(Syntax and Semantics) Patricia Cabredo Hofherr, Anne Zribi-Hertz (Eds.) - Crosslinguistic Studies On Noun Phrase Structure and Reference-Brill (2013)

Crosslinguistic Studies on Noun Phrase Structure and Reference
Syntax & Semantics
Series Editor
Jeffrey T. Runner (University of Rochester)
Editorial Board
Judith Aissen, University of California, Santa Cruz Peter Culicover, The Ohio
State University Elisabet Engdahl, University of Gothenburg Janet Fodor,
City University of New York Erhard Hinrichs, University of Tubingen Paul M.
Postal, Scarsdale, New York Barbara H. Partee, University of Massachusetts
William A. Ladusaw, University of California, Santa Cruz Manfred Krifka,
University of Texas Pauline Jacobson, Brown University
volume 39
The titles published in this series are listed at brill.com/sas

Crosslinguistic Studies on
Noun Phrase Structure and
Reference
Edited by
Patricia Cabredo Hofherr

Anne Zribi-Hertz
LEIDEN | BOSTON
This publication has been typeset in the multilingual Brill typeface. With over 5,100 characters covering
Latin, ipa, Greek, and Cyrillic, this typeface is especially suitable for use in the humanities. For more
information, please see www.brill.com/brill-typeface.
issn 0092-4563
isbn 978 90 04 26082 5 (hardback)
isbn 978 90 04 26144 0 (e-book)
Copyright 2014 by Koninklijke Brill nv, Leiden, The Netherlands.

Koninklijke Brill nv incorporates the imprints Brill, Global Oriental, Hotei Publishing, idc Publishers and
Martinus Nijhoff Publishers.
All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or
transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise,
without prior written permission from the publisher.
Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill nv provided
that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910,
Danvers, ma 01923, usa. Fees are subject to change.
This book is printed on acid-free paper.

Contents
Acknowledgements vii
Biographies viii
Introduction 1
Patricia Cabredo Hofherr and Anne Zribi-Hertz
part 1
Noun Phrase Syntax and Interpretation:
In Search of Crosslinguistic Regularities
Information Structure and (In)definiteness 23

Nomi Erteschik-Shir
On Number and Numberlessness in Languages with and without
Articles 52
Asya Pereltsvaig
The Cognitive Basis of the Mass-Count Distinction: Evidence from Bare
Nouns 73
Edit Doron and Ana Mller
The Turkish NP 102
eljko Bokovi and Serkan ener
part 2
Definiteness and Definiteness Markers across Languages
The Morphology, Syntax and Semantics of Definite Determiners in Swiss

German 143
Rebekka Studler
Reduced Definite Articles with Restrictive Relative Clauses 172
When Determiners Abound: Implications for the Encoding of
Definiteness 212
Marika Lekakou and Kriszta Szendri
The Semantics and Syntax of Japanese Adnominal Demonstratives 239
Makoto Kaneko
vi contents
From Noun to Name: On Definiteness Marking in Modern

Martinik 269
Anne Zribi-Hertz and Loc Jean-Louis
Reference Resolution in French Sign Language: The Effects of the
Visuo-Gestual Modality 316
Brigitte Garcia and Marie-Anne Sallandre
part 3
Noun Phrase Interpretation and Second-Language Acquisition
When Articles Have Different Meanings: Acquiring the Expression of

Genericity in English and Brazilian Portuguese 367
Tania Ionin, Elaine Grolla, Silvina Montrul and Hlade Santos
Index 399
Acknowledgements
We gratefully acknowledge the support of the project Le calcul de la rfrence

nominale dans les langues avec et sans articles, sponsored by the Fdration
TUL (Typologie et universaux du langage, FR 2559) of the French CNRS. The
project got further financial help from the UMR 7023/SFL (Structures formelles
du langage, CNRS/Paris-8) and from the Scientific Board of Universit Paris-8,
whose support enabled us to organise a series of workshops on the construal
of nominal reference from a cross-linguistic perspective.
Biographies
eljko Bokovi
(Ph.D. University of Connecticut, 1995) is Professor of Linguistics at the Uni-
versity of Connecticut. He is the author of The Syntax of Nonfinite Complemen-
tation: An Economy Approach (MIT Press, 1997), On the Nature of the Syntax-
Phonology Interface: Cliticization and Related Phenomena (Elsevier, 2001), and
Minimalist Syntax: The essential readings (with H. Lasnik, Blackwell 2007). He
has also published articles in a number of journals, including Linguistic Inquiry,
Natural Language and Linguistic Theory, Lingua, and Syntax. He has supervised,
or is in the process of supervising, over 40 Ph.D. dissertations.

is a researcher at the UMR 7023 Structures formelles du langage (CNRS & Paris
8). She works on verbal plurality, definiteness, impersonal pronouns and on the
syntax-morphology interface. Her publications include Layers of Aspect (2010)
and Verbal plurality and distributivity (2012), both co-edited with Brenda Laca.
Her current research focuses on the syntax and semantics of R(eferential)-
impersonal pronouns cross-linguistically.
Edit Doron
is Professor of Linguistics at the Hebrew University of Jerusalem. Her research
interests include the semantics of voice, the semantics of predication and the
semantics of aspect and habituality. She has also written on the interpretation
of resumptive pronouns, predicate nominals, verbal templates, apposition,
bare nouns, definiteness, and adjectival passives.
Nomi Erteschik-Shir
is Professor of Linguistics in the Department of Foreign Literatures and Lin-
guistics at Ben Gurion University, Israel. Her publications include The Dynam-
ics of Focus Structure (1997), Information Structure: The Syntax-Discourse Inter-
face (2007), The Syntax of Aspect (2005) co-edited with Tova Rapoport and The
Sound Patterns of Syntax co-edited with Lisa Rochman. She is currently extend-
ing her work on the syntax-phonology interface to the interaction between
tonal properties and word order in Scandinavian languages.
Brigitte Garcia
is professor of Linguistics at the University Paris-8 Vincennes-St Denis, special-
ising in French Sign Language (LSF) linguistics. She is director of the research
biographies ix
group Langues des signes et gestualit (Sign languages and gesture) of the
research unit UMR 7023 Structures formelles du langage (CNRS-Paris 8). Her
main research interests are the epistemology of sign language linguistics, the
methodology of annotation for large discourse corpora in French Sign Lan-
guage, the development of a graphic representation for French Sign Language
and the relationship between writing and deafness. She is particularly inter-
ested in the processes involved in the emergence of signs in French Sign Lan-
guage.
Elaine Grolla
is an Assistant Professor at the University of So Paolo. She received her PhD
from the University of Connecticut in 2005 with a thesis on the acquisition of
pronouns. Her current research interests are on how children acquire several
aspects of Brazilian Portuguese grammar, such as the interpretation of pro-
nouns and anaphors, the comprehension and production of relative clauses,
wh-questions, and passive structures. Besides her work on first language acqui-
sition, she also develops research on second and third language acquisition,
investigating learners knowledge about the interpretation of different types of
noun phrases.
Tania Ionin
is an Associate Professor of Linguistics at the University of Illinois at Urbana-
Champaign. She received a PhD in Cognitive Science from MIT in 2003, with
a dissertation on how speakers of article-less languages acquire articles when
learning English as a second language. In her research, she uses experimental
methodology to examine the interpretation of noun phrases with and with-
out articles in the grammars of both native speakers and second and third
language learners of several different languages, including English, Spanish,
Russian and Brazilian Portuguese. She has published articles in such jour-
nals as Second Language Research, Bilingualism: Language and Cognition, Lin-
guistic Inquiry, Natural Language Semantics, Language Acquisition and Lan-
guage Learning. She is an associate editor of the journal Language Acquisi-
tion.
Loic Jean-Louis
is presently an MA student in linguistics at Universit Paris-8. Working out the
grammar of Martinik (one of his two native languages) with Anne Zribi-Hertz
(one of his teachers) has been his choice and part of his linguistic training since
he began his B.A.
x biographies
Makoto Kaneko
is an Associate Professor at the University of Okayama, Japan. He received his
PhD from Paris 8 University in 2002 with a thesis on the syntax and semantics
of thetic judgments in Japanese and in French. He has also written on nomi-
nal exclamative clauses, free choice items, epistemic indefinite expressions in
Japanese and in French. His current research focuses on the syntax and seman-
tics of additive and associative plural markers and definiteness/indefiniteness
markers in Japanese and cross-linguistically.
Marika Lekakou
received her BA from the Department of Philology at the University of Athens
in 1999, and her doctorate from University College London in 2005. Her PhD
dissertation investigated the semantics of middle constructions and its mor-
phosyntactic realization across languages. In 2005 she joined the ESF-funded
European Dialect Syntax project as a post-doctoral researcher. In 2009 she was
elected Assistant Professor at the University of Ioannina. She has worked on
several topics at the syntax-semantics interface, adopting both a macro- and
a micro-comparative perspective: argument structure, syntactic doubling in
questions and in compound tenses, definiteness, and the interaction between
tense, aspect and modality.
Silvina Montrul
is Professor and Head of the Department of Spanish, Italian and Portuguese and
Professor of Linguistics at the University of Illinois at Urbana-Champaign. She
is author of The Acquisition of Spanish (Benjamins, 2004) and Incomplete Acqui-
sition in Bilingualism. Re-examining the Age Factor (Benjamins, 2008), as well
as numerous articles in journals such as Bilingualism: Language and Cognition,
The International Journal of Bilingualism, Language Learning, The Heritage Lan-
guage Journal, Studies in Second Language Acquisition, Language Acquisition,
Second Language Research. She is co-editor of the journal Second Language
Research. Her research focuses on linguistic and psycholinguistic approaches
to adult second language acquisition and bilingualism, in particular syntax,
semantics and morphology. She also has expertise in language loss and reten-
tion in minority language-speaking bilinguals, or heritage speakers.
Ana Mller
is Assistant Professor of Linguistics at the University of So Paulo, Brazil. Her
primary research areas are formal semantics of Brazilian Portuguese and of
Karitiana, a Brazilian native language that belongs to the Tupi stock. Her re-
search interests include the semantics of number, the semantics of pluraction-
biographies xi
ality and the semantics of distributivity. She has also worked on the semantics
of bare nouns and on the semantics of genericity.
Asya Pereltsvaig
has taught linguistics at Yale and Cornell, and most recently at Stanford Univer-
sity. Her specialty is in the syntax of Slavic languages, particularly Russian, and
in cross-linguistic variation and typology worldwide. She is particularly inter-
ested in such issues as the structure of noun phrases and their role in the syntax
and semantics of clauses. She is the author of two books (Copular Sentences in
Russian and Languages of the World: An Introduction) and numerous articles
on syntax, semantics, and geolinguistics. She is also a lead linguistics author on
GeoCurrents.info.
Marie-Anne Sallandre
is associate professor in Linguistics at the University Paris-8 Vincennes-Saint
Denis, specialising in French Sign Language (LSF). She teaches general lin-
guistics and sign language linguistics to deaf and hearing students. Her main
research interest is the reference to discourse entities in French Sign Language,
in particular spatial reference, person reference and constructions with non-
conventional units such as role shifts. Her work is based both on developmental
data and on adult corpora. Central questions to her research are the role of dif-
ferent types of iconicity in sign languages and the typology of sign languages,
comparing different sign languages as well as signed and vocal languages.
Hlade Scutti Santos

graduated in Spanish and Portuguese from the University of So Paulo, Brazil,
in 2000. In 2005, she received an M.A. degree in Spanish from the same insti-
tution, where she conducted research on language attitudes/representations.
From 2004 to 2007 she was an Assistant Professor of Spanish at the Universi-
dade Federal de So Carlos, Brazil, where she conducted research on language
attitudes/representations. She is currently a PhD candidate in Hispanic Lin-
guistics at the University of Illinois at Urbana-Champaign. Her dissertation
research focuses on the acquisition of Portuguese as a third language by learn-
ers who already know English and Spanish. Her work contributes to a better
understanding of the acquisition of additional languages beyond the second
language.
Serkan ener
completed his PhD in the Department of Linguistics at the University of Con-
necticut and is currently an Assistant Professor of Linguistics in the Depart-
xii biographies
ment of Turkish Language and Literature at Yeditepe University, Istanbul. His

main field of study is syntax, and he has done work on syntax-information
structure interface, case and agreement, and ellipsis.
Rebekka Studler
is a senior teaching and research associate at the German Department at the
University of Basel. She received her PhD from the University of Zurich in 2008
with a thesis on the determiner system in Swiss German dialects. At present
she is a visiting scholar at CUNY Graduate Center, New York and at UC Berkeley.
Her research interests include dialect syntax, DP structure, relative clauses, as
well as perceptual dialectology. In her current research she focuses on language
attitudes towards standard and nonstandard varieties in Switzerland.
Kriszta Szendri
graduated from the Department of Theoretical Linguistics at ELTE University
(Hungary) in 1998, and obtained a doctorate at University College London in
2001 with a dissertation on focus and the syntax-phonology interface. She was a
VENI postdoctoral research fellow at Utrecht University until her appointment
at UCL in 2006, where she is now Senior Lecturer. She performs experimentally
informed theoretical work and theoretically informed experimental work. Her
recent projects concern the syntax of definiteness, the interpretation of king
of France-sentences, the interpretation of quantifier raising by adults and chil-
dren, the acquisition of focus and intonation in autism.
Anne Zribi-Hertz
is Professor of Linguistics at Universit Paris-8, and a member of the UMR
Structures Formelles du Langage (a research centre supported by Universit
Paris-8 and the French CNRS). She has written two books (Dcouvrir la gram-
maire franaise, with L. Picabia, 1981; Lanaphore et les pronoms: une introduc-
tion la syntaxe gnrative, 1996), edited or co-edited a number of collective
volumes, and published many articles on the morphosyntax and semantics
of an arrray of languages including French, English, Malagasy, Attie, Bambara,
Sango, Wolof, Korean and French-lexifier creoles. She has contributed to the
Grande Grammaire du Franais, to appear shortly.
Introduction
Patricia Cabredo Hofherr and Anne Zribi-Hertz
The papers collected in this volume explore the factors determining the ref-
erential interpretation of noun phrases across a wide array of typologically
unrelated languages. The languages discussed include Armenian, Brazilian Por-
tuguese, Catalan, Danish, French Sign Language (LSF), several West-Germanic
languages, Modern Greek, Japanese, Karitiana, Martinique creole (Martinik),
Modern Hebrew, Norwegian, Russian, Spanish, Tatar and Turkish. The indi-
vidual papers approach this question from complementary angles, covering
morphology, syntax, semantics, information structure and acquisition.
The research reported here was inspired by the project Languages With
and Without Articles: calculating nominal reference. By comparing languages
with and without articles, the aim of this project was to place definite and
indefinite articles in the wider context of grammatical devices constraining the
construal of nominal reference. In what follows we use the abbreviation NP for
Noun Phrase, taking this term as neutral with respect to the precise syntactic
analysis of the constituent (e.g. NP, NumP, DP ) which may be proposed in
the aftermath of Abney (1987).
As is well known, the referential properties of nominal expressions are
constrained by a number of grammatical factors, both external and internal to
the noun phrase (cf. Kramsky 1972, Lyons 1999). External factors include, e.g.,
information structure, word order, case, verbal aspect, while internal factors
include determiners, number, quantity and quantifiers, classifiers, noun type
(e.g. count, mass, collective). The definition of definiteness is complicated by
the fact that it is variably viewed as a syntactic property or feature arising from a
structural position: D, or as a semantic property involving the way the reference
of a noun phrase is construed in its sentence and discourse (cf. section 1.1.
below).
1 Issues in the Analysis of Definiteness
Determiners and definiteness have been extensively studied in formal linguis-

tics, and this introduction does not attempt to provide an exhaustive survey of
the vast relevant bibliography. In what follows we single out a few key issues
addressed in current research on definiteness and (in)definite articles, in order
to place the contributions gathered in this volume in a wider theoretical per-
spective:
koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_002

2 cabredo hofherr and zribi-hertz
(i) To what extent are (in)definite determiners necessary for argument-

hood?
(ii) What are the semantic features expressed by (in)definite determiners
and how are these semantic categories expressed in languages without
articles?
(iii) What is the relationship between marking by a definite determiner and
semantic definiteness?
1.1 Nominal Reference and Definiteness

Many studies regard the occurrence of definite determiners as licensed by two
semantic properties: Uniqueness (or Maximality), and Familiarity.
The property of Uniqueness is illustrated by examples such as (1a) below,
where the singular definite noun phrase the pear identifies the unique pear ref-
erent provided by the discourse context, or (1b), where the definite noun phrase
the moon identifies an entity a priori thought of as unique in our human world
(cf., e.g., Lbner 1985, Corblin 1987). In a context such as (1c), where a singular
referent fails to be presupposed as unique, the definite article is banned. Since
the term uniqueness is not quite appropriate to account for the definite article
in plural NPs such as the pears in (1d) (Hawkins 1978:158), the term maximal-
ity has been proposed to capture the semantic effect of definite determiners
regardless of number. Under this view, a definite determiner signals that its NP
identifies the maximal set which, in the discourse context, satisfies the descrip-
tive content of the head noun (see e.g. Link 1983). For a set reduced to a single
member, this boils down to uniqueness.
(1) a. There was a pear on the table. John took the pear.
(unique referent/maximal set
of pear)
b. The moon stood still, (unique referent/maximal set
on Blueberry Hill. of moon)
c. There were a few pears on the table. John took a/#the pear.
(non unique referent/non
maximal set of pear)
d. There were a few pears on the table. John took the pears.
(maximal set of pear)
Familiarity has been modelled as the contrast between newly introduced dis-
course referents and previously introduced referents (Kamp 1981, Heim 1982).
In (2), for example, the indefinite noun phrases a man, a woman, a hat intro-
duce new discourse referents, while the definite noun phrase the man refers
introduction 3
back to a familiar individualfamiliar to the speaker and hearer, since previ-

ously introduced by the indefinite NP a man:
(2) A man and a woman came in. The man was wearing a funny hat.
This dichotomy between previously introduced discourse referents for def-

inites, and newly introduced discourse referents for indefinites is, however,
underdetermined (cf. Kaneko, this volume, for a discussion of the link between
Uniqueness and Familiarity). In particular, various approaches rely on a range
of definitions of given and new information making use of such notions as
specificity, topicality, salience, or accessibility (cf. Gundel & Fretheim 2005; for a
discussion of definiteness in French Sign Language in these terms see Garcia &
Sallandre this vol.).
Specificity has been linked to an existence presupposition: while the specific

indefinite in (3a) presupposes the existence of an individual being looked for,
the non-specific indefinite in (3b) does not carry such a presupposition (Fodor
and Sag 1982, En 1991):
(3) a. He is looking for a (certain) secretary.

(specific: there exists a (certain) secretary such that he is looking for
her)
b. He is looking for a secretary.
(nonspecific: anybody qualified as secretary can apply).
Topicality has been linked to the definiteness issue since Kurodas (1965, 1979,
a.o.) and Kunos (1973, a.o.) work on topic markers in Japanese. As they point
out, the phrase marked by the Topic marker wa in Japanese (or (n)eun in
Korean) necessarily has a definite interpretation. This constraint is illustrated
below by the Korean examples in (4), where the argument marked as subject
by the particle ga in (4a) may be construed as preidentified (definite) or not
(indefinite), while the phrase marked as topic by the particle neun in (4b) nec-
essarily points to a uniquely identified referent pre-activated by the immediate
discourse context:
(4) a. Beoseu -ga o-goiss -da.1 (Korean)

bus -subj come-prog-dec
(i) Theres {a/the} bus coming.
(ii) Its {a/the} bus thats coming.
b. Beoseu -neun o -goiss-da.

bus -top come -prog-dec
(i) (What about the bus? >) The bus is coming. [simple topic]
(ii) (What about the bus and the taxi? >) The BUS is coming (not the
TAXI). [contrastive topic]
This link between topicality and some definition of familiarity is widely as-
sumed in works on Topicalitycf. Gundels (1988) Topic Familiarity Con-
dition and Lambrechts (1994) Principle of Separation of Reference and
Role.
(5) a. Topic-familiarity condition (Gundel 1988)

An entity, E, can successfully serve as a topic, T,
iff both speaker and addressee have previous knowledge of or
familiarity with E.
b. Principle of Separation of Reference and Role (Lambrecht 1994:185):
Do not introduce a referent and talk about it in the same clause.
Similar views are formalised by Erteshik-Shir (1997) in her own model of Infor-
mation Structure, which leads her to assume that every utterance must contain
a Topiccovert if not overtinstantiating a presupposed or old discourse
referent.
(6) Utterances are conceived of as a set of instructions by a speaker to a

hearer to update and organize a file so that the file will contain all the
information the speaker intends to convey. The file consists of indexed
cards which represent existing discourse referents. Information is entered
on these cards according to well-defined principles. Each card has an
indexed heading and information pertaining to this heading can be
1 Abbreviations used in the Korean glosses: dec = declarative sentence; prog = progressive
aspect; subj = subject marker; top = topic marker.
introduction 5
entered on the card. Common ground information is thus ordered accord-

ing to the topics defined by each discourse referent.
erteschik-shir 1997:17
Conversely, existential constructions are generally associated with new dis-

course referents, i.e. indefinite NPs, an effect known as the definiteness effect
(Milsark 1979 and Keenan 2003 for there-constructions in English, Leonetti 2008
for Romance).
(7) a. There was a child in the garden.

There were {children / some children / few children} in the garden.
b. *There was the child in the garden.
*There were {the children / most children/ all the children} in the
garden.
(8) a. Hay {algunos / dos / muchos / pocos / } perros. (Spanish)

have some / two / many / few / dogs
There exist some / two / many / few dogs.
b. *Hay {l / el perro / ese perro / Fido}.

have 3MSG / def dog / dem dog / Fido
Lit. There exists {it / the dog / that dog / Fido}. (adapted from Leonetti
2008)
As Erteschik-Shir (this volume) shows, apparent counterexamples to the cor-

relation between topicality and definiteness, on the one hand, and existential
constructions and indefiniteness, on the other, can be explained in a theory
that allows subordinate information structures.
Accessibility (cf. Ariel 1990) restricts the availability of discourse referents as

antecedents for anaphoric expressions in the discourse. The conditions on
accessibility have been variously discussed in terms of salience, discourse acti-
vation and givenness (cf. Lewis 1979, Gundel et al. 1993, Walker et al. 1998). The
Givenness Hierarchy, for instance, links the referential status of NPs to their
different available forms in a given language (Gundel et al. 1993):
(9) Givenness Hierarchy (Gundel, Hedberg and Zacharski. 1993 and refer-
ences therein)
in focus > activated > familiar > uniquely identifiable > type identifiable
it this N that N the N aN
Needless to say, the terms specificity, topicality and salience are themselves given
varying definitions across different studies (cf. v. Heusinger 2002 on specificity,
Gundel and Fretheim 2005, Neeleman and Vermeulen 2012 on the definition of
topic).
1.2 (In)definite Articles and Argumenthood

It has been proposed that determiners are required to turn a predicate into a
referential argument (Higginbotham 1985). A syntactic phrasing of this hypoth-
esis is proposed by Longobardi (1994, 2000) who assumes that argumental
noun phrases are syntactically DPs (Determiner Phrases). This hypothesis has
triggered extensive discussion, rephrasings and counter-proposals (cf. Chier-
chia 1998, Cheng and Sybesma 1999, Coene and D Hulst eds, 2003, Dobrovie-
Sorin and Laca 2003, Bokovi 2008, a.o.). Chierchia (1998) proposes that lexi-
cal nouns are not predicative across languages but must be parameterized for
every language as argumental and predicative, citing Mandarin Chinese as
a language with inherently argumental/ kind-denoting nouns. Under this the-
ory, determiners are viewed as spell-outs of the Down operator which derives
Kinds (arguments) from Properties (predicates), and are thus only necessary
in languages whose nouns are lexically predicative (thus. in English but not in
Chinese).
Bokovi (2008, 2012) argues that argumental noun phrases are not uni-
versally DPs, proposes a list of systematic syntactic discrepancies between
(determiner-less) NP-languages and DP-languages (cf. Bokovi and ener, this
volume, for an analysis of Turkish along these lines). According to this theory,
the structure of bare NPs in article-less languages is different from that of DPs
in languages which have at least an overt definite article.
In the wake of Abneys (1987) dissertation arguing for a syntactic parallel
between noun phrases and clauses, a structure of NPs acknowledging various
functional layers between the maximal phrase (DP) and its lexical component
(NP) has been explored: Number (cf. Ritter 1991), Quantity and Classifier (cf.
Doetjes 1997, Cheng and Sybesma 1999, Borer 2005), Noun-hood (cf. Kihm
2003). Under these assumptions, the hypothesis put forward by Higginbotham
(1985) can be syntactically rephrased as follows: syntactic NPs are predicates,
and in order to function as arguments they must be embedded under at least
one functional projection (cf. Pereltsvaig 2007, 2013, this volume). It has been
introduction 7
proposed in particular that NPs introduced by the indefinite determiner a(n)

in English are not DPs but (syntactically smaller) Number Phrases (Lyons 1999:
3335).
A link between number and definiteness is suggested in some languages
where plural marking may only occur in NPs otherwise marked as definitea
situation typically illustrated by French-lexifier creoles, cf. Zribi-Hertz and
Jean-Louis (this volume).
1.3 Semantic Subtypes of Definite Noun Phrases

As recalled above, definite noun phrases are often characterised as expressions
which denote entities already introduced into the (mental) index file listing
the discourse referents available to the speaker and hearer, and indefinite noun
phrases as expressions introducing new discourse referents (Kamp 1981, Heim
1982).
However, this characterisation does not suffice to cover the semantic sub-
classes of definite Noun Phrases observed cross-linguistically, for many lan-
guages turn out to have two paradigms of definite determiners (cf. Eberts
1970, 1971 seminal study of two definite articles in a variety of Frisian). In such
languages with two definite articles, morphologically reduced definite deter-
miners mark semantically unique entities such as the king in (10a), while
morphologically full definite determiners mark entities construed as unique
via discourse linking, such as the young man in (10b).
(10) a. A kning kaam tu bischk. (Fehring Frisian)

detred king came to visit
The king came for a visit. (Ebert 1971: 83, ex 30)
b. Matje hee al wler an ni bridj. Di gast kn

M. has again a new bride detfull young-man can
a ns uk wel i fol fu.
detred nose also prt not full get
Matje has a new girl friend again. The young man cant get enough, it
seems. (Ebert 1971: 108, ex 12)
This distinction can be overtly signalled by morphologically full definite deter-

miners for pragmatic (viz. discourse-linked) definiteness (Lbner 1985), and
by morphologically reduced definite determiners for semantic definiteness
(ibid.). Similar oppositions can also be instantiated in languages that have a lex-
ical determiner specialised in pragmatic definiteness, and bare nouns ambigu-
ous between indefinite and semantically-definite readings (cf. Breu 2004 on
Upper Sorbian; Zribi-Hertz and Jean-Louis, this volume on Martinik). In some

languages (e.g. French or English), pragmatic and semantic definiteness may be
signalled by the same morphology (the definite article) and must therefore be
told apart on syntactic and semantic grounds (cf. Aguilar and Zwarts 2010 on
English; Corblin 2001, 2011 on French).
As argued by Lbner (1985, 2011) semantic and pragmatic definiteness may
be regarded as the two end-points of the Definiteness Scale in (11):
(11) Scale of uniqueness (Lbner 2011:320): pragmatic >> semantic defi-

niteness2
a. deictic with sortal common NPs (e.g. this book)
b. anaphoric with sortal common NPs (e.g. a man came in the man)
c. sortal common NPs with establishing RC (e.g. the house that
I live in)
d. functional common NPs with explicit definite possessor (e.g. his
mother)
e. definite associative anaphors (a carthe motor)
f. individual common NPs (e.g. the sun, the king)
g. proper names (e.g. the Nile, the Alps)
While languages with two articles are uniform with respect to article choice
at the end-points of the scale (cf. (11a/b) above), article choice for the inter-
mediate types is more variable (see (12), cf. Cabredo Hofherr this volume,
Studler this volume and references therein).
(12) Det / At iast buk, wat hi skrewen hee, docht

detfull / detred first book rel he written has is-worth
niks. (Fehring Frisian)
nothing.
The first book he wrote is no good. (Ebert 1970:169, ex 33)
Differences in the distribution of definite determiners can also be observed

between languages that only have one definite determiner. As is well known,
plural generic NPs appear article-less in English but obligatorily with a definite
article in Romance languages, illustrated here by French:
2 Sortal nouns are unary predicate terms, of type e,t, such as table, tree. Individual
introduction 9
(13) Les pandas sont vgtariens. (French)

def pandas are vegetarians.
The plural definite determiners in English and French are therefore not equiv-
alent (see Ionin et al., this volume, for a study of L2 acquisition of articles in
Brazilian Portuguese, English and Spanish in the expression of genericity).
Part of the contrast in (13) could possibly be explained by different degrees
of grammaticalisation of definite determiners in English and French. From a
diachronic viewpoint, definite determiners indeed spread along the scale in (11)
from deictic and anaphoric uses to semantically-unique uses like associative
anaphora, individual nouns (the sun) and generic NPs (the Panda) (cf. De
Mulder and Carlier 2010). Notice, however, that singular and plural definite
determiners need not proceed along the same grammaticalisation path. In
English, for instance, the singular definite determiner is possible in generic NPs
(the Panda) while the plural definite determiner is not.
1.4 Definite Determiners and Definiteness

The study of definiteness is further complicated by the fact that the distribu-
tion of (in)definite determiners is not limited to contexts where the expression
of (in)definiteness is at stake. At least three problematic cases deserve to be
mentioned: (i) the occurrence of definite determiners where no uniqueness
presupposition seems present, (ii) the possible occurrence of multiple defi-
nite determiners in a single noun phrase, (iii) the absence of determiners with
certain types of nouns which in other languages would call for a definite deter-
miner.
Case (i) is illustrated by the so-called weak definites of, e.g. English (Poesio
1994, Carlson et al 2006, Klein and al. 2009, Aguilar and Zwarts 2010)cf. ex.
(13)or French (Corblin 2001, 2011)cf. ex. (14):
(13) a. He took the train to come here, and so did Mary.

[true even if they took different trains]
b. He usually spends his summers at the seaside.
[true if he spends his summer vacation in a different place every year]
nouns are individual terms, of type e such as pope, US president, sun. Relational nouns
are binary predicate terms, of type e,e,t such as brother, sister. Functional nouns are
unary function terms, of type e,e such as mother, father, head. For details see Lbner (2011,
pp. 280282).
(14) Marie est lhpital, et Jean aussi.

Mary be.prs.3sg at def.sg hospital and John too
Mary is in (the) hospital, and so is John.
[true if they are in different hospitals]
For Malagasy, Paul (2009) argues that the semantic contribution of the deter-
miner is limited to contexts that allow a choice between presence and absence
of the determiner; in contexts where the determiner is either required or
banned, the interpretation of DPs is underdetermined.
The second type of mismatch between definite articles and definiteness can
be found in constructions containing multiple definite determiners. Multiple
definite marking is found in such examples as (15a) (Modern Greek) and (15b)
(Modern French):
(15) a. i asimenia i pena vs. a. i asimenia pena (Modern Greek)

def silver def pen def silver pen
the silver pen (adapted from ex 2 Lekakou and Szendri this volume)
b. la nia ms viva vs. b. la fille la plus intelligente

def girl more intelligent def girl def more intelligent
the most intelligent girl (Spanish) the most intelligent girl (French)
Lekakou and Szendri (this volume) argue that the distribution of the definite
determiner in (15a/a) corresponds to a syntactic difference in Modern Greek.
The contrast between Spanish and French in (15b/b) is as yet unexplained.
Well-known examples of missing determiners are provided by mass nouns

in a language like English, which only requires overt determiners for singular
count nouns in argument positions (16a/b). Mass nouns have been shown to
pattern with plurals (Jespersen 1909, Carlson 1977):
(16) a. He bought oil for the car

b. He bought *(a) parasol for the house.
(17) a. Oil is expensive.

b. *(A) parasol is expensive.
c. Parasols are expensive.
introduction 11
The nature and sources of the mass/count distinction are a topic of on-going
research (cf. Chierchia 2010; Massam, ed., 2012, and references therein, Doron
& Mller this volume).
1.5 Articles and (In)definiteness from a Crosslinguistic Perspective

In a nutshell, the conceptual issues which underly todays research on noun
phrases include the following: What are the sources of (In)definiteness effects
associated to noun phrases? How do morphology, syntax and semantics inter-
act in the expression of Number? What are the sources of the so-called Mass/
Count distinction, and is it universal? Must argument noun phrases contain
a covert Determiner in Languages Without Articles? Can Definiteness be
viewed as a universal, language-independent, cognitive category? As regards
Languages With Articles, why and to what extent do definite articles vary as
to their morphology, distribution, and semantic effects? To what extent is the
acquisition of determiner systems by L2 learners influenced by their L1 gram-
mar? Are there mechanisms of reference construal common to all natural
grammars? Does the vocal or visual nature of the signifier have a crucial inci-
dence on such mechanisms?
2 This Volume
The eleven articles selected for this book each contribute partial answers to
some of the above questions. The proposed analyses are based on first-hand
data from sixteen typologically diverse languages or dialectal groups.
The four papers grouped in Part I (Noun Phrase syntax and interpretation:
in search of crosslinguistic regularities) seek to bring out interpretive and
morphosyntactic invariants in noun phrases, beyond the occurrence or non-
occurrence of articles: the first text bears on Information Structure, the second
on Number, the third on the Mass/Count distinction, and the fourth on the syn-
tactic structure of noun phrases in Languages Without Articles.
Nomi Erteschik-Shir (Information Structure and (In)definiteness) dis-
cusses the two most prominent examples of the interaction between Definite-
ness and Information Structure (abbreviated IS in what follows): Topicaliza-
tion has been associated with Definiteness (specificity) and existentials with
Indefiniteness (the definiteness effect). Both phenomena exhibit seemingly
idiosyncratic exceptions to the assumed correlations. This paper demonstrates
that these exceptions are resolved by a careful analysis in terms of IS. Section
2 defines the primitives of IS, topic and focus, and shows how subordinate ISs
afford an explanation of the fact that specific indefinites can provide topics.
Section 3 shows how topicalization is constrained differently in Danish, Nor-

wegian, Hebrew, Catalan and Russian in view of their different canonical ISs
as well as other language-particular properties. Although the initial position in
Germanic languages is generally dedicated to topics, non-topics in this posi-
tion also impact IS by forming thetic sentences. Section 4 offers an analysis of
the definiteness effect in existential clauses and gives some evidence that here
again differences in canonical IS account for syntactic and morphological vari-
ation across languages.
Asya Pereltsvaig (Number and Numberlessness in Languages With and
Without Articles) is concerned with the representation of number in article-
less languages, focusing on two distantly related languagesRussian and Ar-
menianand an unrelated language, Tatar. It is argued that morphological
number and semantic number are mediated by syntactic number, encoded
even in article-less languages via a dedicated functional projection, NumP.
Thus, an argument is made against the strongest anti-DP position that denies
any functional projections inside a nominal in an article-less language. Instead,
it is shown that at least the projection of NumP must be assumed even for
article-less languages.
The focus of this paper is on the so-called number-neutral nominals, i.e.
nominals that denote one or more X. Semantically, such nominals are nei-
ther singular (one X) nor plural (more than one X). Pereltsvaig argues that the
semantic number-neutrality of such nominals results from the lack the syntac-
tic number feature, normally hosted in NumP, which she shows to be absent
from such nominals. Depending on the language, such number-neutral nomi-
nals can be morphologically either singular or plural. Pereltsvaig further shows
that the morphological expression of number neutrality does not correlate with
whether a given language has articles or not.
Edit Doron and Ana Mller (The cognitive basis for the Mass/Count dis-
tinction: evidence from bare nouns) seek to tighten the link between the
Mass/Count distinction and its cognitive basis. They first discuss Karitiana,
where the Mass/Count distinction is semantically active although it fails to be
signalled by syntax or morphology, then Hebrew, a language which has plural
morphology but where countability is argued to arise from the semantic iden-
tification of stable units (in the sense of Chierchia 2010, who regards the cogni-
tive contrast stable-unit as the basis of the count/mass distinction), rather
than from morphological number. On the basis of further hitherto undiscussed
data from Hebrew involving mass nouns with atomic structure, Doron and
Mller argue that the cognitive model of the Mass/Count distinction sketched
by Chierchia (2010) should be improved so as to include what Chierchia calls
fake mass nouns among regular mass nouns, whose atomic structure cru-
cially involves unstable units.
introduction 13
eljko Bokovi and Serkan ener (The Turkish NP) argue for an analysis
of the Noun Phrase in Turkisha Language Without Articlesinvolving no
Determiner Phrase above the NP node, an assumption in line with Bokovis
general theory of Noun Phrase structure developed in his previous works (cf.
Bokovi 2008), and which runs against the DP Hypothesis as developed
by, e.g., Longobardi (2000). Empirical evidence in support of Bokovis NP
Hypothesis is provided for Turkish by the order of constituents within the
Noun Phrase, and by some interesting constraints on interpretation. Bokovi
and ener show that Turkish disallows stranding of possessors, demonstratives,
numerals, and adjectives under ellipsis, a constraint expected under the NP
Hypothesis, since under this theory these elements are part of the NP itself,
hence cannot survive NP ellipsis. The authors however argue that a functional
projection is present above NP in classifier constructions. Classifier construc-
tions allow internal ellipsis within the Noun Phrase, with the elements located
within the Classifier Phrase, hence outside of NP, surviving ellipsis. Bokovi
and ener finally explore the possibility of a functional projection in predi-
cate constructions and demonstrate that several cases which appear to involve
internal ellipsis do not actually do so.
The six articles grouped in Part II take a close look at definiteness
its nature and markersin five typologically different languages or language
groups: West Germanic, Greek, Japanese, Martinik creole, and French Sign
Language (LSF). From a crosslinguistic perspective definiteness appears as a
heterogeneous concept with respect to both morphology (definite articles may
be full, reduced, expletive, cliticised or prefixed) and to semantics, since the
term covers a range of different interpretations depending on the chosen mark-
ers. In one language (LSF), the relevance of semantic Definiteness for linguis-
tic description is overtly questioned.
Rebekka Studler (The morphology, syntax and semantics of definite deter-
miners in Swiss German) scrutinizes the three possible translations of English
the in Swiss German: a strong article, a weak article, and a proximal demon-
strative, all three historically derived from the same demonstrative morpheme.
These three determiners are definite to the extent that they all signal the
referent as uniquely identifiable, but they differ as to their distribution and
interpretations. The strong definite selects nominals construed as anaphori-
cally unique, the weak definite, nominals construed as inherently unique (e.g.
proper names, inalienables, superlatives.), and the proximal demonstrative,
nominals construed as deictically unique. Studler argues that each definite
determiner heads its own syntactic projection within the larger noun phrase.
Patricia Cabredo Hofherr (Reduced definite articles with restrictive rela-
tive clauses) further discusses the competition between full and reduced def-
inite articles in West-Germanic languages (Austro-Bavarian German, Fehring

Frisian, Swiss German dialects), and the traces of such a system in Standard
German, focusing on those definite Noun Phrases which contain a restrictive
relative clause. In the literature on definite determiners in Standard German
and Austro-Bavarian, it is claimed that restrictive relative clauses cannot com-
bine with reduced definite determiners in their Noun Phrase. In other West-
Germanic varieties, however, this restriction does not hold. Cabredo Hofherr
shows that these diverging conclusions are due to two interrelated factors:
(i) the systematic incompatibility between reduced definite determiners and
restrictive relative clauses only concerns contrastive restrictive relative clauses;
(ii) the examples of restrictive relative clauses considered in the different stud-
ies are not of the same type. She argues that while all the languages under
consideration select the full definite determiner with contrastive restrictive rel-
atives, languages differ with respect to definiteness marking with other types
of restrictive relatives. A general tendency is that relative clauses construed as
non-specific favour the selection of the reduced definite article.
Marika Lekakou and Kriszta Szendri (When determiners abound: impli-
cations for the encoding of definiteness) investigate the grammar of definite-
ness in Greek. Their point of departure is the so-called polydefinite construc-
tion, whereby an adjective modifying a noun bears its own definite determiner,
resulting in double (or even multiple) definite determiners within the same
noun phrase. The analysis proposed by the authors crucially draws a paral-
lel between polydefinites and close appositives (e.g. my sister the dancer (not
the writer)), which also involve multiple definite determiners in Greek. The
authors central claim is that multiple definite determiners are semantically
expletive: they instantiate a type of syntactic agreement, while semantic def-
initeness arises from an empty functional head dominating DP. Independent
evidence for this analysis is drawn from pseudo-partitives and PP-modifiers.
Possible counterarguments against the proposed analysis, involving proper
names and nominal co-ordination, are discussed and dismissed.
Makoto Kaneko (The semantics and syntax of Japanese adnominal demon-
stratives) discusses the grammatical properties of the Japanese adnominal
demonstratives a-no, ko-no and so-no. As regards interpretation, he claims that,
while conveying familiarity (an assumed ingredient of semantic definiteness)
by means of the demonstrative prefixes a-, ko- and so-, they lack uniqueness or
maximality (another assumed ingredient of semantic definiteness), and that
the whole demonstrative phrase is existentially quantified. As regards syntax,
he proposes to analyse Japanese adnominal demonstratives as NP-adjuncts, an
assumption supported by three morpho-syntactic properties: (i) the demon-
strative prefixes, ko-, so-, a- systematically display the same morphology as that
introduction 15
of the WH-prefix do-; (ii) the Japanese demonstratives may be preceded by a

restrictive modifier, like other adjunct modifiers; (iii) they behave with respect
to the ellipsis of the following NP as other no-marked expressions clearly identi-
fied as adnominal adjuncts. Kaneko argues that these hypotheses further shed
light on some data from L2 acquisition.
Anne Zribi-Hertz and Loc Jean-Louis (From Noun to Name: on definiteness
marking in Modern Martinik) explore the morphosyntax of definite noun
phrases in Martinique creole, where three different overt morphemes qualify
as definiteness markers since they unambiguously identify a unique referent.
They however differ from one another as to their morphology, distribution, and
interpretive effects. The enclitic determiner -la signals the referent of its DP as
pragmatically definite, in Lbners (1985) sense, viz. as crucially identified by
means of anchoring to the discourse context or utterance situation. Two other
morphemes, l(a)- and l(-), are shown to form semantically definite DPs (in
Lbners sense) which unambiguously identify individual terms independently
of the discourse context and utterance situation: l(a)- is a word-level prefix,
while l(-) occurs either as a free morpheme (with common nouns) or as a pre-
fix (with some country names). The distributional and semantic properties of
l(a)- and l(-) DPs make them similar to definite proper names. The authors
argue that l(a)- and l(-) form a subtype of definite DPs they call Names, char-
acterised by their syntactic and semantic properties regardless of the proper
or common nature of their head noun.
Brigitte Garcia and Marie-Anne Sallandre (Reference resolution in French
Sign Language: the effects of the visuo-gestual modality) seek to identify the
linguistic units which contribute to the construal of nominal reference in
French Sign Language (LSF). A central observation is that these units include
not only lexical signs, but also another type of items the authors characterise
as non-conventional, which frequently occur in actual signed discourse and
involve unlimited creativity on the part of the signer. This second class of
units has been acknowledged in all the works on sign languages reviewed by
Garcia and Sallandre, but the analysis of these items fails to be consensual
among researchers. After having laid out the main assumptions available at
this stage in the specialised literature, the authors present their own semio-
logical approach to non-conventional units. In contradistinction with other
authors working on noun phrases in sign languages, who only take into account
conventional lexical signs, Garcia and Sallandre argue, on the basis of a cor-
pus of attested data from LSF, that a linguistic description of sign languages
should take into account both lexical and non conventional units, as well as
the different ways in which the two types of units alternate and combine in
discourse.
The single article which makes up Part III examines the noun-phrase issue from
the perspective of second-language acquisition, focusing on the means used to
trigger generic or Kind interpretations.
Tania Ionin, Elaine Grolla, Silvina Montrul and Hlade Santos (When arti-
cles have different meanings: acquiring the expression of genericity in English
and Brazilian Portuguese) report on an experimental study of the expression
of genericity in the acquisition of English by native speakers of Spanish and
Brazilian Portuguese, and in the acquisition of Brazilian Portuguese by native
speakers of English and Spanish. English, Spanish and Brazilian Portuguese dif-
fer as to which noun-phrase types are open to generic and kind interpretations.
On the basis of these discrepancies, specific, testable predictions are made
regarding the effects of cross-linguistic influence on the expression of generic-
ity in second-language acquisition. These predictions are tested in a small-scale
study, by means of a written, context-based Acceptability Judgment Task. The
results show that transfer from the learners native language has a limited effect
and is overridden by considerations of register and/or input frequency.
References
Abney, Steven. 1987. The English Noun Phrase in its sentential aspects. PhD diss.,
MIT.
Aguilar, Ana, and Joost Zwarts, 2010. Weak definites and reference to kinds. In Pro-
ceedings of SALT 20, edited by Nan Li and David Lutz, 179196. Ithaca, NY: Cornell
University.
Ariel, Mira. 1990. Accessing Noun-Phrase antecedents. London: Routledge.
Bokovi, eljko. 2008. What will you have, DP or NP? Proceedings of NELS 37, edited
by Emily Elfner and Martin Walkow, 101114. University of Illinois/Urbana-Cham-
paign.
Bokovi, eljko. 2012. On NPs and clauses. In Discourse and grammar: From sentence
types to lexical categories, edited by Gnther Grewendorf and Thomas Ede Zimmer-
mann, 179245. Berlin: de Gruyter.
Bokovi, eljko, and Serkan ener. This volume. The Turkish NP, 102140.
Borer, Hagit. 2005. In name only. Structuring sense, vol. I. Oxford: Oxford University
Press.
Breu, Walter. 2004. Der definite Artikel in der obersorbischen Umgangssprache. In
Slavistische Linguistik 2002, edited by Marion Krause and Christian Sappok, 957.
Mnchen: Otto Sagner.
Cabredo Hofherr, Patricia. This volume. Reduced definite articles with restrictive
relative clauses, 172211.
introduction 17
Carlson, Greg. 1977. A unified analysis of the English bare plural. Linguistics and
Philosophy 1:413457.
Carlson, Greg, Rachel Sussman, Natalie Klein, and Michael Tanenhaus. 2006. Weak
definite noun phrases. In Proceedings of NELS 36, edited by Christopher Davis, Amy
Rose Deal, and Youri Zabbal. University of Massachusetts/Amherst.
Cheng, Lisa, and Rint Sybesma. 1999. Bare and not-so-bare nouns and the structure of
NP. Linguistic Inquiry 30:509542.
Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language
Semantics 6:339405.
Chierchia, Gennaro. 2010. Mass nouns, vagueness and semantic variation. Synthese
174:99149.
Coene, Martine, and Yves DHuls, eds. 2003. From NP to DP, volume 1: The syntax
and semantics of noun phrase, vol. 2: The expression of possession in noun phrases.
Amsterdam/Philadelphia: John Benjamins
Corblin, Francis. 1987. Indfini, dfini et dmonstratif. Constructions linguistiques de la
rfrence. Geneva: Droz.
Corblin, Francis. 2001. Dfini et gnitif: le cas des gnitifs dfectifs. In Cahiers Jean-
Claude Milner, edited by Jean-Marie Marandin, 1954. Paris: Verdier.
Corblin, Francis. 2011. Des dfinis para-intensionnels: tre lhpital, aller lcole.
Langue franaise 171: 5575.
De Mulder, Walter, and Anne Carlier. 2010. The grammaticalization of definite articles.
In The Oxford Handbook of Grammaticalization, edited by Heiko Narrog and Bernd
Heine. Oxford: Oxford University Press.
Dobrovie-Sorin, Carmen, and Brenda Laca, 2003, Les noms sans dterminant dans les
langues romanes. In Les langues romanes: problmes de la phrase simple, edited by
Danile Godard, 235279. Paris: CNRS.
Doetjes, Jenny. 1997. Quantifiers and selection, The Hague: HIL.
Doron, Edit, and Ana Mller. This volume. The cognitive basis of the mass-count
distinction: evidence from bare nouns, 73101.
Ebert, Karen H. 1970. Zwei Formen des bestimmten Artikels. In Probleme und Fort-
schritte der Transformationsgrammatik, edited by Dieter Wunderlich, 159173. T-
bingen: Max Hueber Verlag.
Ebert, Karen H. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem
nordfriesischen Dialekt (Fehring). Brist/ Bredstedt: Nordfriisk Institut, 1971.
En, Mrvet. 1991. The semantics of specificity. Linguistic Inquiry 22:126.
Erteschik-Shir, Nomi. 1997. The dynamics of focus structure. Oxford: Oxford University
Press.
Erteschik-Shir, Nomi. This volume. Information Structure and (in)definiteness, 2351.
Fodor, Janet, and Ivan Sag. 1982. Referential and Quantificational Indefinites. Linguis-
tics and Philosophy 5:355398.
Garcia, Brigitte, and Marie-Anne Sallandre. This volume. Reference resolution in

French Sign Language: the effects of the visuo-gestual modality, 316364.
Gundel, Jeannette. 1988. The Role of Topic and Comment in Linguistic Theory. Ph.D. diss.
U. Texas at Austin.
Gundel, Jeannette, and Thorstein Fretheim. 2005. Topic and Focus. In The Handbook
of Pragmatics, edited by Laurence R. Horn and Gregory Ward, 175196. Oxford:
Blackwell.
Gundel, Jeannette, Nancy Hedberg and Ron Zacharski. 1993. Cognitive status and the
form of referring expressions in discourse. Language 69:274307.
Hawkins, John. 1978. Definiteness and Indefiniteness: A study in reference and grammat-
icality prediction. London: Croom Helm, 1978.
Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Ph.D. diss.,
University of Massachusetts.
Higginbotham, James T. (1985). On semantics. Linguistic Inquiry 16:547593.
von Heusinger, Klaus. 2002. Specificity and definiteness in sentence and discourse
structure. Journal of Semantics 19:245274.
Ionin, Tania, Elaine Grolla, Silvina Montrul and Helade Santos. This volume. When
articles have different meanings: acquiring the expression of genericity in English
and Brazilian Portuguese, 367397.
Jespersen, Otto. 1909. A Modern English Grammar on historical principles. London: Allen
& Unwin.
Kamp, Hans. 1981. A Theory of Truth and Semantic Interpretation. In Formal Methods
in the Study of Language, edited by J. Groenendijk, T. Janssen, and M. Stokhof (eds.),
277322. Amsterdam Center. Amsterdam.
Kaneko, Makoto. This volume. The semantics and syntax of Japanese adnominal
demonstratives, 239268.
Keenan, Edward. 2003. The Definiteness Effect: Semantics or Pragmatics? Natural
Language Semantics 11: 187216, 2003.
Kihm, Alain. 2003. Quy a-t-il dans un nom? Genre, classes nominales, et nominalit.
In Typologie des langues dAfrique et universaux de la grammaire, vol. I, edited by
Patrick Sauzet and Anne Zribi-Hertz, 3964. Paris: LHarmattan.
Klein, Natalie, Whitney Gegg-Harrison, Greg Carlson, and Michael Tanenhaus. 2009.
Special but not unique: weak definite noun phrases. In Semantics and Pragmatics:
from experiment to theory, edited by Ulrich Sauerland and Kazuko Yatsushiro, 264
275. London: Palgrave Macmillan.
Kramsky, Jiri. 1972. The Article and the Concept of Definiteness in Language. The Hague:
Mouton.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA: MIT
Press.
Kuroda, Shige-Yuki. 1965. Generative studies in the Japanese language. PhD diss., MIT.
introduction 19
Kuroda, Shige-Yuki. 1979. The semantics of the Japanese topic marker wa. Lingvisticae
Investigationes 3, 7585.
Lambrecht, Knud. 1994. Information structure and sentence form: topic, focus and the
mental representations of discourse referents. Cambridge UK: Cambridge University
Press.
Lekakou, Marika and Kriszta Szendri. This volume. When determiners abound: im-
plications for the encoding of definiteness, 212238.
Leonetti, Manuel, 2008. Definiteness effects and the role of the coda in existential
constructions. In Essays on Nominal Determination, edited by Henrik Heg Muller
and Alex Klinge, 131162. Amsterdam: John Benjamins.
Lewis, David. 1979. Scorekeeping in a language game. Journal of Philosophical Logic
8:339359.
Link, Godehard. 1983. The logical analysis of plurals and mass terms: A lattice the-
oretical approach. In Meaning, use, and the interpretation of language, edited by
Reiner Buerle, Christoph Schwarze, and Arnim von Stechow, 302323. Berlin: de
Gruyter.
Lbner, Sebastian. 1985. Definites. Journal of Semantics 4:279326.
Lbner, Sebastian. 2011. Concept types and determination. Journal of Semantics 28:
279333.
Longobardi, Giuseppe. 1994. Reference and proper names: a theory of N movement in
syntax and Logical Form. Linguistic Inquiry 25:609665.
Longobardi, Giuseppe. 2000. The structure of DPs: some principles, parameters and
problems. In The handbook of contemporary syntactic theory, edited by Mark Baltin
and Chris Collins, 562603. London: Blackwell.
Lyons, Christopher. 1999. Definiteness. Cambridge, UK: Cambridge University Press.
Massam, Diane (ed.). 2012. Count and Mass across languages. Oxford: Oxford University
Press.
Milsark, Gary. 1979. Existential sentences in English. New York: Garland.
Neeleman, Ad and Reiko Vermeulen. 2012. The Syntactic Expression of Information
Structure. In The Syntax of Topic, Focus and Contrast, edited by Ad Neeleman &
Reiko Vermeulen, 138. Berlin: de Gruyter.
Paul, Ileana. 2009. On the Presence versus Absence of Determiners in Malagasy. In
Determiners: Universals and Variation, edited by Jila Ghomeshi, Ileana Paul, and
Martina Wiltschko, 215242. Amsterdam: John Benjamins.
Pereltsvaig, Asya. 2007. On the Universality of DP: A View from Russian. Studia Lin-
guistica 61: 5994.
Pereltsvaig, Asya. 2013. Noun Phrase Structure in Article-less Slavic languages: DP or
not DP? Language and Linguistics Compass 7:201219.
Pereltsvaig, Asya. This volume. On number and numberlessness in languages with and
without articles, 5272.
Poesio, Massimo. 1994. Weak definites. In Proceedings of SALT 4, edited by Mandy

Harvey and Lynn Santelmann, 282299. Ithaca, NY: Cornell: University.
Ritter, Elizabeth. 1991. Two functional categories in Noun Phrases: evidence from
Modern Hebrew. In Syntax and Semantics 25: Heads and licensing. Perspectives on
phrase structure, edited by Susan Rothstein, 3762. New York: Academic Press.
Walker, Marilyn A., Aravind K. Joshi, and Ellen F. Prince, eds. 1998. Centering in Dis-
course. Oxford: Oxford University Press.
Zribi-Hertz, Anne, and Loc Jean-Louis. This volume. From Noun to Name: on definite-
ness marking in Modern Martinik, 269315.
part 1
Noun Phrase Syntax and Interpretation:

In Search of Crosslinguistic Regularities

Information Structure and (In)definiteness
Nomi Erteschik-Shir
1 Introduction
In this paper, I discuss the two most prominent examples of the interaction
between definiteness and Information Structure (IS): Topicalization has been
associated with definiteness (specificity) and existentials with indefiniteness
(the definiteness effect). Both these phenomena exhibit seemingly idiosyn-
cratic exceptions to this association. This paper demonstrates that these excep-
tions are resolved by a careful analysis in terms of IS. Section 2 defines the
primitives of IS, topic and focus, in terms of their effect on a file system rep-
resenting the discourse manipulation of referents in the common ground. It is
shown how subordinate ISs afford an explanation of the fact that specific indef-
inites can provide topics.
It is well known that IS has an impact on word order. I have argued elsewhere
(e.g. Erteschik-Shir 2007, 2005b) for an account of this interaction in terms of
PF linearization constrained by the canonical IS of a language.1 This is the topic
of section 3 where I demonstrate how topicalization is constrained differently
in Danish, Norwegian, Hebrew, Catalan and Russian in view of their different
canonical ISs as well as other language particular properties. I also show that
although the first position in Germanic languages is generally dedicated to
topics, non-topics in this position also impact IS by forming thetic sentences.
Section 4 offers an analysis of the definiteness effect in existentials. It also gives
some evidence (citing Romance data from Leonetti 2008) that here again there
are differences in canonical IS as well as morphological differences between
the languages.
2 What Is a Topic?
Topics are what the sentence is about and the truth value of a sentence is deter-
mined with respect to them (Reinhart 1981, Strawson 1964). Since sentences
1 For arguments against the idea that IS functional features trigger movement (e.g. Rizzi 1997)
see Erteschik-Shir 2007, 86101.

24 erteschik-shir
may have more than one topic, the main topic (often the syntactically high-
est one, i.e., a subject or one that is topicalized) is the pivot for truth value
assessment. Depending on context, however, any one of the topics in a sen-
tence can play this role. Only referential expressions serve as topics. Topics are
prototypically referential DPs with a discoursal antecedent. Weak (unstressed)
pronouns are therefore by definition topics and can be used to tell which con-
stituent types may function as such. Personal pronouns, temporal and locative
pronouns (then, there) show that DPs and spatio-temporal expressions may
function as topics. Although topics are necessarily given or presupposed, not
all presupposed elements are topics.
Languages mark topics in a variety of different ways. Topics can be marked
by topicalization, by a (clitic) pronoun, morphologically, by topic drop or by
intonation (including destressing). Most languages use several of these options.
In Danish, for example, topicalization is prevalent, but topics can optionally
remain in situ (Erteschik-Shir 2007). Different types of topics may therefore
have different properties cross-linguistically. The following two kinds of topics
are commonly distinguished: continued topics, which refer back to an already
mentioned referent, and shifted topics, which are derived from a restrictive
(d-linked) or contrastive set.2 In Catalan, this distinction applies as follows:
topicalization is reserved for shifted topics but continued topics are postposed
rather than dropped (Barker 2007).3 It has been claimed that dropped topics
are continued topics (Schulz 2003). In the case of languages that employ sev-
eral ways of marking topics, for example both topicalization and topic drop,
there may be a division of labour such that the former applies to shifted top-
ics, whereas the latter applies to continued topics. In some languages, how-
ever, topics selected from restrictive or contrastive sets are distinguished from
continued topics. (Erteschik-Shir, Ibnbari, and Taube 2013) argue that Topi-
calization applies to the former and Topic drop to the latter in both Russian
and Hebrew. (1)(3) ((60)(63) in Erteschik-Shir et al. op. cit.) illustrate this for
Hebrew.
(1) Dani hevi xalav me-ha-super ve-sam ba-mekarer

Dani brought milk from-the-supermarket and-put in-the-fridge
Dani brought milk from the supermarket and put it in the fridge.
2 Frascarelli and Hinterhlzl 2007 distinguish Familiar Topics, Aboutness-shift Topics and Con-
trastive Topics. These are parallel to continued, shifted and contrastive topics respectively.
3 Catalan is discussed in section 3.3 below.
information structure and (in)definiteness 25
(2) Dani hevi xalav ve-tapuxim me-ha-super.

Dani brought milk and-apples from-the-supermarket
Dani brought milk and apples from the supermarket.
(3) et ha-xalav hu sam ba-mekarer

ACC the-milk he put in-the-fridge
He put the milk in the fridge.
In (1) topic drop applies to the continued topic in the second conjunct which
refers back to the object of the first conjunct. The topicalized example in (3)
is licensed in a context such as (2) which introduces the set {milk,apples}, but
cannot occur in the context of the first (italicized) sentence in (1). Topic drop
is also blocked in a context such as (2).
Since the topic is the pivot for truth value assessment, every sentence must
contain at least one topic. This must also be the case for all focus sentences.
Following Gundel 1974 and Erteschik-Shir 1997, such sentences are analysed
as having an implicit or overt stage topic indicating the spatio-temporal
parameters of the sentence (the here-and-now of the discourse).
2.1 Indefinite Topics

Following Kratzer 1995, 1989, stage level predicates may, but need not have a
spatio-temporal argument shown in Erteschik-Shir 1997 to be a stage topic:
(4) Firefighters are (always) available.
(4) can be asserted out-of-the-blue indicating that it is predicated of an implicit

stage topic. It can also be interpreted with the subject as a topic in which case
the predicate is interpreted as a property of the subject. This is not the case
for Individual-level predicates which cannot be interpreted as predicated of a
stage topic:
(5) Dogs are intelligent.
Since no stage topic is available here, the only candidate for topichood in in-
transitive individual-level predicates such as (5) is the subject. Intransitive in-
dividual-level predicates therefore provide an excellent test for topichood. Any
element that can function as a subject in such sentences must qualify as a topic.
As expected, definites are possible topics.4
4 Definiteness is viewed here as a semantic property involving familiarty (see below).

26 erteschik-shir
(6) a. The little boy is intelligent.

b. He is intelligent.
c. John is intelligent.
As shown in (7), indefinites are also possible topics. Only singular indefinites
are excluded:
(7) a. #A little boy is intelligent.

b. Dogs/a dog are intelligent. (only generic)
c. A student I know is intelligent. (specific)
d. A DOG is intelligent, a CAT is not. (contrastive)
e. TWO/SOME (of the) students are intelligent. (partitive)
These facts can be accounted for within a theory of information structure (IS)
which is sensitive to definiteness, keeps track of those discourse referents that
are given and can be topics, and also allows for the introduction of new
potential topics.
Following Reinhart 1981, the common ground is represented by a set of file
cards. Each file card represents a discourse referent. These cards are orga-
nized so that the most recently activated cards are to be found on top of the
stack of cards. These are the discourse referents which provide potential top-
ics in the discourse. In order to get to the top of the stack, the card (the ref-
erent it represents) is focused. This follows implicitly from the definition of
focus:
(8) The Focus of a sentence S = the (intension of a) constituent c of S which

the speaker intends to direct the attention of his/her hearer(s) to, by
uttering S. (Erteschik-Shir 1973, Erteschik-Shir and Lappin 1979)
If the attention of the hearer is drawn to (the referent of) X, then the hearer
(metaphorically) selects the card for X and puts it in a place of prominence,
namely on top of his stack of file cards. The Heimian (Heim 1982) distinction
between definites as old and indefinites as new is incorporated into the filing
system as follows:
(9) a. The card is selected from among the already existing file cards if it is
definite and therefore represents an existing referent.
b. The hearer is required to make out a new card for an indefinite.
The file system thus involves locating cards on top of a stack (topics) or posi-
tioning them there (foci). Additionally, each card is updated with the informa-
tion predicated of it in the sentence. Certain cards are permanently available
on top of the file. These include the card for the speaker and the card for the
hearer and the current stage (the spatio-temporal parameters of the discourse
situation) since these referents are available in any discourse situation. Let me
illustrate with the sequence of sentences in (10):
(10) a. Itop [know a student]foc

b. Shetop [is intelligent]foc
The first person topic of (10a) is located on top of the file and is therefore
licensed as a topic. The focus rule applies to a student, a referential element
within the focus domain. Since this is an indefinite, a new card is made out
for this referent and is then positioned on top of the file. This card therefore
licenses the topichood of the coreferential subject of (10b).
In this system the notions topic and focus are defined discoursally. Partici-
pants in a discourse update their common ground according to the rules of
IS outlined here. Topics and focus in this framework do not project syntac-
tic structure la Rizzi 1997, but are rather integrated at the PF interface. For
discussion of various aspects of this issue see Erteschik-Shir and Lappin 1987,
Erteschik-Shir 2005b, Erteschik-Shir 2006a, b.
Topics, as defined above, are the pivot for truth value assessment. It follows
that topics necessarily take wide scope. The scopal consequences of this view
are discussed in Erteschik-Shir 1997, 1999. (Endriss 2009), a recent proponent
of this view, offers a comprehensive account of the quantificational properties
of topics tying together their semantic, structural, and prosodic properties.
The interpretation of Foci differs from that of Rooth (Rooth 1985, 1992) for
whom a focus (informally) involves selection from a set of alternatives. In
Erteschik-Shir 1997, I argue that only restrictive foci (see below) range over a
discourse defined set of alternatives, but that such foci must be distinguished
from nonrestrictive foci which have different distributional properties.
The main difference between the approach advocated here and syntac-
tic and semantic approaches to IS proposed elsewhere, is the requirement I
impose that all IS properties (syntactic, semantic and prosodic) be derivable
from the two IS primitives, topic and focus as defined here. (These are the only
IS primitives required. Elements which are unmarked for topic or focus, do not
have any status with respect to IS. This is the case for eat in Itop ate an applefoc
in the context of What did you eat?) As shown in the next section, this neces-
sitates subordinate information structures.
28 erteschik-shir
2.2 Subordinate Information Structure

We are now ready to examine the distribution of actual topics in (6) and
(7). The definite referents are acceptable in the examples in (6) if they have
been introduced as foci in the discourse previous to the utterance of the sen-
tences and following (9a), have been selected from the existing file cards and
positioned on top of the stack. Similarly, (7b) is acceptable with the generic
reading, since generics, like names, are definite. Since no card is available for
the singular indefinite in (7a), however, it cannot provide a topic for the sen-
tence.
The distinction between singular indefinites which do not provide valid
topics and the specific, contrastive and partitive indefinites in (7ce), follows
naturally from the rules assuming that more than one topic and focus can be
assigned within a sentence. The next section is devoted to this topic.
Whereas indefinites generally are new to both speaker and hearer and there-
fore do not qualify as topics, one type of specific indefinite is known to the
speaker alone. These specific indefinites contain a modifier which minimally
indicates that the speaker has a particular referent in mind.5 This modifier is
what allows this type of specific indefinite to be a topic as well as a focus.
With this in mind let us examine the information structural properties of
the specific indefinite subject of (7c), repeated here:
(11) A student I know is intelligent.
Since this is an individual-level predicate, the subject is the only possible topic
rendering the IS in (12).
(12) [A student I know]top [is intelligent]foc
The question arises as to how a card for such an indefinite subject can be
placed on top of the file, a requirement for topichood. Once a subordinate IS is
assigned to this constituent an explanation is readily found:
(13) Itop [know a student]foc
5 For more examples of this type see Erteschik-Shir 1997, 4142. Cases of unmodified spe-
cific indefinites occur as well. In such cases the discoursal connection is accommodated. A
detailed discussion of such examples in terms of IS is offered in Erteschik-Shir op. cit., 6167.
As demonstrated in (10a), the first person subject qualifies as a topic and

a new card is made out for the indefinite a student contained in the focus.
This card is updated with the information provided by the relative clause and
positioned on top of the file. This new card is now available as the topic of the
sentence as a whole. The way (12) is derived is by processing it as two separate
sentences, the first derived from the subordinate IS of the subject, the second
equivalent to (12) as shown in (14).
(14) Itop [know a student]foc Shetop [is intelligent]foc
The same IS is assigned as a subordinate IS to the subject of (12).
(15) [[A student]foc Itop know]top [is intelligent]foc
Note that the focushood of a student is evidenced by the fact that it is stressed.
Specificity is therefore accounted for by the information structure assigned
to the modifier. The IS of this clause requires the introduction of a new card
for a student and its placement on top of the file and it also specifies some
information that the speaker has about the student, distinguishing the spe-
cific indefinite from the nonspecific one. Once the subordinate f-structure is
processed, the new card for the indefinite is to be found on top of the file and
therefore qualifies as the topic of the sentence as a whole.6
Partitives (e.g., (7e)) are another type of specific indefinites and are derived
in a similar fashion:
(16) TWO of the students are intelligent.
The utterance of a partitive requires that a set of students is contextually given.

It follows that the set of students provides a topic and that a card for this set
must be available on top of the file. Another property of indefinite partitives is
that the quantifier must be stressed.
The selection of two students from this set is performed by focusing on the
two members of the given set, creating a new card defining two members of the
given set. The following IS results:7
6 Note that a similar set of operations would derive the topic the student I know, the difference
being that the definite requires an existing card for the student in the available stack of cards.
7 The idea that topics may contain foci is introduced already in von Fintel 1994, 58 and Krifka
1994, Krifka 1998, 94,99.
30 erteschik-shir
(17) [[[TWO]foc of [the students]top]]top [are intelligent]foc
All types of specific indefinites are derived by subordinate IS and the resulting
manipulation of the file system. The specific interpretation, the potential for
both topichood and focusability, the contextual requirements as well as the
stress patterns are all derived as a single package.8
The connection between topicality and specificity has a fairly long history
fraught with different views of what a topic is on the one hand and what speci-
ficity is on the other. Important contributions include Gundel 1988, Ward and
Prince 1991, and more formal approaches such as, Cresti 1995, Portner and
Yabushita 2001, von Heusinger 2002 and Endriss 2009. It is not possible in a
short paper to review these contributions although several of them are com-
patible with the analysis I propose. My intention here is merely to demonstrate
that in principle the properties associated with specificity can be derived from
IS as viewed here.
3 Topicalization
It is well know that in the unmarked case, topics (old/given elements) precede
foci (new elements). However, there is little agreement as to what exactly the
relevant information-structural elements are. Often this ordering of elements is
considered to be determined by the relative status of the constituents.9 Under
the view presented here, the various types of topic and focus (contrastive,
restrictive, etc.) are derived from subordinate ISs employing solely the basic
notions of topic and focus employed above. In this way the manipulation of
the file-system is accurately defined and the potential contexts of sentences
with particular ISs is derived.
8 A detailed discussion of other types of specifics and their analysis in terms of IS is offered in
Erteschik-Shir 1997, 6167. As pointed out by an anonymous reviewer it is not obvious that
the subject of (i) should be analysed as a topic.
(i) Less than 10 people on this planet are intelligent.
Here again a subordinate IS is imposed on the subject so that the quantifier less than 10
operates on the given set of people on this planet. See also Erteschik-Shir 1997, 182183 for
an analysis of German data from Krifka 1994 including the scopal properties of similar topics.
9 E.g., Birner and Ward 2009, 1172.
3.1 Danish10
The same constituent types which can be topics in English (6)(7) can be
topicalized in Danish, including pronouns, definites, generics, contrastive ele-
ments, specific indefinites but not non-specific indefinites:
(18) a. Hende mdte jeg i gr.

her met I yesterday
I met her yesterday.
b. Pigen mdte jeg i gr.

girl-the met I yesterday
I met the girl yesterday.
c. Blomster ser man om forret.

flowers sees one in spring-the
Flowers, one sees in the spring.
d. Pigen mdte jeg i gr, drengen mdte jeg frst i dag.

girl-the met I yesterday, boy-the met I first today
The girl I met yesterday, the boy I only met today.
e. En pige som jeg mdte i gr gav jeg en god bog til.

A girl that I met yesterday gave I a good book to
A girl I met yesterday I gave a good book to.
f. *En pige mdte jeg i gr

a girl met I yesterday
I met a girl yesterday.
g. *Hospitalet tog Peter p

hospital-the went Peter on
Peter went to the hospital.
Pronouns are by definition topics, hence they can be topicalized as in (18a).

Definites can (but need not be) topics and can therefore also be topicalized,
cf. (18b). Bare plurals can be topicalized under their generic interpretation
since, according to Cohen and Erteschik-Shir 2002, bare plural topics (18c) are
10 The material in this section is to a large extent drawn from Erteschik-Shir 2006a.
32 erteschik-shir
interpreted generically and bare plural foci are interpreted existentially.11 (18d)
illustrates topicalized contrast. (18e) shows that specific DPs can be topical-
ized whereas nonspecific, singular indefinites (18f) cannot. (18g) finally demon-
strates that weak definites do not tropicalize.12
A surprising constraint on Danish topicalization is that Danish does not tol-
erate ambiguity. V-second and topicalization result in a potentially ambiguous
string: DP1 V DP2, where DP1 could in principle be either the subject or the
object. In fact, in cases of potential ambiguity, Danish allows only one interpre-
tation, the one in which DP1 is the subject and DP2 is the object. (19), therefore
cannot be an instance of topicalization of the object.13
(19) *Peter mdte Sara

Peter met Sara
Sara met Peter.
Even in a context which enhances the object reading of the initial DP, (19),
under this reading, is not licensed:
(20) Hvem var det Sara mdte, Peter eller Thomas?

Who was it Sara met, Peter or Thomas?
The context forces a contrastive reading of the topicalized object and still
allows for a topic reading of the subject, yet Danish informants reject the
sentence and necessarily interpret Peter as the subject.
Danish pronouns are case-marked as they are in English. Pronouns therefore
identify subjects and objects and if either the subject or the object or both are
pronouns, topicalization is licensed:
(21) a. Ham mdte Sara igr.

him met Sara yesterday
Sara met him yesterday.
11 Singular indefinites can also be employed generically as topics. The different cross-linguistic
expressions of genericity will not be dealt with here. Still it is predicted that generics can
function as topics cross-linguistically.
12 According to Carlson et al. 2006, weak definites are not in fact semantically definite but are
rather akin to bare count singulars.
13 Note that the presence of auxiliaries and negation disambiguates the DP V DP string. See
Raviv 2005 and the references cited therein for details.
b. ?Peter mdte jeg igr.

Peter met I yesterday
I met Peter yesterday.
c. Ham mdte hun igr.

Him met she yesterday
She met him yesterday.
These data show that topicalization in Danish is not only restricted to con-
stituents that qualify as topics, the result must also lead to an unambiguous
parse. This constraint cannot be defined in terms of features of the fronted ele-
ment in view of the fact that it is not necessarily the fronted element itself that
bears case-marking: The case-marked subject pronoun (21b) licenses topical-
ization even though the fronted element itself is not case-marked. Note that
the slight degradation of this sentence is likely to be due to the fact that the dis-
ambiguating DP comes later in the sentence in this case, so the correct parse is
signaled late in the sentence, resulting in a garden path effect with the fronted
element parsed as the subject.
Topicalization also does not render ambiguity when one of the arguments is
inanimate:
(22) a. Den skuffe malede drengen igr.

That drawer painted boy-the yesterday
b. ?Peter forskrkkede lynet meget.

Peter frightened lightning-the a lot.
Here again, when the disambiguating inanimate is the subject as in (22b), the
sentence is somewhat degraded since the same garden path effect is triggered
as in (21b).
The requirement of an unambiguous parse cannot be accounted for by syn-
tax and is most naturally construed as a parsing constraint on the identification
of arguments at the interface with the articulatory-perceptual system, PF. I pro-
pose the following constraint for Danish:
(23) In a string, X Y, where X and Y are arguments, ID X as subject and Y as

object if neither is marked otherwise.
Parsing would be facilitated even further if no dislocation were to take place.

Dislocation must therefore also serve some function. In Danish, topicalization
34 erteschik-shir
serves to identify the topic. (Focus is marked intonationally as in English). It

is the need to identify both syntactic roles (subject, object) and information-
structural roles (topic, focus) which lays a heavy burden on the parser in a
language such as Danish which lacks both agreement and case-marking mor-
phology.
A comparison with parallel data in Norwegian renders a surprising result. As
demonstrated by Raviv 2005, ambiguous topicalized sentences in Norwegian
are licensed and contextually resolved. This difference between the syntacti-
cally similar Norwegian and Danish, according to Raviv, is due to the morpho-
logical differences between the pronoun systems of Norwegian and Danish.
Whereas the Danish pronoun system distinguishes nominative and accusative
pronouns morphologically, this is not the case in Norwegian (from Raviv op.
cit. 57):
(24) Han/hun/dere s Jonas.

Him/her, you.pl saw Jonas
Jonas saw him/her/you.pl
He/she/you saw Jonas
The morphological ambiguity of the Norwegian pronouns allows both read-

ings. It follows that the two languages are distinct with respect to their toler-
ance for ambiguity. Raviv argues that the Norwegian tolerance for ambiguity
both with respect to pronouns and with respect to proper names and DPs,
all unmarked for case, follows from the ambiguity in the pronoun system in
Norwegian. Pronoun topicalization is very common in both languages and
results in an unambiguous parse in Danish, but not in Norwegian. If Norwe-
gian were subject to the parsing constraint in (23), topicalization would be very
restrictedan unhappy result if topicalization plays a critical role in marking
IS in the language. Still, even in Norwegian, the unmarked word order is SVO
and this reading will be assigned (as it is in Danish) in ambiguous strings which
are not contextually disambiguated. Contextual disambiguation is, however,
licensed in Norwegian in contrast to what we saw in Danish.
Although only Danish adheres to (23), the preverbal argument is the un-
marked topic in both languages. Subjects in the preverbal position also provide
topics, all else being equal. Since topics are defined as what the sentence is
about, a response to tell me about X will necessarily have X as its topic. The fol-
lowing interchange shows that subjects are indeed unmarked topics in Danish:
(25) A. Fortl mig om Peter/Hvad med Peter?

Tell me about Peter/How about Peter?
B. Han er forelsket i Marie.

He is in love with Marie.
B. ??Marie er forelsket i ham.

Marie is in love with him
B. I ham er Marie forelsket.

With him is Marie in love
What is wrong with B is that the topic him is not preverbal.

I use the term canonical information structure to identify the unmarked
alignment of syntactic structure and IS in a particular language. The canonical
IS for Danish (as well as Norwegian) is:
(26) Canonical IS (Danish):

Xtop V [ Y ]foc
The preverbal element (be it a subject, or a topicalized element) is the topic

and the focus is postverbal.14
As pointed out by an anonymous reviewer, the response B in (25) accords
with the canonical IS of Danish, since the topic is included in the topicalized
PP I ham.
Canonical IS thus constrains word order and so avoids an ambiguous parse
of the linear string. Canonical word order can be violated on condition that the
sentence is disambiguated by morphology (Danish) or by context (Norwegian).
3.2 Topicalization in Hebrew

Hebrew topicalization is possible in two different word orders, OSV (27a) and
OVS (27b).
(27) a. et hasefer moshe kana. (et marks definite objects)

the-book Moshe bought
b. et hasefer kana moshe.
14 Note that the topic can also be a spatio-temporal phrase:
(i) I morgen kommer Peter.

Tomorrow comes Peter
Peter is coming tomorrow.
36 erteschik-shir
The examples in (28a) and (28b) show that indefinite (nonspecific, noncon-
trastive) DP subjects which, as shown above, cannot provide topics, must be
postverbal:
(28) a. *et hasefer yeled exad kana

the-book boy one bought
Some boy bought the book.
b. et hasefer kana yeled exad.
(29a) and (29b) show that a (non-deictic) subject-pronoun cannot occur post-
verbally.
(29) a. et hasefer hu kana.

the-book he bought
b. *et hasefer kana hu
These data reveal that the alignment between syntactic and IS structure is
such that the topic must be preverbal and the focus postverbal as shown in
(30).
(30) Canonical IS (Hebrew)

Xtop V [ Y ]foc
The canonical IS of Hebrew is therefore parallel to the one argued for in Danish
(26). There is an important difference between the two languages, however:
Danish requires V-2, whereas Hebrew allows OSV as well.15
3.3 Topicalization in Catalan

Catalan topicalization has in common with Danish topicalization that topical-
ized elements must be given, yet only new topics can be fronted. Vallduv 1992
refers to these as Links. Old topics are detached to the right (Vallduvs Tails).
The following examples are from Vallduv 1992:
15 In both languages topicalization is optional. Hebrew topicalization is in most cases con-

trastive. The two word orders (27a) and (27b) differ in that the subject is in focus in (27b),
but not in (27a).
(31) [written on an aerogram; first line on the extra space overleaf]

a. Amb-aquest-tros-de-paperet1 ja no hi1 comptava t1.
with-this-little-piece-of-paper anymore no obl 1s-impf-count-on
This-little piece-of-paper I wasnt counting on anymore.
b. Quant al Joan i la Isidora no tho s dir, doncs

as-for the J. and the I. no iobj.obj 1s-know to-say since
el Joan1 el1 veiem t1 ben poc.
the J. obj 1p-see quite little
As for Joan and Isidora I cant say, since Joan we see very little of.
Catalan topicalization has in common with Danish topicalization that the

topicalized element must be old, yet, according to Vallduv, there is a differ-
ence. In Catalan only elements that are not topics in the previous sentence
are topicalized. Such new topics are called shifted topics. Vallduv calls these
fronted elements Links and views them as address pointers in a file sys-
tem in which new information is listed under the address specified by the
Link. Since a Link is a command to go to an existing address, Links only
appear when there is a change of address. In the examples in (31), the fronted
(underlined) elements are all new topics in this sense. The Link in (31a) is
made deictically available, the one in (31b) is restrictively selected from the set
{Joan, Isidora}. Prime cases of Links are contrastive or members of restrictive
sets.
A strategy parallel to the left dislocation of Links is employed in sentences
introduced by as for X in English and similar constructions in other lan-
guages.
(32) As for X, X is nice.

LINK
As for X can only be used in a context in which X is a member of a contextually

available set. The initial phrase selects X from this set and focuses it, thus
making it a (pronominal) topic of the following sentence.
Brunetti 2009 makes the same point concerning Italian and argues that If
it turned out that preverbal subjects that function as links are left dislocated,
then Italian would display a perfect matching between the IS construction and
the syntactic construction. (759) In other words, the canonical IS for Italian
(and Catalan) is one in which shifted topics are left dislocated, and topics
which have an antecedent in the discourse are detached to the right. In the
38 erteschik-shir
theory of IS promoted here, shifted topics have a subordinate IS and are easily
distinguishable from old topics which do not, and for which cards must be
available on top of the file.
3.4 Topicalization in Russian

Languages like Russian which dont mark definiteness morphologically employ
word order for this purpose. Topicalization marks definiteness and focusing
marks indefiniteness.16 The examples in (33)(35) illustrate topicalized DPs not
marked for definiteness.
(33) Devuku Petja vstretil vera.

girl.acc Peter met yesterday
Peter met the girl yesterday. (not: Peter met a girl yesterday)
(34) vety my vidim tolko vesnoj.

Flowers we see only in-the-spring
We see flowers only in the spring.
(35) Devoke kotoruju ja vstretil vera ja dala xorouju knigu

girl.dat who.acc I met yesterday I gave good book
The girl who I met yesterday I gave a good book to?
(33) receives only a definite reading, even in a contrastive context. In (34),

flowers must be interpreted generically and in (35), even though the fronted DP
is modified, it must be interpreted definitely and does not receive the reading
of a specific indefinite. In Russian IS (via word order) therefore plays a critical
role in determining definiteness.
According to (King 1995):2 the various surface orders found in Russian are
predictable from discourse factors although word order alone is not responsible
for encoding grammatical functions. King employs the following well-known
example from (Jakobson 1971) to illustrate the fact that context may overrule
word order in determining grammatical functions:
(36) Mat ljubit do

Mother.nom/acc loves daughter.nom/acc
Mother loves (her) daughter.
16 See Erteschik-Shir and Strahov 2004.

Generally, in Russian, case marking disambiguates the grammatical func-

tions of the arguments. This sentence, however, is potentially ambiguous since
the cases of both the preverbal subject and the postverbal object NPs are
ambiguous between nominative and accusative. In principle, the sentence
should therefore be interpretable with either argument as subject or object. In
fact, the unmarked interpretation follows the unmarked SVO word order with
mother as the subject. According to King, however, in an appropriate context,
the sentence can also be understood with daughter as the subject. Russian
thus employs the same strategy as Norwegian, in which the unmarked reading
takes the initial DP as the subject and only context can license the interpreta-
tion in which the initial DP is interpreted as a topicalized object.
3.5 Parameters
Topicalization in Russian plays a more critical role than in the other languages
mentioned here, in that it facilitates the interpretation of NPs in terms of def-
initeness.17 This is probably why a specific reading of the topicalized element
in (35) is excluded. Here the connection between the lack of definiteness mor-
phology and the role of topicalization is transparent. In languages such as Dan-
ish, the role of topicalization is limited to marking IS. Indefinites can therefore
topicalize as long as they are specific or contrastive. The properties of topi-
calization cross-linguistically are therefore not easy to parameterize. Not only
do morphological properties of the language play a role (definiteness marking,
case marking, pronoun paradigms, etc.), so does V-second syntax and prosody
as argued by Speyer 2005.
According to Speyer, the verb-second constraint was lost in English in the
course of the Middle English Period. During the same time frame, the rate at
which direct object noun phrases topicalize also declines. Speyer poses the
question as to why the rate of topicalization should decline parallel to the loss
of V2. Speyer notes that topicalization is motivated by pragmatic reasons and
that it is unlikely that the conditions of language usage change over time. The
decline in topicalization is therefore surprising. Speyer found that the decline
of topicalization with full DP subjects is continual but the decline of topical-
ization with pronoun subjects is less pronounced and stops with the transition
from Old English to Middle English Grammar. Since the decline affects pro-
noun subjects and full noun-phrase subjects differently, Speyer figures that
17 Whether or not Russian has DPs is debatable (e.g., Bokovi and Gajewski 2011, Pereltsvaig
2007), therefore, for convenience, I use NP for Russian.
40 erteschik-shir
prosody must be a factor. He makes the following comparison between topical-

ization in German and English:
(37) a. Hans hasst Bohnen. Erbsen hasst Maria.

b. John hates beans. Peas, Mary hates.
Topicalized elements are generally selected from a contextually evoked set and
are therefore accented. This is the case in both languages. The German sen-
tence (37a) is unobtrusive. The English sentence (37b) is awkward. It requires
a little break between the two accents. This looks as ifat least in Englisha
weak element between two accents is compulsory. Speyer calls this require-
ment the Trochaic Requirement. He views inversion in German as a handy
way to avoid violation of the Trochaic Requirement. Modern English, since
it has lost V-second, no longer has this option. The loss of topicalization,
according to Speyer, therefore follows naturally from the loss of V-second since
topicalization without V-second violates the Trochaic Requirement. Modern
English therefore uses topicalization sparingly compared to the other lan-
guages reviewed here. Instead, a strict alignment between syntactic structure
and IS is required resulting in the canonical IS for English shown in (38) (where
s and t stand for the spatial and temporal parameters, respectively in sTopt in
(38b)).
(38) Canonical IS (English)

a. SUBJECTtop []foc
b. sTOPt []foc
In the canonical structure in (38a), the subject aligns with the topic and the
predicate with the focus. (38b) is an all-focus sentence with an implicit or overt
stage topic. Syntactic constituent structure is again aligned with IS. The only
marked case is one in which the object is the topic. Evidence for this being the
case is given in the dialogue in (39), parallel to the Danish in (25).
(39) A. Tell me about John:

B. He is in love with Mary.
B. ??Mary is in love with him.
Similarly, a response to What happened? triggers the out-of-the-blue, all-focus

reading in (40)
(40) John is in love with Mary.

3.6 Fronting Non-Topics

According to Frey 2006, not all initial elements are topics. Such cases are illus-
trated in (41). Here, according to Frey, the fronted elements are not referential
and require contextualization:
(41) a. Fast jeden Kollegen findet der berhmte Linguist

nearly every.acc colleague thinks the famous linguist
sympathisch
(is) nice
The famous linguist thinks nearly every colleague is nice.
b. In einem Garten hat Maria den Hund gefttert

in a garden has Mary the.acc dog fed
Mary fed the dog in a garden.
According to Frey, (41a) in the context (42) is perfect:
(42) Hans fhlt sich wohl an seinem neuen Arbeitsplatz.

Hans feels refl fine at his new working place
Note that the preceding sentence introduces the place of work into the con-
text, allowing it, and any of its natural parts to be the topic of the following
sentence. Colleagues are clearly a natural part of a place of work and therefore
qualifies as the topic of (41a). This follows if elements which are subsets of pre-
viously mentioned constituents are defined as topics, a definition not adopted
by Frey. Frey does not supply a context for (41b) but assumes that the preposed
locational PP cannot be a topic. Yet in a framework which allows stage topics, a
PP may in fact play the role of a topic. (41b) cannot answer the question: Where
did Maria feed the dog? It follows that the PP does not play the role of focus.
Moreover, the fact that (43) is well-formed indicates that the PP in fact must be
the topic, since the other sentence constituents are indefinite and therefore do
not qualify as potential topics, and a sentence must have at least one topic to
be interpreted.
(43) In einem Garten hat ein Mdchen einen Hund gefttert.

in a garden has a girl a dog fed
It follows that (41b) and (43) are predicated of a stage topic in which a garden
restricts the location defined by the discoursally-available current stage.
Yet, arguments for topichood do not extend to (44) and similar examples
42 erteschik-shir
in which the fronted elements are argued by Frey to have no information-

structural impact.
(44) Leider hat keiner dem alten Mann geholfen.

unfortunately has nobody theDAT old man helped
Unfortunately, nobody has helped the old man.
(44) exemplifies a sentence adverbial which cannot be topical. That adverbs

such as leider are not likely topics is uncontroversial, yet it is not obvious that
sentences in which such an adverb is fronted and sentences in which it is not
are equivalent from an information structural perspective. A straightforward
conclusion is that fronting an element which does not function as a topic
indicates that none of the other elements in the sentence is to be interpreted as
a topic. This would leave the sentence topicless, not an option in a framework
in which every sentence must have a topic for truth-value assignment to take
place. The only other option is that sentences such as these must be predicated
of a stage topic. The sentences in (45) show that an initial adverb is possible
in sentences in which all the arguments are indefinite and therefore cannot
be interpreted as topics. This proves that these sentences are indeed to be
interpreted as having a stage topic.
(45) a. Glcklicherweise hat ein Mdchen eine Leiter mitgebracht.

Fortunately has a girl a ladder brought
Fortunately a girl brought a ladder.
b. Leider hat ein Hund einen alten Mann gebissen.

Unfortunately has a dog an old man bitten
Unfortunately, a dog bit an old man.
I conclude that fronting a non-topic marks the sentence as having a stage

topic in German. It follows that, counter Frey, fronting adverbs to the left
periphery does have information-structural impact. This impact however is not
associated with the fronted element itself. Note that interpreting a sentence
with a stage topic does not exclude additional topics in the sentence. (44), for
example, could be uttered in a context in which dem alten Mann is also a topic.
The following Danish data gives the same results:
(46) a. Desvrre kom Hans/han ikke til selskabet.

unfortunately came Hans/he not to the party
Unfortunately Hans/he didnt come to the party.
b. Hans/han kom desvrre ikke til selskabet.

