b. MOSTmass (sx. water(x), sx. liquid(x))
In (58¢), s is the sigma operator, the maximalizing operator that applies to a set
and picks up the maximal element in that set. Sigma is comparable to Chierchia’s
(1998) Down operator. According to (58¢)a–b, mass quantifiers denote relations
between two objects, more precisely two maximal sums, e.g., the maximal sum of
32
On the use of the Down operator in the analysis of kind referring mass nouns, see Dayal (2004).
Chierchia himself uses the Down operator only for kind-referring plural DPs, e.g., bare plurals in
English.
33
Quantification over amounts (or portions) of substance is assumed, implicitly or explicitly, by
most theorists of mass quantification. Thus, according to Gillon (1992: 632), in examples of the
type Most water is liquid, ‘[the quantifiers] range over elements in […] the greatest aggregate in
the domain of discourse of which the mass noun is true.’ This type of analysis may be adequate if
the pragmatic context makes obvious a certain partition of water, e.g., there are five buckets of
water, three of which containing liquid water and the other two frozen water. But note that the
partition provided by the pragmatic context allows the mass noun to function as a count noun. In
other words, this is a case of covert mass-to-count coercion.
gold in the world and the maximal sum of yellow stuff in the world. The formal
analysis of mass quantifiers is quite complex and cannot be presented here. Crucial
for our present concerns is only the fact that under Higginbotham’s (1994) analysis,
the restriction of mass quantifiers is not filled with a set but rather with an object, a
maximal sum (see sx. gold(x) and sx. water(x) in (58¢)), which can be measured: in
order to assign a truth value to (58¢b) we compare the measure of the maximal sum
of portions of water with the measure of the maximal sum of portions of liquid
water. Crucially, in order to evaluate LFs of the type in (58¢), we need to know the
size of the object in the restriction of the quantifier but we do not need to count the
number of parts of that object under a given partition.
This analysis explains why Q-adverbs allow kind-referring mass DPs in their
restriction: kind-referring mass DPs are maximal intensional objects and the
Q-adverb tells us what proportion of the kind satisfies the main predicate (in other
words, the Q-adverb compares the size of the kind with the size of the part of it that
satisfies the main predicate).
Turning now to Q-adverbs built with kind-referring plural terms (English bare
plurals and French definite plurals), we may extend Higginbotham’s (1994) analysis
of mass quantifiers to quantifiers over plural domains:
(59) Cats are intelligent.
(59¢) GENmass (sx. cats(x), sx. intelligent(x))
The only difference between mass quantification over mass and plural domains is
the measure unit that is used in order to evaluate the relevant LFs: since number is
inherent to plural domains, the default measure unit for plural domains is number.
7.6 Conclusions
Summarizing, the generic reading of indefinites depends on an LF representation

in which a Q-adverb binds the variable supplied by the indefinite. These
configurations, in which the quantifier denotes the relation between two sets, are
legitimate with singular indefinites and plural indefinites built with symmetric
nouns because such indefinites supply variables over elements that are not ordered
by the part-whole relation. Mass indefinites and plural sortal indefinites cannot
receive generic readings because these indefinites supply variables over elements
ordered by the part-whole relation, which cannot be bound by a Q-adverb (ICQ).
Adverbial quantification is compatible with plural and mass definites (or names of
kinds) because these expressions refer to objects that can be measured and there-
fore mass quantification can apply. In these cases, overt or covert Q-adverbs are
analyzed as mass quantifiers.
Chapter 8
Dependent Indefinites in Donkey Sentences
Indefinite DPs occurring in the so-called ‘donkey sentences’ are interpreted as

universally rather than as existentially quantified. In order to account for this old
observation, Heim (1982) and Kamp (1981) proposed that indefinites are to be
analyzed as free variables that get bound, depending on the context, either by an
adverb of quantification (as in donkey sentences) or by existential closure. This
proposal constitutes an important step towards a solution of the puzzles related to
donkey sentences but is confronted with the so-called ‘proportion’ problem. We
will therefore be led to depart from the DRT analysis of indefinites and assume
instead Steedman’s (2003) view according to which indefinite DPs occurring in
donkey sentences should be analyzed as dependent referential terms.
Pursuing the line of analysis adopted in previous chapters, we show that once
we implement dependency relations in functional terms, we can account for donkey
sentences without resorting to an analysis in terms of scope. By treating both
indefinite DPs and pronouns occurring in donkey sentences as functional terms,
this approach will be shown to provide a solution to the proportion problem.
8.1 Dependency and Donkey Sentences
8.1.1 Indefinite DPs and Universal Quantification
Sentences like (1) and (2), originally discussed in medieval texts and subsequently
examined in detail by Geach (1962), are known as ‘donkey sentences’:
(1) Every farmer who owns a donkey beats it.
(2) If a farmer owns a donkey, he beats it.

222 8 Dependent Indefinites in Donkey Sentences
These sentences raise two interrelated problems. The first one is that the
indefinite DPs a donkey in (1) and a farmer and a donkey in (2) are construed as
universally quantified rather than as existentially quantified DPs. According to
Geach (1962), the sentences in (1) and (2) have the LFs in (3):
(3) ∀x∀y ((farmer(x) ∧ donkey(y)) → beat(x,y))

In this representation, the indefinite DPs in (1) and (2) are translated as universal
quantifiers and the pronouns it in (1) and he and it in (2) are analyzed as bound
variables: both x and y in the consequent of (3) are bound by the universal quantifiers
∀x∀y, which take scope over the implication. The first problem with this account is
that indefinite DPs are analyzed, depending on the context, either by means of exis-
tential quantification, as illustrated by the LF in (4¢), or by means of universal
quantification, as in (3) above.
(4) John saw a donkey.
(4¢) ∃x (donkey(x) ∧ saw(John,x))
The second problem raised by the analysis in (3) is a syntax-semantics mismatch:
the pronouns he and it in (1) and (2) are interpreted as bound by their respective
antecedents, a farmer and a donkey, but they are not c-commanded1 by those
antecedents; in other words, donkey pronouns are peculiar in that they are out-
side the scope of their antecedent. In (1), the indefinite DP a donkey is embedded
in a relative clause and therefore does not c-command the pronoun it occurring in
the main clause. Likewise, in (2), the indefinite DPs a farmer and a donkey do not
c-command the pronouns he and it in the main clause, as they are embedded in a
conditional clause.
Summarizing, there are two properties that characterize donkey sentences: (i) the
presence of an indefinite DP interpreted as a universal quantifier; (ii) the presence of
a pronoun that is analyzed as a variable bound by the universal quantifier corre-
sponding to the indefinite DP but is not c-commanded by it.2 These sentences thus
raise two intertwined questions, which concern the quantificational force of the
indefinite DP and the status of the pronoun. What are the conditions on the universal
1
We will not discuss the details of the debate concerning the precise definition of c-command.
The only relevant point here is the fact that the indefinite DPs, which are the antecedents of these
anaphoric pronouns, appear in a position that does not allow binding: cf. Every mani loves hisi
mother vs. *‘Hisi mother loves every mani.’
2
According to Heim (1982), we may talk about donkey sentences only for examples that are char-
acterized by both of these two properties. Which means that the following sentences, though
apparently similar, are in fact different:
(i) If someone lives in Paris, he does not live outside the capital.
(ii) Someone who lives in Paris does not live outside the capital.
While (i) is a donkey sentence, (ii) is not, because there is no anaphoric pronoun outside the rela-
tive clause. Other authors, e.g., Steedman (2003), adopt a less strict view, considering that only one
of these conditions needs to be satisfied.
8.1 Dependency and Donkey Sentences 223
reading of an indefinite DP and what is the analysis of the pronoun? Both of these
questions arise if we assume that indefinite DPs are to be analyzed as quantified
expressions. It is however clear that indefinite DPs differ from other quantified
expressions in more than one respect. In particular, it can be shown that there is a
genuine difference between indefinite DPs and universally quantified DPs with
respect to anaphoric processes (cf. Chierchia 1995c:2–10). In (1) and (2) for exam-
ple, the indefinite DP a donkey cannot be replaced with quantified DPs built with
every, each or no.
(5) a. *Every farmer who owns every/each/no donkey beats it.
(6) a. *If a farmer owns every/each/no donkey, he beats it.
Likewise, the indefinite DP a farmer in (2) cannot be replaced with every farmer.
Indeed, (7) is not a possible paraphrase of the sentence in (2):
(7) ??If every farmer owns a donkey, he beats it.
If we abandon the hypothesis that indefinite DPs are quantified expressions, we
are no longer confronted with the scope problem either: since the indefinite DP is
not analyzed as a quantified expression, the pronoun is not analyzed as a variable
bound by a quantifier and as such, it need not be in the c-command domain of the
quantifier.
8.1.2 Indefinite DPs, Free Variables and Unselective Binding
Kamp (1981) and Heim (1982) argued in favor of an analysis of indefinite DPs in
donkey sentences as free variables rather than as quantified expressions. On this
view, (i) indefinites do not have any inherent quantificational force, (ii) the
quantificational force of indefinites is provided by the first binder available in the
sentence; this is a case of unselective binding of a free variable by a quantifier
present in the context, (iii) a Q-binder introduces a tripartite structure of the form
Q (A) [B], where A is the restriction of the quantifier Q, and B, its nuclear scope
and (iv) a rule of existential closure assigns default existential force to unbound
indefinite DPs.
According to this analysis, indefinites have a default existential value. This is
illustrated in (8), where the indefinite DP a man contributes a free variable and the
condition man on that variable. Applying rule (iv) to the resulting formula, we
obtain (8¢), where the free variable x supplied by the indefinite is existentially
bound.
(8) A man is smoking.
(8¢) ∃x (man(x) ∧ smoke(x))
In certain contexts, however, indefinite DPs may acquire values other than
existential. For instance, in when/if-clauses, indefinite DPs can be indirectly
bound by adverbs of quantification such as rarely ((9a)) or sometimes ((9b)),

translated as FEW and SOME in the Logical Forms given in (9¢) a–b:
(9) a. Rarely, when a man owns a donkey, he beats it.
b. Sometimes, when a man owns a donkey, he beats it.
(9¢) a. FEWx,y (man(x) ∧ donkey(y) ∧ own(x,y)) [beat(x,y)]
b. SOMEx,y (man(x) ∧ donkey(y) ∧ own(x,y)) [beat(x,y)]
This analysis is reminiscent of Lewis (1975), who was concerned with examples
of the type in (10), which show that the variables supplied by indefinites can be
bound not only by adverbs of quantification but also by quantificational determiners
(see also (1) above).
(10) a. Usually/often/sometimes, if an Italian affords to drink, he is happy.
b. Most/many/some Italians who afford to drink are happy.
In the absence of an explicit quantificational adverb in an if/when-clause, a
generic operator GEN is introduced by default:
(11) If John owns a donkey, he beats it.
(11¢) GENx (donkey(x) ∧ own(John,x)) [beat(John,x)]
The major advantage of this kind of analysis (currently referred to as the DRT
analysis) is that it abandons the idea that indefinite DPs are quantified expressions.
As a consequence, the anaphoric relation between the pronoun and its indefinite
antecedent in donkey sentences is no longer problematic. More specifically, the
pronouns in donkey sentences are analyzed as instances of free variables, indirectly
bound by the quantifier binding the indefinite DP. The scope problem observed by
Geach disappears because the quantifier c-commands both the indefinite DP and the
pronoun and therefore can bind both of them.
Thus although indefinite DPs are not themselves analyzed as quantified
expressions, they end up being interpreted as quantificational expressions at the
end of the evaluation procedure. There are three ways in which the variable
introduced by an indefinite DP can be bound: (i) by existential closure, in which
case the indefinite acquires an existential reading, (ii) by a quantifier introduced by
the DP on which the indefinite depends (every farmer in (1)), in which case the
indefinite receives a universal reading and (iii) by the quantification associated
with the if-operator, as in (2), which, once again, yields a universal reading of the
indefinite DP.3
Summarizing, the indefinite DPs in donkey sentences are always associated
with logical formulas where a quantifier binds the variables that the indefinites
introduce. However, an analysis of this kind cannot account for all possible scenarios.
3
If there is a Q-adverb in the main clause of the if/when clause, we do not get universal quantification
but e.g., proportional quantification as in (i):
(i) When an Italian drinks, he is rarely sad.
8.1 Dependency and Donkey Sentences 225
The unselective binding hypothesis advocated by Kamp and Heim predicts that in a
sentence like (12), the generic operator GEN quantifies over < farmer, donkey > pairs.
(12) If a farmer owns a donkey, he beats it.
(12¢) GENx,y (farmer(x) ∧ donkey(y) ∧ own(x,y)) [beat(x,y)]
Yet, in addition to these so-called symmetric readings, several authors (Partee
(1984), Kadmon (1987), Kratzer (1995), among others) have pointed out the existence
of so-called asymmetric readings, which Kamp and Heim cannot account for. This
is known as the proportion problem.
8.1.3 Symmetric and Asymmetric Readings

of Donkey Sentences
The proportion problem arises for DRT analyses of donkey sentences as soon as we
consider situations in which some farmers own more than one donkey. Does (13)
mean that each farmer beats each of the donkeys he owns or that he beats at least
one of his donkeys?
(13) a. Every farmer who owns a donkey beats it.
b. If a farmer owns a donkey, he beats it.
Certain authors, e.g., Cooper (1979) argue that sentences of this kind only refer
to farmers that own a single donkey; they do not say anything about other farmers.
But the LF in (12¢), which Kamp and Heim propose for sentences like (13), does not
allow this restriction to farmers that own only one donkey. (12¢) quantifies over
< farmer, donkey > pairs, which means that for each farmer who owns more than one
donkey, we will have as many pairs as the number of donkeys he owns. Thus, ten
donkeys owned by the same farmer count ten times more than one donkey owned by
one farmer.
The issue of what and how we count is even more relevant for sentences like
(14), built with the quantifier most:
(14) Most farmers who own a donkey beat it.
The question that (14) raises is whether the counting should take into
account < farmer, donkey > pairs, in which case there is a one-to-one mapping
between farmers and donkeys, or introduce an asymmetry between the two sets. Let
us imagine there are only three farmers in the universe: one that owns one hundred
donkeys and beats them all and two others who own one donkey each but neither
beats his donkey. If we consider < farmer, donkey > pairs, we obtain one hundred
and two pairs and the sentence in (14) is true. But if we count only the farmers, the
sentence in (14) is false: only one out of three farmers beats his donkeys.
A sentence like (14) can thus yield three distinct readings, according to whether
we consider < farmer, donkey > pairs, only the farmers (who own a donkey), or only
the donkeys (that each farmer owns). The first case gives rise to a symmetric reading
(cf. (15a)) and the last two cases give rise to asymmetric readings. We further
distinguish between a subject-asymmetric reading, illustrated in (15b) and an object-
asymmetric reading, in (15c). The LFs corresponding to these three readings are
given in (15) below.
(15) a. MOSTx,y (farmer(x) ∧ donkey(y) ∧ own(x,y)) [beat(x,y)]
b. MOSTx (farmer(x) ∧ ∃y (donkey(y) ∧ own(x,y))) [beat(x,y)]
c. MOSTy (donkey(y) ∧ ∃x (farmer(x) ∧ own(x,y))) [beat(x,y)]
According to the unselective binding hypothesis advocated by Kamp (1981)
and Heim (1982), the sentence in (14) has the representation in (15a), where the
indefinite DP a donkey introduces a free variable y, which is bound by the unse-
lective quantifier most. The unselective binding hypothesis does not predict the
existence of asymmetric readings because the LFs in (15b–c) cannot be obtained
via unselective binding. These LFs are moreover problematic because of the bind-
ing of the pronoun: the existential quantifiers that occur in the restriction do not
c-command the variables y in (15b) and x in (15c), which occur in the nuclear
scope ([beat(x,y)]).
The choice between the three possible readings (symmetric, subject-asymmetric
and object-asymmetric) is determined by different factors, which have been dis-
cussed by Heim (1982), Kadmon (1987), Rooth (1985) and Kratzer (1995), among
others. Rooth (1985) and Kadmon (1987), for instance, have shown that in English,
different readings obtain depending on whether the indefinite DPs in the antecedent
of a conditional are stressed4 or not. Thus, while (16a) gives rise to a subject-
asymmetric reading, glossed as in (16b), (17a) gives rise to an object-asymmetric
reading, glossed as in (17b).
(16) a. If a farmer owns a DONkey, he usually beats it.
b. Most of the farmers who own a donkey beat it.
(17) a. If a FARmer owns a donkey, he usually beats it.
b. Most of the donkeys that belong to a farmer are beaten by this farmer.
In (16a), the counting does not concern the < farmer, donkey > pairs but rather the
farmers who own donkeys. Each farmer is counted once, regardless of whether he
owns one or more donkeys. By contrast, in (17a), the counting concerns the donkeys
(that are owned by the farmers).
Kratzer (1995) showed that, in addition to intonation and information structure,
other parameters such as unaccusativity, type of predicate (stage-level or individual-
level), scrambling, or the presence/absence of anaphoric pronouns in the consequent,
can influence the preferred reading. For instance, when two indefinite DPs and no
pronoun are present in the antecedent of a conditional as in (13b), there is a prefer-
ence for the symmetric reading, whereas when there is one anaphoric pronoun in
addition to the two indefinite DPs in the antecedent, the asymmetric reading is
4
The stressed syllables are written in capitals.
8.2 Dependent Indefinites 227
favored, as in (18)–(19), due to Comorovski (1997), where the object-asymmetric