Hans/he came unfortunately not to the party
Hans/he unfortunately didnt come to the party.
Here again, only (46a) can be employed out of context. (46b), however, requires
that the subject is interpreted as a topic, and is therefore a good response to Tell
me about Hans. (46a) is not a possible continuation in this context. We can
therefore conclude that in these Germanic languages, when the initial element
does not qualify as a topic, the sentence is interpreted with an implicit stage
topic. Fronting a non-topic signals a particular IS, namely one in which none
of the overt elements is a topic.
Svenonius 2004 claims that the initial position is not a simple topic position
as argued above, he claims it is a switch topic. He includes in this category
contrastive foci, speaker-oriented adverbials, discourse connectives, scene-
setting adverbials and actual switch topics. If no switch topic is available, a con-
tinued topic, often the subject, is placed in initial position, and if neither a shift
topic nor a continued topic is available, an expletive may appear. Svenoniuss
description of the elements in initial position can be captured by the follow-
ing generalization: The initial element in Germanic is either a topic or else the
sentence is interpreted as having a stage topic. Since the class of topics includes
continued topics, switch topics, contrast, and overt stage topics, Topicalization
in Germanic in these cases can be seen as motivated by the movement of a
topic. It is only when the fronted element is itself not a topic that such motiva-
tion fails.
. Kiss 2004 demonstrates that the placement of speaker oriented sentential
adverbials in English also determines IS (her (32):117):
(47) a. *[TPA baby boy luckily was born]

b. Luckily [TP a baby boy was born]
c. John luckily [TP was born on time]
(47a) shows that a non-specific subject cannot be followed by a sentence

adverbial and (47b) shows that the adverb must precede such a subject. (47c)
shows that a specific subject may precede the adverb. According to . Kiss, the
preverbal subject in this example is a topic. (47b), however, is predicated of an
implicit stage topic. This data is similar to the German data in (45). There too
an initial sentence adverb indicates a stage topic. The position of the adverb
therefore identifies the IS of sentences in English as well with ramifications for
the definiteness of the subject. A preadverbial subject must qualify as a topic
and be definite (or at least specific), but a postadverbial subject need not be.
44 erteschik-shir
All the cases listed in this section are illustrations of a left-peripheral ele-
ment which does not itself have information-structural properties, yet sig-
nals a particular information structure, namely one with a stage topic. These
structures also have another property which may explain their information-
structural status: Due to the fact that the left-peripheral element is not an
argument, the subject is necessarily postverbal, a position in which it cannot
be interpreted as a topic. The motivation for these constructions may there-
fore be to oust the subject from initial position in order to enable its inter-
pretation as a non-topic, triggering, in this case, an out-of-the-blue interpre-
tation.
One of the issues that all authors who discuss left-peripheral elements such
as these, is how to motivate their movement, in view of the fact that the ele-
ment that moves cannot itself be identified with particular IS properties. A
different view on this problem, which raises different theoretical issues, is to
take seriously the particular alignment properties of the languages in ques-
tion and to examine whether the resulting alignments are in fact canonical.
If so, the motivation for a particular word order would be to promote a canon-
ical alignment. Clearly, such motivation is non-syntactic, forcing IS-motivated
word order to occur at PF where both IS properties and linear order are visible.
This view has the advantage of simplifying syntax and allowing linear reorder-
ing at PF.
4 Focusing the Subject: Existentials
One way of marking a sentence as being all-focus and having a stage topic is
therefore for a non-topic to occupy the left peripheral position. According to
. Kiss 2004 and Holmberg 2000, existentials employ exactly this strategy. As
argued in Erteschik-Shir 2007, the outcome is an all-focus sentence predicated
of a stage topic. (A parenthesized spatial or temporal index is one which is
missing contextually):
(48) a. (s)Topt[There is a/*the dog in my garden]foc

b. sTop(t)[There is a/*the meeting at two oclock]foc
c. (s)Topt[There are many/*all people who like icecream]foc
In such an IS, the full sentence is entered on the card for the current here-
and-now which provides the stage topic and an all-focus sentence is derived.
What is special about the stage topic in existentials is that it is lacking in con-
textual definition: either the place or the time are not contextually available
and a new stage is defined by adding these parameters to the stage. This can
be seen in (48). In (48a), the location is not given contextually and in (48b),
the time is missing in the context. In (48c), no locative parameter is contex-
tually available, yet this parameter is not provided in the sentence either, the
new stage is accommodated to mean the whole world. The definition of a new
stage requires new inventory. Definites presuppose a referent associated with a
location. Located referents are therefore incompatible with the interpretation
of a new stage. This is the explanation for the definiteness effect in existen-
tials.18
The definition of a new stage in this way also provides an explanation for
when the definiteness effect applies. (49) illustrates examples in which it does
not hold:19
(49) a. Theres city hall, the museum, and the park.

b. Theres the meeting at 2 oclock and the office event at 4.
Such existentials generally provide a list of elements contained in a certain

place, or time: (49a) could be a description of the sights in a given town. (49b)
could be a response to a request for the days schedule at the office. In both
cases, the context must include reference to the stage in question, namely the
town, and the office events, respectively, but whats special about these stages
is that they are unpopulated. The inventory which is listed in the existential
may be given, yet it is new to the stage in question. An obvious difference
between the sentences in (48) and (49) is that the former lack at least one of the
18 The same definiteness effect is also found in locatives such as (i) and (ii) but not in possessives
such as (iii):
(i) My souptop [has a/*the fly in it]foc

(ii) Johntop [has a/*the hat on]foc
(iii) Johntop [has a/the hat (in his hand)]foc
In (i) and (ii) the subjects are interpreted as locations and therefore function as stage topics.
Their IS is therefore parallel to that of the sentences in (48) in that these stage topics also
require the filling in of the location by a prepositional phrase. (iii) differs in that the subject
is interpreted as a possessor and not as a location. The definiteness effect does not apply and
the addition of a locational prepositional phrase is optional.
19 The literature on the definiteness effect originating with Milsark 1974 is vast and will not be
reviewed here. Leonetti 2008 offers an excellent review of literature on the definiteness effect
as well as work that characterizes definites in existentials.
46 erteschik-shir
parameters of the stage, the latter require full contextual specification of the
stage (e.g., for (49a), a particular city, and for (49b), a particular day at work).
Since the stage is not new, the inventory on it needs not be new either. The
definiteness effect is therefore predicted to hold only of new stages.
The contextual difference between existentials of the first type in which the
DE holds and those of the second type in which it doesnt, also plays a role in the
IS of the sentence as a whole. Whereas the first type is predicated of an (at least
partially) unindexed stage, one for which the spatio-temporal parameters are
not contextually specified, the stage topic of the second type is fully specified
contextually. It follows that as part of the focus in the first kind, the missing
spatio-temporal parameter(s) must be specified, which is why such sentences
are incomplete without their coda. This is illustrated in (50) for the examples
in (48).
(50) a. *(s)Topt[There is a dog]foc

b. *sTop(t)[There is a meeting]foc
c. *(s)Topt[There are many people]foc
(50c) is somewhat different from the other two. It can easily be completed
by a locative, but the coda, in (48c), is a relative clause. What is wrong with
(50c) is therefore not that a missing locative must be filled in, but that without
some added information the sentence is incomplete, it is missing a contentful
focus. One way to remedy this is to add a location, another is to add a relative
clause, and a third is to supply a contrastive context in which many people is
contrasted with few people, in which case many will be stressed. Existential
sentences which are subject to the DE therefore generally include a coda as
part of the focus.
As shown in (49a), this is not a requirement for existentials of the second
kind. Here the location is part and parcel of the stage, and the focus intro-
duces the inventory on this unpopulated but given stage. No coda is therefore
required. In (49b), a coda is (optionally) present. This coda is however pack-
aged differently with respect to IS. Compare (51) and (52):
(51) sTopt [Theres [the meeting at 2 oclock]] foc

(52) sTop(t) [There is a meeting at 2 oclock]foc
Leonetti 2008 cites Rando and Napoli 1978 as distinguishing the codas in the
two cases. With the definite in (51) the (optional) postnominal constituent is
parsed as a nominal modifier and is not a real coda. What this means in the
current framework is that he meeting at 2 oclock is the element introduced on
stage. In the existential in (52), however, what is introduced on the new stage
is a meeting, the coda at 2 oclock functions to specify the missing temporal
parameter of the stage.
The more fine-tuned view of the properties of stage topics developed here
provides a way of distinguishing the different types of existentials and their
properties. This is missing in Leonetti 2008s inspiring paper. Leonetti demon-
strates that not all properties of existentials can be derived from the IS of the
construction. Other cross-linguistic factors play a crucial role as well, in par-
ticular the language particular encoding of information structure: Taking into
account the principles of information structure in each language is essential
for our understanding of the link between syntactic positions and definiteness.
(p. 139) This is very much in line with the view taken above with respect to top-
icalization.
Leonetti addresses the seeming non-adherence of Italian and Catalan to the
definiteness effect illustrated in (53) and (54) respectively (p. 134).
(53) C un cane. / C il cane. / C Gianni.

Cl-is a dog. / Cl-is the dog / Cl-is John
(54) Hi ha un gos. / Hi ha el gos. / Hi ha en Joan.
Cl has a dog / Cl has the dog / Cl has the John
The first point Leonetti makes and argues for is that the Definiteness Effect is in
fact operative in Italian and that constructions with esserci such as C Gianni
conflate two different constructions: the existential construction and a locative
construction. The Definiteness Effect shows up only in the former.
The second and more significant point he makes is the observation that the
presence of the locative coda inside the VP blocks the insertion of definite DPs:
these are excluded unless the locative coda is itself (right/left-) dislocated (or
removed). (Coda Constraint, p. 142).20 It follows that the Definiteness Effect
shows up in these languages as long as the Coda is information-structurally
integrated with the DP.
Languages, according to Leonetti, differ with respect to their coda effects:
Italian, Catalan and French, disallow the insertion of definites in existentials
(they adhere to the Coda Constraint), whereas Spanish and possibly Romanian
allow it. According to Leonetti, this is because thetic (all-focus) sentences resist
20 As shown in section 3.3, Catalan marginalizes topics by left-dislocating shifted topics and
right-dislocating continued topics.
48 erteschik-shir
subordinate ISs within them. Languages which nevertheless allow internal

topic-focus partitions within such a focus, do so, because they do not have such
marginalizing devices.
Leonettis account of the different cross-linguistic expressions of the DE in
Romance languages is therefore very much along the lines of discussion of
cross-linguistic differences in topicalization discussed here. In both cases the
differences boil down to differences in canonical IS as well as other idiosyn-
cratic differences between the languages.
5 Architecture
Cross-linguistic variations in Topicalization constraints and the Definiteness

effect in existentials are viewed here as being due to differences in the mapping
of Information Structure to Syntactic Structure in the languages examined.
This analysis requires a view of Topic and Focus assignment as part and parcel
of the externalization system, and argues against syntactic approaches such
as the cartographic approach which allows for syntactic projections of topic
and focus features. It therefore provides support for the view expressed in
Berwick and Chomsky 2011 (among others) that the externalization system
(PF) is responsible for at least microvariation. It also follows that displacement
is constrained by the externalization system and not by the computational
system as generally thought. This approach is particularly appropriate since
by tying the effects to information structure, it allows for the great variety in
empirical findings both within and across languages.
References
Barker, Chris. 2007. Parasitic scope. Linguistics and Philosophy 30 (4):407444.

Berwick, Robert, and Noam Chomsky. 2011. Biolinguistics: The current state of its
evolution and development. In Biolinguistic investigations, edited by Anna Maria
Di Sciullo and Cedric Boeckx, 1941. Oxford: Oxford University Press.
Birner, Betty J., and Gregory Ward. 2009. Information Structure and Syntactic Struc-
ture. Language and Linguistics Compass 3 (4):11671187.
Brunetti, Lisa. 2009. On links and tails in Italian. Lingua 119 (5):756781. doi: 10.1016
/j.lingua.2008.10.005.
Definite Noun Phrases. In Proceedings of NELS 36, edited by Christopher Davis, Amy
Rose Deal and Youri Zabbal, 179196. UMass/Amherst: GLSA.
Cohen, Ariel, and Nomi Erteschik-Shir. 2002. Topic, Focus and the Interpretation of
Bare Plurals. Natural Language Semantics 10:125165.
Cresti, Diana. 1995. Indefinite topics. Ph.D. dissertation, Department of Linguistics &
Philosophy, University of Massachusetts, Amherst.
. Kiss, Katalin. 2004. The EPP in a Topic-Prominent Language. In Subjects, Expletives,
and the EPP, edited by Peter Svenonius, 107124. Oxford: Oxford University Press.
Endriss, Cornelia. 2009. Quantificational Topics: A Scopal Treatment of Exceptional Wide
Scope Phenomena Berlin: Springer.
Erteschik-Shir, Nomi. 1973. On the nature of island constraints. Ph.D. Dissertation, M.I.T.,
Cambridge.
Erteschik-Shir, Nomi. 1997. The Dynamics of Focus Structure. Cambridge: Cambridge
University Press.
Erteschik-Shir, Nomi. 1999. Focus Structure and Scope. In The Grammar of Focus,
edited by Georges Rebuschi and Laurice Tuller, 119150. Amsterdam: John Ben-
jamins.
Erteschik-Shir, Nomi. 2005b. What is Syntax? Theoretical Linguistics 31 (12):263274.
doi: doi:10.1515/thli.2005.31.12.263.
Erteschik-Shir, Nomi. 2006a. On the architecture of topic and focus. In The architecture
of focus, edited by Valria Molnr and Susanne Winkler, 3357. Berlin: Mouton de
Gruyter.
Erteschik-Shir, Nomi. 2006b. Whats what? In Gradience in Grammar, edited by Car-
oline Fry, Gisbert Fanselow, Matthias Schlesewsky, Ralf Vogel, 317335. Oxford:
Oxford University Press.
Erteschik-Shir, Nomi. 2007. Information Structure: The Syntax-Discourse Interface. In
Syntax and Morphology, edited by Robert Van Valin. Oxford: Oxford University Press.
Erteschik-Shir, Nomi, Lena Ibnbari, and Sharon Taube. Erteschik-Shir, Nomi, Lena
Ibnbari, and Sharon Taube. 2013. Missing objects as Topic Drop. Lingua. doi: http:/
/dx.doi.org/10.1016/j.lingua.2013.07.009.
Erteschik-Shir, Nomi, and Shalom Lappin. 1979. Dominance and the functional expla-
nation of island phenomena. Theoretical Linguistics 6:4185.
Erteschik-Shir, Nomi, and Shalom Lappin. 1987. Dominance and modularity. Linguis-
tics 25 (4):671686. doi: 10.1515/ling.1987.25.4.671.
Erteschik-Shir, Nomi, and Natalia Strahov. 2004. Focus structure architecture and
P-syntax. Lingua no. 114 (3):301323.
Frascarelli, Mara, and Roland Hinterhlzl. 2007. Types of topics in German and Italian.
In On Information Structure, Meaning and Form, edited by Susanne Winkler and
Kerstin Schwabe. Amsterdam: John Benjamins.
Frey, Werner. 2006. Contrast and movement to the German prefield. In The archi-
tecture of focus, edited by Valria Molnr and Susanne Winkler, 235264. Berlin:
Mouton de Gruyter.
50 erteschik-shir
Gundel, Jeanette K. 1974. The role of topic and comment in linguistic theory. Ph.D. disser-
tation, University of Texas, Austin.
Gundel, Jeanette K. 1988. Universals of topic-comment structure. In Studies in syntac-
tic typology, edited by Michael Hammond, Edith A. Moravcsik and Jessica R. Wirth,
209239. Amsterdam / Philadelphia: John Benjamins Publishing Co.
Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Ph.D. disserta-
tion, University of Massachusetts, Amherst, Amherst.
Holmberg, Anders. 2000. Scandinavian Stylistic Fronting: How Any Category Can
Become an Expletive. Linguistic Inquiry 31 (3):445483.
Jakobson, Roman. 1971. Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der
Russischen Kasus. In Selected Writings II, 2371. The Hague: Mouton. Original edi-
tion, TCLP, VI (1936).
King, Tracy Holloway. 1995. Configuring Topic and Focus in Russian. Stanford: CSLI.
Kratzer, Angelika. 1989. Stage-level and individual-level predicates. In Papers on Quan-
tification. Amherst: Department of Linguistics, University of Massachusetts.
Kratzer, Angelika. 1995. Stage-Level and Individual Level Predicates. In The Generic
Book, edited by Greg N. Carlson and Francis J. Pelletier, 125175. Chicago: University
of Chicago Press.
Krifka, Manfred 1994. Focus and operator scope in German. In Intonation and syntax,
edited by Peter Bosch and Rob van der Sandt, 133152. Heidelberg: IBM Deutscland
Informationssysteme, GMBH Scientic Center.
Krifka, Manfred. 1998. Scope Inversion under the Rise-Fall Contour in German. Lin-
guistic Inquiry 29 (1):75112.
Leonetti, Manuel. 2008. Definiteness Effects and the Role of the Coda in Existential
Constructions. In Essays on nominal determination: from morphology to discourse
management, edited by Alex Klinge and Henrik Heg Mller, 131162. Amsterdam:
John Benjamins.
Milsark, Gary, L. 1974. Existential sentences in English. Ph.D. dissertation, M.I.T., Cam-
bridge.
Portner, Paul, and Katsuhiko Yabushita. 2001. Specific Indefinites and the Information
Structure Theory of Topics. Journal of Semantics 18 (3):271297. doi: 10.1093/jos
/18.3.271.
Rando, Emily, and Donna Jo Napoli. 1978. Definites in There-sentences. Language 54
(1):300313.
Raviv, Anne Sofie 2005. Identifying and processing topicalization in Danish. M.A. thesis,
Foreign Literatures and Linguistics, Ben-Gurion University, Beer Sheva.
Reinhart, Tanya. 1981. Pragmatics and linguistics: an analysis of sentence topics. Philo-
sophica no. 27:5394.
Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery. In Elements of Grammar,
edited by Liliane Haegeman, 281337. Dordrecht: Kluwer.
Rooth, Mats. 1985. Association with focus. Ph.D. dissertation, reproduced by GLSA,
University of Massachusetts, Amherst.
Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics
1:75116.
Schulz, Barbara. 2003. Crossing the borders of functional and formal linguistics: an
optimality theoretic account of German topic drop. Paper read at Selected papers
from the seventh college-wide conference for students in languages, linguistics and
literature, at University of Hawaii, Manoa.
Speyer, Augustin 2005. Topicalization and the Trochaic Requirement. Penn Working
Papers in Linguistics 10 (2):243256
Strawson, Peter Frederick. 1964. Identifying reference and truth-values. Theoria 30:86
99.
Svenonius, Peter. 2004. Subject positions and the placement of adverbials. In Subjects,
Expletives, and the EPP, edited by Peter Svenonius, 201242. Oxford: Oxford Univer-
sity Press.
Vallduv, Enric. 1992. The informational component. New York: Garland.
von Fintel, Kai. 1994. Restrictions on Quantifier Domains. Ph.D. dissertation, Depart-
ment of Linguistics, University of Massachusetts, Amherst.
von Heusinger, Klaus. 2002. Specificity and Definiteness in Sentence and Discourse
Structure. Journal of Semantics 19 (3):245274.
Ward, Gregory, and Ellen F. Prince. 1991. On the topicalization of indefinite NPs.
Journal of Pragmatics 15 (8):167178.
On Number and Numberlessness in Languages
with and without Articles*
Asya Pereltsvaig
1 Introduction
This paper is concerned with the representation of number in article-less lan-

guages, focusing on two distantly related languagesRussian and Armenian
and an unrelated language, Tatar. It is argued that morphological number and
semantic number are mediated by syntactic number, encoded even in article-
less languages via a dedicated functional projection, NumP. Thus, an argument
is made against the strongest anti-DP position that denies any functional pro-
jections inside a nominal in an article-less language (cf. Fukui 1986, 1988; Chier-
chia 1998; Baker 2003: 113). Instead, it is shown that at least the projection of
NumP must be assumed even for article-less languages (cf. also Pereltsvaig 2001,
2006, 2007a, b, 2008, 2010; Rutkowski 2002a, b, 2006, 2007, in press; Bailyn 2004).
For a review of the arguments for and against the DP analysis of nominals in
article-less Slavic languages, the reader is referred to Pereltsvaig (2013).
The focus of this paper is on the so-called number-neutral nominals, that is

nominals that denote one or more X: semantically, such nominals are neither
singular (one X) nor plural (more than one X). We argue that the semantic
number-neutrality of such nominals results from the lack of the syntactic
number feature, normally hosted in NumP, which we show to be absent in
such nominals. Depending on the language, such number-neutral nominals
can be morphologically either singular or plural. Furthermore, we show that
* All Tatar data in this paper comes from Ekaterina Lyutikovas fieldwork in the village of
Kutlushkino (and pertains to the local dialect only). I am indebted to Katya for sharing these
materials and for most productive and inspiring discussions of theoretical issues. I am also
grateful to Kaya Borthen, eljko Bokovi, Greville Corbett, Carmen Dobrovie-Sorin, Scott
Grimm, Olga Kagan, Beth Levin, Karine Megedoomian and David Pesetsky for very inspiring
and helpful discussions. I am also thankful to the reviewers and audiences at Berkeley
Linguistic Society and the Paris Workshop on Languages with and without Articles, and the
three reviewers who critiqued the draft of this paper for the present volume. All errors are
mine.

on number and numberlessness 53
the morphological expression of number neutrality does not correlate with

whether a given language has articles or not.
The paper is organized as follows: number-neutral nominals are introduced in

section 2 and their syntactic properties are discussed in more detail in section
3. In section 4, we show based on Tatar data that semantic number-neutrality
is indeed associated with the amount of functional structure present in the
nominal: only nominals that lack the NumP projection (and consequently also
the DP projection as well) can be number-neutral. In section 5 number-neutral
nominals in article-less languages are compared with those in languages with
articles. Section 6 fleshes out the analysis and discusses its consequences for
pluralia tantum nouns in Russian.
2 Number-Neutral Nominals
It has long been noted that some morphologically plural nominals in English,
such as the boldfaced nominal in (1a), do not necessarily denote plural individ-
uals. Thus, the question in (1a) can be truthfully and felicitously answered as in
(1b) and not as in (1c):1
(1) a. Do you have children?

b. Yes, I have one.
c. #No, I have one.
A semantico-pragmatic analysis has been proposed for such instances of num-

ber-neutrality (also referred to as inclusive plurals) in Sauerland (2003) and
Sauerland, Anderssen & Yatsushiro (2005). According to them, the denotation
of morphologically plural nominals includes both atomic individuals and plu-
ral individuals, but under normal circumstances the atomic interpretation is
ruled out by pragmatic principles; however, in certain semantically or pragmat-
ically defined contexts (such as downward entailing contexts) these pragmatic
principles do not apply, making the atomic interpretation available.
While such semantico-pragmatic approach may be appropriate for number-
neutral nominals in English, we will show that it is not suitable for their
1 Note that such number-neutrality in English is restricted by some (so far poorly understood)
pragmatic factors: one is unlikely to enquire Do you have husbands? to find out about some-
ones marital status.
54 pereltsvaig
counterparts in Russian or Armenian. In those languages, one also finds num-

ber-neutral nominals in contexts similar to those in the English examples:2
(2) a. U vas est deti? (Ru)

to you there-is children
Do you have (one or more) children?
b. Bezdig unis? (W Ar)

child have.2sg
Do you have (one or more) children? (cf. Bale et al. 2010)3
Two things are noteworthy here. First, while in Russian number-neutral nomi-
nals are morphologically plural, much like in English (cf. (1a) above), in Arme-
nian number-neutral nominals are morphologically singular (unmarked). Sec-
ond, unlike in English, where number-neutral nominals are restricted to certain
semantically/pragmatically defined contexts, as mentioned above, in Russian
and Armenian (as well as in Tatar, as we shall see below) number-neutral nom-
inals can also be found in other contexts, which are defined syntactically rather
than semantically, as discussed below.
In Russian number-neutral nominals are found in a number of peculiar con-
structions: as complements of intensive reflexives (cf. Tatevosov 2006; Kagan
& Pereltsvaig 2011a, b), as in (3); as complements of the preposition v into in
the v-prezidenty construction (cf. Bailyn 2002; Pereltsvaig 2006), as in (4); and
as complements of syntactic compounds (cf. Trugman 2008; Pesetsky 2010), as
in (5). In the remainder of this paper, we will focus on the first two of those
three constructions. As in (2a) above, the number-neutral nominals in these
constructions are morphologically plural, yet denote one or more X.4
2 The following abbreviations are used for the languages: E. Ar. = Eastern Armenian, W. Ar. =
Western Armenian, Ru = Russian, Ta = Tatar.
Abbreviations in the glosses 1 = first person, 2 = second person, 3 = third person, acc
= accusative, aor = aorist tense, atr = attributivizer, aux = auxiliary verb, cl = classifier,
defsuff = definite suffix, gen = genitive, indef = indefinite, masc = masculine, neut = neuter,
nom = nominative, perf = perfective, pl = plural, pr = present tense, pst = past tense, refl =
reflexive, sg = singular, sp = specific.
3 Throughout this paper, Armenian is illustrated with examples from both Western and Eastern
Armenian. The two varieties are very similar when it comes to number-neutral nominals but
differ significantly in other respects.
4 One of the reviewers for the present volume noted that according to his/her native judgment,
(3) Lena na-jela-s kotlet. (Ru)

Lena na-ate-sja cutlets.gen
Lena ate her fill of cutlets. =
= Lena ate (one or more) cutlets & Lena doesnt want to eat (one or more)
cutlets anymore.
(4) Medvedeva vybrali v prezidenty. (Ru)

Medvedev they-elected into presidents
Medvedev has been elected president.
(5) klonirovanie ivotnyx / remont avtomobilej (Ru)

cloning animals / repairing cars
cloning of animals/an animal; repair of cars/a car
Similarly, in Armenian and Tatar number-neutral nominals are not limited to

downward entailing contexts (Armenian examples below are from Dum-Tragut
(2009: 106) and Megerdoomian (2011)):
(6) a. Ara-n girkh a a-el. (E. Ar.)

Ara-nom book aux.3sg.pr buy-perf
Ara has bought (one or more) books.
b. Dursek a poc mard tes a.

go out aux.1sg.aor street person.nom see aux.1sg.aor
I went out on the street and saw (one or more) persons.
(7) Marat kzl alma aa-d. (Ta)

Marat red apple eat-pst
Marat ate (one or more) red apple(s).
Note that number-neutral nominals in Armenian and Tatar are uniformly mor-
phologically singular (or rather unmarked). In fact, morphologically plural
nominals in Armenian and Tatar can denote only plural individuals and can-
not have the inclusive plural reading of English plurals as in (1a). This is shown
with a Western Armenian example from Bale et al. (2010):
the sentence in (3) cannot mean that Lena has eaten only one cutlet. Other native speakers
agree with our judgment above. We have no explanation for this variation in judgments.
56 pereltsvaig
(8) Bezdig-ner unis? (W. Ar.)

child-indef.pl have.2sg
Do you have (two or more) children?
In the next section, we show that number-neutral plurals in Russian and num-
ber-neutral singulars in Armenian are but two sides of the same coin not only
because of their common interpretation (inclusive, or denoting both atomic
and plural individuals), but because both types of nominals exhibit the same
syntactic properties.
3 Properties of Number-Neutral Nominals in Russian and Armenian
The number-neutral nominals in Russian and Armeniandespite differing in

morphological numbershare a set of common syntactic properties: they are
syntactically selected, but not syntactically incorporated (cf. Baker 1988), and
they fit the syntactic profile of a Small Nominal (cf. Pereltsvaig 2006). (Due
to space limitations, we do not present parallel Tatar data in this section, but
return to this language in the following section.)
While the contexts in which number-neutral nominals can appear in Rus-
sian and Armenian cannot be defined in purely semantic or pragmatic terms, as
mentioned above, such nominals are clearly syntactically selected. For exam-
ple, in Russian they can be selected by a preposition such as the preposition v
into; see (4) above. Moreover, number-neutral nominals can be selected by
certain types of verbal morphology: specifically, by the combination of the
accumulative prefix na- and the reflexive suffix -sja, which (as discussed in
Kagan and Pereltsvaig 2010, 2011) does not amount to the sum of the two parts.5
In Armenian number-neutral nominals can appear in the predicative position
(i.e., as complements of the copula) or as objects (i.e., complements of certain
verbs). Number-neutral nominals in subject position are quite restricted: typ-
ically, they appear as subjects of unaccusative verbs or the copula, that is as
internal arguments of a verb.
However, it is not possible to analyze number-neutral nominals are being
syntactically incorporated in the sense of Baker (1988). First, number-neutral
nominals need not be adjacent to the selecting verb and can be separated from
5 Pereltsvaig (2006) shows that verbal morphology (such as the accumulative prefix na-) can
impose selectional restrictions on what appears to be the complement of the verb.
it by various elements including an adverbial as in the Russian examples in (9)

below, or an auxiliary as in the Armenian example (6b) above.
(9) a. Lena na-jela-s s utra kotlet. (Ru)

Lena na-ate-sja from morning cutlets.gen
Lena ate her fill of cutlets first thing in the morning.
b. Lena na-jela-s do otvala kotlet.

Lena na-ate-sja to falling.off cutlets.gen
Lena ate cutlets up to the brim.
Secondly, number-neutral nominals can contain more than just a bare noun:
they can contain adjectival, adverbial or prepositional modifiers, as in (10ac),
or complements, as in (11). (Additional examples from Tatar are found in sec-
tion 4 below.)
(10) a. Lena na-jela-s rybnyx kotlet. (Ru)

Lena na-ate-sja fish cutlets.gen
Lena ate her fill of fish cakes.
b. Lena na-jela-s kotlet po-poarski.

Lena na-ate-sja cutlets.gen Pozharsky-style
Lena ate her fill of cakes Pozharsky-style.
c. Lena na-jela-s kotlet s podaristoj korokoj.

Lena na-ate-sja cutlets.gen with fried-up crust
Lena ate her fill of cakes with a fried-up crust.
(11) Gorbava vybrali v prezidenty SSSR. (Ru)

Gorbachev they-elected into presidents USSR.gen
Gorbachev was elected president of the USSR.
Thus, we must conclude that number-neutral nominals are phrasal. However,

as we show immediately below, they are structurally deficient, lacking room
for certain pre-nominal elements. In other words, these nominals fit the syn-
tactic profile of a Small Nominal, more specifically of a bare NP. More specif-
ically, number-neutral nominals lack the room for expressions of specificity.
For instance, neither the complements of intensive reflexives nor the com-
plements of v into in Russian allow such modifiers as opredelnnye certain,
specific, leading to the ungrammaticality of the following examples (other
58 pereltsvaig
adjectives, e.g. v potnye prezidenty lit. into honorary presidents, are gram-
matical):
(12) a. *Ja najelas opredelnnyx kotlet. (Ru)

I na-ate-sja specific cutlets
b. *On izbiraetsja v opredelnnye prezidenty.

He being-elected into specific presidents
Similarly, in (Eastern) Armenian number-neutrality obtains only with non-

specific morphologically singular objects. If a specificity marker is added, the
nominal is no longer number-neutral (example from Megedoomian 2011):
(13) Ara-n girkh- a-el a. (E. Ar.)

Ara-nom book-sp buy-perf aux.3sg.pr
Ara has bought {the / a specific} book. (not books!)
Furthermore, number-neutral nominals have no room for expressions of quan-

tity. For example, a numeral or a measure noun cannot be added to such nom-
inals in Russian, leading to the ungrammaticality of examples in (14) and (15);
similarly, in Armenian a singular object with an overt numeral is not number-
neutral (example in (16) is from Megerdoomian 2011).
(14) a. *Ja najelas pjati kotlet. (Ru)

I na-ate-sja five cutlets
b. *Ja napilas stakana vody.

I na-drank-sja glass water
(15) *Obama i Medvedev byli izbrany v dva prezidenta. (Ru)

Obama & Medvedev were elected into two presidents
Obama & Medvedev were elected into presidents.
(16) yerex-erk- me hat muk en ar-el. (E. Ar.)

child-pl-nom one cl mouse aux.3pl.pr find-perf
The children have found a mouse. (not mice!)
Moreover, number-neutral nominals have no room for higher adjectival mod-

ifiers, in the sense of Svenonius (2008), Beauseroy and Knittel (2008), such as
evaluative adjectives. This is true for complements of intensive reflexives, as
shown in (17a), and complements of v into in Russian, as shown in (17b).
(17) a. *Ja najelas otvratitelnyx kotlet. (Ru)

I na-ate-sja terrible burgers
b. *On reil ballotirovatsja v dostojnye gubernatory.

he decided to-run into worthy governors
He decided to run for a worthy governor.
In addition to lacking room for expressions of specificity or quantity and for

higher adjectival modifiers, number-neutral nominals pattern with other Small
Nominals, discussed by Pereltsvaig (2006), in that they are inert for move-
ment either for scope or for focus. As shown by Kagan and Pereltsvaig (2011),
complements of intensive reflexives can only have surface scope: for exam-
ple, the sentence in (18) means that Lena has seen an eyeful of black-and-
white movies in general and, crucially, cannot mean that there is a specific
set of black-and-white movies such that Lena has watched these movies to
the limit. Thus, the genitive NP cannot refer to a specific set of movies which
the speaker has in mind or which have been previously mentioned in the con-
text.
(18) Lena nasmotrelas erno-belyx filmov. (Ru)

Lena na-watched-sja [black-and-white movies].gen
Lena has watched black-and-white movies to the limit.
Pereltsvaig (2006) shows that complements of the preposition v into are like-
wise scopally inert and cannot have a wide scope with respect to a universal
quantifier (which is possible with complements of the preposition v in its other
meanings). For instance, the sentence in (19a) is unambiguous, whereas the one
in (19b) is ambiguous between the wide and the narrow scope readings for gen-
eralov generals.
(19) a. Vsjakij soldat dolen metit v generaly.

every soldier.nom must aim into generals
Every soldier must aim to become a general.
[soldier > general]
b. Vsjakij soldat dolen metit v generalov.

every soldier.nom must aim into generals.acc
Every soldier must aim to shoot {the/some} generals.
[soldier > generals OR generals > soldier]
60 pereltsvaig
Finally, bare singular objects in Armenian are likewise scopally inert: they
cannot take scope over another quantified noun phrase:
(20) amen mi yerexa girkh a a-el. (E. Ar.)

all one child book aux.3sg.pr buy-perf
Every child has bought a book/books. (unambiguous: Karine Megedoo-
mian, p.c.)
Nor can number-neutral objects in Armenian move outside the vP for focus.
Thus, as discussed in detail by Megedoomian (2011), bare singular, number-
neutral objects appear in the vP, whereas specific singular objects (which are
not number-neutral, as discussed above in connection with (13)) can move out
of the vP. Although the word order in the following two examples is exactly the
sameS-O-Aux-VMegedoomian argues that they have different structure:
the vP is bracketed in both sentences below.
(21) a. Ara-n [girkh a a-el]. (E. Ar.)

Ara-nom book aux.3sg.pr buy-perf
Ara has bought {a book/books}.
b. Aram- girkh- a [a-el].

Aram-nom book-SP aux.3sg.pr buy-perf
It is the (specific) book that Aram bought.
According to Megedoomian, the auxiliary can appear in Armenian in one of

two positions: (a) as a second position clitic in the vP, with the element imme-
diately preceding it marking the left edge of the vP, or (b) in FocP. These two
possibilities are instantiated by the two sentences in (21ab), respectively. In
addition to the difference in interpretation (i.e., whether the object is focused,
as in (21b), or not, as in (21a)), the structural differences between these sen-
tences can be shown by placement of manner adverbials such as arag fast: it
can be placed between the auxiliary and the lexical verb in (22b), but not in
(22a)):
(22) a. *Ara-n [girkh a arag a-el]. (E. Ar.)

Ara-nom book aux.3sg.pr fast buy-perf
b. Aram- girkh- a [arag a-el].

Aram-nom book-sp aux.3sg.pr fast buy-perf
It is the (specific) book that Aram bought fast.
The manner adverbial appears in Armenian at the left edge of the vP; hence,
it can appear in the position preceding the lexical verb in (22b). Note that
the auxiliary in this sentence appears outside the vP, in FocP, to be precise. In
contrast, in (21a) the adverbial is not at the left edge of the vP, which leads to
the ungrammaticality of this sentence. When both a manner adverbial and a
bare singular object appear in vP, the former must precede the latter, with the
auxiliary following the first element in the vP, in this case the manner adverbial:
(23) Ara-n [arag a girkh a-el]. (E. Ar.)

Ara-nom fast aux.3sg.pr book buy-perf
Ara has bought {a book/books} fast.
To recap, the bare singular object, which is number-neutral, cannot appear

outside the vP in Armenian. This is in line with the generalization that number-
neutral nominals are inert to movement.
4 Number-Neutral Nominals Lack the NumP Projection
In order to buttress our argument that semantic number-neutrality follows

from the lack of the NumP projection (rather than from the semantic properties
such as mass or collective interpretation of the nominal, or from some semantic
properties of the contexts in which such nominals appear, as was suggested by
one of the reviewers), lets consider the structural types of nominals which can
number-neutral, focusing on Tatar.
First, for number-neutrality to be available, the nominal must be unmarked for

morphological number (i.e. lacking the plural suffix -lar) and, in the case of
direct objects, unmarked for accusative case. We assume that both the plural
suffix and the accusative case marker -n are hosted by dedicated functional
projections inside a nominal, NumP and KP, respectively. It then follows that
number neutral nominals must lack those projections.
(24) a. Marat kzl alma-lar aa-d. (Ta)

Marat red apple-pl eat-pst
Marat ate red apples. (#a red apple)
b. Marat kzl alma-n aa-d.

Marat red apple-acc eat-pst
Marat ate a/the red apple. (#red apples)
62 pereltsvaig
Second, number-neutral nominals can be found as complements to N in

some but not all syntactic constructions. Specifically, complements in the
ezafe-1 construction and in structures involving the attributivizing suffix -le are
number-neutral, whereas complements in ezafe-2 (or ezafe-3, which we will
not consider here due to space limitations, but see Lyutikova & Pereltsvaig 2013
for a more detailed discussion) cannot be number-neutral. Each of these con-
structions places its limitations on the amount of functional architecture avail-
able in the complement: in ezafe-1 the noun selects for a bare N complement
(these are essentially N-N compounds), while structures with the attributivizer
-le contain NP complements. Only ezafe-2 involves complements as large as
NumP.
(25) a. ezafe-1: [N] N (Ta)

ta-(*lar) jrt
stone-(*pl) house-pst
stone house
b. attributivizer -le: [NP] N

altn jezek(*-ler)-le kz
gold *ring(*-pl)-atr girl
a girl with {a gold ring/gold rings}
c. ezafe-2: [NumP] N
krsak-l xatn-nar kijem-e
belly-atr woman-pl clothing-3
clothing for pregnant women
Besides facts involving number-neutrality, the amount of structure in each

type of complement is revealed by the types of elements that are or are not
allowed. Thus, in ezafe-1, the complement may not contain any modifiers (as
in (26a)) or complements of its own; in line with our assumption above that
the plural suffix is hosted by NumP, this suffix is likewise not allowed in ezafe-1
(as in (26b)). The complement generally has a number-neutral interpretation,
although only the plural or singular interpretation is called for, based on general
world knowledge.
(26) a. (*n) altn jezek (Ta)

(*real) gold ring
a gold ring (impossible: a ring of true gold)
b. ta(*-lar) jrt
stone(*-pl) house
stone house
The complement in structures with the suffix -le is larger than a bare N, as it can
contain modifiers (as in (27a)) and can be an ezafe-1 structure (as in (27b)): altn
jezek gold ring, but it cannot contain the plural suffix (as in (27a)) and cannot
be an ezafe-3 structure (which contains a genitive possessor and whose head
agrees with that possessor in person and number), as in (27c). As the ezafe-3
construction has been analyzed as a fully projected DP, we can conclude that
the complement in the structure with -le must be less than a DP and less than
a NumP.
(27) a. kk k(*-lr)-le aka (Ta)

blue flower(*-pl)-atr cup
a cup with a blue flower/flowers
b. altn jezek-le kz
gold ring-atr girl
a girl with a gold ring(s)
c. *Tukaj-n rsem-e-le kitap

Tukaj-gen picture-3-atr book
intended: a book with a picture of Tukaj
In contrast to the two nominal constructions examined above, the ezafe-2

construction (whose head agrees with the complement in person and num-
ber) contains a complement that is a NumP. The complement can contain
not only modifiers (as with the -le construction above) but also the plural
suffix (as in (28a)); the complement of ezafe-2 can itself be an ezafe-2 struc-
ture (as in (28b)). However, the complement of ezafe-2 cannot itself be ezafe-
3, as shown in (28c), meaning that it is maximally a NumP, but not a DP.
Furthermore, as shown in (28b), complements in ezafe-2 cannot be number-
neutral.
(28) a. krsak-l xatn-nar kijem-e (Ta)

belly-atr woman-pl clothing-3
clothing (for) pregnant women
64 pereltsvaig
b. bala-lar xastaxan-se tabib-

child-pl hospital-3 doctor-3
a doctor from a childrens hospital
(impossible: a doctor from childrens hospitals)
c. *uku-n dftr-lr-e papka-s

pupil-gen notebook-pl-3 folder-3
intended: a folder for a/the pupils notebooks
The various nominal constructions are summarized below:
(29) a. ezafe-1: [N] N

b. ATR -le: [NP] N
c. ezafe-2: [NumP] N
d. ezafe-3: [DP] N
Crucially, only complements in the constructions in (29ab), which are struc-

turally smaller than NumP can be number-neutral. Hence, we conclude that
number-neutrality is indeed associated with the lack of the NumP projection.
5 Number-Neutral Nominals in Languages with Articles
As mentioned in section 2 above, number-neutral nominals are found also

in languages with articles; for the purposes of this paper, we are particularly
interested in number-neutral nominals whose occurrence is not determined by
semantico-pragmatic factors, as it is the case in English. In fact, such number-
neutral nominals are found in many languages with articles, including Nor-
wegian (Borthen 2003), Brazilian Portuguese (Schmitt & Munn 2002, Pires de
Oliveira & Rothstein 2011, among others), Catalan (Espinal & McNally 2011), as
well as Spanish and Romanian (cf. Dobrovie-Sorin et al. 2006); the first three
languages listed are illustrated below:
(30) a. Per har hatt hund i ti r. (No)

Per has had dog in ten years
Per has had dog(s) for ten years. (Borthen 2003)
b. Eu acho que vi livro espalhado pelo cho. (BrP)

I think that saw book spread on floor
I think that I saw book(s) spread on the floor. (Schmitt & Munn 2002)
c. Busco pis. (Cat)

look.1sg apartment
Im looking for {an apartment/apartments}. (Espinal & McNally 2011:
93)
What is crucial here is that number-neutral nominals in Norwegian, Brazilian

Portuguese and Catalan share with their counterparts a number of properties
that tend to cluster together (Rullmann 2011). First, they are bare in the sense
of lacking functional syntactic structure. In particular, they have no room for
such elements as demonstratives, higher adjectives or quantity expression. The
latter point is illustrated by the following examples from Pires de Oliveira &
Rothstein (2011):
(31) a. Comprei cadeira. Elam estavam na promoo. (BrP)

bought chair they were on sale
I bought (one or more) chairs. They were on sale.
b. *Comprei duas cadeira.

bought two chair
intended: I bought two chairs.
Second, these nominals are, as mentioned above, number-neutral in that their

denotation includes both atomic and plural individuals. Third, they are not
incorporated (in the sense of Baker 1988), that is, they are not morpho-
syntactically integrated with the verb more tightly than regular objects. Fourth,
they are scopally deficient in the sense of obligatorily having the narrowest
scope possible. Finally, they are weakly referential in the sense of having only a
reduced ability to license discourse anaphora.6
These properties are illustrated below with the Norwegian bare singulars
(which are, by definition, bare; cf. Borthen 2003). That these bare singulars are
number-neutral is highlighted by the fact that they can be picked up by plural
anaphors such as alle all (example from Borthen 2003: 146).
(32) Per har hatt hund i ti r. (No)

Per has had dog in ten years
6 Grimm (to appear) also relates number-neutrality to weak referentiality, but he defines
weak referentiality not as a reduced ability to license discourse anaphora but as lack of
presupposition of existence of any particular referent.
66 pereltsvaig
Alle har vrt svrt snille.

all have been very kind
Per has had dog(s) for ten years. They all have been very kind.
Note also the plural pronoun used to refer back to the number-neutral bare sin-
gular in the Brazilian Portuguese example in (31a) above; as shown by Pires de
Oliveira & Rothstein (2011), a singular pronoun can not be used in this situation.
Furthermore, Norwegian bare singulars are not syntactically incorporated;
in fact, they can be multi-word phrases and not just single (bare) nouns, as
shown by the example below from Borthen (2003: 164).
(33) Ola nsker seg kopp med bilde av Mikke Mus. (No)
Ola wants refl cup with picture of Mickey Mouse
Ola wants a cup with a picture of Mickey Mouse.
Like other number-neutral nominals considered in this paper, Norwegian bare

singulars are scopally deficient, so that the sentence in (34) unambiguously
means All the children tried on some jacket or other. In other words, the
bare singular cannot have wide scope with respect to the universal quantifier
denoted by alle all (cf. Borthen 2003: 24).
(34) Alle barna prvde jakke. (No)

all children-defsuff tried jacket
All the children tried on some jacket or other.
Finally, Norwegian bare singulars are weakly referential in that they license the
type-anaphor dt that but not the token-anaphor den it (cf. Borthen 2003:
3941).
(35) a. Per nsker seg ny bt,

Per wants refl new boat.masc,
men dt fr han nok aldri. (No)
but that-neut gets he probably never
Per wants a new boat, but he probably will never get that.
b. Jeg nsker meg sykkel til jul.

I want refl bike to Christmas.
??Den er bl.
??it is blue
I want a bike for Christmas. It is blue.
To account for this clustering of properties, we assume that the central prop-
erty on which the others hinge is bareness. In particularly, bareness implies that
there is no structural space for determiners or other quantificational elements,
which in turn implies the properties of scopal deficiency and weak referen-
tiality. Furthermore, we take bareness (in the relevant sense) to be a necessary
but not a sufficient condition for syntactic incorporation into the verb. Finally
and most importantly for the present paper, bareness implies that there is no
syntactic number projection, which in turn results in number-neutrality. The
parallel clustering of these properties in languages with and without articles
suggests that syntactic number and the lack thereof (which translates into
semantic number-neutrality) is to be analyzed in a parallel fashion in both
types of languages. Ergo, article-less languages have a dedicated functional pro-
jection for number, NumP. This argument is fleshed out in the following sec-
tion.
6 Proposal and Consequences
So far, we have argued that number-neutral nominalsincluding those in

languages with and without articlesare syntactically bare, meaning they lack
the functional projection hosting number and numerals, NumP. In other words,
such number-neutral nominals are bare NPs (cf. Dobrovie-Sorin et al.s 2006
analysis of bare singulars in Spanish).
However, you will recall from our discussion in sections 2 and 3 above that
in Russian number-neutral nominals are realized morphologically as plurals.
This means that morphological plurality does not always reflect the presence
of [singular] feature in syntax.7 Similar views are espoused by Pesetsky (2010)
and Alexiadou (2011). For Pesetsky, however, it is morphological singularity
that is not always a reflection of [+singular] feature in syntax, as he takes
singular complements of paucal numerals to be numberless. As we show in this
paper, truly number-neutral nominals in Russian are morphologically plural;
for an alternative analysis of the apparently singular complements of paucal
numerals, see Pereltsvaig (2010, 2011). Alexiadou (2011) takes a similar view to
ours, based on a different set of data. She discusses plural mass nouns in Greek
(the English counterpart of which is something like The river discharges its
7 For the sake of presentation, we will assume that the number feature is [singular]. The
essence of our proposal does not change if the [plural] feature is used instead.
68 pereltsvaig
waters into the lake) and comes the conclusion that the plural morphology of
such plural mass nouns is not associated with NumP.
Following Pesetsky (2010) and Alexiadou (2011), we propose here the num-
ber feature is introduced into the computation by the NumP. Furthermore, the
head of NumP establishes an agreement relation with the noun. Thus, if the
NumP has a [singular] feature, the N receives the same value of the number
feature and appears in the plural morphological form. Conversely, if the NumP
has a [+singular] feature, the N receives the same value of the number feature
and appears in the singular morphological form. In the absence of NumP, the
noun does not receive the value for its number feature. The default morpholog-
ical realization of such numberless nouns is plural in Russian (but singular in
Armenian, where number-neutral nominals are morphologically singular; see
the discussion surrounding (7) above).
This analysis has an interesting consequence for the so-called pluralia tan-
tum nouns in Russian, such as nonicy scissors, brjuki trousers, sutki 24-hour
period, etc. Unlike other nouns, whose number specification is determined
syntactically, through agreement with NumP, as described above, pluralia tan-
tum nouns are specified for number lexically. In other words, they are specified
as [singular] in the lexicon. As a result, pluralia tantum nouns are not com-
patible with paucal numerals, such as dva two, tri three and etyre four.
(36) *tri {sutki / sutok } (Ru)

three 24-hour.period.pl.nom / 24-hour.period.pl.gen
3 24-hour periods (i.e., 72 hours)
Paucal numerals introduce their own (paucal) number specification, which,

following Bailyn and Nevins (2008), we will take to be [singular, augmented].
Pluralia tantum nouns, however, have no paucal form; they are lexically spec-
ified as plural (which, taking into account the existence of paucal in Russian,
might be rendered as [singular, +augmented]). This clash between the paucal
number introduced by the paucal numeral and the plural number of the plu-
ralia tantum noun itself results in the ungrammaticality of examples like (35).
Note that there is no problem combining pluralia tantum nouns with non-
paucal numerals, which in Russian include higher numerals (5 and up):
(37) pjat sutok

five 24-hour.period.pl.gen
Moreover, non-paucal numerals in Russian include also the so-called collec-

tive numerals (e.g., dvoe two, troe three, etvero four, pjatero five, etc.). Note
that the set of collective numerals includes those that encode the same lower
numerosities as paucal numerals: for example, semantically the collective nu-
meral troe and the paucal numeral tri are identical in that both encode the same
numerosity, three. Yet, the collective numeral is compatible with a pluralia tan-
tum noun, as shown in (38) below, whereas non-collective, paucal numeral is
not, as shown in (36) above.
(38) troe sutok

three 24-hour.period.pl.gen
3 24-hour periods (i.e., 72 hours)
Note further that apart from their appearance with pluralia tantum nouns,
collective numerals are limited to masculine human nouns as in (39a) and
are impossible with either feminine human nouns (as shown in (39b)) or
non-human nouns (as shown in (39c)).
(39) a. troe parnej

three chap. pl.gen
3 chaps
b. *troe devuek (OK: tri devuki)

three girl.pl.gen three girl.sg.gen
three girls
c. *troe dnej (OK: tri dnja)

three day.pl.gen three day.sg.gen
three days
To recap, our analysis of number as being introduced into the computation by

the NumP allows us to account for the peculiar combination possibilities of
pluralia tantum nouns with different types of numerals in Russian.
7 Conclusion
In this paper, we have shown that numberless nominals are not limited to
downward entailing contexts in several unrelated (or distantly-related) lan-
guages: Russian, Armenian, Tatar. Instead, numberless nominals are found in
various syntactic contexts where they are selected by certain lexical or func-
tional heads. We have argued that the lack of semantic number is due to the lack
70 pereltsvaig
of the functional projection which hosts number and numerals and therefore
encodes number, that is NumP. In other words, such number-neutral nominals
are bare NPs. Such numberless numerals have room for (certain kinds) of adjec-
tival modifiers, however, as such modifiers are merged low enough, in the NP.
8 References
Alexiadou, Artemis. 2011. Plural Mass Nouns and the Morpho-syntax of Number.
In Proceedings of the 28th West Coast Conference on Formal Linguistics, edited by
Mary Washburn, Katherine McKinney-Bock, Erika Varis, Ann Sawyer, and Barbara
Tomaszewicz, 3341. Somerville, MA: Cascadilla Proceedings Project.
Bailyn, John Frederick. 2002. Overt Predicators. Journal of Slavic Linguistics 10:2352.
Bailyn, John Frederick. 2004. The Case of Q. In Proceedings of FASL 12, edited by Olga
Arnaudova, Wayles Browne, Mara Luisa Rivero, and Danijela Stojanovic, 136. Ann
Arbor, MI: University of Michigan Press.
Bailyn, John Frederick and Andrew Nevins. 2008. Russian Genitive Plurals are Impos-
tors. In Inflectional Identity, edited by Andrew Nevins and Asaf Bachrach, 237270.
Oxford: Oxford University Press.
Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chi-
cago, IL: University of Chicago Press.
Baker, Mark C. 2003. Lexical Categories. Verbs, Nouns, and Adjectives. Cambridge, UK:
Cambridge University Press.
Bale, Alan, Michal Gagnon, and Hrayr Khanjian. 2010. Cross-linguistic Representa-
tions of Numerals and Number Marking. Proceedings of SALT 20: 115.
Beauseroy, Delphine and Marie-Laurence Knittel. 2008. Nombre et adjectifs: le cas des
noms abstraits (= Number and adjectives: the case of abstracts nouns). Paper pre-
sented at the Workshop on nominal and verbal plurality, Paris, CNRS, 78 November
2008.
Borthen, Kaja. 2003. Norwegian bare singulars. Ph.D. thesis, Trondheim: NTNU.
Chierchia, Gennaro. 1998. Reference to Kinds Across Languages. Natural Language
Semantics 6:339405.
Dobrovie-Sorin, Carmen, Tonia Bleam and M. Teresa Espinal. 2006. Bare Nouns, Num-
ber and Types of Incorporation. In Non-definiteness and plurality, edited by Svetlana
Vogeleer and Liliane Tasmovski, 5181. Amsterdam: John Benjamins.
Dum-Tragut, Jasmine. 2009. Armenian. Amsterdam: John Benjamins.
Espinal, M. Teresa and Luise McNally. 2011. Bare nominals and incorporating verbs in
Spanish and Catalan. Journal of Linguistics 47:87128.
Fukui, Naoki. 1986. A Theory of Category Projection and its Aplications. Ph.D. thesis,
Massachusetts Institute of Technology.
Fukui, Naoki. 1988. Deriving the differences between English and Japanese: A case
study in parametric syntax. English Linguistics 5:249270.
Grimm, Scott. To appear. Plurality is Distinct from Number Neutrality. To appear in
Proceedings of NELS 41.
Kagan, Olga and Asya Pereltsvaig. 2011a. Bare NPs and Semantic Incorporation: Objects
of intensive reflexives at the syntax-semantics interface. In Formal Approaches to
Slavic Linguistics 18, edited by Wayles Browne, Adam Cooper, Alison Fisher, Esra
Kesici, Nikola Predolac, and Draga Zec, 226240. Ann Arbor, MI: Michigan Slavic
Publications.
Kagan, Olga and Asya Pereltsvaig. 2011b. Syntax and Semantics of Bare NPs: Objects of
Intensive Reflexive Verbs in Russian. In Empirical Issues in Syntax and Semantics 8,
edited by Olivier Bonami and Patricia Cabredo Hofherr, 221238. http://www.cssp
.cnrs.fr/eiss8/kagan-pereltsvaig-eiss8.pdf
Lyutikova, Ekaterina and Asya Pereltsvaig. 2013. Elucidating Nominal Structure in
Articleless Languages: A Case Study of Tatar. In Proceedings of Berkeley Linguistics
Society 39.
Megerdoomian, Karine. 2011. Focus and the auxiliary in Eastern Armenian. Paper
presented at the Berkeley Linguistic Society, Berkeley, February 2011.
Pereltsvaig, Asya. 2001. On the Nature of Intra-Clausal Relations: A Study of Copular
Sentences in Russian and Italian. Ph.D. thesis, McGill University.
Pereltsvaig, Asya. 2006. Small nominals. Natural Language and Linguistic Theory 24:
433500.
Pereltsvaig, Asya. 2007. Copular Sentences in Russian. A Theory of Intra-Clausal Rela-
tions. Springer.
Pereltsvaig, Asya. 2007. On the Universality of DP: A View from Russian. Studia Lin-
guistica 61:5994.
Pereltsvaig, Asya. 2008. Split Phrases in Colloquial Russian. Studia Linguistica 62:538.
Pereltsvaig, Asya. 2010. As easy as two, three, four? In Formal Approaches to Slavic Lin-
guistics 18, edited by Wayles Browne, Adam Cooper, Alison Fisher, Esra Kesici, Nikola
Predolac, and Draga Zec, 417434. Ann Arbor, MI: Michigan Slavic Publications.
Pereltsvaig, Asya. 2011. On numberlessness and paucal numerals in Russian. Paper
presented at the FASL 2011 meeting, MIT, May 2011.
Pereltsvaig, Asya. 2013. Noun Phrase Structure in Article-less Slavic languages: DP or
not DP? Language and Linguistics Compass 7:201219.
Pesetsky, David. 2010. Russian Case Morphology and the Syntactic Categories. Ms.,
MIT.
Pires de Oliveira, Roberta and Susan Rothstein. 2011. Bare singular noun phrases are
mass in Brazilian Portuguese. Lingua 121:21532175.
Rullmann, Hotze. 2011. Scopal Deficiency and Number. Paper presented at the Discus-
sion Workshop on Weak Referentiality, Utrecht, March 2011.
72 pereltsvaig
Rutkowski, Pawe. 2002. Noun/pronoun asymmetries: evidence in support of the DP

hypothesis in Polish. Jezikoslovlje 3:159170.
Rutkowski, Pawe. 2002. Numerals as grammaticalised nouns: a generative approach.
Interlingstica 13:317328.
Rutkowski, Pawe. 2006. Why Polish numerals should not be analyzed as nouns. In
Minimalist Views on Language Design: Proceedings of the 8th Seoul International
Conference on Generative Grammar, edited by Changguk Yim, 249263. Seoul: Han-
kook/Korean Generative Grammar Circle.
Rutkowski, Pawe. 2007. The Determiner Phrase hypothesis as a tool of syntactic analysis
of Polish nominal phrases. Ph.D. thesis, Warsaw University.
Rutkowski, Pawe. In press. The syntax of floating intensifiers in Polish and its impli-
cations for the Determiner Phrase hypothesis. In Proceedings of the Thirty-Second
Annual Meeting of the Berkeley Linguistics Society, edited by Zhenya Antic, Molly
Babel, Charles Chang, Jisup Hong, Michael Houser, Fang-Chun Liu, Maziar Toosar-
vandani, and Yao Yao. Berkeley: University of California.
Sauerland, Uli. 2003. A New Semantics for Number. In Proceedings of SALT 13, edited
by Robert B. Young and Yuping Zhou, 258275. Ithaca, NY: CLC Publications.
Sauerland, Uli, Jan Andersen, and Kazuko Yatsushiro. 2005. The plural is semantically
unmarked. In Linguistic Evidence, edited by Stephan Kepser and Marga Reis, 413
434. Berlin: Mouton de Gruyter.
Schmitt, Christina and Alan Munn. 2002. The Syntax and Semantics of Bare Arguments
in Brazilian Portuguese. In Linguistic Variation Yearbook, Vol. II, edited by Pierre
Pica and Johann Rooryck, 185216. Amsterdam: John Benjamins.
Svenonius, Peter. 2008. The Position of Adjectives and other Phrasal Modifiers in the
Decomposition of DP. In Adjectives and Adverbs, edited by Louise McNally and
Christopher Kennedy, 1642. Oxford: Oxford University Press.
Tatevosov, Sergei. 2006. Measuring individuals, partitioning events: Semantics of cu-
mulative verbs in Russian. In Linguistic Investigations into Formal Description of
Slavic Languages, edited by Peter Kosta and Lilia Schurcks, 529544. Frankfurt: Peter
Lang.
Trugman, Helen. 2008. Modifiers of bare nouns in Russian. Paper presented at FDSL
7.5. Moscow Independent University, December 2008.
The Cognitive Basis of the Mass-Count Distinction:
Evidence from Bare Nouns1
Edit Doron and Ana Mller
1 Introduction
The naive view of the linguistic mass-count distinction has been that it reflects
a cognitive distinction between homogeneous matter which lacks units for
counting, and discrete entities which form atomic units and thus can be
counted. The naive view has often been questioned in the literature, most
recently when Gillon (1992) and Chierchia (1998) discussed mass nouns which
denote discrete entitiessuch as jewelry, clothing, furniture, mail. To consider
one example, a chair is an atomic unit of furniture, since part of a chair is not
furniture. Thus furniture is not homogeneous; nevertheless, it is a mass noun.
Conversely, Rothstein (2010) discussed the fact, first pointed out by Mittwoch
(1988), that there are count nouns which denote homogeneous entitiessuch
as fence, line, cloud, bouquet. Two clouds which come together form a cloud,
demonstrating the homogeneity of the count noun cloud. As a result of the dis-
crepancy between the mass-count linguistic contrast and the homogeneous-
atomic cognitive contrast, the distinction between mass and count nouns
emerges in the work of these scholars as partly arbitrary and language spe-
cific.
Indeed Chierchia (1998) constructs a theory of the mass-count distinction
which views it as a linguistic distinction, only partly cognitively based. In a
sophisticated twist, it actually presents those mass nouns with atomic structure
such as jewelry, clothing, furniture, mail, to be prototypical mass nouns. The idea
is that the denotation of all mass nouns contains discrete units, for example
particular quantities of water in the case of the mass noun water, but these units
are not linguistically accessible. Later, Chierchia (2010) abandons this view. One
reason is the observation due to Roger Schwarzschild whereby units of mass
1 We are grateful to the organizers and audiences of the Workshop on Bare NPs at Bar-Ilan
University, 18 October 2010, and the CNRS & Paris 8 Journes dtude Langues avec et sans
articles, 34 March 2011. We thank three anonymous referees for their comments, which were
very helpful in the reformulation of some of our claims. Dorons research was supported by
the Israel Science Foundation grant #1157/10. Mllers research was supported by the Conselho
Nacional de Pesquisa grant #303407/2009-3, and by FAPESP grant #2011/51408-4.

74 doron and mller
nouns are linguistically accessible after all, since one can for example predi-
cate size of them in the phrase the big furniture, where big is the size of units of
furniture. Chierchia (2010) readopts the view whereby the mass-count classifi-
cation reflects a cognitive distinction between types of units. Mass nouns are
vague nouns with unstable units: within the same context (or actually within
precisifications of the context), entities in the denotation of a mass noun might
at the same time be both a unit and an aggregate of units. Only mass nouns
which actually have stable units, like furniture, now treated as fake mass nouns,
reflect an arbitrary linguistic decision.
Our aim in this paper is to tighten the connection between the mass-count
distinction and its cognitive basis. In section 2 we dicuss Karitiana, a language
that does not have nominal pluralization and does not have any formal mass-
count distinction in the structure of nouns or noun phrases, yet semantically
distinguishes nouns which can be counted from nouns which cannot. In sec-
tion 3, we will bring data from Modern Hebrew, a language which has plural
nominal morphology, but where, like in Karitiana, countability is not reflected
by pluralization, but rather by a semantic identification of stable units. Follow-
ing Chierchia (2010), we view mass nouns as denoting entities with unstable
units: within the same context, an entity is at the same time both a unit and
an aggregate of units. Count nouns on the other hand have stable units in a
given context. We discuss a new example of mass nouns with atomic structure,
found in Modern Hebrew and hitherto undiscussed in the literature. The anal-
ysis of this new example will substantiate the (2010) model, as it demonstrates
that even fake mass nouns fit non-arbitrarily into the mass-count classification.
Thus we believe that the claim that the mass-count distinction reflects a cog-
nitive distinction can be extended to its limit and include fake mass nouns.2
In the system of Chierchia (2010) there is no need to assume, as he does, that
fake mass nouns reflect the arbitrary linguistic decision to ignore their exist-
ing atomic structure. Rather, we will show a principled reason for their mass
nature.
(A) We claim regarding such mass nouns as furniture that they are bona-fide
mass terms, since what counts as a unit of furniture in a given context is
not stable; it could be the whole sitting room set or just one of its parts.
Accordingly, it may be felicitous in a given context to utter This living room
set is so much furniture!
2 A different type of approach for the substantiation of the same claim is found in Grimm and
Levin (2011).
the cognitive basis of the mass-count distinction 75
(B) Conversely, count nouns have stable units in each given context. For
example, cloud is bona-fide count, since considering parts of a cloud to
be separate clouds necessitates changing the context. In order to view a
cloud both as a unit and as several units at the same time, a gestalt switch
is required which changes the context mid sentence in #This cloud is so
many clouds!
Point (B) has already been argued in Nicolas (2002) and Chierchia (2010), and in
this paper we therefore concentrate on substantiating point (A), the instability
of the units of such mass nouns as furniture.
But first we argue, on the basis of Karitiana, for the general point that
countability is independent of a formal linguistic mass-count distinction.
2 Karitiana
Karitiana is a Tupi-Arikm language spoken in Rondnia, in the western Brazil-

ian Amazonic region. The language has around 400 speakers, most of them
living in a demarcated reservation in Rondnia. The mass-count distinction
is not formally encoded in Karitiana in any way, yet the language semantically
distinguishes nouns which can be directly counted from nouns which cannot.
A similar claim has been made by Wilhelm (2008) for Dne Sulin.
The mass-count distinction is not formally encoded in Karitiana in any way.
First, there is no nominal number morphology in the language that could
set apart mass from count nouns (see Mller et al. 2006). The word pikom
(monkey) in sentence (1a) below is entirely undefined as for whether the
number of monkeys eaten is one, more than one, or even parts of one or various
monkeys. In (1b) oho is a bare singular referring to a kind:3
3 The data from Karitiana was collected by Mller during fieldwork. The examples are pre-
sented as follows1st line: orthographic transcription of the Karitiana sentence; 2nd line:
morphological segmentation; 3rd line: morpheme by morpheme gloss; 4th line: transla-
tion.
Abbreviations used in the glosses are as follows: abs = absolutive; abs.agr = absolutive
agreement; anaph = anaphor; ass = assertive mood; caus = causative; cop = copula; cop.agr
= copula agreement; decl = declarative mood; deic = deictic; fem = feminine; ft = future; impf
= imperfective; inv = inverse; masc = masculine; nft = non-future tense; nmz = nominalizer;
obl = oblique; pl = plural; postp =postposition; rdpl = reduplication; sg = singular; sub =
subordinator; tv = thematic vowel; 1, 2, 3 = 1st, 2nd, 3rd person.
76 doron and mller
(1) a. yn nakayt pikom (Karitiana)

yn -naka-y-t pikom
1s 3-decl-eat-nft monkey
I ate (the/a/some) monkey(s).
b. oho atakamat Ora (Karitiana)

oho -a-taka-m-a-t Ora
potato 3-inv-decl-caus-make-nft Ora
Potatoes, Ora created (them)
Nevertheless counting is attested in the language. In sentence (2a), the phrase

myhint pikom (one monkey) is semantically singular, whereas in sentence (2b)
the phrase sypomp pikom (two monkeys) is semantically plural. Yet, the noun
pikom remains uninflected for number in both environments. In addition, Kari-
tiana is not a classifier language, since, as the examples in (2) show numerals
and common nouns combine directly.
(2) a. yn nakayt myhint pikom (Karitiana)

yn -naka-y-t myhin-t pikom
1s 3-decl-eat-nft one-obl monkey
I ate one monkey.
b. yn nakayt sypomp pikom

yn -naka-y-t sypom-t pikom
1s 3-decl-eat-nft two-obl monkey
I ate two monkeys.
Not even personal pronouns are marked for number in the language. Table 1
presents the paradigm of personal pronouns. The 3rd person is clearly non-
variable. On the other hand, 1st and 2nd person plural pronouns do not incor-
porate any morpheme with a plural meaning. They are formed by the suffix-
ation of the 3rd person anaphora ta or by the suffixation of the third person
pronoun i, as shown in second column of Table 1.
Second, measure quantifiers and demonstratives do no distinguish between
mass and count either, as they combine equally with both. The quantifiers
kandat much/many and syyn a little/few co-occurs both with count and mass
nouns. Examples with kandat are given below:
table 1 Personal pronouns in Karitiana (Mller et al 2006)
Pronoun Morphology Person Meaning
yn y+n 1sg I+ participant

na a+n 2sg you + participant
i i 3 other (non-participant)
yjxa y+i+ta 1pl (inclusive) I+other(s)+anaphor
yta y+ta 1pl (exclusive) I+anaphor
ajxa a+i+ta 2pl you+other(s)+anaphor
i i 3 other
(3) a. kandat taso naponpon sojxaty kyn (Karitiana)

kanda-t taso -na-pon.pon- sojxaty kyn
much-obl man 3-decl-shoot.rdpl-nft boar at
Many men shot at boars./Men shot at boars many times.
b. jonso nakaot kandat ese

jonso -na-ot- kanda-t ese
woman 3-decl-get-nft much-obl water
Women brought a lot of water./ Women brought water many times.
Universal quantifiers, which are expressed by relative clauses, combine equally

well with count and mass nouns:
(4) a. taakatyym nakapyyk ombaky

ta-aka-tyym -naka-pyky-t ombaky
3.anaph-cop-sub 3-decl-be.over-nft jaguar
Maria Conga pip (Karitiana)
Maria Conga pip
Maria Conga postp
All jaguars are gone from Maria Conga.
b. taakatyym nakapyyk ouro Maria Conga pip

ta-aka-tyym -naka-pyky-t ouro Maria Conga pip
3.anaph-cop-sub 3-decl-be.over-nft gold Maria Conga postp
All gold is gone from Maria Conga.
78 doron and mller
Similarly, demonstratives too combine both with mass and count nouns:
(5) a. tykat idjiera ahop aka a ese aka (Karitiana)

tykat i-djera- ahop aka a ese aka
impf nmz-cost-nft much cop deic water cop
How much does this water cost?
b. dibm nakatari a w aka

dibm naka-tar-i a w aka
tomorrow decl-leave-ft deic child cop
These boys will leave tomorrow.
Nevertheless, the mass-count distinction manifests itself in what Chierchia

(2010) has called the signature property, which is the marked status of a mass
noun when combined directly with a numeral expression. Count nouns are
naturally modified by numerals, as illustrated in the sentences in (2), and by
sentences (7) and (9) below, whereas mass nouns, if they do so, require contex-
tual information in order to be interpreted, as illustrated by the awkwardness
of sentences (6) and (8) when uttered in out-of-the blue contexts. The contrast
in acceptabiliy between examples (6) and (8), and examples (7) and (9) shows
that the denotation of certain nouns can only be counted if count units are
introduced (explicitly or implicitly).
(6) */# myhint ouro naakat iorot

myhin-t oro na-aka-t i-ot.ot-t
one-obl gold decl-cop-nft nmz-fall.rdp-abs.agr
One gold fell.
(7) myhint kilot ouro naakat iorot

myhin-t kilo-t oro na-aka-t i-ot.ot-t
one-obl kilo-obl gold decl-cop-nft nmz-fall.rdpl-abs.agr
One kilogram of gold fell.
(8) #jonso nakaot sypomp ese

jonso naka-ot-t sypom-t ese
woman decl-bring-nft two-obl water
The woman brought two waters.
(9) jonso nakaot sypomp bytypip ese

jonso naka-ot-t sympom-t byty-pip ese
woman decl-bring-nft two-obl bowl-postp water
The woman brought two bowls of water.
Counting can be encoded in Karitiana by modifiers other than numerals, as in

(10) below. The distributive numerals myhint myhint (one one) and sypomp
sypomp (two two) are sentential adjuncts that distribute individuals over
events in the sentences at hand. The individuals are separated in groups which
have their cardinality determined by the distributive numeral so that, in sen-
tence (10a) boys are grouped one by one, and in sentence (10b), men are
grouped in twos. The distribution of groups of individuals of a given cardinal-
ity presupposes individuation on both sides of the distributive relationin our
case, one boy per event of going to the river or two men per event of arriving.
(10) a. myhint.myhint nakahori w se pip (Karitiana)

myhin-t.myhin-t naka-hot-i w se pip
one-obl.one-obl decl-go.pl-ft child river postp
Boys will go to the river one at a time.
b. sypomp.sypomp naotm taso

sypom-t.sypom-t na-otm- taso
two-obl.two-obl decl-arrive-nft man
Men arrived two at a time.
Distribution then can only operate on count arguments. As expected, distribu-

tive quantifiers applied to mass nouns do not yield grammatical sentences, as
illustrated by the sentences (11) and (12), unless particular contexts are given so
that they introduce feasible measure phrases for the nouns at hand.
(11) */#ese naakaj iorot myhint.myhint (Karitiana)

water na-aka-j i-ot.ot-t myhin-t.myhin-t
water decl-cop-ft nmz-fall.rdpl-abs.agr one-obl.one-obl
Water will fall one at a time.
(12) */#sypomp.sypomp naotm ouro (Karitiana)

sypom-t.sypom-t na-otm- oro
two-obl.two-obl decl-arrive-nft gold
Gold arrived two at a time.
80 doron and mller
Thus in Karitiana the individuability of units is directly reflected for some

nouns, without the mediation of morphology, since the difference between
individuated vs. non-individuated nouns is expressed in their ability to be inter-
preted in certain grammatical constructions and their corresponding semantic
operations. This is not contradicted by examples of what Chierchia (2010) calls
the property of elasticity. One finds mass to count and count to mass coer-
cion in the language. Count nouns may be coerced into mass by the so-called
universal grinder, as illustrated below by the word ep that is count in (13a)
and (14a) (meaning tree), but turns mass in (13b) and (14b) (meaning wood).
According to Chierchia, grinding count nouns into mass nouns seems to involve
the notion of material part of, which is also illustrated by sentence (15) in the
context of a rat being smashed against a wall.4
(13) a. ep itipasagngt Joo (Karitiana)

ep i-ti-pasag.pasag-t Joo
tree 3-inv-count.rdpl-nft Joo
The trees, Joo is counting (them).
b. ep naakat jepyryt
ep -na-aka-t jepyry-t
wood 3- decl-cop-nft club-abs.agr
The club is of wood.
(14) a. myhint.myhint namangat ep Joo (Karitiana)

myhin-t.myhin-t -na-mangat- ep Joo
one-obl.one-obl 3-decl-carry-nft tree Joo
Joo carried trees one by one.
b. myhint namangat kandat ep Joo

myhin-t -na-mangat- kandat ep Joo
one-obl 3-decl-carry-nft much-obl wood Joo
Joo carried a lot of wood at once.
(15) pyryhopiyn mejahygng amby parede sok (Karitiana)

pyry-hop-iyn mejahygng house parede sok
ass-exist-nft rat house wall over
There was rat all over the wall.
4 Notice that Cheng, Doetjes and Sybesma (2008), on the basis of data from Mandarin, argue
that the grinder reading is not similarly available in all languages.
The reverse coercion, of mass nouns into count nouns, is also attested in
Karitiana, by what Lewis called the universal packager, which inserts context
dependent units as in (16):
(16) a. ony sypomp ge aka naakat ipositivot

ony sypom-t ge cop na-aka-t i-positivo-t
deic two-obl blood cop decl-cop-nft nmz-positive-abs.agr
Those two bloods (blood containers) are positive.
b. myjymp him pysyp iorot

myjym-t him pysyp i-ot-ot-
three-obl game meat nmz-fall.rdpl-abs.agr
Three (pieces of game) meat fell/ Three steaks fell
In Karitiana, it appears that all nouns with atomic structure are countable, i.e.
we do not find fake mass nouns in this language. For example, clothes/ clothing,
which is a fake mass noun in some languages, and could have been considered
a mass term in Karitiana as well, since it appears with much in e.g. (17a), can
actually be counted, as shown in (17b):
(17) a. Milena naakat iamyt kandat pykypyty

-na-aka-t i-amy-t kanda-t pykyp-y-ty
Milena 3-decl-cop-nft nmz-buy-nft much-obl clothes-tv-obl
Milena bought a lot of clothes.
b. Milena naakat iamyt sypomp pykypyty

-na-aka-t i-amy-t sypom-t pykyp-y-ty
Milena 3-decl-cop-nft nmz-buy-nft two-obl clothes-tv-obl
Milena bought 2 pieces/units of clothes.
In this section, we have shown that the conceptual distinction of countability

is directly expressed in Karitiana without the mediation of morphological
marking of count nouns.5
5 A reviewer rightly points out that a full understanding of the basis of the mass/count dis-
tinction in Karitiana would require an extensive investigation of all nouns which exhibit
variable-behavior in different languages, e.g. furniture, hair, as well as abstract nouns. We
leave this investigation to future research.
82 doron and mller
3 Modern Hebrew
Unlike Karitiana, Modern Hebrew has plural nominal morphology. Yet as in

Karitiana, it is not plural morphology which distinguishes count from mass
nouns in Hebrew.
3.1 Plural Morphology in Modern Hebrew

Among nouns which pluralize, we can distinguish broadly between three noun
classes in Modern Hebrew, according to their plural morphology. These classes
only partly overlap with gender distinctions (cf. Bat-El 1989, Faust 2011, Ritter
1995, Schwarzwald 1991). Class I nouns, which are mostly masculine, have the
suffix -im in the plural. Class II nouns are often feminine, and are inflected
in the plural by the the suffix -ot. Class III nouns mostly denote members of
natural pairs, and are inflected in the plural by the suffix -yim.6 These are
shown in (18ac) respectively:7
(18) a. Class I nouns (plural suffix -im)

atul / atul-im dbor-a / dbor-im
cat.masc cat-pl bee-fem bee- pl
b. Class II nouns (plural suffix -ot)

tmun-a / tmun-ot alon / alon-ot
picture-fem picture-pl window.masc window-pl
c. Class III nouns (plural suffix -yim)

maga / maga-yim greb / garb-yim
boot.masc boot-pl sock.fem sock-pl
The nouns illustrated in (18) above are all count nouns. Most mass nouns in
Hebrew do not pluralize:8
6 If attached to nouns which do not denote members of natural pairs, the suffix -yim may be
interpreted as dual rather than plural, but we will not be interested here in the dual.
7 In the Hebrew transcriptions, stress is marked with an accent whenever it deviates from
the unmarked word-final stress of the language. The notation b, , , marks the fricative
allophones of the stops b, k, p, respectively. is a voiceless pharyngeal fricative; in word-final
position, it typically triggers the insertion of a preceding epenthetic a. Abbreviations used
in the glosses are: fem = feminine; masc = masculine; pl = plural; ASN = Argument Structure
Nominal.
8 Some plural forms in (19) are found when these mass nouns are coerced to count readings,
(19) rez *oraz-im acac *acac-im

rice.masc rice-pl gravel.masc gravel-pl
avir *avir-im boc *boc-im

air.masc air-pl mud.masc mud-pl
mi *mei-im em-a *em-ot

silk.masc silk-pl butter-fem butter-pl
kutn-a *kutn-ot alud-a *alud-ot

cotton-fem cotton-pl rust-fem rust-pl
But there are also quite a few mass nouns in Hebrew which are plural, where
plural morphology does not mark a count reading but retains the mass inter-
pretation. First, there are mass nouns which are pluralia-tantum. These nouns
are obligatorily inflected with a plural suffix from one of the three classes I
III:9
(20) haris-ot *haris-a atiq-ot *atiq-a imur-im *imur

ruins antiquities canned goods
mar-im *mer m-yim *ma am-yim *ama

yeast-pl water-pl sky-pl
qur-ey10 / *qur akabi kl-ey / *kli mita maca-im maca

web-pl spider linen-pl bed bed-linen-pl bed
spider webs bed linen
Second, there are also mass nouns which have a morphological contrast be-
tween singular and plural forms:
(21) gem / gam-im leg / lag-im

rain.masc rain-pl snow.masc snow-pl
by e.g. the universal packager or subkind coercion. These are always the default forms, i.e.
Class I for masculine nouns and Class II for feminine nouns.
9 Some of these singular forms exist as deverbal nominalizations, e.g. haris-a destruction,
imur preservation.
10 -ey is the construct-state form of the plural suffixes -im and -yim.
84 doron and mller
de / da-im ed / ed-im
grass.masc grass-pl steam.masc steam-pl
kse / ksa-im dam / dam-im

money.masc money-pl blood.masc blood-pl
adam-a / adam-ot ap-a / ap(-at)-ot

land-fem land-pl rubbish-fem rubbish-pl
ol / ol-ot rua / ru-ot

sand.masc sand-pl wind.fem wind-pl
merab / merab-im
space.masc space-pl
Semantically, the plural form of mass terms, when it contrasts with the singu-
lar, denotes abundance plural, similarly to what has been reported for other
languages (Corbett 2000, Ojeda 2005, Tsoulas 2006, Acquaviva 2008, Alexiadou
2011).
We conclude that overt plural morphology does not distinguish count from
mass nouns. Rather, as in Karitiana, the distinction between count and mass
nouns depends on the availability of counting, i.e. cooccurrence with cardinal-
ity modifiers.
3.2 The Cardinality Modifier Criterion

Count nouns combine with cardinality modifiers such as one, two, several.
(22) a. yled ead yelad-im aad-im ney yelad-im

child.masc one.masc child-pl one-pl two.masc child-pl
one boy several boys two boys
b. yald-a aat yelad-ot aad-ot tey yelad-ot

child-fem one.fem child-pl one-pl two.fem child-pl
one girl several girls two girls
Cardinality modifiers do not combine with mass nouns, irrespective of whether

these mass nouns are singularia tantum, pluralia tantum, or alternate in plural-
ity. In particular, this indicates that the plural mass terms illustrated in (20) and
(21) above are indeed mass terms: similarly to singular mass terms, they do not
co-occur with cardinality modifiers.
(23) *rez ead *dam-im aad-im

rice one blood-pl one-pl
*one rice *several bloods
*alud-a aat *tey ru-ot

rust-fem one.fem two.fem wind-pl
*one rust *two winds
*ol-ot aad-im *ney mar-im

sand-pl one-pl two.masc yeast-pl
*several sands *two yeasts
3.3 The Measure Quantifier Criterion

Like in Karitiana, there are measure quantifiers in Modern Hebrew such as a lot,
a little, which basically measure quantities, and they co-occur both with mass
nouns and with count nouns. Count nouns combined with these quantifiers
are interpreted as pluralities, both in Karitiana and in Modern Hebrew, but in
a language such as Modern Hebrew, their plurality must be morphologically
marked. This gives rise, in languages with plural morphology, to an additional
distributional criterion for the mass-count distinction: plurality is imposed on
count nouns but not mass nouns for the purpose of measure quantification.
This criterion has been emphasized in the semantic literature at least since
Pelletier (1975) and Link (1983), as it demonstrates the semantic affinity of
plural count nouns to mass nouns.
In Hebrew, examples of measure quantifiers are harbe a lot of, meat a little,
de/ yoter miday too much, klo a kilo of etc. They co-occur both with count
nouns and mass nouns, but in the case of count nouns, they only co-occur
with plural forms of the noun. This is illustrated by the contrast between (24a),
where the count nouns are plural, and the ungrammatical (24b), with singular
count nouns:
(24) a. harbe yelad-ot meat atul-im klo tapu-im

a lot child-pl a little cat-pl kilo apple-pl
many girls few cats a kilo apples
de botn-im
too much peanut-pl
too many peanuts
86 doron and mller
b. *harbe yald-a *meat atul *klo tapa

a lot child-fem a little cat kilo apple
*de bten
too much peanut
When combining with mass nouns, measure quantifiers allow singular mor-
phology (though plural morphology is also an option for mass nouns which
have plural forms, preserving the abundance plural reading):
(25) harbe ol/ ol-ot meat rez klo um

a lot sand/sand-pl a little rice kilo garlic
much sand a little rice a kilo garlic
de gem/gam-im
too much rain/rain-pl
too much rain
3.4 The Paradox of Flexible Nouns

The two distributional criteria described in sections 3.2 and 3.3 above mostly
yield consistent results separating between mass nouns (ol sand, rez rice,
um garlic, gem rain etc.) and count nouns ( yald-a girl, atul cat, tapa
apple, bten peanut etc.). But as already mentioned above for Karitiana, there
is elasticity in the system, and as a result there are nouns which these two
criteria fail to classify. Some are of the types which are familiar crosslinguis-
tically. First, nouns which are usually taken to basically be count nouns and
are coerced to mass readings by the universal grinder, e.g. kbes lamb also
interpreted as meat, o bird also interpreted as chicken meat, ec tree, also
interpreted as wood:
(26) a. loa kbas-im amia o-ot

three.masc lamb.masc-pl five.masc bird.masc-pl
three lambs five birds
ney ec-im
two.masc tree.masc-pl
two trees
b. de kbes klo o harbe ec

too much lamb kilo chicken a lot wood
too much lamb a kilo of chicken a lot of wood
Second, nouns which are usually considered to be basically mass nouns and
are coerced to count readings by the universal packager, e.g. stone, rope, beer,
soap,
(27) a. tona ben harbe bel yoter miday br-a

ton stone a lot rope too much beer-fem
a ton of stone a lot of rope too much beer
b. ame aban-im ney abal-im

five.fem stone.fem-pl two.masc rope.masc-pl
five stones two ropes
alo br-ot
three.fem beer.fem-pl
three beers
or by the subkind coercion whereby alo brot three beers means three
kinds of beer.11,12
But there is an additional class in Hebrew which we will call flexible nouns,
which the two criteria fail to classify. According to the first criterion, co-occur-
rence with cardinality modifiers, these are count nouns. The examples in (28a)
below show that flexible nouns co-occur with cardinality modifiers. Yet these
nouns are found in the singular with measure quantifiers, as in (28b), and are
thus classified as mass nouns by the second criterion.13
(28) a. amia gzar-im / iba bcal-im / tey s-ot / dla-ot

five carrots / seven onions / two lettuces / pumpkins
mona cnon-im / tras-im / krub-im / tut-im

eight radishes / corncobs / cabbages / mulberries
11 Some authors (e.g. Barner and Snedeker 2005) do do not consider stone and rope to be
basically mass terms.
12 Subkind coercion has been shown to be at work even in languages that do not have a plural
morphology (Chung 2000), but we were not able to find such examples in Karitiana.
13 Flexibility cannot be attributed to the absence of grammatical number (Kwon and Zribi-Hertz
2004), which is present in Modern Hebrew.
88 doron and mller
b. klo gzer / bacal / tut / s-a / let

kilo carrot / onion / mulberry / lettuce / turnip
de cnon / tras / krub / dlat / umar

too many radish / corn / cabbage / pumpkin / fennel
The mass interpretation in (28b) is not the result of coercion by the universal
grinder, since it is not necessarily e.g. mashed carrot substance but individu-
ated carrots which are measured.14 Similarly, though it is possible to interpret
tras corn as corn grains in (28b), it is also interpretable as individuated corn
cobs. Also, whereas de kbes too much lamb cannot be interpreted as too
many lambs (for which the plural would have to be used in Modern Hebrew as
well as in English), de gzer too much carrot can be interpreted as too many
carrots despite of the singular form of gzer carrot. Moreover, if the mass inter-
pretation of flexible nouns were the effect of the universal grinder, we would
expect the same interpretation for the count nouns in (29) below such as apple
and peanut, but this is not the case. (29) includes bona fide count nouns which
are not flexible, i.e. they usually appear in the plural with measure phrases:15
(29) a. klo tapu-im / agas-im / acil-im / qiu-im / ekoliy-ot

kilo apples / pears / aubergines / courgettes / grapefruits
harbe ada-im / anab-im / zeyt-im / botn-im / dubdeban-im

many lentils / grapes / olives / peanuts / cherries
de agbaniy-ot / ezi-im / tapuz-im / pitriy-ot

too many tomatoes / plums / oranges / mushrooms
b. *klo tapa / agas / acil / qiu / agbaniy-a

kilo apple / pear / aubergine / courgette / tomato
14 Other fruits/vegetables, typically large ones, only have a coerced mass reading, where the
mass term denotes substance of the fruit denoted by the count noun: milon melon, abata
watermelon.
15 We leave out borrowed nouns such as mmi (Arabic) apricot, nanas (French) pineapple,
anna (Latin) sweetsop, aboqdo avocado, batta sweet potato, fijya feijoa, and also
singularia tantum nouns, which resist plural morphology, both in the context of counting
and in the context of measuring, yet are nevertheless clearly count nouns, as they appears
with the same number morphology in both environments, e.g. ney eseq two loquats, kilo
eseq a kilo loquats.
*harbe ada-a / anab / zyit / bten / dubdeban

many lentil / grape / olive / peanuts / cherry
*de ekolit / ezi / tapuz / pitriy-a

too many grapefruit / plum / orange / mushroom
Neither is the count interpretation in (28a) the result of coercion by the uni-
versal packager. Unlike stone, rope, beer in (27) which we consider to be mass
nouns that may give rise to standardized units, the flexible nouns carrot, onion,
mulberry etc in (28) have very salient natural units just like bona fide count
nouns such as apple, pear, olive in (29).
Flexible nouns are found in the singular in additional contexts where the
plural is normally required with counts nouns, such as in (30a) below, where a
plurality of units is intended, yet the singular can be used with flexible nouns. A
relevant context would be the planning of a shopping expedition to the market,
where one does not normally buy single fruits and vegetables. With bona fide
count nouns, as in (30b), singularity gives rise to an anomalous interpretation
in this context, since the only possible interpretation would be one where the
addressee is asked to buy a single exemplar at the market. Yet apples, just
like carrots, are typically not purchased by the unit at the market but by the
kilo:
(30) (In the context of shopping)

a. tiqni gzer / tut / umar / bacal
buy carrot / mulberry / fennel / onion
Buy carrots/ mulberries/ fennels/onions.
b. #tiqni tapa / agas / tapuz / agbaniya / ezi

buy apple / pear / orange / tomato / plum
#Buy an apple / a pear / an orange / a tomato / a plum.
In the same shopping context, it is possible to form the comparative on the

singular of carrot but not apple:
(31) (In the context of shopping)

a. hi qanta yoter gzer / cnon / bacal mi ma e-
she bought more carrot / radish / onion than what that
biqanu
we-asked
She bought more carrots / radishes / onions than we asked.
90 doron and mller
b. *hi qanta yoter tapa / agbaniya / ezi mi ma e-

she bought more apple / tomato / plum than what that
biqanu
we-asked
In partitive and existential examples where there isnt a contextual preference

for plural readings and both singular and plural interpretations are in prin-
ciple felicitous, the singular form of a count noun in (32b33b) below only
gives rise to a single unit interpretation, whereas the singular form of the flex-
ible noun in (32a33a) also refers to a plurality in addition to singular refer-
ence.
(32) a. rob ha-gzer raquv

most (of) the carrot is rotten
Most of the carrot is rotten.
Most of the carrots are rotten.
b. rob ha-tapa raquv

most (of) the apple is rotten
Most of the apple is rotten.
(33) a. ye gzer b-a-tiq

there (is) carrot in-the-bag
There is a carrot in the bag.
There are carrots in the bag.
b. ye tapa b-a-tiq
there (is) apple in-the-bag
There is an apple in the bag.
The examples below in (34)(35) below demonstrate that flexible nouns give
rise to amount relatives in the singular, unlike ordinary count nouns (Carlson
1977):
(34) b-a-ayim lo naclia le-gadel et ha-gzer e hem

in-the-life not we-succeed to-grow acc the-carrot that they
criim bibil mial-ha-imurim el-ahem
need for the-canning-factory of-theirs
Not even in a life-time will we succeed to grow the carrots that they need
for their canning factory.
(35) b-a-ayim lo naclia le-gadel et ha-tapu-im/*tapa e

in-the-life not we-succeed to-grow acc the-apples/*apple that
hem criim bibil mial-ha-imurim el-ahem
they need for the-canning-factory of-theirs
Not even in a life-time will we succeed to grow the apples that they need
for their canning factory.
The examples in (36)(37) below demonstrate that reference to kinds also

distinguishes between count nouns and flexible nouns, which can have the
distribution of mass nouns. Singular reference to kinds is impossible in the
environments in (36)(37) for count nouns, but is possible for mass nouns
(Doron 2003). Flexible nouns appear in the singular in these environments,
like mass nouns and unlike count nouns. For example, a bare singular noun
in the object position of love can denote a kind, which is possible for singular
flexible nouns, similarly to mass nouns, but not for count nouns, as shown in
(36). Additionally, count nouns such as apple and tomato must be pluralized in
the compounds apple-juice and tomato-soup in Modern Hebrew, whereas the
flexible noun carrot and onion are singular in the same compounds, as shown
in (37).
(36) ani ohbet gzer / *tapa / tapu-im