reading is the preferred reading.
(18) When an article on hisi private life harms a politiciani, hei usually tries to get
it censored.
(19) If a picture depicting himi flatters a politiciani, hei usually tries to get it
published.
In (18) and (19), usually quantifies over politicians, not over articles or pictures.
Chierchia (1995c) suggests that there are interesting contrasts between donkey
sentences with relative clauses (13a) and donkey sentences with conditionals (13b).
More precisely, the former involve quantification over the individuals denoted by
the head of the relative (i.e., farmers in (13a)). For this type of donkey sentences, the
symmetric and object-asymmetric readings are unavailable.
An additional problem comes from the fact that asymmetric readings can be
further divided into weak and strong readings (cf. Rooth, (1987) and Comorovski
(1999) among others). On a weak reading, the sentence makes reference to those
farmers who beat at least one of their donkeys, as illustrated by the LF in (20a). On
a strong reading, the sentence refers to the farmers who beat all their donkeys, as
illustrated by the LF in (20b).
(20) a. MOSTx (farmer(x) Ù $y (donkey(y) Ù own(x,y))) [$y (donkey(y) Ù
own(x,y) Ù beat(x,y))]
b. MOSTx (farmer(x) Ù $y (donkey(y) Ù own(x,y))) [∀y ((donkey(y) Ù
own(x,y)) → beat(x,y))]
The example in (18) above can be analyzed in a parallel way. On a strong asym-
metric reading, the sentence means that a politician tries to censor all the articles on
his private life that harm him. By contrast, on a weak asymmetric reading, it means
that a politician tries to censor one of the articles on his private life that harms him.
In sum, DRT analyses cannot account for the asymmetric readings of donkey
sentences and a fortiori they cannot capture the difference between their weak vs.
strong interpretations.
8.2 Dependent Indefinites
In what follows, we will develop a different analysis of donkey sentences, which

provides a solution to the problems mentioned above. Following Steedman (2003),
donkey indefinites will not be treated as variables bound by a quantifier but rather as
dependent referential expressions. Likewise, the anaphoric pronoun will not be ana-
lyzed as the second occurrence of a variable bound by a quantifier but rather as a
discourse pronoun co-referring with a dependent referential term.
Dependent referential terms will be represented as Skolem terms, notated f(x),
i.e., as individuals obtained by applying a Skolem function to a variable that ranges
over individuals. Skolem functions are used in logical representations in order to
eliminate existential quantifiers. In case an existential quantifier is not in the scope

of any quantifier, it is deleted and the variable bound by it is replaced by a constant.
In case the existential quantifier is in the scope of another quantifier, it is deleted and
the variable bound by it is replaced by a Skolem term notated f(x). Thus, after
Skolemization, a formula such as ∀x $y P(x,y) becomes ∀x P(x, f(x)).
8.2.1 Dependency on a Situation
In the previous chapter, we have seen that generic quantification over individuals
should be distinguished from generic quantification over situations or events:
(21) A dog has four legs.
(22) When John invites a friend, he (always) cooks dinner.
The sentence in (21) has the representation in (21¢), in which the default GEN
operator quantifies over individuals (i.e., dogs):
(21¢) GENx (dog(x)) [have-four-legs(x)]
The generic reading of (22), on the other hand, comes about as the result of
quantification over situations: always translates as a universal quantifier over the
situations described by the when-clause (which functions as the restrictor of the
quantifier)5:
(22¢) ∀e (invite (e, John, f(e)) Ù friend (f(e))) [cook-dinner (e, John)]
The generic reading of the indefinite is not due to quantification over friends but
rather to a representation in which the adverb quantifies over situations in which
John invites a friend. It is the dependency on situations that is responsible for the
pseudo-generic reading of the indefinite DP. This dependency is represented in
functional terms in (22¢), where the value of the indefinite a friend depends on the
inviting event; f(e) is a referential term, whose denotation varies depending on the
value of e. The representation in (22¢) can be paraphrased as in (23):
(23) Every time John invites a friend, he cooks dinner.
It should be stressed that the analysis in (22¢) goes against DRT type analyses
according to which the variables supplied by indefinites are bound by unselective
quantificational adverbs; on our analysis, they are referential terms (represented as
Skolem terms) dependent on situations that are quantified over by a Q-adverb.
5
The LF in (22¢), as well as other LFs in this chapter, is imprecise insofar as it does not explicitly
indicate that the events in the restriction and in the nuclear scope are not identical but rather over-
lapping sub-events. This simplification does not affect the main argument.
8.2 Dependent Indefinites 229
Consider now (25), in which the matrix clause contains a pronoun that picks up
the same referent as the situation-dependent indefinite:
(25) When John invites a friend, he cooks dinner for her.
The pronoun her in (25) refers to the same individual as the functional term f(e),
i.e., the friend that John invites in situation e. This interpretation is associated with
the LF in (25¢), where the first two occurrences of f(e) correspond to the dependent
indefinite a friend, whereas the third corresponds to the pronoun her.
(25¢) GENe (invite (e, John, f(e)) Ù friend (f(e)))
[cook-dinner (e, John, f(e))]
In (25¢), the referential term f(e) is dependent on inviting events e and as such f(e)
is indirectly bound by the GEN operator (GEN is supplied by default whenever no
overt Q-adverb is present in the matrix clause associated with a when-clause), which
quantifies over inviting events.
Any occurrence of f(e) bound by the GEN operator must be in its scope. This
analysis correctly predicts that the use of an anaphoric pronoun further in the dis-
course is ungrammatical.6
(26) When John invites a friend, he cooks dinner for her. *She enjoys it.
In (26), the pronoun she, which is anaphoric on the dependent indefinite DP,
should be translated as f(e). But this occurrence of f(e) appears outside the scope of
GEN; e is thus free and, as a result, the term f(e) that translates she cannot be inter-
preted as situation-dependent.
The analysis of the indefinites as dependent referential terms solves the problem
raised by anaphoric pronouns in donkey sentences. According to the traditional
logic account, indefinites are quantified expressions and anaphoric pronouns are
variables bound by the quantifier that these expressions introduce. This approach
leads to the well-known problem, first observed by Geach (1962) and later read-
dressed in the literature: a supposedly quantified DP (a friend in (25)) seems to
bind a variable (her), which it does not c-command. On the analysis adopted here,
this problem does not arise because dependent indefinite DPs are not treated as
quantified expressions but rather as dependent referential terms. The referential
nature of the dependent indefinite explains why it can co-refer with a pronoun
occurring outside the minimal clause containing the indefinite. Moreover, the fact
that the indefinite is dependent on an event explains why the anaphoric pronoun
6
For some speakers, the sequence in (26) is grammatical and has the same reading as the sentence
in (i).
(i) When John invites a friend, he cooks dinner for her and she enjoys it.
On this reading, the example is acceptable because she appears in the matrix of the when-clause,
i.e., in the scope of GEN.
must occur in the scope of the quantifier over events e.7 Crucially, the pronoun need
not be c-commanded by the indefinite DP itself but instead must be c-commanded
by the operator that quantifies over the events on which it depends.
8.2.2 Dependency on a Quantified DP
An indefinite DP can be dependent on a quantified DP. In (27), for example, the

indefinite expression a donkey is dependent on the quantified DP every farmer. This
dependency can also be represented by means of a functional term (see f(x) in (27¢)).
(27) Every farmer who owns a donkey is rich.
(27¢) ∀x [(farmer(x) Ù own(x, f(x)) Ù donkey (f(x))) → rich(x)]
The reference of the indefinite a donkey, translated as f(x), depends on the value
associated with the variable x, which is bound by the universal quantifier corre-
sponding to every farmer. As indicated by the fact that (27) cannot be paraphrased
as in (28) below, the indefinite a donkey is clearly not a universally quantified
expression.
(28) Every farmer who owns every donkey is rich.
Since it acts as a referential element, a functional term can serve as an antecedent
for discourse anaphora. Accordingly, a functional term can have more than one
occurrence within the same formula. This is precisely the case of donkey sentences
such as (1), repeated here as (29) and translated as (29¢):
(29¢) ∀x [(farmer(x) Ù own(x, f(x)) Ù donkey(f(x))) → beat(x, f(x))]
The indefinite a donkey translates as the functional term f(x) in the representation
in (29¢), in which the Skolem function f assigns a different donkey to each farmer x.
The indefinite DP serves as an antecedent for the anaphoric pronoun it, which is
associated with the third occurrence of f(x) in (29¢).
As already observed above in connection with the example in (26), with an event-
dependent indefinite, any occurrence of f(x) must be in the scope of the quantifier
that binds x. Hence the ungrammaticality of the second sentence in (30), where the
anaphoric pronoun is outside the scope of every:
(30) Every farmer who owns a donkey beats it. *It is maltreated.
7
Besides if/when clauses, quantification over events and correlative dependency of participants on
the event may be triggered by modal and intensional verbs or tenses such as the generic present.
All the examples paraphrasable by “in all events in which…” involve quantification over events
and dependency of the participants on the event.
8.3 Dependency and Proportion 231
Let us insist on the fact that the relation between a donkey and it in (29) should
be kept distinct from the type of relation illustrated in (31) and represented in (31¢),
between a quantified DP and a bound variable :
(31) Every farmer says that he is happy.
(31¢) ∀x [farmer(x) → x says that x is happy]
In (31), he is interpreted as a variable bound by the universal quantifier; in (31¢),
he corresponds to the second occurrence of the variable x in the consequent of
(31¢).
In (29), on the other hand, it is an anaphoric pronoun whose antecedent is a ref-
erential term dependent on a quantified DP (a donkey, which depends on every
farmer). Both the antecedent indefinite DP and the donkey pronoun are notated f(x),
which corresponds to the donkey(s) owned by the farmers x.
We can thus distinguish two types of pronouns: (i) pronouns that function as
bound variables, which must be in the scope of the quantifier that binds them and
(ii) pronouns that are anaphoric on a dependent indefinite DP. It is this latter kind of
pronouns that are called ‘donkey pronouns’.
8.3 Dependency and Proportion
8.3.1 Symmetric and Asymmetric Readings
The so-called ‘proportion problem’ can be accounted for in terms of dependency

relations. Let us consider the sentence in (32), which contains more than one depen-
dent indefinite DP: a donkey and a farmer.
(32) If a farmer owns a donkey, he usually beats it.
As already observed in Sect. 8.1.3, the sentence in (32) is ambiguous. On its
symmetric reading, both indefinite DPs are represented as Skolem terms (f(e) and
g(e), which respectively correspond to a farmer and a donkey), which depend on the
event variable introduced by the conditional clause and bound by MOST, corre-
sponding to the quantificational adverb usually.
(32¢) a. MOSTe (own(e) Ù Agent(e, f(e)) Ù farmer(f(e)) Ù Theme(e,g(e))
Ù donkey(g(e))) [beat (e, f(e), g(e))]
According to this representation, the adverb of quantification quantifies over
events and indirectly over pairs of farmers and donkeys. The two indefinite DPs
depend on the situation e but are independent with respect to each other, which
explains why the interpretation is ‘symmetric’.
(32¢a) has the same truth-conditions as the LF in (33a), which relies on the
hypothesis that the Q-adverb quantifies over pairs of individuals, along the lines of
Lewis (1975) and DRT analyses.
(33) a. MOSTx,y (farmer(x) Ù donkey(y) Ù own(x,y)) [beat(x,y)]

Besides the symmetric reading, (32) can also have asymmetric readings, arising
when there is a dependency between the indefinite DPs. There are two different
asymmetric readings, according to whether the subject depends on the object or vice
versa.
On the subject-asymmetric reading, a donkey is dependent on the indefinite DP
a farmer, which is itself dependent on the situation e. Consequently, a farmer, which
is situation-dependent, is represented in (32¢b) by the Skolem term f(e) and a don-
key, which exhibits a double dependency, is represented by the Skolem term g(f(e)),
where the function g marks the dependency of a donkey on a farmer and the func-
tion f marks the dependency of a farmer on the situation:
(32¢) b. MOSTe (own(e) Ù Agent (e, f(e)) Ù farmer (f(e))
Ù Theme (e,g(f(e))) Ù donkey (g(f(e))) [beat (e, f(e), g(f(e)))]
The formula in (32¢b) is equivalent to that in (33b), where the dependency on the
situation is not explicitly marked and the dependency of one DP on the other is
marked in functional terms.
(33) b. MOSTx (farmer(x) Ù donkey(f(x)) Ù own(x,f(x))) [beat (x,f(x))]
As for the object asymmetric reading, it corresponds to the case where the
indefinite DP a farmer is dependent on the indefinite a donkey, which is itself depen-
dent on the situation. This time, it is the DP a farmer that exhibits a double depen-
dency: it depends on a donkey, which depends on the situation:
(32¢) c. MOSTe (own(e) Ù Agent (e, g(f(e))) Ù farmer (g(f(e))) Ù
Theme (e,f(e)) Ù donkey (f(e)) [beat (e, g(f(e)), f(e))]
The formula in (32¢c) can be simplified by leaving the dependency on the situa-
tion implicit, as in (33c):
(33) c. MOSTy (farmer(f(y)) Ù donkey(y) Ù own(f(y),y))
[beat (f(y),y)]
8.3.2 Weak and Strong Asymmetric Readings
As already observed in Sect. 8.1.3, asymmetric readings can be either ‘weak’ or

‘strong’. We must thus distinguish between the two subject-asymmetric interpreta-
tions of the example in (32), repeated below in (34), which would correspond
respectively to the LFs in (35a) and (35b):
(34) If a farmer owns a donkey, he usually beats it.
8.3 Dependency and Proportion 233
(35) a. MOSTx (farmer(x) Ù $y (donkey(y) Ù own(x,y)))