I love carrot / apple / apples
I love carrots/apples.
(37) a. mic gzer / *tapa / tapu-im

juice carrot / apple / apples
carrot/apple juice
b. maraq bacal / *agbaniy-a / agbaniy-ot

soup onion / tomato / tomatoes
onion/tomato soup
To summarize this section, we have seen that flexible nouns are distinguishable
from count nouns. Flexible nouns appear in the singular in environments
where count nouns are typically plural:
A. In the environment of measure quantifiers such as harbe a lot of, meat a

little, de/ yoter miday too much, klo a kilo of
B. In the context of shopping
C. In partitive and existential constructions
92 doron and mller
D. With amount relatives

E. In reference to kinds
On the other hand, flexible nouns are similar to mass nouns such as petruzlya
parsley, amir dill, tred spinach, uiyt bean, um garlic. Mass nouns, like
flexible nouns, are singular when combined with measure quantifiers, as shown
in (38a). But unlike flexible nouns, mass terms are not countable, as shown in
(38b):
(38) a. klo tred harbe petruzly-a meat amir de uiy-t

kilo spinach much parsley a little dill too much bean
b. *tred ead *tey petruzly-ot *amir ead *alo uiy-ot
one spinach two parsley-pl one dill three bean-pl
3.5 Fake Mass Nouns

The flexible nouns introduced in the last section have a lot in common with
what Chierchia (2010) called fake mass nouns, a term which he coined for
nouns like English furniture, jewelry, mail, mass nouns which have recognizable
atomic units. The flexible Modern Hebrew nouns carrot, fennel, onion, mul-
berry, cabbage etc have the characteristics of fake mass nouns: on the one hand
they have the distribution of mass nouns, and on the other hand they have rec-
ognizable atomic units. There is one difference between flexible nouns and fake
mass noun, and it is that the former, unlike the latter, also have the distribution
of counts nouns. We will return to this difference below, but, based on the sim-
ilarities, we will henceforth consider flexible nouns in Modern Hebrew to be
fake mass nouns.
We propose that what semantically characterizes fake mass nouns is that
they have units which can be individuated in many contexts, yet speakers are
actually not normally interested in these units. The reason is that the typical
context for the use of these terms normally involves other units, which, in the
case of the English fake mass nouns, are typically aggregates of the atomic
units. These aggregates are unstable, in that modifiying them in the process
of context precisification changes their status as units. Consider the English
fake mass nouns furniture, clothing, bed-linen, mail, silverware. These predicates
have perceptible atomic units like a chair, a knife, a letter, a shirt, a sheet. Yet
in most everyday contexts we are not interested in these units, but in other
units which are aggregates of these units: a set of tableware, a living-room set,
a combination of clothes, a set of bed linen, the contents of a mailbox. These
aggregates are not themselves stable units, since we could include poorer or
richer aggregates in more precise contexts, which may result in the original
aggregates loosing their status as units. Accordingly, these concepts lack stable
units, and the type of predicate which denotes them is indeed a mass term.16 A
similar point is independently made by Landman (2010) regarding the nature of
fake mass nouns. According to him, fake mass nouns, which he calls neat mass
nouns, have possibly overlapping generators, that is, both atomic individuals
and their groupings may count as one.
Turning to Modern Hebrew fake mass nouns, the examples we have consid-
ered so far all name fruits and vegetables. Examples parallel to the English fake
mass nouns exist as well; we will return to them in the next section. The fruit
and vegetable fake mass nouns, like the English-type fake mass nouns, have
natural atomic units. Yet in the context of preparation of food, we are not nor-
mally interested in the natural units of these particular fruits and vegetables,
but typically in edible serving-size units. What characterizes these particular
fruits and vegetables seems to be their texture, which determines the ease with
which serving size units can be constructed. Fruits and vegetables with uni-
form texture easily lend themselves to have parts or aggregates considered to
be food portions. On the other hand, one cannot indiscriminately carve food
portions out of apples, plums and oranges, because their texture is not uni-
form and contains corks, pits, sections, etc. The same consideration extends
to courgettes, cucumbers and aubergines, which are not uniform in texture
since some of their parts are packed with seeds and others are free of seeds.
These are therefore bona fide count nouns. Carrots, turnips and radishes, on
the other hand, have uniform texture, and thus avail themselves to be carved
into portions, or have portions constructed from parts of different natural units.
Similarly for onions, cabbage, lettuce, fennel, which also carve out naturally out
into indiscriminate parts. All these are fake mass nouns in Modern Hebrew. It
is predicted that though minuscule fruit never form serving-size portions by
themselves, they are not all categorized in the same way. Fruits with pits, such
as olives and cherries do not have uniform texture and do not allow indiscrim-
inate formation of serving-size portions. They are therefore classified as count
nouns. Strawberries and mulberries are uniform in texture, and are thus fake
mass terms.
16 In the case of the fake mass noun change, each coin is a unit, yet at the same time its monetary
value has different units, e.g. a two-Euro coin is counted just like two one-Euro units for the
purpose of paying. Thus change inseparably involves both coins and their values, and though
both types of units are stable, the existence of two sets of equally salient units in the same
context prevents using either for counting.
94 doron and mller
We thus propose that fake mass nouns are nouns which naturally allow
for an additional mode of individuation in parallel to their natural atomic
structure, within the same context. This additional mode of individuation is the
one typically relevant to speakers, and it determines the distribution of these
nouns. The units of this mode of individuation are unstable, which is a property
that characterizes mass nouns in Chierchias (2010) system.17
3.6 The Collective-Singulative Alternation

We now return to the difference between English and Modern Hebrew fake
mass nouns, i.e. Modern Hebrew fake nouns also have the distribution of count
nouns. We attribute this difference to the fact that Modern Hebrew does, and
English doesnt, have singulative morphology which marks the selection of
natural units, and the shift of the type of these nouns from mass to count.
Singulative morphology differs from the universal packager in that it does not
derive standardized units of mass nouns in general, but it only applies to fake
mass nouns which have natural units to begin with.
Singulative morphology (nomen unitatis) in Modern Hebrew, and in Semitic
languages in general, e.g. Arabic (Wright 1859: 147) and Neo Aramaic (Khan
2008: 343), is homonymous to collective morphology, both expressed by the
feminine suffix, cf. Moscati et al. 1964: 86. This type of polar morphology is an
example of the phenomenon of morphological reversal, whereby two opposite
processes make use of the same exponent (Baerman 2007). In Modern Hebrew,
the actual use of the feminine exponent for these processes is relatively rare,
but crucially it is found in both directions. (39) illustrates the direction in which
the singulative is marked as feminine, and (40)the direction in which the
collective is marked as feminine.18
17 There are language specific factors that determine which units are linguistically encoded
beyond the atomic units of nouns which denote discrete entities. Languages may choose to
disregard aggregates as units for some nouns which have natural atomic units. At the limit, as
pointed out to us by a reviewer, there are languages such as Modern Greek, which disregard
aggregates as units in the case of all nouns that have natural units; such languages thus have
no fake mass nouns at all.
18 The fact that both directions are marked may present a problem for unidirectional views such
as Borer (2005) whereby roots are interpreted as mass, and count nouns are derived from
roots by additional structure, hence it is count nouns which should be marked relative to
mass nouns.
(39) The singulative alternation (singulative is derived by fem suffix)
Basic mass noun Singulative count noun Plural
sear saar-a saar-ot

hair.masc a-hair-fem hairs
sar sear-a sear-ot

turbulence.masc storm-fem storms
mer imr-a imr-ot

speech.masc saying-fem sayings
sa si-a si-ot
discourse.masc conversation-fem conversations
mvet mit-a mit-ot

death.masc a-death-fem deaths
(40) The collective alternation (collective is derived by fem suffix)
Basic count noun Plural Collective mass noun19
a. dag dag-im dag-a

fish.masc fish-pl fish-fem
b. ale al-im alv-a

leaf.masc leaves foliage-fem
19 The pattern in (40) may account for the fact that though Modern Hebrew Class I nouns
(nouns pluralizing with the suffix -im) are normally masculine, they also include a limited
subclass of feminine nouns with the suffix -a. Such feminine nouns, e.g. dbor-a bee in (18a)
can be considered a backformation from an original collective mass noun dbor-a, which
historically belonged in the third column of (40). This would have been a collecitve mass
noun related to the plural masc count noun dbor-im, similarly to the situation in rows (40ef)
where the singular count noun is mising. Eventually, the collective mass noun dbor-a was
reinterpreted as the missing singular count noun, which was facilitated by the fact that
96 doron and mller
(40) The collective alternation, Continued
Basic count noun Plural Collective mass noun
c. gole gol-im gol-a

expatriate.masc expatriates diaspora-fem
d. aun aun-im aun-a

a-pea.masc peas pea-fem
e. it-im it-a
wheat plants wheat-fem
f. seor-im seor-a
barley plants barley-fem
In the derivation of collective mass nouns in Modern Hebrew, the fem suffix
is often replaced by Argument Structure Nominal morphology (ASN). As was
shown by Grimshaw (1990), ASNs have the distributional properties of mass
nouns:
(41) Allomorphy in the collective alternation
a. bged bgad-im bigud

garment garments clothing-ASN
b. nal naal-yim hanala

shoe shoes footwear-ASN
singulative morphology is identical to collective morphology. Evidence is provided by the

collective nature of many of the feminine nouns which pluralize in Class I: nemal-a / nemal-im
ants, kin-a / kin-im lice, yon-a / yon-im pigeons, ada-a/ ada-im lentils, teen-a/ teen-im figs
etc. There are other languages in which nouns of this sort have singulative morphology (cf.
Schwarzwald 1991).
c. rehit rehit-im rihut

piece of furniture20 pieces of furniture furniture-ASN
d. pra pra-im pria

flower flowers bloom-ASN
e. mircet mirca-ot ricu

tile tiles tiling-ASN
Both mass nouns with collective morphology and mass nouns related to count
nouns with singulative morphology are fake mass nounsmass nouns which
nevertheless have natural units. What is special about the flexible nouns dis-
cussed in the previous sections is that there is no morphological distinction
between the mass noun and the corresponding count noun. Thus it is not
clear whether they belong to the collective or to the singulative alternation,
if there is alternation in their case at all. We will sidestep this issue in the
present work by saying that they belong to a collective-singulative alterna-
tion:
20 The difference between the Modern Hebrew count noun rehit and the corresponding English
count term which does not include any sortal has semantics repercussions. Wereas the
following Modern Hebrew sentence is true, its English translation is normally taken to be
false, since a sofa-bed is one piece of furniture, not two:
(i) sapa nitat mehava ney rehit-im be- rehit ead

sofa-bed constitutes two furniture.count-pl in furniture.count one
A sofa-bed consists of two pieces of furniture in one.
In examples where both languages have count nouns, both are judged equally for truth:
(ii) ele ney mair-im be- mair ead

these (are) two gadgets in gadget one
These are two gadgets in one.
98 doron and mller
(42) The collectivesingulative alternation (flexible nouns)
Collective mass noun Singulative count noun Plural
gzer gzer gzar-im

carrot carrot carrots
bacal bacal bcal-im

onion onion onions
tut tut tut-im etc

mulberry mulberry mulberries21
4 Conclusion
We have argued that fake mass nouns do not distort after all the correspon-
dence between a clear cognitive distinction and the mass-count linguistic dis-
tinction. Though fake mass terms, e.g. furniture, clothing, mail, jewelry, denote
entities with natural atomic units, these units are nevertheless irrelevant since
in many given contexts, it is natural to rather view parts or aggregates of these
units as units. The instability of these latter units is what makes these nouns
mass. We have given examples of fake mass terms in Modern Hebrew which
have not so far been brought up in the literature: carrot, onion, strawberry, mul-
berry, etc. We have shown that these nouns denote units that are found in
nature, but, due to their homogeneous texture, also denote at the same time
serving-size units in the context of food preparation. With these units in mind,
such nouns emerge as vague, since the size of edible portions changes in the
process of context precisification in a way which changes their status as units.
Accordingly, they too exhibit unit instability and are treated as mass nouns. The
view developed in this paper explains the different properties of these nouns
in comparison to what might otherwise look like an indistinguishable class, e.g.
apple, tomato, orange, cherry, but which actually belongs with count nouns.
These nouns do not lend themselves, due to their texture, to a level of vague
food portions, and thus remain countable even in the context of food prepa-
21 A couple of examples of this sort exist in English as well, hair, grain, seed; we are indebted to
Malka Rappaport Hovav for this observation.
ration. Chierchias (2010) analysis of mass nouns as vague nouns with unstable
units has shaped the present approach, which in turn extends the limits of his
analysis to include fake mass nouns as well.
We have not found examples of fake mass nouns in Karitiana, a language
where nouns are number-neutral. It appears that the role of plural morphology
is crucial for constructing different types of mass nouns, and for distinguishing
different types of units, stable and unstable, of which only the former are
available for counting.
References
Acquaviva, Paolo. 2008. Lexical plurals: a morpho-semantic approach. Oxford: Oxford

University Press.
Alexiadou, Artemis. 2011. Plural Mass Nouns and the Morpho-syntax of Number. Pro-
ceedings of the 28th West Coast Conference on Formal Linguistics, 3341. Somer-
ville, MA: Cascadilla Proceedings Project.
Baerman, Matthew. 2007. Morphological Reversals. Journal of Linguistics 43:3361.
Barner, David and Jesse Snedeker. 2005. Quantity judgments and individuation: evi-
dence that mass nouns count. Cognition 97:4166.
Bat-El, Outi. 1989. Phonology and Word Structure in Modern Hebrew. PhD Thesis.
UCLA.
Borer, Hagit. 2005. Structuring Sense Volume I: In Name Only. Oxford: Oxford University
Press.
Carlson, Greg N. 1977. Amount Relatives. Language 53:520542.
Cheng, Lisa L., Jenny Doetjes and Rynt Sybesma 2008. How universal is the Universal
Grinder? Linguistics in the Netherlands 2008, 5062.
Chierchia, Gennaro. 1998. Reference to Kinds Across Languages. Natural Language
Semantics 6:339405.
Chierchia, Gennaro. 2010. Mass nouns, vagueness and semantic variation. Synthese
174: 99149.
Chung, Sandra. 2000. On reference to kinds in Indonesian. Natural Language Seman-
tics 8:157171
Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press.
Doron, Edit. 2003. Bare Singular Reference to Kinds. Proceedings of Semantics and
Linguistic Theory 13:7390.
Faust, Noam. 2011. Forme et fonction dans la morphologie nominale de lhbreu moderne.
PhD Thesis, Universit Paris Diderot.
Gillon, Brendan. 1992. Toward a common semantics for English count and mass
nouns. Linguistics and Philosophy 15:597639.
100 doron and mller
Grimm, Scott and Beth Levin. 2011. Furniture and Other Functional Aggregates: More
and Less Countable than Mass Nouns. Paper presented at Sinn und Bedeutung 16,
University of Utrecht, Utrecht, The Netherlands, September 68, 2011.
Grimshaw, Jane. 1990. Argument Structure. Cambridge, MA: MIT Press.
Khan, Geoffrey. 2008. The Neo-Aramaic Dialect of Barwar. Leiden: Brill.
Kwon, SongNim and Anne Zribi-Hertz. 2004. Number from a syntactic perspective:
Why plural marking looks truer in French than in Korean. In Empirical Issues in
Formal Syntax and Semantics 5:133158, edited by Olivier Bonami & Patricia Cabredo
Hofherr.
Landman, Fred. 2010. Count nouns, mass nouns, neat nouns, mess nouns. In For-
mal Semantics and Pragmatics: Discourse Context and Models, edited by Jurgis Skil-
ters. Riga: The Baltic International Yearbook of Cognition, Logic and Communica-
tion.
Link, Godehard. 1983. The logical analysis of plural and mass terms. In Meaning,
Use and Interpretation of Language, edited by Reiner Buerle, Christoph Schwarze
and Arnim von Stechow, 302323. Berlin: de Gruyter. Reprinted in G. Link. 1998.
Algebraic Semantics in Language and Philosophy, 1133. Stanford: CSLI Publi-
cations.
Mittwoch, Anita 1988. Aspects of English aspect: on the interaction of perfect, progres-
sive and durational phrases. Linguistics and Philosophy 11:203254.
Moscati, Sabatino, Anton Spitaler, Edward Ullendorff and Wolfram von Soden. 1964.
An introduction to the comparative grammar of the Semitic languages: phonology and
morphology. Wiesbaden: Otto Harrassowitz.
Mller, Ana, Luciana Storto, and Thiago Coutinho da Silva. 2006. Number and the
mass/count distinction in Karitiana. In Proceedings of the Eleventh Workshop on
Structure and Constituency in Languages of the Americas, edited by Atsushi Fujimori
and Maria Amelia Reis Silva, 122135. Dept. Linguistics, UBC.
Nicolas, David. 2002. Do mass nouns constitute a semantically uniform class? Kansas
Working Papers in Linguistics 26:113121.
Ojeda, Almerindo E. 2005. The Paradox of Mass Plurals. UC Davis ms.
Pelletier, Jeffry. 1975. Non-singular reference: some preliminaries. Philosophia 5: 451
465.
Ritter, Elisabeth. 1995. On the Syntactic Category of Pronouns and Agreement. Natural
Language and Linguistic Theory 13: 405443.
Rothstein, Susan. 2010. Counting and the mass-count distinction. Journal of Semantics
27:343397.
Schwarzwald, Ora. 1991. Grammatical vs. Lexical Plural Formation in Hebrew. Folia
Linguistica 25:577608.
Tsoulas, George. 2006. Plurality of mass nouns and the grammar of Number. Paper
presented at the 29th GLOW colloquium in Barcelona.
Wilhelm, Andrea. 2008. Bare nouns and number in Dne Sin. Natural Language
Semantics 16:3968.
Wright, William. 1859. A Grammar of the Arabic Language [1967. Third edition contain-
ing both volumes]. Cambridge: Cambridge University Press.
The Turkish NP*
eljko Bokovi and Serkan ener
This paper investigates the Turkish NP, with emphasis on the structural posi-
tion of possessors (both overt and null possessorsthey will be shown to be
located in different positions), demonstratives, numerals, and adjectives as well
as the interpretation of possessors in different contexts. We also investigate in
detail ellipsis within Turkish NPs, which turns out to be a particularly useful
tool for probing NP structure, due to rather strong constraints on such ellip-
sis that Turkish displays. Thus, in stark contrast with English, Turkish disallows
simple possessor-stranding ellipsis, a fact which we will show has important
consequences for the structure of NP in Turkish. Turkish in fact displays rather
complex paradigms regarding word order and interpretation of NP-internal ele-
ments, as well as NP-internal ellipsis, which make it a perfect testing ground for
various approaches to these phenomena.
The paper is organized as follows. We start by considering in section 1 how
Turkish, an article-less language, fares with respect to Bokovis (2008a, 2012a)
generalizations regarding DP/NP languages. In principle, even if some article-
less languages dont have DP, it may not be out of the question that some could
have it, hence we first test Turkish regarding Bokovis generalizations. We
show that Turkish patterns with NP, not DP languages, which has an impor-
tant impact on the analysis we develop in later sections based on c-command
tests, linear order, interpretation, and ellipsis of NP-internal elements. Sections
2 and 3 are the main parts of the paper. Section 2 establishes an outline of NP
structure in Turkish, based on c-command tests, word order, and the interpre-
tation of possessors, which is subjected to further testing with respect to ellipsis
in section 3. Section 4 is the conclusion.
1 NP/DP Generalizations and Turkish
Bokovi (2008a,2012a) argues for a no-DP analysis of languages without def-

inite articles, based on a number of cross-linguistic generalizations acknowl-
edging the fact that the presence/absence of articles in a language plays a
* This material is based upon work supported by the National Science Foundation under Grant
BCS-0920888. We thank three anonymous reviewers and the participants of a seminar at the
University of Connecticut for helpful comments.

the turkish np 103
crucial role. These generalizations show that there is a fundamental difference

between the traditional Noun Phrases (TNPs)1 of languages with and without
articles that cannot be reduced to phonology (overt vs null articles) since the
generalizations involve syntactic and semantic, not phonological, phenomena.
Below, we will test Turkish with respect to these generalizations, which are
given in subsection headings. (The generalizations are only briefly summa-
rized here; see Bokovi 2012a for detailed discussion of these generalizations,
including the precise definitions and additional cross-linguistic illustrations of
the phenomena referred to in these generalizations as well as their deductions
under the DP/NP analysis, which provides a uniform account of all the cross-
linguistic differences noted below, where a single factor is responsible for all of
them.)2
1.1 Article-Less Languages Disallow Clause-Mate NPI-Licensing under

Negative Raising, Article Languages Allow It
With negative raising (NR), negation behaves as if it were lower than where it
surfaces, as confirmed by clause-mate Negative-Polarity Items (NPIs). The NPI
in (1) (in at least two years) requires a clause-mate Negation, as witnessed by
(2), involving the non-NR verb claim. Negation must therefore be present in
the embedded clause of (3) when the NPI is licensed:
(1) John hasnt/*has visited her in at least two years.

(2) *John doesnt claim [that Mary has visited her [NPI in at least two years]]
(3) John doesnt believe [that Mary has visited her [NPI in at least two years]]
Bokovi (op. cit.) notes that whether or not a language allows clause-mate NPI-
licensing under NR out of indicative clauses depends on whether it has arti-
cles, establishing 1.1. (Thus, article-less languages such as Serbo-Croatian (SC),
Czech, Slovenian, Polish, Russian, Ukrainian, Korean, Japanese, and Chinese
disallow such licensing, while English, German, Spanish, French, Portuguese,
1 TNP is a neutral term that does not take a stand on the potential presence of functional
projections in this domain.
2 Although we will here investigate Turkish with respect to the generalizations in question, it
should be noted that since these generalizations involve typological arguments, they really
cannot be dissolved by looking at a single language even if the language should turn out
to provide some exceptions; this would merely turn some of the generalizations below into
strong tendencies, which would still call for an explanation. (Needless to say, even showing
conclusively that some of the generalizations below are incorrect would not affect other
generalizations.)
104 bokovi and ener
Romanian, and Bulgarian, which have articles, allow it.) Note that the general-
ization concerns only indicatives (i.e. finite complement clauses), not gerunds,
subjunctive(-style) and infinitival clauses. Turkish clause-mate NPIs cannot be
licensed long-distance within typical indicatives, even under typical raising
verbs like san think/believe. Turkish thus patterns with NP languages here.
(4) Pelin Mete-yi en az iki yl-dr ziyaret et-me-di/*et-ti.

P.-nom M.-acc at least two year-for visit do-neg-past/do-past
Pelin hasnt/*has visited Mete in at least two years.
(5) Mete [Pelin-/-i (*en az iki yl-dr) Timbuktu-ya git-ti]

M.-nom P.-nom/-acc at least two year-for T.-dat go-past
san-m-yor.
think-neg-pres
Mete doesnt think Pelin went to Timbuktu in at least two years.
1.2 Article-Less Languages Disallow Transitive Nominals with Two

Lexical Genitives
This section concerns the availability of structures where both the external
(not simply a possessor, but a true external argument) and the internal argu-
ment of a noun are genitive, with the genitive realized via a clitic/suffix or a
dummy preposition. Such cases are disallowed in article-less languages (which
dont otherwise allow multiplication of the same case like Japanese; e.g. Polish,
Czech, Russian, Latin, SC, Ukrainian, Chinese, and Quechua disallow it). The
same holds for Turkish.
(6) a. *Osmanllar-n stanbul-un feth-i

Ottomans-gen stanbul-gen conquest-3sg.poss
Ottomans conquest of Istanbul.
b. cf. stanbul-un feth-i
1.3 Only Article-Less Languages May Allow Scrambling

Bokovi (2008a, 2012a) observes that traditional scrambling languages (e.g.
Chukchi, Chichewa, Hindi, Japanese, Korean, Latin, Serbo-Croatian, Polish,
Czech, Slovenian, and Warlpiri) all lack articles.3 As is well-known, Turkish is a
scrambling language, hence fits 1.3 as an NP language.
3 What is meant by scrambling here is the kind of movement referred to as scrambling in

the turkish np 105
1.4 Radical Pro-Drop is Possible Only in Article-Less Languages

Bokovi (2012a) defines radical pro-drop as the productive pro-drop of sub-
jects and objects in the absence of rich verbal agreement. This type of pro-drop
differs from pro-drop in Spanish, which is licensed by verbal morphology. As a
result, since Spanish has subject but not object agreement, pro-drop is allowed
only with subjects. Radical pro-drop is allowed in Japanese, Chinese, Korean,
Kokota, Hindi, Wichita, Malayalam, Thai, Burmese, and Indonesian, all NP lan-
guages. Turkish has subject agreement and Spanish-style subject-drop. How-
ever, it also has productive object-drop although it lacks object agreement (see
ener and Takahashi 2010), hence fits 1.4 as an NP language.
1.5 Negative Concord Reading May Be Absent with Complex Negative

Constituents Only in Article Negative Concord Languages
In some negative concord languages, like Italian, the negative concord reading
is unavailable with complex negative constituents.
(7) a. Non ho visto nessuno.

neg have seen nobody
I didnt see anybody. (Negative Concord only)
b. Nessuno studente ha letto nessun libro.

no student has read no book (Double Negation only)
Bokovi (2012a) shows that while DP languages differ as to whether the double
negation reading is forced in examples like (7b), in NP languages (e.g. SC,
Russian, Polish, Ukrainian, Japanese, and Korean) only the negative concord
reading is allowed here. The same holds for Turkish:
(8) Hibir ocuk hibir kitab- oku-ma-d.

no child-nom no book-acc read-neg-past
No child read any book. (Negative Concord/*Double Negation)
Japanese, which can take place long-distance out of finite clauses, not the type discussed for
German, whose scrambling is a very different operation with very different semantic effects
from Japanese scrambling (importantly, it also cannot take place long-distance).
106 bokovi and ener
1.6 Possessors May Induce an Exhaustivity Presupposition Only in

Article Languages
Partee (2006) notes that while (9) has the presupposition Zhangsan has exactly
three sweaters, (10) in Mandarin doesnt have that exhaustivity presupposition,
although it is definite:
(9) Zhangsans three sweaters
(10) Zhangsan de [san jian maoxianyi]

Z deposs three CL sweater
Zhangsans three sweaters
Bokovi (2012a) shows that we are dealing here with a broader generaliza-
tion, with the exhaustivity presupposition absent from NP languages (e.g. Rus-
sian, SC, Chinese, Japanese, Korean, Hindi, Bangla, Malayalam, and Magahi) in
such contexts. Turkish again patterns with NP languages: (11) doesnt have the
exhaustivity presupposition:4
(11) Can-n bisiklet-i

John-gen three bicycle-3sg.poss
Johns three bicycles
1.7 Only Article-Less Languages May Allow LBE

Bokovi (2012a) shows that only article-less languages may allow AP left-
branch extraction (LBE). (Thus, such extraction is allowed in Russian, Polish,
Czech, Ukrainian, Slovenian, Latin, Mohawk, Southern Tiwa, Gunwinjguan lan-
guages, Hindi, Bangla, Angika, and Magahi, all article-less languages). Note that
the development of a definite article has led to the loss of such LBE in Ancient
Greek and Colloquial Finnish.)
(12) Dorogujui on videl [ti mainu] (Russian)

*Expensivei he saw [ti car]
It seems Turkish should be classified as a non-LBE language (Turkish does allow

possessor extraction).
4 The interpretation of the possessor may vary according to the syntactic context, in Turkish
and Chinese, see section 2.
the turkish np 107
(13) a. Pelin [kaln kitap] oku-du.

P-nom thick book read-past
Pelin read a thick book.
b. *Kaln1 Pelin [t1 kitap] oku-du.
This still wouldnt provide an argument for DP in Turkish, given the one-way
correlation status of 1.7. In fact, Bokovi (2012a) notes that LBE often requires
A-N agreement, which Turkish doesnt have. However, LBE is actually allowed
in Turkish: an adjective can be postposed to the post-verbal field (Kornfilt 2003,
Gksel and Kerslake 2005).
(14) ?Pelin [t1 kitap] oku-du kaln1.
Note that postposing in Turkish only involves non-contrastive elements that

are given in the previous discourse, classified as [contrastive,+discourse ana-
phoric] in ener (2010). Fronting however typically involves [+contrastive] con-
stituents, which may also be [+topic] or [+discourse anaphoric]. As for foci,
they remain in-situthey cannot undergo movement no matter whether they
are + or -contrastive (ener 2010). Kaln in (13)a, for instance, must remain
inside the NP if it is [+focus]. We therefore conclude that adjectival LBE is not
categorically ruled out in Turkish, its impossibility in fronting contexts being
due to discourse factors (such adjectives must be [contrastive,+discourse ana-
phoric] and all movement of such elements in Turkish is to the right).5
1.9 Only Article Languages Allow the Majority Superlative Reading

ivanovi (2008) reports that (15) in Slovenian only allows the plurality reading.
5 See Kural (1997), Kornfilt (2005), and ener (2010) for arguments based on c-command
relations, showing that postposing involves rightward movement (it cannot be handled in
terms of base-generation or remnant movement). Note that Slavic adjectival LBE is also
subject to discourse restrictions. There are also syntactic constraints on it: e.g., it is disallowed
from an NP that is an argument of another noun (Bokovi 2012a, in press b). Significantly,
Turkish patterns with Slavic, which suggests that AP post-posing in Turkish should be indeed
treated in the same way as Slavic adjectival LBE. ((i) is fine if yal modifies the higher N.)
(i) *Pelin [[pro ti teyze-m]-in arkada-]-n gr-m

P-nom aunt-1sg.poss-gen friend-3sg.poss-nom see-evidential.past
yali.
old
Pelin saw the friend of my old aunt.
108 bokovi and ener
(15) Najve ljudi pije pivo.

most people drink beer
Plurality reading (PR): More people drink beer than any other beverage
(though it could be less than half the people).
Majority reading (MR): *More than half the people drink beer.
English most allows both readings, though in different contexts. German most
also has both readings: (16) is ambiguous in this respect (PR requires focus on
beer).
(16) Die meisten Leute trinken Bier.

the most people drink beer.
ivanovi (2008) notes that a broader generalization is at stake here, where

the Majority Reading is allowed only in article languages (ivanovi 2008 and
Bokovi 2012a note that English, German, Dutch, Hungarian, Romanian,
Basque, Arabic, Macedonian and Bulgarian, which have articles, allow the
Majority Reading, while Slovenian, Czech, Polish, SC, Hindi, Chinese, Punjabi,
Angika, and Magahi, which do not have articles, disallow it). Note that 1.9 con-
cerns only traditional determiners, not the cases where the Majority Reading
is expressed by nouns such majority or Turkish ou (as in insanlar-n o-u
majority of (the) people).
Gajewski (2011) shows that Turkish disallows the Majority Reading: the most
natural interpretation of (17) is that events of beer drinking outnumbered
events of drinking any other beverage.
(17) nsanlar en ok bira i-ti.

people-nom most beer drink-past
People drank beer the most.
Gajewski notes that the Majority Reading can be contextually inferred from the
Plurality Reading. Thus, (17) may be interpreted indirectly as counting other
objects with certain background assumptions; i.e., one might get the Majority
Reading as an inference. However, Gajewski shows that the Majority Reading
is unavailable in (17) under the scenario in (18), although it enforces it ((17)
cannot be truthfully uttered in this context). Turkish thus again patterns with
NP languages.
(18) Suppose people at a dinner were allowed more than one beverage. 60 %
of the people had a beer. 75% of the people had a glass of wine.
the turkish np 109
1.10 Inverse Scope is Unavailable in Article-Less Languages (in Some

Examples)
Another generalization from Bokovi (2012a) concerns inverse scope inter-
pretation of examples like (19), with the unmarked word order (not involving
movement) for the language. This interpretation is available in English.
(19) Someone loves everyone.
Bokovi (2012a) observes that while DP languages differ as to whether or not

they allow inverse scope in such contexts (with most DP languages examined
allowing it), this interpretation is as a rule disallowed in NP languages (e.g.
Japanese, Korean, Turkish, Persian, Hindi, Bangla, Chinese, Russian, Polish,
Slovenian, Ukrainian, and SC). As is well known, Turkish exhibits here the
general behaviour of NP languages. Thus, the object cannot scope over the
subject in (20).
(20) iki renci her sandalye-yi kr-m.

two student-nom every chair-acc crush-evidential.past
Two students crushed every chair.
1.11 Number Morphology May Not Be Obligatory Only in Article-Less

Languages
Consider (21), where the N can be interpreted as plural in the absence of plural
morphology.
(21) Susumu-ga hon-o yonda. (Japanese)

Susumu-nom book-acc read
Susumu read a/the book/books.
Bokovi (2012a) notes languages without obligatory number morphology in

non-numeral contexts (where some or all countable Ns can receive plural
interpretation without number morphology, as in Japanese, Korean, Chinese,
Hindi, Bangla, Malayalam, Mohawk, Dyirbal, Warlpiri, Kuku-Yalanji, Warrga-
may, Indonesian, Vietnamese), all lack articles, establishing 1.11. Note however
that 1.11 doesnt conversely require all NP languages to lack number morphol-
ogy. While Turkish may seem to require it, it also productively allows cases like
(22), where book can be interpreted as plural in the absence of plural morphol-
ogy (its accusative counterpart cannot be so interpreted, see Aydemir 2004,
Ketrez 2005, ztrk 2004, ener 2010).
110 bokovi and ener
(22) Can (kaln) kitap oku-mu.

John-nom thick book read-evidential.past
John read a (long) book / (long) books.
Based on the above, we conclude that Turkish patterns with other article-less
languages studied by Bokovi (2012a), which provides motivation for classify-
ing it as an NP rather than a DP language.6 In what follows we will therefore
adopt this position.
2 The Structure of Turkish NPs
Before investigating word order within Turkish TNPs, we will determine the
position of possessors, which we will use as a pivot in the attempt to understand
the distribution of other TNP-elements. The reason for this is that there is a
6 One of the NP/DP generalizations involves adjunct extraction from TNPs, as in *From which
cityi did Peter meet [girls ti]. It is claimed that such extraction may be possible only in NP
languages. Turkish, however, disallows it:
(i) a. *okul-dani Mete [ti kz-lar]-la sinema-ya git-ti.

school-abl M-nom girl-pl-with movies-dat go-past
b. *Mete [ti kz-lar]-la sinema-ya git-ti okul-dani.

Mete went to the movies with girls from school.
However, we are dealing here with a one-way correlation: such extraction cannot be possible
in DP languages, but can be allowed or disallowed in NP languages (the lack of articles is not
the only factor, see Bokovi in press b for discussion of the phenomenon).
A number of Bokovis generalizations are irrelevant because Turkish doesnt have the
relevant constructions. This concerns the generalizations regarding head-internal relatives,
pronominal clitics, multiple-wh-fronting, obligatory classifiers, and focus-movement (see
also a suggestion made in Bokovi 2012a regarding Turkish and a generalization concerning
negative constituents). Regarding the focus-movement generalization, according to which
elements undergoing focus-movement are subject to a V-adjacency requirement only in
DP languages, it should be noted that Turkish requires linear adjacency of (non-D-linked)
wh-phrases/foci to V. However, ener (2010) argues this doesnt stem from the movement of
Wh/Foci and V to the Spec and Head position of a single projection in the left periphery, but is
a consequence of left-peripheral movement of all but Wh and Foci; Wh/Foci remain adjacent
to V in Turkish because only Wh/Foci and the verb dont undergo movement. This makes the
focus-movement generalization irrelevant to Turkish.
the turkish np 111
rather straightforward test proposed in Despi (2011, 2013) that determines the
position of possessors.
In his NP analysis of Serbo-Croatian (SC), Bokovi (2012a) treats SC posses-
sors and demonstratives as NP-adjuncts. One of the arguments for this analysis,
noted by Despi (2011, 2013), is provided by (24), which contrasts with English
(23) in that the pronoun and the name cannot be co-indexed. Given that the
possessor is an NP-adjunct and that SC lacks DP, the possessor c-commands
out of the TNP, which results in Condition B/C violations in (24).7
(23) a. Hisi latest movie really disappointed Tarantinoi.

b. Tarantinois latest movie really disappointed himi.
(24) a. *[NP Kusturicini [NP najnoviji film]] gai je zaista razoarao.

Kusturicas latest movie him is really disappointed
Kusturicais latest movie really disappointed himi.
b. *[NP Njegovi [NP najnoviji film]] je zaista razoarao Kusturicui.

his latest movie is really disappointed Kusturica
Significantly, Turkish patterns with SC, not English. As (25) shows, the pos-
sessor apparently c-commands out of its TNP, as these sentences are clear
violations of Conditions B/C. We therefore assume that possessors are also NP-
adjoined in Turkish, with the DP-layer missing in this language.
(25) a. *[zpeteki-in film]-i oi-nu hayal krklna urat-t.

.-gen movie-3sg.poss he-acc disappoint-past
zpeteks movie disappointed him.
b. *[oi-nun film]-i zpeteki-i hayal krklna urat-t.

he-gen movie-3sg.poss .-acc disappoint-past
His movie disappointed zpetek.
7 Japanese and Chinese pattern with SC (Bokovi 2012a, Cheng 2013, Takahashi 2011). It is,
however, not out of question that in some NP languages possessors could be in SpecNP or
even function as N-complements (like English of-genitives), in which case they wouldnt
c-command out of the TNP. Takahashi notes that for some speakers, relational nouns (like
father) behave differently due to an interfering factor which is not relevant here. Since
contrastive focus can affect binding relations, it also needs to be controlled for, see Bokovi
(2012a) (the co-indexed elements shouldnt be focused). Note also that Condition A cannot
be tested here due to interfering factors, see Despi (2011).
112 bokovi and ener
Now consider word order within Turkish TNPs:8
(26) PossDem(A)Num(A)N
a. Can-n u (eski) (eski) bisiklet-i
John-gen that old three bicycle-3sg.poss
those three old bicycles of Johns
*PossNumDemAN
b. *Can-n u eski bisiklet-i
*NumPossDemAN
c. * Can-n u eski bisiklet-i
*PossADemNumN
d. *Can-n eski u bisiklet-i
*APossDemNumN
e. *eski Can-n u bisiklet-i
(27) DemPoss(A)Num(A)N
a. u Can-n (eski) (eski) bisiklet-i
*DemNumPossAN
b. *u Can-n eski bisiklet-i
*NumDemPossAN
c. * u Can-n eski bisiklet-i
*DemAPossNumN
d. *u eski Can-n bisiklet-i
*ADemPossNumN
e. *eski u Can-n bisiklet-i
8 (26e)/(27d) are possible on the irrelevant interpretation where there are multiple individuals
with the name John and the one that is considered as the former (for whatever reason) is
talked about here.
the turkish np 113
Poss must precede Num/Adj, but can precede or follow Dem. Num and Adj
may shift order but these are the only options for them. As regards Poss/Num/A,
their order follows if Poss is an NP-adjunct, as discussed above, and if Num
and Adj hold NP spec positions. The free order for Num and Adj is then a
consequence of the free order for the specifiers, but they both must follow
Poss.9
Recall that Bokovi (2012a) argues that both possessors and demonstratives
are NP-adjuncts in SC. (27)a, where Dem precedes Poss, can be accounted for if
Dem is also NP-adjoined in Turkish. This provides a simple account for the fact
that Dem precedes Num and Adj, but may precede or follow Poss.10
9 We are treating numerals and adjectives differently from Bokovis account of SC. Numerals
actually have a rather peculiar behaviour in SC, and SC adjectives differ in several respects
from Turkish adjectives (e.g. they are much more mobile and agree in case/phi-features).
We leave open the question whether these differences can be unified with the different
structures for the elements proposed here and in Bokovi (2012a). (However, note that we
follow Bokovis 2009, 2012b analysis of adjectives, where adjectives are located in multiple
Specs of the same phrase, rather than Cinques 1994 approach.)
10 Bir is traditionally claimed to be homophonous between an indefinite article and the numeral
one. Even if bir were an indefinite article, this wouldnt require adopting a DP analysis for
Turkish, given that Slovenian, which clearly has indefinite but not definite articles, behaves
like NP languages in all respects, including in cases where indefinite articles are present
(Bokovi 2008b). Ketrez (2004), however, shows that the two birs have identical syntac-
tic distributions and that they are furthermore distributionally identical with other cardinal
numbers. Thus, like other numerals, bir can precede or follow adjectives. Turkish TNPs in gen-
eral can be interpreted as specific or non-specific, depending on the context and prosody. The
same holds for bir/numeral phrases. Thus, while bir/iki can either follow or precede mavi in
(i), phonological prominence on bir/iki in either position favours the specific interpretation,
while the lack of phonological prominence favours the non-specific interpretation (see also
Ketrez 2004, ztrk 2004; for another parallel, see footnote 16). It thus appears that there is
no reason to give bir a fundamentally different treatment from other numerals (due to space
limitations we cannot discuss bir further here).
(i) a. Ali mavi BR/K bisiklet al-d.

A-nom blue one/two bicycle buy-past
Ali bought one/two blue bicycle(s).
b. Ali bir/iki mavi bisiklet al-d.

114 bokovi and ener
(28)
A significant prediction of this structure is this: Poss should c-command out

of its TNP even when preceded by Dem, since even in this case Poss is not
dominated by the TNP. The prediction is borne out. (SC behaves in the same
way, see Despi 2011, Bokovi 2012a.)
(29) a. *[u zpeteki-in film]-i oi-nu hayal krklna urat-t.

that .-gen movie-3sg.poss he-acc disappoint-past
That movie of zpeteks disappointed him.
b. *[u oi-nun film]-i zpeteki-i hayal krklna urat-t.

that he-gen movie-3sg.poss .-acc disappoint-past
That movie of him disappointed zpetek.
The violations of Conditions B/C in (29) provide clear evidence that the Dem
that precedes Poss is not in a separate projection (it doesnt close off the
c-command domain of Poss).11 We take this to be a strong argument for the
claim defended here that Turkish TNPs lack DP.
11 We leave open the question why (i) is only somewhat degraded.
(i) ?[zpeteki-in u film]-i oi-nu hayal krklna urat-t.

.-gen that movie-3sg.poss he-acc disappoint-past
Notice that (ii) is fully unacceptable (due to a Condition C violation).
(ii) *[oi-nun u film]-i zpeteki-i hayal krklna urat-t.

he-gen that movie-3sg.poss .-acc disappoint-past
That movie of his disappointed zpetek.
the turkish np 115
To complete the paradigm, we provide examples where Poss precedes Num,

Num+Classifier, and Adj, which are all ungrammatical, due to Condition B
violations:
(30) a. *[zpeteki-in iki (tane) film]-i oi-nu

.-gen two CLL movie-3sg.poss he-acc
hayal krklna urat-t.
disappoint-past
Two movies of zpeteks disappointed him.
b. *[zpeteki-in eski film]-i oi-nu hayal krklna urat-t.

.-gen old movie-3sg.poss he-acc disappoint-past
zpeteks old movie disappointed him.
We now return to Generalization 1.6, exploring its relevance for TNP-structure.

Recall that, in contrast to English (9), Turkish (31) and Mandarin (32) dont have
the presupposition Z has exactly three sweaters:
(31) Can-n bisiklet-i

John-gen three bicycle-3sg.poss
Johns three bicycles
(32) Zhangsan de [san jian maoxianyi]

Z deposs three cl sweater
Zhangsans three sweaters
Mandarin also allows the order in (33), which implies that Zhangsan has more
than three books (see Partee 2006):
(33) san ben [Zhangsan de] shu

three cl Z deposs book
three of Zhangsans books
Recall that PossNumN is the only licit order in Turkish. The order can yield
different interpretations in different contexts, which can be seen when NPs
are placed in clauses. Thus, (34), where the possessive NP is embedded in
a sentence with a locative predicate, implies that John has more than three
bicycles:
116 bokovi and ener
(34) Can-n bisiklet-i garaj-da.

John-gen three bicycle-3sg.poss garage-loc
Johns three bicycles are in the garage.
If we take the Mandarin word order to transparently reflect LF, (33) should be
taken as indicating that the more-than-# reading requires Poss to be interpreted
within the scope of Num. If in Turkish NPs with this reading, Poss should also
be interpreted inside Num, there should be a Poss position below Num. We
therefore modify (28) by assuming Poss is merged below Num, presumably as
an N-complement, and then adjoins to NP.12
(35)
On the more-than-# reading, which in Mandarin requires the NumPoss order,

Poss is therefore interpreted in its reconstructed position.
Bokovi (2007) shows that SC possessors can precede or follow adjectives.
The order permutations in (36) have semantic effects. (36b) can only refer to
the pants John formerly owned. To refer to an object John now possesses and
that used to be pants (it could be shorts now) (36a) must be used. Importantly,
Larson and Cho (1999) argue that under the former but not the latter reading,
Poss must be interpreted within the scope of A, which is transparently reflected
in the SC word order.
(36) a. Jovanove bive pantalone

Johns former pants
b. bive Jovanove pantalone
12 This movement violates Bokovis (2005, in press b) version of anti-locality, which requires
movement to cross at least one full phrase. Larson and Cho (1999), however, argue such
examples involve richer structure (a null PP, with Poss starting as a P-complement), in which
case (35) can be modified so that the anti-locality problem doesnt arise.
the turkish np 117
While Turkish only allows the PossA order, (37) is ambiguous. Extending
the above analysis to Turkish requires Poss to start below A, where it would
reconstruct on the reading it shares with (36b).
(37) Can-n eski pantolon-u

John-gen former pants-3sg.poss
Johns former pants
The proposal is then that Poss in Turkish is base-generated low but moves to
a higher position. An interesting prediction arises when this proposal is com-
bined with Takahashis (1996, 2000, 2001) claim that pro doesnt move (see these
works and Ochi 2005 for evidence to this effect). The prediction is that null
possessors inside subject NPs should not induce Condition C violations, unlike
overt possessors, since such possessors wouldnt move to the NP-adjoined posi-
tion (see Kornfilt 1984, Sezer 1991 on pro possessors in Turkish, which are
agreement-licensed). This prediction is borne out (compare (38) and (29b); the
binding violations from (29)(30) are all avoided with pro possessors).
(38) [u proi film]-i zpeteki-i hayal krklna urat-t.

that movie-3sg.poss .-acc disappoint-past
That movie of his disappointed zpetek.
Furthermore, (39) has the implication that I have more than three books
(which is not forced with overt possessors like Pelin-in). This also follows if
pro-poss doesnt move, but stays in the scope of Num.13,14
13 The impossibility of stressing pro interferes with the adjectival interpretation test.
14 Another construction that can be captured by the pro-doesnt-move analysis is (i), where what
appears to be an anaphor does not induce a Condition C effect.
(i) [kendii film]-i zpeteki-i hayal krklna urat-t.

own movie-3sg.poss .-acc disappoint-past
His own movie disappointed zpetek.
Given that kendi can co-occur with overt possessors (cf. (ii)) and given its interpretation, we
analyse it as an emphatic element here. We can then account for the lack of Condition-C
effects in (i) by assuming that the subject NP in (i) has a pro possessor, with kendi right-
adjoined to it (see Aygen 2002 and ener 2008 for such adjunction analyses of emphatic
kendi). Since pro doesnt move, there is no Condition C violation in (i). As expected, an overt
pronominal possessor accompanied by kendi induces a Condition-C violation (iia). (iib), on
the other hand, involves a Condition-B violation.
118 bokovi and ener
(39) [ pro kitab-m]- ev-de brak-t-m.

three book-1sg.poss-acc home-loc leave-past-1sg
I left three of my books at home.
To conclude, we have shown that only the orders listed in (40) are allowed
in Turkish and accounted for this by treating Poss and Dem as NP-adjuncts
and Num and A as NP-Specs. Overt possessors move to this position, while pro
possessors stay in a lower position.
(40) a. PossDem NumAN b. PossDemANumN

c. DemPossANumN d. DemPossNumAN
3 NP/N-Ellipsis
3.1 Impossible Cases

We now turn to TNP-internal ellipsis, referred to below as NP/N-ellipsis. Notice
first that ellipsis inside bare objects with numerals is disallowed.
(41) *Pelin her gn [be elma] ye-r, Can-sa [iki elma]

P.-nom every day five apple eat-aor J.-nom-however two
ye-r.
eat-aor
Pelin eats five apples every day, while John eats two.
Ellipsis inside bare objects with adjectives is also disallowed (the NPs number
is irrelevant):15
(42) a. *Pelin [eski kitap] sat-t, Suzan-sa [yeni kitap] sat-t.

P.-nom old book sell-past S.-nom-however new sell-past
Pelin sold old books, while Susan sold new ones.
(ii) a. *[on-un kendii film]-i zpeteki-i hayal krklna urat-t.

he-gen own movie-3sg.poss .-acc disappoint-past
His own movie disappointed zpetek.
b. *[zpeteki-in kendii film]-i oi-nu hayal krklna urat-t.

.-gen own movie-3sg.poss he-acc disappoint-past
zpeteks own movie disappointed him.
15 See section 3.2.2 for accusative objects.

the turkish np 119
b. *Pelin [eski araba-lar] sat-t, Suzan-sa [yeni araba

P.-nom old car-pl sell-past S.-nom-however new
-ler] sat-t.
-pl sell-past
Pelin sold old cars, while Susan sold new ones.
NP/N-ellipsis also cannot strand possessors:
(43) *[Pamuk-un kitab--n] oku-du-m, ama [Oe-nin kitab--n]

P.-gen book-3sg.poss-acc read-past-1sg but O.-gen
oku-ma-d-m.
read-neg-past-1sg
I read Pamuks book, but I didnt read Oes.
The unacceptability of all the examples in (41)(43) receives a simple unified

account under the present proposal that numerals, adjectives, and possessors
are all NP-specifiers/adjuncts in Turkish, as in (44), given that only phrases (not
segments or bar-level categories) can be elided. We take this to be a strong
argument for the current analysis.16
(44)
16 It is well-known that, in contrast to accusative objects, bare objects must incorporate into
the verb in Turkish. There is a potential alternative analysis of the ellipsis facts that relies
on the assumption that Ns with numerals/adjectives must incorporate, which would make
ellipsis of Ns that strands V impossible. (This analysis is inapplicable to possessive NPs,
which require accusatives as direct objects.) However, a test from Aydemir (2004) shows
that Ns with numerals are not V-incorporated, though Ns with adjectives may be analysed
as V-incorporated. Consider (i), where deleting N leads to ungrammaticality (ia), whereas
ellipsis of the entire nominal complement clause is grammatical (ib). Deleting Bir+NP is
allowed, with or without V-ellipsis (ii).
(i) a. Btn gn kitap oku-du-m, *san-a da kitap oku-ma-n-

all day book read-past-1sg you-dat too read-noml-agr.2sg-acc
tavsiye ed-er-im.
recommend-aor-1sg
I read books all day, I recommend to you to read too.
b. Btn gn kitap oku-du-m, san-a da kitap oku-ma-n- tavsiye ed-er-im.

120 bokovi and ener
We now turn to classifier constructions. Count nouns quantified by a nu-

meral dont require a Classifier-like element (CLL). However, CLL can be pres-
ent, bringing in an individuating function (ztrk 2004).
(ii) Dn bir kitap oku-du-m, san-a da {bir kitap

yesterday one book read-past-1sg you-dat too
oku-ma-n-/ bir kitap oku-ma-n-} tavsiye ed-er-im
read-noml-agr.2sg-acc recommend-aor-1sg
I read a book yesterday, I recommend to you to read (it) too. Aydemir (2004)
This can be taken as a test for obligatory incorporation: given that the bare NP in (i) must
incorporate into the verb it cannot be elided without it. It must then be the case that the bir
NP doesnt (have to) incorporate.
This test indicates numeral NPs dont (have to) incorporate:
(iii) Dn iki kitap oku-du-m, san-a da iki kitap oku-ma-n-

yesterday two book read-past-1sg you-dat also read-noml-2s.poss.acc
tavsiye ed-er-im.
recommend-aor-1sg
I read two books yesterday, I recommend you read them too.
(iv) Dn iki kitap oku-du-m, san-a da iki kitap oku-ma-n- tavsiye ed-er-im.
Bare NPs with adjectives behave differently:
(v) a. Ben souk ay i-ece-im, ??san-a da souk ay

I-nom cold tea drink-fut-1sg you-dat also
i-me-n-i tavsiye ed-er-im.
drink-noml-2sg.poss.acc recommend-aor-1sg
I will drink cold tea, I recommend you drink it too.
b. Ben souk ay i-ece-im, san-a da souk ay i-me-n-i tavsiye ed-er-im.
However, the grammaticality status of (va) is not the same as that of (42). We interpret this as
indicating that an additional factor is at play with (42), which we have discussed above.
Also, ellipsis is impossible in the subject NP in (via), where incorporation is clearly not an
issue. The incorporation analysis therefore cannot be extended to this case. (Partial subject
ellipsis is also impossible with numerals and possessives.)
(vi) a. *[Yal bakc] ocuk-lar azarla-d, ama [gen bakc] onlar-

old caretaker-nom kid-pl-acc scold-past but young they-acc
teselli et-ti.
console-past
The old caretaker scolded the students, but the young one tried to console
them.
the turkish np 121
(45) a. kitap
three book
b. tane kitap
three CLL book
three (items of) books
Importantly, NP/N-ellipsis is possible if Num is accompanied by CLL, in con-

trast to (41), where there is no CLL and ellipsis is disallowed.17
(46) Pelin her gn [ tane elma] ye-r, Can-sa [iki

P.-nom every day three CLL apple eat-aor J.-nom-however two
tane elma] ye-r.
CLL eat-aor
Pelin eats three apples every day, whereas John eats two.
Bokovi (2012a, in press a,b) and Despi (2011) argue that non-adjectival nu-
merals in SC have a distinct projection above NP, namely QP. (SC also has
adjectival numerals which dont project additional structure.) Assuming that
Turkish numerals accompanied by a CLL also project a larger structure than
those without a CLL provides us with a straightforward account for (46). Con-
sider (47).
b. *Bu dnem [ renci] snfta kal-d, ama geen dnem [be

this semester three student flunk-past but last semester five
renci] snfta kal-m-t.
flunk-past.perfect
This semester three students flunked, whereas in the past semester five students
had flunked.
c. *[Pelin-in anne-si] dn gel-di, [Mete-nin anne-si] bugn

P.-gen mother-3sg.poss yesterday come-past M.-gen today
gel-di.
come-past
Pelins mother came yesterday, Metes came today.
17 Such examples require a linguistic antecedent, which confirms that we are indeed dealing
here with ellipsis: (with occasional exceptions, cf. e.g., Elbourne 2005) true ellipsis requires a
linguistic antecedent, cf. Hankamer and Sag (1976). Thus, (i) cannot be used in this situation:
Two people are shopping for groceries. Pointing to apples, A says:
(i) #Pelin her gn tane ye-r.

P.-nom every day three CLL eat-aor
122 bokovi and ener
(47)
Recall numerals are base-generated in SpecNP (28). We suggest that Num

moves to SpecCLLP in the presence of overt classifiers. (46) thus involves
simple NP-ellipsis.18,19
Recall that possessors cannot be stranded under NP/N-ellipsis. Under a DP
analysis of Turkish, possessors are analysed as occupying SpecDP, which natu-
rally predicts that Poss-stranding ellipsis should be possible. The ungrammati-
cality of (48)(49) therefore shows that the DP approach to Turkish TNPs fails:
Turkish differs sharply from English regarding the possibility of Poss-stranding
ellipsis.
(48) *[Auster-n kitab--n] oku-du-m ama [Pamuk-un

A.-gen book-3sg.poss-acc read-past-1sg but P.-gen
kitab--n] oku-ma-d-m.
read-neg-past-1sg
I read Austers book, but I didnt read Pamuks.
18 Note that bare Num+CLL NPs dont incorporate (see footnote 16).
(i) Dn tane kitap oku-du-m, san-a da tane kitap

yesterday three CLL book read-past-1sg you-dat also
oku-ma-n- tavsiye ed-er-im.
read-noml-2sg.poss.acc recommend-aor-1sg
19 We assume that Num is in SpecCLLP and tane in CLL. Another possibility is that Num, which,
being non-branching, is an ambiguous XP/X0 element, is head-adjoined to tane (hence the
ellipsis here wouldnt conform to Saito and Murasugis 1990 Spec-Head agreement require-
ment, but this requirement anyway has exceptions, see Bokovi in press a). CLLP is then
head-initial, which is not implausible given that there are other cases of mixed-headedness
languages (the nominal domain is a separate domain that doesnt necessarily have to show
the exact same properties as the clausal domain in terms of headedness). There are, how-
ever, alternatives where CLLP can be head-final. One alternative is the following: The head of
CLLP is null and tane is in SpecCLLP. Num can be in the outer SpecCLLP, or adjoined to tane
in SpecCLLP. We will compare these options below.
the turkish np 123
(49) *[Auster-n iki kitab--n] oku-du-m, ama [Pamuk-un

A.-gen two book-3sg.poss-acc read-past-1sg but P.-gen
iki kitab--n] oku-ma-d-m.
read-neg-past-1sg
I read two books of Austers, but I didnt read two books of Pamuks.
As discussed above, in contrast to the DP analysis, the impossibility of pos-

sessor-stranding ellipsis in Turkish, and the Turkish/English contrast in this
respect, is straightforwardly accounted for under the current NP analysis of
Turkish, where Turkish possessors are NP-adjoined hence cannot be stranded
under NP-ellipsis. Notice also that the ill-formedness of (49) is absolute in that
stranding Num doesnt change the judgment (see (50)). The same holds for
demonstratives (51).20
(50) *[Auster-n iki kitab--n] oku-du-m, ama [Pamuk-un

A.-gen two book-3sg.poss-acc read-past-1sg but P.-gen
kitab--n] oku-du-m.
three read-past-1sg
(51) a. *[Sartre-n u kitab-n] oku-du-k ama [Beckettin

S.-gen that book-acc read-past-1pl but B.-gen
u kitab-n] oku-ma-d-k.
read-neg-past-1pl
We read that book of Sartres but didnt read that book/one of
Becketts.
b. *[Sartre-n u kitab-n] oku-du-k ama [Beckettin u kitab--n]

oku-ma-d-k.
c. *[u Sartre-n kitab-n] oku-du-k ama [u Beckettin kitab--n]
oku-ma-d-k.
All of this is expected under the current analysis, where (48)(51) in fact receive
a uniform account: NP-ellipsis fails to elide elements that belong to the
NP.
20 (51) (and the same holds for other unacceptable possessor-stranding examples) remain unac-
ceptable if Acc is not elided (see below for case and demonstratives).
124 bokovi and ener
Significantly, Poss-stranding ellipsis is not always disallowed; it is possible in

the presence of CLL.21
(52) ?Pelin [Chomsky-nin tane kitab--n]

P.-nom C.-gen three CLL book-3sg.poss-acc
oku-mu, ama [Foucault-nun iki tane kitab--n]
read-evidential.past but F.-gen two CLL
oku-mu.
read-evidential.past
Pelin read three books of Chomskys, but s/he read two books of Fou-
caults.
(53) cf. * ama Foucault-nun iki oku-mu.
The contrast between (50)/(53) and (52) is accounted for under the present
proposal that CLL projects its own phrase, assuming Poss is CLLP-adjoined.
(Poss must precede Num here, which, as discussed above, can be accounted
for if Poss is adjoined to the phrase whose Spec Num occupies.)
(54)
We take the contrast between (48)/(49) and (52) to provide a strong argument
for the current analysis. The analysis also accounts for (55), where Num+
CL is elided. ((55) would have to involve segment deletion, which is disallowed.)
21 The reduced NP in such examples requires a linguistic antecedent, which shows we are
dealing with ellipsis. Thus, (i) cannot be used in this situation: A and B are in a bookstore.
Pointing to Foucaults books, A says:
(i) #Pelin Foucault-nun iki tane oku-mu

P-nom F-gen two CLL read-evidential.past
the turkish np 125
(55) *Pelin [Chomsky-nin tane kitab--n]

P.-nom C.-gen three CLL book-3sg.poss-acc
oku-mu, ama [Foucault-nun tane kitab--n]
read-evidential.past but F.-gen
oku-ma-m.
read-neg-evidential.past
Regarding the binding properties of possessors, Condition-B/C effects are pre-

dicted to show up in this context too since CLLPs dont change Posss
c-command properties:
(56) a. *[zpeteki-in iki tane film]-i oi-nu

.-gen two CLL movie-3sg.poss he-acc
disappoint-past
zpeteks two movies disappointed him.
b. *[oi-nun iki tane film]-i zpeteki-i

he-gen two CLL movie-3sg.poss .-acc
disappoint-past
His two movies disappointed zpetek.
Recall that adjective-stranding ellipsis is impossible, which is not surprising

under the current analysis: since the adjective is in SpecNP, (57) cannot involve
full phrasal ellipsis.
(57) *Pelin [eski (iki) kitap] sat-m ama Mete [yeni

P.-nom old two book sell-evidential.past but M.-nom new
(iki) kitap] sat-m.
sell-evidential.past
Significantly, adjectives that precede CLL survive ellipsis, which can be easily
accounted for if the adjective is in CLLP here (whether it moves or is base-
generated there is irrelevant).22
22 This kind of reduced NPs also require a linguistic antecedent, which indicates we are dealing
with ellipsis here. Thus, (i) cannot be used in this situation: We are in a bookstore. Pointing to
the books on the shelves, I say:
126 bokovi and ener
(58) Pelin [kaln tane kitap] oku-du, Pnar-sa [ince

P.-nom thick three CLL book read-past P.-nom-however thin
tane kitap] oku-du.
three CLL read-past
Pelin read three long books, but Pnar read three short ones.
Furthermore, (59) is unacceptable. This is also expected: since the adjective is

part of CLLP, other elements in CLLP cannot be elided without the adjective.
(59) *Pelin [kaln tane kitap] oku-du, Pnar-sa [ince

P.-nom thick three CLL book read-past P.-nom-however thin
tane kitap] oku-du.
read-past
Adjectives in this context still must follow Poss, which indicates that they are
located in SpecCLLP. They can either follow or precede Num, which is not
surprising, given that both Num and Adj are Specs. Interestingly, adjectives
cannot intervene between Num and CLL, which may help us tease apart the
options from footnote 19.
(60) a. Pelin-in eski tane masa-s

P.-gen old three CLL desk-3sg.poss
b. *Eski Pelin-in tane masa-s

c. Pelin-in tane eski masa-s
d. *Pelin-in eski tane masa-s
If tane is CLL0, we need to assume that in (60c) we are dealing with a PF re-
ordering (unless eski is NP-adjoined, see below). If tane and Num are in differ-
ent CLLP-Specs, we need to assume that there is a PF-adjacency requirement
between the two, but PF re-ordering is not required for (60c). Finally, if Num is
adjoined to the CLLP-Spec where tane is located, no additional assumptions are
required: eski can then only precede or follow the Num+tane complex, depend-
ing on whether it is located in the higher or lower CLLP-Spec.
The above discussion leads us to assume that eski in (60c) is located in CLLP,
not NP. Independent evidence for this is provided by (61): to survive NP-ellipsis
(i) *Pnar ince tane oku-du.

P-nom thin three CLL read-past
the turkish np 127
here, ince must be located in CLLP. Is SpecNP still an option for adjectives
that follow CLL? This depends on what is elided in (62): deleting kaln+kitap
under NP-ellipsis would require this option to also be available. However, its
not easy to determine what is elided here, given that kitap-deletion doesnt
prevent the interpretation where Pnar read two long books (which is the only
interpretation under kaln+kitap deletion, and is the most natural interpreta-
tion here).
(61) Pelin [ tane kaln kitap] oku-du, Pnar-sa [

P.-nom three CLL thick book read-past P-nom-however three
tane ince kitap] oku-du.
CLL thin read-past
Pelin read three long books, while Pnar read three short ones.
(62) Pelin [ tane kaln kitap] oku-du, Pnar-sa [iki tane {kaln kitap/kitap}]
oku-du.
Returning to possessors, there is one context where possessor-stranding ellipsis

is possible even without CLLP: this is when possessives are used as predi-
cates:
(63) a. Senin kuzeninin bu kaza ok sevdiini biliyorum.

I know your cousin likes this sweater a lot.
O yzden, bu kazak artk [kuzen-i-nin
because-of-that, this sweater-nom now cousin-3sg.poss-gen
kaza-].
sweater-3sg.poss
Because of that, the sweater is now your cousins.
b. Bu benim kazam.
this is my sweater.
Bu da [sen-in/Pelin-in/kuzen-in-in kaza-]
this too you-gen/Pelin-gen/cousin-3sg.poss-gen sweater-3sg.poss
And this is yours/Pelins/your cousins sweater.
Following Bowers (1993) and Koster (1994), we assume that the relevant NPs
here are dominated by a predicate projection, PredP, with the possessor gen-
erated inside the NP and then moved up to SpecPredP (or adjoined to PredP).
This provides a simple account for Poss-stranding under NP ellipsis in this case,
as in (64):
128 bokovi and ener
(64)
However, a number of authors have argued that at least some cases of strand-
ed possessors in predicate positions involve an intransitive use of possessors,
without any kind of null elements (Partee and Borschev 2001, Tremblay 1989,
Zribi-Hertz 1997; note that, as discussed in section 3.2.1, Turkish stranded pos-
sessors allow only the possession reading,23 which is the one argued to involve
no null element and, according to Tremblay, is confined to predicate posi-
tions; it is, e.g., disallowed with objects). Under this approach, (63) doesnt
involve ellipsis, hence there is no need to posit PredP and Poss-movement
here.
In fact, such constructions dont require a linguistic antecedent, which sug-
gests we indeed may not be dealing with ellipsis here (see Zribi-Hertz 1997).
(65) A and B are planning to buy desks for several friends. In a store, A points to
a desk and says:
a. Bu (masa) Pelin-in ol-mal.
this (desk) P.-gen be-deontic.necessity
This desk should be Pelins.
A linguist is doing an experiment. She turns to the colleague and says (turn-
ing the informant over to his colleague):
b. u andan itibaren (o) sen-in.
from-now-on he-nom you-gen
From now on, hes yours.
Finally, we take a brief look at NPs containing demonstratives. Consider (66a),

which we argue below does not involve ellipsis, i.e. it should not be analysed as
in (66c).
23 On the possession reading this book is yours can be paraphrased as this book belongs to you
and on the relational reading as this book is your book (see section 3.2.1).
the turkish np 129
(66) a. [u [iki kitab]]- oku-du-m ama bu-nu oku-ma-d-m.

that two book-acc read-past-1sg but this-acc read-neg-past-1sg
Intended Reading: I read these two books but didnt read those two
books.
b. bu-nlar-
this-pl-acc
these
c. u [iki kitab]]- oku-du-m ama [bu [iki kitap]]-nu

that two book-acc read-past-1sg but this two book-acc
oku-ma-d-m.
read-neg-past-1sg
Intended Reading: I read these two books but didnt read those two
books.
Two forms, bu-nu and bu-nlar-, can be used in the second conjunct. (66a)
presupposes that the entity Dem picks out is a unique/single entity, which
indicates that the second conjunct cannot involve Num+N ellipsis (as in (66c))
because it simply is not semantically identical to the antecedent NP. (66b)
seems like a more likely candidate for an ellipsis analysis but we contend that
it shouldnt be analysed as ellipsis either. In (66b), Dem bears plural marking;
the elided constituent is thus potentially identical to the antecedent. However,
(66b) may mean these ten books in a relevant context; it doesnt have to be
interpreted as these two books (although this option is not excluded since two
is plural).
(67) shows it is impossible to strand a numeral in the second conjunct
NP:24
24 We assume that (67) involves ellipsis (see also (51)), unlike only-demonstrative examples like
(66). Acc is left behind in the elliptical TNP in (67) since demonstrative TNPs generally require
it (i); (67), however, remains unacceptable if Acc is elided.
(i) a. [bu iki kitap]*(-) oku-ma-d-m.

this two book-acc read-neg-past-1sg
b. Bu-*(nu) sev-er-im
this-acc like-aor-1sg
I like this.
130 bokovi and ener
(67) *[u [iki kitab]]- oku-du-m ama [bu iki [kitab]] -yi
that two book-acc read-past-1sg but this two -acc
oku-ma-d-m
read-neg-past-1sg
This is expected under the current analysis: being located in SpecNP, Num
cannot survive NP ellipsis.
Most importantly, stranding Dem+Num is possible with classifiers. As (68)
shows, the presence of CLL makes such ellipsis, which was disallowed without
CLL, possible.
(68) [u [iki (tane) kitab]]- oku-du-m ama [bu iki tane

that two (CLL) book-acc read-past-1sg but this two CLL
[kitap]] -yi oku-ma-d-m
-acc read-neg-past-1sg
I read those two books but didnt read these two.
This is exactly what is expected: the numeral here moves to SpecCLLP, kitap is
then a full NP, hence can be deleted alone under NP-ellipsis.25
Summarizing, simple Poss/Dem/Adj/Num-stranding ellipsis is disallowed
in Turkish. The NP analysis of Turkish provides a uniform account for this
set of constraints: since these elements are part of NP, it is not possible to
elide NP while stranding them. TNP-internal ellipsis is possible in classifier
constructions, where CLL0 takes NP as its complement. NP can then be elided,
with the material located in CLLP surviving ellipsis. We have considered two
possibilities for stranded possessors in predicate positions. Such examples
either involve PredP, with the possessor located in PredP hence outside of NP,
or they do not involve ellipsis at all, i.e. they involve an intransitive use of
possessors. Finally, only-demonstrative TNPs do not involve ellipsis; however,
cases where demonstratives co-occur with other material are not amenable
to such an analysis. While the NP analysis provides a rather straightforward,
uniform account of all the facts discussed in this section, it is very difficult to
see how these facts can be explained (especially in a uniform manner) under
the DP analysis.
25 We assume that, like Poss, Dem is CLLP-adjoined here (Dem can either precede or follow Poss,
but it must precede Num and Adj). Whether it moves to this position or not is immaterial.
the turkish np 131
3.2 No Ellipsis
In this section we discuss two cases which appear to involve ellipsis, arguing
they shouldnt be analysed in this way.
3.2.1 Pronominal ki
We have seen Turkish possessors cannot be stranded under NP/N-ellipsis:
(69) *[Pamuk-un kitab--n] oku-du-m, ama [Oe-nin kitab--n]

P.-gen book-3sg.poss-acc read-past-1sg but O.-gen
oku-ma-d-m.
read-neg-past-1sg
I read Pamuks book, but didnt read Oes.
However, (69) becomes acceptable in the presence of ki, which attaches to Poss:
(70) [Pamuk-un kitab--n] oku-du-m, ama [Oe-nin-ki-ni]

P.-gen book-3sg.poss-acc read-past-1sg but O.-gen-ki-acc
oku-ma-d-m.
read-neg-past-1sg
I read Pamuks book, but didnt read the one by Oe.
This ki has been identified as a pronominal element in the literature, referred to

as pronominal ki (see e.g. Lewis 1967, Gksel and Kerslake 2005).26 Adopting this
analysis, we argue that ki-NPs dont involve ellipsis, which is what Hankamer
(2004) also argues for. Evidence for this analysis comes from the observation
that ki can only attach to a possessor if the NP is reduced; it cannot be used
when the NP is fully represented. This means ki stands for the ellipsis site.
(71) *[Pamuk-un kitab--n] oku-du-m, ama [Oe-nin-ki

P.-gen book-3sg.poss-acc read-past-1sg but O.-gen-ki
kitab-]-n oku-ma-d-m.
book-3sg.poss-acc read-neg-past-1sg
26 Ki also attaches to locative/temporal expressions, turning them into nominal modifiers:
(i) dn-k hava

yesterday-ki weather
yesterdays weather
132 bokovi and ener
The claim that ki is a (pro)nominal element receives support from the

observation that ki-NPs must receive case and can bear plural morphology.
Consider also the following:
(72) [Auster-n iki kitab-]-n oku-du-m ama

A.-gen two book-3sg.poss-acc read-past-1sg but
[Pamuk-un-ki]-ni oku-ma-d-m.
P.-gen-ki-acc read-neg-past-1sg
I read two books by Auster, but didnt read the one by Pamuk.
(73) [Auster-n iki kitab-]-n oku-du-m ama

A.-gen two book-3sg.poss-acc read-past-1sg but
[Pamuk-un-ki-ler]-i oku-ma-d-m.
P.-gen-ki-pl-acc read-neg-past-1sg
I read two books by Auster, but didnt read the ones by Pamuk.
In (72), the ki-object cannot be interpreted like the first object, i.e. as two
books. This is expected under the no-ellipsis analysis. In (73), the ki-object
bears plural marking. Importantly, it is not necessarily interpreted as denoting
two books; there can be any number of books as long as its more than one.
Further support for the no-ellipsis analysis is provided by NPs containing
other nominal elements. The ki-NP in (74) cannot be interpreted as that book
by Beckett; it is simply interpreted as Becketts book.27
(74) [Sartre-n u kitab--n] oku-du-k ama

S.-gen that book-3sg.poss-acc read-past-1pl but
[Beckett-in-ki]-ni oku-ma-d-k.
B.-gen-ki-acc read-neg-past-1pl
We read that book by Sartre, but didnt read the one by Beckett.
27 Ki can only be supported by possessors (along with temporal/locative expressions, though

these could involve a different ki), not by a demonstrative (or Adj/Num/CLL, even if Poss
precedes them). We assume this is a morphological restriction.
(i) *[Sartre-n u kitab-n] oku-du-k ama [Beckett-in {u-nu-ki/u-ki-ni}]

S.-gen that book-acc read-past-1pl but B.-gen that-acc-ki/that-ki-acc
oku-ma-d-k.
read-neg-past-1pl
We read that book by Sartre but we didnt read that one by Beckett.
(ii) [u Sartre-n kitab-n] oku-du-k ama [u Beckett-in-ki-ni] oku-ma-d-k.

the turkish np 133
Finally, ki-NPs dont require a linguistic antecedent, which confirms they

dont involve ellipsis.
(75) I know my friends are sending me presents for my birthday. I come home and
find a number of presents on the table. Having opened one of them, I say:
Bu (hediye) Pelin-in-ki ol-mal.
this (present) P.-gen-ki be-epistemic.modal
This (present) must be Pelins.
Regarding possible crosslinguistic counterparts of ki, ki seems similar to the

Japanese pronoun sore, which can also be modified by possessives (Takahashi
2008).
(76) [Taroo-no taido]]-wa yoi ga, [Hanako-no sore]wa

-gen attitude-top good though -gen it-top
yoku nai
good not
Though Taroos attitude is good, Hanakos isnt.
Another parallel might be provided by English mine. Like ki-NPs, mine occurs
only with ellipsis (this is mine (*book)). However, ki-examples allow only the
relational, not the possessor reading (see Zribi-Hertz 1997 on these readings).
Thus, ki can occur instead of the gap in (63b) (on the relational reading), but
not (63a), which is unambiguously possessive (see Partee and Borschev 2001).
Ki is also inappropriate in (65) and (77), which are also possessive.
(77) [Pelin-in bu kitab]- aslnda Mete-nin-(*ki)

P.-gen this book-3sg.poss in-fact M-gen(-ki)
This book of Pelin actually belongs to Mete.
Ki is similar here to English one (while mine is ambiguous, my one is only

relational, see Allen 2008).
(78) a. antalarmz masada; benim*(-ki-ni) verir misin?

Our bags are on the table; Can you pass me my one/mine?
b. Borlar bittikten sonra ev benim(*-ki) olacak.
After I pay off the mortgage, the house will be mine/*my one.
Interestingly, ki is obligatory in the unambiguously relational (75), which sug-

gests non-ki stranded possessors may be unambiguously possessive. Note also
that ki is impossible in (79), which is possessive, but obligatory in (80).
134 bokovi and ener
(79) Bu kitap kim-in? Ben-im(*-ki)

this book who-gen I-gen-ki
Whose book is this? Mine
(80) Bu kitap-lar-n hangi-si ngilizce? Ben-im*(-ki)

this book-pl-gen which-3sg.poss English I-gen
Which of these books is in English?
At any rate, what is most important for our purposes is that ki-NPs dont involve
ellipsis. We discuss another similar case in the next section.
3.2.2 Stranded Adjectives

We have seen that adjectives cannot be stranded under ellipsis in examples like
(42). In many languages, adjectives can be used in what seem to be ellipsis con-
texts without a noun, but in such cases adjectives are essentially used as nouns,
hence such cases shouldnt be treated as involving ellipsis. Such cases often
require overt manifestation of this special use of adjectives, such as special mor-
phological marking on the adjective or use of an article, as in English the rich.
Bokovi (2005) observes that NP languages seem much more productive in
this respect. (For an explanation of this fact, see Bokovi 2013, who establishes
a semantic conditioni.e., a type-shifting conditionon the nominalization
of adjectives which is easier to satisfy in NP languages, due to a difference in
the semantic type of TNPs in NP and DP languages). Turkish patterns with
other NP-languages in that such usage of adjectives is quite productive.28 How-
ever, as in other languages, it generally requires special morphological marking
(see Bokovi 2013 on this morphological requirement). Thus, such adjectives
in object position must be accusative-case-marked, as in (81), which contrasts
with (42), shows.
(81) a. Pelin eski kitab- sat-t, Suzan-sa yeni-yi sat-t.

P.-nom old book-acc sell-past S.-nom-however new-acc sell-past
Pelin sold the old book while Susan sold the new one.
b. Pelin eski araba-lar- sat-t, Suzan-sa yeni-ler-i

P.-nom old car-pl-acc sell-past S.-nom-however new-pl-acc
sat-t.
sell-past
Pelin sold the old cars while Susan sold the new ones.
28 See Kornfilt (1997), Gksel and Kerslake (2005), who note that many adjectives in Turkish
the turkish np 135
Our claim is then that the reduced NPs above are simply nominalized adjec-
tives; they dont involve ellipsis. That this is indeed the case is confirmed by
the fact that such reduced NPs dont require a linguistic antecedent. Thus, (82)
can be used in the following context: John and Mary are planning to buy a table.
They go into a shop, where there are a number of tables of different shapes and
colours, one old, others new. Pointing to one of them John says: (82) ((83) can be
used in a similar context involving car-buying.)29
can be used as nouns. Turkish adjectives can also quite generally bear typical nominal
morphology.
29 Consider also adjectives stranded in subject position. Note that nominative case has no
morphological exponence in Turkish. Some adjectives in subject position require plural
marking, some are only slightly degraded without it, and some are fully acceptable without it.
(Plural should be taken as present or absent in both conjuncts in each coordination. Note that
the partitivity marker -(s)I improves some degraded cases (ii), which confirms the relevance
of overt nominal morphology.) A linguistic antecedent is not necessary.
(i) Yal bakc-(lar) ocuk-lar- azarla-d, ama gen-*(ler) onlar-

old caretaker-nom-pl kid-pl-acc scold-past but young-nom-pl they-acc
teselli et-ti.
console-past
The old caretaker(s) scolded the students, but the young one(s) tried to console
them.
(ii) Yeni araba-(lar) ok ie yara-d ama eski-?*(ler/si) epeyce sorun

new car-nom-pl very useful-past but old-nom-pl-partitive quite trouble
kar-d.
cause-past
The new car(s) was/were very useful but the old one(s) was/were very troublesome.
(iii) Hzl araba-(lar) biz-i etkile-di, ama yava-?(lar)

fast car-pl-nom we-acc impress-past but slow-pl-nom
disappoint-past
The fast car(s) impressed us, but the slow one(s) disappointed us.
(iv) Gl insan-(lar) biz-i etkile-r, ama zayf-(lar)

strong person-pl-nom we-acc impress-aor but weak-pl-nom
hayal krklna urat-r.
disappoint-aor
Strong people impress us, weak ones disappoint us.
136 bokovi and ener
(82) Eski-yi/Yuvarla-/Yeil-i isti-yor-um.

old-acc/round-acc/green-acc want-pres-1sg
I want the old/round/green one.
(83) Hzl-y isti-yor-um.

fast-acc want-pres-1sg
Although Turkish is much more productive than, e.g., English, in this respect,
some adjectives in Turkish resist nominal usage even under the morphological
conditions noted above.
(84) *Senato gerek soykrm iddialar-n tart-t, ama

Senate-nom true genocide claims-acc discuss-past but
szde-ler-i tart-ma-d.
alleged-acc discuss-neg-past
Lit. The senate discussed the true claims of genocide, but it didnt dis-
cuss the alleged ones.
(85) *Pelin sradan tablo-lar- sev-me-z ama

P-nom ordinary painting-pl-acc like-neg-aor but
muhteem-ler-i sev-er.
magnificent-pl-acc like-aor
Lit. Pelin doesnt like ordinary paintings, but she likes magnificent
ones.
(v) Eski-*(ler) herey-i bil-dik-leri-ni dn-r-ler.

old-nom-pl everything-acc know-noml-3pl.poss-acc think-aor-3pl
The elderly think they know everything.
(vi) Zengin (-ler) Bush-u sev-er.

rich (-pl) Bush-acc like-aor
The rich like Bush.
(viii) Mete is lying on the ground, after being hit by a car. There is a green and a red car parked
in the middle of the street. Pointing to the green car I tell the policeman:
Yeil vur-du o-na
green hit-past he-dat
The green one hit him.
the turkish np 137
Given such cases and examples like (42), which are quite generally unavail-
able, we assume that the cases of stand-alone adjectives noted above involve
adjectives used as nouns, as confirmed by the associated nominal morphology
and the fact that such adjectives can occur in no-ellipsis contexts.
To summarize, we have argued that NP/N-ellipsis in Turkish is only licensed
by CLL0 and (possibly) Pred0; there is no ellipsis licensed by D since there
is no D. This provides a straightforward account for why simple possessor-
stranding ellipsis is disallowed in Turkish, unlike in English, and a strong argu-
ment against a potential DP analysis of Turkish. We have also argued that
several cases that appear to involve ellipsis (in particular, only-demonstrative
TNPs, ki-NPs, nominalized adjectives, and possibly predicate possessives) do
not actually involve ellipsis.
4 Conclusion
We have argued for a no-DP analysis of Turkish TNP. We have provided an

account of word order and interpretation of elements within the Turkish TNP
where Poss and Dem are NP-adjoined and Num and Adj are NP Specs. Overt
possessors move to this position, while pro possessors stay in a lower position,
which provides evidence for Takahashis (2000, 2001) claim that pro does not
undergo movement. We have shown that Turkish disallows Poss/Dem/Adj/
Num stranding under ellipsis, which follows under the NP analysis, given that
these elements are part of NP, hence cannot be stranded under NP ellipsis. We
have argued that a functional projection is present above NP in classifier and
possibly predicate constructions. Such cases allow TNP-internal ellipsis, with
elements located within the functional projection(s) in question (and outside
of NP) surviving ellipsis.
References
Allen, Cynthia. 2008. Genitives in early English: Typology and evidence. Oxford: Oxford
University Press.
Aydemir, Yasemin. 2004. Are Turkish preverbal bare nouns syntactic arguments?
Linguistic Inquiry 35:465474.
Aygen, Glat. 2002. Finiteness, case, and clausal architecture. PhD diss., Harvard
University.
Bokovi, eljko. 2005. On the locality of left branch extraction and the structure of
NP. Studia Linguistica 59:145.
138 bokovi and ener
Bokovi, eljko. 2007. On the clausal and NP structure of Serbo-Croatian. In For-

mal Approaches to Slavic Linguistics: The Toronto Meeting, 2006, edited by Richard
Compton, Magdalena Goledzinowska, and Ulyana Savchenko, 4275. Ann Arbor:
Michigan Slavic Publications.
Bokovi, eljko. 2008a. What will you have, DP or NP? In Proceedings of NELS 37,
edited by Emily Elfner and Martin Walkow, 101114. Amherst: GSLA Publications.
Bokovi, eljko. 2008b. The NP/DP analysis and Slovenian. In Proceedings of the Novi
Sad Generative Syntax Workshop 1, 5373. Novi Sad: Filozofski fakultet u Novom
Sadu.
Bokovi, eljko. 2009. More on the No-DP analysis of article-less languages. Studia
Linguistica 63:187203.
Bokovi, eljko. 2012a. On NPs and clauses. In Discourse and grammar: From sentence
types to lexical categories, edited by Gnther Grewendorf and Thomas Ede Zimmer-
mann, 179245. Berlin: de Gruyter.
Bokovi, eljko. 2012b. On the edge. Ms., University of Connecticut.
Bokovi, eljko. 2013. Adjectival escapades. In Formal Approaches to Slavic Linguis-
tics: The Indiana Meeting, 2012, edited by Steven Franks, Markus Dickinson, George
Fowler, Melissa Witcombe, and Ksenia Zanon, 125. Ann Arbor: Michigan Slavic
Publications.
Bokovi, eljko. in press a. Now Im a phase, now Im not a phase: On the variability
of phases with extraction and ellipsis. Linguistic Inquiry.
Bokovi, eljko. in press b. Phases beyond clauses. In Nominal constructions in Slavic
and beyond, edited by Lilia Schrcks, Anastasia Giannakidou, Urtzi Etxeberria and
Peter Kosta. Berlin: de Gruyter.
Bowers, John. 1993. The syntax of predication. Linguistic Inquiry 24:591656.
Cheng, Hsu-Te. 2013. Argument ellipsis, classifier phrases, and the DP parameter. PhD
diss., University of Connecticut.
Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance
DP. In Paths towards Universal Grammar: Studies in honor of Richard S. Kayne, edited
by Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, Raffaella Zanuttini,
85110. Georgetown: Georgetown University Press.
Despi, Miloje. 2011. Syntax in the absence of determiner phrase. PhD diss., University
of Connecticut.
Despi, Miloje. 2013. Binding and the structure of NP in Serbo-Croatian. Linguistic
Inquiry 44:239270.
Elbourne, Paul. 2005. The semantics of ellipsis. Interdisciplinary Studies on Informa-
tion Structure 3:63109.
Gajewski, Jon. 2011. A Little More on MOST. Talk given at CUNY.
Gksel, Asl and Celia Kerslake. 2005. Turkish: A comprehensive grammar. London:
Routledge.
the turkish np 139
Hankamer, Jorge. 2004. An ad-phrasal affix in Turkish. MIT Working Papers in Linguis-
tics 46:289299.
Hankamer, Jorge, and Ivan Sag. 1976. Deep and surface anaphora. Linguistic Inquiry
7:391428.
Ketrez, Nihan. 2004. -lAr-marked nominals and three types of plurality in Turkish. In
Proceedings of CLS 39, edited by Jon Cihlar, Amy Franklin, David Kaiser, and Irene
Kimbara, 176192. Chicago: University of Chicago.
Ketrez, Nihan. 2005. Childrens scope of indefinite objects. PhD diss., University of
Southern California.
Kornfilt, Jaklin. 1984. Case marking, Agreement and Empty Categories in Turkish. PhD
diss., Harvard University.
Kornfilt, Jaklin. 1997. Turkish. London: Routledge.
Kornfilt, Jaklin. 2003. Scrambling, subscrambling, and Case in Turkish. In Word Order
and Scrambling, edited by Simin Karimi, 125156. Malden: Blackwell.
Kornfilt, Jaklin. 2005. Asymmetries between pre-verbal and post-verbal scrambling
in Turkish. In The free word order phenomenon: its syntactic sources and diversity,
edited by Joachim Sabel and Mamoru Saito, 163181. Berlin: de Gruyter.
Koster, Jan. 1994. Predicate incorporation and the word order of Dutch. In Paths
towards Universal Grammar: Studies in honor of Richard S. Kayne, edited by Gugliel-
mo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, Raffaella Zanuttini, 255277.
Washington: Georgetown University Press.
Kural, Murat. 1997. Postverbal constituents in Turkish and the Linear Correspondence
Axiom. Linguistic Inquiry 28: 498519.
Larson, Richard, and Sungeun Cho. 1999. Temporal adjectives and the structure of
possessive DPs. Proceedings of WCCFL 18, edited by Sonya Bird, Andrew Carnie,
Jason D. Haugen, and Peter Norquest, 299311. Somerville, MA: Cascadilla Press.
Lewis, Geoffrey. 1967. Turkish Grammar. Oxford: Oxford University Press.
Ochi, Masao. 2005. Ga-no conversion and overt object shift in Japanese. Nanzan
Linguistics 2:6181.
ztrk, Balkz. 2004. Case, Referentiality and Phrase Structure. PhD diss., Harvard
University.
Partee, Barbara. 2006. A note on Mandarin possessives, demonstratives, and definite-
ness. In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and
Semantics in honor of Laurence R. Horn, edited by Betty J. Birner and Gregory L. Ward,
263280. Amsterdam: Benjamins.
Partee, Barbara and Vladimir Borschev. 2001. Some puzzles of predicate possessives. In
Perspectives on Semantics, Pragmatics and Discourse: A Festschrift for Ferenc Kiefer,
edited by Istvn Kenesei and Robert M. Harnish, 91117. Amsterdam: Benjamins.
Saito, Mamoru, and Keiko Murasugi. 1990. N-deletion in Japanese. University of Con-
necticut Working Papers in Linguistics 3:87107.
140 bokovi and ener
Sezer, Engin. 1991. Issues in Turkish syntax. PhD diss., Harvard University.
ener, Serkan. 2008. Non-canonical Case licensing is canonical: Accusative subjects of
CPs in Turkish. Ms., University of Connecticut.
ener, Serkan. 2010. (Non-)Peripheral Matters in Turkish Syntax. PhD diss., University
of Connecticut.
ener, Serkan and Daiko Takahashi. 2010. Argument Ellipsis in Japanese and Turkish.
MIT Working Papers in Linguistics 61:325339.
Takahashi, Daiko. 1996. Move-F and Pro. MIT Working Papers in Linguistics 29:255265.
Takahashi, Daiko. 2000. Move F and raising of lexical and empty DPs. In Step by Step:
Essays on Minimalist Syntax in Honor of Howard Lasnik, edited by Roger Martin,
David Michaels, Juan Uriagereka, 297317. Cambridge: MIT Press.
Takahashi, Daiko. 2001. Scrambling and empty categories. MIT Working Papers in
Linguistics 41:4758.
Takahashi, Masahiko. 2008. Pronominal NP ellipsis and its theoretical implications.
Ms., University of Connecticut.
Takahashi, Masahiko. 2011. Some consequences of Case-marking in Japanese. PhD
diss., University of Connecticut.
Tremblay, Mireille. 1989. French possessive adjectives as dative clitics. Proceedings
of WCCFL 8, edited by Jane Fee and Kathryn Hunt, 399413. Stanford, CA: CSLI
Publications.
ivanovi, Sao. 2008. Varieties of most: On different readings of superlative determin-
ers. In Studies in Formal Slavic Linguistics: Contributions from Formal Description
of Slavic Languages (FDSL) 6.5, edited by. Franc Marui and Rok Zaucer, 337354.
Berlin: Frankfurt am Main.
Zribi-Hertz, Anne. 1997. On the dual nature of the possessive marker in Modern
English. Journal of Linguistics 33:511537.
part 2
Definiteness and Definiteness

Markers across Languages

The Morphology, Syntax and Semantics of
Definite Determiners in Swiss German*
Rebekka Studler
In Swiss German there are three paradigms for the definite determiner: a weak
article, a strong article, and a proximal demonstrative. The aim of this paper is
to show that these three paradigms not only differ with respect to their mor-
phological form, but also with respect to their semantic function and their
syntactic structure. Based on several Swiss German data corpora there is evi-
dence for a strong correlation between morphology, semantics, and syntax. I
will demonstrate that every paradigm has its prototypical semantic function.
The weak article is used in inherently unique contexts, the strong article shows
up in anaphoric contexts and the demonstrative is used in deictic contexts.
However, some cases seem to challenge the correlation established in the anal-
ysis, particularly modification structures with relative clauses. However, I shall
show that the use of articles in these cases follows its own strictly semantic
rules. To meet the semantics syntax correlation, I put forward the idea that
every paradigm hasdue to its particular feature structureits own syntac-
tic projection. Therefore, a semantic-syntactic analysis is proposed in which
the semantic-syntactic features of the three paradigms are accommodated by
assuming three functional categories in the nominal phrase for the features in
question, i.e. [DEF] for definiteness, [ANAPH] for anaphoricity, and [DEIKT]
for deixis.
1 Introduction
Like Standard German, Swiss German dialects have a definite determiner

with distinct forms marked for gender, number, and case (for details see
* This research was supported by the Swiss National Science Foundation (SNF). I thank Elvira
Glaser and Helen Christen for the data, as well as my informants for their patience. Versions
of this paper have been presented on many occasions; in particular I thank the audience of
the Syntax Semantics colloquium at the University of Konstanz and of the annual meeting
of the DGfS 2007. I am indebted to Josef Bayer, Ellen Brandner, Jana Hussler, Uli Lutz,
Albert Ortmann, Martin Prinzhorn, Manuela Schnenberger, Susanne Trissler, and ystein
Vangsnes for helpful discussions and valuable feedback. Special thanks go to the reviewers of
this volume for their comments and suggestions to improve this paper.