[$y (donkey(y) Ù own(x,y) Ù beat(x,y))]
b. MOSTx (farmer(x) Ù $y (donkey(y) Ù own(x,y)))
[∀y ((donkey(y) Ù own(x,y)) → beat(x,y))]
In both cases, the indefinite DP a donkey depends on the indefinite DP a farmer.
By representing dependent indefinites as Skolem terms rather than as existentially
closed variables, the LFs in (35a–b) can be rewritten as in (36a–b):
(36) a. MOSTx (farmer(x) Ù donkey(f(x)) Ù own(x,f(x)))
[$y (donkey(y) Ù own(x,y) Ù beat(x,y))]
b. MOSTx (farmer(x) Ù donkey(f(x)) Ù own(x,f(x)))
[∀y ((donkey(y) Ù own(x,y)) → beat(x,y))]
In order to account for the weak and strong asymmetric interpretations,
Comorovski (1999) draws on insights from Lappin and Francez (1994), who sug-
gested that the function f, which establishes the dependency between the two DPs a
farmer and a donkey, instead of ranging over atomic individuals, here the set of
donkeys, ranges over a set that contains not only atomic but also plural individuals.
So f(x) may be either one or several donkeys. In other words, f(x) corresponds to
either the donkey that the farmer x owns, or the donkeys that the farmer x owns. On
a strong asymmetric reading, f maps every farmer x onto the maximal sum of don-
keys that x owns. On a weak reading, f maps every farmer x onto the sum of the
donkeys that x beats and only those donkeys. In the case where a farmer does not
beat all of his donkeys, this plural entity is different from the maximal sum of don-
keys that x owns. In other terms, the weak and strong asymmetric readings of (34)
are analyzed as two pragmatic versions of the same LF representation, (36c). Note
that this representation is identical to that in (33b), the only difference being that the
predicates donkey, own and beat are starred, which indicates that their arguments
may be either atomic or plural entities.
(36) c. MOSTx (farmer(x) Ù donkey*(f(x)) Ù own*(x,f(x))) [beat*(x,f(x))]
Krifka (1996) tried to show that the preference for a weak/strong reading is due
to certain semantic or lexical parameters. However this is not totally correct. Even
if the type of predicate – eventive versus non-eventive (cf. von Fintel (1994)) or
partial versus total (cf. Krifka (1996)) – may favor one reading or the other, Dekker
(2004) and Steedman (2003) have argued that the problem that these readings pose
is not semantic but rather pragmatic. For example, in (37) below, the preference for
the weak asymmetric reading is due to our world-knowledge: one person cannot
wear several masks at the same time:
(37) Every person who owns a mask wears it.
8.4 Dependency and Reference
In this section, we focus on the referentiality of dependent indefinite DPs and

anaphoric pronouns that occur in donkey sentences. The indefinites in donkey
sentences should by no means be analyzed as quantificational expressions, neither
as first-order quantifiers (quantifying over individuals) nor as second-order
quantifiers (quantifying over functions). Contrary to Comorovski (1999), we have
avoided existential quantification over functions f and followed Steedman (2003) in
using Skolem terms8 (rather than variables over Skolem functions, as Comorovski).
With this approach, indefinite DPs in donkey sentences are represented as entity-
referring dependent Skolem terms notated f(x) or f(e),9 in which a name of function
f applies to an individual or an event variable, respectively. Their distinctive feature
is that they do not have fixed reference; their reference co-varies with the value
of the element on which they depend. Skolem terms thus allow us to capture depen-
dency relations.
It is important to note that as far as anaphoric processes are concerned, these
dependent indefinites are more like proper names or definite descriptions than like
universal quantifiers. Thus, the indefinite DP a donkey in donkey sentences is similar
to Dexter in (38) and differs from a quantified DP like every donkey in (39) in that
it allows an anaphoric pronoun that it does not c-command either outside the rela-
tive clause, as in (a), or further in the discourse, as in (b):
(38) a. Every person who knows Dexteri adores himi.
b. All the students know Dexteri. They adore himi.
(39) a. * John who owns every donkeyi adores iti.
b. * John owns every donkeyi. He adores iti.
Moreover, just like definite descriptions, which can serve as antecedents for ana-
phoric definite descriptions (cf. (40)), dependent indefinites may also serve as ante-
cedents for anaphoric definite descriptions (cf. (41)). By contrast, this is never the
case for universally quantified expressions (cf. (42)).
(40) a. (John’s wife)i came. Shei had forgotten heri keys.
b. (John’s wife)i came. The poor womani had forgotten heri keys.
(41) a. Every farmeri who owns a donkeyj feeds itj.
b. Every farmeri who owns a donkeyj feeds hisi animalj.
(42) a. Every farmeri thinks that hei has to feed his donkey.
b. *Every farmeri thinks that the poor mani has to feed hisi donkeyj
The analysis of dependent indefinite DPs as referential terms explains why, with
respect to anaphoric processes, they are closer to proper names or definite descriptions,
8
Steedman (2003) uses a different notation.
9
Steedman insists that Skolem terms do not denote functions (a name of function is merely part of
their representation) but are referential terms (type e).
8.6 Appendix 235
which are also of type e, than to quantified expressions, of type <<e,t>,t>.

Nevertheless, because they are dependent terms, these indefinite DPs do not have all
the anaphoric properties that proper names or definite descriptions have.
It is furthermore interesting to note that like indefinite DPs, definite DPs such as
his mother-in-law or the meal can exhibit a dependent behavior, as illustrated in
(43a–b):
(43) a. When a man does not like his mother-in-law, she does not have an easy life.
b. When a man prepares the meal, it is rarely a success.
8.5 Conclusion
In order to solve the problems raised by donkey sentences, various solutions have
been proposed in the literature. We have reviewed the limits of a quantificational
approach to indefinite DPs in donkey sentences as well as the limits of DRT analy-
ses,10 for which the asymmetric readings are problematic. These accounts are prob-
lematic because they treat indefinite DPs on a par with quantified expressions. We
have instead proposed that indefinite DPs in donkey sentences should be analyzed
as dependent DPs represented as Skolem terms. This type of analysis accounts for
the whole range of available readings (symmetric/asymmetric, weak/strong) and
also explains why they cannot serve as antecedents of anaphoric pronouns in exam-
ples such as (26) or (30) above.
8.6 Appendix
Here we will present two pre-DRT proposals that attempted to solve the problems
raised by donkey sentences. The first is due to Egli (1979), who elaborated an algo-
rithm that translates a fragment of English (including DPs headed by every,
indefinite DPs, proper names and pronouns) into formulas of predicate logic. Egli’s
proposal presents two drawbacks: it predicts some readings that do not exist and it
postulates rules of translation for indefinite DPs that are context-dependent. It turns
out that this semantics of indefinite DP’s is not compositional (see Sect. 8.1). The
second proposal, due to Evans (1980) relies on the idea that pronouns in donkey
sentences are not bound pronouns but rather ‘E(vans)-type pronouns’, which are
disguised definite descriptions, since they are equivalent to definite descriptions
that can be recovered from the context (see Sect. 8.6.1). Cooper (1979) proposed
a formal analysis of pronouns, which can be used to account for donkey sentences.
10
For a presentation of three other accounts of donkey sentences, the interested reader is referred
to the appendix.
The major problem with Cooper’s proposal is that it predicts too many interpreta-
tions due to the fact that the interpretation of the pronoun does not depend on the
syntactic environment of the pronoun but on pragmatic factors (see Sect. 8.6.2).
8.6.1 Egli’s (1979) Solution
Egli (1979) argues that in predicate logic, the formulas in (1a) and (1b) are equiva-
lent, as long as there is no free occurrence of x in j.
(1) a. ($x F) → j
b. ∀x (F → j)
An existential quantifier in the antecedent of a conditional clause is equivalent to
a universal quantifier taking scope over the conditional. This equivalence can explain
the universal interpretation of indefinite DPs in donkey sentences if there is no ana-
phoric pronoun in the consequent. However, since the equivalence in (1) only holds
for cases where there is no free occurrence of x in j, it cannot account for cases
where there is a pronoun in the consequent, which acts as a free variable.
Egli suggests that the logical equivalence in (1) holds in natural language, where
the restriction concerning the variables in j does not apply. According to him, it is
always possible to replace the existential quantifier in the antecedent of a condi-
tional with a universal quantifier having wide scope over the conditional, regardless
of whether or not the consequent contains unbound occurrences of the free variable.
His hypothesis is supported by other cases of binding of pronouns that occur outside
the scope of their antecedent.
(2) a. Somebody left, and he went home.
b. Somebody left. He went home.
c. Somebody left and went home.
In other words, on Egli’s proposal, there is a mismatch between scope in logic
and scope in natural language.
In order to account for the interpretation of indefinite DPs in conditionals, Egli
posits the existence of a specific rule, without, however, taking a clear position on
the precise nature of this rule. Apparently, it is not a grammatical rule but a rule that
applies to an intermediate level, where the meaning of the sentence is computed. For
example, in analyzing (3), Egli first builds the intermediate representation (3¢a) in
which the quantified DP is translated as a universal quantifier that scopes over the
implication. He then applies a conversion rule, which turns the existential $ corre-
sponding to the indefinite DP a donkey into the universal ∀ resulting in the repre-
sentation in (3¢b).
(3¢) a. ∀x [(man(x) Ù x owns a donkey ) → (x beats it )]
b. ∀x∀y [(man(x) Ù donkey(y) Ù x owns y) → (x beats y )]
8.6 Appendix 237
Egli’s solution has several advantages but also a number of shortcomings. One
advantage is that it captures the interpretation of examples such as (4a), which can-
not be paraphrased as in (4b):
(4) a. One of my friends who owns a donkey beats it.
b. For each donkey, if one of my friends owns it, then, he beats it.
(4¢) a. $x (friend-of-mine(x) Ù x owns a donkey Ù x beats it)
b. $x $y (friend-of-mine(x) Ù own(x,y) Ù donkey(y) Ù beat(x,y))
The fact that (4a) cannot be paraphrased as (4b) is due to the nature of the DP on
which the relative depends. The DP one of my friends (unlike every farmer in (3)
above) is not a universally quantified DP; rather, it is an indefinite DP that is associ-
ated with existential quantification. The sentence in (4a) has the intermediate repre-
sentation in (4¢a). Notice that the conversion rule that transforms the existential $
corresponding to the indefinite DP a donkey into the universal ∀ does not apply in
this case, because there is no logical implication in (4¢a) (contrary to what happens
in (3¢a)). The representation in (4¢b) is obtained by replacing a donkey in (4¢a) with
a variable bound by an existential quantifier. To sum up, when the DP on which the
relative clause depends is not universally quantified, we are not in a donkey context
anymore.
However, Egli’s solution raises at least two problems. As we have already said,
it is not clear what the nature of the rule that converts $ to ∀ is. It is also unclear why
the constraint on the absence of free variables disappears in natural language.
Note moreover that in addition to the equivalence in (1), the equivalence in (5)
also holds in predicate logic. According to this equivalence, a universal quantifier
can scope out of the antecedent of a conditional once it is assigned an existential
value, provided there are no free occurrences of x in j:
(5) a. (∀x F) → j
b. $x (F → j)
The question is why the logical equivalence in (5) does not hold for natural lan-
guage sentences (possibly with the difference observed for (1) concerning the vari-
ables in j). Egli observes this difference between predicate calculus and natural
language LFs but does not provide an explanation. He points out that binding is not
possible in examples such as (6) but does not consider sentences such as (7a), which
would come out equivalent to (7b), if the rule in (5) applied. And yet, (7a) and (7b)
have clearly different meanings:
(6) *If every farmer works, he becomes rich.
(7) a. If every farmer vaccinates his donkey, the disease will not spread.
b. There is a farmer such that, if he vaccinates his donkey, the disease will not
spread.
By changing the notion of scope, Egli treats connectives like Ù and → as dynamic
connectives but this provides only a partial and somewhat ad hoc solution to the
problem of donkey sentences. Egli’s solution is not entirely satisfactory because he
analyzes indefinite DPs as quantified expressions and it is precisely this assumption

that needs to be changed.
8.6.2 E-type Analyses
Another way of accounting for donkey sentences is to analyze the pronouns that
occur in these contexts not as bound pronouns but as disguised definite
descriptions.
8.6.2.1 E-type Pronouns: Evans (1980)
Evans (1980) puts forth a different classification of pronouns. The originality of this
approach consists in introducing a novel class of pronouns, which he calls E-type
pronouns, where E is the first letter of his name. E-type pronouns are those pro-
nouns whose antecedent is a quantified expression and which are not in the scope of
the quantifier that the antecedent introduces. To put it differently, Evans makes a
distinction between bound pronouns, which act as variables bound by a quantifier,
and E-type pronouns. The pronoun his in (8) for example is a bound pronoun and
them in (9) is an E-type pronoun.
(8) Every man loves his mother.
(9) John owns some sheep and Harry vaccinates them.
There are two tests that distinguish E-type pronouns from bound pronouns. On
the one hand, unlike what happens in the case of bound pronouns, the antecedent of
an E-type pronoun cannot be replaced with an expression such as no N:
b. *Every farmer who owns no donkey beats it.
(11) a. Every man loves his mother.
b. No man loves his mother.
On the other hand, E-type pronouns, but not bound pronouns, have interrogative
counterparts:
(11) c. ??Who does every man love? Every man loves his mother.
(12) John owns some sheep and what does Harry vaccinate? John owns some
sheep and Harry vaccinates them.
E-type pronouns are reminiscent of Russell’s view of pronouns as abbreviations
of complex nominal expressions, which are used instead of full DPs for stylistic
reasons. It is however difficult to determine what exactly these pronouns replace.
One possibility would be to assume that they replace their antecedent, which would
8.6 Appendix 239
suffice to copy in order to obtain a paraphrase. It is however rarely the case that we
obtain a good paraphrase by replacing the pronoun with its antecedent:
(13) a. John is tall. He is handsome.
b. John is tall. John is handsome.
(14) a. A man walks in the garden. He smokes.
b. A man walks in the garden. A man smokes.
c. A man walks in the garden. This man smokes.
d. A man walks in the garden. The man who walks in the garden smokes.
While (13a) and b are equivalent, this is not the case for (14a) and b. We can
paraphrase (14a) either by replacing the pronoun with a demonstrative DP, as in
(14c), or by reconstructing a definite description, as in (14d).
The problem observed in (14) also appears with E-type pronouns. Substituting
the pronoun with its antecedent changes the meaning of the sentence, as shown in
(15) and (16).
b. Every farmer who owns a donkey beats a donkey.
c. Every farmer who owns a donkey beats (this donkey/the donkey he owns).
(16) a. John owns some sheep and Harry vaccinates them.
b. John owns some sheep and Harry vaccinates some sheep.
c. John owns some sheep and Harry vaccinates (these sheep/the sheep that
John owns).
In these examples, the E-type pronouns were replaced with definite descriptions.
This step is intuitively correct but raises several problems. On the one hand, in order
to make the procedure compositional, one needs to explain how the definite descrip-
tion is built, a non-trivial task considering that the reference of the noun is always
restricted by the relative clause. This is particularly problematic for those sentences
where the antecedent of the E-type pronoun is a quantified expression such as some-
one, as in (17):
(17) a. If someone lives in Paris, he does not live in London.
b. If someone lives in Paris, this someone does not live in London.
c. If someone lives in Paris, the person who lives in Paris does not live in
London.
On the other hand, the reconstructed definite description carries a uniqueness
presupposition, which seems counter-intuitive. Consider example (15) again. In
(15c), the donkey he owns is a definite description that presupposes that the person
in question owns only one donkey. Yet, it seems obvious that the sentence in (15a)
does not presuppose that there is no farmer who owns more than one donkey. It also
seems difficult to assume that the sentence only refers to those farmers who own a
single donkey and does not say anything about the other farmers. Many scholars
have tried to understand what exactly (15a) says about farmers who own more than
one donkey. Is the sentence in (15a) appropriate or inappropriate in a context where
a farmer owns several donkeys and beats only one? This brings us back to the pro-
portion problem, discussed in Sect. 8.1.3 and Sect. 8.3 above.
Let us return to example (17a) and its version in (17b), where the E-type pronoun
was replaced with a definite description. Heim (1982) showed that an E-type pro-
noun analysis induces the presupposition that there is only one person who lives in
Paris, which is untenable. One solution proposed by Davidson, Parsons and others
is to say that the uniqueness presupposition is associated with situations/events,
rather than with donkeys or inhabitants of Paris. We may then consider a minimal
situation where there is only one person living in Paris.
There still remains one problematic case, namely the case of symmetric situa-
tions pointed out by van Eijck and Kamp (1997) and illustrated in (18):
(18) a. If a man lives with another man, he shares the desk with him.
b. If a cardinal meets another cardinal, he blesses him.
These examples are problematic for the minimal situation approach, since the
minimal situation in which the sentence is true contains two individuals and there-
fore the uniqueness presupposition does not hold.
It seems nevertheless quite clear that the pronouns that are used in donkey
sentences are E-type pronouns. In the next section, we review the formal implemen-
tation of this kind of analysis, due to Cooper (1979).
8.6.2.2 A Formal Analysis with Lambda Operators
Cooper (1979) proposed that pronouns should be analyzed as incomplete definite

descriptions. This can be formally represented by means of a lambda operator,
which abstracts over a free variable denoting an n-ary relation. The value of this
n-ary relation, just like the value of the relation holding between the variables v1,…,
vn−1, is determined by the context:
(19) lK $x [∀y [R(v1, v2, …, vn−1, y) ↔ x = y] Ù K(x)]
This implementation has the advantage of giving a uniform account of pronouns:
it accounts for E-type pronouns as well as ordinary pronouns. To allow a pronoun to
pick up the same referent as its antecedent, it suffices to empty the content of the
R-relation by reducing it to, say, an identity relation. On this view, the representa-
tion in (19) is reduced to (20):
(20) l K K(x)
This analysis can also account for the so-called laziness pronouns, as in (21):
(21) John gave his paycheck to his mistress. Everyone else kept it for himself.
8.6 Appendix 241
The pronoun it in the second sentence corresponds to the definite description his
own paycheck and can be formally represented by means of a n-ary relation:
(22) lK $x [∀y [paycheck(v1, y) ↔ x = y] Ù K(x)]
This kind of account also captures the interpretation of E-type pronouns.
Its major drawback is that anaphora resolution is relegated to pragmatics, since
it is the job of pragmatics to determine how the n-ary R relation is instantiated.
Moreover, according to this analysis, all pronouns are analyzed in the same way,
regardless of whether their antecedent is a proper name, a definite, an indefinite or
a quantificational DP. It is important to note that Cooper’s solution is based exclu-
sively on the analysis of pronouns; the representation of the indefinite DP in sen-
tences such as (15a) is taken to be irrelevant. The element that acquires a universal
interpretation (see ∀ in the formula in (19)) is not the indefinite DP but the definite
description that stands for the pronoun. When the reconstructed definite description
is in the singular, the N W carries a uniqueness presupposition and when it is in the
plural, the N Ws reads as all of the N W.
Conclusion
Having reached the end of our investigation of indefinites, which consisted in a

series of specific studies (on bare nouns, existential constructions, weak and strong
interpretations, scope and genericity), we can provide a short summary of our main
results.
Two Types of Entities: Individuals and Amounts
In order to characterize indefinites, we need to assume, following Link (1983), an

enriched ontological space that contains amounts in addition to individuals (and
groups). Just like individuals, amounts are entities but non-individuated ones. Mass
indefinites constitute the paradigmatic example of amount-referring expressions:
200 g of butter does not refer to an identifiable entity but rather to an equivalence
class, the class of all pieces of butter that weigh 200 g. These pieces are not indi-
vidualized portions, unlike John and Mary, who are distinct individuals. We have
shown that both mass nouns and weak plural indefinites refer to amounts: on its
weak reading, two books refers to any quantity of two books.
The Weak/Strong Distinction
We have provided evidence that Milsark’s (1977) two-way distinction between

weak and strong indefinites should be replaced by a three-way distinction: in addi-
tion to the weak reading, we have distinguished two types of strong readings, a
quantificational and a non-quantificational referential one. We proposed that the
difference between weak and strong indefinites is denotational: weak indefinites
denote existential generalized quantifiers over amounts, to be distinguished from
generalized quantifiers over individuals, which correspond to strong indefinites.