144 studler
section 3).1 However, Standard German has only one paradigm for the defi-
nite determiner, while some Swiss German dialects distinguish between two
paradigms. These paradigms correspond to a morphologically reduced article
(1a) and a morphologically full article (1b). Additionally, there exists a proximal
demonstrative, which originates from the same word stem (1c).2
(1) a. de Maa d Frou s Chend

detred man detred woman detred child
b. d Maa di Frou das Chend

detred man detred woman detred child
c. d Maa die Frou daas Chend

this man this woman this child
What are the semantic differences between these determiner paradigms and
how can these differences be explained by syntax? The aim of this paper is to
delineate the similarities and differences between the three paradigms and to
explore how the similarities and differences emerge in morphology, semantics,
and syntax. For this purpose, I will begin with the premise that the three
paradigms not only differ in their morphological form but also in their semantic
function and in their syntactic structure. I shall argue for a strong correlation
between morphology, semantics, and syntax. For each paradigm, it should be
1 Abbreviations in the text: AnaphP = anaphoricity phrase, APP.RC = appositive relative clause,
CP = complementizer phrase, D = determiner, DefP = definiteness phrase, DemP = demonstra-
tive phrase, DxP = deixis phrase, DP = determiner phrase, DRC = descriptive relative clause,
FinP = finiteness phrase, fn = functional noun, FP = functional phrase, IN = individual noun,
NP = noun phrase, nP = little nP, rel.pron = relative pronoun, RESTR.RC = restrictive relative
clause, RN = relational noun, SN = sortal noun, TopP = topic phrase.
Features: [ANAPH] = feature for anaphoricity, [DEF] = feature for definiteness, [DEIKT] =
feature for deixis, [DEM] = feature for demonstrative, [DET] = feature for determination.
Abbreviations in the glosses: ACC = accusative, ADV = adverb, DAT = dative, f. = feminine,
FULL = full (article), GEN = genitive, m. = masculine, n. = neuter, pl. = plural, ps = person, RED
= reduced (article), sg. = singular.
Languages: SADS = Syntactic Atlas of the German-speaking part of Switzerland, AG =
canton of Aargau, BA = canton of Basel, BE = canton of Berne, SO = canton of Solothurn.
2 Cf. Fischer (1989). The form di for the full feminine article is pronounced as [di], the form die
for the feminine demonstrative is pronounced as [die]. The reduced article is glossed with
detred, the full article with detfull. See section 3 for the full paradigms.
definite determiners in swiss german 145
possible to formulate specific conditions concerning its form, function, and

structure. To what extent such a strong correlation holds will be examined in
the following text.3
The semantic function of determiners is, in simplified terms, to determine
the referent, i.e. determiners specify the denotatum of the noun (for details
see the references in Kolde 1996: 2750). There are two ways of referring to
something or somebody: indefinite and definite reference. While the indefinite
article is prototypically used to introduce an object into the discourse (2a), the
definite article is prototypically used to mark an object as established in the
discourse (2b):
(2) a. Nora has got herself a dog.

b. The dog is sweet, but not yet house-trained.
Given this categorization (which is not unproblematic, as pointed out for

example by von Heusinger 2002), the act of definite determination is not yet
completely explained. How is it possible to successfully refer to an object in
the world with a definite expression? This question is subject to controversial
discussions and there are different concepts of reference to deal withstarting
with Russells uniqueness claim to the concepts of inclusiveness, identifiability,
familiarity, and salience, to name the most influential ones (for a detailed
discussion see Lyons 1999). Regardless of which concept one prefers, there
are several options to fulfill the requirement of a unique reference: either the
nominal expression refers inherently to a unique object or there is need for
3 The results of my study of the definite determiner in Swiss German are founded on the
analysis of several data corpora (for details see Studler 2011). There are traditional grammar
books and monographs for most capital Swiss German dialects and various short grammars
and small monographs on small-scale dialect regions. Even though these grammar books
offer systematic and detailed overviews, they have several shortcomings for the present
purpose. First, they are descriptive. For most phenomena, they do not provide any additional
explanation. Secondly, they are normative in the sense that they are designed to be a guide
for good and correct Swiss German. Thirdly, they are based on slightly outdated evidence
and do not include recent language changes. Finally, they do not give enough attention
to the important connection between definite determiners and modifications within the
noun phrase. For these reasons, I analyzed additional data corpora, such as the data from
the SADS (Syntactic Atlas of the German-speaking part of Switzerland, with a sample of
approximately 3,200 persons in almost 400 places in Switzerland), including subsequent
detailed questionings I implemented with a small sample of six to ten persons. In addition
to this elicited data, I analyzed the data from Christen (1998), a transcript of forty-two
spontaneous narrative interviews, which contain about five thousand article occurrences.
146 studler
contextual information to support the identification of the referent. In the first

case, our common knowledge about the world is constituted in a way that
we identify the nominal expression as inherently unique, e.g. unica, like the
sun, the father. In the second case, we need additional intratextual informa-
tion, for example anaphorically accessible lexical material, or extralinguistic
information like pointing gestures and deictic words, which indicate the refer-
ent. Lbner (1985 and recently 2011) describes this distinction by the notion
of semantic definiteness resp. uniqueness and pragmatic definiteness resp.
uniqueness.4
In various German dialects, as well as other languages, the difference be-
tween semantic definite reference and pragmatic definite reference is reflected
in the use of two different article paradigms.5 While a strong (i.e. morpholog-
ically full) article is used with semantic definites, a weak (i.e. morphologically
reduced) article is used with pragmatic definites. The same holds true for the
Swiss German article system (for details see section 4).
To meet the correlation between semantic function and determiner para-
digms pragmatic uniqueness has to be subdivided in anaphoric and in deictic
uniqueness. Thus, I shall use a detailed categorization with inherent unique-
ness, textual information, and deictic information. While the reduced article
is normally used in cases of inherent uniqueness, the full article appears in
cases of textual information and the demonstrative in cases of deictic infor-
mation.
Having settled the definite determiners morphology and semantics, one
might ask how the correlation between the semantic concept of definite-
ness and the syntactic feature of definiteness can be explained. According
to Lyons (1999), definiteness is a universal semantic-pragmatic entity that is
syntactically represented by the feature [DEF] (see also Felix 1988 and Alexi-
adou et al. 2007). The feature [DEF] is therefore the syntactic correlate to the
semantic concept of definiteness, which can be expressed in different ways; as
4 Besides the fact that there are typical cases for both types, e.g. so called monosemantica for
semantic uniqueness (the sun, the pope) and anaphoric/deictic DPs for pragmatic uniqueness,
it is assumed that the distinction between semantic and pragmatic uniqueness is not clear-
cut but rather that semantic and pragmatic uniqueness need to be thought of as the endpoints
of a scale (for details see Lbner 2011).
5 For the German dialects see e.g. Heinrichs (1954), Hartmann (1982), Scheutz (1988), Eroms
(1989), Brugger & Prinzhorn (1995), and recently Schwarz (2009), for Fering, the dialect of
North Frisian, Ebert (1971), for the Upper Sorbian colloquial language Breu (2004), Scholze
(2007), and for the Scandinavian languages Delsing (1993), Vangsnes (1999, 2001), Julien
(2005), Hankamer & Mikkelsen (2002), and Heck et al. (2008).
in Swiss German, a prototypical realization is a definite determiner. In order

to explain the correlation between the semantic functions and the syntactic
features of our three determiner paradigms, it is worth disentangling their
feature structure more thoroughly. Taking into account the feature structure
of the different determiners in the grammaticalization process proposed by
Lehmann (2002) and Himmelmann (1997), it is possible to distinguish the
three paradigms within the framework of generative grammar. I argue for
three different features[DEF] for definiteness, [ANAPH] for anaphoricity,
and [DEIKT] for deixiswhich are reflected in different syntactic structures
resp. in different syntactic projections.
This paper is organized as follows: In section 2, the semantic function of
definite determination will be discussed. Apart from the chosen concept of def-
inite reference as uniqueness, familiarity, or salience, there are three strategies
to retrieve the reference of a definite DP: by inherent uniqueness, by textual
information, or by deictic information. I will characterize these three options in
detail while providing the relevant examples for every subcategory. In section
3, the morphological forms of the three paradigms will be presented, and I will
discuss very briefly some morphological resp. phonological variants. In section
4, I will demonstrate the correlation between morphology and semantics for
the three paradigms of definite determiners in Swiss German. Also, I will dis-
cuss cases that rebut a strong correlation between morphology and semantics.
On one hand, there are cases showing a language change in the use of definite
articles in Swiss German; on the other hand, there are cases that, prima facie,
seem to violate the correlation thesis. In section 5, I will propose a syntactic
structure for the Swiss German noun phrase within the framework of gener-
ative grammar, which should explain the distribution of the three paradigms.
The morphological and semantic distribution is shown to have a syntactic cor-
relation insofar as different syntactic positions for the paradigms in question
are proposed.
2 Semantics: The Function of Definite Determiners
Russell (1905) famously advocated the idea that definite descriptions exhibit
a complex logical form: the present King of France must be analyzed as a
conjunction of the claim that there is at least a present King of France and
of the claim that there is only one present King of France (x(Fx & y(Fy
y=x))). Therefore, for a definite description to refer, two conditions must be
fulfilledan existential condition and a uniqueness condition. Since Russells
theory for definite descriptions is criticized, especially for the shortcoming that
148 studler
only singular count nouns can be accounted for, a variety of alternative con-
cepts have been proposed to explain the notion of definitenessas inclusive-
ness (Hawkins 1978), familiarity (Christophersen 1939), identifiability (Strawson
1950; Searle 1969), and salience (Lewis 1979 and von Heusinger 1996, 1997). For a
successful act of reference, there needs to be a discourse-relevant referent and
the referent must be uniquein the sense of non-ambiguous or, depending on
the chosen concept, familiar, identifiable, or salient. For our purpose there is no
need to take a stand on this debate (for details see Lyons 1999) instead we will
concentrate on the classification of definite types.
There are two ways to successfully fix the uniqueness of the reference: by
nominal expressions that refer inherently to a unique referent or by context
dependent expressions that refer to a unique referent. While the former is
inherently unique, the latter needs more information to single out the ref-
erent. As Lbner (1985, 2011) points out, the difference can be explained by
the concepts of semantic uniqueness and pragmatic uniqueness (see for similar
approaches Christophersen 1939, Ebert 1971, Hawkins 1978, Himmelmann 1997,
among others). While semantic uniqueness is fulfilled by individual nouns (IN;
like the sun) and functional nouns (FN; like the father), pragmatic uniqueness
takes effect with sortal nouns (SN; like the man) and relational nouns (RN; like
the sister). For SNs and RNs to fulfill the uniqueness condition, a shift to IN
or FN must take place by means of the given context (for details see Lbner
2011 and Ortmann in press). There are two options to enrich SNs and RNs with
contextual information: either by anaphoric (textual) information or by deictic
information.
To summarize, there are the following possibilities of unique reference:
(3) I) semantic uniqueness inherent-unique reference IN or FN

II) pragmatic uniqueness context-unique reference SN or RN
a. textual information anaphoric-unique reference SN/RN shift to IN/FN
b. deictic information deictic-unique reference SN/RN shift to IN/FN
I will sketch these options while providing the relevant examples in the follow-
ing paragraphs. Since for our purpose it is meaningful to disentangle the two
types of pragmatic uniqueness (II a. and II b.), I will concentrate on the three-
fold distinction inherent-unique, anaphoric-unique, and deictic-unique.
Inherent-unique reference: By using common knowledge for the process of sin-

gling out the referent, the definite description refers to something common or
familiar. The knowledge of the hearer about the world in general or a specific
situation makes an effortless interpretation of the definite description avail-
able. The semantic uniqueness of the reference can be reached in different

ways. Based on Christophersen (1939), Ebert (1971), Hawkins (1978), Lbner
(1985), and Himmelmann (1997), I distinguish four different kinds of unique
reference: absolute-unique (4), situative-unique (5), associative-anaphoric (6),
and generic (7).
(4) a. Sara went to the cinema yesterday.

b. The moon has vanished behind a cloud.
(5) The groom was wearing a red tie.
(6) a. When I was admitted to the hospital, the nurse gave me an injection.
b. I baked a cake yesterday. The sugar icing worked perfectly.
(7) a. The dragonfly is endangered.

b. We estimate the team spirit as very important.
While in English and Standard German proper names are used without an
article (see ex. 4a), in Swiss German (and other German dialects) proper names
are used with a definite article (see section 4). In the situative-unique use there
is theoretically more than one available referent, but due to the situation it is
easy to single out the intended referent. Because of our common knowledge,
we know that (normally) there is only one groom at a wedding (see ex. 5). In
the associative-anaphoric use, the unique reference of a nominal expression
is guaranteed by knowledge of frames (hospital) or scripts (baking a cake).
Even so there is an indirect anaphoric link or a so-called bridging in the text
(hospital, cake) the reference fixing is made by the common knowledge of the
hearer and not by the textual information (see ex. 6). The fourth subfunction
is the generic use (see ex. 7) in which the referent is not a unique object
but rather a class (the dragonfly) or a concept (the team spirit). Here, it is
referred to as a totality, which corresponds to an inclusive set, i.e. the totality
of the elements of a class (see also the natural kind terms for classes in Kripke
1980).
Anaphoric-unique reference: With textual information the unique reference

is not reached by the nominal expression alone but also by additional lex-
ical material. Neither our common knowledge nor the nature of the nomi-
nal expression can fix the reference. Only by intratextual information does
the identification of the intended referent work. There are two instances of
150 studler
references involved with the intratexual information: anaphoric textual refer-

ence (8a), where the referential anchor is found in an antecedent, and auto-
phoric textual reference (8b), where the referential anchor is found in the nom-
inal expression itself (cf. Ortmann in press: 7).
(8) a. There is a church at the end of our street The church burned down
ten years ago.
b. The book I bought yesterday is worthless junk.
Deictic-unique reference: If it is impossible to identify the referent by the nomi-

nal expression alone, we may give an informational hint by a deictic expression,
such as a demonstrative together with an ostensive act.
(9) This piece of cake belongs to John. (plus pointing gesture)
For clarification, there are often locative adverbs in use (10a), especially in the
case of distinguishing two referents in a contrastive context (10b).
(10) a. This cake here (on the table) belongs to John.

b. This cake here belongs to John. The one there (in the storage) is for
you.
Figure 1 summarizes the different functions of definite determiners.
figure 1 The semantic functions of definite determiners

Before I turn to the correlation between the semantic functions and the
three determiner paradigms in Swiss German, I will present the morphological
forms of the three paradigms and very briefly discuss some remarkable variants.
3 Morphology: The Paradigms of the Definite Determiner in Swiss

German Dialects
In Swiss German dialects there are two different paradigms for the definite
determiner the: a morphologically reduced form de, d, s, detred m./f./n.
and a morphologically full form d, di, das detfull m./f./n.. The proximal
demonstrative originates from the same word stem as the stressed form d,
die, daas (this m./f./n.).
3.1 Three Morphological Forms

The three determiner paradigms in Swiss German arose from the morphologi-
cal forms of the original demonstrative:
(11) a. a stressed form (d, die, daas)

b. a weakened, but morphological full form (d, di, das)
c. a morphological reduced form (de, d, s)
In Swiss German grammars, these three paradigms are described for Bernese
German (Marti 1985; Hodler 1969), Basel German (Suter 1992), Zurich Ger-
man (Weber 1987), Lucerne German (Fischer 1989), parts of the canton Valais
(Bohnenberger 1913; Wipf 1910), and parts of the canton Grisons (Meinherz
1920; Hotzenkcherle 1934). For Bernese German, the three forms are also dis-
cussed in the studies of Nbling (1992) and Penner (1993). The data corpora I
used for my study show that the two article paradigms exist in other dialects,
too. In other regions of Switzerland, including Central and Eastern Switzerland,
there are two articles in useeven if the grammars for these two regions do
not give evidence for two article paradigms but distinguish only between one
article and a proximal demonstrative.
In accordance with other literature on two-article languages, I call the
stressed form (11a) demonstrative and the unstressed forms articles; the mor-
phological full form (11b) is called full article and the morphological reduced
form (11c) is called reduced article. These labels are meant to reflect the various
semantic functions of articles and demonstratives (see section 2), despite the
fact that most Swiss German grammars categorize the full article as demonstra-
tive.
152 studler
In the next subsections, I will present the morphological forms of the three
paradigms. The presented forms are from Lucerne German, as according to
Fischer (1989). I choose this dialect, because it most clearly distinguishes the
differences between the paradigms.
3.2 The Reduced Article

The forms of the reduced article are as follows (see Fischer 1989: 227).6
table 1 Reduced articles in Lucerne German
Masculine Feminine Neuter
nom./acc. sg. de d(i) s

dat. sg. em de em
nom./acc. pl. d(i) d(i) d(i)
dat. pl. de de de
The forms of the reduced article across Swiss German dialects are the result of a
wide range of morphological reduction processes. Dependening on the dialect,
there are the following variants (see Marti 1985; Suter 1992; Weber 1987; Fischer
1989; Bossard 1962): either the initial sound (12a) or the final sound (12b) or the
vowel (12c) is reduced.
(12) a. initial sound, e.g.: nom./acc. sg. n.: s, dat. sg. m./n.: em, dat. sg. f.: er
b. final sound, e.g.: nom./acc. sg. m.: de, dat. sg. f.: de
c. vowel, e.g.: nom./acc. sg. m.: dr, nom./acc. sg. n.: ds, dat. sg. f.: dr
In all forms with the e-vowel, an unstressed schwa-like vowel is involved, which
is realized as a vowel oscillating between front and central and between open-
mid and open depending on the dialect ([], [], [], [a]).
6 In Swiss German the accusative forms are identical to the nominative forms. This formal
identity holds true not only for the article but also for all declinable word classes, except for
personal pronouns. Furthermore, the genitive is replaced in almost every context, e.g. by a
dative construction:
(i) Em Vater si Huet vs. Vaters Huet

the father.dat his hut fathers.gen hut
3.3 The Full Article

The full article has the following forms (see Fischer 1989: 227).
table 2 Full articles in Lucerne German
nom./acc. sg. d di(e) das

dat. sg. dm dr dm
nom./acc. pl. di(e) di(e) di(e)
dat. pl. dne dne dne
The forms of the full article are characterized by morphological completeness

and unstressed pronunciation. The identification of the full article is guided by
the variants of the reduced article. If the reduced article has lost the initial d-,
the full article can be identified by the initial d- (13a); if the final sound of the
reduced article is lost, the full article is marked by the final sound (13b); if the
vowel of the reduced article is faded, the full article has a strong vowel (13c).
(13) a. initial sound, e.g.: dat. sg. m./n.: full dem instead of reduced em
b. final sound, e.g.: dat. sg. f.: full dr instead of reduced de
c. vowel, e.g.: nom./acc. sg. m.: full d instead of reduced de
3.4 The Demonstrative

The forms of the demonstrative are as follows (see Fischer 1989: 227).7
table 3 Demonstrative in Lucerne German
nom./acc. sg. d die daas

dat. sg. dmm dr(e) dmm
nom./acc. pl. die die die
dat. pl. dnne dnne dnne
7 The -e in the form die is not a mute lengthener of the preceding vowel as in Standard German,
but a fully pronounced vowel.
154 studler
The demonstrative is always stressed and lengthened. The lengthening is

realized either by lengthening the vowel or by doubling the consonant.8 The
option used depends on the form in question: nominative and accusative forms
are subject to vowel lengthening (14a) and dative forms often show a consonant
doubling (14b):
(14) a. vowel lengthening: nom./acc.sg.m./f./n.: d, die, daas

nom./acc.pl.m./f./n.: die
b. consonant doubling: dat.sg.m./n.: dmm
dat.pl.m./f./n.: dnne
The next section is dedicated to showing a strong correlation between the

morphology and the semantics of the definite determiners in Swiss German.
4 Correlation of Form and Function: The Semantic Distribution of

the Three Paradigms
In section 2, I have shown the three options of reference with a nominal

expression. Is there a correlation between these three options and the three
determiner paradigms reduced article, full article, and proximal demonstrative?
The data corpora I analyzed confirm the assumption widely found in the
literature of a one-to-one correlation on weak and strong definite articles. In
the following subsection, I will briefly demonstrate this correlation of form and
functionspecial cases are discussed in section 4.2.
4.1 General Distribution

4.1.1 The Reduced Article in Inherent-Unique Use
The reduced article is typically used in cases where the nominal expression
refers inherent-uniquely. This is the case with absolute unica (15a)including
proper names (15b), superlatives (15c), and inalienables (15d), situative unica
(15e), anaphoric-associative unica (15f), and generic expressions in the form of
classes (15g) or abstracta (15h).9
8 The lengthening by doubling the consonant is rarely realized in writing. While Fischer (1989)
correctly transcribed the double consonant, Weber (1987) called attention to the doubling
but did not transcribe the second consonant for the sake of the typeface.
9 The examples in this section are taken from the standard Swiss Central Plateau dialect but
are confirmed for the other Swiss dialects, too.
(15) a. De Mond schiint ht aber weder schn.

detred moon shines today but again nice
The moon is shining very nice today.
b. E ha d Lisa geschter gfrogt.

I have detred L. yesterday asked
I have asked Lisa yesterday.
c. De T het s grscht Schtck Chueche gnoo.

the Tim has detred biggest piece cake taken
Tim has taken the biggest piece of cake.
d. De Kbi het d Nase broche.

the Jack has detred nose broken
Jack has broken his nose.
e. De Schlagzger gseet besser uus as de Snger.

detred drummer looks better out than detred singer
The drummer looks better than the singer.
f. I sere Schtrooss stoot e Chile. I de Sakrischtei hets

in our street stands a church in detred sacristy has=it
letscht Joor brnnt.
last year burnt
There is a church in our street. In the sacristy, it has burnt last year.
g. De Li esch de Kchnig under de Tiere.

detred lion is the king among detred animals
The lion is the king of animals.
h. De Tiimgeischt werd be s grooss gschrebe.

detred team-spirit is by us big written
Team spirit is important to us.
Notice that the characteristic use of the reduced article matches the above-
described inherent-unique use of definite determiners. The reduced article is
typically inherent-unique.
The inherent-unique reference of the reduced article is unproblematic and
confirmed by the examples and the descriptions in the grammar books and
monographs of Swiss German as well as by the studies on the article system in
156 studler
Swiss German (Nbling 1992; Penner 1993). The analyzed data of the interview
transcript, and especially of the SADS data (with at least 96 % occurrences
of the reduced article in these contexts), confirm this result (for details, see
Studler 2011).
4.1.2 The Full Article in Anaphoric-Unique Use

The full article is not allowed in cases in which the nominal expression refers
inherent-uniquely. The two articles are in complementary distribution. While
the reduced article appears in inherent-unique contexts, the full article is used
in cases in which the reference of the noun is not completely determined.
The unique reference is achieved by intratextual information, either anaphoric
(16a) or autophoric (16b).
(16) a. Uf em Tesch liit es Buech. Das Buech wot i lse.

on the table lies a book detfull book want I read
A book is lying on the table. I want to read the book.
b. Das Buech, won i geschter gchauft ha,

detfull book wo.rel.pron=n I yesterday bought have
han i scho glse.
have=n I already read10
detfull book which I bought yesterday I have already read.
The anaphoric-unique use of the full article matches the above-described ana-
phoric-unique use of determiners. The full article is prototypically anaphoric-
definite.
The anaphoric use of the full article is confirmed by the grammar books and
monographs of Swiss German as well as by Nbling (1992) and Penner (1993).
The analysis of the data shows that the full article is the most used determiner
in this context, but only at a rate of 67%. The reduced article occurs surprisingly
often, and the demonstrative is also possible in some cases (see section 4.2).
The autophoric use (e.g. with establishing relative clauses) is only partially
confirmed by the literature. While some of the grammar books with two article
paradigms present the full article as being typical with relative clause construc-
tions (e.g. Bernese German and Basel German), some grammars suggest that
the reduced article may also be used in these contexts (e.g. Zurich German and
10 The -n functions as a linking n between two vowels, i.e. if the following word begins with a
vowel, the relative pronoun wo becomes won, ha (have.1ps.sing) becomes han.
Lucerne German). In the analyzed data corpora, I found that the full article was
predominantly used (for example, in at least 85% of the SADS data and in 90 %
of the interview transcript); see details in section 4.2.
4.1.3 The Demonstrative in Deictic-Unique Use

The demonstrative is used to fix the reference extralinguistically. It is the
paradigmatic realization of the deictic-unique use. In Swiss German, the proxi-
mal demonstrative d, die, daas may be used for the reference with a pointing
gesture.
(17) Die Toorte han i slber gmacht. (plus pointing gesture)

this cake have=n I self made
I made this cake by myself.
The proximal demonstrative can be completed with a locative adverb (18a) and
contrasted with the distal demonstrative (18b).
(18) a. Die Toorte do han i slber gmacht.

this cake here have=n I self made
I made this cake here by myself.
b. Die Toorte han i slber gmacht, desi (em Cheschrangk)

this cake have=n I self made, that in=the fridge
esch vom Beck.
is from=the bakery
I made this cake by myself, that in the fridge is from the bakery.
The prototypical use of the demonstrative in deictic-unique contexts is con-

firmed by all grammar books and monographs as well as in the studies of
Nbling (1992) and Penner (1993). However, the data corpus of the SADS and my
subsequent questioning show that the demonstrative is not as easily verifiable
in elicited data as the prototypical realization in deictic-unique use. In fact, the
full article appears in these contexts as naturally as the demonstrative (similar
to the stressed article dr in Standard German). In translations from Standard
German into Swiss German with a demonstrative in a deictic-unique context,
the full article was applied in 53% of the cases and the demonstrative in 47 %
of the cases. In translations from Swiss German to Standard German with a
demonstrative or an article in a deictic-unique context, however, the proximal
demonstrative dieser, diese, dieses (this) was used in 83 % of the cases and the
full article in 17% of the cases.
158 studler
In summary, there is a correlation between the three determiner paradigms

and the three semantic functions of determiners, even though they are not all
equally strong. The reduced article is used in inherent-unique contexts when
the nominal expression refers inherent-uniquely. The full article is used in
anaphoric-unique contexts when the uniqueness of the nominal expression is
reached by additional information in the text. The demonstrative is used in
deictic-unique contexts when the uniqueness is reached by deictic elements
or a pointing gesture. In the next section, I will discuss some cases in which
these correlations are called into question.
4.2 Variations in the Distribution

In addition to their prototypical use, the three paradigms appear in contexts in
which they are not expected. The reduced article may be used in anaphoric-
unique contexts, the full article appears in deictic-unique contexts, and the
demonstrative appears in a somewhat inherent-unique context.11
4.2.1 The Reduced Article in Anaphoric-Unique Use

The strict correlation between the full article and the anaphoric-unique use is
challenged by the following examples.
(19) a. Es esch emou e Kchnig gsi. De Kchnig het e Tochtr

it is one a king been detred king has a daughter
ghaa. BE
had
Once upon a time there was a king. The king has had a daughter.
b. D Lt, won i vo frener kenne, wone aui

detred people, wo.rel.pron=n I from before know live all
nmme z Oute. SO
no longer in Olten
The people I know from yore live all no longer in Olten.
In (19a), the article anaphorically refers to the indefinite noun phrase in the
former sentence. Even though this is the prototypical occurrence of the full
article, in Swiss German the reduced article is possible in sentences like (19a). In
11 Data in this section are from my questionnaire and from the interview transcript. The respec-
tive dialect is indicated in the examples (AG = dialect of canton of Aargau, BA = Basel German,
BE = Bernese German, SO = dialect of canton of Solothurn), however, the examples are not
exclusively confirmed for these dialects but for other dialects, too.
(19b), the article is linked with the relative clause modification of the noun. This
use belongs to the domain of the full article as well. Nevertheless, the reduced
article may show up with restrictive relative clause constructions.
Since the correlation between the article and the restrictive relative clauses
is controversial, I analyzed my data corpora regarding all relative clause occur-
rences. The result of the analyzed sentences with a restrictive relative clause
from the SADS shows a clear picture: 85% of the sample used the full arti-
cle. Moreover, the full article is preferred in every Swiss German dialect. In
the interview transcript, the reduced article is almost always used with appos-
itive relative clauses (which are not relevant for reference fixing), but it is only
marginally used with restrictive relative clausesthat is, in only 12 % of the
cases. Furthermore, if one takes a closer look at the relative clauses resp. the
nouns that are modified by the relative clause, it seems that relative clauses
are used to specify a concept rather than to identify the referent (in the sense
of Lehmann 1984 and Ebert 197112). The phrase in (20a), for example, could be
paraphrased by residents and the phrase in (20b) could be paraphrased by
nub or core:
(20) a. d Lt, wo jez i dem Gebiet woone BA

detred people wo.rel.pron now in the area live
the people who are living now in this area
b. i de Phngkt, wo s druf aa chunt AG

in detred points wo.rel.pron it on at come
in the points which matter
In regards to the use of the reduced article with restrictive relative clauses, one
can observe that the use of the reduced article is limited to non-establishing
restrictive relative clauses. Establishing relative clauses (Hawkins 1978) oras
Gunkel 2007 names them defining restrictive relative clausesdefines an object
not yet specified (for details see Gunkel 2007). Defining restrictive relative
clauses are attributive in the sense of Donnellan (1966). In every case and for
all dialects, they show up only with the full article:
(21) Jede bechonnt di Frau, won er verdient. AG

Everybody gets detfull woman wo.rel.pron=n he deserves
Everyone gets the woman he deserves.
12 [] Nomen des Matrixsatzes und Relativsatz [bilden] zusammen einen Begriff [] the
160 studler
While the anaphoric use of the reduced article marks a language change,
insofar as the reduced article intrudes on the domain of the full article, the use
of the reduced article with relative clauses only partially marks a takeover. First,
the reduced article is often used with relative clauses that are not merely used
to identify the referent but also to build a new concept with the noun. Secondly,
when the referent of the noun has not been established the reduced article is
not possible with defining restrictive relative clauses (see Hawkins 1978, Gunkel
2007).
4.2.2 The Full Article in Deictic-Unique Use

In addition to its typical function as an anaphoric-unique determiner, the full
article is often used in deictic-unique contexts (with or without a reinforcing
locative adverb):
(22) Wo su i di Buder do heschtue? BE

where should I detfull painitings here put
Where should I put these paintings here?
Thus, the full article encroaches on the domain of the demonstrative and
weakens the correlation between the demonstrative and the deictic-unique
use. This shift towards the full article seems to mark a language change similar
to the reduced articles takeover of anaphoric contexts; however, in both cases
the changes are not arbitrary but rather strictly regular.
The same phenomenon exists in Standard German. Occasionally, the full
article is used in deictic-unique contexts instead of the demonstrative.
(23) Wo soll ich diese/die Bilder da hinstellen?

where should I these/the paintings here put
Where should I put these/the paintings here?
4.2.3 The Demonstrative in Inherent-Unique Use

The demonstrative not only appears in deictic-unique contexts, but also in con-
texts that seem similar to the inherent-unique contexts of the reduced article.
noun of the matrix sentence and the relative clause build together one concept (Ebert 1971:
143, emphasis in the original). A similar approach is targeted in Wiltschko (to appear: 37)
where descriptive relative clauses are distinguished from restrictive relative clauses: It is for
this reason, that DRCs [descriptive RCs]like phrasal compoundsmay serve to create
new concepts.
The purpose of this function is not to deictically fix the reference, because
the reference is already fixed by the specific knowledge that exists between
the speaker and the hearer (similar to the reference fixing with the reduced
article). Different from the inherent-unique use of the reduced article, whereby
the reference counts as unproblematic, the inherent-unique demonstrative is
used to problematize the reference, e.g. the speaker may indicate that he/she
knows the referent but at the moment does not know its name, as in example
(24).
(24) Es get doch die Creme gge Bibeli; wie heist si scho
it give however this cream against pimples; how called she yet
weder ? AG
again
There is this cream against pimples; what is its name again ?
A variation of the inherent-unique use of the demonstrative is the emotional

use.13 The purpose of this function is also not to fix the reference but to express
an emotion. The emotion may be positive (25a) or negative (25b).
(25) a. Aso, i dmm Barcelona, i sg ich, eschs so schn gsi. AG

well in this Barcelona I tell you is=it so nice been
Well, in Barcelona, I tell you, it was so nice.
b. D bld Cheib het doch vou mis Outo zu Schrott

this stupid moron has effectively prt my car to scrap
gfaare. SO
driven
This stupid moron has effectively smashed my car.
These two functions remain unmentioned in the grammars of Swiss German.

The standard literature on determiners in various dialects and languages, how-
ever, presents functions that seem similar to the above-described functions of
problematized and/or emotional inherent-unique usesee for example Ebert
13 I borrowed the name of this function from Hartmann (1982). In Breu (2004), this function
is named emphatic use. One of my reviewers suggested to not interpret these examples as
counter examples to the default distribution but to assume that they are cases of anaphoric
use with an emotional upgrade as pragmatic Grician effect.
162 studler
(1971), Auer (1981), Prince (1981), Hartmann (1982), Bisle-Mller (1991), Himmel-
mann (1997), Vangsnes (1999, 2001), and Breu (2004).
In conclusion, there areapart from the normal distribution of the definite
determinerssome displacements to record, which indicate a unidirectional
language change. First, the reduced article shows up in anaphoric-unique con-
texts, regularly in anaphoric contexts, and with limitation in autophoric con-
texts. Secondly, the full article can be used in deictic-unique contexts. Thirdly,
the demonstrative emerges in specific inherent-unique contexts. In addition
to the takeovers, there are strict barriers: the reduced article never shows up
in deictic-unique function; the full article never shows up in inherent-unique
function; the demonstrative may not be used in anaphoric-unique function.
5 Syntax: The Syntactic Distribution of the Three Paradigms
In the last section, I will scrutinize whether the correlation between morphol-
ogy and semantics has an analogue in syntax. A correlation between all three
componentsmorphology, semantics, and syntaxwould require three dif-
ferent syntactic structures for the three paradigms. Therefore, I will present a
syntactic structure within the framework of generative grammar, which accom-
modates the desired correlation. As a first step, the feature structure of the
three paradigms will need to be clarified. As a second step, I will show that the
corresponding feature structures resp. the three relevant features of the three
paradigms constitute three different positions in the syntactic structure of the
noun phrase.
5.1 The Feature Structure of Determiners

By applying the grammaticalization process of determiners, as pursued in
Lehmann (2002), Himmelmann (1997), or Lyons (1999), we will be able to
distinguish the determiner paradigms by means of their feature structure.
Lehmann (2002: 33) assumes that three features are relevant for the grammat-
icalization from demonstrative to article: a deictic element [DEIKT], a def-
initeness element combined with a demonstrative feature [DEM/DEF], and
a categorial element [DET]. While the deictic element [DEIKT] may weaken
to an anaphoric element [ANAPH], the definiteness element loses its [DEM]
feature. The categorial element [DET] sustains for all determiner paradigms
within the grammaticalization process. Based on these features, the following
positions in the grammaticalization path are assumed (naming from Lehmann
2002: 49, and naming of the feature combinations in the style of Himmelmann
1997: 27):
demonstrative pronoun DEIKT DEM/DEF DET

anaphoric pronoun ANAPH DEM/DEF DET
anamnestic pronoun DEM/DEF DET
demonstrative determiner DEIKT /DEF DET

weakly demonstrative definite determiner ANAPH /DEF DET
definite article /DEF DET

affixal article /DEF DET
noun marker / DET
These positions in the grammaticalization path mirror the three paradigms

in their different uses. The first three positions are reserved for the demon-
strativethe demonstrative is the only one with a demonstrative element.
The deictic feature weakens for the anaphoric-unique use to [ANAPH] and
disappears for the inherent-unique use (or as Himmelmann names it, the
anamnestic use). For the full article, there are two positions:14 the demonstra-
tive position with a deictic feature for the deictic-unique use of the full arti-
cle and the weakly demonstrative position with an anaphoric feature for the
anaphoric-unique use. The reduced article in its inherent-unique use is marked
by the feature [DEF] alone. In its anaphoric-unique use, it coincides with the
full article in its anaphoric-unique use. The position affixal article is repre-
sented by the reduced article, which fuses with prepositions, as in im Garte
in-the garden. The position noun marker is not represented in the Swiss Ger-
man determiner system.
5.2 The Syntactic Structure of Noun Phrases

One might ask whether the feature structure of the three paradigms has an
effect in a feature-driven minimalist syntax theory and how the syntactic
structure of the Swiss German noun phrase might look. I propose a structure
with three positions for the three relevant determiner features definiteness,
anaphoricity, and deixis.
In the discussion about the structure of the noun phrase, several categories
are proposed for the functional features. How many features and which
featuresin addition to the functional category D(et)are required is a mat-
14 The naming of these two positions is somewhat awkward since they certainly have a feature
[DEIKT] or [ANAPH], but no [DEM] feature. The [DEM] feature is reserved for the demon-
strative.
164 studler
ter of controversial debate (see Abney 1987). Similar to the left periphery of the
verb (Rizzi 1997), a left periphery with morphosyntactic, semantic, modifying,
and information structural features for the noun phrase is proposed (for details
see Alexiadou et al. 2007 and references cited there). For our three determiner
paradigms, the projection structure of the three relevant features[DEF],
[ANAPH] and [DEIKT]is central. Definiteness is assumed to be encoded in a
FinP-like DefP. Anaphoricity marks old information and therefore encoded in
TopP, similar to the projection for old information in the verb phrase (see for
example Ihsane & Pusks 2001). For the deictic component of demonstratives,
a DxP is proposed (see for example Vangsnes 1999, 2001).
5.3 Three Positions for the Three Paradigms

Related to the preliminary work on the noun phrase structure, I propose for the
Swiss German noun phrase three relevant projections over the NPa DefP,
an AnaphP, and a DxP. These projections need to project depending on the
particular determiner features. The proposed noun phrase structure is shown
in (26).
(26)
The reduced article in its inherent-unique use is equipped only with a feature
[DEF] for definiteness. The article is generated in the head position of D(ef)P,
where its definiteness feature can be checked. In the case of the reduced article,
the noun phrase itself refers inherent-uniquely; therefore, the article is not used
to make the noun phrase definite. The reduced article is the default article
in Swiss German and is therefore not provided with referential force. This is
reflected in its low position and in the flat structure with only a D(ef)P as
functional projection.
The full article in its anaphoric-unique use (and also the reduced article or
the demonstrative in anaphoric-unique use) has, in addition to the definiteness
feature, a feature [ANAPH] for anaphoricity. Since the asserted TopP is prob-
lematic in many respects, I assume a less controversial AnaphP over D(ef)P.15
The full article is raised to AnaphP to check its [ANAPH] feature. The higher
position reflects the referential force of the full article and its wider scope.
The demonstrative in its deictic-unique use (and also the full article in
its deictic-unique use) is equipped with a feature [DEIKT]. It is presupposed
that deixis is always expressed at the leftmost periphery of the noun phrase.
With the assumption of a leftmost situated DxP, the [DEIKT] feature can be
checked. Based on the data of other languages, like Greek or Scandinavian,
it is assumed that the demonstrative is not base-generated in this position
but rather in a lower position between DP and NP (Bernstein 1997; Brug
2002; Giusti 2002), e.g. in a projection DemP, where the [DEM] feature can be
checked. From DemP, the demonstrative cyclically moves through DefP to DxP.
If the demonstrative is used (problematized/emotionally) inherent-uniquely,
it moves only to DefP to check the definiteness feature. The referential force
of the demonstrative is also mirrored in its positionthe more features it
possesses, the more referential force it has and the higher the landing site
projection.
Consequently, I propose an analysis of relative clauses as follows. In accor-
dance with the standard analysis that different kinds of relative clauses match
different projections (see Jackendoff 1977; Demirdache 1991; Alexiadou et al.
2000), I assume for every kind of relative clause a different position in the noun
phrase structure. Appositive relative clauses, which are not relevant for the
reference fixing and hence do not need to be in the scope of the article, are
adjuncts of a high position in the noun phrase structure, e.g. of DxP. Defining
restrictive relative clauses, which are only in the scope of the full article and
hence ungrammatical with the reduced article, are adjuncts of DefP. Concept-
forming restrictive relative clauses, which forms a concept with the noun, are
adjuncts of a low position, namely of NP. In this position, they are in the scope
of the reduced article, and therefore the reduced article is grammatical. The
proposed noun phrase structure and the three options for relative clauses are
as follows.
15 TopP is problematic for at least three reasons. First, it should stand for old information,
but Topic is not equated with old information (see Molnar 1993; Jacobs 1992). Secondly, the
concept of old information might be adequate with the anaphoric reference of the article but
not with the autophoric reference. Thirdly, the parallelism between a TopP in the noun phrase
and a TopP in the verb phrase is too vague and not asserted. Recent proposals advocate for a
little nP as an alternative option, in accordance with the little vP in the verb phrase. Like my
proposed AnaphP, the little nP opens a position for the full article.
166 studler
(27)
Wiltschko (2012, to appear) has a very similar approach. Based on Chinese data,
she proposes to distinguish descriptive relative clauses (to compare with my
concept-forming relative clauses) from restrictive relative clauses (to compare
with my defining relative clauses). To explain the distribution of these relative
clause types with two article paradigms in Austro-Bavarian she assumes, par-
allel to my approach, that the articles select for distinct projections and that
the relative clauses have distinct adjunct positions. The full article selects for
nP (over NP), whereas the reduced article selects for NP. Since restrictive rela-
tive clauses are assumed to be adjuncts to nP, they are compatible with the full
article. Because descriptive relative clauses are assumed to be adjuncts to NP,
they are compatible with the reduced article.16 While the two articles are in the
same position in Wiltschkos analysis, namely in D, I assume that the two arti-
cles areaccording to their feature structurein different positions (see also
Studler 2004). The reduced article remains in D(ef) and the full article moves
to Anaph.
16 I thank both my reviewers for pointing out to me Wiltschko (2012, to appear).

6 Conclusion
The aim of this paper was to describe the three determiner paradigmsthe
reduced article, the full article, and the proximal demonstrativein Swiss Ger-
man and to evince their similarities and differences with respect to their mor-
phological form, semantic function, and syntactic structure. For this purpose, I
assumed a strong correlation between morphology, semantics, and syntax inso-
far as every difference in form correlates with a difference in function and with
a difference in structure.
I demonstrated how morphological forms characterize the three paradigms,
and I showed which semantic functions must be considered for determiners.
Based on several data corpora, it could be shown that there is a correlation
between form and function resp. between morphology and semantics: the
reduced article is used in inherent-unique contexts; the full article appears
in anaphoric-unique contexts; the demonstrative is used in deictic-unique
contexts. The exceptional cases to this standard distribution turned out to
be strictly regular rather than arbitrary: the reduced article may appear in
anaphoric-unique function, the full article in deictic-unique function, and the
demonstrative in inherent-unique function. Moreover, strict barriers could also
be detected: the reduced article never shows up in deictic-unique function and
the full article is never used in inherent-unique function; the demonstrative
may not appear at all in anaphoric-unique function.
To explain the distribution of the determiner paradigms in Swiss German,
I proposed, based on the respective feature structures, three functional pro-
jections over the NPa D(ef)P to check the definiteness feature, an AnaphP
over the D(ef)P to check the feature for anaphoricity, and a DxP at the left-
most periphery to check the deictic feature. The syntactic distribution with
three functional projections is motivated by the feature structure of the deter-
miners and by common assumptions about the noun phrase structure. The
final landing sites of the determiners may explain the distribution with differ-
ent kinds of relative clauses. Given that relative clauses could be matched to
different projections, and hence generate different scope relations, I assumed
that appositive relative clauses are adjuncts of DxP, defining restrictive relative
clauses are adjuncts of D(ef)P, and concept-forming restrictive relative clauses
are adjuncts of NP.
The presented noun phrase structure mirrors the characteristic properties
of the three paradigms in Swiss German: the more morphological material the
determiner has, the more features are involved; the more features involved, the
higher the landing site projection; the higher the landing site projection, the
wider the scope and the referential force of the determiner.
168 studler
References
Abney, Steven. 1987. The English noun phrase in its sentential aspect. PhD. diss., Cam-
bridge, MA.
Alexiadou, Artemis, Paul Law, Andr Meinunger, and Chris Wilder. 2000. Introduction
to The syntax of relative clauses, edited by Artemis Alexiadou, Paul Law, Andr
Meinunger, and Chris Wilder, 151. Amsterdam: Benjamins.
Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. 2007. Noun phrase in the
generative perspective. Berlin/New York: Mouton de Gruyter.
Auer, Peter. 1981. Zur indexikalittsmarkierenden Funktion der demonstrativen Ar-
tikelform in deutschen Konversationen. In Sprache. Verstehen und Handeln, edited
by Gtz Hindelang, and Werner Zillig, 301311. Tbingen: Niemeyer.
Bernstein, Judy. 1997. Demonstratives and reinforcers in Romance and Germanic
languages. Lingua 102:87113.
Bisle-Mller, Hansjrg. 1991. Artikelwrter im Deutschen. Semantische und pragmatische
Aspekte ihrer Verwendung. Tbingen: Niemeyer.
Bohnenberger, Karl. 1913. Die Mundart der deutschen Walliser im Heimattal und in den
Aussenorten. Frauenfeld: Huber.
Bossard, Hans. 1962. Zuger Mundartbuch. Grammatik und Wrterverzeichnisse. Ein Weg-
weiser zur guten Mundart. Zrich: Schweizer Spiegel Verlag.
Slavistische Linguistik 2002. Referate des XXIII. Konstanzer Slavistischen Arbeitstre-
ffens Bochum, 10.-12.9.2002, edited by Marion Krause, and Christian Sappok, 957.
Mnchen: Sagner.
Brug, Laura. 2002. The positions of demonstratives in the extended nominal projec-
tion. In Functional structure in DP and IP. The cartography of syntactic structures,
volume 1, edited by Guglielmo Cinque, 1553. Oxford: Oxford University Press.
Brugger, Gerhard, and Martin Prinzhorn. 1995. Some properties of German determiners.
ms., Los Angeles/Vienna: University of South California/Universitt Wien.
Christen, Helen. 1998. Dialekt im Alltag. Eine empirische Untersuchung zur lokalen Kom-
ponente heutiger schweizerdeutscher Varietten. Tbingen: Niemeyer.
Christopherson, Paul. 1939. The articles. A study of their theory and use in English.
Copenhagen: Munksgaard.
Delsing, Lars-Olof. 1993. The internal structure of the noun phrases in the Scandinavian
languages. PhD diss., University of Lund.
Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives, and
dislocated structures. PhD. diss., MIT, Cambridge, MA.
Donnellan, Keith. 1966. Reference and definite descriptions. Philosophical Review
75:281304.
Ebert, Karen. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem Nord-
friesischen Dialekt. PhD diss., Universitt Kiel.
Eroms, Hans-Werner. 1989. Artikelparadigma und Artikelfunktionen im Dialekt und

in der Standardsprache. In Bayerisch-stereichische Dialektforschung. Wrzburger
Arbeitstagung 1986, edited by Erwin Koller, Werner Wegstein, and Norbert Richard
Wolf, 305328. Wrzburg: Knigshausen & Neumann.
Felix, Sascha. 1988. The structure of functional categories. Groninger Arbeiten zur
Germanistischen Linguistik (GAGL) 29:3762.
Fischer, Ludwig. 1989. Luzerndeutsche Grammatik. Ein Wegweiser zur guten Mundart.
Zrich: Schweizer Spiegel Verlag.
Giusti, Giuliana. 2002. The functional structure of noun phrases. A bare phrase struc-
ture approach. In Functional structure in DP and IP. The cartography of syntactic
structures, edited by Giulielmo Cinque, 5490. Oxford: Oxford University Press.
Gunkel, Lutz. 2007. Nominalphrasen des Typs derjenige (N) + Relativsatz in den
europischen Sprachen. in Deutsche Grammatik im europischen Dialog. Beitrge
zum Kongress Krakau, edited by Norbert Fries, and Christiane Fries. http://krakau
2006.anaman.de/beitraege/gunkel.pdf
Hankamer, Jorge, and Line Mikkelsen. 2002. A morphological analysis of definite nouns
in Danish. Journal of Germanic Linguistics 14.2:137175.
Hartmann, Dietrich. 1982. Deixis and anaphora in German dialects. The semantics and
pragmatics of two definite articles in dialectal varieties. In Here and There, edited by
Jrgen Weissenborn and Wolfgang Klein, 187207. Amsterdam/Philadelphia: Ben-
jamins.
Hawkins, John A. 1978. Definiteness and indefiniteness. London: Croom Helm.
Heck, Fabian, Gereon Mller, and Jochen Trommer. 2008. A phase-based approach to
Scandinavian definiteness marking. In Proceedings of WCCFL 26, edited by Charles
B. Chang, and Hannah Haynie, 226233. Sommerville, MA: Cascadilla Press.
Heinrichs, Heinrich Matthias. 1954. Studien zum bestimmten Artikel in den germanis-
chen Sprachen. Giessen: Wilhelm Schitz.
Heusinger, Klaus von. 1996. Definite Kennzeichnungen, Anaphora und Salienz. Lin-
guistische Berichte 163:197226.
Heusinger, Klaus von. 1997. Salienz und Referenz. Der Epsilonoperator in der Semantik
der Nominalphrase und anaphorischer Pronomen. Berlin: Akademie Verlag.
Heusinger, Klaus von. 2002. Specificity and definiteness in sentence and discourse
Himmelmann, Nikolaus. 1997. Deiktikon, Artikel, Nominalphrase. Zur Emergenz syntak-
tischer Struktur. Tbingen: Niemeyer.
Hodler, Werner. 1969. Berndeutsche Syntax. Bern: Francke.
Hotzenkcherle, Rudolf. 1934. Die Mundart von Mutten. Laut- und Flexionslehre. Frauen-
feld: Huber.
Ihsane, Tabea, and Genoveva Pusks. 2001. Specific is not definite. Generative Gram-
mar in Geneva 2:3954.
170 studler
Jackendoff, Ray. 1977. X-bar syntax. A study of phrase structure. Cambridge, MA: MIT
Press.
Jacobs, Joachim. 1992. Informationsstruktur und Grammatik. Opladen: Westdeutscher
Verlag.
Julien, Marit. 2005. Nominal phrases from a Skandinavian perspective. Amsterdam: Ben-
jamins.
Kolde, Gottfried. 1996. Nominaldetermination. Eine systematische und kommentierte
Bibliographie unter besonderer Bercksichtigung des Deutschen, Englischen und
Franzsischen. Tbingen: Niemeyer.
Kripke, Saul. 1980. Naming and necessity. Cambridge, MA: Harvard University Press.
Lehmann, Christian. 1984. Der Relativsatz. Typologie seiner Strukturen, Theorie seiner
Funktionen, Kompendium seiner Grammatik. Tbingen: Narr.
Lehmann, Christian. 2002. Thoughts on grammaticalization. Erfurt, Arbeitspapiere des
Seminars fr Sprachwissenschaft der Universitt Erfurt 9.
Lbner, Sebastian. 2011. Concept types and determination. Journal of Semantics
28:279333.
Lewis, David. 1979. Scorekeeping in a language game. In Semantics of different points
of view, edited by Rainer Buerle, Urs Egli, and Arnim von Stechow, 172187. Berlin:
Springer.
Lyons, Christopher. 1999. Definiteness. Cambridge, MA: Cambridge University Press.
Marti, Werner. 1985. Berndeutsch-Grammatik. Bern: Francke.
Meinherz, Paul. 1920. Die Mundart der Bndner Herrschaft. Frauenfeld: Huber.
Molnar, Valeria. 1993. Zur Pragmatik und Grammatik des TOPIK-Begriffes. In Wortstel-
lung und Informationsstruktur, edited by Marga Reis, 155202. Tbingen: Niemeyer.
Nbling, Damaris. 1992. Klitika im Deutschen. Schriftsprache, Umgangssprache und ale-
mannische Dialekte. Tbingen: Narr.
Ortmann, Albert. in press. Definite article asymmetries and concept types: seman-
tic and pragmatic uniqueness. In Proceedings Workshop 2009, Universitt Dssel-
dorf.
Penner, Zvi. 1993. The earliest stage in the acquisition of the nominal phrase in Bernese
Swiss German. Syntactic bootstrapping and the architecture of language learning.
Bern, Arbeitspapier 30, Institut fr Sprachwissenschaft, Universitt Bern.
Prince, Ellen. 1981. On the inferencing of indefinite-this NPs. in Elements of discourse
understanding, edited by Aravind Krishna Joshi, Bonnie Lynn Webber, and Ivan
Andrew Sag, 231250. Cambridge, MA: Cambridge University Press.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar,
edited by Liliane Haegeman, 282337. Dordrecht: Kluwer.
Russell, Bertrand. 1905. On Denoting. Mind 14:479493.
Scheutz, Hannes. 1988. Determinantien und Definitheitsarten im Bairischen und Stan-
darddeutschen. In Festschrift fr Ingo Reiffenstein zum 60. Geburtstag, edited by

Peter K. Stein, Andreas Weiss, and Gerold Hayer, 231258. Gppingen: Kmmerle.
Scholze, Lenka. 2007. Das grammatische System der obersorbischen Umgangssprache.
Unter besonderer Bercksichtigung des Sprachkontakts. PhD diss., Konstanz: Kon-
stanzer Online-Publikations-System (KOPS).
Schwarz, Florian. 2009. Two types of definites in natural languages. PhD diss., UMass at
Amherst.
Searle, John. 1969. Speech acts. An essay in the philosophy of language. Cambridge,
MA/London: Cambridge University Press.
Strawson, Peter. 1950. On Referring. Mind 59:320344.
Studler, Rebekka. 2004. Voller und reduzierter Artikel in der schweizerdeutschen DP.
In Linguistische Studien im Europischen Jahr der Sprachen. Akten des 36. Linguis-
tischen Kolloquiums in Ljublijana 2001, edited by Stojan Brai, Darko uden, Saa
Podgorek, and Vladimir Poganik, 625635. Frankfurt a.M.: Peter Lang.
Studler, Rebekka. 2011. Artikelparadigmen. Form, Funktion und syntaktisch-semantische
Analyse von definiten Determinierern im Schweizerdeutschen. Elektronische Hoch-
schulschrift Universitt Zrich.
Suter, Rudolf. 1992. Baseldeutsch-Grammatik. Basel: Merian.
Vangsnes, ystein. 1999. The identification of functional architecture. PhD diss., Univer-
sity of Bergen.
Vangsnes, ystein. 2001. On noun phrase architecture, referentiality, and article sys-
tems. Studia Linguistica 55.3:249299.
Weber, Albert. 1987. Zrichdeutsche Grammatik. Zrich: Hans Rohr.
Wiltschko, Martina. 2012. What does it take to host a (restrictive) relative clause?
Working Papers of the Linguistics Circle of the University of Victoria 21.2:100145.
Wiltschko, Martina. to appear. Descriptive relative clauses in Austria-Bavarian Ger-
man. Canadian Journal of Linguistics.
Reduced Definite Articles with
Restrictive Relative Clauses*
The present study examines the interaction between the occurrence of relative
clauses (RCs) in noun phrases, and the choice of full or reduced definite articles
on the relativised (head) nouns. The data are drawn from West Germanic
languages that have two series of definite articles (Austro-Bavarian German,
Bavarian, Fehring Frisian, Swiss German dialects), or traces of such a system
(Standard German).
In the literature on definite determiners in Standard German (Hartmann
1980 and others), Austro-Bavarian (Brugger and Prinzhorn 1996; Wiltschko
2010) and Bavarian (Weiss, 1998) it is claimed that restrictive RCs cannot com-
bine with reduced definite determiners on the head noun (see 1a). In other West
Germanic varieties this restriction does not hold, however, (Ebert 1970, 1971 on
Fehring Frisian, Studler 2008: 95108, 310 for a range of Swiss German varieties)
(see 1b).1
(1) a. I hobs Buach, des (was) da Chomsky gschriem hot,

I have-detred book that (which) the Chomsky written has
nit glesn (ABav)
not read
I didnt read the book, which Chomsky wrote.
(non-restrictive reading of the RC only, Brugger and Prinzhorn
1996:1213)
* I am very grateful to an anonymous reviewer of a previous paper who pointed out to me that,
contrary to what I claimed at the time, contracted P+D amalgams do allow restrictive relative
clauses in German. I thank Gerhard Schaden, Rebekka Studler and Helmut Weiss for help
with the data, Nomi Erteschik-Shir, Lutz Gunkel and Klaus von Heusinger for comments and
useful references. I am particularly grateful to Matthew Baerman for his valuable comments
and questions on a previous version of this paper. All remaining errors and misinterpretations
are mine.
1 The following abbreviations are used for the languages: ABav = Austro-Bavarian, Bav =
Bavarian, FFr = Fehring Frisian, Ge = German, Sw = Swedish, VAr= Vorarlberger dialect.
Abbreviations in the glosses: comp= complementiser, cop= copula, dem= demonstrative,
det= determiner, expl = expletive, inf = infinitive, pl= plural, prs= present, prt= particle,
pst= past, rel= relative complementiser, rel.pr = relative pronoun.

reduced definite articles with restrictive relative clauses 173
b. Det/ At iast buk, wat hi skrewen hee, docht

detfull/ detred first book rel he written has is-worth
niks. (FFr)
nothing.
In what follows I show that these diverging observations are due to two inter-
related factors. First, the incompatibility between reduced definite determin-
ers and restrictive RCs only concerns a subset of restrictive RCs. Secondly, the
examples of restrictive RCs considered in the different studies are not of the
same type.
The paper is structured as follows. In section 2 I briefly summarise the prop-
erties of reduced and full definite determiners found in West Germanic. I then
present the claims made in the literature about the interaction of restrictive
RCs with the choice of full or reduced definite articles on the head noun, as
well as counterexamples to these claims from Standard German and the Vorarl-
berger dialect.
In section 3, I discuss the definition of restrictive RCs, showing that the
restrictive RCs considered in the literature are only a subclass of restrictive
relatives. I argue that we should acknowledge at least four sub-types of restric-
tive relatives. In section 3.5, I examine further semantic factors that have been
shown to influence the choice of full or reduced definite determiners on rela-
tivised nouns.
In section 4, I discuss data from the Vorarlberger dialect, Standard German
and Fehring Frisian in the light of the proposed classification of RCs.
In section 5, I show that different studies rely on different types of examples. I
show that the restriction on restrictive RCs stated in the literature does not con-
cern restrictive relatives in general, but only one of the subtypes identified here,
namely contrastive RCs. I further show that other types of restrictive relatives
co-occur with reduced definite determiners in Fehring Frisian as described
by Ebert (1970, 1971), and in the Vorarlberger dialect (an Austrian Alemannic
dialect).
2 Restrictive Relative Clauses and Article Choice on the Head Noun
2.1 Reduced and Full Definite Determiners

Many West Germanic dialects have two distinct paradigms of definite deter-
miners, often referred to as weak and strong definite determiners (Heinrichs
1954, Ebert 1970, 1971, a.o., for similar contrasts in Upper Sorbian see Breu 2004).
174 patricia cabredo hofherr
These definite determiners differ in that they are morphologically reduced

(detred) or un-reduced (detfull).2 While in many West Germanic dialects the
reduced determiner is a reduced form of the full determiner, this need not be
so, as shown by Fehring Frisian, where the reduced form a is not transparently
a reduced form of the full definite determiner dn (see Ebert 1970, 1971). To
avoid confusion with weak definite NPs as discussed in Carlson and Suss-
man (2005); Carlson et al. (2006) I will call the two types of definite determiners
reduced and full definite determiners. In what follows, I give a brief summary
of the semantic contrasts between reduced and full definite determiners, essen-
tially following Eberts seminal study (Ebert, 1971). The morphological distinc-
tion between reduced and full definite determiners correlates with a semantic
distinction which can be summarised as follows:3
(3) a. The full definite forms have anaphoric and deictic function.
b. The reduced definite forms are used for non-anaphoric and non-
deictic definite readings. They mark entities which are generally
unique (Eberts Unika):
i. Absolute uniqueness
the sun, the moon, the pope
ii. Situational uniqueness
within a flat: the kitchen, the bathroom,
within the social setting: father, mother, the priest, the doctor,
on a ship: the rudder, the stern, the engine room.
However, the two contexts in (3) are not exhaustive but rather at opposite
ends of the scale from pragmatic to semantic uniqueness (i.e. from discourse-
situation uniqueness to inherent uniqueness as proposed, e.g., in Lbner 2011).
2 Morphologically reduced and un-reduced definite determiners are also called weak and
strong definite determiners in the literature. I avoid this terminology as it leads to
confusion with an independent phenomenon, calle weak definite NPs.
Morphologically un-reduced determiners differ from demonstratives in that, like English
the, they cannot be used in contexts like (ia):
(i) a. #The man came and the man didnt come.
b. Compare:
That man came and that man didnt come.
This man came and this man didnt come.
3 As Schiering (2002) points out, this distinction co-incides with Lbners distinction between
semantic and pragmatic definiteness (Lbner 1985, 2011).
(4) Scale of uniqueness (Lbner 2011:320) pragmatic semantic4

a. deictic with sortal common NPs (e.g. this book)
b. anaphoric with sortal common NPs (e.g. a man the man)
c. sortal common NPs with establishing RC (e.g. the woman that John
married)
d. functional common NPs with explicit definite possessor (e.g. his
mother)
e. definite associative anaphors (a carthe motor)
f. individual common NPs (e.g. the sun)
g. proper names (e.g. the Nile, the Alps)
Furthermore, as Wiltschko (2012) points out, the term situationally unique

is not sufficiently precise since there is a sense in which the referent of an
anaphoric or deictic DP is also situationally unique. Wiltschko stresses that
the use of the reduced definite determiner detred has to do with the way
uniqueness is established:
(5) detred is used if the uniqueness of the discourse referent does not need
to be established. It is in this sense that the use of detred does not depend
on the discourse context. As such it is crucial that we distinguish between
the common ground that is independent of the current conversation (we
may call it the common background) and the conversational common
ground, which is sensitive to and manipulated by the ongoing discourse
(see Krifka 2008).
wiltschko 2012:91
Cases of nominal anaphora are an unambiguous case of reference to a discourse

antecedent, and consequently the full determiner is obligatory:
(6) Da Petr isch gescht do gsi.

detred Peter is yesterday here been.
4 Sortal nouns are are unary predicate terms, of type e,t, such as table, tree. Individual
nouns are individual terms, of type e such as pope, US president, sun. Relational nouns
are binary predicate terms, of type e,e,t such as brother, sister. Functional nouns are
unary function terms, of type e,e such as mother, father, head. For details see Lbner (2011:
280282).
Dea trottl hot scho widr min reagaschiam ipackt. (VAr)

detfull idiot has again my umbrella taken
(Cabredo Hofherr and Schaden 2012b)
In the contexts in (3)/(6), the distribution of reduced and full determiners is

consistent across different dialects. In other contexts, however, for example
with adjectives between D and N, variation can be observed:
(7) a. In Swiss German dialects attributive adjectives force the reduced deter-
miner, except for the feminine and the plural, where the full form is
used.
studler 2008
b. In Bavarian reduced and full definite articles are both possible with
adjectives. This is visible for all forms of the masculine and the neuter
singular article and the dative feminine singular. In the nominative and
accusative feminine singular and in all cases in the plural the full form
of the article has to be used.
weiss 1998:9193
These examples of variation in determiner choice are not of the same type as
the examples of variation with RCs in (1). At least for some languages, the form
of the determiner when followed by an adjective has been argued to be mor-
phologically conditioned (see Weiss 1998:93 for Bavarian). When examining the
choice of definite determiner in Bavarian below, we therefore have chosen com-
binations of case, number and gender that mark the difference between full
and reduced determiners.
2.2 Determiner Choice and Relative Clauses

The claim that reduced definite determiners cannot combine with restric-
tive RCs has been made for Standard German and for (Austro-)Bavarian Ger-
man (for Standard German see Hartmann (1980), Raffelsiefen (1987), Nbling
(2005: 112), Eisenberg et al. (2006: 624), Puig-Waldmller (2008: 148) Schwarz
(2009, Ch. 2.1, Ch. 6.4.2); for Austro-Bavarian see Brugger and Prinzhorn (1996),
Wiltschko (2012: 72), for Bavarian see Weiss (1998: 91)).5
5 Brugger and Prinzhorn (1996) and Wiltschko (2012) argue that reduced determiners require
contextual uniqueness and that they are therefore incompatible in principle with restrictive
RCs as they presuppose the existence of one or more entities in the domain of discourse that
satisfy the head noun but not the restriction of the RC. I will come back to this argument
below in section 5.
However, there are counterexamples for this claim for Standard German. In
this language, the reduced definite determiner is present in a number of gram-
maticalised P+D amalgams such as im in+ detred.dat.sg.m/n (see Cabredo
Hofherr 2012a,b and references therein) and PPs introduced by such amalgams
are compatible with restrictive RCs, as the following examples show.
(8) Im Institut, in dem ich vorher gearbeitet habe, war das

in+ detred institute in rel.pr I before worked have was that
kein Problem. (Ge)
no problem
In the institute that I worked in before that wasnt a problem.
(example due to an anonymous reviewer)
(9) Es drfe nicht sein, dass Anwohner, die eine Nachtparkgebhr

it should not be that inhabitants that a night-parking-fee
zahlen, im Wohnquartier, in dem sie leben, keinen
pay in+detred quarter in that they live no
freien Parkplatz mehr finden,
free parking-space anymore find
It is not acceptable that inhabitants that pay a parking fee for night
parking cant find a parking space in the quarter they live in anymore.
A97/SEP.22788 St. Galler Tagblatt, 08.09.1997
(10) Im Betrieb, in dem ich Werkzeugmacher gelernt habe, gab

in+detred factory, in that I tool-maker learned have, gave
es 1000 Mitarbeiter.
expl 1000 employees.
In the factory where I did my apprenticeship to become a toolmaker,
there were 1000 employees. BRZ08/APR.11354 Braunschweiger Zeitung,
22.04.2008
(11) Der Besuch des ehemaligen Kellners im Privatclub,

the visit det.gen former waiter.gen in+detred private-club
in dem sie arbeitete, endete fr die Frau in
in that she worked ended for the woman in
Todesangst.
fear-for-her-life
The visit of the former waiter to the private club, in which she worked,
ended in her fearing for her own life. M98/DEZ.93016 Mannheimer Mor-
gen, 09.12.1998
These counterexamples cannot be explained by the fact that reduced deter-

miners in Standard German are a grammatical remnant that only exists in
amalgamated P+D forms. In fact, this type of example can be found in the
Vorarlberg dialect that has two full paradigms of definite determiners, showing
that it is not exceptional as such for reduced definite determiners to combine
with restrictive RCs.
(12) a. s/ *des inschtitut wo i frhr gschaffat ho isch bessr

detred/ detfull institute rel I before worked have is better
organisiert gsi. (VAr)
organised been.
The institute that I used to work in was better organised.
b. mir hot s/ des inschtitut wo i frhr gschaffat ho

to-me has detred/ detfull institute rel I before worked have
bessr gfalla.
better pleased.
I preferred the institute that I used to work in before.
The observation that P+D amalgams are compatible with restrictive RCs con-
verges with the observations in Studler (2008: 95108, 310) for Swiss German
dialects and Ebert (1970) for Fehring Frisian that restrictive RCs can appear with
a reduced definite determiner on the head noun. Studler shows for a range of
Swiss German varieties that while restrictive RCs appear with the full definite
determiner significantly more often, it cannot be maintained that the reduced
definite determiner is excluded in these contexts (Studler, 2008: 95108, 310).
For Fehring Frisian, Ebert (1970) explicitly points out that for certain types of
relativised NPs a reduced D is compatible with a restrictive relative (see (1b)
above).6
6 Ebert (1971:160) points out that in example (i)differing minimally from (1b)the identify-
ing information is given inside the RC and that this does not license the reduced article on
the head noun.
(i) Det / *At buk, wat hi tuiast skrewen hee, docht niks. (FFr)
detfull / detred book rel he first written has is-worth nothing
The book that he wrote first is no good.
3 The Classification of Relative Clauses
In the literature on RCs, a distinction is drawn between appositive and restric-

tive RCs. However, it has long been pointed out that this binary classification
is not sufficient. In particular, different authors have provided convincing evi-
dence that non-appositive RCs do not form a uniform class (Ebert, 1971, 1973;
Carlson, 1977; Grosu and Landman, 1998; Gunkel, 2007) and proposed various
distinctions among them (see discussion below). More recently, Cinque (2008)
has argued in detail that appositive RCs do not form a uniform class either.7
Cinque shows that integrated and non-integrated appositive clauses do
not have the same syntactic and semantic properties in Italian. On the basis of
these observations, he concludes that this difference provides an insight into
the cross-linguistic variation found with appositive RCs.
In what follows, I concentrate on the classification of RCs that are not
appositive. I first examine the distinction between restrictive and appositive
RCs (section 3.1). I argue that the notion of contrast implicit in many studies
on restrictive RCs is not co-extensive with the class of restrictive RCs. A special
case of non-contrasted RCs is provided by functional RCs, which I discuss in
section 3.2. In section 3.3, I briefly discuss restrictive RCs in the context of the
third-type relatives argued for by Carlson (1977) and Grosu and Landman
(1998).
In section 3.5, I present further semantic factors besides the appositive/rela-
tive distinction that have been shown to affect article choice with RCs in West
Germanic languages with two definite articles.
3.1 Restrictive vs Appositive Relative Clauses

In the literature on RCs, two types are commonly distinguished: restrictive and
appositive relatives. Appositive RCs are often referred to as non-restrictive (see
e.g. Ebert (1971:134) and references given therein).
The following definition of the two types of relatives is given in Kempson
(2003):
(13) On the one hand, RC sequences may be used to restrict the range of
variables over which the determiner is presumed to quantify; on the other
hand they may, non-restrictively, solely add additional information about
the entity picked out by the determiner-noun sequence alone:
The linguists who were drunk spoiled the party.
7 See also Martin (to appear) on non-restrictive modifiers within the DP.
The restrictive reading is generally taken to correspond to (14a), while the

appositive reading roughly corresponds to a paraphrase such as (14b).
(14) a. The linguists that were drunk spoiled the party. The sober linguists
were blameless. (restrictive reading of the linguists who were drunk)
b. The linguists spoiled the party. They were drunk. (appositive reading
of the linguists who were drunk)
In the literature, non-restrictive is often equated with appositive. However, the

term non-restrictive RCs only makes sense if two conditions are fulfilled: (i)
there is a clear definition of what is understood by a restrictive RC, and (ii)
there is a binary classification of RCs. As there are arguments to posit a distinct
third category of RCs (Carlson 1977, Grosu and Landman 1998), I will avoid the
term non-restrictive and use the following positive definition of appositives in
what follows.
(15) Appositive RCs are RCs that combine with a DP whose referent is already
identified by the DP without the RC.
Given this definition, appositive RCs only provide supplementary information

on a referent that is already identified by the DP without the RC. The clearest
cases of appositive relatives in this sense are RCs that appear with proper
names.
(16) Einstein, who discovered relativity, was a professor at Princeton.
Let us now turn to the second part of the definition in (13) above, namely RCs
that restrict the range of the variable that the determiner quantifies over. In the
formulation of Bach (1974):
(17) a restrictive RC presupposes the existence of entities of which the descrip-

tion given in the RC is not true.
bach, 1974
In many studies, (17) is taken to mean that a definite description with a restric-
tive RC has to satisfy two conditions: (i) it refers to a unique salient entity
in the domain of discourse that satisfies the description combining the head
noun and the restrictive RC and (ii) it also presupposes the existence of one
or more entities in the domain of discourse that satisfy the description of the
head noun, but not the property given by the restrictive RC.
This view corresponds to the reading exemplified in (14a): the restrictive RC

divides the set of linguists in the domain of discourse into those who were
drunk and those who were not, contrasting the two subgroups. This contrasting
reading is taken to be the characteristic reading of restrictive RCs in many
studies.8
In what follows, I argue that this understanding of restrictive relatives is
too narrow. In particular, I will show that different types of restrictive rela-
tives correspond to different information structures. The contrasting reading
of restrictive relatives presented above is only one of at least two possible read-
ings.
First, as pointed out by Ebert (1971), there are cases such as (18) for which
the classification of RCs as restrictive or appositive is less clear cut.9 On the
one hand, given the definition of appositive RCs in (15), these examples do not
contain appositive RCs since the information in the RC is necessary to establish
the referent of the overall DP. On the other hand, the RCs in these examples are
not of the same type as the RC exemplified in (14a) above, which contrasts the
8 de Vries (2006:229) gives a characterisation of maximalising relatives as different from restric-

tive relatives in that there is no contrast with other instances of [the head noun], yet the RC
is essential for the meaning of the sentence.
Wiltschko (2012) discusses the Austro-Bavarian example in (i) below pointing out that
In this context, the mailman is situationally unique []. As such, the RC does not serve to
identify the discourse referent under discussion. This is consistent with the fact that in [(i)]
nothing is said about other mailmen (i.e., mailmen who did not deliver our mail). Thus, the
RC in [(i)] cannot be considered a restrictive RC.
(i) da Briaftroga (wos bei uns austrogn hot)
detred mailman comp at us delivered has
the mailman that used to deliver our mail
9 As Ebert (1971:135) points out, RCs on rhematically introduced noun phrases as in the
example (i) are also variously classified either as restrictive or as appositive. In fact these
examples have both readings (see (ii) and (iii)). Under the reading (ii) the example charac-
terises the subject as the kind of person that owns a messy garden, under the reading (iii) the
example characterises the subject as someone owning a garden, adding some information
about the state of the garden. As these cases have an indefinite article on the head noun, I
will leave them aside here.
(i) Er besass einen Garten, der sehr ungepflegt war.
he owned a garden that very neglected was
He owned a garden that was very neglected (messy). (Eberts ex. 19)
(ii) He owned a very messy garden. (He is the type of person to have a messy garden.)
(iii) He owned a garden. By the way, the garden was very messy.
referent of the relativised DP with other referents in the discourse satisfying

the head noun, i.e. house in (18).
(18) a. Anna bought the house that Ina had inherited

b. Anna bought the house that Ina is now repairing.
In languages like German and English, the fact that the RC is necessary to iden-
tify the referent of the overall DP is reflected in the article choice that appears
in paraphrases. While the paraphrase of the appositive reading given in (14b)
relies on a definite NP and a second clause providing additional information,
the paraphrases of first-mention definite DPs with RCs rely on an indefinite NP
introducing the referent and a second clause providing additional information
that is necessary for the identification of the referent. The same is true for the
examples in (18) paraphrased as in (19).
(19) a. Ina inherited a house. Anna bought it.

Anna bought a house. Ina had inherited it. (Paraphrases for (18a))
b. Anna bought a house. Ina is now repairing it.
Ina is repairing a house. Anna had previously bought it.
(Paraphrases for (18b))
Furthermore, the examples in (18) differ from the appositive reading exempli-
fied in (14a) in their information structure (IS in what follows). In restrictive
RCs of the type exemplified in (14a), a contrastive focus between two subgroups
in the domain of discourse is implied (between linguists that were drunk and
those that werent). The examples in (18) are different: the house that Ina inher-
ited is not necessarily contrasted with other houses that she didnt inherit.
Examples like (18) have an Information Structure corresponding to question-
answer focus, as in the following exchange:10,11
10 Using the formalism based on file-card management proposed in Erteschik-Shir (1997), the
examples can be formalised as a layered information structure:
(i) a. house-TOP [Ina inherited x from her grandmother]-FOC
b. Embedded focus structure: Ina-TOP [inherited house from her grandmother]-FOC
In the embedded focus structure (ib), the head noun of the relativised DP is in focus, calling
for a paraphrase with an indefinite DP. Since house is in focus, a card for house-inherited
from grandmother is opened and put on top of the file, allowing for its use as a definite head
of the relative. This processing of the subordinate IS of the sentence mimics the processing
of the sentence as two independent sentences, as in the paraphrases in (19).
11 The heuristic use of a paraphrase as in (19) is not always applicable. In examples such as
(20) A: Anna bought a house.

B: Which house?
A: The house that Ina had inherited.
Summarising, restrictive modification is defined as modification that is neces-

sary for the identification of the discourse referent of the modified DP (see e.g.
Fabricius-Hansen 2012 and literature cited therein.) Consequently, restrictive
relative clauses should be defined not in terms of contrast, but rather as RCs
that add restrictive modification, i.e. RCs that are necessary for the identifica-
tion of the discourse referents of the DPs containing them. In other words the
restrictive RC, in such cases, restricts the range of the variable that the deter-
miner of the overall DP is quantifying over.
Given these assumptions, we have seen that restrictive RCs can appear with
at least two different types of information structure.12 In one case, the restric-
tion takes the form of an additional property comparable to question-answer-
focus, while in the second case, the head noun with the restrictive relative is
contrasted with elements in the domain of discourse that do satisfy the prop-
erty denoted by the head noun but not the property denoted by the restrictive
relative.
As Fabricius-Hansen (2012) points out, narrow focus (contrastive focus)
implies restrictive modification, so contrastive restrictive RCs are always re-
strictive. The inverse does not hold, however: restrictive modification need not
be contrastive.
(iab), the RC is interpreted as an answer to a question, but these examples do not allow a
paraphrase with an indefinite DP as in (19), as shown in (iia-b). Notice that a paraphrase with
a definite DP is also excluded.
(i) a. What day was she born?
She was born on the day that Trinidad became independent.
b. What type of birds do they hunt?
They hunt the birds that fly south in autumn.
(ii) a. #She was born on a/ the day. Trinidad became independent on it.
b. #They hunt some/ the birds. They fly south in the autumn.
12 The two types of RC can be disambiguated using contrastive stress. Notice, however, that the
contrastive reading seems to be available even without such marked prosody.
3.2 Functional Restrictive Relatives

A particular case of non-contrastive restrictive RCs is provided by examples
that Hawkins (1978) calls establishing RCs. Establishing RCs allow the use
of the relativised DP as a first-mention definite (22a). This contrasts with the
example in (22b), which is not an establishing RC. This type of example is
optimal when the predicates of the head-noun and the RC allow a reading that
plausibly entails that only one referent satisfies the description, i.e. a functional
reading (22a):
(22) a. Whats wrong with Bill?Oh, the woman he went out with last night
was nasty to him. ((3.16) in Hawkins 1978:101)
b. Whats wrong with Bill?#Oh, the woman who was from the south
was nasty to him. ((3.18) in Hawkins 1978:102)
It has to be stressed that the head noun woman in Hawkins example (22a)
is not lexically functional in the way the relational noun mother is for one
of its arguments (see Lbner (2011) above on lexically functional nouns). The
functional relationship between an argument of the RC and the relativised DP
arises through the world knowledge shared by the discourse participants. Only
under the assumption that a man goes out with only one woman per night,
this example allows a functional inference. This contrasts with the examples in
(22b): here the predicate in the RC does not allow us to construe a functional
relationship between Bill and the woman described, hence the DP with the
relativised RC cannot be used as a first-mention definite (marked by # in the
example).
In the examples of functional RCs below, we similarly see that while cinema
and institute are not lexically functional nouns, the following predicate+N
combinations are functional in that they are true of a single referent:13
(23) a. the cinema that x saw Shrek in

f cinema+see film (x,y) the cinema x saw film y in
b. the institute that x used to work in
f institute+work (x) the institute x worked in
13 Notice that the work-place reading of institute I used to work at also has a temporal variable in
the same way as the noun king while being lexically unique at a given time still has a temporal
variable (see (27) below).
It is independently known that functional relationships14 are relevant for

reduced definite head nouns. Referents that are functionally associated with
a given referent use the reduced definite determiner (illustrated here with
Austro-Bavarian, but this is equally true for other languages with two definite
determiners):
(24) a. is/ *des Zentrum vom Universum (ABav)

detred/ detfull centre of-detred universe
the centre of the universe
b. a/ *dea Biagamasta von Hintertupf

detred/ detfull major of Hintertupf
the mayor of Hintertupf
c. Da Hans hotn/ *den strktsten Mann von Los Feliz

the Hans has-detred/ detfull strongest man of Los Feliz
gschlogn
beaten
Hans beat the strongest man of Los Feliz.
(adapted from Brugger and Prinzhorn, 1996:11, ex 53)
Notice that a functional relationship only implies that each argument is asso-
ciated with a single entity by the function in questionthe identity of the
individual need not be known (25a) and the singular definite is compatible
with the existence of several individuals if a quantifier is embedded (25b):
(25) a. The mayor of a small town came to our meeting.

b. The winner of every heat goes through to the final.
Cases in which the [head noun + RC] sequence is interpreted functionally are
restrictive, since only the information provided by the RC allows us to narrow
the range of the variable to a unique entity in the domain of discourse.
(26) a. #Penguin published the first book.

b. Penguin published the first book that Saki wrote.
14 Functional relationships are predicates that assign a single entity to each argument: centre
assigns to its argument a unique centre, mayor is a unique function, each city only having
one mayor, and superlatives select the highest element on a gradable scale as in the strongest
man in X.
In the example in (26b), one reading is obtained by applying the function

first book written by x to a particular author, Saki, yielding a unique entity.15
Unlike DPs that are interpreted as intrinsically unique such as the sun, the
moon, the referent of a functional relativised DP co-varies as a function of
the content of the RC. This co-variation with the RC can be due to a nominal
argument of the RC predicate as in (26b), but also to other elements such as the
temporal argument (see Wespel 2008 for a discussion of temporal arguments
in semantically unique nouns):
(27) The king that ruled France in 1590 was Henri IV.
Cases in which we find a functional interpretation of the RC with respect to the

head noun I will call functional relativised DPs in what follows.
3.3 Restrictive Relatives and Amount/Third-Type Relatives (Carlson 1977,

Grosu and Landman 1998)
In much recent work on RCs, a three-way distinction among RCs is assumed
(Alexiadou et al., 2000; Bianchi, 2000; Grosu, 2002), following Grosu and Land-
man (1998). Developing an insight in Carlson (1977), Grosu and Landman (1998)
argue that a third type of RCs has to be acknowledged, which is neither appos-
itive nor restrictive, and which they call maximalising relatives. Maximal-
ising relatives differ from restrictive relatives in that there is no contrast with
other instances of [the head noun], yet the RC is essential for the meaning of
the sentence (Vries, 2006).
(28) (I spilled) the coffee that there was in the pot. (maximalizing)
(de Vries 2006, ex. 1c)
Given this definition, it is possible that maximalising relatives could be the

same as the non-contrastive restrictive relatives identified above.
Carlson (1977) shows that one syntactic context that distinguishes amount
relatives from restrictive relatives in English is relativisation from a there-
insertion context; however, this test cannot be generalised to other languages
such as German:
15 Another functional reading would take the predicate as a variable: first book P by Saki,
with P ranging over wrote, illustrated, reviewed, As a first approximation I assume that
the lambda-abstracts available for focus interpretation are also available for functional RC
interpretation. Whether this is correct has to be confirmed in further research.
(29) In [(29a)] we see that an RR [restrictive relative] cannot relativize the

logical subject of a context where There-Insertion has applied. In [(29b)],
however, we find a grammatical sentence, where the quantifier has been
changed from some to every. We might at first think of [(29b)] as an RR,
but I will claim that it is rather an AR [amount relative]: Carlson (1977)
a. *Some man there was on the life-raft died.
b. Every man there was on the life-raft died.
Carlson (1977, 522) proposes an analysis of amount relatives that parallels the
analysis of comparative clauses proposed by Bresnan (1973, 1975), involving a
silent noun amount. Carlson argues that for English, amount readings of RCs
are found in examples such as the following:
(30) We will take a year to drink the champagne that they spilled yesterday.
However, the amount reading is not a reliable indicator of amount relatives.