Language and Linguistic Theory 85, DOI 10.1007/978-94-007-3002-1,
244 Conclusion
The difference between strong referential and strong quantificational indefinites is

not a matter of ambiguity, the choice between the two options being determined by
the context.
The strong or weak interpretation of indefinites is determined by their internal
structure: the presence or absence of a determiner, the lexical properties of the
determiner and the noun and singular or plural number markings. Certain indefinites,
in particular singular indefinites and cardinal indefinites, are ambiguous between a
weak and a strong reading, the choice between the two being determined by the
syntactic context: certain contexts, e.g., existential sentences, allow only weak
indefinites, whereas other contexts, e.g., the position of the external subject, allow
only strong DPs. Other indefinites can be only weak (bare plurals in Romance lan-
guages other than French, des indefinites in French and mass indefinites crossling-
uistically) or only strong (certain in English and certains in French). Unacceptability
arises when the reading determined by the internal structure of an indefinite DP
clashes with the requirements imposed by the syntactic context in which the
indefinite occurs.
Two Types of Predicates
Among the contextual factors that bear on the denotation of indefinites are the lexical
properties of the main predicates of which indefinites are arguments: certain predi-
cates, which we labeled ‘existential predicates’, e.g., run, dance or eat, write, etc.,
allow both weak and strong readings of indefinites occurring in their subject or object
positions, whereas other predicates, which we labeled ‘entity predicates’, e.g., intel-
ligent, sad or love, hate, allow only strong readings. All theories of indefiniteness
share the hypothesis that the weak reading of indefinites is allowed only if the main
predicate can be represented as ‘existential’, i.e., as supplying an existential quantifier
over the position of the indefinite. However, the lexical characterization of the rele-
vant classes of predicates is still under debate. We show that Carlson’s (1977a, c)
distinction between s-level and i-level predicates does not exactly parallel the distinc-
tion between existential and entity predicates needed for the account of weak vs.
strong readings and we follow Dobrovie-Sorin (1997a) in proposing that space local-
ization is the criterion that distinguishes between the two classes.
Weak Indefinites as Generalized Quantifiers over Amounts
In terms of formalization, the main original contribution of this book is the repre-
sentation of weak DPs (in particular bare plurals and bare mass nouns): they are not
property-referring expressions (as in van Geenhoven (1996) or Dobrovie-Sorin and
Beyssade (2004)) but rather existential generalized quantifiers over amounts. Qua
generalized quantifiers over amounts, weak indefinites cannot combine with entity
No Existential Closure 245
predicates, which can only combine with DPs that denote either individuals (type e)
or generalized quantifiers over individuals (type <<e,t>t>). They can only combine
with existential predicates (or appear in there is or il y a sentences), the role of which
is to supply an existentially bound variable over individuals, which is identified with
the amount variable introduced by the generalized quantifier over amounts.
Three Types of Indefinites
Besides denoting generalized quantifiers over amounts (which is allowed only

for indefinites, on their weak reading) indefinites may take the other canonical
types of denotation for predicates and arguments. In predicate positions,
indefinites denote properties (type <e,t>), whereas in argument positions, strong
indefinites denote either individuals (type e) or generalized quantifiers over indi-
viduals (type <<e,t>,t>).
By adopting this analysis, we have largely departed from the tradition going back
to the theories of Heim and Kamp, which hold that indefinite DPs are to be analyzed
as free variables, a proposal that we adopted at the beginning of our investigation
(see in particular Dobrovie-Sorin and Beyssade (2004)). In this book we followed
Steedman (2003) in assuming that strong indefinites are not free variables but
rather Skolem terms of the form f(x) or f(e), where f is a constant function that
applies to an individual or event variable, which is itself bound by a quantifier. All
indefinite DPs that refer to an individual were treated as Skolem terms, both in the
case of dependent DPs, whose reference varies with the context and of specific
DPs, which have a fixed but undetermined or underspecified reference, which
depends on the speaker.
Summarizing, the three types of indefinites distinguished in this book correspond
to three distinct denotational types:
(a) Weak indefinites denote existential generalized quantifiers over amounts.
(b) Non-quantificational strong indefinites are referential expressions that are
represented as Skolem terms.
(c) Quantificational strong indefinites are generalized quantifiers, which we have
represented in terms of tripartite configurations.
No Existential Closure
The use of Skolem terms enables us to dispense with the rules of existential closure
proposed in Heim (1982). Thus, in the Logical Form of a sentence like A man has
entered the room, the indefinite DP a man is represented as a Skolem term f(s),
where f is a constant function that applies to the situation of utterance and yields a
specific individual that ranges over the set of men. Since f is not a variable, no existential
quantifier is needed for the representation of this sentence. Similarly, the sentence
246 Conclusion
Every farmer who owns a donkey beats it is associated with the Logical Form ∀x
[(farmer(x) ∧ owns(x, f(x)) ∧ donkey (f(x))) → beats (x,f(x))], which also lacks an
existential quantifier. The indefinite DP a donkey, as well as the anaphoric pronoun
it, are represented as the Skolem term f(x), whose value varies according to the value
of the variable x, which ranges over the set of farmers and is bound by the universal
quantifier introduced by the quantified NP every farmer. Finally, the Logical Forms
of sentences like Whenever John invites a friend, he cooks dinner for him, which
involve quantification over events, make use of a Skolem term f(e), with an event
variable bound by the generic quantifier GEN: GENe (invitation(e) ∧ Agent(e, John)
∧ Theme(e, f(e)) ∧ friend(f(e))) [make-dinner(e, Jean, f(e))].
Diesing’s (1992) rule of VP-level existential closure is also dispensed within our
account of weak indefinites and its import is taken over by the use of an existential
predicate in the lexical representation of certain predicates, as in Carlsonian kind-
based accounts or in semantic incorporation accounts. However, as already explained
above, our analysis of weak indefinites as generalized quantifiers over amounts
differs from both kind-based and property-based accounts of bare NPs (and more
generally, of weak indefinites).
Generic Indefinites
The original contribution of the chapter devoted to generic indefinites concerns the
analysis of plural des indefinites and mass de la/du indefinites in French. We pro-
pose that the strict restrictions to which they are subject are due to an Individuation
Constraint on Quantification. This proposal has two far-reaching consequences. On
the one hand, the generic readings of bare plurals or bare mass Ns (see English in
particular but we believe this is a general, possibly universal generalization) cannot
be attributed to an indefinite-like nature of bare plurals and bare mass nouns (as in
Diesing (1992) or Kratzer (1995)): because they refer to amounts, indefinite-like
bare plurals and bare mass Ns (see Romance languages other than French or
Brazilian Portuguese) cannot take generic readings, on a par with des and de la/du
indefinites in French. The generic readings of bare NPs can therefore only be attrib-
uted to kind reference, as in Carlson (1977a, c). On the other hand, quantification
over mass domains, analyzed as in Higginbotham (1994), should be extended to
plural domains and should be clearly distinguished from what we may call ‘distribu-
tive’ quantification (see determiners built with singular NPs and indefinite DPs
mapped onto the restriction of adverbs of quantification).
Comparison with Other Approaches
In order to assess the specificity and the advantages of our account, let us briefly
compare it with other approaches in the literature, such as the Theory of Generalized
Quantifiers, DRT and analyses in terms of properties.
Comparison with Other Approaches 247
Our proposal differs from the Theory of Generalized Quantifiers, which

assumes a uniform treatment of DPs as generalized quantifiers. We have tried to
establish the need to keep apart referential and quantified expressions and have
shown how a reformulation in terms of dependency allows a better account of
scope phenomena.
The present account also departs from analyses pursued in the DRT framework,
which assume that indefinite DPs1 always introduce a discourse referent.2 Even
though indefinite DPs are not treated as quantified expressions per se, at the end of
the evaluation procedure, any discourse referent introduced by an indefinite NP in a
given DRS is analyzed as a quantified expression (bound via default existential
closure, or via unselective binding of a quantifier in the context). The analyses pro-
posed in this book are instead based on the hypothesis that indefinites allow for the
three types of denotation listed in (a)–(c) above.
Regarding the analysis of weak indefinites, we have departed from the widely
assumed mechanism of ‘semantic incorporation’ according to which weak
indefinites denote properties (type <e,t>), despite occurring in argument posi-
tions. This hypothesis cannot explain a number of empirical generalizations, such
as (i) the contrast between bare plurals and count bare singulars in Romance lan-
guages like Romanian, Spanish or Catalan, (ii) argument-predicate contrasts regard-
ing the use of bare NPs (in French, bare NPs are allowed in post copular positions
but not in argument positions) and (iii) the fact that adjectives cannot occur in
argument positions. Our account avoids all these problems: (i) bare plurals (as
well as bare mass nouns and more generally all weak indefinites) denote general-
ized quantifiers over amounts; (ii) bare plurals and bare mass nouns denote prop-
erties when they occur in predicate positions but generalized quantifiers over
amounts when they occur in argument positions; (iii) adjectives denote properties
and as such cannot occur in argument positions. Our account of bare plurals (and
bare mass Ns) allows us to explain the contrasts observed by Carlson (1977a, c)
between bare plurals and singular indefinites without assuming Carlson’s view
that existential bare plurals are kind-referring. This hypothesis cannot be correct
for Romance languages (other than Brazilian Portuguese), in which bare NPs can-
not function as names of kinds.
1
Standard DRT (Kamp 1981; Kamp and Reyle 1993) does not distinguish between weak and
strong indefinites and the proposed accounts, summarized here, concern only strong indefinites.
DRT accounts of weak indefinites, which are relatively recent (Kamp and Bende Farkas 2001;
Farkas and de Swart 2003), rely on unification and seem to be DRT implementations of the prop-
erty analysis.
2
This discourse referent can be (i) introduced at the upper level of the DRS, in which case it is
interpreted as existentially quantified; (ii) be introduced in a DRS embedded under the main DRS,
in which case it is indirectly bound and thus analyzed as a quantified DP.
248 Conclusion
What We Have Set Aside
The set of phenomena relevant for the study of indefinites is extremely wide, which
means we were forced to set aside a certain number of issues, such as the relations
between indefinites and negation (negative polarity items), indefinites and interroga-
tive phrases, or the behavior of free-choice indefinites like English any. The
indefiniteness of so-called indefinite pronouns (somebody, something, nobody) would
have also been relevant, as well as the study of relative and interrogative pronouns,
which in some languages seem to exhibit a strong vs. weak distinction. Nevertheless,
we hope that the formal tools and the theoretical distinctions proposed in this book
may contribute to the future investigation of these empirical areas.
Bibliography
Abusch, D. 1994. The scope of indefinites. Natural Language Semantics 2: 83–136.

Arregi, K., Z. Fagyal, S. Montrul, and A. Tremblay, eds. 2010. Romance linguistics 2008.
Interactions in Romance. Selected papers from the 38th linguistic symposium on Romance
Languages (LSRL), Urbana Champaign, April 2008, John Benjamins.
Attal, P. 1976. A propos de l’indéfini des: Problèmes de représentation sémantique. Le français
moderne 44(2): 126–142.
Attal, P. 1994. Indéfinis et structures sémantiques. Faits de langues 2(4): 187–195.
Bach, E., E. Jelinek, A. Kratzer, and B.H. Partee, eds. 1995. Quantification in natural languages.
Dordrecht: Kluwer.
Barwise, J., and R. Cooper. 1981. Generalized quantifiers and natural language. Linguistics and
Philosophy 4(2): 159–219.
Bäuerle, R., U. Egli, and A. von Stechow, eds. 1979. Semantics from different point of view. Berlin:
Springer.
Bäuerle, R., C. Schwartze, and A. von Stechow, eds. 1983. Meaning, use and interpretation of
language. Berlin: De Gruyter.
Beghelli, F. 1995. The phrase-structure of quantifier scope. Ph.D. dissertation, University of
California, Los Angeles.
Beghelli, F., and T. Stowell. 1997. Distributivity and negation: The syntax of each and every. In
Ways of scope taking, ed. A. Szabolcsi, 71–107. Dordrecht: Kluwer.
Ben-Shalom, D. 1993. Object wide scope and semantic trees. Los Angeles: MS, UCLA.
Beyssade, C., and C. Dobrovie-Sorin. 2005. A syntax-based analysis of predication. In Proceedings
of semantics and linguistic theory 15, ed. E. Georgala and J. Howell, 44–61. Ithaca: CLC
Publications.
Beyssade, C., and C. Dobrovie-Sorin. 2009. Predication and identity in copular sentences, ms.
CNRS-LLF and Institut Jean Nicod. Paper presented at LSRL 2009 and glow in Asia 2009.
Beyssade, C., R. Bok-Bennema, F. Drijkoningen, and P. Monachesi, eds. 2002. Romance
languages and linguistic theory 2000. Amsterdam: John Benjamins Publishing Company.
Bleam, T. 1999. Leísta Spanish and the syntax of clitic doubling. Ph.D. dissertation, University of
Delaware.
Borthen, K. 2003. Norwegian bare singulars. Ph.D. thesis, Norwegian University of Science and
Technology.
Bosque, I., ed. 1996. El sustantivo sin determinación. Presencia y ausencia de determinante en la
lengua española. Madrid: Visor.
Bosque, I., and V. Demonte, eds. 1999. Gramática descriptiva de la lengua española. Madrid:
Espasa Calpe.
Bosveld de Smet, L. 1997. On mass and plural quantification. The case of French des/du NPs.
Thèse de doctorat, Gröningen.

250 Bibliography
Bouchard, D. 1998. The distribution and interpretation of adjectives in French: A consequence of

bare phrase structure. Probus 10(2): 139–183.
Büring, D. 1997. The 59th street bridge accent: The meaning of topic and focus. London: Routledge.
Burton-Roberts, N. 1977. Generic sentences and analyticity. Studies in Language 1: 155–196.
Cabredo-Hofherr, P. 2011. Bare singulars and bare habituals. In Genericity, ed. C. Beyssade,
A. Mari, and F. Del Prete. Oxford: Oxford University Press.
Cardinaletti, A., N. Munaro, G. Giusti, and C. Poletto, eds. to appear. Functional heads. New York:
Oxford University Press.
Carlier, A. 1989. Généricité du syntagme nominal sujet et modalités. In Généricité, spécificité
et aspect, ed. M. Wilmet, 33–56. Paris: Editions Duculot.
Carlier, A. 2000. Les articles du et des en synchronie et en diachronie. Revue Romane 35(2):
177–206.
Carlson, G.N. 1977a. A unified analysis of the English bare plural. Linguistics and Philosophy 1:
413–457.
Carlson, G.N. 1977b. Amount relatives. Language 55(3): 520–542.
Carlson, G.N. 1977c. Reference to kinds in English. Ph.D. dissertation, University of
Massachusetts.
Carlson, G.N. 1989. On the semantic composition of English generic sentences. In Properties,
types and meaning, vol. II, ed. G. Chierchia, B.H. Partee, and R. Turner, 167–191. Dordrecht/
Boston/London: Kluwer Academic Publishers.
Carlson, G.N. 2003. Weak indefinites. In From NP to DP: On the syntax and pragma-semantics of
noun phrases, vol. 1, ed. M. Coene and Y. D’Hulst, 195–210. Amsterdam: John Benjamins.
Carlson, G.N., and F.J. Pelletier, eds. 1995. The generic book. Chicago: The University of Chicago Press.
Chierchia, G. 1995a. Individual-level predicates as inherent generics. In The generic book, ed.
G.N. Carlson and F.J. Pelletier, 176–223. Chicago: University of Chicago Press.
Chierchia, G. 1995b. Plurality of mass nouns and the notion of “Semantic Parameter”. Milan: Ms,
University of Milan et DIPSCO.
Chierchia, G. 1995c. Dynamics of meaning, anaphora, presupposition and the theory of grammar.
Chicago/London: The University of Chicago Press.
Chierchia, G. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339–405.
Chierchia, G., B.H. Partee, and R. Turner, eds. 1989. Properties, types and meaning. Dordrecht: Kluwer.
Chung, S., and W. Ladusaw. 2003. Restriction and saturation. Cambridge: MIT Press.
Coene, M., and Y. d’Hulst, eds. 2003. From NP to DP. Amsterdam: John Benjamins Publishing
Company.
Cohen, A. 2001. On the generic use of indefinite singulars. Journal of Semantics 18:183–209.
Cohen, A., and N. Erteschik-Shir. 2002. Topic, focus and the interpretation of bare plurals. Natural
Language Semantics 10: 125–165.
Cole, P., ed. 1981. Radical pragmatics. New York: Academic.
Comorovski, I. 1995. On quantifier strength and partitive noun phrases. In Quantification in natural
languages, ed. E. Bach, E. Jelinek, A. Kratzer, and B.H. Partee, 145–177. Dordrecht: Kulwer
Academic Publishers.
Comorovski, I. 2007. Constituent questions and the copula of specification. In Existence: Semantics
and syntax, ed. I. Comorovski and K. von Heusinger, 49–77. Dordrecht: Springer.
Comorovski, I., and K. von Heusinger, eds. 2007. Existence: Semantics and syntax. Dordrecht:
Springer.
Comorovski, I. 1997. Functional indefinites and the proportion problem. Paris: CSSP.
Comorovski, I. 1998. Topic, adverbial quantification, and backwards anaphora. In G. Bouma, G.-J.
M. Kruij, and R. T. Oehrle (eds.), Proceedings of the Joint Conference on Formal Grammar,
Head Driven Phrase Structure Grammar and Categorial Grammar, Saarbrûcken: Germany,
46–54.
Comorovski, I. 1999. Functional indefinites and the proportion problem. In F. Corblin, C. Dobrovie-
Sorin, and J.M. Marandin (eds.) Empirical Issues in Formal Syntax and Semantics II, La Haye:
Holland Academic Graphics, 17–37.
Bibliography 251
Condoravdi, C. 1992. Weakly and strongly novel noun phrases. In C. Barker, and D. Dowty (eds.),
Proceedings of SALT 2, Working Papers in Linguistics 40, Columbus: Ohio State university,
17–37.
Condoravdi, C. 1994. Descriptions in context. Ph.D. dissertation, Yale.
Condoravdi, C., and G. Renardel, eds. 2001. Logical perspectives on language and information.
Stanford: CSLI.
Cooper, R. 1979. The interpretation of pronouns. In Syntax and semantics: Selections from the
third Groningen round table, ed. F. Heny and H. Schnelle, 61–97. New York: Academic.
Cooper, R. 1983. Quantification and syntactic theory. Dordrecht: Reidel.
Corblin, F. 1987. Indéfini, défini et démonstratif. Constructions linguistiques de la référence.
Genève: Droz.
Corblin, F. 1997. Les indéfinis: Variables et quantificateurs. Langue Française 116: 8–32.
Corblin, F. 2001. Où situer certains dans une typologie sémantique des groupes nominaux? In
Typologie des groupes nominaux, ed. G. Kleiber, B. Laca, and L. Tasmowski, 99–117. Rennes:
Presses Universitaires de Rennes.
Corblin, F. 2002. Représentation du discours et sémantique formelle. Paris: PUF.
Corblin, F., and H. de Swart, eds. 2004. Handbook of French semantics. Stanford: CSLI.
Corblin, F., D. Godard, and J.-M. Marandin, eds. 1997. Empirical issues in formal syntax and
semantics. Bern: Peter Lang Publishers.
Cornilescu, A. 2000. Notes on the interpretation of the prepositional accusative in Romanian, vol. 2,
91–106. Bucharest Working Papers in Linguistics, University of Bucharest.
Danon, G. 2002. Case and formal definiteness: The licensing of definite and indefinite noun phrases
in Hebrew. Ph.D. dissertation, Tel Aviv University.
David, J., and G. Kleiber, eds. 1986. Déterminants: Syntaxe et sémantique. Paris: Klincksieck.
Davidson, D., and G. Harman, eds. 1972. Semantics of natural language. Dordrecht: Reidel.
Dayal, V. 2003. A semantics for pseudo incorporation. Newark: Ms., Rutgers University.
Dayal, V. 2004. Number marking and (in)definiteness in kind terms. Linguistics and Philosophy
27: 393–450.
de Hoop, H. 1992. Case configuration and noun phrase interpretation. Ph.D. dissertation, pub-
lished in 1997 by Garland, New York.
de Swart, H. 1991. Adverbs of quantification: A generalized quantifiers approach. Ph.D. disserta-
tion, Rijksuniversiteit Gröningen.
de Swart, H. 1992. Genericity, conditionals and the weak-strong distinction. In Proceedings of the
Eleventh West Coast Conference on Formal Linguistics, 135–147. Stanford: CSLI.
de Swart, H. 1996. (In)definites and genericity. In Quantifiers, deduction and context, ed.
M. Kanazawa, C. Piñon, and H. de Swart, 171–194. Stanford: CSLI Publications.
de Swart, H., and D. Farkas. 2005. Généricité et (in)définitude. Une analyse dans la théorie de
l’optimalité. In C. Dobrovie-Sorin (ed.). Noms nus et généricité, Paris: Presses Universitaires
de Vincennes, 97–126.
de Swart, H., Y. Winter, and J. Zwarts. 2007. Bare nominals and reference to capacities. Natural
Language and Linguistic Theory 25: 195–222.
Declerck, R. 1987. A puzzle about generics. Folia Linguistica 21: 143–153.
Dekker, P. 2004. The pragmatic dimension of indefinites, Research on Language and Computation
2(3): 365–399.
Delfitto, D. 2002. Genericity in language. Issues of syntax, logical form and interpretation.
Alessandria: Dell’Orso.
den Dikken, M. 2006. Relators and linkers: The syntax of predication, predicate inversion and
copulas. Cambridge: MIT Press.
Diesing, M. 1992. Indefinites. Cambridge: MIT Press.
Diesing, M., and E. Jelinek. 1996. Distributing arguments. Natural Language Semantics 3(2):
123–176.
Dobrovie-Sorin, C. 1987. Syntaxe du roumain. Chaînes thématiques. Thèse de doctorat, Université
Paris 7.
252 Bibliography
Dobrovie-Sorin, C. 1990. Clitic doubling, wh-movement and quantification in Romanian.

Linguistic Inquiry 21(3): 351–397.
Dobrovie-Sorin, C. 1993. The syntax of Romanian: Comparative studies in Romance. Berlin:
Mouton de Gruyter.
Dobrovie-Sorin, C. 1995. On the denotation and scope of indefinites, vol. 5, 67–114. Venice
Working Papers in Linguistics, University of Venice.
Dobrovie-Sorin, C. 1997a. Types of predicates and the representation of existential readings. In
Proceedings of SALT VII, ed. A. Lawson. Ithaca/New York: Cornell University Press,
117–134.
Dobrovie-Sorin, C. 1997b. Classes de prédicats, distribution des indéfinis et la distinction thétique-
catégorique. Le Gré des Langues 12: 58–97.
Dobrovie-Sorin, C. 2003. Adverbs of quantification and genericity. In Proceedings of CSSP, Paris,
2001; electronic version in Y.N. Falk, ed. Proceedings of Israel Association of Theoretical
Linguistics 17, Jerusalem: The Hebrew University of Jerusalem, Hebrew University of
Jerusalem English Department. http://linguistics.huji.ac.il/IATL/17.
Dobrovie-Sorin, C. 2004. Generic plural indefinites and (in)direct binding. In Corblin, F. and H. de
Swart (eds.), Handbook of French semantics, Stanford: CSLI, 55–70.
Dobrovie-Sorin, C., ed. 2005. Noms nus et généricité. Paris: Presses Universitaires de Vincennes.
Dobrovie-Sorin, C. 2007. Existential bare plurals: From properties back to entities. Lingua 117:
296–313.
Dobrovie-Sorin, C. 2010. Number neutral amounts and pluralities in Brazilian Portuguese. Journal
of Portuguese Linguistics 9(1): 53–74.
Dobrovie-Sorin, C. to appear a. Generic plural and mass indefinites. In Mari, A., C. Beyssade, and
F. Del Prete (eds.). Oxford University Press.
Dobrovie-Sorin, C. to appear b. Number as a feature. In Cardinaletti, A., N. Munaro, G. Giusti, and
C. Poletto (eds.). Oxford University Press.
Dobrovie-Sorin, C., and C. Beyssade. 2004. Définir les indéfinis. Paris: Editions du CNRS.
Dobrovie-Sorin, C., and B. Laca. 1996. Generic bare NPs. manuscript. University of Paris 7/
Strasbourg 2.
Dobrovie-Sorin, C., and B. Laca. 1999. La généricité entre la référence à l’espèce et la quantification
générique. In Langues et grammaire III. Syntaxe, ed. P. Sauzet, 163–177. Paris: Presses
Universitaires de Vincennes.
Dobrovie-Sorin, C., and B. Laca. 2003. Les noms sans déterminant dans les langues romanes. In
Les langues romanes, ed. D. Godard, 235–281. Paris: Ed. du CNRS.
Dobrovie-Sorin, C., and B. Laca. 2010. Bare nouns in the romance languages. In Fundamental
issues in the Romance languages, ed. D. Godard, 221–262. Stanford: CSLI.
Dobrovie-Sorin C., and A. Mari. 2007a. Generic plural indefinites: Sums or groups? In E. Elfner
and M. Walkow (eds.) Proceedings of NELS 37. Amherst: GLSA, 205–218.
Dobrovie-Sorin, C., and A. Mari. 2007b. Constraints on quantificational domains: Generic plural
des-indefinites in French. In Proceedings of Sinn und Bedeutung 11, ed. E. Puig-Waldmüller,
165–179. Barcelona: University of Barcelona.
Dobrovie-Sorin, C., and R. de Pires Oliveira. 2008. Reference to kinds in Brazilian Portuguese:
Definite singulars vs bare singulars. In Proceedings of Sinn und Bedeutung 12, ed. A. Grønn,
107–121. Oslo: Department of Literature, Area Studies and European Languages, University
of Oslo.
Dobrovie-Sorin, C., and R. Pires de Oliveira. 2010. Generic bare singulars in Brazilian Portuguese.
In Arregi, K., Z. Fagyal, S. Montrul & A. Tremblay (eds.), Romance linguistics 2008. Inter-
actions in Romance. Selected papers from the 38th linguistic symposium on Romance
Languages (LSRL), Urbana Champaign, April 2008, John Benjamins, 203–216.
Dobrovie-Sorin, C., T. Bleam, and M.T. Espinal. 2005. Noms nus, nombre et types d’incorporation.
In Noms nus et généricité, ed. C. Dobrovie-Sorin, 129–157. Paris: Presses Universitaires de
Vincennes.
Dobrovie-Sorin, C., T. Bleam, and M.T. Espinal. 2006. Bare nouns, number and types of incorpo-
ration. In Non-definiteness and plurality, ed. S. Vogeleer and L. Tasmowski, 51–79. Amsterdam:
John Benjamins.
Bibliography 253
Dölling, J., T. Heyde-Zybatow, and M. Schäfer, eds. 2007. Event structures in linguistic form and
interpretation. Berlin/New York: Mouton de Gruyter.
Dowty, D. 1979. Word meaning and Montague grammar. Dordrecht: Reidel.
Dowty, D., R. Wall, and S. Peters. 1981. Introduction to montague semantics. Dordrecht: Reidel.
Egli, U. 1979. The Stoïc concept of anaphora. In R. Bäuerle, U. Egli, and A. von Stechow (eds.),
Semantics from different point of view. Berlin: Springer, 266–283.
Elfner, E., and M. Walkow, eds. 2007. Proceedings of NELS 37. Amherst: GLSA.
Enç, M. 1991. The semantics of specificity. Linguistic Inquiry 22(1): 1–25.
Engdahl, E. 1986. Constituent questions. Dordrecht: Reidel.
Espinal, M.T., and L. McNally. 2011. Bare singular nouns and incorporating verbs in Spanish and
Catalan. Journal of Linguistics 47: 87–128.
Evans, E. 1980. Pronouns. Linguistic Inquiry 11: 337–362.
Farkas, D. 1978. Direct and indirect object reduplication in Romanian. Papers from the fourteenth
regional meeting, CLS, Chicago Linguistic Society, University of Chicago, Chicago, IL.
Farkas, D. 1981. Quantifier scope and syntactic islands. In Proceedings of the Chicago linguistic
society, vol. 7, 59–66. Chicago: Chicago Linguistic Society, University of Chicago.
Farkas, D. 1985. Intensional descriptions and the Romance subjunctive. New York: Garland.
Farkas, D. 1994. Specificity and Scope in L. Nash and G. Tsoulas (eds.), proceedings of Langues
et Grammaire 1, 119–137.
Farkas, D. 1997a. Dependent Indefinites. In Empirical issues in formal syntax and semantics, ed.
F. Corblin, D. Godard, and J.-M. Marandin, 243–268. Berne: Peter Lang Publishers.
Farkas, D. 1997b. Evaluation indices and scope. In Ways of scope taking, ed. A. Szabolcsi, 183–215.
Dordrecht: Kluwer Academic Publishers.
Farkas, D. 2001. Dependent indefinites and direct scope. In Logical perspectives on language and
information, ed. C. Condoravdi and G. Renardel, 41–72. Stanford: CSLI Publications.
Farkas, D., and H. de Swart. 2003. The semantics of incorporation: From argument structure to
discourse transparency. Stanford: CSLI.
Farkas, D., and H. de Swart. 2007. Article choice in plural generics. Lingua 117: 1657–1676.
Farkas, D., and A. Giannakidou. 1996. How clause-bounded is the scope of universals? In
Proceedings of SALT VI, ed. T. Galloway and J. Spence et al., 35–52. Ithaca: CLC Publications,
Cornell University.
Farkas, D., and Y. Sugioka. 1983. Restrictive if/when clauses. Linguistics and Philosophy 6(2):
225–258.
Fodor, J.D., and I.A. Sag. 1982. Referential and quantificational indefinites. Linguistics and
Philosophy 5: 355–398.
Fradin, B. 1988. Approche des constructions à détachement. La reprise interne. Langue Française
78: 26–56.
Frege, G. 1971. Ecrits logiques et philosophiques. Traduction C. Imbert. Paris: Le Seuil.
Gabbay, D., and F. Guenther, eds. 1989. Handbook of philosophical logic. Dordrecht: Reidel.
Galloway, T., and J. Spence, eds. 1996. Proceedings from semantics and linguistic theory (SALT)
6. Ithaca: CLC Publications, Cornell University.
Galmiche, M. 1986. Référence indéfinie, événements, propriétés et pertinence. In Déterminants:
syntaxe et sémantique, ed. J. David and G. Kleiber, 41–71. Paris: Librairie Klincksieck.
Gamut, L.T.F. 1991. Logic, language and meaning. Chicago: The University of Chicago Press.
Gardenfors, P., ed. 1987. Generalized quantifiers: Linguistic and logical approaches. Dordrecht:
Kluwer.
Geach, P.T. 1962. Reference and generality. Ithaca/New York: Cornell University Press.
Georgala, E., and J. Howell, eds. 2005. Proceedings of semantics and linguistic theory 15. Ithaca/
New York: CLC Publications.
Gerstner, C., and M. Krifka. 1993. Genericity. In Handbuch der Syntax, ed. J. Jacobs, A. von
Stechow, W. Sterne, and T. Vennemann, 966–978. Berlin: Walter de Gruyter.
Gillon, B.S. 1992. Towards a common semantics for English count and mass nouns. Linguistics
and Philosophy 15: 597–639.
Gillon, B.S. 1996. Collectivity and distributivity internal to English noun phrases. Language
Sciences 18(1–2): 443–468.
254 Bibliography
Giry-Schneider, J. 1988. L’interprétation événementielle des phrases en il y a. Linguisticae