First, Grosu and Landman (1998) point out that amount relatives must have
an individual reading as well (p. 133, 141). And secondly they point out that the
identity of quantity readings [] are not generally available but require con-
textual triggers (often modals and generics). While the identity of quantity-
reading is available for Geld money in (31a), it does not seem to be available in
an example like (31b), modelled on (30):
(31) a. Ich arbeite ein Jahr um das Geld zu verdienen, das dein
I work one year to the money to earn that your
Bruder an einem Wochenende ausgibt. (Ge)
brother on one weekend spends
I work a whole year to earn the money that your brother spends on
one weekend.
b. #Wir brauchen den Rest unseres Lebens um den Champagner

we need the rest of our lives to det champagne
zu trinken, den die gestern verschttet haben. (Ge)
to drink that they yesterday spilled have
It will take us the rest of our lives to drink the champagne that they
spilled yesterday. (token-reading only for champagne, no amount
reading in German)
The German example (31b) is pragmatically anomalous since the only read-
ing is the object reading where the spilt champage is the one that will be
drunk.16 I will therefore use the distributional criteria proposed by (Carlson,

1977) and Grosu and Landman (1998, 126) given in (33 / 34) to establish whether
the non-contrastive restrictive RCs should be analysed as amount-relatives.
(33) Carlsons criteria to distinguish Amount Relatives (ARs) from Restrictive

Relatives (RRs) (Carlson, 1977)
a. ARs can only take a subset of determiners, RRs combine freely with
all of them: the, all, what, that, any, every but not most, several, many,
some, each a (p. 525)
b. No wh-forms with ARs. (p. 526)
*Every man WHO there was disagreed.
c. No count nouns (p. 526)
i. Those/ The/ What men (that) there were in Austria like Bob.
ii. That/ The/ What meat there was was soon eaten by the cougar.
iii. *That/ The/ What man there was in Austria likes Bob.
d. cf. parallel with
i. that much sand (mass N)
ii. that many birds (plural count N)
iii. *that much/many lamp (singular count N)
e. ARs but not RRs can be used in contexts that require a cardinality
expression:
The movie lasted two/several/many/*all/ *the/ *most/*these hours
past my bedtime. (p. 530)
(34) Grosu and Landman (1998, 126): Third-type relatives do not stack.
It is not clear to me if these criteria can be applied to other languages. For

my present purposes it suffices to observe that the English equivalents of the
non-contrastive restrictive relatives discussed above clearly can appear with
singular count nouns; therefore, non-contrastive restrictives cannot be reduced
to amount relatives in English given the above criteria.
16 The same is true for the Vorarlberger dialect, with both versions of the definite article:
(i) Mir wearan da rscht fo seram leaba dafr brucha, zum dean / da
we will the rest of our life for-it need to detfull / detred
sekt zum sufa, wo dia gescht zobad frschttet hond. (VAr)
champagne to drink rel they yesterday evening spilt have
(I thank Gerhard Schaden for this example).
Grosu and Landman (1998) argue that third-type relatives are maximalising.
As we will see below, this makes the relationship between the head-noun and
the relativised DP a functional relationship of the type discussed in section 3.2.
I will therefore consider that maximalising RCs are special cases of functional
relativised DPs and therefore a subcase of restrictive RCs as defined here.
3.4 Tests for Appositive vs Restrictive Relative Clauses

In the previous section I have argued that two types of restrictive RCs have to
be distinguished. In what follows I briefly review the criteria proposed in the
literature to distinguish between appositive and restrictive relatives, and I show
that the two types of restrictive RCs distinguished above pattern together with
respect to the criteria. Carlson (1977) gives the following criteria to distinguish
between restrictive and appositive RCs in English:
(35) Carlsons criteria (Carlson 1977)

a. wh-forms: Appositive relatives, unlike restrictive relatives, must be
introduced by a wh-form, and never by that (or zero):
i. George, (who, *that) likes no one, enjoys Handel.
b. Proper names: Appositive relatives, unlike restrictive relatives, may
co-occur with certain types of head nouns (e.g. proper names):
i. *George that likes no one enjoys Handel.
c. Restrictions on quantifiers: Appositive relatives, unlike restrictive
relatives, may not co-occur with certain quantifiers:
i. *Any lion, which eats small animals, is cowardly.
ii. Any lion that eats small animals is cowardly.
d. Comma intonation for appositives but not for restrictive relatives
Restrictive and appositive RCs differ with respect to variable binding by an

operator: variable binding into the RC is possible for restrictives but not for
appositives (Jackendoff, 1977:176, Wiltschko 2012):
(36) a. Everyone bought a suit that suited him. [RRC]

b. *Everyone bought a suit, which suited him. [ARC]
(Jackendoff (1977, 176), cited apud de Vries (2002))
Furthermore, Carlson claims that appositivesunlike restrictivesdo not

stack (37).17
17 Vries (2002, 197) claims on the basis of example (i) that this is not true in the general case:
(37) Appositive relatives, unlike restrictive relatives, may co-occur on the same
head only if they are conjoined (i.e., they may not stack)
a. The tiger that I saw that I wanted to buy was five weeks old.
b. *The lion, which was five weeks old, which was fed twice a day, ate
only fillet of salmon. (Carlson 1977)
With respect to these criteria, the non-contrastive and functional relatives

identified above pattern with restrictive relatives:
(38) wh-forms: Non-contrastive and functional RCs pattern with restrictive

RCs in allowing that and zero:
a. Anna bought the house (that) Ina had inherited. (=18)
b. The woman (that) Bill went out with was nasty to him.
c. Penguin published the first book (that) Saki wrote.
(39) Restrictions on quantifiers: Non-contrastive and functional RCs pattern

with restrictive RCs in allowing any quantifier on the head noun:
a. Q: What vegetables do you use for this recipe?
A: You can use any vegetable that is in season.
b. Any athlete who wins the selection race will gain automatic selection.
(attested)
(40) Comma intonation: Non-contrastive and functional RCs pattern

with restrictive RCs in not having a comma intonation.
(41) Variable binding into the RC:non-contrastive and functional RCs pattern
with restrictive RCs in allowing binding into the RC:
a. Everyonei used the vegetables that grew in hisi garden.
b. (After the creative writing workshop ) Everyonei bought the book
that hisi child wrote.
c. Everyonei bought the first book that hei wrote.
(i) Ik woon in Amsterdam, dat 750000 inwoners heeft, waar bovendien vele
I live in Amsterdam, which 750000 inhabitants has, where moreover many
toeristen komen. (Dutch)
tourists come (ex 67c in Vries 2002, 197)
However, this argument is not conclusive since the difference may be due to the presence
of bovendien moreover, as adding moreover also makes (37b) acceptable (I thank Matthew
Baerman for pointing this out to me).
Notice that functional head nouns do not readily allow other determin-
ers than the definite determiner. The examples in (42b/c) suggest, however,
that this is a general property of functional head nouns, independently of
RCs:
(42) a. #Any first book that Saki wrote / #Any first book that an author writes
b. #Any first book of an author vs. The first book of any author
c. #any middle of a circle vs. the middle of any circle
Summarising: I will consider the following subtypes of non-contrastive restric-

tive RCs in what follows:18
18 The two types of restrictive RCs in (43a,bii) correspond to different patterns of definite-
ness marking in Haitian Creole as observed in Zribi-Hertz and Glaude (2007) (see Wespel
(2008:131136) for discussion):
(i) a. N + RC + det (establishing RC)
In these examples the referent can be identified by the addressee, if the descriptive
content of the noun and the RC is taken into account:
Moun (ke) ou te rele a te la. (Haitian Creole)
person rel you pst call def pst there
The person that you called was there. (Wespels ex. 84, taken from Damoiseau
2005:46)
b. N + RC (functional head noun)
Moun ki te envante dinat se te Charles Babbage.
man rel pst invent computer cop pst C B (E.F.5.7.18b)
The man who invented the computer was Charles Babbage. (Wespels ex. 90)
The missing determination pattern corresponds to appositive RCs for which the referent of
the head noun is given in the discourse situation, but not the property attributed to it in the
RC:
(ii) N + det + RC
Samdi dnye, manman m te kit pmdet. Pmdet yo, ke papa m
Saturday last mother my pst cook potato potato pl rel father my
te achte mache, te tr bon. (Haitian Creole)
pst buy market pst very good
Last Saturday, my mother cooked potatoes. The potatoes, which my father had bought
at the market, were very good. (Wespels ex. 81)
(43) Non-contrastive restrictive RCs

a. Establishing RCs (first mention definites Hawkins 1978)
b. Functional restrictive RCs
i. Inferred
ii. Functional head noun (Ebert 1970)
c. Maximalising RCs (Grosu and Landman 1998)
3.5 Semantic Factors in Article Choice with Relative Clauses

Above I have shown in detail that restrictive RCs cannot be equated with
contrastive restrictive RCs as is implicitly assumed in some studies. However,
the distinction between contrastive restrictive, non-contrastive restrictive and
functional restrictive relatives, discussed above, does not suffice to describe
definite article choice with RCs. Here I will point out a further semantic factor
that has been shown to influence the choice of definite article with restrictive
RCs.
3.5.1 Reference to an Intensional Class

In her study of article choice with RCs in Fehring Frisian, Ebert (1971:136150)
shows that the choice of definite article can mark a difference in interpretation
which she interprets as a choice between specific and non-specific reference to
a class (Ebert 1971: 147148):19
(44) a. i. a mensken wat suart hiar haa (detred) (FFr)

ii. dn mensken wat suart hiar haa (detfull)
detred / detfull people.pl that black hair have
(i) people who have black hair
(ii) those people that have black hair20
19 Ebert (1971:147): Der Unterschied zwischen spezifischer und nicht-spezifischer Referenz auf
eine Klasse kann also nur in der Relativsatzkonstruktion signalisiert werden. The difference
between specific and non-specific reference to a class can only be signalled in a RC construc-
tion. [my translation]
20 Eberts original translations are
(i) alle mglichen Menschen, die schwarzhaarig sind
all possible people that have black hair
(ii)genau die Menschen, die schwarzhaarig sind
exactly those people that have black hair
b. i. a lidj wat fr frihaaid stridj (detred)

ii. dn lidj wat fr frihaaid stridj (detfull)
detred / detfull people.pl who for freedom fight
(i) the people who fight for freedom (non-specific)
(ii) the people who are fighting for freedom (specific)
Eberts observation that the reduced definite article is used for intensional ref-
erence21 to a class is confirmed by the data discussed in Gunkel (2007). I will
avoid Eberts terms (non-)specific since conflicting definitions of these terms
have arisen since (see von Heusinger (2002) for discussion), and I will refer to
the two uses discussed by Gunkel as atemporal and anchored intensional
RCs. Gunkel shows that intensional RCs, be they atemporal or anchored/epi-
sodic, do not allow the determiner jener on the head noun while they allow the
forms of the determiner derjenige (45/46):22
(45) Atemporal intensional RCs

a. Nach den Spielregeln gewinnt derjenige/ #jener (Spieler),
after the rules-of-play wins.3sg derjenige/ 3sg.jener player
der am Ende die meisten Punkte hat.
that at-the end the most points has.
According to the rules, the player who has the most points at the end,
wins.
b. Nach den Wettkampfregeln erreichen diejenigen/#jene

after the rules-of-competition gets-into.3sg. derjenige/ 3sg.jener
21 Here I use intensional reference to mean a description of a class that includes potential as well
as actual members, and allows counterfactual reasoning.
22 Gunkel defines defining RCs (definitorische Relativstze) as follows
(i) definitorische Relativstze haben definitorischen Charakter; sie sorgen dafr, das vom
Sprecher anvisierte Referenzobjekt im Diskursuniversum zu etablieren.
Defining RCs have defining character: they establish the reference object intended by the
speaker in the universe of discourse.
The examples Gunkel considers as defining RCs correspond to what is termed atemporal
intensional RCs here. Temporally-anchored intensional RCs are not considered as defining
RCs by Gunkel: he treats them as episodic sentences with non-specific reference for the
relativised DP (Gunkel 2007:220).
(Teilnehmer) die nchste Runde, die mehr als 1000 Punkte

participant the next round that more than 1000 points
haben
has.
According to the rules, the participants that have more than 1000
points get into the next round.
(46) Anchored intensional RCs

a. Diejenigen / #Jene (Kandidaten), die in der nchsten
3sg.derjenige / 3sg.jener candidates that in the next
Runde ausscheiden, bekommen noch eine zweite Chance.
round are-eliminated get prt a second chance
The candidates that are eliminated in the second round still get a
second chance.
b. Derjenige / #Jener (Kandidat), der in der nchsten

3sg.derjenige / 3sg.jener candidate that in the next
Runde ausscheidet, bekommt noch eine zweite Chance.
round is-eliminated gets prt a second chance
The candidate that is eliminated in the second round gets a second
chance.
Gunkel (2007:232) points out that the contrast between intensional and
non-intensional DP+RC combinations also bears on the choice of definite
article in Swedish. The enclitic article is obligatory with appositive RCs.
With restrictive RCs, a more complex pattern emerges: the enclitic article is
optional with restrictive RCs if the overall DP is temporally anchored to a
particular, fixed day (47a) and dispreferred if the overall DP+RC is atempo-
ral (47b/48) (Gunkel cites Delsing (1993:119), Holmberg (1987, 1993) and Julien
(2005), see also Dahl (1978)). As Gunkel stresses, the contrast in (47a/b) is not
plausibly reducible to a difference between a restrictive and a non-restrictive
RC.
(47) a. Den dag-en jag mste sluta med mitt

dem day-def I had-to.prs/pst close.inf with my
pensionat. (Sw)
guesthouse
The day that I had to close my guesthouse
(past, temporally anchored particular day)
b. Den dag jag mste sluta med mitt pensionat

dem day I had-to.prs/pst close.inf with my guesthouse
The day that I have to close my guesthouse
(present/future, atemporal)
(examples from Perridon 1989:208, cited apud Gunkel 2007)
(48) Den spelare(?-n) som fr hg-st-a siffra-n brjar. (Sw)

dem player-def rel gets highest number-det begins
The player that gets the highest number begins. (atemporal)
(Gunkel 2007:233, ex 65a)
3.5.2 Relativising Verb+Noun Expressions

Another type of examples of RCs that are not appositive but favour a weak
determiner on the head noun are discussed in Studler (2008). In her study
of article use across Swiss German dialects, she (Studler 2008:105) points
out that examples like the following appear with weak definite determiners
more frequently than with other RCs, even though the RCs are not apposi-
tive:
(49) a. d Bedngke, wo si hnd (Swiss German)

detred doubts rel they have
the doubts that they have
b. s Wsse, wo d muesch haa

detred knowledge that you have to.have
the knowledge that you have to have
c. i de Phngkt, wo s druf aa chunt

in detred points where it on at comes
the point that is important
(Studler 2008:105, exs 46ac)
The common feature of these examples is that the verb and noun form a
complex expression, where the verb can be light as in (49a/b) but need not
be (49c). In some such examples, the RC can contribute the arguments of the
noun (50a/b) or modify the abstract noun (compare (50c) with (49c) above):
(50) a. their doubts (compare 48a)

b. the impression that he makes on me
c. the important point (compare (48c))
Even though the RC in relativised V+N constructions is not appositive

since it is obligatorywe cannot simply classify all relativised V+N construc-
tions as restrictive RCs. While some examples are restrictive relatives (49a), at
least in a subset of cases the noun is not clearly referential (see e.g. (49c/50b)),
therefore the RC involved does not seem to restrict the reference of the noun.
Summarising, in this section I have argued that apart from contrastive re-
strictive RCs, four types of non-contrastive restrictive RC must be distin-
guished, with a further semantic distinction concerning atemporal or anchored
intensional DPs and relativised V+N constructions.
4 Types of Relative Clauses and Article Choice
In what follows, I use the classification of restrictive RCs arrived at above, as

well as relativised V+N constructions. I therefore examine the choice of definite
article in the following cases of restrictive RCs:
(51) a. Contrastive restrictive RCs

b. Establishing RCs (first mention definites Hawkins (1978))
c. Functional restrictive RCs
i. Inferred
ii. Functional head noun (Ebert 1970)
d. Maximalising RCs (Grosu and Landman 1998)
e. Intensional relativised DP
f. Relativised V+N constructions
4.1 The Vorarlberg Dialect

In the Vorarlberger dialect the choice of a full definite article on the head
noun is clearly correlated with a contrastive reading of the RC (the data on
the Vorarlberger dialect are due to Gerhard Schaden and taken from Cabredo
Hofherr and Schaden 2012a,b):
(52) a. D/ Dia frouw mit dera da Hans gescht

detred/ detfull woman with whom detred H. yesterday
ufs fscht ku isch, isch fo Neuseeland. (VAr)
to-the party come is is from NZ
The woman that Hans came to the party with yesterday is from
NZ.
b. Da / Dea maa wo sera bojler gflickt hot, hot gsejt,

detred / detfull man rel our boiler repaired has has said
dass mr d rohr usfrba lo sllan.
that we the flue sweep let should
The man who repaired our boiler said we should have the flue swept
out.
These examples contain establishing RCs and both definite articles are
possible. However, when the full article is used, the referent described by the
DP+RC is contrasted with another potential referent of the DP: in ex (52a/b)
there would have to be another woman/ man in the context who does not ful-
fill the description of the RC (contrastive restrictive reading of the RC).
Functional restrictive RCs take the reduced definite determiner in the Vor-
arlberger dialect:
(53) a. s/ ?des inschtitut wo i frhr gschaffat ho isch

detred/ detfull institute where I before worked have is
bessr organisiert gsi. (VAr)
better organised been
The institute where I worked before was better organised.
b. Schtimmt, abr s kino i deam mr Shrek gsea hond isch

true but detred cinema in rel.pr we S seen have is
no vil schlimmr.
prt much worse
True, but the cinema that we saw Shrek in was even worse.
(adapted from the discussion in Studler 2008:153, ex 43a)
c. Functional head N:
I hob gat s/ des easchte buach vom Musil gleasa.
I have just detred/ detfull first book of+detred Musil read
I just read the first book by Musil.
Recall from (12) that the pre- or post-verbal position of the relativised DP plays
a role in article choice, (the reason is not currently understood). With amount
relatives, we also observe that the position of the relativised DP plays a role for
the choice of article, with a semantic effect which can be characterised in terms
of contrast. In pre-V2 position, both articles are semantically equivalent (54a),
while in post-V2 position the full article implies a contrast with other quantities
of tea (54b):
(54) Amount relatives

a. Da / dea tee wo i dr kanne isch isch kolt. (VAr)
detred / detfull tea rel in detred tea-pot is is cold
The tea in the tea-pot is cold.
no meaning difference
b. Jetz isch da / dea tee wo i dr kanna isch, kolt.

now is detred / detfull tea rel in detred pot is, cold
strong D: contrastive interpretation [there is some tea elsewhere]
In examples with reference to a class, both articles are possible; I have not been
able to establish whether the meaning difference between an intensional and
an extensional reference to a class noted for Fehring Frisian by Ebert arises here
as well.
(55) a. d/ dia lt wo rouchat sind mojschtens uzfrida. (VAr)

detred/ detfull people rel smoke are mostly unhappy
People who smoke are mostly unhappy.
b. d/ dia vgl wo im heabscht in sda zinad

detred/ detfull birds rel in+detred autumn in south go
kummad im frajor widr.
come in+detred spring back
Birds which fly south in the fall come back in the spring.
(adapted from exs in Ebert 1971)
With intensional DPs, both articles are possible for both atemporal (56a) and
temporally-anchored readings (56b):
(56) a. Da/ dea spiler wo am mojschta punkt hot, deaf i dr

detred/ detfull player rel the most points has may in the
nkschta runda afanga.
next round begin
The player who has the most points can open the game in the next
round. (atemporal)
b. Da/ dea fahrer wo des renna gwinnt, wiat

detred/ detfull driver rel detfull race wins becomes
olympiasiegea.
olymic.champion
The driver who wins this race will become Olympic champion.
(temporally anchored)
Relativised V+N constructions are compatible with a reduced definite deter-

miner even when modified by a RC:
(57) a. es isch da punkt, uf dean as akut (VAr)

that is detred point on that it on-comes
that is the point that matters
b. da easchte idruck won i fo eam hob isch ganz guat

detred first impression rel I of him have is quite good
c. da/ dea punkt, uf dean as akut isch doch, dass er

detred/ detfull point on that it on-comes is prt that he
a arbat finda muass
a work find must
The point that matters is that he needs to find work.
The choice of definite article interacts with the position of the DP in pre- or
post-V2 position: in pre-V2 position, both articles are acceptable (57c), while in
post-V2 position, the weak article is preferred in (57a).
4.2 Standard German

German P+D amalgams also appear in all five RC contexts considered here:
(58) a. Establishing RC
Im Haus, das an unseren Garten angrenzt, wohnt ein
in+detred house rel.pr at our garden border lives a
lteres Ehepaar. (Ge)
older couple
An older couple lives in the house that borders on our garden.
b. Inferred functional RC
Im Hotel, in dem sie wohnen, gibt es ab 7
in+detred hotel in rel.pr they stay gives expl from 7
Uhr Frhstck.
oclock breakfast
In the hotel that they are staying in, breakfast is served from 7
oclock.
c. Functional head noun

Karen wird im ersten Buch, das jemals ber Britische
Karen is in+detred first book, rel.pr ever over British
Gospelmusik geschrieben wurde, erwhnt.
gospel-music written was mentioned
Karen is mentioned in the first book that was ever written on British
gospel music.
d. Amount RC
Wir sind begeistert vom Kaffee, den diese Maschine
we are thrilled by+detred coffee, rel.pr this machine
produziert.
produces
We are thrilled by the coffee that this machine makes.
e. Relativised V+N construction

Das ndert wenig am Eindruck, den das Epos als
that changes little at+detred impression rel.pr det epic as
Ganzes vermittelt.
whole gives
That hardly changes the impression that the epos makes as a whole.
4.3 Fehring Frisian

In the discussion above, we have seen that Fehring Frisian allows reduced
definite articles with certain types of restrictive RCs (59a/b) (see also Keenan
and Ebert 1973). According to Ebert (1971), however, establishing RCs license
the full definite article (59c).
(59) a. Det/ At iast buk, wat hi skrewen hee, docht niks. (FFr)
detfull/ detred first book rel he written has is-worth nothing.
b. Intensional atemporal relativised DPs: reduced definite article

i. a lidj wat fr frihaaid stridj (detred)
ii. dn lidj wat fr frihaaid stridj (detfull)
detred / detfull people.pl rel for freedom fight
(i) the people who fight for freedom (non-specific)
(ii) the people who are fighting for freedom (specific)
c. Establishing RCs: full definite article

Anne hee det hs keeft, wat Inna arewt hee.
Anne has detfull house bought rel Inna inherited has
Anne has bought the house that Inna inherited. (Ebert 1971:137, ex 23)
Notice, however, that the example in (59c) is not entirely conclusive since here
as in the other examples of this type given by Ebert (1971), the RC is extraposed
after the verb, and Ebert shows independently that in other contexts extrapo-
sition is not compatible with a reduced article. In the following examples, the
particle tj23 marks the right boundary of the main clause:
(60) a. A kning nategt a/ dn mensken, wat

detred king took-advantage detred/ detfull humans rel
aarem wiar, tj. (FFr)
poor were prt
b. A kning nategt *a/ dn mensken tj, wat
detred king took-advantage detred/ detfull humans prt rel
aarem wiar.
poor were
The king took advantage of the poor. (lit. the people that were poor).
5 Analysis
Analyses that claim an incompatibility between restrictive relatives and re-

duced definite determiners on the head noun are based on a binary classifi-
cation that distinguishes restrictive and appositive RCs.
In section 3, I have shown that if we define restrictive RCs as RCs standing as
restrictive modifiers, this claim cannot be maintained. I have shown that in the
literature the term restrictive RC is narrowed down to only include contrastive
restrictive relatives. As shown above, however, we have to acknowledge the
existence of non-contrastive restrictive RCs, which cannot be reduced to the
23 This particle is part of a particle verb nategt tj uses out (cf. German ausnutzen). In Main
clauses only the verbal part (nategt in the present ex.) moves to the V2 position, with
the particle (tj) staying in clause-final position. Any clause that follows the particle tj is
therefore outside the main clausal domain. In the present example, the RC following the
particle is therefore either right-extraposed or stranded by leftward movement of the nominal
head, depending on the analysis of clausal syntax in Germanic that is adopted.
third type of RCs called maximalising or amount relatives (see Grosu and
Landman 1998; Alexiadou et al. 2000; Bianchi 2000; Grosu 2002). I propose that
the generalisation concerning article choice with RCs is rather:
(61) Contrastive restrictive RCs are incompatible with a reduced definite D on

their head noun.
This claim is supported by the independent observation in Wiltschko (to ap-

pear) that the reduced definite determiner is not compatible with focus in the
RC:
(62) #Da Brieaftroga wos bei UNS austrogt is imma pnktli. (ABav)
detred mailman comp at us delivers is always on.time.
The mailman who delivers in OUR neighbourhood is always on time.
(Wiltschko to appear, ex 59)
Following, Wiltschko (2012), I assume that the reduced definite determiner

signals that the referent does not need to be introduced explicitly, but can be
construed from the discourse context. Wilschko points out that situationally
unique does not suffice to capture the conditions of occurrence for reduced
definite articles, since anaphoric and deictic DPs are also situationally unique.
Wiltschko proposes the following characterisation of the reduced determiner.
(63) a. The reduced determiner is used if the uniqueness of the discourse

referent does not need to be established.
b. The use of the reduced determiner does not depend on discourse
context.
As Wiltschko convincingly argues, this view requires a model that distinguishes

the discourse context (or conversational common ground) (in (64a)) from the
common background (in (64b)), as proposed in Krifka (2007):
(64) a. common background: i.e. the common ground independent of the

current conversation
b. the conversational common ground which is sensitive to and ma-
nipulated by the ongoing discourse
Contrary to Wiltschko, however, I do not consider that there is evidence for

a syntactic distinction among relativised DPs, correlating with the type of
definite determiner. I propose a uniform syntactic analysis of RCs with both
types of definite determiner. For concreteness, I assume that the RC is adjoined

to the complement of D, with D hosting the definite determiner.
(65) [DP D [NP NP [CP RC ]]

the tea that is left in the cup
In her own analysis, Wiltschko (2012, to appear) considers that non-appositive

RCs occurring with reduced determiners on their head noun as in (66) are not
restrictive RCs:
(66) Wiltschko (2012, ex 3) Context: the mailman who has been delivering mail
in the neighborhood for the last 10 years is retired. Everyone knows this
mailman. A and B have been living in this neighborhood. A tells B.
Wasst eh, da Briaftroga (wos bei uns austrogn hot) is jetz in
Know prt detred mailman comp at us delivered has is now in
Pension. (ABav)
retirement
You know, the mailman (who used to deliver our mail) is now retired.
In the classification proposed here, the examples discussed by Wiltschko in-

volve non-contrastive restrictive RCs with functional head nouns. Wiltschko
proposes that RCs like the one in (66) are not restrictive, but belong to a dif-
ferent type, namely descriptive RCs. This assumption builds on the analysis
proposed by Del Gobbo (2005) for Mandarin Chinese restrictive and descrip-
tive RCs.
I do not follow Wiltschko in this assumption, since according to Del Gobbo
(2005), Chinese descriptive RCs should be analysed as generic or i-level modi-
fiers, while Chinese restrictive RCs can be classified as deictic or s-level modi-
fiers. This means that Del Gobbos use of the term restrictive RC for Mandarin
refers to a sub-class of restrictive relatives, as generally assumed. In fact, the
property described by a restrictive RC can be temporary (s-level) or permanent
(i-level): in English, the RCs in (67) are both restrictive, even though the prop-
erty they express is i-level in (67a) and s-level in (67b).
(66) a. the children that have blue eyes / were born in summer24 (i-level)
b. the children that just came in / are running over there (s-level)
24 The intended reading is one where the restrictive modifier picks out a subgroup of children
out of a larger group, not the reading identifying a class of children, available for bare nouns:
children that were born in summer.
It has further been argued that reduced definite determiners are incompat-
ible with restrictive RCs on semantic grounds. Brugger and Prinzhorn (1996,
14), for example, claim that reduced definite determiners impose a condition
that the cardinality of the referent of the NP is 1 in the domain of discourse, a
condition they claim to be incompatible with Bachs (1974) observation that a
restrictive RC presupposes the existence of entities of which the description
given in the RC is not true (see (17) above). Wiltschko (2012) adopts a vari-
ant of this argument assuming a different characterisation of the semantics of
reduced definite deteminers. According to Wiltschko (2012), reduced definite
articles are incompatible with restrictive RCs, since the reduced article can only
be used if uniqueness of the referent does not need to be established.
This argument for a semantic incompatibility between restrictive RCs and
reduced definite determiners is not convincing, however. Semantically, the
composition of the article proceeds only once the restrictive RC has combined
with the head noun; hence, what the reduced determiner marks as unique is
not the referent of the head-noun but that of the N+RC complex, as shown in
(68a), without this leading to any semantic contradiction. This is clearly the
semantic composition arising from an example such as (68b):
(68) a. the house that Ina inherited

detred [house [that Ina inherited]]
b. Every house [that Ina inherited] was renovated.

Every [house [that Ina inherited]]
This view of semantic composition in RCs follows Partees analysis of RCs

(Partee 1973). Partee points out that the meaning of the boy is not part of the
relativised DP the boy that is running over there since the semantic composition
first combines head noun and RC, and only the resulting N+RC complex is
combined with the determiner.
Furthermore, as explicitly pointed out by Brugger and Prinzhorn (1996:14),
nouns modified by restrictive adjectives are compatible with the reduced def-
inite determiner. This shows that restrictive modification of the head noun is
not incompatible with the semantics of the reduced definite determiner.25
While I follow Wiltschkos proposal that the reduced determiner semanti-
cally signals recoverable uniqueness, I do not follow her in assuming a syntactic
25 According to Studler (2008), the reduced article is even preferred in such cases in many Swiss
German dialects that have two definite determiners.
difference between relativised DPs depending on the type of definite deter-

miner. Wiltschko (2012) proposes the following structural difference between
reduced and full determiners, where the context-variable is assumed to be asso-
ciated with nP:
(69) a. full definite determiners select an [nP[NP]] complement (Wiltschkos

24a)
b. reduced definite determiners select an [NP] complement (Wiltschkos
24b)
Wilschkos main evidence for a syntactic difference between full and reduced
definite determiners is based on their contrasting behaviours with respect to
extraction (due to Brugger and Prinzhorn 1996: 6):
(70) [Von wem]i host du [s / *des Possbild ti] nit

of whom have you detred/ *detfull passport.foto not
gsegn? (ABav)
seen
Whose did you not see the / *that passport picture of? (Wiltschko 2012
exs 25a/b)
However, I propose that the ill-formedness of the strong definite D in this

example should be attributed to the fact that strong definite Ds do not allow
binding /co-varying readings independently of extraction, as shown by the
following examples from the Vorarlberger dialect (see Cabredo Hofherr and
Schaden 2012b):26
(71) jedsmol wnn i a nje pfanna kouf, (VAr)

everytime when I a new pan buy,
26 Notice that with an overt emphatic element, the emphatic reading of the full determiner is
forced and binding becomes possible:
(i) strong definite D with overt emphatic element co-varying reading ok
lot mia dea faschissene schtil noch ujnr wucha ab. [ok: ok: ]
let me detfull bloody handle after one week off
(Everytime I buy a new pan), the bloody handle falls off after one week. (VAr)
A similar contrast exists for that driver / that bloody driver for some speakers of English (I
thank Anne Zribi-Hertz for pointing this out to me).
a. lot mr noch ujnr wucha da schtil ab. [ok: , ?*: ]

let me after one week detred handle off
b. strong definite D no co-varying reading for NP

lot mr noch ujnr wucha dea schtil ab. [*: , ok: ]
let me after one week detfull handle off
Every time I buy a new pan, the handle falls off after one week.
The following examples clearly show that languages distinguishing two defi-
nite determiners do not pattern uniformly with respect to article choice with
restrictive RCs. In the following examples from Bavarian (Helmut Weiss p.c.),
the full definite determiner is required with a larger range of RCs than in the
Vorarlberger dialect (cf. section 4.1).
(72) a. Establishing RC (Bav)

dMaria renoviert grod des Haus, des dInna
detred=Maria renovates just detfull house rel.pr detred=Inna
vo ihra Oma gerbd hod
of her granny inherited has
Mary is renovating the house that Inna inherited from her granny.
b. Inferred functional RCs

(was isn mitm Maral los?)
what is-prt with-detred Mary wrong?
Dea man, mit dems gesdan
detfull man with rel.pr yesterday
(What is wrong with Mary?) The man she went out with last night was
mean to her.
c. Amount RCs
Wie hosd du den Tee, der noa in da kann war, so
how have you detfull tea that still in detred tea-pot was so
schnei tringa kind?
quickly drink could
How could you drink the tea that was in the tea-pot so quickly?
d. V+N relatives
dea punkt, auf den (wo)s ankimd
detfull point on rel.pr where=expl on-comes
the point that matters (Helmut Weiss, p.c.)
There is an independent factor however, that complicates direct comparison

of relativised DPs in Vorarlberger and Bavarian. Bavarian, like Austro-Bavarian,
has three possibilities for introducing a RC (Brugger and Prinzhorn 1996): []
a d-pronoun can optionally be followed by a wh-pronoun or the complemen-
tizer is realized only by a wh-pronoun. Notice that Wiltschkos example (62)
involves a RC introduced by wo, (lit. where) while the examples in (72) involve
RCs introduced by a d-pronoun.
Future research will have to establish what the semantic contribution of
each of the different RC types (der / der wo / wo) in Austro-Bavarian is.27
6 Conclusion
I have shown that only contrastive restrictive relatives are incompatible with
reduced determiners in all the cases considered in this study.
The impression of complementarity of article choice with restrictive RCs
mainly comes from the fact that non-contrastive RCs are systematically mis-
classified as non-restrictive. The observation that article choice correlates with
RC type is thus partially accurate and partially circular.
Once non-contrastive restrictive RCs are taken into account, however, the
impression of clear complementarity disappears.
I have proposed a classification of RC-types that have been shown to influ-
ence article choice. On the basis of observations from Ebert (1970) and Gunkel
(2007), I identify temporal anchoring as a further potential factor in the choice
of definite determiner.
I argue that the arguments supporting a syntactic distinction among DPs
correlating with the type of definite determiner are not convincing, and con-
clude (contra Wiltschko) that a uniform syntactic representation should be
preferred. However, Wiltschkos syntactic assumptions with respect to RCs are
not crucial to her semantic analysis of reduced definite determiners. In particu-
lar, the present analysis adopts this authors valuable insight that in the analysis
of reduced and full definite determiners, two tiers of common ground have to
be distinguished, as in the theory put forward in Krifka (2007).
27 Notice that a similar contrast between RCs introduced by a d-pronoun and those introduced
by a wh-relativizer exists in Fehring, see Ebert (1971).
References
Alexiadou, Artemis, Paul Law, Andr Meinunger, and Chris Wilder. 2000. The syn-
tax of relative clausesIntroduction. In The Syntax of Relative Clauses, edited by
Artemis Alexiadou, Paul Law, Andr Meinunger, and Chris Wilder, 152. Amster-
dam/ Philadelphia: John Benjamins.
Bach, Emmon. 1974. Syntactic theory. New York: Holt, Rinehart and Winston.
Bianchi, Valentina. 2000. The Raising Analysis of Relative Clauses: a reply to Borsley.
Slavistische Linguistik 2002, edited by Marion Krause, and Christian Sappok, 957.
Brugger, Gerhard, and Martin Prinzhorn. 1996. Some properties of German determin-
ers. ms.
Cabredo Hofherr, Patricia. 2012a. Preposition-determiner amalgams in German and
French at the syntax-morphology interface. In Comparative Germanic Syntax: The
State of the Art, edited by Peter Ackema, Rhona Alcorn, Caroline Heycock, Dany
Jaspers, Jeroen van Craenenbroeck, and Guido Vanden Wyngaerd, 99131. Amster-
dam/Philadelphia: John Benjamins,
Cabredo Hofherr, Patricia 2012b. Verschmelzungsformen von Prposition und Artikel.
Deutsch und Franzsisch kontrastiv. In Deutsch im Sprachvergleich. Grammatische
Kontraste und Konvergenzen. Jahrbuch des Instituts fr deutsche Sprache 2011, edited
by Lutz Gunkel, and Gisela Zifonun, 217238. Berlin: de Gruyter.
Cabredo Hofherr, Patricia, and Gerhard Schaden. 2012a. Definite articles and restric-
tive relative clauses: evidence from the Vorarlberg dialect. Talk at the Workshop
Definite articles and restrictive modification, Paris, September 2012.
Cabredo Hofherr, Patricia, and Gerhard Schaden. 2012b. Restrictive relative clauses
and weak definite determiners: evidence from the Vorarlberg dialect. Talk
at the Rencontres dAutomne de Linguistique Formelle (RALFe), Paris, November
2012.
Carlson, Gregory. 1977. Amount relatives. Language 53:520542.
Carlson, Gregory, and Rachel Sussman. 2005. Seemingly indefinite definites. In Lin-
guistic Evidence: Empirical, Theoretical, and Computational Perspectives, edited by
Stephan Kepser, and Marga Reis, 7186. Berlin: Mouton de Gruyter.
Carlson, Gregory, Rachel Sussman, Natalie Klein, and Michael Tanenhaus. 2008. Weak
Definite Noun Phrases. In Proceedings of NELS 36: 179196.
Cinque, Guglielmo. 2008. Two types of non-restrictive relatives. In Empirical issues
in Syntax and Semantics 7, edited by Olivier Bonami, and Patricia Cabredo Hofherr,
99137.
Del Gobbo, Francesca. 2005. Chinese relative clauses: restrictive, descriptive or appos-
itive? In Proceedings of the 30th Incontro di Gramtica Generativa, edited by Laura

Brug, Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano,
287305. Venezia: Cafoscarina.
Delsing, Lars-Olof. 1993. The internal structure of Noun Phrases in the Scandinavian
Languages. Ph.D. diss., University of Lund.
Ebert, Karen H. 1970. Zwei Formen des bestimmten Artikels. In Probleme und Fort-
schritte der Transformationsgrammatik, edited by Dieter Wunderlich, 159173. T-
bingen: Max Hueber Verlag,
Ebert, Karen H. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem
nordfriesischen Dialekt (Fehring). Brist/ Bredstedt: Nordfriisk Institut.
Ebert, Karen. 1973. Functions of relative clauses in reference acts. Linguistische Be-
richte 23:111.
Eisenberg, Peter, et al., editors. 2002. Duden. Die Grammatik. Mannheim, Wien, Zrich:
Dudenverlag. 7th edition.
Erteschik-Shir, Nomi. 1997. The dynamics of focus structure. Cambridge: Cambridge
University Press.
Fabricius-Hansen, Catherine. 2012. (Non-)Restrictivity from a discourse perspective.
Handout Semantic and Pragmatic Properties of (Non)Restrictivity, Stuttgart.
Grosu, Alex. 2002. Strange relatives at the interface of two millennia. Glot international
6:145167.
Grosu, Alex, and Fred Landman. 1998. Strange relatives of the third kind. Natural
Language Semantics 6:125170.
Gunkel, Lutz. 2007. Demonstrativa als Antezendentien von Relativstzen. Deutsche
Sprache 35:213238.
Hartmann, Dietrich. 1980. ber Verschmelzungen von Prposition und bestimmtem
Artikel. Zeitschrift fr Dialektologie und Linguistik 47:160183.
Hawkins, John. 1978. Definiteness and Indefiniteness: A study in reference and grammat-
icality prediction. London: Croom Helm.
Heinrichs, Heinrich Matthias. 1954. Studien zum bestimmten Artikel in den germani-
schen Sprachen. Giessen: Wilhelm Schmitz.
von Heusinger, Klaus. 2002. Specificity and definiteness in sentence and discourse
Holmberg, Anders. 1987. The structure of NP in Swedish. Working Papers in Scandina-
vian Syntax 33:123.
Holmberg, Anders. 1993. Two subject positions in IP in Mainland Scandinavian. Work-
ing Papers in Scandinavian Syntax 53:2941.
Jackendoff, Ray. 1977. X-bar syntax: A Study of Phrase Structure. Cambridge, MA: MIT
Press.
Julien, Marit. 2005. Nominal phrases from a Scandinavian perspective. Amsterdam/
Philadelphia: John Benjamins.
Keenan, Ed, and Karen H. Ebert. 1973. A note on marking transparency and opacity.
Kempson, Ruth. 2003. Nonrestrictive relatives and growth of Logical Form. In Proceed-
ings of WCCFL 22:301314.
Krifka, Manfred. 2007. Basic notions of information structure. In Interdisciplinary
Studies on Information Structure 6, edited by Caroline Fry, and Manfred Krifka,
1356. Potsdam: Universittsverlag.
Lbner, Sebastian. 2011. Concept Types and Determination. Journal of Semantics 28:
279333.
Martin, Fabienne. to appear. Decomposing (non-)restrictivity. Evaluative modifiers in
post-head positions. Lingua.
Nbling, Damaris. 2005. Von in die ber inn und ins bis im: Die Klitisierung von Pr-
positionen und Artikel als Grammatikbaustelle. In Grammatikalisierung im Deut-
schen, edited by Torsten Leuschner, Tanja Mortelmans, and Sarah de Groodt, 103131.
Berlin: de Gruyter.
Partee, Barbara. 1973. Some transformational extensions of Montague Grammar. Jour-
nal of Philosophical Logic 2:509534.
Puig-Waldmller, Estela. 2008. Contracted preposition-determiner forms in German:
semantics and pragmatics. Ph.D. diss., Universitat Pompeu Fabra, Barcelona.
Raffelsiefen, Renate. 1987. Verschmelzungsformen in German: a lexical analysis. Lin-
guistic Analysis 17: 123146.
Schiering, Ren. 2002. Klitisierung von Pronomina und Artikelformen. Eine empirische
Untersuchung am Beispiel des Ruhrdeutschen. Kln: Institut fr Sprachwissenschaft,
Universitt zu Kln.
Schwarz, Florian. 2009. Two types of definites in natural language. Ph.D. diss., UMass at
Amherst.
Studler, Rebekka. 2008. Artikelparadigmen Form, Funktion und syntaktisch-semantische
Analyse von definiten Determinierern im Schweizerdeutschen. Ph.D. diss., Universitt
Zrich.
Vries, Mark de. 2002. The Syntax of Relativization. Ph.D. diss., U. Amsterdam.
Vries, Mark de. 2006. The syntax of appositive relativisation: On specifying coordina-
tion, false free relatives, and promotion. Linguistic Inquiry 37:229270.
Weiss, Helmut. 1998. Syntax des Bairischen. Studien zur Grammatik einer natrlichen
Sprache. Tbingen: Niemeyer.
Wespel, Johannes. 2008. Descriptions and their domains. The patterns of definiteness
marking in French-related creole. Ph.D. diss. U. Stuttgart. Stuttgart: Working Papers
of the SFB 732 Incremental Specification in Context.
Wiltschko, Martina. 2012. What does it take to host a (restrictive) relative clause?
Working papers of the Linguistics Circle of the University of Victoria 21:100145.
Wiltschko, Martina. to appear. Descriptive relative clauses in Austro-Bavarian Ger-

man. Canadian Journal of Linguistics.
Zribi-Hertz, Anne, and Herby Glaude. 2007. Bare NPs and Deficient DPs in Haitian
and French: From Morphosyntax to Referent Construal. In Noun phrases in Creole
languages: a multi-faceted approach, edited by Marlyse Baptista, and Jacqueline
Guron, 265299. Amsterdam/ Philadephia: John Benjamins.
When Determiners Abound:
Implications for the Encoding of Definiteness*
Marika Lekakou and Kriszta Szendri
1 Introduction
The topic of this paper is the encoding of definiteness in Modern Greek. Mod-
ern Greek has a definite article, which at first sight seems to be performing the
regular function of a definite determiner, in terms of contributing semantic
definiteness. Definite noun phrases in Modern Greek obligatorily require the
definite article, as indicated in (1a).1 In fact, the determiner is required even on
proper names in argument position. This is shown in (1b):2
(1) a. *(O) kathijitis eftase protos.

the teacher arrived first
The teacher arrived first.
b. *(O) Janis ine kathijitis.

the John is teacher
John is a teacher.
The picture of definiteness in Modern Greek is, however, more complicated

than what the above facts would suggest. This becomes obvious when we
* We thank the audience of the Workshop on languages with and without determiners, the
anonymous reviewers and the editor for comments, questions and discussions that helped
us clarify the ideas presented here. All remaining errors are ours.
1 Bare noun phrases are possible in Greek, both singulars and plurals. For detailed recent
discussion, see Alexopoulou & Folli (2011). These authors argue that these NPs are arguments
and that they do not involve a null D. The interpretation of these nominals is not definite, and
according to Alexopoulou & Folli (op. cit.), it is also not identical to indefinite NPs preceded
by the numeral enas-mia-ena (one). We return briefly to bare NPs in section 4.
2 Modern Greek distinguishes grammatically between 3 genders (masculine, feminine and
neuter). The distinction is also reflected in the shape of the determiners. This fact is irrelevant
for our purposes and will be ignored in the discussion and the glossing of the examples.
Another property not reflected in our glossing is the case sharing inside the DP in Modern
Greek. For discussion of the role of case in connection to polydefinites, see Lekakou &
Szendri (2012).

when determiners abound 213
consider, in addition to more or less straightforward cases such as (1), the

phenomenon of determiner spreading or polydefinitenesss. Polydefinites are
cases where a noun is modified by an adjective, and noun and adjective are
each accompanied by a definite determiner, as illustrated in (2a) and (2b).
Polydefinites exist in Modern Greek alongside monadic definites, i.e. instances
of regular adjectival modification inside a definite nominal, illustrated in
(2c).3,4
(2) a. i asimenia i pena

the-fem.nom silver the-fem.nom pen
b. i pena i asimenia
the-fem.nom pen the-fem.nom silver
c. i asimenia pena
the-fem.nom silver pen
the silver pen
As we will see in detail in section 3, polydefinites are not semantically polydef-

inite. For example, the polydefinite in (2a) and (2b) refers to a single unique
entity, and in particular one at the intersection of the set of silver entities and
the set of pens. This means that in (2a/b), it cannot be the case that both deter-
miners make a semantic contribution. In other words, polydefinites are only
polydefinite in the morphosyntax, not in the semantics. Despite existing differ-
ences between monadic definites and polydefinites, reviewed briefly in section
2, in terms of definiteness the constructions are equivalent: they both contain
only one source of definiteness. Moreover, note that, morphologically, there is
no distinction between the multiple determiners in (2): both within the poly-
definite construction and across polydefinites and monadic definites the shape
of determiners is identical. This applies in all cases, i.e. across all case-number-
gender combinations. In other words, from the point of view of morphology,
we are dealing with one and the same element in all these instances.
Given these facts, the following questions are raised for Modern Greek: (a)
how is definiteness achieved in polydefinites, and (b) what is the nature of
definiteness more in general, in light of the polydefinite construction? What
3 The terms polydefinite and monadic definite are due to Kolliakou (2004).
4 Abbreviations used: acc= accusative, dial= dialectal, def= definite, fem= feminine, gen= gen-
itive, masc= masculine, neg= negative element, neut= neuter, nom= nominative, nonact=
non-active, pl= plural, sg= singular, stan= standard.
214 lekakou and szendri
enables the definite article in Modern Greek to occur in polydefinites as well

as monadic definites? These are the questions we focus on in this work. To the
best of our knowledge, the question of definiteness across polydefinites and
monadic definites has not been explicitly addressed in the existing literature.
The paper is structured in the following way. In section 2, we briefly present
the properties of the polydefinite construction and the analysis we assume for
it. In section 3 we turn to the implications of this analysis for the encoding
of definiteness in the language. We will argue that the semantic effects usu-
ally associated with definite determiners (e.g. existence and uniqueness asser-
tion/presupposition) are not achieved in Modern Greek through the overtly
realized definite article(s). The overtly realized determiners are merely reflexes
of a phonologically null operator that scopes over the Modern Greek DP and
contributes an iota operator (cf. Zeijlstra 2004 on the encoding of negation in
negative concord languages). In other words, the Modern Greek definite deter-
miner never makes a semantic contribution in terms of definiteness. In section
4 we address two potential problems for our proposed view of definiteness in
Modern Greek. In section 5 we conclude.
2 Polydefinites
2.1 The Core Properties

It has been well-established that polydefinites display a number of properties
not shared by their monadic counterparts (see in particular Kolliakou 2004;
Campos & Stavrou 2004; Alexiadou 2006). We discuss here the most important
properties of polydefinites, and briefly review the account we rely on to capture
these properties. For detailed exposition, see Lekakou & Szendri (2012).
First and foremost, the obvious fact about polydefinites is the multiple
occurrence of the definite determiner. Deriving this property is a far from trivial
task, as extensively demonstrated in Lekakou & Szendri (2012). (The question
of the interpretation of the multiple determiners will preoccupy us in the
following section.) Secondly, there is an ordering freedom in the polydefinite
construction, as seen in (2) above, which is not available in the monadic
definite. As (3) shows, monadic definites only allow the adjective in prenominal
position.5
5 The ordering freedom persists when more than one adjective is present, as discussed in
Androutsopoulou (1995), Alexiadou & Wilder (1998). All possible (six) word orders are accept-
able in those cases.
(3) a. i asimenia pena

the silver pen
the silver pen
b. *i pena asimenia
the pen silver
Thirdly, contrary to adjectives in monadic definites, adjectives in the poly-

definite construction are obligatorily interpreted restrictively. The example
in (4) from Kolliakou (2004) illustrates this: because as a matter of fact all
cobras are poisonous, the adjective dilitiriodis poisonous cannot be inter-
preted restrictively when applied to the noun kobres cobras, and therefore
determiner spreading is illicit. This restriction also entails that non-subsective
adjectives like former, see (4b), and non-intersective interpretations of oth-
erwise ambiguous adjectives like beautiful, see (4c), are unavailable in the
polydefinite construction (see also Campos & Stavrou 2004).
(4) a. Idame tis dilitiriodis (#tis) kobres.

saw.1pl the.acc poisonous the.acc cobras
We saw the poisonous cobras.
b. O proin (*o) proithipurgos pethane.

the.nom former the.nom prime minister died.3sg
The former prime minister died.
c. Ides tin orea ti xoreftria?

saw-3sg the.acc beautiful the.acc dancer
Did you see the beautiful dancer? (intersective reading only)
Finally, only the definite determiner may spread. There is no counterpart of

the polydefinite construction with indefinites (cf. Alexiadou & Wilder 1998,
Stavrou 2009, Velegrakis 2011):
(5) a. *mia pena mia asimenia

a pen a silver
b. *mia asimenia mia pena

a silver a pen
2.2 An Account in Terms of DP-Intersection

In a series of papers (Lekakou & Szendri 2007, 2009, 2012) we have highlighted
a parallel between polydefinites and close appositivesa parallel also noted
by Stavrou (1995); Kolliakou (2004); Panagiotidis & Marinis (2011)and have
proposed a unified account of both phenomena. Close appositives systemati-
cally pattern like polydefinites with respect to the properties identified in the
previous section. Consider the close appositive in (6) (from Stavrou 1995).
(6) a. o aetos to puli

the eagle the bird
b. to puli o aetos
the bird the eagle
the eagle that is a bird
O aetos to puli is a possible close appositive in Modern Greek, in virtue of the

fact that in this language the word for eagle is homophonous to the word for
kite. Using a close appositive helps disambiguate the intended referent of o
aetos.6
Like polydefinites, close appositives in Modern Greek allow multiple deter-
miners (unlike close appositives in e.g. English). Note that here too this con-
cerns the morphosyntax and not the semantics: in (6) reference is made to
a unique entity that is a member of the intersection of two sets. A second
shared property, as shown in (6), is the ordering freedom: both possible orders
are allowed in close appositives in Modern Greek. Moreover, like polydefinites,
close appositives involve a restrictive interpretation. As we saw in (6), one nom-
inal in the close appositive restricts the denotation of the other one. When
this is not possible, the close appositive is ill-formed. For instance, consider
the example in (7a) from Stavrou (1995), which involves a dialectal and the
standard Modern Greek word for the blueberry tree. It is impossible to form
a close appositive out of these two elements, because the two referents within
the whole appositive are identical. This makes it impossible for one subpart of
the appositive to restrict the other. The same effect can be observed, of course,
if the two items belong to the same dialectal variety, as in (7b):
6 For most speakers, polydefinites admit more than one adjective. Close appositives pattern
alike. For discussion of the iteration of the operation that we suggest derives both construc-
tions, see Lekakou & Szendri (2012) (especially section 3.1).
(7) a. *i sikaminja i murja (Stavrou 1995)

the bluberry treedial the blueberry treestand
b. *i sikaminja i sikaminja
the bluberry treedial the blueberry treestand
Finally, close appositives are also only possible with the definite determiner, as
observed by Stavrou (1995):
(8) a. *enas aetos ena puli

one eagle one bird
b. *ena puli enas aetos

one bird one eagle
Summing up, it turns out that, as exotic as they may seem from a cross-linguistic
perspective, polydefinites look much less alien from within Modern Greek:
close appositives share the core properties identified for polydefinites in the
previous section.
In the analysis of Lekakou & Szendri (2012), definite determiners head DPs
and thus polydefinites and close appositives are complex DPs consisting of
multiple DPs. This is illustrated in (9) on the next page (to be revised in the next
section).7 The only way in which polydefinites differ from close appositives is
that one of the two DPs contains noun ellipsis.8
The operation that combines the two DPs is identification of R(eferential)-
roles. We follow the relevant literature in assuming that the R-role is the exter-
nal thematic role of nouns and the element that enables nominals to refer
(Williams 1981, Zwarts 1993, Baker 2003). We follow Higginbotham (1985), who
first discussed identification between thematic roles in the context of attribu-
tive modification, in assuming that the interpretation of thematic identi-
7 See Lekakou & Szendri (2012) for discussion of the lack of evidence in favour of syntactic
asymmetry within the polydefinite, and for discussion of the lack of a unique head that
projects at the highest DP level in (9a) and (9b).
8 Noun ellipsis has several effects (e.g. ensuring that it is the adjectival DP that is restrictive
on the other one), which we cannot go into here. Some of these effects have been taken,
erroneously in our view, to argue for a FocusPhrase inside the DP. See Lekakou & Szendri
(2007, 2012), Szendri (2010) for extensive discussion of this question from both a theoretical
and an empirical perspective.
(9)
fication involving the R-role of nominals is tantamount to set intersection; this

seems reasonable, given that o aetos to puli is something that is both an eagle
and a bird. The operation of R-role identification is schematically illustrated in
(10):
(10)
The operation of R-role identification does not apply freely. It is restricted

by a ban against vacuous application (a ban which can be thought of as a
kind of economy principle): R-role identification applies only when its output
is not identical to (part of) its input. This derives the restrictive interpreta-
tion involved within the larger constituent (polydefinite/close appositive). For
detailed discussion of the operation and its restriction, see Lekakou & Szendri
(2012).
3 Expletive Determiners in Modern Greek
3.1 Determiners in Polydefinites, Monadic Definites and Proper Names

In the previous section we argued that polydefinites and close appositives alike
consist of DP subparts, and that between the two DP subparts set intersection
takes place. For this to be possible, it has to be the case that DPs denote sets
and not individuals (at least in Modern Greek). In other words, the determiner
heading each sister DP in polydefinites/close appositives has to be doing very
little semantic work, and in particular it has to not contribute an iota operator.
This is exactly what we think is going on in Modern Greek: the definite deter-
miner is expletive. This is a conclusion that is forced upon us independently
of the particular analysis that we are advocating, once we take a closer look at
the interpretation of the constructions at hand, and in particular of the definite
determiners within them.
For concreteness, consider the following example of a polydefinite from

Kolliakou (2004). The example contains an exchange between two speakers,
one of whom has been considering several objects as Christmas presents for
common friends. Among the candidates are a silver pen, a golden pen, and
a golden bracelet. The final decisions have been made, and the following
dialogue ensues:
(11) a. Speaker A: Ti pires tu Janni ja ta christujena?

what took.2sg the Jannis.gen for the Christmas
What did you get Jannis for Christmas?
b. Speaker B: (Tu pira) tin asimenia pena.

him.gen took.1sg the silver pen
(I got him) the silver pen.
c. Speaker A: Ti pires tis Marias?

what took.2sg the Maria.gen
What did you get for Maria?
d. Speaker B: (Tis pira) ti chrisi tin pena.

her.gen took.1sg the golden the pen
(I got her) the golden pen.
What interests us is the polydefinite ti chrisi tin pena the golden the pen in
(11d). The context is set up in such a way that there does not exist a unique pen,
but rather two pens. This means that the definite determiner on the noun in the
polydefinite in (11d) cannot be semantically contentful. In fact, given that the
adjective is always interpreted restrictively in the polydefinite construction, it
will always be the case that the noun-referent cannot be unique. So, at least one
determiner in the construction cannot be the one contributing semantic def-
initeness. What about the determiner on the adjective? The context contains
two golden entities. So it is not the case that the determiner on the adjective
is semantically real either.9 Since there is no unique pen in this context, nor is
there a unique golden entity, neither overt determiner can be responsible for
the semantic effect of uniqueness.
9 Since the noun in the polydefinite construction is not necessarily restrictive on a previously
mentioned noun, the polydefinite in (11d) would be felicitous even if the context included
only one golden entity.
If none of the overtly realized determiners is semantically contentful, where

does definiteness reside in the polydefinite? We propose that semantic defi-
niteness is contributed by a phonologically null operator, which is hosted in
a projection above the big DP of polydefinites and close appositives. We dub
this projection DefP, standing for Definiteness Phrase. It is in Def that the iota
operator, taking properties and returning individuals, resides. Overtly realized
D heads make no relevant semantic contribution; they simply encode the iden-
tity function (T,T). NP projections denote sets (type e,t), as is standardly
assumed. Our proposal is illustrated in (12) (which is a revised version of (9)
above):10
(12)
The kind of approach to definiteness that we are pursuing here has been pro-
posed by Zeijlstra (2004) for negation in strict negative concord languages (like
Modern Greek), where multiple negative elements do not cancel each other
out but contribute a single semantic negation. For these languages, it is argued
by Zeijlstra that overtly negative elements are not semantically negative, but
they simply mark the presence of a covert semantic negator in the clause.
10 In line with its minimal semantic content, the definite determiner in Modern Greek can
co-occur with the numeral (sometimes considered, erroneously in our view, as the indefinite
determiner), as well as with other quantificational elements, such as all, many and few.
The following examples illustrate this point. Example (i) is from Lekakou & Szendri (2012),
example (ii) is a Modern Greek proverb:
(i) O enas drastis sinelifthi.

the one perpetrator arrested.nonact.3sg
One of the perpetrators was arrested.
(ii) Ta pola loja ine ftoxia.

the many words are poverty
(lit.) Too many words is poverty.
With this as the analysis of definite determiners in polydefinites, what can

be said for monadic definites? We see no reason not to assume that what we
have just argued to be the case in the polydefinite is generalized to the monadic
case. No instance of the definite determiner in Modern Greek makes a semantic
contribution of definiteness. The source of semantic definiteness is always a
phonologically null element scoping over DP. The picture that emerges for
monadic definites is given in (13):
(13)
Is there any independent evidence for D being semantically expletive in Mod-

ern Greek, and for the concomitant Def-D split? The answer is positive. Recall
that proper names in Modern Greek obligatorily require the definite deter-
miner. The determiner is morphologically identical to the one accompanying
common nouns (contrary to e.g. the Catalan preproprial determiner):
(14) *(O) Janis ine kathijitis.

the John is teacher
John is a teacher.
In line with Kripke (1980) (and contra most recently Elbourne 2005 and Matu-
shansky 2009), we assume that proper names refer rigidly, and are thus of type
e. The determiner they combine with cannot be of type e,t,e, as that would
lead to a type mismatch. We need a determiner that has very minimal seman-
tic content, which is what we postulate for the Modern Greek determiner in
general. Since the Modern Greek definite determiner evidently can be seman-
tically inert, given its co-occurrence with proper names, it is best to assume that
it must be inert, i.e. that it is always inert, and that something else contributes
definiteness whenever that is the case.11 In terms of language acquisition, the
obligatory presence of articles on proper names is sufficient to trigger a split
Def-D structure in the language learners grammar.
11 For proper names, in other words, we do not assume that DefP is projected, since the
name is itself inherently definite. In this case, D marks the definiteness contributed by the
proper name. The Modern Greek definite determiner can thus combine with both predicate
nominals (i.e. common nouns) and with individual-denoting nominals (i.e. proper names).
This kind of flexibility is not unexpected, given the minimal semantics we assign to it.
To summarize, in this section we have provided a proposal for the encoding

of definiteness in Modern Greek that is consonant with the semantics of poly-
definites, namely with set intersection taking place among DP categories and
with the fact that neither determiner is real in the construction. We have pro-
posed that overt definite determiners in Modern Greek do not encode semantic
definiteness, but rather definiteness is due to the workings of a phonologically
null head. We take the obligatory presence of the definite determiner in proper
names as independent evidence for our proposal that the determiner is seman-
tically inert in Modern Greek. We extend our proposal to monadic definites,
which also employ a covert source of definiteness and a semantically exple-
tive D head. The three instances of the definite determinerwith monadic
definites, with polydefinites, and with proper namesare all given the same
treatment. This level of generality is not only desirable conceptually. It also
makes sense from an empirical, and in particular from a morphological point
of view: in all these guises, the determiner is morphologically one and the same
creature.
In what follows, we present a particular syntactic implementation of this
approach to definiteness, as well as more evidence in its favour, in terms of more
constructions with multiple determiners yet single referents. Before doing this,
however, it is worth contemplating what the alternative to the proposed view
of definiteness could be. It could be argued that something special is going on
with the D heads in polydefinites and close appositives (possibly, but not nec-
essarily linked to the obligatory presence of determiners with proper names),
but otherwise the definite determiner is generally semantically definite in Mod-
ern Greek. In brief, Modern Greek determiners are lexically ambiguous. This
has been, implicitly or explicitly, the prevalent view in the literature on poly-
definites: for most, if not all, existing treatments of polydefinites, the idea has
been that one determiner is a (semantically real) D head, and the other is
either semantically expletive (as in, presumably, the analyses of Alexiadou &
Wilder 1998, Kariaeva 2004) or realizes a distinct syntactic head (as in the anal-
yses of Androutsopoulou 1995, Campos & Stavrou 2004, or Ioannidou & den
Dikken 2009). However, the distribution of the alleged two types of determiner
within the polydefinite construction, and also in monadic definites, has to be
stipulated and even so, it is unclear that the observed semantic effects can be
accounted for.12
12 To be fair, most existing analyses of polydefinites do not aim at providing a semantic treat-
ment of the determiners, but at capturing the properties of the construction. However,
explaining the multiplicity of determiners in polydefinites is obviously linked to the issue of
Take for instance the analysis of Alexiadou and Wilder (1998), where polydef-
inites underlyingly involve a reduced relative clause. One determiner is exter-
nal to the relative clause structure, and an additional one is inside it, occupying
the subject position. This is shown in (15a). Predicate raising within the rela-
tive clause (which must be obligatory, as (15a) reflects an ungrammatical base
order) delivers one order of the polydefinite, cf. (15b), and raising of the relative-
clause-subject to the edge of the outer DP delivers the other order, as shown in
(15c).
(15) a. [DP the D [CP [IP [DP the book] [AP red]]]]
b. [DP the D [CP [AP red] [IP [DP the book] tAP]]]
predicate raising
c. [DP [DP the book] the D [CP [AP red] [IP tDP tAP]]]
DP-raising to SpecDP
Regarding the encoding of definiteness, it seems that what needs to be stip-

ulated is that the external D is the semantically real one and the internal
one is the expletive head. (This would perhaps be welcome on independent
grounds: it could be thought of as bringing the analysis more in line with
the original Kaynean analysis of relative clauses, which crucially featured NP
and not DP subjects inside the reduced relative clause.) This will work for
(15a) and (15b), but the scope of definiteness is not right in (15c), where the
external head is now situated too low. In sum, it is impossible for one and
the same D to be the real one across (15).13 It seems that an additional head
is required, which scopes over the whole construction. In that case, all D
heads in (15) will be semantically expletive. This is exactly what we have pro-
posed.
The biggest shortcoming of the alternative view briefly considered here, that
the Modern Greek definite determiner is systematically ambiguous between
their semantic contribution. In other words, it is important to evaluate the different syntactic
claims made also from the perspective of the theory of definiteness.
13 The problem is aggravated in the case of polydefinites that involve more than one adjective,
with a concomitant increase in determiners; which of the two adjectival ones would be the
real one in e.g. (i), and why?
(i) to podilato to kokino to kenurjo

the bicycle the red the new
the new red bicycle
a semantically expletive and a semantically real one, is that the postulated

ambiguity receives no independent justification in the language. Given the
complete morphological overlap between the alleged two sets, it is hard to see
how a child may successfully acquire them. This, in our view, constitutes a real
challenge for an alternative to what we have been pursuing here.
3.2 An Implementation in Terms of Definiteness Agreement

Polydefinites and close appositives constitute one kind of construction where
the morphosyntax and the semantics of definiteness ostensibly part ways:14 if
we are right, the locus of semantic definiteness in these cases is in one place,
but its morphosyntactic reflexes are elsewhere. This is precisely because D in
Modern Greek does not host material that is semantically definite. In some
sense, in other words, we are dealing with definiteness agreement (cf. Kariaeva
2004). We can think of Def as hosting a [+interpretable, +def] feature, and as
agreeing with one or more D head, which realize(s) the feature combination
[interpretable, +def]. We can formulate the following generalization as regu-
lating the distribution of these heads:
(16) Definiteness concord generalisation (DCG): In Def-D split languages,

any nominal element in the scope of a definite operator must be marked
for definiteness by the presence of the syntactic marker for definiteness,
D.
The generalisation is schematically represented as follows:
(17) *Def [D NP *(D) NP]
This says that in Def-D split languages, the presence of D marks definiteness
and the absence of D marks lack of definiteness. So, the latter (i.e. absence of
D) is not possible in the scope of Def. To give an example of how the DCG works,
consider polydefinites. Here, the locus of semantic definiteness in Def takes two
nominals in its scope. Given the DCG in (16), both must bear a definite article.
So, all the nominal elements in a polydefinite must be marked for definiteness,
i.e. bear a definite article.
In fact, the DCG applies in other structures as well. For instance, in pseu-
dopartitives (PsP), it has been independently acknowledged (see e.g. Alexi-
14 For monadic definites we have not included empirical, but only theoretical arguments in
favour of the same state of affairs holding.
adou, Haegeman & Stavrou 2007) that there is a single referent. In PsP, the more
substantial nominal is sometimes the second noun (N2) (which delivers the
so-called quantity reading), while other times the real head is the first noun
(N1) (which yields the so-called container reading). These two options are illus-
trated in (18) from English:
(18) a. The cup of sugar was strewn onto the floor. (quantity reading)
b. The cup of sugar smashed on the floor. (container reading)
Regardless of which N is the more substantial, the two nominals in a PsP do not
refer independently. Indeed, as expected if in languages like English or Dutch
the locus of semantic definiteness is the article itself, no article may occur on
N2 inside the construction. This is because the presence of the article would
turn NP2 into an independently referring nominal and the construction would
no longer be pseudo-partitive.
(19) a. the bottle of (*the/*my) wine

b. de fles (*de) wijn
If the DCG is correct, we expect determiner spreading to show up in pseudo-

partitives in Modern Greek. This is indeed the case as Alexiadou et al. (2007)
note:
(20) To bukali *(to) aroma epese ke espase. (container reading)

the bottle the perfume fell and broke
The bottle of perfume fell down and broke.
(21) To bukali *(to) aroma xithike sto patoma. (quantity reading)

the bottle the perfume spilled on.the floor
The bottle of perfume spilled on the floor.
We remain agnostic as to what the internal structure of pseudo-partitives is (see

Alexiadou et al 2007 for extensive discussion of the options). What is crucial
for any syntactic analysis adopted is that it reflects that the whole construction
picks out a single referent. In our terms, this means that the construction is in
the scope of a single reference-assigning head, D in English, or Def in Modern
Greek.
As Alexiadou et al (op. cit.), further note, determiner spreading occurs in
other domains too, such as PP modifiers like (22)(23) and even certain geni-
tives, (24)(25).
(22) O anthropos me *(ta) jalja bike sto katastima.

the person with the glasses entered in.the shop
The person with the glasses entered the shop.
(23) Enas anthropos me (*ta) jalja bike sto katastima.

a person with the glasses entered in.the shop
A person with glasses entered the shop.
(24) to sinolo *(ton) gramatikon katigorion

the set the.gen grammatical.gen categories.gen
the set of grammatical categories
(25) ena sinolo (*ton) gramatikon katigorion

a set the.gen grammatical.gen categories.gen
a set of grammatical categories
What unifies all these constructions, again, is that semantically there is a single
referent, although morphosyntactically we have multiple Ds.15 Our split Def-D
analysis, placing the locus of reference assignment in a position above D,
together with the DCG, account for all these cases. In all of these structures,
a single Def takes the whole construction in its scope (even if internally these
complex DPs that are complements to Def do not have identical structurewe
do not want to commit ourselves to a specific analysis of the constructions
above). This allows for an interpretation involving a single referent, and by the
DCG, will give rise to determiner spreading.16
15 The construction in (22)/(23) is the focus of Stavrou & Tsimpli (2009), who first make the
observation that agreement in terms of definiteness is required in this construction. These
authors also offer experimental support in favour of this generalization. However, they dis-
card the option that this kind of multiple definite marking is similar to that found in poly-
definites/close appositives, because in their view ta jalja the glasses in example (22) above
introduces a discourse referent. We disagree with this, and follow Danon (2008) instead, who
explicitly argues that glasses is a property-denoting noun in this case: the entire DP has a
single referent, the unique glasses-bearing individual. This aligns the construction with poly-
definites, close appositives and PsPs, in terms of definiteness.
16 Our approach in terms of the DCG has as its starting point the account proposed by Danon
(2008), who proposed something similar on the basis of Hebrew data such as the following:
(i) a. ha-seret al *(ha-)milxama lo matim le-yeladim.

the-movie about the-war neg suitable to-children
The movie about a/the war is not suitable for children.
4 Apparent Challenges
There are two ways in which our proposal could be shown to be inadequate:
one would involve arguing that the definite determiner does, in general, make
the relevant semantic contribution (even though it doesnt make it in poly-
definites). The other way would involve arguing that proper names should be
analysed differently and in particular more in line with definite descriptions;
this would undermine the rationale that proper names in Modern Greek can
provide independent evidence for the expletive nature of the determiner. We
discuss how each kind of counterargument could be constructed, and refuted,
in turn.
4.1 Joint and Split Readings under Coordination

Longobardi (1994) argued that the number of determiners in Italian equals the
number of referents in examples like (26a) and (26b), where subject agreement
on the verb tracks the number of referents. On the basis of examples such
as these, there seems to be a one-to-one correspondence between definite
determiners and referential expressions.
(26) a. La mia nuova efficiente segretaria e tua ottima

the my new efficient secretary and your excellent
collaboratrice *stanno/ sta uscendo
collaborator are/ is left
My new efficient secretary and your excellent collaborator has left.
b. La mia nuova efficiente segretaria e la tua ottima

the my new efficient secretary and the your excellent
collaboratrice stanno/ *sta uscendo
collaborator are/ is left
My new efficient secretary and your excellent collaborator have left.
b. seret al (ha-)milxama lo matim le-yeladim.

movie about the war neg suitable to-children
A movie about a/the war is not suitable for children.
Modern Greek DPs differ in a number of respects from Hebrew DPs (more limited distribution
of bare singulars in Modern Greek, lack of generic readings for bare singulars in Modern
Greek, etc). Moreover, there seem to exist some differences in the two paradigms of multiple
(in)definiteness, which we will not address here.
The pattern in (26) would receive a straightforward explanation if indeed

the D-head would be directly responsible for creating a referring expression.
However, Heycock & Zamparelli (2005) have shown that the situation is more
complicated than this. Conjunction of nominal phrases allows in principle
two different kinds of readings: a joint reading, where a unique (singular or
plural) individual instantiates different properties, and a split reading, where
multiple referents are being picked out. Many languages allow only a joint
reading when singular noun phrases are conjoined, but when conjunction
operates on plural noun phrases split readings become possible, even with
a single determiner present. In terms of the split-joint distinction, therefore,
(26a) involves a joint reading and (26b) a (trivial) split reading. Italian allows
a split reading with plurals, even if coordination takes place under a single
determiner. An example of this is given in (27), from Heycock & Zamparelli
(op. cit., ex (38)).
(27) a. I {numerosi/ pochi/ venti} generali americani e

the numerous/ few/ twenty generals americans and
diplomatici yugoslavi alla conferenza concordavano su
the diplomats Yugoslavian at-the conference agreed on
un solo punto.
a single point.
The numerous/few/20 American generals and Yugoslavian diplomats
at the conference agreed on a single point.
b. {Molti / Vari / Parecchi} amici di Carlo e parenti di

many / various / several friends of Carlo and relatives of
Francesca si incontrarono per la prima volta al
Francesca refl met for the first time at
matrimonio.
the wedding.
Many/various/several friends of Carlo and relatives of Francesca met
for the first time at the wedding.
So, the number of determiners does not directly correlate with reference in the
noun phrases; this holds for a number of languages including Italian, French,
Spanish and German, at least for plurals. Some languages even allow split
readings for singular cases: English, Dutch, and Finnish are such languages.
So, cross-linguistically, it does not seem to be the case that the right way to
analyse the unavailability of split readings under the determiner is due to the
unavailability of referring expressions under the determiner.
To account for the split and joint readings without postulating a cross-
linguistically lexically ambiguous coordinator, Heycock & Zamparelli (op. cit)
put forward an account, which is technically based on the idea that the coor-
dinator gives rise to set product. Without going into the technicalities of the
proposal, the direct consequence of this account is that when the coordina-
tor applies to predicative categories, it will mimic the operation of set inter-
section. So, joint readings arise. This is how we can account for examples
like My [best friend and colleague] is sitting next to the director. The coor-
dinator applying to the predicative nominals delivers (in a technically non-
trivial way) a meaning where the individual in question must have both the
property of being a friend and a colleague. In contrast, when the coordinator
applies to (sets of) individuals, it will create a set product based on the sets
corresponding to the two denotations of the conjuncts. Thus, split readings
arise.
Let us now turn to the corresponding Modern Greek data. Adapting Lon-
gobardis (1994:620) Italian examples for Modern Greek, Alexiadou et al. (2007:
6768) argue that the number of determiners equals the number of referents in
examples like (28a) and (28b). So, Modern Greek, like the Romance languages,
does not allow singular split readings under the definite article.
(28) a. Irthe/ *irthan o andiprosopos tis dikastikis arxis ke

came.3sg/ came.3pl the delegate the.gen court and
proedros tis eforeftikis epitropis.
chair the.gen elective committee
The representative of the court and chair of the elective committee
has arrived.
b. Irthan/ *irthe o andiprosopos tis dikastikis arxis ke

came.3pl/ came.3sg the delegate the.gen court and
o proedros tis eforeftikis epitropis.
the chair the.gen elective committee
The representative of the court and the chair of the elective committee
have arrived.
In our terms, (28a) involves co-ordination of NPs, i.e. below the Def-D struc-
ture, whereas (18b) involves co-ordination of two DefPs. It comes as no sur-
prise that the former involves a joint reading and the latter a split reading.
What is interesting about Modern Greek is that it seems to be quite unique
in completely disallowing split readings under the definite determiner, i.e. also
with co-ordination of plural nominals. This has been acknowledged (but not
accounted for) in the relevant literature (Heycock & Zamparelli 2005; King and
Dalrymple 2004). Example (29) illustrates this state of affairs.17
(29) a. *i gates ke kotes

the-pl cats and chickens (Heycock & Zamparelli 2005: ex 116a)
b. *I fili ke exthri tu Jani simfonisan se ena simio.

the-pl friends and enemies the.gen John agreed.3pl on one point
In Modern Greek, split readings can only apply if coordination takes place at
the highest level, among DefPs. Lower in the structure, coordination leads to
joint readings. Recall that Heycock & Zamparelli proposed that coordination
is uniformly set product applying to sets of individuals. But when it applies
to predicative categories, it mimics set intersection. Recall also that on our
proposal, Modern Greek DPs do not denote individuals, but predicates. Indi-
viduals are only available if DefP is present. Thus, we predict that for split
readings to obtain, coordination should only apply to DefPs. This explains the
unavailability of split readings for plural noun phrases under a single deter-
miner, i.e. (29). Singular split readings, such as (28a) are also excluded in the
same way.18
So, on our proposal the lack of split readings in the case of plural defi-
nite coordination in Modern Greek is easily accommodated. This is important
because previous analyses have not been able to account for this (Heycock &
Zamparelli 2005; King & Dalrymple 2004). But, unfortunately, this cannot be
the whole story. This is because even though Greek does not allow split readings
under the definite determiner, it has been observed that it allows split readings
with what Heycock & Zamparelli (op. cit.) call vague adjectival numerals, i.e.
expressions like several, (a) few, some, etc. Compare the grammatical exam-
ples in (30), which contain such expressions, to the corresponding ungrammat-
ical ones with the definite determiner in (29).
17 The examples in (29) can of course receive a (pragmatically unlikely) joint reading.
18 One may reasonably wonder whether co-ordination at the DP-level, below a single DefP pro-
jection, is possible, i.e. [DEFP Def [&P DP & DP]]. This would be similar to polydefinites/close
appositives, except for the presence of conjunction. The expected interpretation, given our
semantic treatment of DPs in Modern Greek as predicates, is a joint one. However, such
examples are not possible. Presumably, such a construction is blocked by the availability of
conjunction at the NP level, which produces the same effect. A similar filter is used in Hey-
cock & Zamparelli (2005: 244, ex 101) to account for the lack of a split reading of conjoined
singular nouns in Italian.
(30) a. merikes gates ke kotes

some-pl cats and chickens
(adapted from Heycock and Zamparelli 2005: ex 116b)
b. Meriki fili ke exthri tu Jani simfonisan se ena simio.

some friends and enemies the.gen John agreed on one point
Some of Johns friends and enemies agreed on one thing.
Although we do not have a full explanation, we would like to suggest that the
key to understanding this data comes from understanding indefinites and in
particular bare nominals in Modern Greek in general, which is currently a mat-
ter of some controversy. According to Alexopoulou & Folli (2011), Modern Greek
bare nouns, singular and plural, are nominal arguments, albeit functionally
impoverished as compared to definite DPs: they are NumPs, and do not involve
a phonologically null (definite or indefinite) D head (see also Stavrou 2003 for
a similar analysis of partitives).19 However, the precise conditions that license
bare nouns (and especially bare singulars) in Modern Greek are very much
under investigation at the moment, as is the overall question of whether the rel-
evant data cannot be handled by semantic incorporation, along the lines of for
instance Espinal & McNally (2011), who have discussed bare singulars in Cata-
lan and Spanish. This question is addressed in Lazaridou-Chatzigoga (2011).20
If bare nouns in Modern Greek involve semantic (or pseudo-)incorporation,
they denote properties, and not individuals. In that case, something else, sit-
uated higher in the extended nominal structure, must be responsible for indi-
viduation. This is compatible with our analysis, which maintains a predicate
denotation of Modern Greek DPs. If, however, Num is responsible for individ-
uation in the nominal domain (and delivers argumenthood for bare nouns), as
argued by Alexopoulou & Folli, it can presumably also supply the plural indi-
viduals necessary for the split reading in examples like (29) and (30). However,
19 The claim advanced by Alexopoulou & Folli (op. cit.) is that D is not required to turn nominal
predicates into arguments/individuals in Greek, because in this language Number is doing
that work. Modern Greek is thus minimally different from Italian, in terms of the typology
proposed in Chierchia (1998): Greek Num is doing the work performed by Italian D.
20 See also Gehrke & Lekakou (2013) for an analysis of Modern Greek bare nouns in so-called
P-drop contexts (Ioannidou & den Dikken 2009, Terzi 2010) as involving incorporation. On
this analysis, at least some bare nouns in Modern Greek denote properties and not individu-
als. The landscape of Modern Greek bare nouns appears thus to be mixed, and in any event
constitutes an area that has only recently started to be systematically explored.
only (30), without a D layer, allows a split reading. Thus, it seems to us inevitable
that noun phrases involving vague numerals have a different syntax when they
occur bare than when they occur under a definite D. We leave an elaboration
of this issue for future research.
We thus claim that in Modern Greek definite noun phrases, coordination at a
level lower than Def would lead to set intersection. Only at the level that e-type
individuals are created, i.e. at the DefP-level, can we obtain a split reading by
set product. Hence the cross-linguistically unexpected unavailability of split
readings for plural coordinate noun phrases under a single determiner. In
addition, the syntax and semantics of indefinites must differ from that of
definites in ways that allow for the availability of split readings with indefinites
involving vague numerals.
4.2 The Denotation of Proper Names

We have been following the philosophical tradition that treats proper names as
e-type individuals. Thus, unlike common nouns, proper names are not predica-
tive. The obligatory determiner on proper names is thus semantically vacuous.
It is a syntactically necessary marker without a type shifting function.
However, as Dora Alexopoulou (p.c.) brought to our attention, Modern
Greek allows proper names in predicative positions. One such example is given
in (31). Here, the determiner must be absent:
(31) I Dora den ine (*i) Xristina, na vafi ke na

the Dora neg is the Christina subj paint-3sg and subj
stolizi pasxalina avga me tis ores.
decorate-3sg easter eggs with the hours
Dora is not like Christina, to spend hours painting and decorating Easter
eggs.
In fact, she gives the following minimal pairs. In (32a) Evropi Europe is used
predicatively, while (32b) is an identificational copular sentence.
(32) a. I Galia den ine Evropi.

the France not is Europe
France is not (like) Europe.
b. I Galia den ine i Evropi.

the France not is the Europe
France is not (the sum of) Europe.
As Alexopoulou points out, the predicative use of the proper name disal-
lows the presence of the article, while the identificational use requires it. Recall
that we assume that the denotation of proper names is type e. In the pred-
icative use of proper names we propose that an operator is present, for con-
creteness, Partees (1986) IDENT, taking individuals (type e) and lifting them
to the singleton set containing them (type e,t) or to the property of being
that entity (Partee 1986: 122). Possibly, this type-shifter competes syntactically
with the definite determiner, i.e. is merged directly with the proper name NP,
whence the lack of the definite determiner in (32). The obtained interpre-
tation is the right one. In (32a) the meaning is that France is not Europe-
like.
It turns out that proper names with this x-like meaning seem to behave
syntactically like common nouns (see Marmaridou 1989 for this observation
and a similar analysis). They can appear under the indefinite article as in
(33a) and they can even become definite descriptions as in (33b). Crucially, the
meaning of o Iudas the Judas in (33b) is the unique individual in the context
that has Judas-like properties, i.e. the traitor among us.
(33) a. O Nikos ine enas Iudas.

the Nikos is a Judas
Nikos is a Judas/traitor!
b. Irthe o Iudas tis pareas.

arrived-3sg the Judas the.gen company
The Judas of our company [the traitor among us] arrived.
This, we propose is derived by applying the operator IDENT to the proper name,
and then subsequently applying the Def operator:
(34) [DefP [DP the [IDENT Judas]]] = the unique individual in the context
with Judas-like properties
This is not the only possible way to account for the data. One may go the oppo-
site way and assume that the denotation of proper names is predicative and the
definite article is a type shifter that turns it into an e-type individual. So, (32a)
would simply be an example of a predicative use of the proper name, while
(32b) involves an e-type individual created by the iota operator associated with
the definite determiner. However, for such an analysis to take shape, we need
to look at specific proposals in the literature that advocate a predicative deno-
tation for proper names. One such proposal was put forward by Matushansky
(2009). The starting point of her analysis is the syntax and semantics of nam-
ing constructions. She notices that in many languages naming predicates select
small clauses, where the name itself acts as the predicate of the small clause,
while the named individual is the subject. Based on this she ascribes to the
proper name Alice of (35) the meaning in (36):21
(35) I baptized the girl Alice.
(36) [Alice] = x De. Re, n, t. R (x) (/lis/)

where n is a sort of the type e (a phonological string)
In this theory, proper names are two-place predicates, taking as arguments an

individual and a naming convention R (in (35) the matrix verb baptize specifies
the naming convention). It is easy to see that this analysis does not derive the
meaning of (32): the article-less proper name in (32) does not make reference
to the phonological form, but rather to the property of being Europe(-like). So,
even in a theory that treats proper names as predicates, the analysis of such
examples must involve an extra operator like IDENT.
Besides, there are discrepancies between proper names and common nouns,
which are unexpected under the view of the former as definite descriptions.
In identificational copular constructions, coordination between two definite
descriptions involving common nouns is well-formed. So is coordination of
bare common nouns under a single definite articlethe familiar joint
reading, so long as the properties denoted by the common nouns are closely
associated with each other (e.g. friend and colleague; secretary and collabora-
tor; etc.), see (37b). At the same time, proper names can only be coordinated
high. This is illustrated by (38) where the context is that several aliases identify
the same spy, Spiros Alexiou.
21 We find the semantics proposed by Matushansky (2009) correct for naming constructions.
But we have doubts that naming predicates would be the right source for the semantics of
proper names in general. It seems to us that naming constructions are special cases, where
indeed the phonological form of the name is salient. But outside presentational or naming
contexts the phonological string does not seem to be accessible. Compare:
(i) Zygismund took the parcel to the post office. #(Incidentally,) I LIKE names with three
syllables.
(ii)My new partner is called Zygismund. I LIKE names with three syllables.
(37) a. I Maria ine i gramateas ke i sinergatis mu.

the Maria is the secretary and the collaborator me-gen
Maria is my secretary and my collaborator.
b. I Maria ine i gramateas ke sinergatis mu.

the Maria is the secretary and collaborator me-gen
Maria is my secretary and collaborator.
(38) a. O Spiros Alexiou ine o Petros Dimitriou ke o

the Spiros Alexiou is the Petros Dimitrou and the
Alexis Nikolaou.
Alexis Nikol.
Spiros Alexiou is Petros Dimitriou and Alexis Nikolaou.
b. *O Spiros Alexiou ine o Petros Dimitriou ke Alexis Nikolaou.

the Spiros Alexiou is the Petros Dimitriou and Alexis Nikolaou.
If proper names have the same denotation as common nouns, the

discrepancy between (37b) and (38b) needs to be explained.
5 Conclusion
Relying on our analysis of polydefinites as an instance of close apposition, we

have proposed that the Modern Greek determiner is semantically expletive in
the sense that it does not contribute an iota operator. This solution has the
advantage of treating definiteness in monadic and polydefinite constructions
in a uniform way, and of not relying on ad hoc lexical ambiguity for the Modern
Greek definite determiner. After presenting our analysis of definiteness based
on our treatment of polydefinites, we widened the empirical coverage of our
proposal to include pseudo-partitives, PP-complements and agreeing genitives;
three constructions that share the characteristic of involving multiple nomi-
nal phrases corresponding to a single referent. Finally, we put our analysis of
definiteness in Modern Greek to the test by considering two potentially prob-
lematic data sets: one regarding the nature of the link between reference assign-
ment and the definite article (Longobardi 1994, Heycock & Zamparelli 2005);
the other concerning naming predicates (Matushansky 2009). We suggested
how the data can be handled while maintaining the approach to definiteness
in Modern Greek that we have pursued here.
References
Androutsopoulou, Antonia. 1995. The licensing of adjectival modification. In Proceed-

ings of WCCFL 14:1731.
Alexiadou, Artemis. 2006. On the cross-linguistic distribution of (in)definiteness
spreading. Talk presented at the LT Syntax Workshop, 8.-9.12.06, Universitt Kla-
genfurt.
Alexiadou, Artemis & Chris Wilder. 1998. Adjectival modification and multiple deter-
miners. In Possessors, Predicates and Movement in the DP, edited by A. Alexiadou &
C. Wilder, 303332. Amsterdam: John Benjamins.
Alexiadou, Artemis, Liliane M.V. Haegeman and Melita Stavrou 2007. Noun phrase in
the generative perspective. (Studies in Generative Grammar 71.) Berlin: Mouton de
Gruyter.
Alexopoulou, Dora & Raffaella Folli. 2011. Topic-strategies and the internal structure of
nominal arguments in Greek and Italian. Ms. University of Cambridge and Univer-
sity of Ulster.
Baker, Mark. 2003. Lexical Categories. Cambridge: Cambridge University Press.
Campos, Hector. & Melita Stavrou. 2004. Polydefinites in Greek and Aromanian. In
Balkan syntax and semantics, edited by Olga Tomic, 137173. Amsterdam: John Ben-
jamins.
Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language
Semantics 6: 339405.
Danon, Gabi. 2008. Definiteness agreement with PP modifiers. In Current Issues in
Generative Hebrew Linguistics, edited by Sharon Armon-Lotem, Gabi Danon and
Susan Rothstein, 137160. Amsterdam: John Benjamins.
Elbourne, Paul. 2005. Situations and individuals. Cambridge, MA: MIT Press.
Espinal, Maria-Theresa & Louise McNally. 2011. Bare nominals and incorporating verbs
in Spanish and Catalan. Journal of Linguistics 47: 87128.
Gehrke, Berit & Marika Lekakou. 2013. How to miss your preposition. Studies in Greek
Linguistics 33: 92106. Available at: http://ins.web.auth.gr/index.php?option=com
_content&view=article&id=554&Itemid=182&lang=en
Heycock, Caroline and Roberto Zamparelli. 2005. Friends and colleagues: Coordina-
tion, plurality, and the structure of DP. Natural Language Semantics 13: 201
270.
Higginbotham, James 1985. On Semantics. Linguistic Inquiry 16, 547594.
Ioannidou, Alexia & Marcel den Dikken. 2009. P-drop, D-drop, D-spread. In Proceedings
of the 2007 Worskhop in Greek Syntax and Semantics at MIT, edited by Claire Halpert,
Jeremy Hartmann and David Hill, 393408. MIT Working Papers in Linguistics 57.
Kariaeva, Natalia. 2004. Determiner spreading in Modern Greek: split-DP hypothesis.
Ms. Rutgers University.
King, Tracy Holloway and Mary Dalrymple 2004. Determiner Agreement and Noun
Conjunction. Journal of Linguistics 40(1), 69104.
Kripke, Saul. 1980. Naming and necessity. Blackwell, Oxford.
Kolliakou, Dimitra. 2004. Monadic definites and polydefinites: their form, meaning and
use. Journal of Linguistics 40:263333.
Lazaridou-Chatzigoga, Dimitra. 2011. The distribution and interpretation of bare singu-
lar count nouns in Greek. Presented at the Weak Referentiality Workshop, Utrecht
9 September 2011.
Lekakou, Marika & Kriszta Szendri. 2007. Eliding the noun in close apposition, or
Greek polydefinites revisited. UCL Working Papers 19:129154.
Lekakou, Marika & Kriszta Szendri. 2009. Close apposition with and without noun
ellipsis: an analysis of Greek polydefinites. In Proceedings of 29th Meeting of the Lin-
guistics Department of Aristotle University of Thessaloniki, edited by Melita Stavrou-
Sifaki, Despina Papadopoulou & Maria Theodoropoulou, 151166. Thessaloniki:
Manolis Triantafyllidis Foundation.
Lekakou, Marika & Kriszta Szendri. 2012. Polydefinites in Greek: ellipsis, close appo-
sition, and expletive determiners. Journal of Linguistics 48(1), 107149.
Longobardi, Giuseppe. 1994. Reference and proper names: a theory of N-movement in
syntax and Logical Form. Linguistic Inquiry 25:609665.
Marmaridou, Sophia. 1989. Proper names in communication. Journal of Linguistics
25:355372.
Matushansky, Ora. 2009. On the linguistic complexity of proper names. Linguistics and
Philosophy 31(5), 573627.
Panagiotidis, Phoevos. 2003. Empty nouns. Natural Language and Linguistic Theory 21:
381432.
Panagiotidis, Phoevos, & Thodoris Marinis. 2011. Determiner spreading as DP-predi-
cation. Studia Linguistica 65: 268298.
Partee, Barbara. 1986. Noun phrase interpretation and type-shifting principles. In Stud-
ies in Discourse Representation Theory and the Theory of Generalized Quantifiers,
Jeroen Groenendijk & Martin Stockhof, 115143. Dordrecht: Foris.
Stavrou, Melita. 1995. Epexegesis vs. apposition. Scientific Yearbook of the Classics
Department. Aristotle University of Thessaloniki.
Stavrou, Melita. 2003. Semi-lexical nouns, classifiers and the interpretation(s) of the
pseudopartitive construction. In From NP to DP, edited by M. Coene & Y. DHulst,
329354. Amsterdam: John Benjamins.
Stavrou, Melita. 2009. Postnominal adjectives in Greek indefinite noun phrases. To
appear In Functional Heads. Papers in honour of Guglielmo Cinque for his 60th anni-
versary, Laura Brug, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, Cecilia
Poletto.
Stavrou, Melita & Ianthi Tsimpli. 2009. Definite agreement in complex noun phrases.
In Proceedings of 29th Meeting of the Linguistics Department of Aristotle University

of Thessaloniki, edited by Melita Stavrou-Sifaki, Despina Papadopoulou & Maria
Theodoropoulou (eds.), 193206. Thessaloniki: Manolis Triantafyllidis Foundation.
Szendri, Kriszta 2010. A flexible approach to discourse-related word order variations
in the DP. Lingua 120(4): 864878.
Terzi, Arontho. 2010. On null spatial Ps and their arguments. Catalan Journal of Lin-
guistics 9: 167187.
Velegrakis, Nikolaos 2011. The syntax of Greek polydefinites. Doctoral dissertation, UCL.
Williams, Edwin. 1981. Argument structure and morphology. The Linguistic Review
1:81114.
Williams, Edwin. 1989. The anaphoric nature of theta-roles. Linguistic Inquiry 20:425
256.
Zeijlstra, Hedde. 2004. Sentential negation and negative concord. PhD Diss., University
of Amsterdam.
Zwarts, Joost. 1993. X-bar Syntax, S-bar Semantics. PhD Diss, University of Utrecht.
The Semantics and Syntax of
Japanese Adnominal Demonstratives
Makoto Kaneko
This paper proposes some new ideas about the semantics and syntax of the
Japanese adnominal demonstratives a-no, ko-no and so-no. Semantically it
is claimed that, while conveying familiarity by means of the demonstrative
prefixes a-, ko- and so-, they lack uniqueness or maximality, and that the
whole demonstrative phrase is existentially quantified; -no either marks parti-
tivity (without excluding maximality) in the deictic and anaphoric uses, orin
bridging usesserves to fill an argument slot (lexically encoded inside the
following NP or created contextually). This analysis is supported by (i) the
availability of sluicing, (ii) the distribution of numeral classifiers, and (iii) the
similar behavior of French partitive constructions with respect to the consis-
tency test. Syntactically, Japanese adnominal demonstratives are analyzed as
NP-adjuncts, an assumption supported by three morpho-syntactic properties:
(i) The demonstrative prefixes, ko-, so-, a- systematically display the same mor-
phology as that of the WH-prefix do-; (ii) The Japanese demonstratives may
be preceded by a restrictive modifier, like other adjunct modifiers; (iii) They
behave with respect to the ellipsis of the following NP as other no-marked
expressions clearly identified as adnominal adjuncts. These hypotheses further
shed light on some data from L2 acquisition.
1 Introduction
This paper* aims at clarifying the semantics and syntax of the Japanese adnom-
inal demonstratives a-no, ko-no and so-no, with respect to some recent research
on demonstratives and definite determiners. Although demonstratives are a
much-discussed topic in Japanese linguistics, very little is known about their
* I would like to thank Patricia Cabredo Hofherr and Tania Ionin for their valuable comments
on earlier versions of this paper. Special thanks are due to Anne Zribi-Hertz for her kind help
to improve both the style and content. I am responsible for all the remaining problems. This
research is partially supported by a grant from the Japan Society for the Promotion of Sciences
(No. 23520463). This paper is an attempt to extend the ideas previously advanced for so-no in
Kaneko (2012) to the other two Japanese demonstratives, by partially modifying the previous
analysis.