Investigationes XII(1): 85–100.
Glasbey, S. 1998. Progressives, states, and backgrounding. In Events and grammar, ed.
S. Rothstein, 105–124. Dordrecht: Kluwer Academic Publishers.
Godard, D., ed. 2003. Les langues romanes. Problèmes de la phrase simple. Paris: CNRS
Editions.
Godard Danièle, ed. 2010. Fundamental issues in the Romance languages. English translation of
Godard Danièle (ed.) 2003, Stanford: CSLI.
Godard, D., and J. Jayez. 1993. Towards a proper treatment of coercion phenomena. 6th EACL,
168–177.
Greenberg, Y. 2007. Exceptions to generics: Where vagueness, context dependence and modality
interact. Journal of Semantics 24(2): 131–167.
Groenendijk, J., and M. Stokhof. 1984. Studies on the semantics of questions and the pragmatics
of answers. Ph.D. dissertation, University of Amsterdam.
Groenendijk, J., T. Janssen, and M. Stokhof, eds. 1981. Formal methods in the study of language.
Amsterdam: Mathematical Centre.
Groenendijk, J., D. de Jongh, and M. Stokhof, eds. 1987. Studies in discourse representation
theory and the theory of generalized quantifiers. Dordrecht: Foris.
Groenendijk, J., M. Stokhof, and F. Veltman. 1996. Coreference and modality. In The handbook of
contemporary semantic theory, ed. S. Lappin, 179–213. Oxford: Blackwell Publishers.
Grønn A. ed.. 2008. Proceedings of Sinn und Bedeutung 12. Oslo: Department of Literature, Area
Studies and European Languages, University of Oslo.
Gupta, A. 1980. The logic of common nouns. New Haven: Yale University Press.
Haïk, I. 1984. Indirect binding. Linguistic Inquiry 15(2): 185–223.
Harvey, M., and L. Santelmann, eds. 1994. Proceedings of SALT IV. Ithaca: CLC Publications,
Cornell University.
Heim, I. 1982. The semantics of definite and indefinite noun phrases. Ph.D. dissertation, University
of Massachusetts, Amherst. Published in 1988, Garland, New York.
Heim, I. 1987. Where does the definiteness restriction apply? Evidence from the definiteness of
variables. In The representation of (in)definiteness, ed. E. Reuland and A. ter Meulen, 21–42.
Cambridge: MIT Press.
Heim, I. 1990. E-type pronouns and donkey anaphora. Linguistics and Philosophy 13: 137–178.
Heim, I., and A. Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell.
Heny, F., and H. Schelle, eds. 1979. Syntax and semantics, vol. 10. New York: Academic.
Heycock, C., and A. Kroch. 1999. Pseudocleft connectedness: Implications for the LF interface
level. Linguistic Inquiry 30: 365–397.
Heyd, S. 2002. Prédication et interprétation générique des syntagmes nominaux en des en position
sujet. Communication au colloque Indéfinis et prédication en français. Paris: Sorbonne.
Higginbotham, J. 1985. On semantics. Linguistic Inquiry 16(4): 547–593.
Higginbotham, J. 1987. Indefiniteness and predication. In The representation of (in)definiteness,
ed. E. Reuland and A. ter Meulen, 43–70. Cambridge: MIT Press.
Higginbotham, J. 1994. Mass and count quantifiers. Linguistics and Philosophy 17: 447–480.
Higgins, R.F. 1979. The pseudocleft construction in English. New York: Garland.
Hintikka, J. 1974. Quantifiers vs quantification theory. Linguistic Inquiry 5: 153–177.
Hintikka, J. 1986. The semantics of a certain. Linguistic Inquiry 17(2): 331–336.
Hintikka, J., J. Moravcsik, and P. Suppes, eds. 1973. Approaches to natural language. Dordrecht/
Holland: Reidel.
Jackendoff, R. 1974. An introduction to the X-bar convention. Bloomington: Ms, Indiana University
Linguistics Club.
Jacobs, J., A. von Stechow, W. Sterne, and T. Vennemann, eds. 1993. Handbuch der Syntax. Berlin:
Walter de Gruyter.
Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris.
Kadmon, N. 1987. On unique and non-unique reference and asymmetric quantification. Ph.D.
dissertation, Amherst.
Bibliography 255
Kallulli, D. 1999. The comparative syntax of Albanian. On the contribution of syntactic types to
propositional interpretation. Ph.D. thesis. University of Durham.
Kamp, H. 1981. A theory of truth and discourse representation. In Formal methods in the study of
language, ed. J. Groenendijk, T. Janssen, and M. Stokhof, 277–322. Amsterdam: Mathematisch
Centrum.
Kamp, H., and A. Bende-Farkas. 2001. Indefinites and binding: From specificity to incorporation.
Lecture notes from the 13th ESSLI summer school in Helsinki. Ms, Stuttgart University.
Kamp, H., and U. Reyle. 1993. From discourse to logic. Dordrecht: Kluwer.
Kanazawa, M., C. Piñon, and H. de Swart, eds. 1996. Quantifiers, deduction, and context. Stanford:
CSLI.
Kanouse, D.E. 1972. Verbs as implicit quantifiers. Journal of Verbal Learning and Verbal Behavior
11: 141–147.
Keenan, E.L., ed. 1975. Formal semantics of natural language. Cambridge: Cambridge University
Press.
Keenan, E.L. 1987. A semantic definition of indefinite NP. In The representation of (in)definiteness,
ed. E. Reuland and A. ter Meulen, 287–317. Cambridge: MIT Press.
Keenan, E.L. 1996. The semantics of determiners. In The handbook of contemporary semantic
theory, ed. S. Lappin, 42–63. Oxford: Blackwell Publishers.
Keenan, E.L., and J. Stavi. 1986. A semantic characterization of natural language determiners.
Linguistics and Philosophy 9: 253–326.
Kennedy, C. 1999. Gradable adjectives denote measure functions, not partial functions. Studies in
the Linguistic Sciences 29(1): 65–80.
Kiss, K. 1994. Generic and existential bare plurals and the classification of predicates. Working
Papers in the theory of grammar, vol. 1. Budapest University (ELTE), Budapest.
Kleiber, G. 1981a. Problèmes de référence. Descriptions définies et noms propres. Paris:
Klincksieck.
Kleiber, G. 1981b. Relatives spécifiantes et relatives non-spécifiantes. Le français moderne 49:
216–233.
Kleiber, G. 2001. Lecture existentielle et lecture partitive. In Typologie des groupes nominaux, ed.
G. Kleiber, B. Laca, and L. Tasmowski, 47–99. Rennes: Presses Universitaires de Rennes.
Kleiber, G., B. Laca, and L. Tasmowski, eds. 2001. Typologie des groupes nominaux. Rennes:
Presses Universitaires de Rennes.
Kratzer, A. 1988. Stage-level and individual-level predicates. In M. Krifka (ed.), Genericity in
Natural Language. Proceedings of the 1988 Tübingen Conference. SNS-Bericht 88–42.
University of Tübingen, 247–284.
Kratzer, A. 1990. How Specific is a Fact. In Proceedings of the Conference on Theories of Partial
Information, Center for Cognitive Science, University of Texas at Austin.
Kratzer, A. 1995. Stage-level and individual-level predicates. In The generic book, ed.
Kratzer, A. 1996. Severing the external argument from its verb. In Phrase structure and the
lexicon, ed. J. Rooryck and L.A. Zaring, 109–137. Dordrecht/Boston: Kluwer Academic
Publishers.
Kratzer, A. 1998. Scope or pseudoscope? Are there widescope indefinites? In Events and grammar,
ed. S. Rothstein, 163–196. Dordrecht/Boston: Kluwer Academic Publishers.
Kratzer, A. 2007. On the plurality of verbs. In Event structures in linguistic form and interpretation,
ed. J. Dölling, T. Heyde-Zybatow, and M. Schäfer, 269–300. Berlin: Walter de Gruyter.
Krifka, M. 1987. An outline of generics. SNS-Bericht 87-125. University of Tübingen.
Krifka, M., ed. 1988a. Genericity in natural language. In Proceedings of the 1988 Tübingen
Conference. SNS-Bericht 88-42. University of Tübingen.
Krifka, M. 1988b. The relational theory of genericity. In Genericity in natural language, ed.
M. Krifka, 285–312. Tübingen: Seminar für natürlich-sprachliche Systeme d. Univ. Tübingen.
Krifka, M. 1992a. Definite NPs aren’t quantifiers. Linguistic Inquiry 23(1): 156–164.
Krifka, M. 1992b. Thematic relations as links between nominal reference and temporal constitu-
tion. In Lexical matters, ed. I. Sag and A. Szabolsci, 29–53. Stanford: CSLI.
256 Bibliography
Krifka, M. 1995. Focus and the interpretation of generic sentences. In The generic book, ed.
Krifka, M. 1996. Pragmatic strengthening in donkey sentences and plural predications. In
Proceedings from SALT VI, ed. T. Galloway and J. Spence, 136–153. Ithaca: CLC Publications,
Cornell University.
Krifka, M. 1999. At least some determiners aren’t determiners. In The semantics/pragmatics
interface from different points of view, ed. K. Turner, 257–291. Oxford: Elsevier.
Krifka, M., F.J. Pelletier, G.N. Carlson, A. ter Meulen, G. Link, and G. Chierchia. 1995. Genericity:
An introduction. In The generic book, ed. G.N. Carlson and F.J. Pelletier, 1–124. Chicago:
University of Chicago Press.
Kupferman, L. 1979. Les Constructions Il est un médecin/C’est un médecin: Essai de solution.
Cahiers linguistiques 9: 131–164.
Laca, B. 1990. Generic objects: Some more pieces of the puzzle. Lingua 81: 25–46.
Laca, B. 1996. Acerca de la semántica de los plurales escuetos en español. In El sustantivo sin
determinación: la ausencia de determinante en la lengua española, ed. I. Bosque, 241–268.
Madrid: Visor Libros.
Laca, B. 1999. Presencia y Ausencia de Determinante. In Nueva gramática descriptiva de la lengua
española, ed. I. Bosque and V. Demonte, 891–928. Madrid: Espasa Calpe.
Laca, B., and L. Tasmowski. 1994. Le pluriel indéfini de l’attribut métaphorique. Lingvisticæ
Investigationes 18(1): 27–47.
Laca, B., and L. Tasmowski-de Ryck. 1996. Indéfini et quantification. Recherches linguistiques de
Vincennes 25: 107–128.
Ladusaw, W. 1994. Thetic and categorical, stage and individual, weak and strong. In Proceedings
of semantics and linguistic theory IV, ed. M. Harvey and L. Santelmann, 220–229. Ithaca: CLC
Publications, Cornell University.
Lambrecht, K. 2002. Topic, focus and secondary predication. The French presentational relative
construction. In Romance languages and linguistic theory 2000, ed. C. Beyssade, R. Bok-
Bennema, F. Drijkoningen, and P. Monachesi, 171–212. Amsterdam: John Benjamins.
Landman, F. 1989a. Groups I. Linguistics and Philosophy 12(5): 559–605.
Landman, F. 1989b. Groups II. Linguistics and Philosophy 12(6): 723–744.
Landman, F. 2000. Events and plurality. Studies in linguistics and philosophy. Dordrecht: Kluwer.
Landman, F. 2003. Predicates-argument mismatches and the adjectival theory of indefinites. In The
syntax and semantics of noun phrases, vol. 1, ed. M. Coene and Y. d’Hulst, 211–237.
Amsterdam: John Benjamins.
Lappin, S., ed. 1996. The handbook of contemporary semantic theory. Oxford: Blackwell.
Lappin, S., and N. Francez. 1994. E-type pronouns, I-sums and donkey. Linguistics and Philosophy
17: 391–428.
Lawler, J. 1972. Generic to a fault. Papers from the 8th regional meeting, Chicago Linguistic Society.
Lawler, J. 1973. Studies in English generics. University of Michigan papers in linguistics I:1. Ann
Arbor: University of Michigan Press.
Lawson, A., ed. 1997. Proceedings of SALT VII. Ithaca/New York: CLC Publications.
Lepore, E., ed. 1987. New directions in semantics. London: Academic.
Lewis, D. 1975. Adverbs of quantification. In Formal semantics of natural language, ed.
E.L. Keenan, 3–15. Cambridge: Cambridge University Press.
Link, G. 1983. The logical analysis of plurals and mass terms: A lattice theoretic approach. In
Meaning, use and interpretation of language, ed. R. Bauerle, C. Schwartze, and A. von Stechow,
302–323. Berlin: Walter de Gruyter.
Liu, F.H. 1990. Scope and dependency in English and Chinese. Ph.D. dissertation, University of
California, Los Angeles.
Longobardi, G. 1994. Reference and proper names. Linguistic Inquiry 25(4): 609–665.
Longobardi, G. 2002. How comparative is semantics? A unified parametric theory of bare nouns
and proper names. Natural Language Semantics 9: 335–369.
Lønning, J.T. 1987. Mass terms and quantification. Linguistics and Philosophy 10: 1–52.
Bibliography 257
Lopez Palma, H. 2007. Plural indefinite descriptions with unos and the interpretation of number.
Probus 19: 235–266.
Lumsden, M. 1988. Existential sentences: Their structure and meaning. London: Croom Helm.
Maier, E., C. Bary, and J. Huitink, eds. 2004. Proceedings of Sinn & Bedeutung 9. Nijmegen:
NCS.
Mari, A., C. Beyssade, and F. Del Prete, eds. to appear. Genericity: New perspectives. Oxford:
Oxford University Press.
Matushansky, O., and B. Spector. 2004. Tinker, tailor, soldier, spy. In Proceedings of SuB9, ed.
E. Maier, C. Bary, and J. Huitink, 241–255. Nijmegen: Nijmegen Centre of Semantics.
May, R. 1977. The grammar of quantification. Ph.D. dissertation, MIT.
May, R. 1985. Logical form: Its structure and derivation. Cambridge: MIT Press.
McCawley, J.D. 1972. A program for logic. In Semantics of natural language, ed. D. Davidson and
G. Harman, 498–544. Dordrecht: Reidel.
McNally, L. 1995a. Bare plurals in Spanish are interpreted as properties. In G. Morrill & R. Oehrle
(eds.), Proceedings of the 1995 ESSLLI Conference on formal grammar, Barcelona: Spain,
197–222.
McNally, L. 1995b. Stativity and theticity. Columbus: Ms, Center for Cognitive Science, Ohio
State University.
McNally, L. 1998. Existential sentences without existential quantification. Linguistics and
McNally, L., and V. van Geenhoven. 1998. Redefining the weak/strong distinction. Ms, Universitat
Pompeu Fabra & Max Planck Institut Nijmegen, Barcelona/Nijmegen.
Milner, J.-C. 1982. Ordres et raisons de langue. Paris: Le Seuil.
Milner, J.-C. 1989. Elements pour une science du langage. Paris: Le Seuil.
Milsark, G. 1977. Towards the explanation of certain peculiarities of existential sentences in
English. Linguistic Analysis 3: 1–29.
Moeschler, J., and M.J. Béguelin, eds. 2000. Référence temporelle et nominale. Berne: Peter
Lang.
Moltmann, F. 1997. Parts and wholes in semantics. New York: Oxford University Press.
Moltmann, F. 1998. Part structures, integrity, and the mass-count distinction. Synthese 116:
75–111.
Montague, R. 1974. Formal philosophy. Selected papers of Richard Montague, ed. R.H. Thomason.
New Haven: Yale University Press.
Moravcsik, J. 1973. Mass terms in English. In Approaches to natural language, ed. J. Hintikka, J.
Moravcsik, and P. Suppes, 263–285. Dordrecht: Reidel.
Moro, A. 1997. The raising of predicates: Predicative noun phrases and the theory of clause
structure. Cambridge: CUP.
Morrill, G., and R. Oehrle, eds. 1995. Proceedings of the 1995 ESSLLI conference on formal
grammar.
Mostowski, A. 1957. On a generalization of quantifiers. Fundamenta Mathematicae 44: 12–36.
Müller, A.P. 2002. The semantics of generic quantification in Brazilian Portuguese. Probus 14:
279–298.
Munn, A., and C. Schmitt. 2005. Number and indefinites. Lingua 115: 821–855.
Musan, R. 1997. Tense, predicates, and lifetime effects. Natural Language Semantics 5(3):
271–301.
Parsons, T. 1990. Events in the semantics of English. Cambridge: MIT Press.
Partee, B.H. 1984. Nominal and temporal anaphora. Linguistics and Philosophy 7: 243–286.
Partee, B.H. 1987. Noun phrase interpretation and type-shifting principles. In Studies in discourse
representation theory and the theory of generalized quantifiers, ed. J. Groenendijk, D. de
Jongh, and M. Stokhof, 115–144. Dordrecht: Foris Publications.
Partee, B.H., and M. Rooth. 1983. Generalized conjunction and type ambiguity. In Meaning, use,
and interpretation of language, ed. R. Bauerle, C. Schwartze, and A. von Stechow, 115–143.
Berlin: Walter de Gruyter.
258 Bibliography
Pesetsky, D. 1987. Wh-in-situ: Movement and unselective binding. In The representation of (in)
definiteness, ed. E. Reuland and A. ter Meulen, 98–130. Cambridge: MIT Press.
Postal, P.M. 1971. Cross-over phenomena. New York: Holt, Rinehart and Winston.
Prince, E. 1981. Towards a taxonomy of given-new information. In Radical pragmatics, ed.
P. Cole, 223–256. New York: Academic.
Puig-Waldmüller, E., ed. 2007. Proceedings of Sinn und Bedeutung 11. Barcelona: Universitat
Pompeu Fabra.
Rando, E., and D.J. Napoli. 1978. Definiteness in there-sentences. Language 54: 300–313.
Reinhart, T. 1987. Specifier and operator binding. In The representation of (in)definiteness, ed.
E. Reuland and A. ter Meulen, 130–167. Cambridge: MIT Press.
Reinhart, T. 1995. Interface strategies. OTS Working Papers, Utrecht.
Reinhart, T. 1997a. Quantifier scope. How labor is divided between QR and choice functions.
Linguistics and Philosophy 20: 335–397.
Reinhart, T. 1997b. Wh-in-situ in the framework of the minimalist program. Natural Language
Semantics 6(1): 29–56.
Renzi, L. 1988. L’articolo. In L. Renzi, G. Salvi, and A. Cardinaletti (eds.), Grande grammatica
italiana di conzultatione, vol. 1, chap. 7, Bologna: Il Mulino, 357–423.
Reuland, E., and A. ter Meulen, eds. 1987. The representation of (in)definiteness. Cambridge: MIT
Press.
Roberts, G. 1989. Modal subordination and pronominal anaphora in discourse. Linguistics and
Roeper, P. 1983. Semantics for mass terms with quantifiers. Noûs 17: 251–265.
Rooryck, J., and L.A. Zaring, eds. 1996. Phrase structure and the lexicon. Dordrecht: Kluwer
Academic Publishers.
Rooth, M. 1985. Association with focus. Ph.D. dissertation, University of Massachusetts, Amherst.
Rooth, M. 1987. Noun phrase interpretation in Montague grammar, file change semantics, and situation
semantics. In Generalized quantifiers, ed. P. Gardenfors, 237–268. Dordrecht: Reidel.
Rooth, M. 1995. Indefinites, adverbs of quantification and focus semantics. In The generic book,
ed. G.N. Carlson and F.J. Pelletier, 265–299. Chicago: University of Chicago Press.
Rothstein, S., ed. 1998. Events and grammar. Dordrecht: Kluwer.
Roy, I. 2001. Predicate nominals in French. Los Angeles: Ms, University of Southern California.
Roy, I. 2006. Non verbal predications: A syntactic analysis of copular sentences, Ph.D. disserta-
tion, University of Southern California.
Russell, B. 1905. On denoting. Logic and Knowledge 1956: 39–56.
Ruys, E. 1992. The scope of indefinites. Ph.D. dissertation, University of Utrecht.
Safir, K. 1982. Syntactic chains and the definiteness effect. Ph.D. dissertation, MIT, Cambridge,
MA.
Safir, K. 1987. What explains the definiteness effect? In The representation of (in)definiteness, ed.
E. Reuland and A. ter Meulen. Cambridge: MIT Press.
Sag, I., and A. Szabolsci, eds. 1992. Lexical matter. Stanford: CSLI.
Sasse, H.-J. 1987. The thetic/categorical distinction revisited. Linguistics 25: 511–580.
Sauzet, P. ed. 1999. Langues et Grammaire 3, Syntaxe: Communications présentées au colloque
Langues et Grammaire III (juin 1997). Saint-Denis: Université de Paris 8.
Scha, R. 1981. Distributive, collective, and cumulative quantification. In Formal methods in the
study of language, ed. J. Groenendijk, T. Janssen, and M. Stokhof, 483–512. Amsterdam:
Mathematisch Centrum.
Schmitt, C., and A. Munn 1999. Against the nominal mapping parameter: Bare nouns in Brazilian
Portuguese. In Proceedings of NELS 29, University of Massachussetts, Amherst.
Schubert, L.K., and J.F. Pelletier. 1987. Problems in the representation of the logical form of
generics, plurals, and mass nouns. In New directions in semantics, ed. E. Lepore, 385–451.
London: Academic.
Schubert, L.K., and J.F. Pelletier. 1988. An outlook on generic statements. In Proceedings of the
Tübingen conference on generics SNS-Bericht 88-42, ed. M. Krifka, 357–372. Tübingen:
Universität Tübingen.
Bibliography 259
Schwarzschild, R. 1991. On the meaning of definite plural NPs. Ph.D. dissertation, Amherst.
Schwarzschild, R. 1992. Types of plural individuals. Linguistics and Philosophy 15: 641–675.
Simons, P. 1987. Parts. A study in ontology. Oxford: Clarendon.
Steedman, M. 2003. Scope alternation and the syntax/semantics interface. Paris: CSSP.
Steedman, M. 2006. Surface-compositional scope-alternation without existential quantifiers. Ms
available on http://homepages.inf.ed.ac.uk/steedman/papers.html.
Strawson, P.F. 1950. On referring. Mind, 320–344. Translation in Etudes de logique et de linguistique,
Seuil, 1978.
Strawson, P.F. 1959. Individuals: An essay in descriptive metaphysics. London: Routledge.
Szabolcsi, A., ed. 1997. Ways of scope taking. Dordrecht: Kluwer.
Tasmowski, L., and B. Laca. 2000. Le pluriel indéfini et les référents de discours. In J. Moeschler
and M. J. Béguelin (eds.), Référence temporelle et nominale, Bern: Peter Lang, 191–207.
Thomason, R.H. 1974. Formal philosophy: Selected papers of Richard Montague. New Haven:
Yale University Press.
Turner, K., ed. 1999. The semantics/pragmatics interface from different points of view. Current
research in the semantics/pragmatics interface, vol. 1. Oxford: Elsevier.
Ulrich, M. 1985. Thetisch und Kategorisch, Romanica Monacensia 24, Tübingen: Narr.
Vallduvi, E. 1992. The informational component. New York: Garland.
van Benthem, J., and A. ter Meulen, eds. 1997. Handbook of logic and language. Cambridge: MIT
Press.
van Eijck, J., and H. Kamp. 1997. Representing discourse in context. In Handbook of logic and
language, ed. J. van Benthem and A. ter Meulen, 179–237. Amsterdam: Elsevier.
van Geenhoven, V. 1996. Semantic incorporation and indefinite descriptions: Semantic and syntactic
aspects of noun incorporation in West Greenlandic. Ph.D. dissertation, Tübingen. Published in
1998 by CSLI.
van Geenhoven, V. 2004. For-adverbials, frequentative aspect, and pluractionality. Natural
Language Semantics 12: 135–190.
van Geenhoven, V. 2005. Atelicity, pluractionality, and adverbial quantification. In Perspectives on
aspect, ed. H. Verkuyl, H. De Swart, and A. Van Hout, 107–125. Dordrecht: Springer.
Verkuyl, H., H. De Swart, and A. Van Hout, eds. 2005. Perspectives on aspect. Berlin: Springer.
Villalta, E. 1995. Plural indefinites in Spanish and distributivity. Talk given at Going Romance
1994.
Vogeleer, S., and L. Tasmowski, eds. 2006. Non-definiteness and plurality, Linguistik Aktuell/
linguistics today series. Amsterdam: Benjamins.
von Fintel, K. 1994. Restrictions on quantifier domains. Ph.D. dissertation, Amherst.
von Heusinger, K. 2002. The Cross-Linguistic Implementations of Specificity. In: K. Jaszczolt &
K. Turner (eds.), Meaning through Language Contrast. Vol. 2. Amsterdam: Benjamins,
405–421.
von Heusinger, K. 2011. Specificity. In: K. von Heusinger & C. Maienborn & P. Portner (eds.),
Semantics: An International Handbook of Natural Language Meaning. Vol 2. Berlin: de
Gruyter, 1024–1057.
Wasow, T. 1972. Anaphoric relations in English. Ph.D. dissertation, MIT.
Wasow, T. 1979. Anaphora in generative grammar. E. Story-Scientia, Ghent.
Westerstähl, D. 1985. Logical constants in quantifier languages. Linguistics and Philosophy 8:
387–413.
Westerstähl, D. 1989. Quantifiers in formal and natural languages. In Handbook of philosophical
logic, vol. 4, ed. D. Gabbay and F. Guenther, 1–131. Dordrecht: Reidel.
Wilkinson, K. 1991. Studies in the semantics of generic noun phrases. Ph.D. dissertation, University
of Massachusetts, Amherst.
Wilkinson, K. 1995. The semantics of the common noun kind. In The generic book, ed.
Wilmet, M., ed. 1989. Généricité, spécificité et aspect. Paris: Duculot.
Winter, Y. 1997. Choice functions and the scopal semantics of indefinites. Linguistics and
Author Index
A Corblin, F., 144, 152, 154, 173, 174, 205,

Abusch, D., 28, 174, 175 209, 214
Attal, P., 26, 71, 144 Cornilescu, A., 155
B D
Barwise, J., xiv, 7–9, 19, 127 Danon, G., 161
Beghelli, F., 28, 178 Dayal, V., 41, 50, 56–58, 219
Bende-Farkas, A., 247 Declerck, R., 164
Ben-Shalom, D., 178, 181 de Hoop, H., 155
Beyssade, C., 18, 29, 34, 52, 76, 78–80, Dekker, P., 187, 233
83, 88, 89 Delfitto, D., 52, 216
Bleam, T., 34, 41, 46, 50, 57, 155, 159 den Dikken, M., 79, 82
Borthen, K., 41 de Pires Oliveira, R., 33
Bosveld de Smet, L., 71, 147 de Swart, H., 32, 41, 50, 56, 75, 90, 92, 191,
Bouchard, D., 37 193, 199, 217, 247
Büring, D., 166 Diesing, M., 11, 13, 17, 34, 58, 129–131, 139,
Burton-Roberts, N., 201 140, 164, 169, 177, 179, 194, 197, 204,
208, 246
Dobrovie-Sorin, C., 17, 18, 29, 33, 34, 41, 46,
C 50, 52, 53, 57, 64–66, 70, 76, 78–80,
Cabredo-Hofherr, P., 199 83, 88, 89, 102, 103, 131, 155, 169,
Carlier, A., 211, 214 181, 194, 198, 202, 204, 206, 208, 210,
Carlson, G.N., 28, 34, 38, 39, 50–52, 55, 211, 216, 217, 244
57–59, 61–63, 93, 116–118, 120, 121, Dowty, D., 39, 57
130, 190, 191, 193, 194, 196, 199, 204,
208, 216–218, 244, 246, 247
Chierchia, G., 34, 50–52, 191–194, 196, E
216–219, 223, 227 Egli, U., 235–238
Chung, S., 17, 50, 57, 58, 132 Enç, M., 155, 158, 161
Cohen, A., 203 Engdahl, E., 185
Comorovski, I., 81, 101, 103, 109, 227, 233, 234 Erteschik-Shir, N., 203
Condoravdi, C., 66 Espinal, M.T., 33, 34, 41–42, 46, 50, 57
Cooper, R., xiv, 7–9, 19, 127, 225, 235, 240 Evans, E., 235, 238–240