240 kaneko
semantic and syntactic peculiarities, in comparison with demonstratives and

definite determiners in other languages.
Recently, Bokovi (2009) put forward the assumption that determiner-like
expressions (possessives, demonstratives, etc.) in article-less languages, like
Japanese, are not structurally located within the DP projection, but stand as
modifiers adjoined to NP. As regards semantics, it is reported that demonstra-
tives in Salish languages (cf. Matthewson 1999) and a determiner in Malagasy
(cf. Paul 2009) lack at least one of the two features of definiteness (familiarity
and uniqueness / maximality).
Inspired by these previous studies, I will claim i) that while Japanese adnom-
inal demonstratives include familiarity in their semantics, they lack a unique-
ness or maximality presupposition, hence allowing partitive interpretation,
and ii) that their lack of uniqueness or maximality is syntactically captured if
we analyze them as forming a demonstrative phrase (DemP) occurring in NP-
adjoined position, whose specifier and head are respectively occupied by the
demonstrative prefixes ko/so/a, and by the stem -no.
The discussion is organized as follows. Section 2 presents preliminary re-
marks about the system of Japanese adnominal demonstratives and points out
their similarities with English definite articles and demonstratives. Section 3
describes their semantic peculiarities on the basis of Lbner (2011), and shows
similar phenomena in Salish languages and in Malagasy. Section 4 introduces
my own assumptions regarding the semantics and syntax of Japanese adnom-
inal demonstratives, in the wake of some recent proposals on demonstratives,
and assesses the validity of these assumptions with respect to some observa-
tions made in the field of L2 acquisition and from a cross-linguistic perspective.
Section 6 summarizes the main results of the paper.
2 Preliminary Remarks
As regards the differences between the three adnominal demonstratives a-no,

ko-no and so-no, Hoji et al. (2003, 115) remark that a ko-NP is marked as
[Proximal]; a a-NP is marked as [Distal], while a so-NP is neither [Proximal]
nor [Distal]; as for so-no, a linguistic antecedent is necessary (idem. 103),
which is not the case for a-no and ko-no. So-no may surely be deictically used,
like a-no and ko-no in (1a,b), to refer to something close to the hearer, as in (1c).1
1 In each example, a-no and ko-no are respectively translated by English that and this, and so-no,
by that if this translation is appropriate. Otherwise, the most appropriate English expression
the semantics and syntax of japanese adnominal demonstratives 241
(1) deictic
a. A-no hito-wa amerikazin desu. (Hoji et al. 2003, 108)
a-no person-top American cop
That person is an American. [pointing to someone standing 10 meters
away]
b. Ko-no hito-wa amerikazin desu. (ibid.)

ko-no person-top American cop
This person is an American. [pointing to someone the speaker has
his/her arm around]
c. Sumimasen ga, so-no hon-o totte kudasai.

excuse.me but so-no book-acc take please2
Excuse me, but could you get me that book (next to you)? (idem. 111)
Hoji et al. (2003) however argue that the deictic use of so-no is relevant only
when a conflict occurs between the speakers and hearers viewpoints such that
the speaker construes the relevant object as distal, and the speaker thinks that
the hearer would construe the relevant object as proximal (idem. 113). They
further suggest that a marked operation creates, on the basis of visual contact
with an object, what corresponds to a linguistic expression that can serve as
an antecedent [] and this is what underlies the deictic use of so-NPs (ibid.).
This analysis of deictic so-no as derived from its anaphoric use is confirmed
by Okazaki (2010), who shows that the anaphoric use of so- is observed in the
earliest stage of the history of Japanese, while its deictic use is developed only
in later stages.
Being essentially anaphoric, Japanese so-no shows some similarities with
both English the and that (those). The latter allows a co-variable anaphoric
use, as in (2a). English the is further acceptable in bridging contexts, as in
(2b). English that (those) requires high saliency of the referent in the relevant
context, and does not easily allow bridging uses, as in (2b) where the relevant
accompanist is not sufficiently salient. Wolter (2006) however observes that
will be adopted in each context. The NP following demonstratives is italicized, while possible
antecedents are underlined.
2 The abbreviations used in this paper are the following: ACC: accusative; AT: actor topic;
CL: classifier; COMP: complementizer; COP: copula; DAT: dative; DET: determiner; DIMIN:
diminutive; ERG: ergative; GEN: genitive; LOC: locative; M: masculine; NEG: negation; NOM:
nominative; PL: plural; PROG: progressive; PST: past; Q: question marker; SG: singular; TOP:
topic; TR: transitive; TT: theme topic.
242 kaneko
a bridging use of demonstratives is allowed in examples like (2c) where the

contrast between on reserve and in bookstore evokes the salient existence of
books associated with the antecedent the reserve section.
(2) a. co-variable anaphoric

Of every house in the area that was inspected, it was subsequently
reported that {the / that} house was suffering from subsidence
problems. (adapted from Lyons 1999, 32)
b. bridging
Every singer complained that {the / #that} accompanist played too
loudly. (idem. 273)
c. The girls received individualized reading lists with sections labeled
on reserve and in bookstore. Every girl went to the reserve section of
the library and read those books first. (Wolter 2006, 52)
The essentially deictic forms a-no and ko-no in Japanese do not ordinarily allow
these two readings,3 while so-no allows (i) co-variable anaphoric uses, as in
(3a), and (ii) bridging uses, as in (3b) where the referent of so-no ko-gaisya
subsidiary is identified through the associative relation with the quantified
antecedent, do-no zidoosya-gaisya every automobile-company.
(3) a. co-variable anaphoric

Do-no zidoosya-gaisya-mo so-no zidoosya-gaisya-no
which automobile-company- so-no automobile-company-gen
ko-gaisya-o suisensita (Hoji et al. 2003, 104)
subsidiary-acc recommended
Every automobile-company recommended one, some or all of that
automobile-companys subsidiaries.
b. bridging
Do-no zidoosya-gaisya-mo so-no ko-gaisya-o suisensita
which automobile-company- so-no subsidiary-acc recommended
Every automobile company recommended one, some or all of its sub-
sidiaries.
3 Hoji et al. (2003) however point out that ko-no may exceptionally have a covariant reading, as
does English this in (i).
(i) Every family who has a George thinks this George is genius.
But Japanese so-no (as well as a-no and ko-no) and English the and that
(and other definite determiners) display some fundamental differences, to be
discussed in the next section.
3 Semantic Differences between Japanese Demonstratives and

English Definite Determiners
3.1 Congruent and Incongruent Definite Determination

In order to elucidate differences between Japanese adnominal demonstratives
and English definite determiners, I refer to Lbners (2011) theory of definite-
ness. Lbner (2011) discusses the interaction between basic noun types and
determination types. He first distinguishes four basic noun types (sortal, indi-
vidual, relational and functional): (i) sortal nouns are unary predicate terms (of
type e,t) and include prototypical nouns, like man; (ii) individual nous are
individual terms (of type e) uniquely identified in a context of utterance, like
sun, US president, etc.; (iii) relational nouns are binary predicate terms (of type
e,e,t) and characterize their referents in terms of a particular relation (not
necessarily one-to-one) to some other object, like brother (brother of x is not
necessarily uniquely determined with respect to x); (iv) functional nouns are
unary function terms (of type e,e) involving the possessor argument whose
value constitutes the uniquely determined referent, like father ( father of x
is uniquely determined with respect to x). By representing inherent unique-
ness and inherent relationality respectively with the binary features [U] and
[R], Lbner (2011) characterizes (i) sortal nouns as [U][R], (ii) individual
nouns as [+U][R], (iii) relational nouns as [U][+R], (iv) functional nouns as
[+U][+R].
This author next shows which mode of determination is natural with each
type of common NPs. The singular definite determiner is natural with [+U]
NPs (i.e. individual and functional nouns). On the other hand, [U] NPs (i.e.
sortal and relational nouns) must be type-shifted through contextual enrich-
ment to be compatible with the singular definite. To generalize such natural
and forced associations between NP types and determination types, Lbner
introduces a distinction between congruent and incongruent determina-
tion: determination is congruent if it does not change the NP type; otherwise, it
is incongruent. A bridging use of the definite article associated with functional
nouns is both congruent and incongruent since these nouns are specified as
[+U] but their possessor argument should be implicitly fulfilled by means of
an anaphoric relation with the antecedent. Lbner further points out a gen-
eral tendency according to which incongruent determination receives more
244 kaneko
salient expression, such as strong v. weak marking, marking v. non-marking,

additional morphemes (p. 29) and proposes the following scale.
(4) incongruent definite determination (adapted from Lbner 2011, 42)

deictic with sortal or relational nouns
> anaphoric with sortal or relational nouns
> bridging with functional nouns
> individual nouns
congruent definite determination
In English, demonstratives may bear contrastive stress and are thus charac-
terized as incongruent definite markers, while the definite article the, which
cannot be stressed, is essentially a congruent definite marker, although it cov-
ers a wide array of uses ranging from the left to the right edge of the scale in
(4). In its deictic use, as in (5a), the definite article requires uniqueness of the
referent in the relevant situation, which is not the case for demonstratives: in
contexts where the positive and negative form of the same predicate apply to
the same subject (the consistency testLbner 2011, 15), demonstratives, but
not definite the, are acceptable, as in (5b,c).
(5) a. deictic uses

Dont go there, the dog will bite you. (Yang and Ionin 2009, 4)
b. This man is dumb and this man isnt. (Lbner 2011, 18)
c. *The man is dumb and the man isnt.
In anaphoric uses too, a difference is observed between demonstratives and the

definite article. As Wolter (2006) points out in (6), the antecedent of a definite
description is determined in the global domain of discourse, while that of a
demonstrative is determined in the local domain of the immediate context.
(6) Definite descriptions are interpreted relative to a default situation [asso-

ciated with discourse context reporting about a global discourse topic];
demonstrative determiners require that their descriptive content is inter-
preted relative to a non-default situation [immediate salient situation
distinct from the discourse context]
wolter 2006, 6364
Turning to Japanese demonstratives, essentially deictic a-no and ko-no are

basic incongruent markers. Essentially anaphoric so-no clearly covers a wider
range, since it takes on the bridging use, but should nevertheless be classified
among incongruent markers. In Japanese, congruent definite determination

is zero-marked. Thus, just like English that, so-no is not natural with [+U]
individual or functional nouns, as witnessed by (7a,b). In deictic uses, Japanese
so-no (like ko-no and a-no) doesnt induce, like English demonstratives and
contrary to English the, uniqueness of the referent in the relevant situation, as
in (8).
(7) Congruent definite

a. (#So-no) taiyoo-ga kagayai-teiru. [individual noun]
so-no sun-nom shine-prog
(#That/The) sun is shining. (adapted from Yang and Ionin 2009, 4)
b. (#So-no) watasitati-no ie-no yane-wa

so-no us-gen house-gen roof-top
amamori-teiru [functional noun]
leak-prog
(#That/The) roof of our house is leaking. (ibid.)
(8) Incongruent definite [deictic]

{So-no} otoko-wa manuke-da, sikasi so-no otoko-wa kasikoi.
so-no man-top dumb-cop but so-no man-nom clever
That man is dumb and that man is clever.
Table (9) summarizes the semantic domains covered by definite determin-

ers in English and by adnominal demonstratives and the zero form in Japa-
nese.
(9) [congruent] [+congruent]

deictic > anaphoric > bridging with functional > individual
a. this, that the [English]
b. a-no, ko-no so-no zero [Japanese]
Aside from the differences described above, Japanese adnominal demonstra-

tives and English definite determiners display another crucial contrast, dis-
cussed below.
3.2 Lack of Uniqueness / Maximality

First, in its bridging uses, English the is normally only possible with [+U]
functional nouns, and not with [U] relational nouns: (10a) is acceptable since
a one-to-one correspondence is established between the antecedent a truck,
246 kaneko
and the NP following the definite article, hood (a truck has only one hood),
which is not the case for the unacceptable (10b) (a truck has four hubcaps).
(10) a. I bought a truck. The hood was scratched. (Barker 2005, 93)
[functional]
b. I bought a truck. #The hubcap was scratched. (ibid.)
[relational]
Contrastively, bridging so-no4 may be attached not only to [+U] functional

nouns, e.g. hyoosi cover, associated to the antecedent zassi magazine in (11a)
(a magazine has only one cover), but also to [U] relational nouns, such as
tyosyo writing, associated to the antecedent sensee Professor in (11b) (a pro-
fessor may publish more than one writing). In the latter case, so-no allows a
non-maximal, partitive reading, as in (11b) where Bs reply meaning which writ-
ing? confirms that a unique piece of writing is not presupposed. Recall that
Japanese does not have obligatory plural marking, that Japanese nouns, such
as tyosyo writing, may convey either singular or plural readings, and that, as
observed by Nitta (1992, 597600),5 a-no / ko-no / so-no + non-human animate
or inanimate nouns may denote one or more than one referent.
(11) a. Boku-ga aidokusi-teiru zassi-ga atte [] kondo so-no

me-nom adore-prog magazine-nom exist, this time so-no
hyoosi-ni []
cover-loc
I adore a magazine, and this time, on its cover [functional]
(Iori 2007, 159)
b. A: Ko-no aida, gakkai-no kaizyoo-de sensee-ga so-no

Last day meeting-gen place-loc professor-nom so-no
tyosyo-ni me-o toosi-teorare-ta yo. [relational]
writing-dat eye-acc pass-prog-pst you know
4 Gerhard Schaden (p.c.) correctly points out that, contrary to bridging so-no, the bridging use
of the definite article cannot have its antecedent in the same clause. In view of Wolters (2006)
generalization in (6), Schadens observation may be due to the fact that the antecedent of
Japanese so-no, which is basically a demonstrative, is traced back in the local domain of
the immediate context, while that of the definite article is found in the global domain of a
discourse.
5 Nitta (1992, 597600) points out that for example, ko-no hon (ko-no book) and ko-no neko
(ko-no cat) may denote either singular or plural entities.
B: E, do-no tyosyo? (Iori 2007, 146)

Oh, which-gen writing
A: The other day, at the meeting, Professor was reading one or some
writing(s) of hisB: Oh, which writing?
First, plurality of animate nouns in Japanese may be emphasized by the suffix

-tati. But so-no + NP-tati does not necessarily induce maximality, as in (12),
where so-no, referring back to Professor Hata, is attached to the noun gakusee
student lexically type-shifted from sortal to relational, and so-no gakusee-tati
can refer to some or all of Prof Hatas students.6
(12) Hata kyoozyu to so-no gakusee-tati-wa [] KG broadband

Hata professor and so-no student-pl-top KG broadband
station-nituite happyoosimasu.
station-about give.a.talk.
(http://www.jearn.jp/2003conference/news/kwansei.html)
Professor Hata and some or all of his students will give a talk about KG
broadband station
Second, in anaphoric uses of so-no, the default interpretation is admittedly

a definite unique or maximal reading. However, an indefinite partitive read-
ing is not excluded. Thus, in (13), it is unlikely that Speaker A should buy all
of the previously introduced seven puppies. Consequently, the referent of so-
no koinu(-tati) so-no puppy (puppies), is not maximally identified with the
antecedent, as confirmed by Bs question meaning how many?.7
6 In (2c), which involves a bridging use of English those, a [R] sortal noun book is similarly
type-shifted to a [+R] relational noun. But unlike Japanese so-no, the use of those conveys
that every girl read all the books that she found in the reserve section.
7 In (13A), the plural form of so-no, sore-ra-no, is possible. But, as emphasized by Nakanishi and
Tomioka (2004), the Japanese plural markers -tati and -ra basically convey heterogeneous
plurality, where the individual members of the set are not uniform in nature. In the same
vein, Kobayakawa (2004, 42) observes that, when they denote plural referents, sono+NP
represents a group of entities conceived as belonging to the same category, while sore-
ra-no+NP represents a group of entities conceived as belonging to different subcategories
of the same category. As regards (13A), if the speaker regards the relevant puppies as uni-
form, she makes use of so-no, whereas if she wants to emphasize their diversity, sore-ra-no is
used.
248 kaneko
(13) A: Pet shop-ni totemo kawaii koinu-ga nana-hiki imasita.

pet-shop-loc very pretty puppy-nom seven-cl were
Watasi-wa so-no koinu(-tati)-o kaimasita.
me-top so-no puppy-pl-acc bought
B: Nan-biki katta-no desu-ka?
what-cl bought-comp cop-q
A: There were seven very pretty puppies in the pet shop. I bought (one,
some or all) of those puppiesB: How many (puppies) did you buy?
Third, in deictic uses, the default reading of demonstrative phrases is also

that the propositional contribution is uniquely or maximally identified with
the demonstratum or demonstrata. But such uniqueness or maximality may
be cancelled, as in (14): in this example, the propositional contribution of
ko-no/a-no/so-no koinu(-tati) is not maximally identified with the seven pup-
pies indicated by As gesture, as confirmed by Bs question meaning how
many?
(14) [In a pet shop, a client A talks to a shop assistant B, pointing out seven
very similar puppies]
A: {Ko-no/A-no/So-no} koinu(-tati)-o kaimasu.
{ko-no/a-no /so-no} puppy-pl-acc take
B: Nan-biki desu-ka?
what-cl cop-q
A: Ill take (some or all) of {these / those} puppies!B: How many?
In sum, this section has shown that the bridging, anaphoric and deictic uses
of Japanese demonstratives do not convey uniqueness or maximality.8 We yet
need to understand why uniqueness or maximality is inferred by default in
their anaphoric and deictic uses, but not in their bridging use.
3.3 Comparison with Salish Demonstratives and with a Malagasy

Determiner ny
A use of demonstratives or definite determiners in contexts lacking uniqueness
or maximality is also reported in other languages. According to Matthewson
(1999), one of the demonstratives in Sttimcets (Lilooet Salish), ti a (which
8 Yanagida (2011) also argues that, among the three Japanese adnominal demonstratives, so-no
admits an indefinite reading, since it allows co-variable readings, as in (3a,b) above. But this
argument is not convincing since English the and that also allow co-variable readings, as
shown in (2a).
carries a [present] or [proximal] feature), may be used deictically in situations

where uniqueness is not required, as in (15): the ti a string serves here to
refer to one of the three birds present in the utterance situation, but one which
is no more salient than the other two birds.
(15) [context: there is a tree outside the window. There are three birds on the
tree.]
[Tkem i sqayqqyecw-a] atsx-en-tas [ti spzz7-a]
All det.pl man(dimin)(pl)-det] see-tr-3pl erg [det bird-det]
All (the) boys saw a / *the bird. (Matthewson 1999, 107)
Consultants comment: theyre all seeing the same one.
But unlike Japanese demonstratives, ti a does not necessarily convey famil-

iarity, since it may be used in brand-new contexts. Thus, in (16a), a totally
new referent is introduced by means of ti a, and the same referent is fur-
ther reidentified by the same demonstrative phrase: the first occurrence of
ti smmlhats-a is translated in English by a girl, and the second, by the
girl.
(16) Hy-lhkan ptakwlh, ptkwkh-min lts7a

going.to-1sg.subject tell.story tell.story-applicative here
[ti smmlhats-a] wa7 ku7 tal lti7 [ti
det girl-det imperfective report cry there det
smmlhats-a]. (idem. 108)
girl-det
Im going to tell a legend, a legend about a girl. The girl was crying
there.
Furthermore, ti a may denote different entities in anaphoric contexts. Thus,

in (17), we observe two occurrences of ti swwh-a which denote different
referents, as shown by the translation. Such a reading is excluded for Japanese
demonstratives.
(17) Wa7 1ts7a pankph-a [ti swwh-a] mta7 wa7 lku7

Be here Vancouver-det det cougar-det and be there
llwat-a [ti swwh-a] tit. (idem. 106)
Mt. Curie-det det cougar-det also
There is a cougar here in Vancouver and there is also a cougar there in
Mt. Currie. Consultants comment: There are two different cougars.
250 kaneko
Distributions more similar to Japanese adnominal demonstratives are exem-

plified by a Malagasy determiner ny. According to Paul (2009, 237), this deter-
miner uniformly encodes familiarity. [] If there is a relevant discourse ref-
erent present, then the DP must refer back to that referent. Moreover just as
anaphoric uses of Japanese demonstratives, ny+NP is interpreted by default
as denoting a unique or maximal referent: all of the examples of ny+NP in
object position9 collected by Paul (2009) denote familiar and unique / maxi-
mal entities. (idem. 227). But uniqueness or maximality may be cancelled, and
in the constructed example (18), while ny akondro det banana is anaphorically
associated with voankazo fruit, ny akondro does not mean the bananas, but
rather some of the bananas (idem. 228). It should be noted that in Malagasy,
nominal plurality is not necessarily encoded, so that a bare nominal like akon-
dro banana may be understood as denoting either an atomic entity or a sum.
(18) Nandeha tany an-tsena aho omaly ary nividy

at.go there acc-market 1sg(nom) yesterday and at.buy
voankazo. Nihinana ny akondro ny zanako
fruit at.eat det banana det child.1sg(gen)
(fa tsy nohaniny ny rehertra).
(but neg tt.eat.3(gen) det all)
I went to the market yesterday and bought fruit. My child ate (some of)
the bananas (but not all of them). (idem. 228)
4 Proposals
I now present my own assumptions regarding Japanese adnominal demonstra-

tives, on the basis of a comparison of their semantic (4.1) and syntactic (4.2)
properties with those of English demonstratives.
4.1 Semantics
In order to clarify the semantics of Japanese demonstratives, I refer to Elbour-
nes (2008) analysis of demonstratives, according to which demonstratives take
three arguments, index, Relation and NP, as in (19): i) index (signaled by i)
is a salient individual on the basis of which the actual interpretation of the
9 According to Paul (2009), the occurrence of this determiner is, unlike in object position,
obligatory in subject position where a non-familiar reading is available. The possibility of non-
maximal reading of ny acondro in (18) is confirmed by my Malagasy informant, Rasatranabo
Razakanivony Aina Anthony.
demonstrative is computed;10 ii) Relation (signaled by R) constrains the rela-

tion between index and the propositional contribution of the demonstrative
phrase. The semantics of this cat in (20a) is thus represented by (20b), where
the value of index is signaled by an assignment function g and determined by
gesture as Felix (in anaphoric uses, the value of index is determined by the
immediate antecedent). The propositional contribution of this cat is related to
the value of index (signaled by g(i) and determined as Felix) by the Identity
relation (noted by =), as in (20b). The referent of the demonstrative phrase is
uniquely determined by the iota operator.
(19) [DP [that i] R] NP] (adapted from Elbourne 2008, 430)
(20) a. This cat [gesture at Felix] laughs. (idem. 433)

b. [[this cat]]g = x (cat(x) & x = g(i))
(21) a. This donkey [gesture at field A] is healthier than that donkey [gesture
at field B]. (idem. 431)
b. [[this donkey]]g = x (donkey(x) & Localized-in(x)(g(i)))
Elbourne (2008) also notes that the propositional contribution of a demonstra-

tive phrase is not always equivalent to the value of the assignment function. In
(21a), Relation is contextually determined as the Localized-in relation: x (= the
denotation of this donkey) is localized in g(i) (= the denotation of field A)
In the case of bridging uses, as in (22a), index is provided by the antecedent:
the value of the assignment function g(i) is defined here as the reserved section
on the library where a girl y went, y being universally quantified; Relation is
contextually determined as the localized-in relation (x is localized in g(i)).11
(22) a. The girls received individualized reading lists with sections labeled
on reserve and in bookstore. Every girl went to the reserve section of
the library and read those books first. (= (2d))
b. [[those books (in (22a))]]g = x (books(x) & Localized-in(x)(g(i)))
Now, to make sense of the lack of uniqueness or maximality in Japanese adno-

minal demonstratives, I claim that they do not include the iota operator and
10 In deictic uses, index may further be spatially specified as [proximal] or [distal].

11 Elbourne (2008) analyzes demonstratives in the framework of situation semantics. I however
do not integrate situation variables in the semantic representations for simplicity of exposi-
tion.
252 kaneko
further lack their inherent quantificational force, unlike their English counter-
parts, as represented in (23).
(23) semantic hypothesis for Japanese adnominal demonstratives

[[a-no / ko-no / so-no NP]]g =Petx [P(x) & Relation (x)(g(i))]
(24) deictic and anaphoric uses

a. [[so-no hon (in (1c))]]g=x [book(x) & x=g(i)] [g(i) is an atom]
b. [[so-no koinu (in (13)/(14))]]g=x [puppy(x) & x g(i)] (g(i) is a sum)
c. [[(13)/(14)]]g =x [puppy(x) & x g(i) & take(A)(x)]
In their deictic or anaphoric uses, when the value of index is an atom, as in (1c),
Relation is Identity, as in (24a). Consequently, the propositional contribution
of so-no hon so-no book boils down to being uniquely identified. On the other
hand, when the value of index is a sum, as in (13) and (14) (where it corresponds
to the seven puppies either introduced by the preceding discourse or indicated
by As gesture), I crucially claim that Relation is not Identity, unlike with
English definite determiners, but Part-whole, as in (24b).12 The semantics of
(13) and (14) is computed via a contextually introduced existential quantifier,
as in (24c), which says that among the relevant seven puppies deictically or
contextually determined, one, some or all member(s) is / are such that A takes
it / them.
This analysis is based on Heim (2011), who suggests, as in (25), that indefinites
and definites form a scale of competing alternatives: I buy the seven puppies thus
entails I buy some of the seven puppies. The Gricean quantity principle states
that in article-languages, the use of partitive indefinites (ex. some of the seven
puppies) implicates the falsity of the stronger proposition conveyed by the def-
inite article (ex. the seven puppies). On the other hand, in article-less languages
like Japanese, indefinites have a wider range of felicitous uses because they do
not compete with definites and therefore do not induce the same implicature.
Thus, Japanese demonstratives allow a wider range of readings running from
indefinite partitivity to definite maximality, depending on the context.
(25) Ambiguous DPs [between definite and indefinite readings] in such lan-
guages [lacking definite and indefinite articles] are simply indefinites.
12 The semantic analysis proposed here departs from that of Kaneko (2012), where I assumed
that in anaphoric and deictic uses of Japanese demonstratives, Relation is Identity, just like
in anaphoric and deictic uses of English demonstratives.
They are semantically equivalent to English indefinites, but have a wider

range of felicitous uses because they do not compete with definites and
therefore do not induce the same [quantity] implicature.
heim 2011, 1006
The above analysis amounts to assuming that, in the deictic and anaphoric
cases, -no in a-no / ko-no / so-no marks partitivity, and that a / ko / so implic-
itly includes the same NP as the one following -no. In this respect, this analysis
is similar to Barkers (1998) for possessive partitive constructions, as in (26a),
whose semantics is represented by (26b).13 The partitive analysis of -no in deic-
tic or anaphoric uses is supported by the fact that adnominal demonstratives
may be followed not only by a NP as in (27a), but also by a numeral classifier
phrase, as in (27b): in the latter case, -no is clearly interpreted as partitive, and
so should implicitly include the NP koinu puppy.
(26) a. tools of Johns (Barker 1998, 701)

b. y [tools(y) & y Johns tools] (ibid.)
(27) [There were seven puppies in the pet shop. And ]

a. Watasi-wa so-no koinu-o katta.
me-top so-no puppy-acc bought
I bought (some or all) puppies of those (puppies).
b. Watasi-wa so-no 2-hiki-o katta.

me-top so-no 2-cl-acc bought
I bought two of those (puppies).
Next, bridging uses are divided into two cases: (i) one case includes a lexically
[+R] relational noun, such as typsyo writing in (11b). The semantics of this
case is represented in (28a), where the [+R] relational noun tyosyo writing
has two arguments, the external one of which is filled by g(i) provided by the
antecedent sensee Professor. The Relation component is lexically encoded in
the NP and construed as the Write relation (g(i) writes x); (ii) another case
includes nouns type-shifted from sortal ([R]) to relational ([+R]), as in so-no
13 The difference is that the English partitive construction should convey a proper partitivity
and excludes the possibility of signaling the maximality of the subsets, which is not the
case for Japanese adnominal demonstratives because of the lack of a definite / indefinite
distinction in this language.
254 kaneko
gakusee-tati students in (12). The semantics of this case is implemented by

means of a contextually salient Relation variable (Barker 2011, 1114), which is
specified in (12) as the Supervise relation (g(i) supervises x), as in (28b). Here,
the semantic function of -no in so-no is similar to that of English possessive
rather than partitive of, while so fills an argument slot which is either lexically
encoded inside the NP or created contextually.
(28) bridging use

a. [[so-no tyosyo (in (11b))]]g = x [writing(x)(g(i))]
b. [[so-no gakusee-tati (in (12))]]g = x[students(x) & Supervise
(g(i))(x)]
We can now understand why deictic and anaphoric uses of Japanese demon-
stratives induce maximality by default, while such is not the case in bridging
uses: according to the above analysis, in the former case, the Relation linking
the propositional contribution of the demonstrative phrase and the value of
index is defined in terms of partitivity (viz. partial or maximal identity). Now,
as noticed by Elbourne (2008) in (29), a cooperative speaker intends to make a
demonstrative phrase most easily interpreted for the hearer, and maximal iden-
tity with the demonstratum or the denotation of the antecedent requires less
effort from the hearer than partial identity. Therefore, although the demonstra-
tive phrase as a whole is existentially quantified and is a priori interpreted as
indefinite, the maximal-identity interpretation (which apparently corresponds
to the definite interpretation) is preferred by default unless there are obvious
reasons to make it impossible
(29) Since a cooperative speaker will intend that the interpretation [i.e. propo-
sitional contribution of a complex demonstrative] be the one that is most
easily identified in terms of its relation to the demonstratum, this kind of
interpretation [i.e. identify with the value of index] will always win out,
unless there are obvious reasons to make it impossible.
elbourne 2008, 443
On the other hand, in bridging uses, the value of index is related to the proposi-
tional contribution of a demonstrative phrase only indirectly, by filling an argu-
ment slot. The whole demonstrative phrase is existentially quantified, which
naturally yields an indefinite interpretation.
The indefinite partitive analysis of Japanese demonstratives is supported
by at least three phenomena. First, we know that the noun phrase which is
correlated with the wh-remnant of sluicing cases can only be an existential
(Matthewson 1999, 107), as illustrated in (30). Now, both anaphoric and bridging
uses of Japanese demonstratives allow sluicing, as witnessed by (31a,b).
(30) John is looking for {a book / *the book / *this book}, but I dont know
which.
matthewson 1999, 107
(31) a. watasi-wa huta-ri-no gakusee-o sidoosi-teiru. so-no

me-top 2-cl-gen student-acc supervise-prog so-no
gakusee-ga watasi-o home-tekure-ta rasii ga,
student-nom me-acc praise-give-pst I heard but
watasi-wa dotira-ka wakara-nai. [anaphoric use]
me-top which one-Q know-neg
I supervise two students. I heard that one of them spoke well of me.
But I dont know which one.
b. Chomsky to so-no gakusee-ga ronbun-o kaita sooda.

Chomsky and so-no student-nom paper-acc wrote I heard.
Demo, dono-gakusee-ka wakara-nai. [bridging use]
but which student-Q know-neg
I heard that Chomsky and one or some student(s) of his had written a
paper. But I dont know which student(s).
Second, the distribution of numeral classifiers specifying the cardinality of a

demonstrative phrase supports the partitive analysis. In Japanese, a numeral
classifier phrase may occupy at least two positions:14 (i) following the case-
marked NP, as in (32a); (ii) following the adnominal demonstrative and with
genitive-marking, as in (32b). My hypothesis is that so-no koinu so-no puppy
in (32a) itself induces partitivity by means of partitive -no. Another possible
14 The numeral classifier may follow the no-marked NP and be case-marked, as in (i). The second
-no is interpreted as a partitive marker. So-no koinu should here refer back to all of the seven
puppies previously introduced. I assume that this maximality effect is due to the general
principle requiring that the superset of a partitive be definite, rather than to the semantics of
the adnominal demonstrative itself.
(i) So-no koinu-no 2 hiki-o katta

so-no pupp-no 2-cl-acc bought
I bought two of those puppies.
256 kaneko
hypothesis is that so-no koinu is a definite expression and denotes the maximal
referents previously introduced, with partitivity inferred in some NP-external
way. The latter hypothesis is disconfirmed and the former supported by the fact
that when maximality is overtly conveyed by another genitive-marked numeral
classifier in pre-nominal position, the acceptability of another numeral classi-
fier forcing partitivity is degraded, as in (32c).
(32) [There were seven puppies in the pet shop. And ]

a. So-no koinu-o ni-hiki katta.
so-no puppy-acc 2-cl bought
I bought two puppies of those (puppies).
b. So-no ni-hiki-no koinu-o katta

so-no 2-cl-gen puppy-acc bought
I bought two puppies of those (puppies).
c. *So-no nana-hiki-no koinu-o ni-hiki katta.

so-no 7-cl-gen puppy-acc 2-cl bought
(intended) I bought two puppies of those seven (puppies).
Third, the analysis of Japanese demonstrative phrases as basically indefinite

seems to be refuted by the consistency test, showing that two occurrences
of the same nominal can denote different objects in the same context if it is
indefinite, but cannot do so if it is definite (Sebastian Lbner p.c.). In effect,
two occurrences of the partitively-interpreted indefinite DP a puppy in (33)
denote different entities, while such an interpretation is difficult to construe
with two occurrences of the also partitively-interpreted so-no koinu so-no
puppy in (34a). It should however be noticed that if we add a numeral classifier,
acceptability is improved, as in (34b).
(33) There were seven puppies in the pet shop. A puppy was white and a puppy
was black.
(34) [There were seven puppies in the shop. And ]

a. *So-no koinu-wa siro de, so-no koinu-wa kuro
so-no puppy-top while cop, so-no puppy-top black
datta. [anaphoric use]
was
(intended) Some puppies of those (puppies) were white, and some
puppies of those (puppies) were black.
b. So-no 1 pikki-no koinu-wa siro de, so-no 1 pikki-no

so-no 1 cl-gen puppy-top while cop, so-no 1 cl-gen
koinu-wa kuro datta.
puppy-top black was
One puppy of those (puppies) was white, and one puppy of those
(puppies) was black.
Curiously, a similar phenomenon is observed in French bare partitive con-

structions: in this language, partitivity may be expressed simply by the prepo-
sition de of + DP denoting the superset, without any quantifying expression,
as in de tes macarons of your macaroons in (35a) (see Kupferman 2004, Zribi-
Hertz 2006, among others). Such bare partitive constructions do not pass the
consistency text, as shown in (35b). But when a numeral is inserted, acceptabil-
ity is improved, as in (35c). The consistency test therefore supports, rather than
refutes, the partitive analysis of Japanese demonstratives.
(35) a. Jai repris deux fois de tes macarons.

(lit) I re-took two times of your macaroons.
(I took two extra helpings of your macaroons.)
b. *Jai repris de tes macarons, mais pas de tes macarons.
(lit) I re-took of your macaroons, but not of your macaroons.
c. Jai repris un de tes macarons, mais pas un (autre) de tes macarons.
(lit) I re-took one of your macaroons, but not (another) one of your
macaroons.
(I took back one of your macaroons, but not (another) one of your
macaroons.)
4.2 Syntax
Some recent studies argue that demonstratives are not unanalyzable cate-
gories, but are decomposed into three morphological parts: definite, deictic
and noun. Kayne and Pollock (2010) thus decompose the English demon-
strative pronoun that into th (definite), at (deictic) and an unpronounced
noun glossed THING. Along this line of analysis, Yanagida (2011) decomposes
Japanese adnominal demonstratives into two parts, respectively occurring in
the specifier and head of the Demonstrative Phrase (DemP), as in (36a). She
further suggests that DemP is generated in the specifier of Number Phrase
(NumP), as in (36b).
(36) a. [DemP so [Dem no]] (Yanagida 2011, 3)

b. [DP [NumP [Dem so-no] [NP koinu (puppy)]]]
258 kaneko
The structure in (36b) straightforwardly accounts for the indefinite reading

of Japanese adnominal demonstratives, discussed in Section 4. But the exis-
tence of NumP is not established for Japanese nominal structure and further-
more, Japanese adnominal demonstratives do not seem to be specified for the
number feature (see footnote 7).15
On the other hand, Bokovi (2009) claims that article-less languages, like
Japanese, Korean, Serbo-Croatian, etc., lack the DP projection, and that de-
terminer-like expressions (ex. possessives, demonstratives, etc.) in these lan-
guages are syntactically adjunctive modifiers. Leu (2008) similarly argues that
demonstratives include a NP-modifier component, on the basis of the mor-
phological properties of demonstratives in Germanic languages.16 This author
further analyzes Japanese demonstratives as corresponding to his modifier
15 In support of number specification in Japanese adnominal demonstratives, Yanagida (2011)

argues that they trigger number agreement: when they are pluralized and followed by a
nominal, this nominal must also be pluralized, according to Yanagida, as shown by the
contrast between (ia) and (ib) below. Note that the author legitimately analyzes the so-called
third-person pronoun kare as a kind of demonstrative.
(i) a. ??[kare-ra][gakusee]-no kiboo (Yanagida 2011, 7)

him-pl student-gen hope
the hope of them students
b. [kare-ra][gakusee-tati]-no kiboo (ibid.)

him-pl student-pl -gen hope
the hope of them students
Note, however, that kare-ra + bare NP strings are frequently attested, as in (ii).
(ii) [kare-ra] [gaikokuzin gakusee]-no hobo 4-bun-no 3-ga

him-pl foreign student-gen around quarters 3-nom
Around three quarters of them foreign students are (http://www.jmf.or.jp/syuuhou/
html/20100212.html)
Taking such examples into account, I conclude that the existence of number agreement in
Japanese is not firmly confirmed.
16 Leu (2008) observes that in some Germanic languages, demonstratives take the same form
as the one that the definite article takes when followed by an adjective modifier, e.g. di in
Swiss German as illustrated in (ia), and that in colloquial Swedish, demonstratives consist of
definite article + locative here or there, as in (iib). On the basis of these observations, this
author claims that demonstratives have a complex structure consisting of definite article +
implicit or explicit modifier.
component, suggesting that the Japanese way of saying this book is some-
thing like book of here.
Inspired by these previous works, I assume that Japanese adnominal demon-
stratives lack a DP projection responsible for uniqueness / maximality,17 are
decomposed into two parts forming DemP, and should be analyzed as NP-
adjoined phrases, as in (37).
(37) syntactic hypothesis for Japanese adnominal demonstratives

[NP [DemP so [Dem no]] [NP koinu puppy]]
In section 4.1, I claimed that Japanese adnominal demonstratives semantically

serve either (i) to express partitivity in their deictic or anaphoric uses, where
-no assumes a function similar to that of partitive of in English, or (ii) to fill an
argument slot in bridging uses, where the function of -no rather corresponds
to that of possessive of in English. Under this analysis, DemP (formed of a
deictic component, ko / so /a, in its specifier position and of -no in its head)
assumes a function comparable to English prepositional phrases surfacing as
of +NP. At least three syntactic and morphological pieces of evidence support
this assumption.18
4.2.1 Where Adnominal Demonstratives Parallel the Adnominal

WH-Word do-no
The Japanese adnominal demonstratives a-no, ko-no, so-no, are made up of the
demonstrative prefixes a- / ko- /so-, followed by -no. The same decomposition
applies to the adnominal WH-word do-no which, as in (38), which conveys
selection from a set of alternative candidates (therefore includes the notion
of partitivity), and lacks, as well known, its own quantificational force. Do-no
must therefore be combined with some quantificational expression, such as
(i) a. d ros / di rot ros / di- ros (Leu 2008, 19) [Swiss German]
the rose / the red rose / this rose
b. det hr / der dr (idem. 21) [colloquial Swedish]

the here this one / the there that one
17 This idea is due to Lyons (1999) who suggests that the syntactic head D is the locus of the
semantic feature uniqueness / maximality.
18 The first and second properties may equally be explained by assuming, as Yanagida (2011),
that DemP occurs in Spec-NumP, rather than NP-adjoined position. But the third property
does not seem to be accounted for under Yanagidas analysis.
260 kaneko
the interrogative particle -ka in (38). The traditional grammarian Otsuki (1889)
even analyzes do- as a subcategory of demonstratives. Note, furthermore, that
the demonstrative prefixes very systematically display the same morphology
as that of the WH-prefix doin pronominal, locative, directional and adverbial
forms, as in (39ae). These parallel properties suggest that demonstrative pre-
fixes, like the WH-prefix do-, lack their own quantificational force and involve
the idea of partitivity.
(38) Do-no koinu-o kaimasu ka?

which puppy-acc buy Q
Which puppy do you buy?
(39) a. {ko-no / a-no / so-no / do-no} koinu [adnominal]

this / thatdistal / that / which} puppy
b. {ko-re / a-re / so-re / do-re} [pronominal]

this / thatdistal / that / which
c. {ko-ko / aso-ko / so-ko / do-ko} [locative]

{here / theredistal / there / where}
d. {ko-tira / a-tira / so-tira /

{this direction / thatdistal direction / that direction /
do-tira} [directional]
which direction}
e. {ko-o / a-a / so-o / do-o} [adverbial]

{in this way / in thatdistal way / in that way / in which way}
4.2.2 Preceding Modifiers Allowed

In English and even in Serbo-Croatian, another article-less language, demon-
stratives cannot be preceded by other modifiers, as illustrated in (40a,b). Con-
trastively, as noted by Kamio (1977) and Miyamoto (2009), among others, Jap-
anese demonstratives may be preceded by a modifier, as witnessed by (41a).
And although it is sometimes claimed that modifiers preceding a demonstra-
tive must be interpreted as non-restrictive (ex. Kamio 1977), (41b) shows that
they may be interpreted as restrictive: here, the superlative, itiban the most,
inside the relative clause, forces a restrictive reading. The fact that it may fol-
low a restrictive modifier clearly indicates that so-no in (41b) is not a determiner
closing a nominal projection.
(40) a. *expensive this car (Bokovi 2009, 195) [English]

b. ova skupa kola / *skupa ova kola [Serbo-Croatian]
this expensive car / expensive this car (Bokovi 2009, 194)
(41) a. [hahaoya-o nakusita] a-no ko (Kamio 1977, 154)

mother-acc lost thatdistal child
that child who lost his / her mother
b. Toyota-wa [itiban gyoosekinoyoi] so-no ko-gaisya-o

Toyota-top [most productive] so-no subsidiary-acc
suisensita
recommended
Toyota recommended its most productive subsidiary.19
4.2.3 NP Ellipsis Disallowed

Saito, Lin and Murasugi (2008) argue that an adnominal no-marked phrase
in Japanese is situated either (i) in argument position when ellipsis of the
following NP is allowed, or (ii) in adjunct position when ellipsis is disallowed.
19 Anaphoric so-no does not seem to easily allow a preceding restrictive modifier, as shown by
the contrast between (ia) and (ib) below. I assume that this difficulty is due to a pragmatic
constraint requiring that the antecedent of so-no be as near as possible (Iori 2007). It is to be
noticed that so-no may refer to the content of an immediately preceding modifier, as in (ii).
When the intended antecedent is in a different sentence and a modifier intervenes between
it and so-no, as in (ib), the interpretation of so-no becomes ambiguous. On the other hand,
in bridging uses, as in (41b), so-no simply fills an argument slot, and its interpretation is not
disturbed by an intervening modifier.
(i) [There were seven puppies in the pet-shop, and]

a. Watasi-wa so-no [itiban tiisana] koinu-o kaimasita.
me-top so-no [most little] puppy-acc bought
b. ??Watasi-wa [itiban tiisana] so-no koinu-o kaimasita.

me-top [most small] so-no puppy-acc bought
I bought the smallest puppy (among them).
(ii) Anata nasi dewa iki-rare-nai to itteita sono Junko ga ima

you without if alive-can-neg comp was.saying sono Junko-nom now
hoka-no otoko-no kodomo-o huta-ri mo un-deiru.(adapted from Iori 2007, 98)
another guy-gen child-acc two-cl even give.birth.-Resultative
The same Junko who used to say that she could not be alive without me gave birth to two
children with another guy.
262 kaneko
Thus, Hanako-no Hanakos in (42a) (which allows ellipsis of the following NP,
taido attitude) is in an argument position, while ame-no rainy in (42b) (which
doesnt allow ellipsis of the following NP, hi day) is in an adjunct position.
Now, as shown by (43a,b), Japanese adnominal demonstratives do not license
ellipsis of the following NP in deictic and bridging uses, which indicates that
they behave as NP-adjuncts.
(42) a. [Taro-no taido]-wa yoi ga, [Hanako-no taido]-wa

Taro-no attitude-top good but Hanako-no attitude-top
yoku-nai
good-neg
Taros attitude is good, but Hanakos isnt. (Saito, Lin and Murasugi
2008, 253)
b. *[Hare-no hi]-wa yoi ga, [ame-no hi]-wa otikomu. (ibid.)

clear-no day-top good but rain-no day-top feel depressed
(intended) Clear days are ok, but I feel depressed on rainy days.
(43) [Pointing at a slide shown in a conference]

a. *[Ko-no kasetu]-wa tadasii ga [ko-no kasetu]-wa
ko-no hypothesis-top right but ko-no hypothesis-top
tadasiku-nai.
right-neg
(intended) This hypothesis is right, but this hypothesis is not right.
b. *Minsyusyugi-wa [so-no kati]-o usinat-tei-nai, kihonteki

democracy-top so-no value-acc lost-perfect-neg fundamental
zinken -mo [so-no kati]-o usinat-tei-nai
human rights -also so-no value-acc lose-perfect-neg
(intended) Democracy has not lost its value. Fundamental human
rights similarly have not lost their value.
4.3 Data from the Field of Acquisition and Reflections from a

Cross-Linguistic Perspective
The assumption that Japanese demonstratives lack uniqueness or maximal-
ity is consistent with an observation from the field of L2 acquisition. Kaneko
(1996) observes that L1 Japanese learners overuse the English definite article
the in indefinite partitive contexts. Thus in (44), pencil is interpreted as one of
the pencils introduced by the antecedent, some pencils. In such partitive con-
texts, native English speakers choose the indefinite article, a, while Japanese
learners tend to misuse the. It is to be noticed that anaphoric and bridging

uses of English the are sometimes translated in Japanese by so-no, as in (45a,b),
found in English-Japanese dictionaries: in (45a), the dog is co-referent with the
antecedent a dog; in (45b), the mark is understood, through bridging inference,
as the mark left on the telegraph pole after the traffic accident described by the
first sentence. If we assume that the acquisition of English the by L1 Japanese
learners is somehow influenced by L1 transfer due to Japanese so-nos lack of
maximality, we may naturally account for the Japanese learners overuse of
English the in indefinite partitive contexts.
(44) anaphoric
Once there was a boy. He wanted to write a letter. He went to his mother.
She showed him some pencils. So he took (a / the / ) pencil. And he wrote
his letter.
kaneko 1996
(45) a. anaphoric
We keep a dog, and are all fond of the dog. (the translated with so-
no in Kenkyusyas English-Japanese Dictionary for the General Reader,
2246)
b. bridging
His car struck a telegraph pole; you can still see the mark on the
pole. (the translated with so-no in Genius English-Japanese Dictionary,
1940)
A similar misuse is reported for another article-less language, Korean. Ko,

Ionin and Wexler (2010) show, on the basis of a systematic empirical inves-
tigation, that L1-Korean L2-English learners similarly misuse English the in
indefinite partitive contexts both in anaphoric and bridging cases. To account
for this observation, they first assume that not only definiteness, but also
existential presuppositionality (which boils down to familiarity) are semantic
universals provided by Universal Grammar: definiteness is defined as a com-
bination of existential presupposition + uniqueness/maximality presupposi-
tion, while existential presuppositionality does not necessarily correlate with
uniqueness/maximality. They further assume that L2 learners have access to
semantic universals provided by Universal Grammar, just like child L1
learners (idem. 214), and that, fluctuating among possible feature settings,
they mis-set English the as a marker of existential presuppositionality, rather
than definiteness. It should be noted that Korean has three demonstratives
i, ce, ku, each of which roughly corresponds to a-no, ko-no and so-no in
264 kaneko
Japanese.20 One possibility is that Korean demonstratives, like Japanese de-

monstratives, are not specified for [uniqueness/maximality] unlike English
definite determiners, but only for [existential presuppositionality], and that
this feature setting of the native language influences the aforementioned mis-
use of English the by L1 Korean learners.21
In another article-less language of East Asia, Chinese, demonstratives show
syntactic distributions similar to those of Japanese. Partee (2006) points out
that Chinese adnominal demonstratives, formed by na (+numeral)+classifier,
may be either preceded or followed by a possessive phrase,22 as in (46a,b), and
that both orders fail to convey the maximality presupposition that Zhangsan
has only three books; Henrietta Yans comment is reproduced in (47).
(46) a. Zhangsan de [na san ben shu] (Partee 2006, 10)

Zhangsan depossessive that three cl book
(lit) Zhangsans that three books
b. na san ben [Zhangsan de shu] (ibid.)

that three cl Zhangsan depossessive book
(lit) that three books of Zhangsans
(47) If the speaker knows that Zhangsan has exactly three books, she would be
more likely to use [(46a)], and if she knows that Zhangsan has more than
three books, she would use [(46b)]. If the speaker doesnt know, or if the
question is totally irrelevant, either could be used.
idem. 11
20 For more details, see Kinsui, Okazaki and Jo (2002).

21 In order to verify this analysis, we should first examine if Korean demonstratives similarly dis-
play the semantic and syntactic characteristics discussed in sections 4.1 and 4.2 for Japanese
demonstratives.
22 In another article-less language, Latin, demonstratives may, like those of Japanese, be pre-
ceded by modifiers or head nouns, as in (i). Giusti and Iovino (2012) however argue that Latin
has a DP projection, and analyze the liberal word orders, as in (i), as resulting from DP-internal
movement.
(i) liber iste quem mihi misisti

book.nom.m.sg. this.nom.m.sg. thatacc.m.sg. me.dat. send.2.sg.perfect
this book that you sent me (Cicero, fam. 15,21,2 / cited by Giusti & Iovino 2012)
It is the set topic of a future study to examine whether Korean and Chinese
demonstratives should be analyzed in the same way as Japanese demonstra-
tives, both semantically and syntactically.
5 Concluding Remarks
In this paper, I have proposed some new ideas about the semantics and syntax
of the Japanese adnominal demonstratives a-no, ko-no and so-no. As regards
semantics, I have essentially claimed (i) that while conveying familiarity by
means of the demonstrative prefixes a, ko and so, they lack uniqueness or
maximality, and that the whole demonstrative phrase is existentially quantified
in its context, and (ii) that -no either marks partitivity (without excluding
maximality) in the deictic and anaphoric uses, orin bridging usesserves
to fill an argument slot (lexically encoded inside the following NP or created
contextually). This indefinite analysis of Japanese demonstratives is supported
by (i) the availability of sluicing, (ii) the distribution of numeral classifiers, and
(iii) the similar behaviour of French partitive constructions with respect to the
consistency test.
I have further claimed that these semantic properties were adequately cap-
tured if we should analyze Japanese adnominal demonstratives as forming
a demonstrative phrase (DemP) located in NP-adjoined position and whose
specifier and head are respectively occupied by the demonstrative prefixes
ko/so/a, and by -no. The adjunction hypothesis is supported by three morpho-
syntactic properties: (i) the demonstrative prefixes, ko-, so-, a- systematically
display the same morphology as that of the WH-prefix do-; (ii) the Japanese
demonstratives may be preceded by a restrictive modifier, like other adjunct
modifiers and unlike definite determiners in other languages; (iii) they behave
with respect to the ellipsis of the following NP as other no-marked expressions
clearly identified as adnominal adjuncts.
Finally, I have noted that the proposed hypotheses might shed light on
some data from L2 acquisition, and might be extended to Korean and Chinese
adnominal demonstratives.
Bibliography
Barker, Chris. 1998. Partitives, Double Genitives and Anti-Uniquness. Natural Lan-
guage and Linguistic Theory 16: 679717.
Barker, Chris. 2005. Possessive Weak Definites. In Possessives and Beyond: Semantics
266 kaneko
and Syntax, edited by Ji-yung Kim, Yury A. Lander and Barbara H. Partee, 89113.
Amherst, MA: GLSA Publications.
Barker, Chris. 2011. Possessives and Relational Nouns. In Semantics: An International
Handbook of Natural Language Meaning. Volume II, edited by Klaus von Heusinger,
Claudia Maienborn and Paul Portner, 11091130. Berlin: de Gruyter Mouton.
Bokovi, eljko. 2009. No-DP analysis of article-less languages. Studia Linguistica
63.2: 187203.
Elbourne, Paul. 2008. Demonstratives as Individual Concepts. Linguistics and Philos-
ophy 31: 409466.
Genius English-Japanese dictionary 2006. 4th printing. Tokyo: Taisyukan.
Giusti, Giuliana and Rossela Iovino. 2012. Latin as an Articleless DP Languages. Paper
presented at Journe organise par le projet Calcul de la rfrence nominal: langue
avec et sans articles,Paris,March1516.Handout available fromhttp://www.umr7023
.cnrs.fr/sites/sfl/IMG/pdf/lsalaa2012GiustiIovino.pdf
Heim, Irene. 2011. Definiteness and Indefiniteness. In Semantics: An International
Handbook of Natural Language Meaning. Volume II, edited by Klaus von Heusin-
ger, Claudia Maienborn and Paul Portner, 9961025. Berlin: de Gruyter Mou-
ton.
Hoji, Hajime, Satoshi Kinsui, Yukinori Takubo and Ayumi Ueyama. 2003. The Demon-
stratives in Modern Japanese. In Functional Structure(s), Form and Interpretation,
edited by Yen-hui Audrey Li and Andrew Simpson, 97128. London and New York:
Routledge.
Iori, Isao 2007. Nihongo-ni okeru tekisuto-no kessokusee-no kennkyuu (Study of textual
cohesion in Japanese). Tokyo: Kurosio Pub.
Kamio, Akio. 1977. Restrictive and Non-restrictive Relative Clauses in Japanese. In
Descriptive and Applied Linguistics 10: 147168. Tokyo: International Christian Uni-
versity.
Kaneko, Yumina 1996. Knowledge of the English article system in second language
learning: To the or not to the. Undergraduate thesis. Smith College, Northamp-
ton, MA.
Kaneko, Makoto 2012. Japanese Demonstrative so-no as a Modifier Lacking Definite-
ness. In Proceedings of Sinn und Bedeutung 2, edited by Ana Aguliar Guevara, Anna
Chemilovskaya and Rick Nouwen, 335348. Cambridge. MA: MIT Working Papers in
Linguistics.
Kayne, Richard and Jean-Yves Pollock 2010. Notes on French and English demonstra-
tives. manuscript. New York University and Universit Paris Est, EA4120. available
from: http://as.nyu.edu/docs/IO/2652/KaynePollock1109NotesOnFrenchAnd
EnglishDemonstratives.pdf
Kenkyusyas English-Japanese Dictionary for the General Reader. 1986. 5th printing.
Tokyo: Kenkyusya.
Kinsui, Satoshi, Tomoko Okazaki & Mikyon Jo 2002. Sizisi no rekisiteki / taisyoogen-
gogakuteki kenkyuu (A Historical and contrastive study on demonstrativesJapa-
nese/ Korean/ Turkish). In Taisyoo Gengogaku (Contrastive Linguistics), edited by
Naoki Ogoshi, 217247. Tokyo: Tokyo University Pub.
Ko, Heejeong, Tania Ionin and Ken Wexler. 2010. The Role of Presuppositionality in
the Second Language Acquisition of English Article. Linguistic Inquiry 41.2: 213
254.
Kobayakawa, Satoru. 2004. Nihongo-no fukusuu-hyoogensore-ra-no+meesi to
byoogo-meesi (Plurality in Japanesesore-ra-no+noun and nominal reduplica-
tion). Mind and Language 3: 3550. Aichi: University of Human Environments.
Kupferman, Lucien. 2004. Le mot de. Paris: Duculot.
Leu, Thomas. 2008. The Internal Syntax of Determiners. PhD diss., New York Univer-
sity.
Lbner, Sebastian. 2011. Concept Types and Determination. Journal of Semantics 28.3:
279333.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.
Matthewson, Lisa. 1999. On the Interpretation of Wide-Scope Indefinites. Natural
Language Semantics 7: 79134.
Miyamoto, Yoichi. 2009. On the Nominal-Internal Distributive Interpretation in Japa-
nese. Journal of East Asian Linguistics 18: 233251.
Nakanishi, Kimiko and Satoshi Tomioka. 2004. Japanese Plurals are Exceptional. Jour-
nal of East Asian Linguistics 13: 113140.
Nitta, Yoshio. 1992. Nihongo meesi-no kazu-gainen-no hyoozi-ni tuite (On the number
marking of Japanese nouns). Bunka gengogaku. so-no teigen to kensetu (Cultural
Linguistics Its Recommendation and Construction), 608594. Tokyo: Sanseido Pub.
Okazaki, Tomoko. 2010. Nihongo sizisi-no rekisi-teki kenkyuu (Diachronic Study of Japa-
nese Demonstratives). Tokyo: Hitsuzi Pub.
Otsuki, Fumihiko. 18891996. Goho sinan. Tokyo: Benseisya Pub.
Partee, Barbara H. 2006. A Note on Mandarin Possessives, Demonstratives, and Defi-
niteness. In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics
and Semantics in Honor of Laurence R. Horn, edited by Betty J. Birner and Gregory
Ward, 263280. Amsterdam: John Benjamins.
Paul, Ileana. 2009. On the Presence versus Absence of Determiners in Malagasy.
In Determiners Universal and Variation, edited by Jila Ghomeshi, Ileana Paul and
Martina Wiltschko, 215241. Amsterdam: John Benjamins.
Saito, Mamoru, T.-H. Jonah Lin, and Keiko Murasugi. 2008. N-Ellipsis and the Structure
of Noun Phrases in Chinese and Japanese. Journal of East Asian Linguistics 17:
247271.
Wolter, Lynsey. 2006. Thats that: The Semantics and Pragmatics of Demonstrative
Noun Phrases. PhD diss., University of California.
268 kaneko
Yanagida, Yuko. 2011. Agreement and the Restructuring of the Japanese Pronominal
System. In Japanese / Korean linguistics 20, edited by Bijarke Frellesvig and Peter
Sells, 116. Stanford: CSLI Publications.
Yang, Mei and Tania Ionin. 2009. L2 English Articles and the Computation of Unique-
ness. Proceedings of the 3rd Conference on Generative Approaches to Language.
Acquisition North America (GALANA 2008), edited by Jean Crawford et al., 325335.
Somerville, MA: Cascadilla. Proceedings Project.
Zribi-Hertz, Anne. 2006. Pour une analyse unitaire de DE partitif. In Indfini et prdi-
cation, edited by Francis Corblin, Sylvie Ferrando and Lucien Kupferman, 141154.
Paris: Presses de lUniversit Paris-Sorbonne.
From Noun to Name:
On Definiteness Marking in Modern Martinik
Anne Zribi-Hertz and Loic Jean-Louis
1 Introduction1
This article bears on two functional morphemes written l(a)- and l and pro-
nounced [l(a)] and [le] which have developed in Modern Martinik as definite-
ness markers of a sort, alongside the better known enclitic definite determiner
la,2 which is common to all French-lexifier creoles (cf. Bernab 1983, Gadelii
1997, Lefebvre 1998, Dprez 2007, Zribi-Hertz and Glaude 2007, Alleesaib 2012,
a.o.). We shall argue that la conveys pragmatic definiteness, as defined by
Lbner (1985, 2011), while l(a)- and l form semantically definite DPs denot-
ing individual terms in the manner of definite proper names. Since French
the lexifier languageambiguously marks semantic and pragmatic definite-
ness by means of the same definite article, and since the definite determiner
in Gbe languagesa plausible substratic influence on Caribbean creolesis
restricted to pragmatic definiteness (cf. Aboh 2001), the fact that the grammar
of Martinik should have developed three distinct overt markers of definiteness
is, incidentally, evidence that creolisation cannot be viewed as a simplifica-
tion process, as claimed by McWhorter (2001).
1 Previous stages of the research which led to this article were presented orally to various
audiencesthe International Conference on Bare Nouns and Genericity (Universit Paris
7, October 2010), FACS 2 (Berlin, November 2010), the GRGC seminar (Paris, November
2011), the Genius 3 Conference (Paris, December 2011), the Weak Referentiality Workshop
(Utrecht, March 2012), and the ATIFL montly seminar in Nancy (March 2013), whom we
gratefully acknowledge for their critical ear. We owe a special debt of gratitude to Muhsina
Alleesaib, Claire Beyssade, Patricia Cabredo Hofherr, Joaquim de Carvalho, Maxime Deglas,
Henriette De Swart, Malik Ferdinand, Guillaume Fon Sing, Herby Glaude, Fabiola Henri, Bert
Le Bruyn, Ora Matushansky, Lea Nash, Isabelle Roy, Emmanuel Schang, Elena Soare, Alice ter
Meulen, Florence Villoing, Roberto Zamparelli and Joost Zwarts for their precious feedback,
and to Riona Charlery, Guy Deslauriers, Luc and Thrse Milcent and Losa Paulin for their
judgements on the Martinik data.
2 The form(s) taken by this morpheme vary across creoles, and in some of them according to the
phonological context. The spelling la we adopt here is meant to ignore this morphological
variation, and the spelling -la below, to specifically identify the la morpheme of Martinik,
which crucially behaves as a phrasal enclitic.

270 zribi-hertz and jean-louis
We first summarise (section 2) Lbners (1985) distinction between semantic

and pragmatic definiteness (refined in Lbner 2011), which provides a theo-
retical background for the description to follow. Section 3 summarises some
relevant information on DP syntax in Martinik, and argues that the phrasal
enclitic determiner -la must be semantically characterised as a pragmatic def-
initeness marker, in Lbners (1985, 2011) sense. The next two sections present
the morphological, distributional and semantic properties of l(a)-N (section
4) and l+NP (section 5), arguing that their behaviour echoes that of definite
proper names, regardless of the lexical (proper or common) nature of their
head noun. Section 6 recapitulates and argues that l(a)-N and l+NP in Mar-
tinik instantiate a class of DPs we propose to call Names, characterised seman-
tically as a subtype of semantic definites denoting individual concepts, and
syntactically, by the occurrence of a special functional (Name) projection dis-
tinct from nP.
2 Semantic vs. Pragmatic Definiteness
This distinction is developed by Lbner (1985, 2011), whose theory of Definite-

ness somewhat differs from such classical views as those proposed by Russell
(1919), Strawson (1950), Hawkins (1978) Heim (1982), Kadmon (1990), Abbott
(1999), Roberts (2003), Barker (2005), a.o., which characterise definite descrip-
tions in terms of referential uniqueness, cf:
(1) A use of a definite description is felicitous if and only if there is exactly

one object in the context that satisfies the content of the description.
[informal phrasing from Barker 2005]
Contrary to these authors, Lbner (1985) claims that definiteness involves non-
ambiguity (uniqueness) of identification, rather than uniqueness of reference.3
According to this author, the definite article indicates that the (head) noun
identifies the referent via the unambiguous role it plays in the relevant situa-
tion: It is not uniqueness [of reference], but non-ambiguity which is essential
for definiteness. Non-ambiguity is the property of an expression that allows
for only one interpretation (possibly under additional constraints). Unique-
ness of reference is always an accidental property of a sortal concept ()
3 A rather similar view is developed by Corblin (passim).

from noun to name 271
Non-ambiguity, in contrast, may be an inherent property of (also non-sortal)

concepts. (Lbner 1985: 291). Thus, unlike some other theories of definiteness,
Lbners straightforwardly accommodates such examples as (2a) (where the
italicised DP does not identify a unique referent, but rather a unique relational
concept) and does not consider as basic the deictic and anaphoric uses of
definite descriptions illustrated in (2b,c):
(2) a. He is the son of a famous violinist.

b. Could you pass me the potatoes, please?
c. Mary has a dog and a cat. The dog has fleas but the cat hopefully
doesnt.
Non-ambiguity of reference may be established either independently of, or in

relation to, the immediate situation or context of utterance. Lbner (1985, 2011)
calls the first type semantic definiteness, and the second type, pragmatic defi-
niteness, and crucially considers the first type as basic: Semantic definites refer
unambiguously due to general constraints. Pragmatic definites depend on the
particular situation for unambiguous reference. (Lbner 1985: 299). The most
basic type of semantic definites denote one-place functional concepts (FC1s in
Lbner 1985, individuals of type e in Lbner 2011) such as the moon, the sun, the
truth which only involve a situational argument. More complex types headed
by relational nouns (e.g. son in (2a)) involve more than one argument and thus
instantiate, e.g., two-place functional concepts (FC2s in Lbner 1985). Proper
names are crucially a subtype of FC1 semantic definites: Within a certain range
of situations, proper names refer unambiguously to certain objects. They con-
stitute constant functional concepts, as their value does not vary with their
possible arguments. (Lbner 1985: 299). Lbner (2011) further classifies lexi-
cal nouns into four lexical types (sortal: dog, flower; individual: moon, truth;
relational: brother, friend; and functional: roof, back) and explores the articu-
lation of lexical meaning with determination: determination is congruent if its
function converges with the conceptual content of the noun, viz. if a definite
determiner combines with a noun which inherently identifies a unique indi-
vidual or function.
We find that Lbners theory of definiteness, which straightforwardly inte-
grates proper names, provides us with a convenient background to describe
the determiner system of Martinik, a language which overtly distinguishes
a marker of pragmatic definiteness and two markers of semantic definite-
ness selecting individual terms. A morphological split between semantic and
pragmatic definiteness is reported to exist in other languages, e.g. in various
West-Germanic dialects (cf. Ebert 1970, 1971, Lbner 2011, Cabredo Hofherr
this volume, Studler this volume), as well as in Upper Sorbian (Breu 2004).
As regards French-related creoles, Wespel (2008) has observed that the occur-
rence or non-occurrence of the phrase-final definite determiner la (discussed
below), correlates with the semantic contrast between what Lbner calls prag-
matic and semantic definiteness. In what follows we shall show that alongside
its enclitic marker of pragmatic definiteness, -la, Martinik has two overt mark-
ers of semantic definiteness, morphologically distinct from -la. Furthermore,
while the pragmatic and semantic definites of Germanic dialects often involve
full and morphologically reduced forms of a single diachronic source, the prag-
matic and semantic definite markers of Martinik have developed from mor-
phemes which are historically unrelated.
3 DP Syntax in Martinik: A Quick Preliminary Survey
3.1 Generalised Bare Nouns

Bare nominals are freely licensed in Martinik in argument positions regardless
of their Mass or Count denotation:
(3) a. Roch t ka tonb anl tt li.4

stone ant nonp fall on head 3sg
Stones were falling on his head.
b. Lyon danjr.
lion dangerous
Lions are dangerous.
c. Balenn s mamif.
whale cop mammal
Whales are mammals.
(4) a. Jan manj poul.

John ate chicken
John ate chicken(s).
4 Abbreviations used in the glosses: ant = anterior; cop = copula; det = determiner; dm
= demonstrative; ex = existential; fut = future; loc = locative; neg = negation; nonp =
nonpunctual (aspect); pl = plural; poss = possessive; sg = singular; spf = specific (in Gbe);
1, 2, 3 = first, second, third person.
b. Jan enmen poul.

John like chicken
John likes chicken(s).
c. T ni chyen toupatou anl masonn-nan.

ant ex dog everywhere on wall-det
There {was dog (flesh)/were dogs} all over the wall.
These examples show that bare arguments in Martinik may, context allowing,
be construed as existential or generic (in Carlsons 1977 sense) in both
subject and object positions. We shall see below that, under certain conditions,
bare nominals can also be construed as the type of semantic definites we shall
propose to call Names.
3.2 The Pragmatic Definite Determiner -LA

The item commonly acknowledged as filling the D head in Martinik is the
phrase-final morpheme -la glossed as det in our examples, which displays in
this creole some phonologically-conditioned allomorphy correlating with its
enclitic nature. This determiner is attested across all French-lexifier creoles5
5 The Gbe languages, spoken on the coastal area of Western Africa, also have a determiner
whose phonological structure, linear position in the DP, and semantic effect (pragmatic
definiteness) echo the properties of la in creole. Cf. Aboh (2001), from whom we borrow the
following example:
(i) Kk mn tv c b m n x tv l.
Koku see table 1sg-poss and say 3sg Fut buy table spf
Koku saw my table and said he would buy this table. [adapted from Aboh 2001: 11]
The converging properties of French -l and Gbe l might therefore have contributed to
the development of the strong definite determiner in Atlantic French-lexifier creoles. The
phrase-final la determiner is however also present in Indian-Ocean French-lexifier creoles
(e.g. Seychellois, Mauritiancf. Alleesaib 2012), whose substrate languages are likely to have
been different from those of Martinik. As emphasised by Chaudenson (2007), the emergence
of a property common to all French creoles is least likely to have involved a substratic
input. This diachronic problem cannot be sorted out without a fine-grained comparison of
the distribution and semantic effects of the lo/la determiners in Gbe and in Atlantic and
Indian-Ocean creoles. Should Atlantic la turn out to be semantically more Gbe-like than
Indian-Ocean la, the Gbe substrate could have influenced the recycling of French -l in one
area but not in the other. An open issue.
and has been extensively discussed in the specialised linguistic literature (cf.
Valdman 1978, Bernab 1983, Germain 1983, Gadelii 1997, Pinalie and Bernab
1999, Lefebvre 1998, Zribi-Hertz and Glaude 2007, Dprez 2007, Alleesaib 2012,
Glaude 2012, a.o.).
Martinik:
(5) a. Mari w an chat/chyen (nw).

Mary see a cat dog black
Mary saw a (black) cat/dog.
b. Mari w chat -la / chat nw-a / chyen-an.

Mary see cat -det / cat black-det / dog -det
Mary saw the/this/that {cat/black cat/dog}.
c. Mari w chat taa / chat nw taa / chyen taa.

Mary see cat dm-det / cat black dm-det / dog dm-det
Mary saw this/that cat/black cat/dog.
The creole enclitic determiner -la is historically derived from the French loca-
tive morpheme -l which in Standard French may co-occur with the demon-
strative determiner (6a), in some dialectal varieties of French also with the
definite determiner (6b), and triggers a deictic effect:
(6) a. Passe- moi ce livre- l. (Standard/Dialectal French)

pass 1sg dm book loc
Pass me that book.
b. Passe- moi le livre- l. (Dialectal French)

pass 1sg df book loc
Pass me that book.
The French phrase-final, noninflected, locative -l, as instantiated in (6), has

grammaticalised in creole into an enclitic functional morpheme which may
optionally combine with the demonstrative marker ta (cf. (5c)), but which
on its own, as in (5b), triggers the semantic effect corresponding to what
Lbner calls pragmatic definiteness. Thus the DP chat-la in (5b), like its pro-
posed English translations, points to a cat creature whose nonambiguous iden-
tification crucially depends either on deixis (the referent is in sight of the
speaker/hearer) or on the discourse contextthe referent has been previously
mentioned, or is anchored to another, previously introduced, referent. Unlike

English (and French) definite DPs, Martinik DPs headed by the -la deter-
miner cannot denote intensional kinds pertaining to all possible worlds or
inherently unique functions identified independently of the discourse
situation or context, as illustrated in (7) and (8). In (7a) and (8a), bare
nouns are the only options in Martinik to convey such intensional read-
ings; and -la naturally occurs in associative contexts such as (8c), where
the referent of the enhanced DP is d-linked to the sentence-initial loca-
tive:
(7) a. Lyon danjr.

lion dangerous
Lions are dangerous.
b. Lyon-an danjr.
lion-det dangerous
The lion is dangerous. (the aforementioned or visible lion)
(8) a. L an moun malad, yo ka kriy dokt.

when a person sick 3pl nonp call doctor
When someone is sick, you call {doctors/the Doctor}.
(the doctor category).
b. L an moun malad, yo ka kriy dokt-a.

when a person sick 3pl nonp call doctor-det
When someone is sick, you call the/this/that doctor.
(the doctor over there or previously mentioned)
c. Adan an ti vil, sa fasil trouv dokt-a.

in a small town it easy find doctor-det
In a small town, it is easy to find the doctor.
We assume that the -la determiner of Martinik carries a locative feature and
that the interpretation of -la DPs in Martinik characteristically involves the
spatial anchoring of their referent.
Note in passing that the subject DP of (9a), adapted from Krifka (1995), is
ambiguous in Martinik between a token and a type reading, just like its
English translation; a bare subject DP in this context would trigger, as in (9b), a
nonsingular existential reading, whose temporal anchoring conflicts with the
quantified temporal adverbial clause:
(9) a. Lyon-an ka gwond ly pran lod manj.

lion-det nonp growl when-3sg take smell food
The lion growls when it smells food.
(The aforementioned lion or the lion species of our world)
b. *Lyon ka gwond {ly /l yo} pran lod manj.

lion-det nonp growl when-3sg /when-3pl take smell food
Lit. Lions are growling when it/they smell(s) food.
Under our above descriptive assumptions concerning the -la determiner, the
ambiguity of lyon-an in (9a) suggests that this DP is construed as spatially
anchored regardless of the type or token construal of its lexical component.
In other words, even if lyon-an is understood as type-denoting, it is construed
as anchored to the world we live in. We assume that the contrast between
(9a)where the type-reading is licensedand (7b)where the Kind reading
is barredis syntactically correlated with the presence (9) vs. absence (7b) of
a TMA specification.
Summarising: the enclitic determiner -la, in Martinik, only triggers pragmatic

definite readings, in Lbners sense, a restriction we may correlate to the loca-
tive feature inherited from its French lexifier, possibly reinforced by the conver-
gent semantics of Gbe l (see fn. 4). -la, in Martinik, indicates that the referent
of its DP is unambiguously identified via its anchoring to the discourse situa-
tion or context.
3.3 Number
Lexical categories are uninflected in Martinik: no TMA or number inflection
on lexical roots, no morphological gender involved in agreement relations.
Functional markers are mostly, though not only (as illustrated below), realised
as free morphemes. Depending on context and lexical choices, Martinik bare
nouns may translate in English as singular semantic definites, as possible in
(8a) above, or as plural or number-neutral nominals, as in (10):
(10) Mari pt gato pi lt.

Mary bring cake and milk
Mary brought cake(s) and milk.
The Martinik lexicon however contains a plural marker for the DP, s ([se]),
which occurs prenominally and only in combination with the specific deter-
miner -la, as witnessed by the minimal pair in (11):
(11) a. Mari pt s (d) gato-a.

Mary bring pl two cake-det
Mary brought the (two) cakes (discourse-linked or visible).
b. *Mari pt s (d) gato.
3.4 Syntactic Structure

Two main lines of analysis have been explored in the syntactic literature to
account for the phrase-final linear position of the specific determiner. Under
one view (Gadelii 1997; Lefebvre 1998), the DP is parameterised as head-final in
French-lexifier creoles. The main objection to this view is that it lacks generality
since other phrases (VP, PP, CP) are overtly head-initial in these languages.
Under an alternative approach (Lyons 2000, Zribi-Hertz and Glaude 2007,
Dprez 2007) consistent with Kaynes (1994) Antisymmetry theory, phrases
are universally head-initial, and the complement of the -la determiner must
therefore raise up to the spec of DP, as shown in (12):
(12) a. lt-la the letter; s lt-la the letters
4 L(a)-N
We now turn to the l(a)- morpheme which initially motivates this work, which,
unlike the enclitic specific determiner -la discussed above, crucially occurs
as a nominal prefix. This l(a)- is historically derived from the French proclitic
definite singular article spelt out le, la or l, depending on gender specification
and on the phonological context. Although inflectional gender is absent from
creole, the French article has entered the creole lexicon by attaching to a
number of lexical roots.6 We must first distinguish the instable l(a)-prefix we
want to focus on, which only occurs on the noun in certain contexts, from
the stable word-initial l(a) syllable or segment, which occurs in some nouns
regardless of context and must therefore be regarded as part of the lexical root
rather than as a morpheme of its own. We then proceed to show that prefixed
l(a)-Ns (in short: l(a)-N) behave, syntactically and semantically, as semantic
definites denoting individual concepts, viz. individuals of type e.
A number of Martinik nouns, a sample of which are listed below in (13),
have incorporated the segment l or syllable l(a) into their lexical root:7
(13) agglutinated/stable l(a): a small sample8
MC French English MC French English

noun noun translation noun noun translation
lachas chasse hunting labitid habitude habit

laft fte party laj ge age
lafwa foi faith lajan argent money
lajl gele jail lanmou amour love
lanm mer sea ld aide help
lannuit nuit night lgzamen examen exam
lapt porte door lkl cole school
lari rue street lil le island
larivy rivire river lizin usine factory
latjizin cuisine kitchen lonn honneur honour
lavwa voix voice lous ours bear
6 This morpheme displays allomorphy characteristic of morphological attachment in Mar-

tinik: la (la-plaj (at/to/from) the beach), lan (lan-miz Misery), l (l-enjistis Injustice).
7 Baker (1984) and Ndayiragije (1984) suggest that its integration to the creole nominal lexicon
may have been favoured by a Bantu substratic influence. This line of thought is challenged by
the fact that the agglutination of the definite article is observed in both Caribbean and Indian-
Ocean French-lexifier creoles (whose substratic influences are likely to have been different,
cf. Chaudenson 2007). It has further been pointed out to us that in Portugueser-based creoles
spoken in the Gulf of Guinea, where article agglutination is also attested, it is most developed
in the creole varieties the least influenced by a Bantu substrate (thanks to Emmanuel Schang
for this latter piece of information).
8 As suggested by J. Zwarts (p.c.), it is possible that the French definite article l(a) got incorpo-
The l or la initial cannot be identified here as a morpheme since it is present

in this class of creole nouns regardless of the syntactic or discourse context:
there are no lexemes jl and ous alongside lajl jail and lous bear in Mar-
tinik:
(14) a. Ni an (nouvo) lajl/lous adan vil taa.

have a new jail/bear in town dm-det
There is a (new) {jail/bear} in this town.
b. *Ni an (nouvo) jl/ous adan vil taa.

have a new jail/bear in town dm-det
Another set of nouns, however, distinguish a bare form (N) and a prefixed
form (l(a)-N) construed as semantically definite, in Lbners sense. Nouns
exhibiting the N/l(a)-N alternation are either common nouns, as in Table
(13), or country-denoting proper names. We shall discuss each type sepa-
rately.
4.1 Common Nouns Exhibiting the N/l(a)N Alternation

4.1.1 Some Examples
The nouns involved in this alternation have various lexical meanings and mor-
phological properties.9 What they have in common is that they may a priori be
construed as sortal concepts (used to classify referents) or as individual con-
cepts (identifying singular terms). Under the sortal reading they may occur as
bare and may combine with any type of determiner (in particular with a car-
dinal, or with specific -la); under the individual reading these nouns occur
as prefixed in Martinik (l(a)-N) and cannot combine with any independent
determiner. The following examples illustrate the complementary distribu-
tions and interpretations of simplex nouns and their l(a)-prefixed counter-
parts:
rated into these creole lexemes because the definite form was the most frequent in the lexifier
language, this high frequency resulting from its semantic congruence (in Lbners sense).
9 The subtypes exemplified in (15) through (18) are mentioned by Bernab (1983) and Cervinka
(1990). The temporal subtype exemplified in (19) is mentioned in passing by Valdman (1978:
153) who illustrates it by a single example from Dominican. The instrumental subtype in
(20) is mentioned by none of these authors in relation with the l(a)-prefix. The exact lexical
extension of l(a)-prefixation in Martinik still needs to be thoroughly checked.
Types of Places10
(15) a. Ni d (*la-)pisin/(*la)plaj/(*la)fak/(*la)montann adan pyi
have two swimming-pool/beach/college/mountain in country
taa.
dm-det
There are two {swimming pools/beaches/colleges/mountains} in this
country.
b. Mari *(la-)pisin / *(la-)plaj / *(la-)fak.

Mary la-swimming-pool / la-beach / la-college.
Mary is {at the swimming pool/at the beach/in college}.
c. Mari ka rt *(la-)montann.
Mary nonp live la-mountain
Mary lives in the mountains.
d. Man pa al *(la-)pisin /*(la-)montann paske man t malad.

1sg neg go la-pool / la-mountain because 1sg ant sick
I didnt go to the {swimming-pool/mountains} because I was sick.
Abstract Properties
(16) a. Pwof taa kmt anlo (*l-)enjistis.
teacher dm-det commit a-lot injustice
Lit. This teacher committed a lot of injustices (was unfair in many
situations).
b. *(L-)enjistis s an bagay tout moun rayi.

la-injustice cop a thing everybody hate
Injustice/Unfairness is something everyone hates.
(17) a. Jan kouyon.

John stupid
John is stupid.
10 The class of l(a)-Ns illustrated in (15) interestingly seems to correspond to a type of examples
discussed for English by Birner and Ward (1994) and for French by Furukawa (2010a,b) and
Corblin (2011, 2013) under the label short weak definites (e.g. go to the beach/bank/station/
post-office/etc.). Empirical evidence however suggests that l(a)-Ns in Martinik cannot be
characterised as having variable readings, as claimed for weak short definites in French
(cf. Corblin 2011, 2013) and English (cf. Aguilar and Zwarts 2010).
b. S (*la-)kouyonni Jan ki mty konsa.

se stupidity John that put-3sg that.way
It is Johns stupidity that landed him where he is.
c. S *(la-)kouyonni ki mt Jan konsa.

se la-stupidity that put John that.way
Stupidity (viz. Human Stupidity) is what landed John where he is.
Miscel.: Law, Misery, etc.

(18) a. Yo vot an nouvo (*la-)lwa.
3pl vote a new law
They passed a new law.
b. Mari ka rspkt *(la-)lwa.

Mary nonp respect la-law
Mary respects the Law.
c. Kit y bat (*lan-)mizy.

let 3sg beat misery-3sg
Lit. Let him/her beat up his/her misery.
(Let him/her manage on his/her own.)
d. Lapli tonb anl nou kon *(lan-)miz sou l pov.

rain fall on 1pl like la-misery on l poor
Rain poured over us like Misery on the poor.
Some Temporal Nouns

(19) a. Jan pas trwa (*la-)jounen pi Mari.
John spend three day with Mary
John spent three days with Mary.
b. Jan ka dmi *(la-)jounen, i ka travay lannuit.11

John nonp sleep la-day 3sg nonp work night
John sleeps in the daytime, he works at night.
11 Unlike lajounen the daytime, lannuit (the) night, which could replace lajounen in (19b), has
an agglutinated la, as witnessed by its compatibility with a cardinal in (i) below:
(i) Jan pas d {lannuit /(*la)jounen} pi Mari.

John spend two night day with Mary
John spent two {nights/days} with Mary.
Nouns Available for Instrumental Marking

(20) a. Fk ni d (*la-)ranm /(*la)vwl adan kannt
necessary have two oar / sail in boat
taa.
dm-det
Two {oars/sails} are needed on this boat.
b. Jan ka travs lanm-a a- *(la-){ranm/ *(la-)vwl}.

John nonp cross sea-det instr la- oar/ la-sail
Lit. John is crossing the sea -la {oar/sail}.
(John is {rowing/sailing} across the sea.)
The individual-concept reading of such nouns is in various cases only licensed

under a certain theta-role (Locative, Instrumental).
4.1.2 Unstable l(a)-: Morphology

Unstable l(a)- is a word-level prefix: no lexical material may be inserted
between it and the adjacent noun stem:
(21) a. Ni d ti plaj adan vil taa.

have two small beach in town dm-det
There are two small beaches in this town.
b. Mari laplaj.
Mary la-beach
Mary is at the beach. (the type of place called Beach)
c. *Mari la ti plaj.
Mary la small beach
d. Mari anl ti plaj -la.

Mary on small beach -det
Mary is on the small beach.
The availability of l(a)-prefixation is a property of a designated set of lexemes

qualifying as nouns: thus plaj beach and pisin swimming-pool have la-forms
while sinma movies does not; doul pain, penn sorrow and miz misery
have la-forms, but bon happiness, dzspwa despair and rim remorse do
not. Non-alternating nouns crucially occur as bare under the targeted
semantic-definite reading:
(la)plaj beach [alternating N] vs. (*la-)sinma movies [non-alternating N]
(22) a. Mari al *(la-)plaj.

Mari go la-beach
Mary went to the beach.
b. Mari al sinma.
Mari go movies
Mary went to the movies.
(la)doul, (la)penn pain [alternating Ns] vs. (*la)rim remorse [non-alter-

nating N]
(23) a. Ni d (*la-)doul: doul fizik pi doul mantal.

have two pain: pain physical and pain mental
Lit. There are two pains: physical pain and mental pain.
b. *(La-)doul prany.
la- pain take -3sg
Lit. Pain took hold of him/her. ((S)he was suddenly in pain.)
(24) a. Ni d rim: rim konsyan pi rim enkonsyan.

have two remorse remorse conscious and remorse unconscious
Lit. There are two remorses (viz. types of remorse): conscious remorse
and unconscious remorse.
b. (*La-)rim prany.
la- remorse take -3sg
Remorse took hold of {him/her}. ((S)he felt a pang of remorse.)
The lexically-constrained alternation between bare nouns and l(a)-DPs under

semantic definite readings, in Martinik, in some cases echoes such minimal
pairs as (25), discussed by Carlson and al. (2006) and Klein and al. (2009) for
English:
(25) a. John dislikes going to {the store/the hospital}/listening to the radio.

b. John dislikes going to {prison/hospital}/watching TV.
The enhanced definite DPs in (25a) are ambiguous between a context-depen-

dent reading (Lbners pragmatic reading), and a context-free reading,
where the referent fails to be anchored to the specific discourse situation. The
enhanced bare singular nouns in (25b) exhibit the same semantic properties
as the definite DPs of (25a) under the context-free reading. Such examples lead
Carlson and al.s (2006) to assume that what some authors call weak definites
(cf. Poesio 1994, Barker 2005, Aguilar and Zwarts 2010, a.o.)a subtype of
Lbners semantic definites, see belowdo not necessarily contain the definite
articlein languages which have one. Under Lbners theory, the alternation
of overt and zero morphology in examples such as (25) is consistent with
the congruent character of definiteness marking whenever the lexical noun
inherently favours an individual-concept denotation.
The paradigm in (17) shows that l(a)- itself is not a nominaliser, as claimed by
Valdman (1978: 153),12 but rather selects a noun to form a semantic definite DP:
kouyon stupid > kouyon-ni stupidity (Noun, available for a functional

reading (Johns stupidity) or for pragmatic definiteness)
> lakouyonni Stupidity (semantic definite DP: the unique individ-
ual concept thus named).
4.1.3 Instable l(a)- with Common Nouns: Distribution and

Interpretation
DPs formed of nouns prefixed by l(a)- (hereunder: l(a)-N) are typically closed
to cardinals and quantity markers (e.g. anlo a lot in (16a)), as already clear from
various examples above. This constraint correlates with the fact that l(a)-N
always denotes an individual concept. It does not extend to agglutinated l(a)
nouns such as those of Table (13), which may a priori denote sortal or individual
concepts.13 Cardinality or quantity is actually the main diagnostic test we use
12 Valdman (1978: 153) analyses la- (his transcription) as a nominalising affix in the following
examples:
(i) fimen smoke (V) lafimen smoke (N)

(ii) souf to be thirsty lasouf thirst
(iii) m dead (predicate) lanm death
Such pairs are however rare. In most cases, la- attaches to lexemes available as nouns without
their prefix (e.g. jistis/lajistis justice, etc.). Valdmans analysis is further disconfirmed by
the paradigm in (17), where the nominalising affix is -ni, not la- an expected finding under
common assumptions regarding affixation.
13 Thus if lajl jail, whose initial la is agglutinated, denotes a sortal concept, it may occur as an
to tell apart prefixal l(a)- from stable l(a): thus la is stable in lajl jail since it
combines with a cardinal in (26a), and prefixal in (la-)lwa law in (18) since it
cannot cooccur with a cardinal, cf. (18a), (26b):
(26) a. Ni d lajl adan vil taa. stable la

have two jail in town dm-det
There are two jails in this town.
b. Yo vot d (*la-)lwa jodi-a. unstable la

3pl vote two law today-det
They voted two laws today.
L(a)-N is crucially incompatible with the enclitic specific determiner -la:
(27) a. Mari anl (*la-)plaj -la.

Mary on la-beach det
Mary is on the beach. (the aforementioned or visible beach)
b. Mari la-plaj (*-la).

Mary la-beach -det
Mary is at the beach. (the kind of place called Beach)
existential bare noun, as in (i-a), or combine with the -la determiner (and the plural marker),
as in (i-b); if it denotes an individual concept it is construed as a singular semantic definite,
just as l(a)-Ns whose l(a)- is prefixal (compare (ic) below with, e.g., (27b)):
(i) a. Yo ka konstrwi lajl toupatou.

3pl nonp build jail everywhere
They are building jails everywhere.
b. Adan vil taa, (s) lajol-la b gar-la.

in town dm-det (pl) jail-det near station-det
In this town the jail(s) is/are near the railway station.
c. Mari lajl.
Mary jail
Mary is in jail.
(28) a. Mari ka rspkt (*la-)lwa-a.

Mary nonp respect law-det
Mary respects the (aforementioned) law.
b. Mari ka rspkt la-lwa (*-a).

Mary nonp respect la-law -det
The complementary distribution of prefixal l(a)- and enclitic -la within a

DP domain is expected under Lbners theory, if we should identify these
two determiners as respectively congruent and non congruent: prefixal l(a)-
attaches to a noun which inherently favours an individual-concept denotation
(congruent definiteness); whereas enclitic -la signals that uniqueness is not
inherently favoured by the lexical concept identified by the noun (noncongru-
ent definiteness).
L(a)-N is incompatible with plural marking, an expected finding since l(a)-N

characteristically denotes an individual (hence singular) concept. Interestingly,
the plural marker s in Martinik must always cooccur with the specific deter-
miner in its DP (cf. section 3.3):
(29) a. Mari ka rspkt s lwa-a.

Mary nonp respect pl law-det
Mary respects the(se) laws.
b. *Mari ka rspkt s la-lwa-a.

Mary nonp respect pl la-law-det
L(a)-N is incompatible with genitive modifiers (cf. (17b), (18c)) and with restric-
tive relativisation (cf. (30)):
(30) a. Mwen pa konnt (*la)lwa-a ou ka pal-a.