262 Author Index
F Kroch, A., 79
Farkas, D., 10, 16, 28, 32, 41, 50, 56, 132, 155, Kupferman, L., 78
157, 171, 173, 174, 176, 193, 202, 247
Fodor, J.D., 27, 138, 139, 173, 174
Fradin, B., 166 L
Francez, N., 233 Laca, B., 34, 41, 50, 52, 66, 70, 75, 88, 137,
Frege, G., 24 164, 165, 202, 204–206, 216, 217
Ladusaw, W., 17, 50, 57, 58, 102, 131, 132
Lambrecht, K., 99, 103, 106, 107
G Landman, F., 95, 183, 207, 211
Galmiche, M., 137, 147–149 Lappin, S., 233
Gamut, L.T.F., 20 Lawler, J., 164, 201
Geach, P.T., 221, 222, 224, 229 Lewis, D., 5, 9, 191–194, 224, 231
Giannakidou, A., 173 Link, G., 12, 183, 191, 193, 194, 196,
Gillon, B.S., 27, 219 207, 243
Giry-Schneider, J., 103, 109 Liu, F.H., 28, 151
Glasbey, S., 65 Longobardi, G., 35, 52, 206, 216
Greenberg, Y., 201 Lønning, J.T., 219
Groenendijk, J., 80, 81, 185 Lopez Palma, H., 137
Gupta, A., 76 Lumsden, M., 112
H M
Heim, I., xiv, 1, 3, 6, 7, 9, 11–13, 18, 55, 58, Marandin, J.-M., 167
96, 100, 111, 118, 120, 121, 129–131, Mari, A., 206, 208, 210, 211
138, 139, 171, 191, 193, 194, 197, 202, Matushansky, O., 75, 76, 85, 90
207, 221–223, 226, 240, 245 May, R., 6, 172
Heycock, C., 79 McNally, L., 11, 17, 18, 29, 33, 34, 41–42, 52,
Heyd, S., 213, 217 53, 57, 65, 66, 96–98, 102, 127,
Higginbotham, J., xiv, 219, 220, 246 131–132, 160
Higgins, R.F., 78 Milner, J.-C., 101
Hintikka, J., 185 Milsark, G., xv, 9, 11, 18, 26, 49, 57, 93, 102,
127, 129, 243
Moltmann, F., 61, 211
J Montague, R., xiv, 5, 7, 18, 99
Jackendoff, R., 101 Moravcsik, J., 219
Jaeggli, O., 155 Moro, A., 79, 82
Mostowski, A., 8
Müller, A.P., 33
K Munn, A., 32, 33, 46
Kadmon, N., 225, 226 Musan, R., 86
Kallulli, D., 41, 50
Kamp, H., xiii, 1, 9–11, 18, 129, 152, 154,
171, 178, 191, 193, 194, 221, 223, N
226, 240, 247 Napoli, D.J., 110
Kanouse, D.E., 164
Keenan, E.L., xiv, 19, 20, 22, 24
Kennedy, C., 76 P
Kiss, K., 64 Parsons, T., 196, 240
Kleiber, G., 64, 65, 136 Partee, B.H., 8, 84, 99, 225
Kratzer, A., 3, 7, 13, 34, 58, 63, 105, 129, 131, Pelletier, F.J., 191, 193, 194, 196
139, 176, 180, 188, 192, 194, 198, 204, Pelletier, J.F., 191, 193, 196
205, 207, 208, 225, 226, 246 Pesetsky, D., 26, 128, 151
Krifka, M., 154, 180, 181, 191, 193, 194, Postal, P.M., 156
196, 233 Prince, E., 100
Author Index 263
R T
Rando, E., 110 Tasmowski, L., 75, 88, 137, 165
Reinhart, T., 13, 15, 18, 25, 161, 162, 175, 180
Reyle, U., 152, 154, 178, 247
Roeper, P., 219 U
Rooth, M., 99, 191, 193–195, 226, 227 Ulrich, M., 108
Roy, I., 75, 76, 78, 85
Russell, B., 24, 238
Ruys, E., 16, 180 V
van Eijck, J., 240
van Geenhoven, V., xiv, 17, 18, 29, 34, 50,
S 52, 53, 58, 61, 102, 127, 131–132,
Safir, K., 121 160, 199, 244
Sag, I.A., 27, 138, 139, 173, 174 Veltman, F., 80, 81
Sasse, H.-J., 108 Villalta, E., 137, 147
Scha, R., 180, 207 von Fintel, K., 233
Schmitt, C., 32, 33, 46 von Heusinger, K., 139, 155
Schubert, L.K., 191, 193, 196
Schwarzschild, R., 27
Simons, P., 61, 211 W
Spector, B., 75, 76, 85, 90 Wasow, T., 156
Stavi, J., xiv, 19, 20 Westerstähl, D., xiv, 151
Steedman, M., 11, 16–18, 29, 136, 139, Wilkinson, K., 112
171, 176, 187, 193, 221, 222, 227, Winter, Y., 11, 13, 18, 75, 90,
233, 234, 245 92, 187
Stokhof, M., 80, 81, 185
Stowell, T., 28, 178
Strawson, P.F., 139 Z
Sugioka, Y., 10, 202 Zwarts, J., 75, 90, 92
Subject Index
A Binding
Adverb of quantification. See Q-adverb direct, 192, 195, 203, 216
Alternative, 20, 36–40, 48, 52, 78, 84, 122, indirect, 191–193, 195, 216, 223–224,
132, 164, 176, 187, 192 229, 247
Amount, xv, 29, 50, 61, 62, 72, 73, 86, 93, unselective, 5–6, 191–193, 203, 223–226,
96, 105, 106, 116, 119, 121, 125, 133, 228, 247
135, 140, 144, 148, 154, 168, 219
generalized quantifiers over amounts, xv,
18, 50, 60–63, 93, 125, 126, 129, 141, C
151, 163, 244–245 Cardinal, 5, 11–13, 21, 24–26, 28, 61, 169,
relative, 117–122, 124, 125 170, 205, 209, 240
Anaphoraanaphoric, xiii, 10, 40–41, 47, modified, 5, 24, 28, 140, 151–155,
72, 152, 166, 168, 188, 196, 177–178, 182–184
222–224, 226–227, 229–231, Cardinality, 13, 14, 22, 61, 135, 149–150, 154,
234–235, 241, 246 209, 214, 215
Argument, xiv, 2, 3, 7, 8, 15, 17, 18, 29, 32, Characterizing sentence, 193, 194, 196–201,
33, 34, 37, 38, 41, 45–47, 51–58, 208, 217
63–65, 67–76, 80, 84, 91–92, 96–103, Choice function, 13–16, 18, 175, 176, 187, 188
105, 106, 108, 111–126, 132, 133, Coda, 96, 102–105, 107–111, 125, 126
148, 150, 172, 176, 177, 183, 194, Collective, 16, 26–27, 145, 147, 149, 152,
208, 210, 233 153, 183
implicit, 53, 132, 210 Compositionality, xiii, xiv, 4, 7, 53, 57, 58, 60,
Aspect, 39–40, 47, 59 63, 79–82, 93, 96–98, 103, 105, 133,
134, 166, 169, 172, 173, 235, 239
Conservative/conservativity, 19–21, 24
B Contrast/contrastive/contrastivity, v, 6, 26,
Bare 33–35, 38, 42, 48, 54–57, 62–63, 69,
noun, v, xiii, 177, 191–220, 243 71–75, 77, 83–89, 92, 100, 114–116,
NP, 31–93, 177, 204, 246, 247 120, 121, 143–153, 162, 164, 180–182,
plural, v, 20, 28, 31, 32, 35–63, 69–71, 201, 203, 204, 211, 217, 227, 234, 247
131, 138, 147, 191, 192, 201–202, 204, Copula, 64, 68, 80, 82, 83, 85, 92
207–209, 216–218, 244, 246, 247 Copular sentence, 31, 75, 78–83, 85, 86, 88,
singular, v, 31–33, 35–50, 56–58, 62, 92, 93, 137
77–79, 82–89, 132, 247 Cumulative/cumulativity, 61, 180, 211

266 Subject Index
D Generalized quantifier, xiii–xiv, xv, 6–9,

Definite 11–13, 37, 50, 80, 82–84, 97, 98, 105,
determiner, 8, 18–19, 77 106, 126, 127, 134, 139, 141, 151, 154,
DP, xiv, 1, 3, 4, 7, 19, 26, 70, 77, 95, 99–104, 163, 246, 247
108, 109, 111, 122, 125, 127–128, 157, Generic
166, 180, 201, 207, 235, 241 indefinite, 191–220, 246
Dependency relation, xv, 29, 171, 176, 180, operator, 191, 196, 199, 216, 224, 225
184, 185, 189, 221, 231 pseudo, 192, 195, 203–205, 212–216, 228
Dependent reading, xiii, xv, 28, 50–52, 59,
pronoun, 188, 227, 229–231, 234 60, 71, 152, 191, 192, 195–197,
reading, 131, 172, 179–181, 184, 188, 196, 199–201, 203–206, 208–217, 219,
228, 232 220, 228, 246
Discourse sentence, 202, 215, 218, 219
referent, 10, 98, 100, 101, 104, 133, 138, Group, 1, 16, 17, 27, 33–34, 42, 76, 97, 128,
152, 227, 230, 247 138, 146, 158, 174, 180, 183, 205, 206,
topic, 110, 111, 166 210, 211, 215, 216
Discourse Representation Theory (DRT), xiii,
xv, 1–2, 10, 12, 13, 15, 16, 55, 136,
175, 176, 221, 224, 225, 227, 228, 231, H
235, 246, 247 Habitual, 192, 194, 196–200, 203, 212, 213
Dispositional, 194, 198, 200, 212 Heim’s constraint, xiii–xiv, 96, 111, 112, 116,
Distributive/distributivity, xiv, 16, 26–27, 138, 117, 120, 123–125
143–150, 152, 171–172, 177, 180–184,
205, 214, 215
Donkey-sentence, 9, 152, 156, 221–241 I
DRT. See Discourse Representation Theory Identificational sentence, 78–82
(DRT) Identity sentence, 62, 79–83, 86–89, 92, 93
if/when clause, 192–195, 202, 223–224, 230
Implicature, 85–86, 144, 145
E Incorporation, v, xiv, 17, 47–49, 56–58,
Equative, 78–83 246, 247
E-type pronoun, 84, 235, 238–241 Indefinite
Event, 5, 51, 57, 62, 69, 76, 77, 96, 106–110, generic, v, xv, 191–220, 228, 246
133, 167, 191–198, 200, 202–206, quantificational, xv, 2, 9–12, 134–135,
212–216, 228–231 139–141, 164–165, 181–184, 192–196,
Existential 221–223
closure, 11–14, 17, 34, 53, 55, 58–60, specific, 13, 28, 61, 158, 167, 169, 173,
129–133, 136, 137, 171, 175, 176, 193, 181–183, 189
209, 221, 223, 224, 245–247 strong, xiv, 71–73, 127–141, 155, 169,
enumerative, 96, 103, 104, 111, 125, 126 212, 245
eventive, 96, 102–104, 106, 108–111, weak, v, xv, xiv, 11, 13, 18, 29, 60, 71,
125, 126 93, 102, 125, 129–134, 137–141, 148,
sentence, 95–129, 133, 136, 140, 150, 150, 153, 160, 169, 171, 177–178,
163, 244 243–247
Individuation constraint on quantification, v,
205–209, 219, 246
F Information structure, 29, 80, 143, 161–169,
Focus, xv, 11, 106, 110, 117, 162, 163, 180, 184, 194, 226
194, 195, 203–205, 234 Intersective/intersectivity, 21, 22, 24, 135, 165,
169, 179, 184
G
Generalized existential quantifier over K
amounts, xv, 18, 50, 60–63, 93, 125, Kind, 34, 50, 51, 60, 61, 63, 67, 208, 216–218,
126, 129, 141, 151, 163, 244–245 220, 247
Subject Index 267
L Q
Localization, 63, 65–67, 69, 105, 110, 137, Q-adverb, 191–194, 196, 199, 200, 203, 204,
172, 173, 244 206–209, 211, 212, 214, 216, 217, 220,
224, 228, 229, 231
Quantification
M generic, 191, 197, 206, 228
Mass mass, 219, 220
noun, 28, 32–36, 45, 47, 61, 69, 70, 73, over events, 191–198, 202, 204, 206, 212,
121, 149, 209, 216, 217, 219, 243, 215, 230, 246
244, 246, 247 Quantificational adverb. See Q-adverb
quantification, 219, 220 Quantificational DP/quantified DP, 2, 7,
Modality, 171, 201, 214 10–11, 24, 26, 27, 95, 111, 116, 135,
Monotone/monotonicity, 23–24 136, 138–140, 145, 149, 150, 153, 172,
180, 181, 189, 241
Quantifier raising, 6, 59, 112, 158, 172–173,
N 178, 179
Nomic/nomicity, 200–203, 214
Novelty condition, 100
R
Reading
O asymmetric, xv, 192, 225–227,
Operator 231–233, 235
GEN, 10, 191, 196, 199, 201–204, 214, existential, 10, 17, 38, 50–52,
217, 218, 224, 225, 229 58–61, 63–66, 69, 103, 110,
HAB, 197, 199 130, 207, 224
symmetric, 225–227, 231–232, 235
Referential
P DP, 1–4, 7, 8, 10, 18–19, 41, 122, 139, 140,
Partitive/partitivity, xiv, 22, 25, 26, 49, 157, 180, 234
127–129, 134–137, 139, 140, 143–151, indefinite, 4, 16, 18, 29, 139, 141, 174,
158, 159, 169, 217 175, 188, 189
Pragmatic/pragmatics, xiv, 100, 101, 163, 166, Relational noun, 76–77, 91–93, 216
205, 206, 214, 215, 219, 233, 236, 241 Restriction, xiv, 6, 10, 15, 41, 47, 57, 91,
Predicate 95–97, 115, 118, 124, 135, 139,
entity, 54, 63–72, 108, 130, 136, 137, 148, 140, 150, 167, 172, 177, 183,
149, 153, 158, 163, 164, 169, 244 184, 186, 194, 198, 203–205, 213,
episodic, 105, 199 214, 216, 220, 223, 225, 226, 228,
existential, xv, 53–55, 57–58, 61–69, 93, 236, 246
105, 106, 108, 116, 123, 126, 130, 133,
177, 244–246
individual level, 63–65, 130, 193, 194, 226 S
permanent, 64, 66, 130 Scope
stage level, 63–65, 107, 130, 137, 194, intermediate, 151, 154, 171, 174–176, 178,
198, 226 184–188
transitory, 63–65, 107, 130, 198, 199, 203 inverse, 177–183
Prepositional accusative, 155–161, 182 narrow, 14, 16, 28, 39, 47–49, 55, 57, 59,
Presupposition/presuppositional, 24–26, 130, 70, 151, 155, 157–161, 171, 173–178,
139–141, 158, 161–164, 239–241 184–188
Property denotation, 11, 18, 50, 52–58, nuclear, 6, 10, 13, 58, 116, 129, 172, 185,
74, 102 186, 194, 203, 223, 226, 228
Proportional quantifier, 14, 138, 139
determiner, 22, 145 wide, 2, 14, 16, 17, 27, 48, 138–139, 151,
expression, 5 155, 156, 158, 160, 161, 171, 173–181,
pseudo-generic, 192, 203–204, 216 184, 188, 189, 236
Proportion problem, 221, 225, 231 Selectional restriction, 91, 96, 97, 118
268 Subject Index
Skolem T
function, 16–17, 184–188, 196, 227–228, Theme, 57, 62, 104, 131, 184, 193, 203, 231,
230, 234 232, 246
term, xv, 11, 16–18, 24, 29, 136, 139, 141, Topic/topical, 110, 111, 122, 131, 161–167,
150, 153, 160, 172, 187, 189, 193, 196, 169, 170
197, 215, 227, 228, 231–235, 245, 246 Tripartite
Specific/specificity, xiv, xiii, 13, 15, 17, 22, 28, configuration, 6, 140, 141, 151, 186, 245
49, 50, 61, 71, 96, 103, 117, 123, structure, 5–6, 19, 26, 135, 139, 148, 150,
137–139, 141, 143–150, 153, 155–161, 151, 223
167–169, 173, 175, 176, 179–184, 188, Type
189, 204, 236, 245, 246 raising, 8, 109
Strong, xv, xiv, 9, 11, 13, 29, 71–73, 82, 93, shifting, 8, 11, 37, 51, 84, 99–102, 115, 218
108, 127–141, 153, 155, 158, 161, 164,
169, 170, 177, 212, 213, 227, 232–233,
235, 243–245, 248 V
Sum, 14, 49, 50, 57, 61, 65, 68, 73, 83, 98, Variable
101, 109, 111, 115, 117, 129, 151, 154, bound, xv, 17, 115, 132, 188, 196, 222,
158, 160, 166, 169, 170, 174, 178, 181, 223, 227, 228, 231, 237, 245, 246
183, 195, 196, 207, 209–211, 214, 215, free, 10, 12–13, 18, 34, 136, 175, 176,
218–220, 227, 233, 237 191–193, 221, 223–226, 236, 237,
Symmetric noun, 209–212, 220 240, 245

Redefining-Indefinites Compress

Uploaded by

Copyright:

Available Formats

Redefining-Indefinites Compress

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

Redefining-Indefinites Compress

Uploaded by

Copyright:

Available Formats

Redefining Indefinites

Studies in Natural Language and Linguistic Theory

Marcel den Dikken, City University of New York

Guglielmo Cinque, University of Venice

For further volumes:

Library of Congress Control Number: 2012934950

© Springer Science+Business Media B.V. 2012

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Redefining indefinites is a new, updated and thoroughly revised version of Définir

1 Why Indefinites? ....................................................................................... 1

2 Bare Noun Phrases .................................................................................... 31

2.7.7 Modified Bare Nouns ............................................................. 89

5.2 Information Structure and the Disambiguation of Indefinites ........... 161

7.4.3 The Pseudo-Generic Reading of Plural Indefinites ................ 212

Bibliography .................................................................................................... 249

Author Index.................................................................................................... 261

Subject Index ................................................................................................... 265

In this introductory chapter we will be concerned with the reference, or denotation,

C. Dobrovie-Sorin and C. Beyssade, Redefining Indefinites, Studies in Natural 1

1.1 Typology of DPs

Since indefinites can be assimilated neither to referential DPs nor to quantificational

1.1.1 Referential DPs

Formally then, a sentence such as (4c) is represented as in (4¢c), where j is an

According to (4¢c), the denotation of John is handsome, which is a syntactically

Proper names as well as definite or demonstrative DPs display remarkable

1.1.2 Quantified DPs

1.1.2.1 Tripartite Structures

We are interested here not in the interpretative flexibility of adverbs (known as

(10¢a) reads as follows: ‘for each individual x, if x is a professor, then x came.

1.1.2.2 Generalized Quantifiers

This analysis of determiners, called the relational analysis, involves a relaxation

Quantificational DPs thus denote functions of type <<e,t>,t>. Put in set-theoretical

mathematical logic due to Mostowski 1957) as a framework for the investigation on

(13) a. John sleeps.

(13″) b. a person (sleeps) lP ($x (H(x) ∧ P(x))) S

(13″) a. John (sleeps) lP (P(j)) S

An important advantage of the generalized quantifier theory is that it allows us to

1.1.3 Indefinite DPs

(16) Generally, if Pedro owns a donkey, he beats it.

(17) a. A man entered the room.

quantifiers. The primitive type e can be ‘lifted’ by type-shifting, yielding a generalized

1.2 The Representation of Indefinite DPs

1.2.1 Indefinites and Existential Quantification

(18) a. A man came.

(19) a. lQ ∀x (N(x) → Q(x))

1.2.2 Indefinites as Free Variables

(21) a. two students

According to this analysis, a plural indefinite refers to a plural individual (made

(22) a. Adjoin an existential quantifier to the nuclear scope of every quantifier.

1.2.3 Indefinites as Choice Functions

(23) a. a student translates as f([[student]])

(24) A student is absent.

(25) Every woman read a book.

(26) a. $x ∀y (woman(y) → (book(x) ∧ read(y,x)))

(27) Max will be furious if I invite a philosopher.

(28) Every woman read a book.

(29¢) b. ∀x [prof(x) → $y (∀z (student(z) → ((book(y) ∧ recom(x,y) ∧

(30) If three relatives of mine die, I will inherit a house.