1sg neg know la-law-det 2sg nonp talk-det
I dont know the law you are talking about.
b. (*La)jistis-la yo ka pratik-a pa menm-lan toupatou.

la-justice det 3pl nonp observe-det neg same-det everywhere
Lit. The justice (= the law) which is observed is not the same every-
where.
c. (*L)esklavaj-la ou ka pal-a, i analiz adan liv

slavery-det 2sg nonp talk-det 3sg analysed in book
Dlakanpann-an.
Dlakanpann-det
The slavery you are talking about, it is analysed in Delakanpanns
book.
L(a)-N is semantically definite and singular since it unambiguously identifies

an individual concept, contrasting in this respect with DPs headed by its non-
prefixed counterpart, which may be construed as indefinite and/or nonsingu-
lar. This point is shown by the translations of our previous examples and further
brought out by the paradigms in (31) through (33):
(31) a. An politik pa ni vrit.

in politics neg have truth
In politics there is/are no truth(s).
b. Tout vrit pa bon pou di.

every truth neg good to tell
Lit. Every truth isnt good to tell.
(Some truths are better left untold.)
c. I di lavrit, tout lavrit, yen ki lavrit.

3sg tell la-truth, all la-truth, nothing but la-truth
He told the Truth, the whole Truth, nothing but the Truth.
[Art teacher speaking]

d. Mwen l prsizyon pi vrit.
1sg want precision and truth
I want (some) precision and (some) truth.
(32) a. Jan pa enmen lenjistis.

John neg like la-injustice
John doesnt like Injustice/Unfairness.
b. Jan pa enmen enjistis.

John neg like injustice
Lit. John doesnt like injustices/unfairnesses.
(John doesnt like unfair decisions/situations.)
As witnessed by these latter sets of examples, the meaning of l(a)-N is con-

veyed in English either by singular definite DPs (cf. (31c)) or by bare singulars
(cf. (32a)). This brings empirical support to Carlson and al.s (2006) claim that
such pairs as the hospital (under one reading)/hospital in (25) pertain to the
same natural semantic class these authors call indefinite definites. But it is
also consistent with Lbners theory of congruence, which can easily accom-
modate bare nominals alongside proper names as a subclass of semantic def-
inites, whose inherent conceptual uniqueness needs not be overtly marked in
morphology.
The properties of (l)a-N in Martinik in some respects echo those of what
Aguilar and Zwarts (2010), Corblin (2011) and Beyssade (2013) call (short14)
weak definites, Carlson and al. (2006) indefinite definites, Furukawa
(2010a,b,) quasi intensional definitesa subtype of Lbners semantic defi-
nites, as recalled above (section 4.1.2.). Such definite DPs are characteristically
incompatible with restrictive modifiers and associated with a sloppy (vari-
able) reading under VP ellipsis. Thus, while the definite DP the store is a priori
ambiguous in (33a) between a strong reading (identifying a discourse-linked
store referent) and a weak reading (identifying the discourse-free unique type
of place also called Supermarket, where people go shopping for food), only the
strong reading is available in (33c) in the presence of the descriptive modi-
fier new; correlatively, VP ellipsis is ambiguous in (33b) between a referential
reading (where Mary and John went to the same supermarket) and a sloppy
reading (where they went to different supermarkets), while (33c) only allows
the referential reading:
(33) a. Mary had to go to the store.

b. Mary had to go to the store, and so did John.
c. Mary had to go to the new store.
d. Mary had to go to the new store, and so did John.
Similarly, l(a)-N in Martinik cannot host an adjectival modifier (21c) nor a

relative clause (30); and we further observe that l(a)-N seems to trigger variable
14 Contrastively, Long weak definites are those which contain a genitive modifier, e.g.:
(i) I met the daughter of a famous artist.
Cf. Poesio (1994), a.o., on English; Milner (1982), Flaux (1992, 1993), Corblin (passim), Furu-
kawa (passim) on French.
readings under VP ellipsis, contrasting in these respects with DPs containing

the specific determiner -la:
(34) a. Mari laplaj, Jan osi.

Mary la-beach, John too
Mary is at the beach and so is John.
[true if Mary is in Corsica and John in Martinique]
b. Mari anl plaj -la, Jan osi.

Mary on beach -det John too
Mary is on this/that beach, and so is John.
[false if Mary and John are on different beaches]
(35) a. Mari ka kout laradyo, Jan osi.

Mary nonp listen la-radio John too
Mary {listens/is listening} to the radio, and so does/is John.
[true if Mary and John listen to different radio sets or channels]
b. Mari ka kout radyo-a, Jan osi.

Mary nonp listen radio-det John too
Mary {listens/is listening} to the/that (preidentified) radio, and so
{does/is} John.
[false if Mary and John listen to different radio sets or channels]
However, the assumption that l(a)-Ns have variable readings does not fare
equally with all instances of l(a)-N in Martinik. We thus understand in (36a,b)
that Mary and John love, and England and France abolished, the same pre-
identified category. Furthermore, the variable-reading assumption does not
account for the occurrence of l(a)-N in subject position, as in (36c):
(36) a. Mari enmen lajistis, Jan osi.

Mary love la-justice John too
Mary loves Justice, so does John.
b. Langlet aboli lesklavaj, Lafrans osi.

England abolish slavery France too
England abolished Slavery, so did France.
c. Ladoul pran Mari, ladoul pran Jan.

pain take Mary pain take John
Lit. Pain took hold of Mary, pain took hold of John.
Lbners theory on the other hand allows us to propose a unified descrip-

tion of l(a)-Ns hosting common nouns, in Martinik: they all denote constant
individual concepts, whose abstract value is unambiguously retrieved from
our collective mental encyclopaedia. The variable effect, when it intuitively
seems to arise, is due to the variable setting of the situational argument: thus
both laradyo the Radio and lesklavaj (Slavery) identify constant individual
concepts, in the manner of proper names. But the variable concrete instantia-
tions (channels, programmes, etc.) of laradyo the Radio are culturally relevant,
while those of lesklavaj are not. Correlatively, we construe (35a) as true if Mary
and John are not listening to the same radio channels in their respective situa-
tional contexts, while we construe (36b) as meaning that England and France
abolished the same legal clause allowing slavery, rather than different situa-
tional variants of this clause.
As regards place-denoting l(a)-Ns, as in (34a), the assumption that they are
not construed as semantic variables in Martinik is supported by the fact that,
unlike their French and English translations, they do not occur in associative
contexts such as (37a,b), nor in cooccurrence with an indefinite genitive mod-
ifier, as in (37c,d):
(37) a. Adan an vil ki ni touris, {*laplaj/plaj-la } toujou plen

in a town that have tourist la-beach/beach-det always full
moun.
people
In a tourist resort, the beach is always crowded.
b. Amsterdam, {*lafak/fak -la} an mitan vil -la.

Amsterdam, la-uni/uni-det} in centre town -det
In Amsterdam, the uni is in the centre of town.
c. (*La-)pisin an gran lotl souvan payan.

swimming-pool a large hotel often not-free-of-charge
The swimming-pool of a large hotel is often not free of charge.
d. (*La-)fak an ti vil pwovens pa ka menn anlo tidyan

uni a small town province neg nonp attract many student
tranj.
foreign
The university of a small provincial town does not attract many foreign
students.
Associative contexts such as (37a,b) call for the discourse-linked definite

determiner -la, while contexts such as (37c,d) call for a bare head noun sup-
porting an indefinite genitive modifier.
Interestingly, for nouns alternating a bare and a l(a)- form, the l(a)- form is
preferred over the bare form in denomination predicates linked to a singular
subject, as in (38):
(38) a. Yo ka kriy la laplaj, men ou ka mand kow

3pl nonp call (t)here la-beach but 2g nonp ask yourself
poutchi.
why
They call this/that place the Beach, but one wonders why.
b. *Yo ka kriy la plaj, men ou ka mand kow poutchi.

Lit. They call this/that place beaches but one wonders why.
This supports Lbners assumption that nouns denoting individual concepts

are introduced in the syntax as default definite DPs, contra Matushanskys
(2008) claim that proper names are introduced in the syntax as predicates
rather than DPs.
4.2 L(a)-N with Proper Nouns

The analysis of l(a)-N as semantically akin to proper names in the examples
discussed above is consistent with the productive occurrence of instable l(a)-
in a subset of Martinik proper names denoting countries. The morphosyntax
of country nouns in this language is a complex issue which deserves a separate
study: we only focus here on those involving an initial l(a). These include two
subsets: in the first subset, comprising, e.g., Lafwans France, Lachin China,
Lend India,15 l(a) is present in all contexts but one: under the locative reading,
where an(n)- fills the initial syllable in the word (39f,g):
(39) a. Wo {Lafwans/Lend}, t legzil. [vocative]

oh France/India land exile
O France/India, land of exile!
b. {Lafwans/Lend} s an bl pyi. [definite subject]

France/India cop a beautiful country
{France/India} is a beautiful country.
15 Such nouns are historically derived from French monosyllabic, feminine nouns (France,
Chine, Inde, etc.).
c. Mari ka vizit {Lafwans/Lend}. [definite DO]

Marie nonp visit France/India
Mary is visiting France/India.
d. Espyon taa ka travay ba {Lafwans/Lend}. [definite PO]

spy dm-det nonp work for France/India
This spy works (is working) for France/India.
e. Ni d {Lafwans/Lend}: [sortal reading]

have two France/India
There are two France(s)/India(s).
{Lafwans/Lend} moun rich, {Lafwans/Lend} moun pov.

France/India people rich France/India people poor
The France/India of the rich, and the France/India of the poor.
f. Mari (ay) {an-Fwans /ann-End}. [locative/ Place or Goal]

Mary go an-France/ann-India
Mary is in (went to) France/India.
g. Mari soti {an-Fwans/ann-End}. [locative/ Source]

Mary return an-France/ann-India
Mary (has) returned from France/India.
h. Yo ka kriy pyi taa {Lafrans/Lend}. [denomination]

3pl nonp call country dm-det la-france/india
They call this country France/India.
(This country is called France/India)
Another subset of Martinik country nouns contain a more instable initial l(a)-
and is illustrated in (40) by the nouns meaning Germany and Sicily. These16
distinguish three context-sensitive forms: a bare form (Almn Germany, Sisil
16 The nouns of this class are historically derived from French feminine nouns pronounced as
bisyllabic (e.g. Allemagne, Sicile). Those derived from feminine trisyllables exhibit variation
among speakers (some speakers align them on bisyllables, some on quadrisyllables). Country
names derived from long French stems (4 syllables or more) are homogeneously left unpre-
fixed by all Martinik speakers, as illustrated below in (46).
Sicily), a l(a)-prefixed form (Lalmn, Lasisil), and a locative form prefixed by

an(n)- (an-Sisil, ann-Almn):
The bare form occurs in the vocative (40), in subject position (41), and if the
noun is construed as sortal (42):
(40) a. Wo {Sisil/ Almn}, t legzil!

oh Sicily/ Germany land exile
Oh {Sicily/Germany}, land of exile!
b. *O {Lasisil/Lalmn}, t legzil!
(41) a. {Sisil/Almn} s an bl pyi.

Sicily/Germany cop a beautiful country
Sicily/Germany is a beautiful country.
b. *{Lasisil/Lalmn} s an bl pyi.
(42) a. Ni d {Sisil/Almn}()
have two Sicily/Germany
There are two Sicilies/Germanies.
b. *Ni d {Lasisil/Lalmn} ()
The l(a)-form occurs in governed positionsV+O (43a), P+O (43b):
(43) a. Mari pa enmen {Lasisil/Lalmn}

Mary neg like la-Sicily/l-Germany
Mary doesnt like {Sicily/ Germany }.
b. Espyon taa ka travay ba {Lasisil/ Lalmn}.

spy dm-det nonp work for la-Sicily/l-Germany
This spy works for {Sicily/ Germany }.
The an(n)-form is definite and locative and contextually construed as Place,

Goal or Source:
(44) a. Mari {an-Sisil/ann-Almn}.

Mary an-{Sicily/ Germany }.
Mary is in {Sicily/ Germany }.
b. Mari ay {an-Sisil/ann-Almn}.
Mary go an-{Sicily/ Germany}.
Mary went to {Sicily/Germany}.
c. Mari soti {an-Sisil/ann-Almn}.

Mary return an-{Sicily/ Germany }.
Mary (has) returned from {Sicily/ Germany }.
As with alternating (N/l(a)-N) common nouns, the l(a)- form is selected over
the bare form in denomination contexts such as (45):
(45) a. Yo ka kriy pyi taa {Lasisil/Lalmn}.

3pl nonp call country dm-det la-sicily/la-germany
b. *Yo ka kriy pryi taa {Sisil/Almn}.
Lit. This country is called Sicilies/Germanies.
This contrast again supports Lbners (2011), rather than Matushanskys (2008)
syntactic analysis of proper names.
As other nouns naturally construed as individual concepts, country names
may be coerced into sortal readings by means of quantity markers or indefinite
determiners, as in (42) above. In such cases, the l(a)-prefix fails to occur, as
expected of a marker of semantic definiteness.
In contradistinction with the cases discussed above, a third set of country

nouns never involve an initial l(a)-, e.g.: Patagoni Patagonia or Endonzi
Indonesia, as illustrated in (46):
(46) a. Wo {Patagoni/Endonzi}, t legzil. [vocative]

oh Patagonia/Indonesia land exile
O Patagonia/Indonesia, land of exile!
b. {Patagoni/Endonzi} s an bl pyi. [definite subject]

Patagonia/Indonesia cop a beautiful country
Patagonia/Indonesia is a beautiful country.
c. Mari ka vizit {Patagoni/Endonzi}. [definite DO]

Marie nonp visit Patagonia/Indonesia
Mary is visiting Patagonia/Indonesia.
d. Espyon taa ka travay ba {Patagoni/Endonzi}. [definite PO]

spy dm-det nonp work for Patagonia/Indonesia
This spy works (is working) for Patagonia/Indonesia.
e. Ni d {Patagoni/Endonzi}: [sortal reading]

have two Patagonia/Indonesia
There are two Patagonia(s)/Indonesia(s).
{Patagoni/Endonzi} moun rich, Patagoni/Endonzi moun pov.

Patagonia/Indonesia people rich Patagonia/Indonesia people poor
The Patagonia/Indonesia of the rich, and the Patagonia/Indonesia of
the poor.
f. Mari (ay) {Patagoni/Endonzi}. [locative: Place/Goal]

Mary go Patagonia/Indonesia
Mary is in (went to) Patagonia/Indonesia.
b. Mari soti {Patagoni/Endonzi}. [locative/ Source]

Mary return Patagonia/Indonesia.
Mary (has) returned from Patagonia/Indonesia.
The alternation of zero morphology and l(a)- marking on definite country

names echoes the alternation of zero and l(a)- marking on Martinik common
nouns discussed in section 4.1 ((La)doul/Rim), and the alternation of zero
and definite marking in English nominals construed as semantically definite
(to jail/to the store; Italy/the Strand).
4.3 Recap: L(a)-Ns, Singular, Semantic Definite DPs

We propose to analyse the nominal prefix l(a)- of Martinik as a semantic defi-
niteness marker selecting for its complement an individual term, in Lbners
(1985, 2011) senseeither common: plaj beach or proper: sisil Sicily. More
precisely, l(a)-N denotes a singular individual concept which unambiguously
identifies a referent independently of the specific discourse situation in which
it occurs. In this respect, l(a)-Ns have the semantic properties of singular defi-
nite proper names, regardless of the common or proper nature of their head
noun: this accounts for the capitalised initials in our English translations of
l(a)-N in section 4.1. As a word-level prefix, l(a)- is always adjacent to its noun
stem and must be listed as an available option for designated lexical entries. As
shown by such examples as those in (17), l(a)- does not itself carry the categorial
feature n, but takes a noun as its complement to derive a semantic definite
DP instantiating what we propose to call a Namea DP denoting an individual

concept of type e. Correlatively, l(a)- is incompatible in its DP with the prag-
matic definite determiner -la (which signals noncongruent definiteness) and
with any expression requiring a sortal or relational construal of the noun (e.g.
cardinals, demonstrative, genitives, relatives).
5 L-NP
5.1 The Collective Predeterminer l

We found the Martinik determiner morpheme l17 mentioned in two text-
books bearing on French-lexifier grammars: Bernab (1983) and Germain
(1983). L shares with l(a)- some properties characteristic of semantic definites,
but differs from l(a)- as regards morphology (l with common nouns is not a
prefix, but a free morpheme) and interpretation (l has a collective flavour
while l(a)- serves to denote an atomic concept). The triplet of examples in (47),
where l+NP is contrasted with s+NP-la and bare nouns in the argument of an
existential predicate, gives a first hint of the semantics of l:
(47) a. An dfil-a, t ni ponpy, majort, konsy-minisipo.

in parade-det ant have fireman majorette councillor-town
In the parade there were firemen, majorettes, town councillors.
b. An dfil-a, t ni s ponpy-a, s majort -la, s

in parade-det ant have pl fireman-det pl majorette det pl
konsy-minisipo-a
councillor town-det
In the parade there was: the(se) firemen, the(se) majorettes, the(se)
town councillors.
c. An dfil-a, t ni l ponpy, l majort, l konsy-minisipo

In the parade there was: the Firemen, the Majorettes, the Town Coun-
cillors.
In (47a) the bare nouns receive an existential Kind reading (in Carlsons 1977
sense), as their analogues in the English translations: each enhanced DP
denotes an undefined quantity of entities respectively assigned to the fire-
17 The grammar of l exhibits some variation among Martinik speakers: the l we describe is
regarded as typical of the Northern variety.
man, majorette and town councillor categories. In (47b), the DPs overtly
specified as plural (s) and specific (-la) are construed as pragmatic defi-
nites: they refer to three sets of entities crucially identified via anchoring to
the discourse or situation context: the aforementioned firemen/town council-
lors/majorettes, those of the town where the parade took place, or those in sight
of the speaker and hearer. In (47c) the enhanced l+NPs denote three collective
entities (groups, in Landmans 1989 terminology) whose unambiguous iden-
tification is crucially independent of the discourse or situation context: the
Fire Brigade, the Majorettes, the Town Councillors are three unique group
concepts listed as such in our mental encyclopaedia, and which may be instan-
tiated in any human collectivity. The group intuition is consistent with the col-
lective label Bernab (1983) attaches to the l determiner. Calling l a collective
determiner on the other hand fails to capture the semantically definite reading
of its including DP, a property common to l and l(a)-, which we signal above
and below in our translations of l+NP by capitalising the initial of the N head.
We now present in further detail the morphosyntactic properties of the seman-

tic definite determiner l and of the DPs it occurs in.
5.2 Morphology
Unlike instable l(a)-, l with common nouns is not a prefix, since lexical mate-
rial may be inserted between it and the following noun:
(48) a. An dfil-a, l vy ponpy t douvan, l jenn ponpy

in parade-det l old fireman- ant in-front l young fireman
t dy.
ant in-back
In the parade, the Old Firemen were in front, the Young Firemen were
behind.
b. Sinma, l dzym wl ka genyen mwens lajan ki l

movies l minor role nonp earn less money than l
prmy wl.
leading role
In the movies, the Minor Roles earn less money than the Leading
Roles.
Further evidence of the nonprefixal status of l in such examples is the fact

that unlike l(a)-, l is not lexically restricted (it may combine with any noun a
priori open to a discontinuous denotation) and exhibits no sandhi signalling
rightward attachment, as witnessed by the contrast between (49) and (50):
(49) a. la + jistis > lajistis

justice Justice
b. la + enjistis > *la enjistis >lenjistis
injustice, unfairness Injustice
c. la + miz > *lamiz >lanmiz
misery Misery, Poverty
(50) a. l + ponpy > l ponpy

fireman the Fire Brigade
b. l + tidyan > l tidyan *>ltidyan/lztidyan
student the Students
c. l + mb > l mb *>len mb
piece of furniture the Furniture18
5.3 Distribution
L+NP denotes an animate group in several of the above examples, but it may
also freely denote inanimate groups:
(51) Adan an jaden potaj, an pwensip l tomat kay an soly

in a garden vegetable in general l tomato go in sun
pi l jiromon alonm.
and l pumpkin in-shade
Lit. In a vegetable garden, the Tomatoes (should) generally go in the
sun and the Pumpkins in the shade.
As pointed out by Bernab (1983), l readily combines with patronyms, as in

(52a); but patronyms may also combine with the plural marker s (and the spe-
cific definite determiner -la): l Lakwa the Lakwas in (52a) is construed as
denoting a family unambiguously identified by the Lakwa patronym indepen-
dently of the discourse context, while s Lakwa-a in (52b) denotes a specific set
of spatiotemporally anchored members of the Lakwa family (the aforemen-
tioned Lakwas, or the Lakwas of our town or neighbourhood):
(52) a. L Lakwa jadiny dpranfis.

l Lakwa gardener for generations
The Lakwas have been gardeners for generations.
18 Mb furniture is a Count noun in Martinik, as its French lexifier meuble.

b. S Lakwa-a jadiny dpranfis.

pl Lakwa-det gardener for generations
The(se) Lakwas have been gardeners for generations.
L+NP may also naturally adjoin to a plural deictic pronoun, as in (53c), in

which case it contrasts with both bare nouns and s N-la:
(53) a. Zt, tidyan, zt t pou konprann.

2pl student 2pl ant mod understand
You students should have understood.
b. Zt, s tidyan-an, zt t pou konprann.

2pl pl student-det 2pl ant mod understand
You the students (of this place/in question), you should have under-
stood.
c. Zt, l tidyan, zt t pou konprann.

2pl l+tudiant 2pl ant mod understand
You the Students (as opposed to, e.g., the Faculty), you should have
understood.
Last but not least, l productively co-occurs with an adjective licensing an

elliptical noun to denote a common-ground group concept unambiguously
identified by the property conveyed by the adjective, independently of the
discourse context:
[Discussing ponies]
(54) a. S gran -an mwen rapid ki s piti-a.

pl big -det less fast than pl small-det
The big ones (among the aforementioned) are slower than the small
ones.
b. An pwensip, l gran mwen rapid ki l piti.

in principle l big less fast than l small
As a rule, big ones are slower than small ones.19
19 Interestingly, Martinik elliptical DPs such as l gran translate in English as bare nominals
(big ones), while the article triggers a pragmatic (d-linking) effect in English (the big ones). In
5.4 Interpretation
The interpretive properties of l+NP mayfor descriptions sakebe decom-
posed into four ingredients: (i) Definite; (ii) Plural; (iii) Group; (iv) Antispecific.
Below we bring out each property separately.
5.4.1 L+NP Is Construed as Definite

The definite effect of l is brought out by such minimal pairs as (55) (a short-
ened variant of (47)), where l-NP is contrasted with a bare noun:
(55) a. An dfil-a, t ni ponpy pi majort.

in parade-det ant have fireman and majorette
In the parade there were firemen and majorettes.
b. An dfil-a, t ni l Ponpy pi l Majort.

in parade-det ant have l fireman and l majorette
In the parade there was: the Firemen and the Majorettes.
In (55a), the bare nominals are construed as existential Kinds, in Carlsons

(1977) terms. In (55b) l leads us to construe the enhanced DPs as names of
common-ground collective entities unambiguously identified independently
of the discourse context, and expected to be uniquely instantiated in any town
parade: the Fire Brigade and the Majorettes.
5.4.2 L+NP Is Construed as Plural

The fact that the pronouns coreferring with l Rigob the Rigoberts in (56)
surface as yo (3pl) rather than i (3sg) is evidence that l+NP is construed as
a set of atomic entities:
(56) Chrch-a f an tid anl l Rigobz.

scientist-det make a study on l Rigobert
I touv k {yo/*i}z ka pt jenn-nan.
3sg find that 3pl/3sg nonp carry gene-det
The scientist made a study on the Rigobz(family).
(S)he found that theyz carry (*it carries) the gene under discussion.
French, plural elliptical definite DPs (les grands (the) big ones) are ambiguous between the
d-linked reading conveyed by s gran-an in Martinik, and the non-d-linked reading conveyed
by l gran.
Men l Rigobz pa sav ki {yo/*i}z ka pt jenn taa.

but l Rigobert neg know that 3pl/3sg nonp carry gene dm-det
But the Rigobz (family) are not aware that theyz carry this gene.
5.4.3 L +NP is Construed as a Group (Collective Reading)

L is the only determiner option in DPs intended to denote common-ground
group concepts, prototypically exemplified by music bands, as in (57), or by
the social classes in Karl Marxs classification (58):
(57) a. L Lopar s {mizisyen/an gwoup} Senpy.

l Leopard it(is) musician/a group St-Pierre
The Leopards are {musicians/a group} from Senpy.
b. #S Lopar-la s {mizisyen/an gwoup} Senpy.

pl Leopard-det it(is) musician/a group St-Pierre
Lit. These Leopards are {musicians/a group} from St-Pierre.
(58) a. Lta ni klas labory pi klas profit.

State have class working and class exploiting
The State comprises working classes and exploiting classes.
b. #Lta ni s klas labory-a pi s klas profit-a.

State have pl class working-det and pl class exploiting-det
Lit. The State comprises these working classes and these exploiting
classes.
c. Lta ni l klas labory pi l klas profit.

State have l class working and l class exploiting
The State comprises: the Working Classes and the Exploiting Classes.
(ex. inspired by Landmans 1989 Das Kapital series)
L is conversely barred from the argument of a Kind-selecting predicate such

as the one translating to become extinct:
(59) a. Konsyj ka dispart.

janitor nonp disappear
Concierges [French janitors] are about to be extinct.
b. *L Konsyj ka dispart.
In DP positions contextually consistent with either collective or distributive

readings, l+NP unambiguously selects the collective (Group) reading. Plural
s -la DPs are, contrastively, ambiguous between the collective and the
distributive readings:
(60) a. S lengwis -la ni plis ki 8,000 liv an-prs.

pl linguist -det have more than 8,000 book in press
The linguists under discussion have over 8,000 books in press.
[ambiguous: 8,000 altogether or 8,000 each]
b. L lengwis ni plis ki 8,000 liv an-prs.

l linguist have more than 8,000 book in press
The Linguists [contrasting with other similar scholarly groups] have
(collectively) over 8,000 books in press.
(61) a. S Lakwa-a pi s Rigob -a pa ka antann.

pl Lakwa-det and pl Rigob-det neg nonp get.along
The(se) Lakwas and the(se) Rigobs dont get along.
[amgiguous: individuals or groups]
b. L Lakwa pi l Rigob pa ka antann.

The Lakwa family and the Rigob family dont get along. (group
reading)
L+NP is barred whenever any expression in the context forces the predicate to
be construed as distributive, e.g.:
ant yo (a reciprocity marker) in (62b):
(62) a. S Rigob -a pa ka antann ant yo.

pl Rigobert det neg nonp get.along between 3pl
The Rigoberts do not get along with each other.
b. *L Rigob pa ka antann ant yo.
pys (negative-polarity distributive quantifier: none of the X) in (63b):
(63) a. Pys s Rigob -a pa ka antann pi nouvo m-a.

none pl Rigobert-det neg nonp get.along with new mayor-det
None of the Rigobs gets along with the new mayor.
b. *Pys l Rigob pa ka antann pi nouvo m-a.
chak (positive-polarity distributive quantifier: each) in (64b):
(64) a. Adan dfil-a, s ponpy-a, yo chak t ka tchenb an

in parade-det pl fireman-det 3pl each ant nonp hold a
kstenkt.
extinguisher
Lit. In the parade, the firemen, they were each holding a fire extin-
guisher.
b. *Adan dfil-a, l ponpy, yo chak t ka tchenb an kstenkt.
5.4.4 L+NP Is Antispecific

L+NP is infelicitous if the identification of the referent is dependent on the dis-
course context or situation. Thus, l goes unlicensed in (65) since the firemen
involved in a completed event necessarily denote a specific bunch of people
anchored in space and time:
(65) a. S ponpy-a ja tir nich mouchamyl -la an

pl fireman-det already remove nest honey-bee -det from
fitay kay -la.
roof house -det
The firemen have already dislodged the honey-bee nest from the roof
of the house.
b. *L Ponpy ja tir nich mouchamyl -la an fitay kay -la.
L is incompatible in its DP with the specific determiner -la:20
20 In some idiolectal varieties of Martinik, l is but a morphological variant of the plural marker
spelt out as s in mainstream Martinik:
(i) a. L boug-la vini.

guy -det come
(Sainte-Marie speaker; quoted by Bernab 1983: 648)
b. S boug-la vini.
pl guy -det come
The(se) guys came/arrived.
(mainstream Martinik; our own example)
(66) a. S lengwis*(-la) ni plis ki 8,000 liv an prs.

pl linguist-det have more than 8,000 book in press
The(se) linguists have over 8,000 books in press.
b. L lengwis (*-la) ni plis ki 8,000 liv an prs.

The Linguists (collectively) have over 8,000 books in press.
L+NP cannot host a cardinality marker:
(67) a. S senk ponpy-a ka mach adan dfil-a.

pl five fireman-det nonp march in parade-det
The five firemen are marching in the parade.
b. *L senk ponpy ka mach adan dfil -a.
With elliptical Ns, l+NP is construed as discourse-free (antispecific), contrast-

ing with specific s la, cf. (53) above.
L+NP cannot host a restrictive relative clause, as illustrated below by (69c),

contrasting with (68b) and (69b):
(68) a. Pou jw j taa fk spar s chif -la pi s

to play game dm-det must separate pl number -det and pl
koul -a.
face-card -det
To play this game you must separate the (those) numbers (in your
pack) from the (those) face cards (in your pack).
b. Pou jw j taa fk spar l chif pi l

to play game dm-det must separate l numbers and l
koul.
face-cards
To play this game you must separate the Numbers from the Face
Cards.
In the grammar we are describing, s and l are not free variants of the plural marker, as
in (i) above, but in strict complementary distribution.
(69) a. Pou jw j taa fk spar kat ki pi

to play game dm-det must separate card which(are) more
piti ki 7 di kat aparti di 7.
small than 7 from card from-7-up
To play this game you must separate cards below 7 from cards from 7
up.
b. Pou jw j taa fk spar s kat-la ki pi

to play game dm-det must separate pl card-det which more
piti ki 7-la di s lzt-la.
small than 7-det from pl the.other(s)-det
To play this game you must separate the cards below 7 from the other
ones.
c. *Pou jw j taa fk spar l kat ki pi

to play game dm-det must separate l card which(are) more
piti ki 7 di l kat apati di 7.
small than 7 from l card from-7-up
L+NP triggers what looks like narrow-scope effects in such examples as (70):
(70) a. L dif pri, fk kriy l ponpy.

when fire break-out must call l fireman
When (a) fire has broken out, one must call the Fire Brigade.
b. L dif pri, fk kriy s ponpy-a.

when fire break-out must call pl fireman-det
When (a) fire has broken out, one must call the(se) firemen.
Unlike (70b), (70a) does not need us to understand that the same actual indi-
viduals are called upon for every fire. This contrast, however, would follow
from the assumption that the definite DP has in both cases wide scope over
the universal quantifier but denotes a specific entity in one case (70b) and
a common-ground category in the other (70a)an assumption in keeping
both with Aguilar and Zwartss (2010) and Beyssades (2013) analysis of weak
definites as Kind-denoting, and with Lbners theory of definiteness predict-
ing the congruent compatibility of definiteness markers with individual con-
cepts.
However, all instances of l are not associated with the seemingly narrow-
scope effect we get in (70a): thus, l Lopar the Leopards (a music band from
Martinique) and l Lakwa the Lacroixs, as definite proper names, have a con-
stant value (wide-scope effect) in a range of situations.
5.5 L+NP in Denomination Predicates

L+NP is selected over bare nouns in denomination predicates whose argu-
ment denotes a group:
(71) a. Yo ka apl group taa L Ponpy.

3pl nonp call group dm-det l fireman
This group is called the Firemen.
b. *Yo ka apl group taa ponpy.

Lit. This group is called firemen.
(72) a. *Yo ka apl s moun taa L Ponpy.

b. Yo ka apl s moun taa ponpy.
3pl nonp call pl person dm-det fireman
These people are called firemen.
These examples are parallel to those in (38) and (45) involving l(a)-N.
5.6 L+NP: Recap

The above data show that l is a free morpheme which signals its DP as defi-
nite but antispecific (semantically definite in Lbners sense), and plural but
group-denoting. The collective effect may be derived from the fact that the
head noun of l+NP (as the head noun of la-N) denotes an individual concept,
rather than a sortal concept; l+NP however contrasts with la-N in that the
group-denoting individual concept contains a plurality of members. The prop-
erties of l+NP in Martinik echo those of French definite plural proper names
such as les Etats Unis (the United States):21
21 L actually also occurs in some Martinik translations of plural country names, in which case
it exhibits a prefixal morphology, signalled by external sandhi:
(i) L-ztazini s an bl pyi.

l-United.States cop a beautiful country
The United States are a beautiful country.
This subclass of cases is left out of the present article for lack of space, but it is quite consistent
with the general analysis we propose for l(a)- and l.
(73) a. *L/Cet Etat Uni a besoin dargent.

df.sg/dm.msg State United have.prs.3sg need of money
Lit. The/this United State needs money.
b. Les Etats -Unis s{ont/*est} un grand pays,22

df.pl United States be.prs.3pl/3sg a big country
et {ils ont/*il a} besoin dargent.
and {3MPL have.prs.3pl/3msg have.prs.3sg} need of money
Lit. The United States are a big country and they need money.
[Plural]
c. *Les Etats-Unis sont en comptition les uns avec

df.pl States United be.prs.3pl in competition df.pl ones with
les autres.
df.pl others
Lit. The United States compete with each other.
[Collective]
d. #Les Etats-Unis que Marie prfre sont riches.

df.pl States United that Mary prefer.prs.3sg be.prs.3pl rich.pl
The United States that Mary prefers are rich.
[Collective]
e. *Ces Etats-Unis-l sont riches.

dm.pl States United-loc be.prs.3pl rich.pl
These/those United States are rich.
22 Standard French contrasts in this respect with Standard English, where singular agreement
is recommended with the subject the United States (> the United States is a big country). Cf.
http://www.visualthesaurus.com/cm/wordroutes/the-united-states-is-or-are/.
6 Conclusion: L(a)- and l as Name Markers
The properties brought out for la-N and l+NP in Martinik lead us to identify
l(a)- and l as markers of semantic definiteness, in Lbners (1985, 2011) sense:
they both signal the unambiguous identification of the referent of their DP as
an individual concept. L(a)- and l differ morphologically in that l(a)- is a lexi-
cally constrained prefix, while l occurs as a free morpheme, except with coun-
try names. And they differ semantically as to number specification: l(a)-N is
straightforwardly singular, while l+NP is both plural and singular: plural since
it triggers plural anaphora (licensed by the plurality of the groups members),
but singular since the group itself stands as a singular entity whose members
are not accessible for distributivity. Some instances of l(a)-N and l+NP seem to
trigger sloppy readings under VP ellipsis, a property regarded in the linguistic
literature as characteristic of short weak definites, but it turns out neither l(a)-
nor l actually license variable readings, as especially witnessed by their com-
patibility with proper nouns (Lasisil, L Lopar). We propose that l(a)- and l
are best characterised as markers of semantic definiteness identifying individ-
ual concepts, as opposed to sortal, functional and relational concepts. In this
respect, l(a)- and l together contrast with the phrase-final enclitic determiner
-la, which signals pragmatic definiteness; and la-N and l+NP are semantically
similar to what is commonly called proper names, regardless of the common
or proper lexical nature of their head noun. We therefore propose to charac-
terise l(a)- and l as Name markers:
(74) a. Mari ka respekt Lalwa. [la+ common n]23

b. Mari pa enmen Lasisil. [la+ proper n]
Mary does not like Sicily.
c. () fok kriy l Ponpy. [l+ common n]
One must call the Fire Brigade.
d. Mari enmen l Lopar. [l+ proper n]
Marie likes the Leopards.
As shown in section 4, singular Names in Martinik may surface as l(a)-N or

as bare nounsa lexical variation. A similar variation is observed in English
between singular Names morphologically marked as definite ((tell) the Truth,
23 The official spelling rule for la-N and l+NP varies in the textbooks we consulted, but we take
it upon ourselves to capitalise their initial to signal their Name status.
(go to) the Beach; (visit) the Taj Mahal, (go to) the Strand), and Names occurring
as bare ((watch) TV, (go to) Jail; (meet) John, (visit) France). Plural Names, on
the other hand, always involve overt definiteness markingby l in Martinik
(l Ponpy the Firemen, l Lopar the Leopards, l Ztazini the United States),
and by the in English (the Working Classes, the Beatles, the Kennedys, the United
States).
The existence of three definiteness markers in Martinik-la, l(a)- and
lcalls for a revision of the DP structure proposed in (12) acknowledging
Names as a type of definite DPs. We submit the derivations in (75) and (76):
(75) Deriving l(a)-N and l+NP (Names)24
df = definiteness NmP = Name Phrase

NbP = Number Phrase nP = noun Phrase (categorial phrase)
24 The diagram in (75) leaves out country names (e.g. l Ztazini), where l must be anaysed as
a prefix, on a par with l(a)-.
(76) Deriving (s) NP-la (pragmatic definite DPs)
df+loc = pragmatic definiteness (spatiotemporal anchoring)
Under the proposed analysis, Names are DPs characterised by (i) the occur-
rence of a specialised (Name) head (Nm) selected by an antispecific (-loc)
definite feature in D, and which takes nP as its complement; and (ii) a set
value (-pl or +pl) for the Number head: the Nm marker is spelt out l(a)- in
the singular, l in the plural. L(a)-, being a prefix, may only take designated
lexical nouns as its complement, while l, as a free morpheme, may a priori
combine with any noun whose semantic construal allows the group effect to
obtain. After attachment, l(a)-N raises up to D, in the spirit of Longobardis
(1994) analysis of bare proper names, surfacing as Laplaj, Lasisil, etc. If Number
is specified as -pl but l(a)- fails to occur in Nm (lexical restriction for Rim
and Pagatoni, syntactic restriction for Sisil), the Name ends up in D with no
prefix (e.g. Rim, Sisil, Patagoni). L is identified in (75) as a Name marker (like
l(a)-);25 however, as a free morpheme, l raises up to D and leaves the noun
below.26 Under the analysis proposed in (75), l(a)- and l are restricted not only
25 This analysis sheds light on the dialectal variation concerning l mentioned in fn. 17 and 20.
In the Northern variety of Martinik we describe, l is merged in Nm, hence selected by a
nonlocative definite in D and incompatible with -LA, and triggers a collective (group) effect
due to the Name head (which restricts denotation to an individual concept). In varieties of
Martinik where l combines with -LA, it is merged in Nb, hence stands as an individual or
dialectal variant of s (cf. section 3.3) in pragmatic definite DPs.
26 We could alternatively assume that l, specified as a [+df, -loc], is directly merged in a D
head whose complement includes the +pl value in Nb and a NmP projection below. Our
to semantically definite DPs (hosting no Locative feature in D), but more pre-
cisely to semantic definite DPs containing a Name projection, hence construed
as denoting individual concepts rather than sortal or relational concepts. These
properties account for the ban on Quantity markers and restrictive modifiers
in la-N and l+NP, and on distributivity for l+NP, since Quantity, restrictive
modification and distibutivity correlate with a sortal or relational construal of
the noun. Assuming that pragmatic definiteness in Martinik is an effect of the
Locative feature in D, we tentatively assume that this feature is what triggers
the raising of NbP to spec, DP in specific (pragmatic definite) DPs: this move-
ment therefore takes place in (76) but not in (75).
The Martinik data presented in this study bring empirical support to the
conceptual distinction drawn by Lbner (1985, 2011) between semantic and
pragmatic definiteness. Like the West-Germanic dialects reported to distin-
guish strong (morphologically full) from weak (morphologically reduced)
definite articles respectively conveying these two types of definite interpreta-
tions (cf. Ebert 1970, Lbner 1985, 2011, Schiering 2002, Studler 2008, Cabredo
Hofherr this vol), the morphology of Martinik distinguishes pragmatic from
semantic definiteness. However, the morphemes l(a)- and l of Martinik are
not licensed in the whole range of semantically definite DPs, but only in a
subset of them denoting individual concepts in the manner of definite proper
names. The assumption that l(a)-N and l+NP contain a Name projection in
their syntactic representation aims at capturing this restriction in the syntax.
Under our analysis, Names thus instantiate a subtype of definite DPs charac-
terised by the presence of a special Name projection, regardless of the lexical
features merged in the n-head (e.g. Sicily or beach). The presence of the Name
phrase may be made morphologically visibleas by l(a)- or l in Martinik
but needs not be, as witnessed by the many examples of bare Names in var-
ious languagesincluding English, French and Martinik itself. While bare
Names always seem construed as semantically singular, at least in the lan-
guages under discussion, plural Names seem to require some overt functional
marking (English the, French les, Martinik l), a restriction reflecting their
marked nature correlating with a collective effect. The fact that both l(a)-N
and l+NP are selected over their bare counterparts in denomination contexts
(cf. (38), (71)) supports Lbners (2011) analysis of proper names, once merged
in syntax, as default definite DPs, rather than Matushanskys (2008) claim that
proper names enter the syntax as bare denominating predicates.
reason for preferring (75) is that it straightforwardly captures the parallel natures of l(a)- and
l, beyond their different surface positions in the structure.
The data presented in this study show that the morphosyntax of definite-
ness is radically different in Martinik and in French, the lexifier language.
While French uses a single definite article to convey pragmatic and seman-
tic definitenessa reminder of this morphemes pronominal origin,27 Mar-
tinik has developed a marker of pragmatic definiteness-lahistorically
derived from a French deictic locative, and two Name markersl(a)- and l
which have retained the uniqueness presupposition of their lexifier (the French
definite article) but not its pronoun-inherited anaphoric feature.
Should French and Martinik be equally listed in typological charts as Lan-
guages With Definite Determiners? We let typologists ponder over this practi-
cal issue.
References
Abbott, Barbara. 1999. Support for a unique theory of definite descriptions. In Pro-
ceedings of SALT 9, edited by Tanya Matthews and Devon Strolovitch, 115. Amherst
University: GLSA.
Aboh, Enoch Olad. 2001. La morphosyntaxe de la priphrie gauche nominale.
Recherches Linguistiques de Vincennes 31: 926.
Aguilar, Ana, and Joost Zwarts, 2010. Weak definites and reference to kinds. In Pro-
ceedings of SALT 20, edited by Nan Li and David Lutz, 179196. Ithaca, NY: Cornell
University.
Alleesaib, Muhsina. 2012. DP syntax in Mauritian. Doctoral dissertation, Universit
Paris-8.
Baker, Philip. 1984. Agglutinated French articles in Creole French: their evolutionary
significance. Te Reo 27: 89129
Barker, Chris. 2005. Possessive weak definites. In Possessives and beyond: semantics
and syntax, edited by Ji-Yung Kim, Yury Lander and Barbara Partee, 89113. Univer-
sity of Massachusetts: BookSurge Publishing.
Bernab, Jean. 1983. Fondal-natal. Paris: LHarmattan
Beyssade, Claire. 2013. Back to uniqueness presupposition: the case of weak definites.
Talk presented at the LSALAA workshop, Paris, 1 March 2013: Centre Pouchet, CNRS.
Birner, Betty; and Gregory Ward, 1994. Uniqueness, familiarity, and the definite article
in English. Berkeley Linguistics Society Annual Meeting (BLS) 20: 93102.
27 Corblin (2013) proposes a unified analysis of the French paradigm, based on the anaphoric
nature of the French definite article. This property has clearly not carried over to creole.
Breu, Walter, 2004. Der definite Artikel in der obersorbischen Umgangssprache. In

Slavistische Linguistik 2002, edited by Marion Krause and Christian Sappok, 957.
Cabredo Hofherr, Patricia. This volume. Reduced definite articles with restrictive
relative clauses, 172211.
Carlson, Greg. 1977. A unified analysis of the English bare plural. Linguistics and
Philosophy 1: 413457.
definite noun phrases. In Proceedings of NELS 36, edited by Christopher Davis, Amy
Rose Deal, and Youri Zabbal. University of Massachusetts/Amherst.
Cervinka, Bernadette. 1990. Quelques remarques propos du morphme la- dans
les couples du type t/lat, lin/lalin, miz/lanmiz, etc. (crole martiniquais). Mas-
ters thesis, Universit de Pointe--Pitre, Guadeloupe. (Published in Espace Crole 9
http://www.potomitan.info/travaux/espacecreole/morpheme.htm)
Chaudenson, Robert. 2007. Le substrat dans la crolisation: mythes et ralits. In
Grammaires croles et grammaire comparative, edited by Karl Gadelii and Anne
Zribi-Hertz, 2748. Saint-Denis, France: Presses Universitaires de Vincennes.
Confiant, Raphal. 2007. Dictionnaire du crole martiniquais. http://www.potomitan
.info/dictionnaire/
Corblin, Francis. 1987. Indfini, dfini et dmonstratif. Constructions linguistiques de la
rfrence. Geneva: Droz.
Corblin, Francis. 2001. Dfini et gnitif: le cas des gnitifs dfectifs. In Cahiers Jean-
Claude Milner, edited by Jean-Marie Marandin, 1954. Paris: Verdier.
Corblin, Francis. 2011. Des dfinis para-intensionnels: tre lhpital, aller lcole.
Langue franaise 171: 5575.
Corblin, Francis. 2013. Locus et telos: aller lcole, tre la plage. Corela, special issue
on spatial expressions in French edited by Benjamin Fagard and Dejan Stosic, http://
corela.edel.univ-poitiers.fr/index.php?id=2722. Universit de Poitiers.
Dprez, Viviane. 2007. Probing the structuring role of grammaticalization: nominal
constituents in French-lexifier creoles. Journal of Pidgin and Creole Languages 22-2:
263307.
Ebert, Karen. 1970. Zwei Formen des bestimmten Artikels. In Probleme und Fortschritte
der Transformationsgrammatik, edited by Dieter Wunderlich, 159173. Tbingen:
Max Hueber Verlag.
Ebert, Karen. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem nord-
friesischen Dialekt (Fehring). Brist/ Bredstedt: Nordfriisk Institut.
Flaux, Nelly. 1992. Les syntagmes nominaux du type le fils dun paysan: rfrence
dfinie ou indfinie? Le franais moderne 1: 113140.
Flaux, Nelly. 1993. Les syntagmes nominaux du type le fils dun paysan: rfrence
dfinie ou indfinie? Le franais moderne 2: 2345.
Furukawa, Naoyo. 1986. Larticle et le problme de la rfrence en franais. Tokyo: France

Tosho.
Furukawa, Naoyo. 2010a. Larticle dfini et le problme dit de lunicit: quantit ou
qualit? Bulletin dtudes de Linguistique Franaise 44: 6582.
Furukawa, Naoyo. 2010b. Article dfini, son emploi intensionnel et nonc tautolo-
gique. Bulletin dtudes franaises 41: 5171.
Gadelii, Karl. 1997. Lesser Antillean French creole and Universal Grammar. Gteborg
University: Gothenburg Monographs in Linguistics 15.
Germain, Robert. 1983. Grammaire crole. Paris: LHarmattan.
Glaude, Herby. 2012. Aspects de la syntaxe de lhatien. Doctoral dissertation, Univer-
sit Paris-8.
Hawkins, John. 1978. Definiteness and indefiniteness: a study in reference and grammat-
icality. London: Croom Helm.
Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Ph.D diss.,
University of Massachusetts.
Kadmon, Nirit. 1990. Uniqueness. Linguistics and Philosophy 13: 273324.
Klein, Natalie, Whitney Gegg-Harrison, Greg Carlson, and Michael Tanenhaus. 2009.
Special but not unique: weak definite noun phrases. In Semantics and Pragmatics:
from experiment to theory, edited by Ulrich Sauerland and Kazuko Yatsushiro, 264
275. London: Palgrave Macmillan.
Krifka, Manfred. 1995. Genericity: an introduction. In The Generic Book, edited by Greg
Carlson and Francis Pelletier, 1124. Chicago: The University of Chicago Press.
Landman, Fred. 1989. Groups, I. Linguistics and Philosophy 12: 559605
Lefebvre, Claire. 1998. Creole genesis and the acquisition of grammar. Cambridge UK:
Cambridge University Press.
Lbner, Sebastian. 1985. Definites. Journal of Semantics 4: 279326.
Lbner, Sebastian. 2011. Concept types and determination. Journal of Semantics 28:
279333.
Longobardi, Giuseppe. 1994. Reference and proper names. A theory of N-Movement
in syntax and Logical Form. Linguistic Inquiry 25: 609655.
Lyons, Chris. 2000. Definiteness. Cambridge: CUP.
Matushansky, Ora. 2008. On the linguistic complexity of proper names. Linguistics and
Philosophy 31: 573627.
McWhorter, John. 2001. The worlds simplest grammars are creole grammars. Linguis-
tic Typology 53/4: 125156.
Milner Jean-Claude. 1982. Ordres et raisons de langue. Paris: Seuil.
Ndayiragidje, Juvnal. 1984. La source du dterminant agglutin en crole hatien. The
Canadian Journal of Linguistics 34: 313317.
Pinalie, Pierre and Jean Bernab. 1999. Grammaire du crole martiniquais en 50 leons.
Paris: LHarmattan.
Poesio, Massimo. 1994. Weak definites. In Proceedings of SALT 4, edited by Mandy

Harvey and Lynn Santelmann, 282299. Ithaca, NY: Cornell: University.
Roberts, Craige. 2003. Uniqueness in definite noun phrases. Linguistics and Philosophy
26: 287350.
Russell, Bertrand. 1905. On denoting. Mind 14: 479493.
Strawson, Paul. 1950. On referring. Mind 59: 320344.
Studler, Rebekka, 2008. Artikelparadigmen Form, Funktion und syntaktisch-seman-
tische Analyse von definiten Determinierern im Schweizerdeutschen. Ph.D. diss.,
Universitt Zrich.
Studler, Rebekka. This volume. The morphology, syntax and semantics of definite
determiners in Swiss German, 143171.
Valdman, Albert. 1978. Le crole: structure, statut et origine. Paris: Klincksieck.
Wespel, Johannes. 2008. Descriptions and their domains. The patterns of definiteness
marking in French-related creoles. Doctoral dissertation, University of Stuttgart.
Zribi-Hertz, Anne and Herby Glaude. 2007. Bare NPs and deficient DPs in Haitian
and French. In Noun Phrases in creole languages, edited by Marlyse Baptista and
Jacqueline Guron, 265298. Amsterdam/Philadelphia: John Benjamins.
Reference Resolution in
French Sign Language (LSF)
Brigitte Garcia and Marie-Anne Sallandre
1 Introduction1
The present study examines the expression of nominal reference in French Sign
Language (LSF). We will show that an analysis of nominal reference in sign lan-
guages (SL) has to take into consideration not only lexical signs but also the
constructions currently referred to as classifier constructions in the litera-
ture (called non-conventional units in what follows to use a theory-neutral
term).
Non-conventional units have been identified in all studied SLs and have
long been recognised in diverse theoretical approaches under different names.
Nevertheless, the description and the analysis of these units remain a major
topic of debate in SL literature. As shown below, the few studies of nominal
reference in SLs do not take Non-conventional units into account. However,
according to our study, nominal referenceand particularly the expression of
definiteness and specificitycrucially require a better understanding of the
interrelations between lexical signs and non-conventional units.
We begin by reviewing the debate regarding the description and status of
non-conventional units in SL literature (section 2). Against this backdrop, we
present our own theoretical framework and explain the centrality of non-
conventional units for the structural economy of SL (section 3). We then discuss
the main studies of nominal reference in a number of SLs, couched in diverse
theoretical approaches (section 4). In light of these studies, we show (section
5) how the interaction between lexical units and non-conventional units in dis-
course allows a better understanding of the expression of nominal reference in
SL.
1 The main abbreviations we use in this chapter are the following: LSF = French Sign Language;
LU = Lexematic Unit; PT = Personal Transfer; SL = Sign Language; SpL = Spoken language; ST
= Situational Transfer; TSS = Transfer of Shape and Size; TU = Transfer Unit. As for glossing
conventions, we follow the habit in SL linguistics, that is: capital letters for conventional
(lexical) units and lower case for non conventional units. Since SLs have no written form,
glossing through the written words of a spoken language has become a general procedure.

reference resolution in french sign language (lsf) 317
2 The Expression of Reference in SL Discourse:

Debates in the Literature
The relatively young2 field of SL linguistics was initially dominated by a formal,

primarily generative, approach (Klima & Bellugi 1979). Since then, particularly
since the 1990s, major theoretical debates have opposed the tenants of for-
mal approaches and cognitive-functional approaches. These debates centre on
two interrelated issues, both directly related to the topic of this chapter. The
first subject of debate concerns the role and the status accorded to the use
of space in SLs, as many SL categories and morpho-syntactic relations, as well
as the introduction of nominal referents and the continued reference to them
in discourse crucially involve the use of space. The second issue concerns the
description of the constructions in SL discourse which cannot be analysed as
conventional lexical signs. Both of these questions are related to the fundamen-
tal question concerning the impact of modality3 on linguistic structures (e.g.
Meier et al. 2002; Woll 2003; Vermeerbergen 2006; Pizzuto et al 2007; Perniss et
al 2007).
In what follows we review the opposing views on both issues. This allows us
to position our own treatment of these phenomena, and more generally, the
framework which we consider necessary for the analysis of nominal reference
in SL.
2.1 Space and Its Use in SL Discourse

As visual-gestural languages, SLs involve a set of manual and non-manual
articulators. In addition to hands and arms, these include gaze, torso and facial
expression. These diverse articulators are used within the space in front of the
signer, known as the signing space. Through various mechanisms a discourse
entity can be associated with a location in the signing space. Association
with a location in space is obtained by the direction of the gaze, manual
pointing, direct placement of the manual lexical sign and body posture, or any
combination of these means.
The first, formal, descriptions of the function of the signing space (e.g.,
Poizner et al 1987) distinguished two distinct uses of signing space. The first
2 The text generally considered seminal is Stokoe (1960) on American Sign Language
(ASL).
3 The term modality is commonly used in SL linguistics to designate the production and recep-
tion channels used by languages: audio-oral for spoken languages (SpL) and visual-gestural
for SL.
318 garcia and sallandre
type involves properly syntactic uses of space mainly ensuring the expression
of categories such as person, number, verbal agreement, and grammatical rela-
tions. This contrasts with other functions labelled topographical or descrip-
tive. Syntactic functions, the main focus of studies so far, have been described
as consisting of an abstract use of space. Abstract use of space involves arbi-
trarily assigning linguistic entities to a location in the signing space, called the
locus, allowing, for example, the identification of verbal arguments. The loca-
tions can then serve as pronouns, allowing renewed reference later in the dis-
course. Thus, syntactic space is taken to constitute a series of arbitrary points,
non-descriptive and non-modal (surface projections of linear structures). In
contrast, the topographical or descriptive function of space involves a non-
arbitrary use of space (i.e., iconic or analogical). This type of descriptive use is
still considered limited to the expression of relative spatial relations between
referential entities.
The syntactic function of space in SLs (in the sense defined above) is advo-
cated by a number of researchers (e.g., Quer 2005; Sandler & Lillo-Martin 2006;
Barbera 2012). This view has however also been criticised since the early 1990s
(e.g., Liddell 1990, 1995, 2003; Meier 1990; Engberg-Pedersen 1993, 2003). At the
heart of this criticism lies the unlimited nature of spatial reference points that
can be activated by manual and non-manual articulators. The debate focused
initially on a class of verbs known as directional or agreement/agreeing
verbs, common in ASL, and easily identifiable in other SLs, in which the per-
son and/or the arguments of the verb are marked by modifying the verbal form
along the parameters of orientation and movement. The disputed question is
whether a formal analysis of person and/or arguments of the verb (the dom-
inant position) is tenable, given that the location of the referent (a person
or an entity physically present in the discourse situation, or an entity spatially
constructed in the discourse) is infinitely variable and context-dependent (sit-
uational or not).
A further issue related to the use of space in SLs involves the marking of
personal pronouns. The long-held view in the analysis of ASL (and later
in other SLs), identified three pronouns (defined as first, second, and third
person). These pronouns are formally characterised as a pointing with the
index toward the signers chest (first person), toward the addressee (second
person), or to the right of the addressee (third person). The first to challenge
this analysis was Meier (1990). Since the addressee may not be physically
facing the signer and may be one of (potentially multiple) participants in an
exchange, Meier stressed that the direction of pointing for both second and
third person may vary indefinitely. As this situation leads to a problem of formal
specification, i.e., of the very existence of these so-called second and third
person pronouns, Meier concludes that there is grammatical opposition only

between first and non-first person in ASL.
In the context of this debate, Liddell (1990, 1995) proposes an analysis that
takes the gradient nature (i.e., non-grammatically specified) of at least some
components of these pronominal or verbal signs to be crucial. According to
Liddell, this gradient nature stems from the characterisation of these signs,
verbal forms (indicating verbs in his terms) and pronouns alike, as a concep-
tual blend of linguistic components (grammatically specified, integrated into
the ASL lexicon) on the one hand, and a gestural component, characterised
as non-grammatical, on the other. Liddells analysis relies on Fauconniers
(1985) theory of mental spaces, which gradually developed into the con-
ceptual blend theory (Fauconnier & Turner 1996). Liddell adopts the view of
linguistic forms as (partial and undetermined) instructions for constructing
interconnected domains with internal structure (Fauconnier 1997: 35, cited
in Liddell 2003: 81) and the idea that the entities people talk about as they
speak are all conceptual entities within conceptual structures called mental
spaces. (Liddell 2003: 80). For him and the authors who support this view,
ASL (and other SLs) in fact recognises only one opposition: first vs. non-first
person. Similarly, spatial modifications of verbal forms cannot be analysed as
inflection or agreement, as they are considered merely gestural. More gener-
ally, the set of directionality phenomena must be understood as pure pointing
gestures towards spatially-grounded conceptual entities that are always refer-
ents, whether they are physically present in the signing space or constructed
in the discourse. Consequently, these SL uses of space, as well as the construed
referents, are excluded from the grammatical linguistic domain.
Barber (2012) contrasts this view of space in SLs, which she terms the
spatial mapping4 view, with that held by the proponents of the formalist
approaches, the R-locus view5 in her terms. In line with this approach, and
following Lillo-Martin and Klima (1990) in particular, she proposes to view
these loci as referential indices rather than as pronouns. The unlimited charac-
ter is thus transferred to the referential indices, which, although non-specified
lexically, are phonologically specifiable in the sense that they are discourse-
4 The definition used by Winston, to whom this expression is attributed, is: spatial mapping is
the association of an area in the signing space with an element from the mental representa-
tion of the signer. []. Spatial mapping creates a location to which a signer can subsequently
point in order to evoke the mental representation of the entity originally mapped at that loca-
tion. (Winston 1995: 90).
5 For referential locus, following Lillo-Martin & Klima (1991).
determined. In this sense, the only difference between signed and spoken
languages, and ultimately the only impact of modality on structure, is that
despite being discrete, the set of referential indices remains overt.
We must stress an essential point of divergence which will allow us to better
explain our own approach (section 3). In the spatial mapping view, the physical,
real universe and the discourse universe are indistinguishable, and the SL use
of space is simply excluded from the linguistic domain. In contrast, the R-Locus
view espouses the idea that the SL use of space is always a linguistic construct,
based on the very production of discourse, as highlighted by Barber: Without
a conversation and without the use of referring expressions directed to it, sign
space does not exist. [] What matters most for the constructions of linguistic
space is that sign interlocutors share the same coordinates in which discourse
is built (Barber 2012: 6162).
The main alternative proposed in the literature, regarding the status of
space and its SL uses, is to view it either as an essentially abstract/arbitrary
and entirely amodal space, or as non-linguistic. This theoretical divergence
also underlies the second topic of disputethe status and description of the
non-conventional units.
2.2 Discourse Units in SLs: Established Signs vs. Classifier Constructions

Two major types of gestural units were identified early in SL research (e.g. for
ASL, Friedman 1977; Klima & Bellugi 1979). The first type is the conventional
signs, lexical signs also frequently referred to as established signs. These
signs were the initial focus of Stokoe (1960).6 His key proposal, which has been
widely accepted since, was that these signs could be broken down into their
various manual components (initially, the handshape, its movement and its
location7). These were described as the equivalents of phonemes. The second
major type of units corresponds to complex, non-conventional constructions
that involve the motion, location, handling, and/or visual-geometric descrip-
tion of nominal referents. The characteristic property of these constructions
is that each component (of the same parametric type as those used for lexi-
cal signs) seems to be directly meaningful. The handshape, for example, varies
regularly depending on the salient features of the referred entity.
6 These signs also show the greatest structural proximity to spoken language lexemes, which is
easy to spot.
7 Two additional parametric components were later identified: hand orientation (Battison
1973) and facial expression (Baker and Padden 1978).
The original analysis of De Matteo (1977) stresses the crucially iconic nature
of these non-conventional constructions. In contrast with this view, the first
formal descriptions proposed for non-conventional constructions8 argued for
their linguistic status. In particular, they proposed that despite the iconicity
of these constructions they could still be broken down into discrete, conven-
tional and lexically specified components. Notice that these analyses are pri-
marily concerned with the handshape, which is considered to semantically
classify referents, and is therefore compared to classifier morphemes found
in some spoken languages (e.g. Frishberg 1975; Kegl and Wilbur 1976; Supalla
1978, 1982, 1986). In the abundant literature on these handshapes and the con-
structions using them, they quickly became known as classifiers or clas-
sifier handshapes and classifier constructions (CL-constructions). In
the various typologies of classifier handshapes, and CL-constructions, there
is relative consensus that CL-constructions can be divided into three types:9
(i) entity classifier constructions, where the handshape represents the
denoted referent, a part thereof, its movement and location representing the
motion and/or location of the entity; (ii) handling classifiers contruc-
tions, where the handshape represents the way the referent is grasped or
handles, and (iii) visuo-geometric classifier constructions, in which the
handshape represents the form of the referent (primarily through tracing).10
The first two categories are illustrated by examples from British Sign Language
(BSL) in Figures 1 and 2 (from Cormier et al 2012: 330).
figure 1 An entity classifier construction, in BSL

(cormier et al 2012: 330)
8 The first formal analyses were proposed in reaction to De Matteos original analysis (e.g.,
Supalla 1978).
9 We mention only the most common terms. For a detailed review of the typologies and terms
proposed, see Schembri (2003).
10 Also known as size and shape specifiers (SASSs).
figure 2 Handling classifier constructions, in BSL

(cormier et al 2012: 330)
The formal approach to non-conventional constructions gave rise to intense

scholarly debate, particularly in the early 1990s. The first debate contests the
very pertinence of the classifier concept, in the relevant spoken languages as
much as in SLs. Several studies (e.g., Slobin et al 2003; Schembri 2003) show
that the so-called classifier handshapes do not categorise semantic classes
of referents, but specify a particular property or aspect of the referred entity
(which may vary even for the same referent). Despite the strong arguments pre-
sented, which are accepted by the majority of SL researchers, and the absence
of an alternative formalist analysis to this very day, these constructions are still
essentially described and designated as classifier constructions. However, the
study of these constructions took a different turn, again following Liddell, who
pointed out the failure of formalist attempts to break them down into a discrete
and finite set of morphemes.
Liddells proposal (1995, 2003) mirrors his analysis of what he calls indi-
cating verbs (commonly known as directional): he proposes to acknowl-
edge that CL-constructions (that he calls depicting verbs) are also a blend
of two types of components, a lexically-specified component (essentially the
handshape) and a gestural component. For Liddell, the unifying property of
depicting and indicating verbs is that both are directional. In Liddells view
these two types of constructions offer the possibility for the different artic-
ulators used in SLs to be positioned and oriented in the signing space while
projecting a symbolic content. According to Liddell, this positioning in space
pertains to the gestural component. What distinguishes depicting verbs from
indicating verbs is that in the first, hand placement creates a spatial relation,
i.e., a topographical space; [] the directionality of depicting verbs depicts
topographical locative information while the directionality of indicating verbs
identifies entities. (Liddell 2003: 268). Thus reduced to their lexically specified
components, depicting verbs constitute for Liddell a long but finite list, with
their own paradigms and combinatorial constraints. Liddell stresses that a full
inventory is yet to be completed, but, in his words, depicting verbs should be
described as a large semi-productive derivational system (2003: 274) based on
verbal roots.
And so for SL literature, these constructions remain problematic to this
day. Formal approaches have offered no alternative to the classifier concept,
despite the acknowledged inadequacies of the analysis; in particular no for-
mal inventory has been proposed for the constituent morphemes making up
the so-called classifier constructions. The functional alternative, based on the
notion of conceptual blends, simply moves a whole range of such constructions
into the domain of gestureconsequently beyond the scope of linguistic anal-
ysis.
The same appeal to so-called gesture is Liddells solution to the other dis-
puted construction, frequently termed role shifts.
2.3 Role Shifts and Constructed Actions

The process called role shifting, following Mandel (1977), and later role shift
(hereuder: RS), occurs when the signer slips into the role of one of the char-
acters in the utterance, presenting the information from the point of view of
that character. Aside from the atypical minority approaches of Friedman (1977),
where role shifts are considered unique to SLs, the first linguistic studies on ASL
considered these (highly iconic) constructions as non-linguistic, and referred
to them as free pantomime (e.g., Klima & Bellugi 1979). The process of RS is
certainly addressed throughout the 1980s, but only from a single perspective,
the formal description of the consequences of change of reference induced by
the role shift, i.e., the rearrangement of the loci associated with referents (e.g
Padden, 1988, 1990; Lillo-Martin and Klima 1991). The type of RS considered is
primarily that which aims at reporting the thoughts and words of the entity
referred to. Basing ourselves on a much wider range of constructions including
RSs, we propose an entirely different perspective on these discursive phenom-
ena.
Starting from the 1990s, several authors working on corpora of narrative
discourse point out that far from being limited to reported dialogues, RSs can
also report the actions, states and attitudes of the referent (Smith, Lentz and
Mikos 1988; Ahlgren 1990, Meier 1990). Winston (1991) proposes to describe
these constructions as constructed actions (CAs), since they do not involve
a mere copy, but the narrators selective reproduction, of the reported action.
Winstons (1991) and Metzgers (1995) work view CAs essentially as one of
a range of processes that allow spatial mapping. Following these authors,
Liddell (Liddell and Metzger 1998, Liddell 2003) offers his own analysis of the
construction.
Extending his theory of conceptual blending, Liddell characterises CAs as a
specific type of blend, unique in that the signer is part of the blend, thereby
creating what Liddell terms a surrogate blend, which creates a surrogate
space (which is therefore a viewer space). Liddell distinguishes two cate-
gories: surrogates combined with linguistic signs, and pure surrogates (with-
out signing). As with indicating and depicting verbs, any element of these
constructions that does not involve grammatically-specified signs (for Liddell,
any process that is gradient in nature), should be considered as gestural. Lid-
dell intends his descriptions to cover both CAs and RSs.11
In recent years, these constructions have sparked renewed interest among
formal researchers (e.g. Lillo-Martin 1995, 2012; Zucchi 2004, Quer 2005; Quer
and Frigola 2006; Quer 2011). These studies are primarily interested in the
shifting of indexicals in the role shifting phenomenon. However, they share
one property with descriptions of CAs and/or depicting signs, viz. the com-
bination of linguistic and gestural elements. Given the existence of so-called
constructed action in addition to individual RSs, these authors identify a
significant part of the manual and non-manual elements involved in these con-
structions as gestural. As highlighted by Quer (2011: 287) for example:
11 However, for Liddell, the role that pointing plays within the blend is its standard, purely
indexical (ostensive) function, so they are not categorised as pronouns (i.e., not grammatical).
Thus, the question of shifting is irrelevant for him.
Here I have put aside from the discussion cases of so-called constructed
action (Metzger 1995), which frequently co-occurs with RS in narrative
discourse. Within constructed action, the signer adopts the role of the ref-
erent in order to reproduce not his/her linguistic discourse, but his/her
actions, postures or gestures in a more or less imitative fashion. This
aspect is a poorly studied one, although it is found regularly in the descrip-
tions of narrative techniques in sign languages (see Quinto-Pozos 2007).
[] These are the most complex cases to account for, as they require
teasing apart what is gestural from what is linguistic in a RS segment.
As a consequence of the language modality, both regularly coexist, either
simultaneously or consecutively.
Thus, like CL-constructions, the constructions involving RSs (be it to report a

characters thoughts or actions) constitute the least frequently modelled aspect
in the literature on SLs. We must stress that the massive relegation of all or part
of these discourse phenomena to the domain of gesturality essentially boils
down to claiming that these sections of discourse cannot be analysed linguisti-
cally. We regard this view as all the more unsatisfactory as these constructions
together comprise a very significant portion of SL discourse. Sallandre (2003)
reports that these kinds of NCUs represent on average 70% of narrative dis-
course in LSF, and 30% of prescriptive discourse. According to Antinoro Piz-
zuto et al (2008), units of non-conventional type constitute the major mode of
anaphoric expression in LSF, LIS and ASL (8095 %). Russo (2004) has found
similar percentages of non-conventional signs in LIS. Analyses from various
theoretical perspectives confirm the high frequency of these kinds of units in
various SLs (e.g, Klima and Bellugi 1979; Liddell 1995, 2003; Winston 1995, in
ASL; Brennan 1990, 2001, in BSL; Johnston and Schembri 1999, 2007 in Auslan;
Meurant 2008 in LSFB12).
As shown below, the Semiological Model takes NCUs to be central and
specific to SLs, proposing a unified account of these phenomena.
12 LIS: Italian Sign Language; Auslan: Australian Sign Language; LSFB: Belgian French Sign
Language.
3 The Semiological Model
3.1 Epistemological Framework

We begin by outlining the major epistemological foundations of our approach.
The model within which we examine LSF, other SLs, and human languages in
general is couched in an enunciative and functional perspective (e.g., Jakobson
1963). As regards the relations between cognition and language, it mirrors
the principles of Cognitive Grammar (Langacker 1987, 1991). Among these
theoretical roots, one aspect, which relies on a European tradition developed
particularly in France, must be clarified, since it clearly distinguishes us from
other SL linguists who follow a cognitive-functionalist approach. This is the
concept of linguistique de lnonciation (e.g. Jakobson 1963; Benveniste 1966;
Culioli 1990; Ducrot 1984; Lyons 1977), according to which the structures of a
language are determined by a necessary anchoring to an act of uttering.
The act of uttering (nonciation, in French) cannot be reduced to what is
commonly termed the context of utterance, i.e. the context understood as the
physical environment and all the circumstances in which an utterance is pro-
duced. The scope of this concept can be precisely captured with respect to a
fundamental opposition proposed by Jakobson (1963), between what we shall
call here the Situation and Utterance domains,13 which revolve around Jakob-
sons well-known notion of shifter. The Situation Domain both links and lin-
guistically co-determines the signer/speaker and his addressee: the very act of
uttering simultaneously establishes them as 1st and 2nd person. This relation-
ship is crucially symmetrical and reversible, as witnessed by the recourse to
shifters, such as person markers (1st and 2nd person).14 The Utterance Domain
is internal to the discourse produced: it connects the characters involved in the
utterance. An essential point is that the Situation Domain, which is highly lin-
guistic, cannot be reduced to the real (physical) participants in the interaction.
A secondary Situation Domain can be opened up in an utterance, for exam-
ple through the various mechanisms of reported discourse in which a charac-
ter in the utterance event becomes an enunciator, generating a co-enunciator
whom (s)he addresses and/or interacts with. This recursive interplay between
the Situation and the Utterance Domains is made possible and embodied
by nested shifters, and thus generates crucially recursive discourse-reference
13 These terms freely translate what French nonciativistes call: plan de lnonciation (> Situation
Domain) and plan de l nonc (< Utterance Domain).
14 Following Benveniste, we take the 3rd person to be the non-person, i.e., the one which is
neither the 1st nor the 2nd person.
frameworks. In response to one of the debates mentioned in 2.1, we discuss

below one of the major contributions of this general framework to our under-
standing of SL structures, with respect to the expression of person.
The absolute, real physical coordinates of the interlocutors are not relevant
from an enunciative perspective. Personal shifters express the necessarily inter-
relational co-determination of the two co-enunciators, and by contrast, that
of the non-person (i.e. 3rd person). In LSF, this marking is expressed by the
interaction of gaze and pointing: the two co-enunciators mark each other as
such by the mutual engagement of gazes that establishes who is taking the floor
(and this may be enhanced for the 1st person, by self-pointing, and for the 2nd
person, by pointing towards the person being looked at). In contrast, the 3rd
person is pointed at by the signer without being looked at.15
3.2 Major Types of Units

Our functional perspective is combined with an important methodological
option, namely the decision, from the very first descriptions of LSF, to base our
analyses on a long discourse corpus (e.g. Jouison 1981 [1995]; Cuxac 1985; 1996;
Bouvet 1996). Thus, in the early 1980s (Cuxac 1985), Cuxac began to develop the
Semiological Model we adopt here, based on an empirical-inductive analysis of
such discursive corpora. This approach naturally led to an early interest in the
range of very highly iconic productions mentioned in sections 2.2 and 2.3, found
pervasively in the corpus data. The specific contribution of the Semiological
Model is as follows. The competing descriptions proposed for the complex
units described as CL-constructions or as depicting verbs, have in common
that they essentially amount to proposing long lists, which are problematic by
virtue of their very length. In contrast, the Semiological Model accounts for all
these constructions on the basis of three major linguistic structures, termed
transfer structures (structures de transfert), which can generate an infinite
number of transfer units (TUs, units de transfert), as is true for any structure.
We present these structures below.
The structure termed transfer of shape and size (TSS, transfert de taille et
de forme) enables the signer to show the shape or size of an entity; situational
transfer (ST, transfert de situation) can show the motion of a mobile entity
(using the dominant hand) against an unchanging background (typically using
the non-dominant hand), while the scene is presented as seen as a whole from
15 For a detailed description, see Cuxac (2000) or Antinoro Pizzuto & Capobianco (2008), who
elaborate on the distinction between pointing in spoken languages and co-verbal gestuality.
a distance; personal transfer (PT, transfert personnel) involves the signer

literally becoming the denoted entity, thereby showing this entitys point of
view, the actions it performs or is subjected to, and its utterances. Figures 3,
4 and 5 below illustrate these three types of transfer structures: TSS (Figure 3),
ST (Figure 4) and PT (Figure 5).
figures 3, 4, 5 Examples of the three main transfers: size and shape transfer (shape of the horses
tail), situational transfer (the horse jumps over the fence) and personal transfer (gal-
loping horse)
(corpus ls-colin, cuxac et al 2002)
These three structures are considered as the structural result (on a phyloge-
netic scale) of the repeated implementation of a particular semiological intent
of the signer (the illustrative intent), whose goal is not only to tell, but to
tell while showing. This possibility of showing, i.e., to structurally exploit the
iconicity (iconicit dimage) as a way of producing linguistic meaning, is made
available by the visuo-gestural modality. We assume that, since this modality is
the only one which can be activated in deaf communicationwhereas hear-
ing communication is bi-modalSLs have developed a linguistic structuring
of iconicity. This idea requires that the description of human languages, SLs and
spoken languages alike, and the forms they may produce, be based on the semi-
ology of the channel. It also means accepting figurativeness among the possible
modes of language manifestation and full linguistic expression. In this chapter,
we cannot go into the details of the empirical and semiogenetic argumenta-
tion supporting this model. We refer the reader, in particular, to Fusellier-Souza
(2006), Cuxac and Sallandre (2007), Cuxac and Antinoro Pizzuto (2010), Garcia
(2010), Garcia and Derycke (2010), Sallandre and Garcia (2013). We will limit our
discussion to highlighting only those aspects of the model that are indispens-
able for our purposes.
A condition for initiating a linguistic interaction in SL is setting the signers

gaze, which is also a clue to his/her semiological intent (saying vs. saying
while showing). The intent to say while showing requires the signers gaze to be
detached from the addressee (the co-enunciator), thereby signalling the tem-
porary removal of the signer from the Situation Domain. In personal transfer,
the signer actually disappears as enunciator and embodies an entity referred to
in the Utterance Domain, his/her gaze becoming that of the transferred entity.
In size and shape transfer, the signer is simply a bodily vector of the emer-
gence (followed by the gaze) of the shape in display. In situational transfer, the
signer is the vector of the motion (followed by the gaze) undergone by the entity
referred to by the dominant hand.
3.3 Combinations of Units and Discourse Economy: A Detailed Example

In discourse, the three types of TUs can combine together, but also with the
other types of units: lexematic units (LUs, called established signs in the
literature), pointing units and dactylological units.16 Sallandre (2003, 2007)
brings out 24 recurrent combinations. These combinations have enabled us
to refine the part of the model which concerns personal transfers, showing
that there is actually no single type of role shift. Rather, there are many types,
categorised according to a number of morphological and semantic properties:
the implication of the signer-enunciator in the uttering act, the number of
entities the signer represents simultaneously with both his/her hands and
body, and finally the level of discourse in which the signer operates (e.g. the
presence or absence of reported discourse in a personal transfer utterance).
The proposed model allows us to highlight the discursive economy achieved
by the repeated alternation between the two modes of saying (with and without
transfer) and the intertwining or embedding of LUs and TUs. Figure 6 provides
a very clear example.
In the sequence in question, the signer describes his career as an LSF teacher
to hearing adults. He therefore uses a personal transfer which presents himself
(teaching awkwardly) as a role at the beginning of his career. While embody-
ing the young teacher he used to be, he uses the LU [TEACH]. In its conven-
tional form, this LU is a directional verb in which both hands, side by side,
16 Institutionalised SLs in alphabetic cultures include signs representing the letters of the
alphabet. These dactylological units, more or less functionally integrated depending on
the SL, enable words from the surrounding spoken language to be introduced into the signed
discourse by finger-spelling.
figure 6 Semi-transfer TEACH awkwardly

(corpus creagest, garcia & lhuillier 2011)
form a symmetric flattened O handshape in front of the signer, at chest height,

and create a repeated horizontal movement, while the fingertips are oriented
towards the teachings beneficiary. However, in this case, the LU is identified
solely by the manual handshapes (the flattened O), while movement and orien-
tation are disorganized and non-symmetrical, non-conventional, thereby con-
veying the meaning of awkwardness.17
This tight embedding of a LU (or part of one) in a broader illustrative context
in which the signer is using a personal transfer18 stems from a very common
structure in LSF, which we term Semi Personal Transfer (semi-transfert per-
sonnel, Cuxac 2000). This is a very economical structure precisely because the
conventional and generic information carried by the LU and the information
conveyed by the mode of saying (saying while showing) overlap, as witnessed
by the manual and non-manual multi-linearity characteristic of the illustra-
tive intent. But complexity is further increased by the play of gaze (and facial
expression), which allows the signer to shift from the Utterance Domain (where
(s)he stands as an embodied entity) to the Situation Domain (where (s)he
17 These parametric modifications must be distinguished from the systematic morphological

modifications of directionality that mark the agent and beneficiary.
18 The micro-sequence described here is part of a larger sequence with a clear illustrative intent
both before and after the chosen example.
interacts with the addressee). In fact, during the personal transfer of himself as
a young teacher (LU handshape [TEACH]), teaching awkwardly (hand move-
ment and orientation), the signers gaze and facial expression are alternatively:
(i) those of the transferred character (himself at the time)his gaze set on
the moving hands, i.e., disconnected from the addressee, thereby signalling the
transfer, and his facial expression depicting the muddled and awkward nature
of the process (of teaching) [Figure 6, left and right images]; (ii) those of the
signer commenting to the addressee on his teaching experience, displaying
self-deprecationhis gaze set on the addressee, with a self-deprecating facial
expression [Figure 6, central image]. Such sequences, whose complexity arises
from the intertwining of lexical and transfer units, alternating between the
two modes of saying (saying vs. saying while showing), and from the interplay
between the Situation and Utterance Domains are very characteristic of LSF.
3.4 Synthesis and Implementation in Relation to the Literature

On the basis of the same linguistically integrated semiology of saying while
showing, the three transfer structures coherently account for various phenom-
ena that have been discussed under diverse yet problematical headings
CL-constructions (or depicting signs) and role shifts (or constructed actions).
Note that what the literature terms classifier handshapes or depicting hand-
shapes are only one component of transfer structures. Far from categoris-
ing the referent, these proforms (according to our own terminology) form
a closed list of handshapes, which can bring out some aspect of the referent,
through a logic of iconicity.19
As regards the status and nature of the use of space in SLs, note how our
model differs from the two conflicting theories discussed in section 2.1. In
agreement with the proponents of the R-Locus view, we consider space in SLs
to always be a linguistic construct, in which reference points are generated
through the act of uttering and the discourse itself (contra Liddell). We do
not subscribe to the notion of mental spaces, which results in excluding from
the sphere of language most of the facts described in sections 2.2 and 2.3.
Nevertheless, in our perspective, the points activated in the linguistic space
are neither arbitrary nor abstract. We argue that the use of space in LSF falls
into three types, which, just like types of units, may be combined with or
embedded in one another: (i) a topographical or descriptive space, which
characterises reference under the illustrative intent; (ii) a diagrammatic space,
19 On these functions and a detailed inventory of proforms, see Cuxac (2000: 97130).
typically involved in the construction and management of reference outside

the illustrative intent (i.e. saying while showing);20 (iii) the uses of space which,
in an enunciative framework, hinges upon the opposition mentioned above
between the Situation and Utterance Domains, and the discursive frame of
reference they generate. This enunciative use of space is mainly conveyed
by the signers gaze and its coupling/uncoupling with manual pointings, as
illustrated above by the expression of person.
Table 1 below provides a summary of the phenomena and types of construc-
tions under discussion in the specialised literature, outlined in sections 2. and
3. The descriptive concepts proposed in the Semiological Model are matched
with the concepts used in the specialised literature at large on the one hand,
and with those developed by Liddell on the other. This is a difficult task in itself,
since the comparative perspective should not mask the conceptual discrepan-
cies in the underlying approaches.
table 1 Equivalences and discrepancies between the descriptive concepts

used in the literature on space-exploiting phenomena and discourse
constructions (adapted from garcia 2010)
Literature Liddell (2003) The Semiological Model

Combination
of gestural
and linguistic or Outside the Under the
pure gestural Illustrative intent Illustrative intent
Personal pronouns Personal pronouns Personal pronouns [Personal pronouns

Issue: 1st /2nd /3rd 1st / non-1st person in PT in reported
person discourse]
Directional verbs Indicating verbs Directional verbs [Directional verbs

in PT in reported
discourse]
20 We employ here Peirces (1978) notion of diagram, defined as one of the icon types in which
only relations are in a likeness relation with the referent. Thus, in agreement with the
R-locus view and contra Liddell, we argue that the absolute coordinates of the loci (the
points activated in space) do not reflect referent loci, but pure discourse constructs. Yet, the
diagrammatic space which iconically includes the relations between these loci (and therefore
between the referents) is a descriptive rather than arbitrary space (if arbitrary is taken to
mean non-iconic).
Literature Liddell (2003) The Semiological Model

Combination
of gestural
and linguistic or Outside the Under the
pure gestural Illustrative intent Illustrative intent
Loci Tokens Activation of

points in space to
indicate referents
= diagrammatical
iconicity
Classifier Handshapes Depicting Proforms

Handshapes (handshapes)
(present in all types of
transfer)
Classifier Depicting verbs TSS (transfer of size
constructions (topographical and shape)
(and diverse diagrammatical ST (situational
terminologies, space) transfer)
see Emmorey 2003)
Role shifts Surrogates with Semi-transfer

Constructed actions signing
(including more or
less constructed
dialogues)
Surrogates without PT (personal transfer)
signing (including PTs in
reported discourse)
Depicting verbs in a Personal transfer
viewer space: Double transfer
a mix of depicting (PT + ST)
verb and surrogate
blend
(Dudis 2004: multiple
references)
3.5 Functional Distribution of Two Types of Gestural Units

Although their respective frequency varies according to discourse genre, the
distribution between LUs and TUs is clearly functional. This distribution
appears to correspond to two major tendencies. On the one hand, since they
are directly connected to the type of intent that conveys them, LUs are typi-
cally preferred in the construction of generic reference, while TUs tend to be
used for specificity.21 On the other hand, the alternation between LUs and TUs
often corresponds to an informational structure of the type Topic (LU) / Focus
(TU). We use a limited definition of the concepts of topic and focus here: the
topic is what the utterance is about, while the focus (or rheme) is what is said
about the topic. Since TUs make use of all manual and non-manual parameters
simultaneously, they as such convey a rich array of new information about the
referred entities. However, this division of labour between LUs and TUs is only
a tendency, precisely because both types of units are used in the same language
and may occasionally swap roles (see Cuxac 2004). However, these two strong
tendencies for a functional division are of primary importance in the construal
of nominal reference, and of definiteness in particular, as it is closely tied to
both specificity and focus.
Bearing in mind the various approaches to the major units of discourse, and
the various uses of space in SL, we proceed to examine a number of studies
which focus on the instantiation of referents and on the linguistic expression
of nominal reference in SLs.
4 Nominal Reference in SLs and the Expression of Definiteness:

Existing Studies
There are relatively few systematic studies of nominal determination in SLs.

Existing studies frequently focus on the expression of definiteness and on the
possible role of pointing. Proposed descriptions adopt the two major conflict-
ing approaches discussed above, with the conceptualisation of the signing
space one of the main points of disagreement.
21 The notions of genericity, specificity and non-specificity in our approach are defined at the
beginning of section 5.
4.1 Formal Approaches

In accordance with the underlying conceptualisation of space (2.1), descrip-
tions conducted from a formal perspective centre on the linear distribution of
constituents. The first study, focusing on nominal determination in ASL (Zim-
mer and Patschke 1990), is based on the earlier assumption (cf. Wilbur 1979)
that a certain type of pointing functions as a determiner in ASL. These authors
identify a type of pointing dedicated to this function (DET). This DET, char-
acterised phonologically as quasi-stationary and directed slightly upward, can
appear in pre- or post-nominal position and even in both at once. However,
it does not mark a binary definite/indefinite opposition. Its syntactic function
is to specify the referent. In support of this assumption, the authors provide
evidence showing that this DET can accompany a newly-introduced referent.
According to the authors, what distinguishes this DET from determiners in
English, for example, is its ability to accompany proper nouns and possessives.
The authors also note as remarkable that this DET points in a direction which
is generally insignificant (Zimmer and Patschke 1990: 209).
Contrastively, in a PhD devoted to a minimalist analysis of ASL DP struc-
ture, MacLaughlin (1997) associates definiteness in ASL with a formal feature
in the structure. She identifies a definite determiner characterised as manual
pointing directed to a locus, appearing exclusively in prenominal position, and
claims that the same pointing in postnominal position functions as a loca-
tive adverbial. Indefiniteness, according to this author, is formally marked by
very similar pointing, also prenominal, but characterised by upward orienta-
tion as opposed to orientation to a reference point in the space. According
to MacLaughlin, the DET signals the status of the referent as definite/indefi-
nite and as specific/non-specific. MacLaughlin relies of the following defini-
tions: an indefinite NP can introduce into the discourse a new entity that the
addressee is not able to identify, while a definite NP presupposes that the ref-
erent is identifiable by the adressee, which implies that it is also specific. More
generally, in MacLaughlins view, non-specific reference is not associated with
any set point in space. She identifies two formal variants of indefinite DET:
one involves manual upward pointing with the palm facing the signer (glossed
[ONE]), the other has the same handshape and the same orientation, but the
index finger and forearm undergo an oscillating circular movementthe
more marked this movement, the less identifiable the referent ([SOMETHING/
ONE]). Note that, although she argues on this basis that ASL has a dual def-
inite/indefinite grammatical marking, MacLaughlin explicitly acknowledges
that neither of the DETs is obligatory. A definite or indefinite interpretation
of a referent is possible even in the absence of one of the DET markers, given
contextual cues (although no details are given on the nature of these cues).
A similar study was conducted on Hong Kong SL (HKSL) by Tang and Sze
(2002). The authors identify the same opposition between a prenominal DET
and a postnominal locative adverbial. However, the indefinite DET is not asso-
ciated with movement, as in ASL, and it is not linked to a degree of referen-
tial identifiability. In addition, placed after the verb, the [DET + noun] unit
triggers an indefinite and non-specific interpretation. In particular, Tang and
Sze identify a particular role of the gaze accompanying the expression of the
definite/indefinite opposition, especially in the case of newly introduced dis-
course referents. If visual contact is maintained with the addressee, the referent
is interpreted as indefinite (and specific), while the definite interpretation is
marked by a gaze towards the spatial reference point. Most significantly, as in
MacLaughlins analysis of ASL, in HKSL the various markers are described as
optional.
Barber (2012), a more recent study of Catalan SL (LSC) from a formal seman-
tics and pragmatics perspective, provides a comprehensive study of the uses of
spatial points (loci) in LSC. A major challenge in this study is to show that the
spatial locations are integrated into the grammar of LSC (and not gestural, c.f.,
2.). This study is not restricted to the sentence level, which generally limits for-
mal analyses on this issue, but considers the discourse level, adopting for this
purpose the framework of Discourse Representation Theory (Kamp and Reyle
1993). Barber finds no formal binary definite/indefinite marking in LSC, and
therefore calls into question MacLaughlins and Tang and Szes conclusions, at
least for Catalan SL. She notes, however, that for these authors indefiniteness is
marked by manual pointing to upper spatial locations on the frontal plane, and
definiteness by pointing to lower spatial locations on the same frontal plane.
She too grants a syntactic significance to the opposition between higher and
lower location of a referent in the signing space, but argues that this opposi-
tion does not mark definiteness, since both definite and indefinite referents
can be assigned to the lower zone of the frontal space. Rather, in LSC these
zones correspond to a binary formal marking of specificitythe lower frontal
space is specific while the upper front space is non-specific. According to Bar-
ber, indefiniteness must be divided into specific and non-specific reference;
indefinite specific reference signifies that the reference is known only to the
signer, while indefinite non-specific reference signifies neither the signer nor
the addressee is familiar with the referent (Barber 2012: 243).
Finally, Barber highlights the association between strong spatial marking
and specificity in the lower zone of the frontal space, and weak spatial mark-
ing and non-specificity in the upper zone. In the first case, manual pointing
to a locus combines with a gaze towards this locus, whereas in the upper zone,
non-specificity is characterised by pointing that is not associated with a locus
and no directed gaze. For Barber, the clarification of this opposition in LSC
is evidence that space in SLs has abstract syntactic functions, which can be
described as an arbitrary system of discrete points organised according to speci-
fiable morpho-phonological properties.
4.2 Functionalist Approaches

Among the studies addressing the functions of space in SL discourse from a
cognitive-functionalist perspective, only a few have focused on its impact on
reference resolution and construal. Two such studies are Engberg-Pedersen on
Danish SL (DSL) and Rinfret on Quebec SL (LSQ).
Engberg-Pedersen (1993, 2003) examines the role of spatialisation in the
expression of a referent not present in the discourse situation, specifically
how such a referent is introduced and taken up in later discourse. According
to Engberg-Pedersen, a referent is introduced by instantiating a locus, which
she defines as a direction in the signing space or the situational context that
represents a referent in signed discourse (2003: 270). The important point for
the author is that any kind of referenceanimate being, object, place, time, or
abstract conceptcan be represented by a locus in this way, even though not
all of these referents are spatialised. She argues that spatialisation is preferred
with concrete referents and for referents that represent higher relevance for
the signer, in the discourse. As noted in other SLs, DSL loci can be created by
a variety of manual and non-manual mechanisms. She calls these pointing
signs: the index finger, the gaze, the orientation of the head and torso; in
addition, for nominals without a body-anchored location parameter, the sign
can be directly produced on a given location in the signing space.
Among the various pointing signs involved in the expression of reference,
Engberg-Pedersen also identifies manual pointing that serves as a determiner
(DET), which can be either pre- or postnominal. This DET can only appear with
specific nominals, and its function is to signal specificity. However, as it can just
as readily be used with nominals referring to newly-introduced discourse ref-
erents, she argues that it is neutral with respect to definiteness. In addition,
Engberg-Pedersen identifies another DET, which differs from the former in not
being directed.22 She concludes that this non-directed DET encodes definite-
ness, corresponding to Lyonss (1977) definition of the function of the definite
article in English, viz., signaling to the addressee that some specific entity is
22 [] an index hand with a hold movement (i.e., no movement) or occurring in a transi-

tion movement, the direction in which the index finger points being irrelevant (Engberg-
Pedersen 1993: 119).
being referred to without however giving him any locative (or qualitative) infor-
mation about it (Lyons 1977: 654) (Engberg-Pedersenn 1993: 119). Nevertheless,
Engberg-Pedersen stresses that this encoding is not obligatory for the expres-
sion of definiteness; the [directed DET + Noun] can also be interpreted as a
definite expression.
Engberg-Pedersen further introduces a semiological perspective on what
she considers an inherent link between spatial anchoring and specificity. In
her view, which accords with Liddells, there is a deictic basis for locus-marked
determiners (as for all directed pointing signs). Thus, it is due to its semiologi-
cal indexical nature that a locus-marked DET signals to the addressee that the
referent is unique, that is, different from other referents in the discourse uni-
verse (1993: 122). She argues that this intrinsic link between locus and specificity
explains why DSL does not use a formal binary marking to encode the defi-
niteness distinction. This specifying aspect inherent to spatial anchoring is
supported by the fact that inversely the non-locus-marked DET appears only
with definite referents:
Such nominals occur when the sender assumes that the receiver can
assign unique reference to the nominal on the basis of factors such as the
referents status in the discourse (the sender has already referred to the
referent earlier in the discourse). That is, when the determiner does not
have a locus marker, it indicates that the nominal is definite.
engberg-pedersen 2003: 122
Rinfrets (2008) study of LSQ also focuses on the role of presence or absence of
spatial assignment of a noun to a locus. Her study centres on the crucial ques-
tion of why the presence or absence of spatial marking of a nominal has so
far not been clearly answered in SL literature. She therefore undertakes a sys-
tematic study of the various strategies of spatial association of a noun in LSQ,
covering manual strategies (pointing, direct localisation of a sign) and non-
manual ones (the gaze, leaning of the torso), and their possible combinations,
in terms of form, distribution, and function (semantic, syntactic, discursive).
She considers definiteness as one of the potentials functions of spatial mark-
ing.
Basing herself on the literature on definiteness in spoken languages on the
one hand, and on competing hypotheses proposed to account for this concept
(identifiability, familiarity, uniqueness) on the other hand, Rinfret concludes
that in languages with specific grammatical markers (e.g. definite/indefinite
articles), there is no simple correlation between the presence or absence of
such markers and the presence or absence of an (in)definite interpretation.
Other types of markers influence definite/indefinite interpretation, and the

role of context is decisive, in spoken languages as in others. Concluding that
there seems to be no clear analysis of definiteness as either a syntactic or a
semantic category, Rinfret opts for an analysis of this concept based on a con-
textual interpretation of grammatical markers.23 Her choice is supported by
a major weakness of formal approaches to reference resolution in SLsthe
fact that the so-called grammatical marking of definiteness and/or specificity,
noted in Zimmer and Patschke (1990), MacLaughlin (1997) and Tang and Sze
(2002) turns out to be optional. These studies note the role of context with-
out offering further explanation. In contrast, following Ariel (1990) and Epstein
(2002), Rinfret advocates a position that goes beyond the syntax/semantics
opposition, and adopts the concept of accessibility, which she argues pro-
vides a better explanatory scope compared to exclusively syntactic or semantic
theories of definiteness. Ariels notion of accessibility is a gradient concept,
which refers to the degree of activation of information about a referent in short-
term and long-term memory. This activation, according to Ariel, depends on a
variety of factors, including salience. Rinfret adopts Epsteins (2002) theoret-
ical framework for two main reasons. First, Epsteinss model satisfies her search
for an explicit conceptualisation of referent activation and secondly, this model
relies on a model of discourse construction that can be applied to the study of
discourse in SLs.
For Rinfret, Epsteins (2002) proposal is interesting in two respects. First,
Epstein extends Ariels concept of accessibility beyond recoverability from
given information in previous discourse. He presents an impressive range of
empirical cases in which the signer chooses to introduce a brand new discourse
referent (not necessarily recoverable through contextual inferences or ency-
clopaedic knowledge) using a definite expression. According to Epstein, this
choice responds to rhetorical and communicative needs (e.g. to signal to the
addressee that the referent will have particular importance in the discourse),
thereby highlighting the need to expand the treatment of definiteness beyond
strict referentiality.24 Secondly, Epstein adopts Fauconniers (1985) theory of
mental spaces as a discourse model in which she elaborates her accessibil-
ity analysis. She refers to Liddells application of Fauconniers theory in the
23 une interprtation contextuelle des marques grammaticales (Rinfret 2008: 123).

24 This idea is already present in Ariels notion of discourse salience (see also Engberg-
Pedersens discussion of higher relevance of a referent in discourse). Epsteins contribution
is in the systematic consideration of this aspect and the fact that (in)definiteness cannot be
treated solely in relation to referential function.
analysis of ASL, and argues that the notion of mental spaces is even more
relevant for SLs, given that signers can directly spatialise referents and their
relations, thus mapping the mental representations constructed in discourse.
Her central hypothesis, combining the concepts of mental spaces and accessi-
bility, is that spatial marking, which is unique to SLs, facilitates accessibility of
discourse referents, as the relative permanence of the spatial association facil-
itates the recoverability of these referents.
Rinfrets results overall confirm her hypotheses. Aside from her conclusion
that definiteness is not grammatically encoded in LSQ, and that any type of ref-
erent can be spatially associated (contra Engberg-Pedersen), Rinfret highlights
a correlation between the use of manual strategies for associating nouns to loci
(manual pointing in particular) and the marking of discourse salience. Yet, she
stresses that this is not the only way to encode salience. Above all, Rinfrets
conclusions confirm the important role of spatial marking as a signal of refer-
ential accessibility, while the absence of spatial association indicates a highly
accessible referent. Among the various strategies for spatial association, man-
ual strategies, pointing in particular, appear to mark low accessibility. Rinfret
argues that these strategies also signal the specificity of the referent, while the
absence of spatial marking signals the genericity of the referent, thereby par-
tially confirming the results of Engberg-Pedersen for DSL (and Winston 1995
for ASL). It is the very genericity of such nouns which explains the absence of
spatial marking, the author argues. Yet the absence of spatial association for
a newly-introduced referent does not necessarily indicate a non-specific refer-
ent, just as spatial association through pointing is not the sole condition for a
specific reading, given that contextual information may suffice. Thus, accord-
ing to Rinfrets results, the concept of accessibility provides an explanatory
and unified account of spatial marking (or lack thereof), while the concept of
definiteness in terms of uniqueness and familiarity is not sufficient to explain
spatial marking in SLQ.
4.3 Remarks on These Analyses

The analyses discussed above are undeniably fruitful in various ways, and many
of the proposed results must be examined against observed LSF discourse phe-
nomena. This is particularly true with respect to the function and distribution
of various types of manual pointing, and more generally, the function and
mechanisms of the spatial association of a nominal. Furthermore, in accor-
dance with Barber, Engberg-Pedersen, and Rinfret, we consider it necessary
to extend the domain of analysis to include the discourse. Following Rinfret,
we also believe that the question of definiteness should be considered beyond
its referential function, by taking into account communicative and enuncia-
tive aspects.25 However, an essential point is that all these approaches to the
issues of reference, reference resolution, and the expression of (in)definiteness
or specificity/non-specificity/genericity, whether formalist or functionalist, do
not take into account the various kinds of constructions noted in sections 2.2
and 2.3 (CL-constructions, role shift/constructed actions), which remain dis-
puted in the literature, despite their frequency in discourse, and their constant
intertwining with LUs.
Several authors have mentioned this choice explicitly. Barber (2012), for
example, states that her interests lie only in the non-descriptive use of space,
excluding the input of CL-constructions, but without further explanation (2012:
287, 365). Similarly, she excludes role-shifts, although noting their significance:
As well as in DSL, role shift in LSC is also a way of assigning discourse

prominence to the entity without regard to whether it is spatially estab-
lished or not []. However, role shift constructions are outside the scope
of this dissertation.
barber 2012: 220, note 5
Engberg-Pedersen considers at least some of these constructions.26 Neverthe-

less, she claims that they cannot be taken into consideration in the analysis of
reference resolution, since due to their indexical and iconic nature, they cannot
be referential:
Classifier constructions may constitute a full clause, but they may also be
preceded by a nominal used to refer to the referent. That is, even though
classifier constructions represent referents iconically and are indexical,
they are not referential.
engberg-pedersen 2003: 283
There is no explicit basis for this position, aside from the causal link between
the juxtaposition of CL-constructions with a preceding or following nominal
(i.e. an established sign or LU in our terms). The assertion that classifier
25 Epsteins novel suggestion not to limit the study of definiteness to the study of reference is
explicitly couched (cf. Epstein 1995) in the linguistics of enunciation, specifically in Jakobsons
functionalist approach (cf. the expressive function).
26 These are CL-constructions signalling the location or motion of an entity represented by the
handshape (see Engberg-Pedersen 2003: 283), which correspond, mutatis mutandis, to what
we analyse as situational transfers.
constructions are not referential in combination with a preceding LU is taken

to imply that they cannot be referential in themselves. The basic intuition
underlying Engberg-Pedersens position seems to be that, like Liddell, she con-
siders depicting verbs, manual pointing and indicating verbs as only par-
tially grammatical. For Engberg-Pedersen, therefore, only non-locus-modified
manual pointing is fully grammatical. In this perspective, anything that is con-
sidered gestural is not seen as referential.
On the basis of different justifications simply given in passing, Rinfret sim-
ilarly chooses to exclude these constructions from her analysis. In particular,
she discusses the common spatial association strategy in which a nominal
(an LU) is referred to by a spatialised classifier handshape. This is analysed as
a case of Winstons (1995: 87) complex spatial associations, which Rinfret
describes (without illustrating) as cases where the signer situates an element
in space using a verb or a classifier handshape, i.e., while establishing a rela-
tion or an action with an entity even before it is named27 (Rinfret 2008: 91). She
very briefly invokes two reasons for the exclusion of these constructions: (i) the
fact that these classifiers have their own meaning, like verbs (ibid. p. 94), and
moreover, (ii) that classifiers seem to be inherently specific (id. 2008: p. 94).
Examination of the data supporting her analysis (2008: 294306) confirms that
the only utterances examined are those we consider LUs, i.e., not pertaining to
the illustrative intent (saying by showing).
In contrast, according to our approach, TUs are fully referential and play
a crucial role in the instantiation and determination of reference in LSF, as
demonstrated in the following section.
5 An Alternative Approach to Nominal Reference in SL: French Sign

Language (LSF)
The following discussion is not intended to provide a comprehensive descrip-

tion of the expression of nominal reference in LSF. Our primary aim is to show
that such a description should take into consideration the role directly played
27 Lassociation spatiale () est faite de faon complexe lorsque le signeur situe un lment
dans l espace l aide d un verbe ou d un classificateur, cest--dire lorsquil tablit, dans
l espace, une relation ou une action avec une entit avant mme de lavoir nomme (Rinfret
2008:9091). [Spatial association is obtained in a complex fashion in situations where the
signer places an element in the signing space by using a verb or a classifier, i.e. when the
signer establishes a relation or an action with an entity before having named it.]
by TUs and their close interaction with LUs and with pointing units. The
concepts of genericity, specificity and non-specificity used in this section are
based on the definitions proposed by Riegel et al (2002), whose approach is
similar to ours. Following Riegel et al., a referent is generic if the referential
counterpart of the expression is considered in its maximal extension (2002:571).
In contrast, non generic reference can be of at least two types: (i) specific
when the referent is presented as existing and identifiable as such in a given
situation, or (ii) non-specific if it refers to any individual who satisfies the
properties denoted by the expression, but with no guarantee as to its existence
in the signers universe of discourse (id. 2002: 572).
The examples chosen as illustrations are highly representative of the data
in our corpora. In the following section, we explain the characteristics of the
corpora from which the examples in this chapter (previous sections included)
are taken.
5.1 Methodology
The three corpora used are LS-COLIN (Cuxac et al., 2002), Creagest (Garcia et al.
2011) and DEGELS (Boutora and Braffort, 2011). We participated in the creation
of the first two corpora; these are large corpora involving many signers,28 which
have become reference corpora for LSF. The DEGELS corpus is much shorter,
and presents the interaction of two deaf signers.29 The LS-COLIN corpus is a
monologue (one signer facing the camera), and proposes productions in the
narrative and explanatory genres. The Creagest and DEGELS corpora are dia-
logues between deaf adults on various topics of daily life (career, educational
background, areas of expertise in science, technology, art, and others). Conse-
quently, they include productions in various genres: argumentative, narrative,
descriptive and metalinguistic.
In order to maintain a certain homogeneity in the phenomena discussed in
this chapter and to focus on referential strategies, we have chosen to concen-
trate on first-mention reference in the beginning of a production. We thus limit
discussion to cases where reference cannot be derived from prior discourse
context. In addition, we do not consider the very special case of unique ref-
erents (such as the sun). When several entities are analysed through a single
example, the first referent is always a case of first-mention.
28 Thirteen signers from various geographical locations in the LS-COLIN corpus, 57 signers from
four regions of France in the Creagest corpus.
29 This corpus, which lasts less than a minute, was established for a national Annotation
Challenge in which all participants annotated the same video excerp.
To facilitate the understanding of some examples, we include below (Fig-

ure 7) the first two of five images that make up the printed support used as the
stimulus for the Horse Story narratives, produced by the thirteen signers of the
LS-COLIN corpus.
figure 7 Printed support for the Horse Story

(images n1 & 2; hickmann, 2003)
The analyses conducted on these corpora have enabled us to classify the main
modes of referent instantiation in LSF discourse. A major finding for us is that
a referent can be introduced through either an LU or a TU. In other words,
either the LU or the TU can be topical in the utterance. In our presentation,
we first discuss introduction of a referent by a LU (section 5.2) and then treat
introduction by a TU separately (section 5.3). As shown below, both types can
have a referential function, in isolation or combined with another type of unit
(LU, TU, pointing).
5.2 The First Unit Is a Lexematic Unit

5.2.1 The Single LU: The Expression of Indeterminacy
The beach is the main theme of the dialogue in DEGELS that we drew our first
example from. The two signers discuss the tourist attractions of Marseille30
(utterance 1). One of the signers, who is unfamiliar with the city, asks his
addressee, a native of the region, where he could find a beach to relax on.
(1) SO pointing ALSOLU MELU KNOWLU WHERELU BEACHLU31 TO RELAXLU
30 A large city in the South of France.

31 In each glossed utterance, the segment illustrated in the photos and described in more detail
is underlined.
Translation: So, you see, me, I would also like to know where I could find a
beach to relax on.
The entity beach is introduced here directly by a LU without prior or
subsequent pointing, without resumption and without spatial anchoring (left
image of Figure 8). Signer 2 responds using the LU BEACH alone. The nominal
here is highly indeterminate (bare noun).
figure 8 LU [BEACH] (top left image, in circle)

(corpus degels, boutora & braffort 2011)
This process is one of the typical expressions of indeterminacy in LSF: a bare LU,
unanchored, triggering the most generic interpretation, lower than any value
of (in)definiteness and specificity/non-specificity.32
This example also displays an interesting detail that merits further research.
The signer accompanies the LU BEACH with a clear labialisation la plage (Fr.
the beach), la being the feminine definite determiner in French.
32 This point joins Engberg-Perdersens analysis of DSL, Rinfrets analysis of LSQ, but also
Barbers analysis of LSC (see section 3).
The same labialisation is repeated with every production of the BEACH sign
in the discourse, regardless of the actual interpretation. This shows that, for
this signer, the DET + noun sequence in French forms a single labial gestalt,
regardless of the definite/indefinite interpretation at any given moment of the
discourse.33
5.2.2 Lexematic Unit with Pointing: Marking of Specificity

In this sub-section, we focus on cases not involving any TU, which are very simi-
lar to the type of constructions examined in the literature. We therefore discuss
the views put forward in the various studies about the (pre- or post-nominal)
position of pointing in relation to LUs in LSF, and about the potential role of
pointing in the expression of (in)definiteness or (non)specificity.
5.2.1.1 Pointing Preceded by a Lexematic Unit

At the beginning of the Horse Story in the LS-COLIN corpus, the signer
describes a wide circle in the signing space to position the fence, using a trans-
fer of shape and size (TSS) depicting the shape of this fence (see Figure 13
in 5.3.1). Thus, she constructs a topographical space in front of herself (i.e. a
descriptive use of space, see above Section 3.). The signer points to the middle
of this space, with her gaze oriented in the same direction (looking down, cf.
Figure 9, left). She then introduces the main character, the horse, using an LU
(Figure 9, right), accompanied by a camera-directed gaze and a clear labialisa-
tion che-val (Fr. horse).
(2) Shape of the fenceTU(TSS) THEREPOINTING HORSELU gallopTU(PT)34
Translation: There is a fence, and there, inside it, there is a horse galloping, like
this.
Note that pre-LU pointing combines two values, locative and specific. The
interpretation of the LU [HORSE] is indefinite but specific, which could be
translated as there is a horse, standing right there. Specificity is primar-
ily expressed by the anchoring of the horse entity to the topographic space
through pointing.
33 See Zribi-Hertz and Jean-Louis (this volume) for a parallel reanalyis of the French feminine
definite determiner as part of some lexical nouns in Martinique creole.
34 Utterance (2) is used to illustrate two phenomena: pointing before the LU, described here,
and an utterance-initial TU (discussed in section 5.3.1).
figure 9 Pointing (left) then LU [HORSE] (right)

5.2.1.2 Pointing Following a Lexematic Unit

(3) noticeTU(PT) from afarTU(TSS) HOUSELU THEREPOINTING L.A.dactylo
(there) PHARMACYLU cross on the buildingTU(TSS)
Translation: And he notices, in the distance, a building with a cross on it,
its a pharmacy.
In utterance (3) above (see Figure 10), the initial pointing THEREPOINTING,
under discussion, is literally made on the HOUSE sign, a spatialised LU whose
iconic value is reactivated35 by gaze. In fact, this sign is seen and made in the
top right area of the signing space, i.e. from the point of view of the character
who sees the house in the distance (a cow, in this case). The first pointing
is followed by a second pointing, of another type, THERE in finger spelling,
made in a neutral location at the centre of the signing space, and strengthened
by a clear labialisation l (Fr. there). The second pointing is of a particular type
which is used to introduce the following LU, PHARMACY, an LU that cannot
be spatialised by itself, for it is obligatorily made on the signers forehead. The
35 The gaze accompanying this LU is sufficient to shift the intent from saying to the iconic saying
while showing intent. The gaze reactivates the LUs latent iconicity, which is irrelevant so long
as the UL is used as such, that is without any showing intent.
signer performs a TSS in the form of a cross appearing at the spot where the LU
HOUSE was placed, directly leading to the locus created by the first pointing.
This explains the purpose of the first pointingto spatially anchor not only the
cross, a prominent and prototypical marker of pharmacies, but also the neutral
LU PHARMACY, which is thus instantiated and becomes highly specified. As in
(1), the specificity of the referent is marked by (i) its anchoring in the descriptive
space, triggered by spatialisation and the switchover into the illustrative intent
of HOUSE, and (ii) by the pointing that accompanies the gaze, thereby carrying
a descriptive locative function.
Consequently, it seems that above all, the function of pointing is to mark
specificity rather than definiteness, its significance being that it is accompanied
by gaze. Finally, its pre- or post LU position does not seem relevant in LSF.
figure 10 Spatialised LU [HOME] (left) followed by pointing (right)

5.2.3 Lexematic Unit Instanciated by a Transfer Unit: A Determination

Function?
We now illustrate cases in which an LU is immediately associated with a
subsequent TU, an extremely common sequence in first-mention contexts in
LSF. These combinations are not taken into account in the literature on SL
reference resolution.
The TU associated with the LU may be any of the transfer structures. The first
example of this section (utterance 4, Figure 11) involves a personal transfer (PT).
The context of this example is the beginning of the story: after the presentation
of the setting, the main character, the horse, is introduced. Personal transfer
both instantiates and specifies the LU HORSE. The utterance illustrates a very
classical LSF structure with the LU in topical function and the TU, specifying
by nature, in focus function.
(4) HORSELU gallopTU HAPPYLU

Translation: There is a horse,36 it gallops like this, happy
figure 11 LU [HORSE] (left) then PT of a horse galloping (right)

In the second example (utterance 5, Figure 12), the TU is a TSS (transfer of shape
and size), or actually a series of TSSs used to describe in detail the shape of
the wooden fence. This function of precise, specifying description is typical of
transfer structures and of the saying while showing register.
(5) WOODLU shape of fencesTU (TSS x 3)

Translation: there are wooden fences, like this, with this shape.
36 A more faithful translation would be it is about a horse, and in focus there is one that gallops
like this, happy.
The signer opens the story with a very detailed description of a long wooden
fence. As is the rule in this type of constructioncharacteristic of the saying-
without-showing mode of telling (see section 3.)the LU WOOD is accompa-
nied by a gaze directed towards the camera (viz. the addressee) and a slight
labialisation of bois (Fr. wood). Then, during the production of the TSSs spec-
ifying (and illustrating) the shape of the fence, the gaze follows the dominant
hand (displaying the shape), while the facial expression depicts the smaller
pickets in the fence structure.
figure 12 LU [WOOD] (left) followed by two transfers of size and shape used to describe the fence
(centre and right)
In these two examples, as in the very common LU/TU combination found in our
corpus, it is really the TU that instantiates and anchors the discourse entity. The
LU as such is a bare noun, i.e., it carries the conceptual and maximally-generic
content (horse, wood) alone, long before any determination or even longer
before any question of (in)definiteness. The basic function of what is called
nominal determination is to delimit the extension of this entity. As such, it
does not seem excessive to suggest that in these constructions, TUs function
as quasi-determiners of the entity referred to by the nominal (LU). This is not
their only function, as they carry additional information (on the action and
mode of action in the PT, and on the details of shape and position of the fence
in the TSS). However, in both cases the entity is highly specified through the
TU alone.
As we will see below, a TU can also introduce a referent directly and mark
its instantiation and referential status in discourse.
5.3 The First Unit Is a Transfer Unit

5.3.1 A Single Transfer Unit Instantiating a Referential Entity
In the beginning of the Horse Story, the signer in the chosen example37
describes a wide circle in the signing space depicting and positioning the fence,
through a TSS (transfer of shape and size) which describes the shape of the fence
(Figure 13, utterance 6). This TU not only illustrates the fence, but also builds a
descriptive (topographical) space in which the various characters can later be
located and specified.
(6) shape of the fenceTU(TSS) THERE pointing HORSELU skip TU(PT)

Translation: There is a (surrounding) fence, and there, inside, there is a
horse skipping, like this.
Significantly, the fence referent is instantiated and specified directly by the

TSS. In other words, a TU can instantiate a referential entity on its own.
figure 13 Transfer of shape and size of the fence

37 This is the same excerpt as in section 5.2.1.1, but the focus here is on the introduction of the
fence entity.
5.3.2 Transfer Units Combined with Pointing

Example (7) illustrates a combination involving pointing framed by two TUs, a
double transfer before and a personal transfer after (utterance 7, Figure 14).
(7) turn around to observe the scene TU (Double Transfer) THERE Pointing rumi-
nate TU (PT)
These two transfers deserve a brief description: the double transfer features
a bird (represented by the proform flattened O signed with the dominant
hand) watching the horse (represented by the signers facial expression in this
role), while the horse is about to jump over the fence (represented by the
non-dominant hand). This unit is particularly complex, both morphologically
and semantically. This is undoubtedly why the signer feels the need to add
pointing just after the unit, to locate the referent cow in the scene before
taking on the role of the cow in the following personal transfer (PT). Notice
that her gaze is not directed at the locus of pointing. The function of pointing
here is only to locate, as in pointings preceding or following LUs. In other words,
the current position of pointing, sandwiched between two TUs, does not seem
to alter the value of pointing.
figure 14 Pointing (middle) surrounded by a double transfer (left) and a personal transfer
(right).
This example illustrates the variety in the distribution of pointing signs and is
evidence that we must take into consideration not only the combinations of
pointing with LUs, but also its less frequent combination with TUs.
5.3.3 Transfer Unit Instantiated by a Lexematic Unit

As noted above, the most common case of LU-TU combinations is an initial,
thematic LU, followed by a focal TU. We propose that the TU in this combi-
nation serves as a determiner for the referent signalled by the LU. Interestingly,
the reverse sequence is also possible, although much rarer, and triggers a strong
rhetorical effect. This is illustrated in example (8) below, in which a TU (here,
two successive TUs, a situational transfer and a personal transfer) is followed
by a LU (Figure 15).
Thus, at the beginning of the Horse Story, immediately after setting up the
rural environment, the signer introduces the main character (the horse), not by
the noun but by depicting the action it performs, seen from two distinct view-
points presented sequentially: first a situational transfer showing the scene
from a distance, then a personal transfer which shows the scene from a closer
vantage point.
(8) to gallop TU (ST) to gallop TU (PT) HORSELU

Translation: (In the country, in a meadow), something gallops, gallops
like this, its a horse.
In this case, the TUs are topical while the LU HORSE takes on the focus func-
tion. This inversion, compared to the more common combination discussed
above, produces a rhetorical zoom effect, creating a little suspense (what-
ever is that thing which is moving like this? Ah, its a horse!).
figure 15 Situational transfer of the galloping horse in the field (left), personal transfer of the
galloping horse (middle), LU [HORSE] (right)
This sharp stylistic effect contrastively confirms the prototypical, unmarked

character of the topical-LU/focal-TU combination. At the same time, this exam-
ple illustrates the expressive resources made available by the two major types
of units and the different ways in which they instantiate a referent.
Lastly, we would like to emphasize the frequent presence of a specific kind
of pointing in LSF discourse, formally very similar to undirected pointing,
which has been analysed as directly involved in the expression of (in)defi-
niteness (see Engberg-Pedersen 1993; MacLaughlin 1997) or specificity/non-
specificity (Zimmer & Patschke 1990; Barber 2012). This point is discussed in
the following section.
5.4 Undirected Pointing: A Modal-Communicative Function

In LSF discourse, we have identified a particular type of pointing which is
not used to activate a point in the signing space (i.e., to create a locus). It
is characterised by very low muscle tension, although it is usually oriented
slightly upwards.38 This pointing therefore displays the characteristics of the
un-directed pointing described in various studies discussed above (see sec-
tion 3.).
Below we propose two examples illustrating the particular role of this type
of pointing. In our opinion this type of un-directed pointing in LSF serves
communicative and expressive functions of language rather than signals the
referential status of discourse referents.
The first example (utterance 9), located at the very beginning of the dialogue
about the beach in Marseille (see ex. 4.3.1), illustrates the classical function of
this type of pointing. It is done with a soft index finger and its direction is
irrelevanteven if the motion is directed upwards (see figure 16), it does not
serve to create a locus. The signers eyes are tightly closed, as if to think better
and draw the attention of the addressee. This pointing serves both to emphasise
the signer-addressee interaction (cf. our translation you see) and to introduce
a topic of discussion, the beach in this case. We therefore refer to this type of
pointing as modal-enunciative pointing.
(9) WELL Pointing ALSOLU MELU KNOWLU WHERELU BEACHLU RELAXLU

Translation: Well, you see, I would also like to know where I can find a
beach to relax on
38 E.g., Cuxac (2000: 282). See also Garcia et al (2011).

figure 16 Modal-enunciative pointing, by signer 1 (left photo)

In the second examplethe response to the preceding utterancesigner 2

repeats the question asked by signer 1 (a beach?) and seems to think about
the best way to reply. He suddenly has a clear idea of how to start describing
the route to the nearest beach. This stage of planning the best way to organise a
suitable response is signalled by several modal-enunciative markers (utterance
10): first the LU [YES], gestured very slowly as if to allow time to think it
over, followed by the topical LU [HARBOUR]. This is where modal-enunciative
pointing occurs (Figure 17), with the formal characteristics noted above (soft,
undirected pointing, with eyes closed). This pointing is followed by a transfer
of shape and size describing the shape of the harbour. This modal-enunciative
pointing calls on the addressees participation and his/her shared knowledge
(cf. our translation you see) in identifying the harbour referent, while the
referent itself is expressed by a TU. Consequently, the interpretation here is
that of a definite and specific referent (the starting point from which the
signer will plot out the route to the desired beach later on in the discourse).
Thus, this pointing, which is directly related to a TU, closely resembles the
cases considered in the literature. While not a definiteness marker in itself, it
contributes to the definite interpretation of the referent specified by the TU.
(10) YES LU HARBOUR LU you seePointing shape of harbour TU (TSS)

Translation: Yes , the harbour you see, the harbour, there, like that,
(etc.)
figure 17 Modal-enunciative pointing by signer 2 (right photo)

6 Conclusions and Research Directions
We have stressed that our aim in this chapter is not to provide an exhaustive
description of the processes involved in the expression of nominal reference,
and even less so of definiteness, in LSF. Rather, our main contribution (and,
in our view, the contribution of the Semiological Model) is to show to what
extent: (i) an essential prerequisite for such a study is the description of the
range of mechanisms available for the instantiation of referential entities in

SL discourse; (ii) this description must take into consideration not only the
interaction between lexematic units and pointing units (which are the focus of
the literature), but also between these types of units and what we call transfer
units.
Our primary goal has been to highlight the wide range of cases illustrating
the instantiation of referential entities in LSF, although we deliberately limited
our discussion to first-mention cases occurring at the beginning of discourse
production. In particular, we have shown that transfer units are fully referen-
tial (and in our view, fully controlled by Grammar, since they may introduce or
instantiate referents on their own). Moreover, the frequent and apparently pro-
totypical (unmarked) close association between LUs and TUs in introducing
new discourse referents is of great interest. In the sequential LU-TU combina-
tion, the LU typically works as a bare noun (maximally generic) while the TU,
in focus, provides the actual instantiation of the referent (discourse anchor),
whose extension it limits and specifies. In light of these two considerations,
it seems reasonable to acknowledge the determining function of TUs for the
nominal entity designated by the LU.
We have noted that the distribution of pointing in relation to LUs is not rel-
evant in LSF for the expression of either definiteness or specificity. In contrast,
we suggest that in further research the distribution of manual pointing in rela-
tion to gaze should be examined more closely. Furthermore, given our intention
to expand reflection on nominal determination beyond the expression of ref-
erence stricto sensu, it seems necessary to develop further in-depth analysis of
un-directed pointing, frequently found in other SLs, and which we defined as
modal-enunciative. More generally, we believe that refining the description of
specificity and definiteness marking in LSF calls for an analysis of the nature of
(topographical/ diagrammatic) constructed spaces and their combinations.
We hope, at the very least, to have captured the complexity of the data which
still need to be analysed, and shown the potential contribution of SL research
to the understanding of the linguistic expression of reference, provided the
specificities of these languages with respect to spoken languages are duly taken
into consideration.
References
Ahlgren, Inger. 1990. Deictic pronouns in Swedish and Swedish Sign Language. In
Theoretical issues in sign language research, edited by Susan D. Fischer and Patricia
Siple, 167174. Chicago, IL: University of Chicago Press.
Antinoro Pizzuto, Elena, and Micaela Capobianco. 2008. Is pointing just pointing?
Unravelling the complexity of indexes in spoken and signed discourse. Gesture
8:82103.
Antinoro Pizzuto, Elena, Paolo Rossini, Marie-Anne Sallandre and Erin Wilkinson.
2008. Deixis, Anaphora and Highly Iconic Structures: Cross-linguistic Evidence
on American (ASL), French (LSF) and Italian (LIS) Signed Languages. In Sign
Languages: spinning and unraveling the past, present and future. TISLR9, forty five
papers and three posters from the 9th. Theoretical Issues in Sign Language Research
Conference, edited by Ronice Mller de Quadros, 475495. Petrpolis/RJ. Brazil:
Editora Arara Azul.
Ariel, Mira, 1990. Accessing Noun-Phrase Antecedents. London: Routledge.
Baker, Charlotte and Carol A. Padden. 1978. Focusing on the nonmanual components
of American Sign Language. In Understanding Language through Sign Language
Research, edited by Patricia Siple, 5990. New York: Academic Press.
Barber, Gemma. 2012. The meaning of space in Catalan Sign Language (LSC): Ref-
erence, specificity and structure in signed discourse. PhD dissertation, Universitat
Pompeu Fabra, Barcelona.
Battison, Robin. 1973. Phonological deletion in American Sign Language. Sign Lan-
guage Studies 5:119.
Benveniste, Emile. 1966. Problmes de linguistique gnrale, volume 2. Paris: Gallimard.
Boutora, Lela and Annelies Braffort. 2011. Corpus DEGELS1, oai:sldr.fr:sldr000767
Bouvet, Danielle. 1996. Approche polyphonique dun rcit produit en LSF. Lyon: Presses
Universitaires de Lyon.
Brennan, Mary. 1990. Productive morphology in British Sign Language. In Current
trends in European Sign Language Research: Proceedings of the third European
Congress on Sign Language Research, edited by Samuel Prillwitz and Thomas Voll-
haber, 205228. Hamburg: Signum Press.
Brennan, Mary. 2001. Encoding and capturing productive morphology. In Sign Tran-
scription and Databases Storage of Sign Information, Sign Language and Linguistics,
volume 4, number 1/2, edited by Ronnie Wilbur, 4762. Amsterdam/Philadelphia:
John Benjamins.
Cormier, Kearsy, David Quinto-Pozos, Zed Sevcikova and Adam Schembri. 2012. Lexi-
calisation and de-lexicalisation processes. In Sign languages: Comparing depicting
constructions and viewpoint gestures, Language and Communication 32:329348.
Culioli, Antoine. 1990. Pour une linguistique de lnonciation. Paris: Ophrys.
Cuxac, Christian. 1985. Esquisse dune typologie des Langues des Signes. Journe
dtudes 10 (special issue: Autour de la Langue des Signes), 3560. Paris: Universit
Ren Descartes.
Cuxac, Christian. 1996. Fonctions et structures de liconicit des langues des signes.
Thse de doctorat dtat. Paris: Universit Paris 5.
Cuxac, Christian. 2000. La Langue des Signes Franaise (LSF). Les voies de liconicit.
Faits de Langue 1516. Paris: Ophrys.
Cuxac, Christian. 2004. Phontique de la LSF: une formalisation problmatique.
Silexicales 4 (special issue) Linguistique de la LSF: recherches actuelles: 93113.
Cuxac, Christian; Annelies Braffort, Annick Choisier, Christophe Collet, Patrice
Dalle, Ivani Fusellier, Gwenalle Jirou, Fanch Lejeune, Boris Lenseigne, Nathalie
Monteillard, Annie Risler and Marie-Anne Sallandre. 2002. Corpus LS-COLIN,
http://cocoon.tge-adonis.fr/exist/crdo/meta/crdo-FSL-CUC021_SOUND
Cuxac, Christian and Marie-Anne Sallandre. 2007. Iconicity and arbitrariness in
French Sign Language: Highly Iconic Structures, degenerated iconicity and diagram-
matic iconicity. In Verbal and Signed Languages: Comparing Structures, Constructs
and Methodologies, edited by Elena Pizzuto, Paola Pietrandrea and Raffaele Simone,
1333. Berlin: Mouton de Gruyter.
Cuxac, Christian and Elena Antinoro Pizzuto. 2010. Emergence, norme et variation
dans les langues des signes: vers une redfinition conceptuelle. In Langage et Socit
131 (special issue Sourds et langue des signes. Norme et variations), edited by Brigitte
Garcia and Marc Derycke, 3753.
DeMatteo, Asa. 1977. Visual Imagery and visual Analogues in American Sign Lan-
guage. In On the other hand: New Perspectives on American Sign Language, edited
by Lynn A. Friedmann, 109136. New York: Academic Press.
Ducrot, Oswald. 1984. Le dire et le dit. Paris: Editions de Minuit.
Dudis Paul. G. 2004. Body partitioning and real-space blends. Cognitive Linguistics 15:
223238.
Emmorey, Karen (ed.). 2003. Perspective on Classifier Constructions in Sign Languages.
Mahwah: Lawrence Erlbaum Associates.
Engberg-Pedersen, Elisabeth. 1993. Space in Danish Sign Language. The Semantics and
Morphosyntax of the Use of Space in a Visual Language. Hamburg: SIGNUM-
Verlag.
Engberg-Pedersen, Elisabeth. 2003. From Pointing to Reference and Predication: Point-
ing Signs, Eyegaze, and Head and Body Orientation in Danish Sign Language. In
Pointing. Where Language, Culture and Cognition Meet, edited by Sotaro Kita, 269
292. Mahwah: Lawrence Erlbaum Associates.
Epstein, Richard. 1995. Larticle dfini en ancien franais: lexpression de la subjectiv-
it. Langue franaise 107: 5871.
Epstein, Richard. 2002. The Definite Article, Accessibility, and the Construction of
Discourse Referent. Cognitive Linguistics 12.4: 333378.
Fauconnier, Gilles. 1985. Mental Spaces. Cambridge, MA: MIT Press.
Fauconnier, Gilles. 1997. Mappings in Thought and Language. Cambridge: Cambridge
University Press.
Fauconnier, Gilles and Mark Turner. 1996. Blending as a Central Process of Grammar.
In Conceptual Structure, Discourse and Language, edited by Adele Goldberg, 113130.

Stanford: CSLI Publications.
Friedman, Lynn A., (ed.). 1977. On the Other Hand. New Perspectives on American Sign
Language. NY/London: Academic Press.
Frishberg, Nancy. 1975. Arbitrariness and iconicity: historical change in American Sign
Language. Language 51:676710.
Fusellier-Souza, Ivani. 2006. Emergence and development of Signed Languages: from
diachronic ontogenesis to diachronic phylogenesis. Sign Language Studies 7:30
56.
Garcia, Brigitte. 2010. Sourds, surdit, langue(s) des signes et pistmologie des sciences
du langage. Problmatiques de la scripturisation et modlisation des bas niveaux en
Langue des Signes Franaise (LSF). Habilitation Thesis, University Paris 8.
Garcia, Brigitte and Derycke, Marc (eds.). 2010, Sourds et Langues des signes. Norme et
variations. Special issue of Langage et Socit, n 131, Fondation Maison des Sciences
de lHomme, Paris.
Garcia, Brigitte and Marie-Thrse LHuillier. 2011. Corpus Creagest, Dialogues en LSF
adulte.
Garcia, Brigitte, Marie-Anne Sallandre, Camille Schoder and Marie-Thrse LHuillier.
2011. Typologie des pointages en Langue des Signes Franaise (LSF) et problma-
tiques de leur annotation. In Proceedings of Dfi Geste Langues des Signes (DEGELS
2011), 107119. TALN 2011 Montpellier, 1 July 2011.
Hickmann, Maya, 2003. Childrens discourse: person, space and time across languages.
Cambridge: Cambridge University Press.
Jakobson, Roman. 1963 [1957]. Essais de linguistique gnrale. Paris: Editions de Minuit.
Johnston, Trevor and Adam Schembri. 1999. On defining lexeme in a sign language.
Sign Language and Linguistics 2:115185.
Johnston, Trevor and Adam Schembri. 2007. Australian Sign Language (Auslan): An
introduction to sign language linguistics. Cambridge: Cambridge University Press.
Jouison, Paul. 1995. Ecrits sur la Langue des Signes Franaise. Edited by Brigitte Garcia.
LHarmattan: Paris.
Kamp, Hans and Uwe Reyle. 1993. From discourse to logic. Introduction to model theo-
retic semantics of natural language, formal logic and discourse representation theory.
Dordrecht: Kluwer Academic Press.
Kegl, Judy and Wilbur, Ronnie. 1976. When does structure stop and style begin? Syntax,
morphology, and phonology vs stylistic variation in American Sign Language. In
Papers from the twelfth Regional Meeting of the Chicago Linguistics Society, edited by
Selikoko Mufwene, Carol Walker and Sanford Steever, 376396. Chicago: University
of Chicago Press.
Klima, Edward and Bellugi, Ursula (eds.) 1979. The Signs of Language. Cambridge, MA:
Harvard University Press.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, vol. 1. Stanford: Stanford

University Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar, vol. 2, Descriptive Appli-
cation. Stanford: Stanford University Press.
Liddell, Scott K. 1990. Four Functions of a Locus: Reexamining the Structure of Space
in ASL. In Sign Language Research: Theoretical Issues, edited by Ceil Lucas, 176198.
Washington D.C.: Gallaudet University Press
Liddell, Scott K. 1995. Real, surrogate, and token space: Grammatical consequences
in ASL. In Language, gesture and space, edited by Karen Emmorey and Judy Reilly,
1941. Hillsdale, NJ: Lawrence Erlbaum Associates.
Liddell, Scott K. 2003. Grammar, Gesture, and Meaning in American Sign Language.
Liddell, Scott K. and Metzger, Melanie. 1998. Gesture in sign language discourse. Journal
of Pragmatics 30:657697.
Lillo-Martin, Diane. 1995. The point of view predicate in American Sign Language. In
Language, Gesture, and Space, edited by Karen Emmorey and Judy Reilly, 155170.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Lillo-Martin, Diane. 2012. Utterance Reports and Constructed Action in Sign and
Spoken Languages. In Sign Language: An International Handbook, edited by Roland
Pfau, Markus Steinbach, and Bencie Woll, 365387. Berlin: Mouton de Gruyter.
Lillo-Martin, Diane and Edward Klima. 1991. Pointing out differences: ASL pronouns in
Syntactic Theory. In Theoretical Issues in Sign Language Research, Vol. 1: Linguistics,
edited by Susan Fischer and Patricia Siple, 191210. Chicago: University of Chicago
Press.
Lyons, John. 1977. Semantics, Volumes 12. Cambridge: Cambridge University Press.
MacLaughlin, Dawn. 1997. The structure of determiner phrases: Evidence from American
Sign Language. Doctoral dissertation, Boston University.
Mandel, Marc. 1977. Iconic devices in American Sign Language. In On the other hand:
New Perspectives on American Sign Language, edited by Lynn A. Friedmann, 57107.
New York: Academic Press.
Meier, Richard P. 1990. Person deixis in American Sign Language. Theoretical Issues
in Sign Language Research., edited by In Susan Fischer and Patricia Siple, 175190.
Chicago and London: University of Chicago Press.
Meier Richard P., Cormier Kearsy and David Quinto-Pozos (eds). 2002. Modality and
Structure in Signed and Spoken Languages. Cambridge: Cambridge University
Press.
Metzger, Melanie. 1995. Constructed dialogue and constructed action in American
Sign Language. In Sociolinguistics in Deaf Communities, edited by Ceil Lucas, 255
271. Washington, DC: Gallaudet University Press.
Meurant, Laurence. 2008. Le regard en langue des signes. Anaphore en langue des signes
franaise de Belgique (LSFB): morphologie, syntaxe, nonciation. Rivages Linguis-

tiques. Presses Universitaires de Namur, Presses Universitaires de Rennes.
Padden, Carol. 1988. Interaction of Morphology and Syntax in American Sign Language.
New York: Garland Publishing.
Padden, Carol. 1990. The Relation between Space and Grammar in ASL Verb Morphol-
ogy. In Sign Language Research: Theoretical Issues, edited by Ceil Lucas, 118132.
Washington D.C.: Gallaudet University Press.
Perniss, Pamela, M., Roland Pfau, and Markus Steinbach. 2007. Visible Variation. Com-
parative Studies on Sign Language Structure. Trends in Linguistics. Berlin: Mouton
de Gruyter.
Peirce, Charles Sanders. 1978. crits sur le signe. Translation by Grard Deledalle. Paris:
Seuil.
Pizzuto, Elena, Paola Pietrandrea and Raffaele Simone (eds). 2007. Verbal and Signed
Languages. Comparing Structures, Constructs and Methodologies. Berlin: Mouton de
Gruyter.
Poizner, Howard, Edward Klima and Ursula Bellugi. 1987. What the Hands Reveal about
the Brain. Cambridge, MA: MIT Press/Bradford.
Quer, Josep. 2005. Context Shift and Indexical Variables in Sign Languages. In Proceed-
ings of SALT XV, edited by E. Georgala and J. Howell, 152168. Ithaca, NY: Cornell
University.
Quer, Josep. 2011. Reporting and quoting in signed discourse. In Understanding Quota-
tion, edited by Elke Brendel, Jrg Meibauer and Markus Steinbach, 277302. Berlin:
Mouton de Gruyter.
Quer, Josep and Santiago Frigola. 2006. The workings of indexicals in role shift struc-
tures in Catalan Sign Language (LSC). Actes del VII Congrs de Lingstica General,
Universitat Barcelona. 18 al 21 dabril de 2006. CD ROM.
Quinto-Pozos, David. 2007. Can constructed action be considered obligatory? Lingua
117:12851314.
Riegel, Martin, Jean-Christophe Pellat and Ren Rioul. 2002. Grammaire mthodique du
franais. Paris: Quadrige/PUF.
Rinfret, Julie. 2008. Lassociation spatiale du nom en Langue des Signes Qubecoise:
Formes, fonctions et sens. PhD Thesis, University of Qubec Montral (UQAM).
Russo, Tommaso. 2004. Iconicity and Productivity in Sign Language Discourse:
An Analysis of Three LIS Discourse Registers. Sign Language Studies 4:164
197.
Sallandre, Marie-Anne. 2003. Les units du discours en Langue des Signes Franaise.
Tentative de catgorisation dans le cadre dune grammaire de liconicit. PhD diss.,
University Paris 8.
Sallandre, Marie-Anne. 2007. Simultaneity in French Sign Language Discourse. In
Simultaneity in Signed Languages: Form and Function, edited by Myriam Vermeer-
bergen, Lorraine Leeson, and Onno Crasborn, 103125. Amsterdam/ Philadelphia:

John Benjamins.
Sallandre, Marie-Anne and Brigitte Garcia. (2013). Epistemological issues in the semi-
ological model for the annotation of sign language. In Sign Language research,
uses and practices, Crossing views on theoretical and applied sign language linguistics
(Sign Language and Deaf Communities), edited by Laurence Meurant, Aurlie Sinte,
Mieke Van Herreweghe and Myriam Vermeerbergen. Berlin / New York: Mouton De
Gruyter, 159177.
Sandler, Wendy and Diane Lillo-Martin. 2006. Sign Language and Linguistic Universals.
Schembri, Adam. 2003. Rethinking Classifiers in Signed Languages. In Perspectives
on classifier constructions in sign languages, edited by Karen D. Emmorey, 334.
Mahwah, NJ: Lawrence Erlbaum Associates.
Slobin, Dan I., Nini Hoiting, Marlon Kuntze, Reyna Lindert, Amy Weinberg, Jennie
Pyers, Michelle Anthony, Yael Biederman and Helen Thuman. 2003. A Cognitive/
Functional Perspective on the Acquisition of Classifiers. In Perspectives on classifier
constructions in sign languages, edited by Karen D. Emmorey, 271296. Mahwah, NJ:
Lawrence Erlbaum Associates.
Smith, Cheri, Ella Mae Lentz, and Ken Mikos. 1988. Signing naturally: teachers curricu-
lum guide, Level 1. Berkeley CA: DawnSign Press.
Stokoe, William C. 1960. Sign Language Structure. Studies in LinguisticsOccasional
Paper, 8 (rev. ed. Linstok Press, Silver Spring, MD, 1978).
Supalla, Ted. 1978. Morphology of Verbs of Motion and Location in American Sign Lan-
guage. In American Sign Language in a Bilingual, Bicultural Context. Proceedings of
the Second National Symposium on Sign Language Research and Teaching, edited by
Frank Caccamise and Doin Hicks (eds), 2746. Coronado, CA: National Association
of the Deaf.
Suppala, Ted. 1982. Structure and acquisition of verbs of motion and location in American
Sign Language. Ph-D dissertation, University of California, San Diego.
Suppala, Ted. 1986. The classifier system in American Sign Language. In Noun Classes
and Categorization: Typological Studies in Language, 7, edited by Colette Craig, 181
214. Amsterdam/ Philadelphia: John Benjamins.
Tang, Gladys, and Sze Felix. 2002. Nominal expressions in Hong Kong Sign Language:
Does modality make a difference. In Modality and Structure in Signed and Spoken
Languages, edited by Richard Meier, Kearsy Cormier and David Quinto-Pozos, 296
320. Cambridge: Cambridge University Press.
Vermeerbergen, Myriam. 2006. Past and current trends in sign language research.
Language and Communication 26:168192.
Wilbur, Ronnie. 1979. American Sign Language and sign systems: research and applica-
tion. Baltimore: University Park Press.
Winston, Elisabeth. 1991. Spatial referencing and cohesion in an American Sign Lan-
guage text. Sign Language Studies 73:397410.
Winston, Elisabeth. 1995. Spatial mapping in comparative discourse frames. In Lan-
guage, Gesture, and Space, edited by Karen Emmorey and Judy S. Reilly, 87114. Hills-
dale, NJ: Lawrence Erlbaum Associates.
Woll, Bencie. 2003. Modality, Universality and the Similarities among Sign Languages:
an Historical Perspective. In Cross-Linguistic Perspectives in Sign Language
Research: selected Papers from TISLR 2000, edited by Ann Baker, Beppie van den
Bogaerde and Onno Crasborn, 1730. Hamburg: Signum Press.
Zimmer, June and Patschke Cynthia, 1990. A Class of Determiners in ASL. In Sign Lan-
guage Research: Theoretical Issues, edited by Ceil Lucas, 201210. Washington DC:
Gallaudet University Press.
Zucchi, Sandro. 2004. Monsters in the Visual Mode? Ms., Universit degli studi di Milano.
http://filosofia.dipafilo.unimi.it/~zucchi/NuoviFile/LISmonsters.pdf
part 3
Noun Phrase Interpretation and

Second-Language Acquisition

When Articles Have Different Meanings:
Acquiring the Expression of Genericity
in English and Brazilian Portuguese*
Tania Ionin, Elaine Grolla, Silvina Montrul, and Hlade Santos
This paper reports on an experimental study of the expression of genericity

in the acquisition of English by native speakers of Spanish and Brazilian Por-
tuguese, and in the acquisition of Brazilian Portuguese by native speakers of
English and Spanish. English, Spanish and Brazilian Portuguese exhibit a three-
way distinction in terms of which NP types are allowed with generic and kind
interpretations. This difference leads to specific, testable predictions for the
effects of cross-linguistic influence on the expression of genericity in second
language acquisition. These predictions are tested in a small-scale study, by
means of a written, context-based Acceptability Judgment Task. The results
show that transfer from the learners native language has a limited effect, and is
overridden by considerations of register and/or input frequency. The findings
pose interesting questions for further research.
1 Introduction
An important question in the field of second language (L2) acquisition is how

L2-learners map syntax to semantics (see Slabakova 2008 for an overview). The
interpretation of articles in the second language is one of the areas in which
this question has been explored. A number of studies have examined how
L2-learners use and interpret articles, and what learners patterns of article
(mis)use and/or (mis)interpretation can tell us about the factors involved in
* We gratefully acknowledge the financial support of the University of Illinois Lemann Institute
for Brazilian Studies for the collaborative research reported here. We are grateful to our
research assistants: Justin Davidson and Gretchen Shaw (testing of native English speakers);
Jenna Kim and Tatiana Luchkina (testing of BrP learners in the U.S.); and Karina Bertolino
(testing in Brazil); and to Mnica Crivos for the testing in Argentina. Thanks to the audience of
the Third International Conference on Bare Nominals (Rio de Janeiro, November 2011), where
this work was presented. We are grateful to Anne Zribi-Hertz and to two anonymous reviewers
for insightful and helpful comments on an earlier version of this paper. All remaining errors
are our own.

368 ionin et al.
L2-acquisition (Huebner 1983; Thomas 1989; Ionin et al. 2004; Trenkic 2008,
among many others). In the case of L2-learners whose first language (L1) lacks
articles, many studies examine what factors (explicit strategies, input and
instruction, access to semantic universals) influence learners acquisition of
article meaning. When the learners L1 does have articles, the role of cross-
linguistic influence, or L1-transfer, is considered, and studies ask to what extent
learners transfer the meaning of articles in their L1 to those in their L2. Most
studies on articles and nominals in the L2 have focused on non-generic con-
texts of article use; some non-generic uses are exemplified in (1), where the
indefinites (a lion, lions) are interpreted existentially, and the definites (the
lion, the lions)anaphorically. Our current research program focuses instead
on generic contexts of article use, exemplified in (2): unlike the sentences in
(1), the ones in (2) talk about lions in general, rather than about any particular
lion(s). (We consider only nominals in subject position, as in (2), leaving the
interpretation of nominals in other syntactic positions for further research).
(1) a. A lion is running towards us. The lion is hungry.

b. I see lions in the distance. The lions are asleep.
(2) a. The lion is a dangerous animal.

b. Lions are dangerous animals.
In a series of recent studies, we have experimentally examined the expression

of genericity with native speakers of English, Spanish, and Brazilian Portuguese
(BrP) (Ionin et al. 2011a), and we have investigated the role of L1-transfer in the
interpretation of NPs with and without articles in generic environments in the
L2-English of Spanish speakers as well as in the L2-Spanish of English speak-
ers (Ionin et al. 2013); we have also looked at how L2-English learners learn
about the relationship between articles and genericity when their L1 (Russian
or Korean) lacks articles (Ionin et al. 2011b). We have recently extended this
investigation to the acquisition of BrP by English and Spanish speakers, as well
as the acquisition of English by BrP speakers. As discussed in the next sec-
tion, English, Spanish, and BrP differ from one another in how they express
generic interpretation for NPs with and without articles; the three-way dif-
ference allows us to make testable predictions, based on L1-transfer, for the
L2-acquisition of one of these languages by speakers of another. Here, we report
on a small-scale study testing these predictions, and show that while L1-transfer
does play a role in the expression of genericity in the L2, it is overridden by such
considerations as input frequency of generic expressions and register. After dis-
cussing the findings, we pose questions for further research.
when articles have different meanings 369
2 The Expression of Genericity in English, Spanish and Brazilian

Portuguese
English allows three different NP types with generic readings, as shown in (3):1
the variants with a definite singular (3a), indefinite singular (3b) and bare plural
(3e) are all sentences about hummingbirds in general; ((3a-b) can also be inter-
preted as being about a specific hummingbird, but this reading is irrelevant for
our purposes). In contrast, English definite plurals normally lack generic read-
ings, so that (3d) can only be interpreted as a sentence about a specific group
of hummingbirds (e.g., the ones outside the window), rather than humming-
birds in general. Finally, bare singulars in English are ungrammatical, whether
used generically (as in (3c)) or otherwise (we focus only on count nouns here,
leaving mass nouns aside).2
(3) a. The hummingbird is a bird. [definite singular, generic]

b. A hummingbird is a bird. [indefinite singular, generic]
c. *Hummingbird is a bird. [*bare singular]
d. The hummingbirds are birds. [definite plural, #generic]
e. Hummingbirds are birds. [bare plural, generic]
Spanish (and many other Romance languages, such as Italian and French)
differs from English in using definite plurals rather than bare plurals for generic
readings, as shown in (4d); unmodified bare plurals in preverbal (subject)
position are generally ungrammatical in Spanish (4e). With regard to singular
generics, Spanish patterns just like English (compare (4a-c) to (3a-c)).
(4) a. El picaflor es un pjaro. [definite singular, generic]

the hummingbird is a bird
1 Throughout this paper, we use the term NP in a descriptive way, to encompass both NPs and
DPs.
2 Bare singulars are acceptable with some English NPs, in specific types of contexts, as in
going to school, being in jail, lying in bed, etc.; see Stvan (1998) and Carlson, Sussman, Klein
and Tanenhaus (2006) for discussion. We are grateful to Anne Zribi-Hertz (p.c.) for bringing
this phenomenon to our attention. Our study did not include any NPs that can occur in
bare singular form in English, so we leave this issue aside at present. In future research, it
would be fruitful to consider how the existence of (lexically and contextually constrained)
bare singulars in English may influence the acquisition of bare singulars in BrP by English
speakers.
370 ionin et al.
b. Un picaflor es un pjaro. [indefinite singular, generic]

a hummingbird is a bird
c. *Picaflor es un pjaro. [*bare singular]

hummingbird is a bird
d. Los picaflores son pjaros. [definite plural, generic]

the hummingbirds are birds
e. *Picaflores son pjaros. [*bare plural]

hummingbirds are birds
Finally, BrP differs from both English and Spanish in its expression of genericity,
as shown in (5) (from Schmitt and Munn 1999, 2002; see also Mller 2002a,b).
First, BrP allows both bare and definite plurals to have generic readings (5d-e);
second, bare singulars are grammatical in BrP, and have generic readings (5c),
on a par with definite and indefinite singulars (5a-b).
(5) a. O beija-flor uma ave. [definite singular, generic]

the hummingbird is a bird
b. Um beija-flor uma ave. [indefinite singular, generic]

a hummingbird is a bird
c. Beija-flor ave. [bare singular, generic]

hummingbird is bird
d. Os beija-flores so aves. [definite plural, generic]

the hummingbirds are birds
e. Beija-flores so aves. [bare plural, generic]

hummingbirds are birds
The judgments reported in (3) through (5) above, about which NP types are
allowed in generic environments in English, Spanish and BrP, received exper-
imental support from a study with native speakers by Ionin et al. (2011a). We
next consider the sources of these generic readings.
2.1 Sentence-Level vs. NP-Level Genericity

As discussed by Krifka et al. (1995), genericity can come either from the sen-
tence level or from the NP-level. At the sentence level, generic sentences are
sentences that describe habitual or characteristic events/behaviours, in con-
trast to episodic sentences, which discuss specific events/behaviours. At the
NP-level, kind-denoting NPs denote kinds rather than individuals. The two
sources of genericity are in principle independent. Generic sentences can de-
scribe characteristic properties or behaviours of specific individuals, as in Rex
barks loudly, or That tiger is very dangerous; these are cases of sentence-level
genericity without NP-level genericity. Conversely, a kind-denoting NP can
occur in an episodic (non-generic) sentence, as in The rat arrived in Australia
in 1788 (from Krifka et al. 1995): this sentence describes an episode in the his-
tory of the rat species, and the rat denotes the rat-kind rather than a specific
rat.
The sentences in (3) through (5) above are all examples of sentence-level
genericity: the generic reading is retained if the subject NP is replaced by a
proper name (e.g., Polly is a bird describes Pollys characteristic property). We
now consider which of the NP types in (3) through (5) can denote kinds at
the NP-level, using Krifka et al.s (1995) diagnostic of compatibility with a kind
predicate, such as be extinct / common / rare.
As shown in (6), in English, kind-reference at the NP-level is available to
definite singulars (6a) and bare plurals (6e), but not to indefinite singulars
(6b) or definite plurals (6d) (for completeness, we include bare singulars (6c),
which are ungrammatical). Spanish differs minimally from English, in that
kind-reference at the NP level is available to definite singulars and definite
plurals (rather than bare plurals, as in English).
(6) a. The hummingbird is rare in the United States. [definite singular]

b. #A hummingbird is rare in the United States. [#indefinite singular]
c. *Hummingbird is rare in the United States. [*bare singular]
d. #The hummingbirds are rare in the United States. [#definite plural]
e. Hummingbirds are rare in the United States. [bare plural]
In BrP, kind-reference at the NP-level is available for definite singulars (7a)

and definite plurals (7d), and unavailable for indefinite singulars (7b). In the
case of bare singulars (7c) and bare plurals (7e), Schmitt and Munn (1999,
2002) as well as Dobrovie-Sorin and Oliveira (2008) argue that kind-reference
is available, while Mller (2002a,b, 2003) argues that it is not. The experimental
study of Ionin et al. (2011a) found that bare plurals were accepted as much as
definite plurals in kind-reference contexts such as (7d-e), supporting Schmitt
372 ionin et al.
and Munns analysis. Ionin et al.s findings on bare singulars (7c) were incon-
clusive: bare singulars were rated relatively low with kind-reference, but they
were also not rated as high as expected in generic sentences (see section 3.4
below for a summary of the results).
(7) a. O beija-flor raro em So Paulo. [definite singular]

the hummingbird is rare in Sao Paulo
b. #Um beija-flor raro em So Paulo. [#indefinite singular]

a hummingbird is rare in Sao Paulo
c. ?Beija-flor raro em So Paulo. [?bare singular]

hummingbird is rare in Sao Paulo
d. Os beija-flores so raros em So Paulo. [definite plural]

the hummingbirds are rare in Sao Paulo
e. Beija-flores so raros em So Paulo. [bare plural]

hummingbirds are rare in Sao Paulo
The ability of BrP bare singulars to have kind-reference was also studied by
Oliveira, Silvo and Bressane (2010), who asked 200 native speakers of BrP to
rate the acceptability of sentences with bare and definite singular NP sub-
jects. They tested both sentences with kind predicates, and sentences with
episodic predicates that impose a kind reading on the subject NP (cf. the
example The rat arrived in Australia in 1788). Definite singulars were found
to be more acceptable than bare singulars, with both types of predicates,
and bare singulars were found to be more acceptable with kind predicates
than with episodic predicates. Overall, the findings of Oliveira et al. (2010),
like those of Ionin et al. (2011a), were somewhat inconclusive: bare singu-
lars were to some extent allowed with kind readings, but not as fully as def-
inite singulars. It is possible that both studies were confounded by register
and/or modality: the studies used formal, written judgment tasks, whereas bare
singulars are observed to be more natural in casual, oral speech (cf. Mller
2002b, Munn and Schmitt 2005). This issue is addressed by the other part of
Oliveira et al. (2010), a corpus study of bare singulars in both oral and writ-
ten corpora. Oliveira et al. found only a single instance of a bare singular NP
subject with a kind predicate in the oral corpora (out of 22 instances total
of kind predicates in the corpora), and several instances in the written cor-
pora, including the example in (8). Thus, the status of bare singulars in BrP
remains inconclusive. The findings that bare singulars with kind readings are
given low ratings in judgment tasks suggest that bare singulars cannot be kind
terms. At the same time, the fact that bare singulars do sometimes (if infre-
quently) appear with kind predicates suggests the opposite. Furthermore, the
fact that bare singulars were rated relatively low in subject position, in a writ-
ten test, even in the absence of a kind predicate (in generic sentences, Ionin
et al. 2011a) suggests that there may be syntactic restrictions on bare singulars
in subject position which are independent of the availability of kind read-
ings, and/or that register rather than grammar is responsible for the low rat-
ings.
(8) Boto chins est extinto, dizem cientistas.

river-dolphin Chinese is extinct, say scientists
The Chinese river dolphin is extinct, scientists say.
2.2 The Well-Defined Kind Restriction

Finally, yet another fact about kind-reference at the NP level has to do with the
so-called well-defined kind (WDK) restriction (term from Krifka et al. 1995;
see also Carlson 1977; Vergnaud and Zubizarreta 1992; Dayal 2004). While both
bare plurals and definite singulars in English can denote kinds (see (6)), the
distribution of the latter is more restricted than the former, as shown in (9)
(example from Carlson 1977, cited in Krifka et al. 1995:11). (9a), which discusses
a well-defined kind, is perfectly fine on the kind interpretation of the definite
singular NP, but (9b), which is not about an existing kind, is ill-formed. In
contrast, bare plurals are fine in both cases (9cd).
(9) a. The Coke bottle has a narrow neck.

b. #The green bottle has a narrow neck. [non-generic reading ok]
c. Coke bottles have narrow necks.
d. Green bottles have narrow necks.
Although the WDK restriction was originally discussed as a fact specific to

English (see Carlson 1977; Vergnaud and Zubizarreta 1992), Dayal (2004) argues
that it is a restriction on definite singular kind terms cross-linguistically. The
experimental findings of Ionin et al. (2011a) support this claim: generic sen-
tences of the form in (9b), with definite singular NPs that denoted non-well-
defined kinds, were rated low in English, Spanish, and BrP. The WDK restriction
did not affect any other NP type.
To sum up, the three languages under discussion behave the same with
regard to definite singulars (which can denote kinds, and are subject to the
374 ionin et al.
WDK restriction) and indefinite singulars (which can occur in generic sen-
tences, and cannot denote kinds). In the case of plural NPs, English allows only
bare plurals to have generic/kind readings, Spanish allows only definite plurals
to have these readings, and BrP allows both. Finally, BrP is the only one of the
three languages that allows bare singulars with generic readings, although the
availability of kind-reference to bare singulars is still unresolved.
2.3 Theoretical Framework

Following Krifka et al.s (1995) analysis of sentence-level genericity, we assume
that indefinite singulars cross-linguistically cannot refer to kinds, and obtain
generic readings as a result of binding by a generic operator. Turning to NP-level
genericity, we adopt the semantic framework of Chierchia (1998) and Dayal
(2004), in which kind readings of plural NPs are derived by the Down operator,
which maps properties to functions from situations to the maximal individ-
ual satisfying that property in that situation. Following Dayal (2004, 2011), we
assume that languages differ in whether the Down operator is lexicalized on
the definite article (Spanish, which uses definite plurals for kind-reference),
applies covertly (English, which uses bare plurals for kind-reference), or is lex-
icalized, but only optionally (BrP, which allows both definite and bare plurals
for kind-reference).3
Turning to definite singular generics, we follow Dayal (2004) in assuming
that, like canonical definites, they are derived by the Iota operator, which
maps properties to the maximal individual satisfying that property. When
the Iota operator applies to a common NP, it returns a unique individual
bearing the property denoted by that NP; when it applies to a taxonomic NP, it
3 In this framework, bare plurals in English are unambiguously kind-denoting in all environ-
ments, while Spanish definite plurals are ambiguous between true definite readings (derived
by the Iota operator) and kind readings (derived by the Down operator). On an alternative
Ambiguity approach, based in the framework of Heim (1982) (e.g., Wilkinson 1991), English
bare plurals are only kind-denoting when they appear with kind predicates. In all other envi-
ronments, bare plurals are indefinite: e.g., in generic sentences like (3e), bare plurals, like
indefinite singulars, are indefinite terms bound a generic operator (on Chierchias approach
adopted in the present paper, the generic operator quantifies over instances of the kind).
Extending the Ambiguity approach to Spanish definite plurals leads to three-way ambigu-
ity between canonical definite readings, kind readings, and indefinite readings (cf. Zampar-
elli 2002). For the sake of simplicity, we adopt the Chierchia/Dayal approach in our paper,
although nothing in our study hinges on this choice. See Ionin et al. (2011a) for more discus-
sion.
returns the unique taxonomic entity bearing the property denoted by the NP.
Thus, the lion can denote either the unique lion in the discourse context, or
the unique taxonomic entity lion. Dayal suggests that the WDK restriction
on definite singular generics is a pragmatic consequence of definite singular
generics denoting taxonomic entities (see (9)): in the absence of special con-
text, green bottles are not taxonomic entities, unlike Coke bottles. In contrast,
plural generics are not subject to the WDK restriction, since they do not denote
taxonomic entities, but rather are derived by the Down operator applying to a
property.
Turning to bare singulars in BrP, there are two possibilities, in light of the
still-unresolved empirical questions discussed above. For Mller (2002a,b),
bare singulars are indefinite terms, which can be bound by a generic operator
and hence occur in generic sentences, but cannot have true kind-reference. An
alternative analysis of bare singulars as kind terms is proposed by Dobrovie-
Sorin and Oliveira (2008): building on the finding that BrP bare singulars are
number-neutral (Schmitt and Munn 2002), they propose that kind readings of
bare singulars are derived by the Down operator applying to a number-neutral
NP. On this analysis, kind readings of bare singulars are analogous to those of
bare plurals and definite plurals (which are derived by the Down operator), and
distinct from those of definite singulars (which are derived by the Iota operator
applying to a taxonomic entity). This view is supported by the fact that defi-
nite singulars in BrP, but not bare singulars, are subject to the WDK restriction
(Dobrovie-Sorin and Oliveira 2008; Ionin et al. 2011a).
2.4 Second Language Acquisition of Genericity in English and BrP

The cross-linguistic differences in the expression of genericity discussed above
allow us to formulate specific predictions for second-language acquisition in
this domain, with reference to transfer from the L1 to the L2.
Most L2-researchers today agree that at least some aspects of L2-acquisition
are influenced by the learners native language, a process known as L1-transfer
(see, e.g., Dechert and Raupach 1989; Odlin 1989; Gass and Selinker 1992;
Schwartz and Sprouse 1994, 1996; Schwartz 1998; Ellis 2006; among many oth-
ers). Within the field of generative approaches to L2-acquisition, a particularly
influential model is the Full Transfer / Full Access (FT/FA) model of Schwartz
and Sprouse (1994, 1996), on which learners initially transfer the properties
of their L1-grammar to their L2, but are also eventually able to acquire cate-
gories and features of the L2-grammar not instantiated in the L1, through direct
access to Universal Grammar (UG). While the FT/FA as originally formulated
focused primarily on morphosyntactic phenomena, recent studies on the L2-
acquisition of semantics have provided evidence that L1-transfer also occurs
376 ionin et al.
on the semantic level. Domains in which L1-transfer of semantics has been

attested include quantifier scope (e.g., Marsden 2004), aspectual interpreta-
tion (e.g., Gabriele 2009) and the interpretation of nominals (e.g., Slabakova
2006; Ionin et al. 2013). Following prior literature, and consistent with the FT/FA
model, we hypothesize that L2-learners initially transfer the semantics of their
L1 onto their L2. That is, during the earlier stages of acquisition, and until
the input informs them otherwise, learners assume that the semantic distinc-
tions lexicalized in the L1 (such as perfective aspect, or definiteness, or kind-
formation) are also lexicalized in the L2, that the distinctions not lexicalized in
the L1 are not lexicalized in the L2, and that semantic operations apply in the
same way in the two languages.
Testing this hypothesis specifically with regard to the semantics of nomi-
nals, we focus on four distinct populations: adult Spanish-speaking and BrP-
speaking learners of English, as well as adult English-speaking and Spanish-
speaking learners of BrP. Comparing two learner groups for each target lan-
guage allows us to isolate the effects of L1-transfer from any developmental
effects not related to the L1. The main goal of our study was to determine the
extent to which L1-transfer influences the expression of genericity when the L1
and the L2 have subtly different article semantics.
As discussed above, English, Spanish and BrP behave the same on definite
singular generics (which denote taxonomic entities) and indefinite singular
generics (which are indefinites bound by a generic operator), so for those NP
types, L1-transfer should lead to fully target-like performance for all learner
groups. In contrast, these three languages differ on the interpretation of plural
NPs in generic/kind contexts, with the Down operator lexicalized obligatorily
in Spanish, optionally in BrP, and not at all in English. If L1-transfer plays a
role here, then English-speaking learners of BrP should prefer bare plurals to
definite plurals for generic/kind readings, while Spanish-speaking learners of
BrP should do exactly the opposite. In the case of learners of English, BrP
speakers should allow bare plural generics (which are grammatical in BrP as
well as in English) more readily than Spanish speakers, while both groups
should also incorrectly allow definite plural generics. Finally, L1-transfer should
lead both English-speaking and Spanish-speaking learners of BrP to reject
bare singulars, while BrP-speaking learners of English should overaccept bare
singulars in generic contexts.
3 Experimental Study
3.1 Participants
In order to test the predictions of L1-transfer outlined above, we have con-
ducted two small-scale studies, one on the acquisition of English and the other
on the acquisition of BrP. The adult learners in both studies took a multiple-
choice cloze test estimating their proficiency in the target language (in the case
of BrP, the cloze test was supplemented by a vocabulary test). The cut-off for
inclusion in the study was a score of at least 50% on the proficiency test (this
was above chance-level, since the test questions had more than two correct
answers).
The participants in the English study were 26 adult Spanish speakers study-
ing English in Argentina (mean proficiency test score 80 %, range 50 % to 98 %),
and 16 adult BrP speakers studying English in Brazil (mean proficiency test
score 80%, range 55% to 98%).4 An independent-samples t-test found no dif-
ference in proficiency test scores between the two groups of learners (p=.97).
Most of the learners had begun their study of English as children or adolescents,
in a formal classroom setting, and had never lived in an English-speaking coun-
try.
The participants in the BrP study were 14 native English speakers studying
BrP in the U.S. (mean proficiency test score 73%, range 58 % to 88 %) and 10
native Spanish speakers, of whom six were studying BrP in the U.S. and fourin
Brazil (mean proficiency test score 80%, range 60% to 90 %). An independent-
samples t-test found no difference in proficiency test scores between the L1-
English vs. L1-Spanish learners of BrP (p=.13). All 24 learners of BrP had begun
the study of BrP as adults, after age 18; the four who were living in Brazil had
only a few months of residence there. Of the 14 English-speaking learners of
BrP, 12 had studied Spanish prior to studying BrP, and one had studied French.
All but one of the Spanish-speaking learners of BrP had studied English prior
to studying BrP (the six Spanish-speaking learners of BrP who were tested in
the U.S. were all highly proficient in English, being students at a U.S. univer-
sity).5 Thus, the learners of BrP were for the most part third language rather
than second language learners of BrP, an issue we come back to in section 3.5.2.
4 The 26 Spanish-speaking learners of English were selected from a larger group of 32 learners,
in order to most closely match the BrP-speaking learners of English in proficiency. The results
of all 32 Spanish speakers performance on bare and definite plurals in English are reported
in Ionin et al. (2013).
5 Five of these six learners arrived in the U.S. as adults, for university study. The sixth was
born in the U.S. to Spanish-speaking parents; since she reported speaking Spanish as her
378 ionin et al.
The control groups in the English and BrP studies were 22 native English
speakers living in the U.S., and 19 native BrP speakers living in Brazil, respec-
tively. The results of these native-speaker participants are reported in Ionin et
al. (2011a); they are repeated here in order to allow a direct comparison to the
learner groups.
3.2 Test Instrument

The test instrument was an Acceptability Judgment Task (AJT) with contexts,
the same one used in Ionin et al. (2011a,b). Two versions of the AJT (identical
in content, format and ordering) were created, one in English and one in BrP.
Both versions were placed on the internet using the survey gizmo tool. The
AJT consisted of 40 items (20 test items and 20 fillers), and each item was a
paragraph-long story followed five different target sentences. The participants
were asked to rate each sentence for its (un)acceptability in the context of the
story, using a scale from 1 (unacceptable) to 4 (acceptable). The participants
were explicitly instructed that they did not have to rank the sentences, and
could give the same rating to two or more sentences. For the test items, the
five sentences were identical except for the form of the subject NP (and the
corresponding agreement on the verb), as shown below. The fillers tested
differences of tense and aspectual interpretation. The 20 target items were
broken down into five categories of four items each; we discuss two of the
categories here, which tested NP-level genericity and sentence-level genericity,
respectively (see Ionin et al. 2011a for discussion of other test categories). The
40 AJT items were arranged into four blocks and randomized for order of
presentation within each block.
The Kind-reference category (NP-level genericity) is exemplified in (10) and
(11) for the two languages. In this category, the subject NP is followed by a kind
predicate. The Generic category (sentence-level genericity) is exemplified in
(12) and (13). In this category, the subject NP denotes a non-well-defined kind
(e.g., green lamp) and occurs in a generic sentence; the violation of the WDK
restriction checks that learners are aware of the WDK restriction on definite
singulars.
first language (and English as her second), she was classified as a Spanish-speaking learner
of BrP for the purposes of the study. See section 3.5.2 for discussion of potential transfer from
English vs. from Spanish for learners of BrP. Interestingly, this learner, despite being bilingual
in English, showed no evidence of transfer from English to BrP: she rated bare plural generics
in BrP very low, and definite plural generics high, just like the rest of the Spanish-speaking
learners of BrP (and unlike the English-speaking learners of BrP).
The sentence variants expected to be rated high by native speakers are

highlighted in the examples below; there was no highlighting in the actual test.
(10) Kind-reference category (English): I really like going to the zoo. Unfortu-
nately, there are many animals that cant be found in a zoo, or anywhere
else. Its very sad. For example
a. The dodo bird is extinct. [definite singular]

b. A dodo bird is extinct. [indefinite singular]
c. Dodo bird is extinct. [bare singular]
d. The dodo birds are extinct. [definite plural]
e. Dodo birds are extinct. [bare plural]
(11) Kind-reference category (BrP): Eu gosto muito de ir ao zoolgico. Infeliz-

mente, h vrios animais que ns no vemos mais no zoolgico ou em
nenhum outro lugar. muito triste! Por exemplo
a. O pssaro dod est extinto. [definite singular]

b. Um pssaro dod est extinto. [indefinite singular]
c. Pssaro dod est extinto. [bare singular]
d. Os pssaros dod esto extintos. [definite plural]
e. Pssaros dod esto extintos. [bare plural]
(12) Generic category (English): My brother has been in a bad mood lately. And
no wonder: his apartment is so uncomfortable, it must be very depressing
to live there. And he has a very dim and unpleasant overhead light. I told
him he should buy a new lamp, something pleasant. For example, I know
that
a. The green lamp is very relaxing. [definite singular]

b. A green lamp is very relaxing. [indefinite singular]
c. Green lamp is very relaxing. [bare singular]
d. The green lamps are very relaxing. [definite plural]
e. Green lamps are very relaxing. [bare plural]
(13) Generic category (BrP): O meu irmo tem estado de mal humor ultima-
mente. No sem motivo: o apartamento dele muito desconfortvel e
deve ser muito deprimente morar l. E ele tem um lustre com uma luz
muito fraca e desagradvel. Eu disse a ele que ele deveria comprar uma
luminria nova: alguma coisa agradvel. Por exemplo, eu sei que
380 ionin et al.
a. A luminria verde muito relaxante. [definite singular]

b. Uma luminria verde muito relaxante. [indefinite singular]
c. Luminria verde muito relaxante. [bare singular]
d. As luminrias verdes so muito relaxantes. [definite plural]
e. Luminrias verdes so muito relaxantes. [bare plural]
If the learners in our study transfer the interpretation of nominals from their
L1, then we expect to see the following patterns of results. In the English study,
both L1-Spanish and L1-BrP learners of English are expected to be accurate with
regard to definite singular and indefinite singular generics ((10a,b) and (12a,b)),
which function similarly in all three languages. Both groups are expected to
incorrectly rate definite plural generics ((10d) and (12d)) high, since these
are fully acceptable in Spanish and BrP; at the same time, BrP speakers are
expected to be more accurate than Spanish speakers in accepting English bare
plural generics ((10e) and (12e)), given that these exist in BrP. Finally, BrP
speakers, but not Spanish speakers, are expected to rate bare singulars relatively
high in the Generic category (10c) and possibly in the Kind-reference category
(12c) (recall that the evidence on the status of bare singular kind terms in BrP
is somewhat inconclusive).
Turning to the BrP study, we again expect both learner groups (L1-English
and L1-Spanish learners of BrP) to be accurate with definite singular and indef-
inite singular generics ((11a,b) and (13a,b)), while giving lower ratings than
native BrP speakers to bare singulars ((11c) and (13c)). With respect to plu-
ral generics, L1-English learners of BrP are expected to give higher ratings to
bare plurals ((11e) and (13e)) than to definite plurals ((11d) and (13d)), while L1-
Spanish learners of BrP are expected to do the opposite. Recall that both types
of plural generics are acceptable for native BrP speakers, so both learner groups
are expected to differ from native speakers on plural generics, but in different
ways.
3.3 Results: English Study

Mean numerical ratings for each sentence type and subject group were entered
into the statistical analysis. The results for the Kind-reference and Generic
categories in the English study are presented in Figures 1 and 2, respectively. As
Figure 1 shows, both learner groups were fairly similar to the native speakers
except for noticeably lower ratings of definite singulars in the Kind-reference
category. Figure 2 shows that learners were quite similar to native speakers in
the Generic category, across all NP types.
We conducted a mixed Analysis of Variance (ANOVA) on the results for
each category, with NP type (5 levels) as the within-subjects variable, and
figure 1 Results for the Kind-reference category, English study (mean ratings)
participant group (3 levels: native speakers vs. L1-Spanish learners of English

vs. L1-BrP learners of English) as the between-subjects variable.6 This analysis
allows us to determine whether, for each category, there were differences in
ratings among the five NP types, as well as among participant groups, and
whether the pattern of ratings was similar among the participant groups. In
both Kind-reference and Generic categories, NP type had a significant (p<.05)
effect on the ratings: i.e., the five different NP types were rated differently. There
was no main effect of participant group, but there was a significant NP-type by
6 For all statistical analyses, we analyzed the Kind-reference and Generic categories separately.
The rationale for this is that the categories were set up very differently, and any differences
between them could be due to lexical material and/or contextual effects rather than the
relevant factor of NP-level vs. sentence-level genericity. In contrast, the sentence types within
each category were identical except for the form of the subject NP, so any differences in ratings
are clearly due to the NP form.
382 ionin et al.
figure 2 Results for Generic category, English study (mean ratings)
group interaction: i.e., the three participant groups did not exhibit the same
pattern of ratings. In order to explore exactly where the differences among
the groups lay, we conducted multiple follow-up comparisons (a Bonferroni
correction was used to avoid inflating the Type I error rate). First, we conducted
five one-way ANOVAs comparing the three groups on each NP type: the goal
of this comparison was to see whether the groups differ on all NP types, or
only some of them. Second, we conducted three repeated-measures ANOVAs
examining the effect of NP type within each group: the goal of this comparison
was to examine the pattern of ratings within each participant group. We report
the results for each category in turn.
3.3.1 Statistical Comparisons on the Kind-Reference Category, English

Study
The one-way ANOVAs comparing the three groups on the five NP types in the
Kind-reference category found that the groups differed significantly in their
ratings of definite singular NPs: native speakers rated them significantly higher
than each of the learner groups, which did not differ from each other. The three
groups also differed marginally on definite plurals, due to the Spanish group
rating definite plurals marginally higher than the native English group; the BrP
group did not differ from either of the other groups. The three groups did not
differ on their ratings of indefinite singulars, bare singulars, or bare plurals.
The repeated-measures ANOVAs showed significant effects of NP type
within each group. For the sake of readability, we do not report on the results
of each pairwise comparison between each pair of NP types in each group, but
instead report the overall patterns. First, all three groups rated bare plurals
above all other NP types; however, the difference in the ratings of bare plu-
rals and definite plurals was significant for the native speaker group, but only
marginally significant for the two learner groups. Second, the native-speaker
group rated definite singulars significantly higher than the three non-target
NP types (indefinite singulars, bare singulars, and definite plurals), although
still significantly lower than bare plurals. On the other hand, both learner
groups rated definite singulars no differently from definite plurals. Third, all
three groups rated indefinite singulars and bare singulars very low, significantly
below all other NP types (the only case where bare singulars were not rated
below all other NP types was as follows: BrP speakers rated bare singulars no
differently from definite singulars, and marginally higher than indefinite sin-
gulars).
3.3.2 Statistical Comparisons on the Generic Category, English Study

Generic category found that the groups differed marginally on their ratings
of indefinite singulars and bare singulars: both learner groups rated indefinite
singulars marginally lower than native speakers did, and the BrP group alone
rated bare singulars marginally higher than native speakers did. The three
groups did not differ from one another in their ratings of definite singulars,
definite plurals, or bare plurals.
within each group. The patterns of results were highly similar across the three
groups. All three groups rated bare plurals and indefinite singulars significantly
higher than the three non-target NP types (while native speakers rated bare
plurals significantly above indefinite singulars, the two learner groups made no
such distinction). The two learner groups gave equally low ratings to the three
non-target NP types (definite singulars, bare singulars, and definite plurals),
while the native speaker group rated bare singulars significantly lower than
the other non-target NP types. Otherwise, performance in the three groups was
nearly identical.
384 ionin et al.
3.3.3 Summary: English Study Results

To sum up, overall both learner groups were quite target-like and similar to
native speakers in their performance. The only clear difference between the
two learner groups and the native group was the relatively low ratings of defi-
nite singulars by the learner groups in the Kind-reference category. This effect
is not explained by L1-transfer, since definite singulars have kind-reference in
Spanish and BrP as well as in English. There were no noticeable L1-transfer
effects: while the two learner groups did rate definite plurals with Kind-
reference higher than native speakers, this difference was only marginal, and
was not replicated in the Generic category. The two learner groups had clearly
acquired the fact that bare plurals, and not definite plurals, are used for generic/
kind readings in English. Transfer effects from bare singulars in BrP were also
very slight, resulting in marginally elevated ratings in the Kind-reference cate-
gory alone.7
3.4 Results: BrP Study

The results for the Kind-reference and Generic categories in the BrP study are
presented in Figures 3 and 4, respectively. The striking result in both figures
is that the two learner groups rated bare plurals well below definite plurals,
in both categories, while native speakers rated the two NP types the same.
The contrast between bare and definite plurals was particularly large for the
Spanish group, but also present for the English group. The second noticeable
difference is that bare singulars were rated very low by both learner groups
in both categories, including the Generic category, where bare singulars were
rated relatively high by native speakers.
We conducted a mixed ANOVA on the results for each category, with NP type
(5 levels) as the within-subjects variable, and participant group (3 levels) as
the between-subjects variable (see the previous section for an explanation of
mixed ANOVAs). In both categories, NP type and group had significant effects
7 Given the small size of our sample, we grouped all learners together regardless of proficiency.
In order to examine whether proficiency had any effect on performance, we computed partial
correlations between the proficiency test score and the mean rating on each category/NP-
type combination, controlling for the learners native language (native English speakers were
excluded from this analysis). Significant positive correlations were found between the pro-
ficiency score and the ratings of bare plurals in the Kind-reference category (r=.32, p<.05)
as well as the ratings of definite singulars in the Kind-reference category (r=.50, p<.01): as
proficiency went up, learners became more target-like in these conditions. Additionally, a
marginal inverse correlation was found between proficiency and the ratings of definite plu-
rals in the Generic category (r=-.26, p=.10).
figure 3 Results for the Kind-reference category, BrP study (mean ratings)
on performance, and interacted with each other (p<.05): the patterns of ratings
across the five NP types were different across the groups. In order to explore this
interaction, we conducted five one-way ANOVAs comparing the three groups
on each NP type, as well as three repeated-measures ANOVAs examining the
effect of NP type within each group (see the previous section for an explanation
of these follow-up comparisons). We report the results for each category in
turn.
3.4.1 Statistical Comparisons on the Kind-Reference Category, BrP Study

Kind-reference category found that the groups differed significantly in their
ratings of bare plurals: native speakers rated bare plurals significantly higher
than did each of the learner groups, and English-speaking learners of BrP rated
bare plurals significantly higher than did Spanish-speaking learners. The three
groups did not differ from one another on any of the other four NP types.
386 ionin et al.
figure 4 Results for the Generic category, BrP study (mean ratings)
The repeated-measures ANOVAs showed significant effects of NP type within

each group. The patterns of results were as follows. First, both the native BrP
group and the English group rated definite plurals no differently from bare
plurals, while the Spanish group rated definite plurals significantly above bare
plurals. Second, all three groups rated definite singulars as high as definite plu-
rals (for the native BrP group, definite plurals were rated marginally higher than
definite singulars). Third, the native BrP group rated bare singulars significantly
above indefinite singulars, but significantly below the other three NP types; in
contrast, the two learner groups rated bare singulars as low as indefinite singu-
lars. The Spanish group, furthermore, rated bare plurals as low as bare singulars.
3.4.2 Statistical Comparisons on the Generic Category, BrP Study

Generic category found that the groups differed significantly on their ratings of
bare singulars and bare plurals: the native BrP group rated both bare singulars
and bare plurals significantly higher than either of the two learner groups.
Additionally, the English group rated bare plurals significantly higher than did
the Spanish group; the two learner groups did not differ on bare singulars. The
three groups did not differ on definite singulars, indefinite singulars, or definite
plurals.
within each group. The patterns of results were as follows (because of the small
sample sizes, numerically large differences did not always come out as sig-
nificant). First, all three groups rated definite plurals and indefinite singulars
equally high. Second, the native BrP group and the English group rated bare
plurals as high as definite plurals, while the Spanish group rated definite plu-
rals marginally higher than bare plurals. Third, the native BrP group rated bare
singulars as high as indefinite singulars, and marginally above definite singu-
lars; in contrast, both learner groups rated bare singulars significantly below
indefinite singulars and no differently from definite singulars. Additionally, the
two learner groups rated bare plurals no differently from bare singulars and def-
inite singulars, while the native BrP group rated bare plurals significantly above
these two NP types.
3.4.3 Summary: BrP Study Results

To sum up, the major difference among the groups had to do with the ratings
of bare plurals, where there was a clear three-way distinction, in both Kind-
reference and Generic categories. The native BrP group rated bare plurals high,
on a par with definite plurals; the English group rated bare plurals numerically
(though not statistically) lower than definite plurals; and the Spanish group
rated bare plurals significantly below definite plurals. The three groups differed
from one another in their ratings of bare plurals, but not in their ratings of def-
inite plurals. The difference between the English and Spanish groups is likely
to be due to L1-transfer since English, unlike Spanish, allows bare plurals with
generic/kind readings. However, L1-transfer cannot explain why English speak-
ers still rated definite plurals above bare plurals in BrP: transfer from English
should have led to the opposite result.
The second major finding has to do with the ratings of bare singulars: in the
Generic category, bare singulars were rated relatively high by native speakers,
but very low by both learner groups. This difference again is consistent with L1-
transfer, since bare singulars are ungrammatical in both English and Spanish.8
8 In the BrP study, as in the English study, we examined the effects of proficiency by computing
partial correlations between the proficiency score and the mean rating for each condition,
388 ionin et al.
3.5 Discussion
Our original hypothesis was that learners would transfer the expression of
genericity from their L1 to their L2. Based on this transfer hypothesis, we
predicted target-like performance on definite singular and indefinite singu-
lar generics, combined with non-target-like performance on plural generics.
Specifically, we expected over-acceptance of English definite plural generics by
both Spanish and BrP speakers, as well as over-rejection of English bare plural
generics by Spanish speakers. In BrP, where both bare plural and definite plu-
ral generics are acceptable, we expected English speakers to rate bare plurals
higher than definite plurals, and Spanish speakers to do the opposite. Finally,
we predicted over-acceptance of English bare singulars by BrP speakers, and
over-rejection of BrP bare singulars by English and Spanish speakers. However,
these predictions were only partially confirmed, as discussed below.
Overall, we found the learners in both the English and the BrP studies to be
quite target-like on most of the conditions tested. The learners were sensitive
to the WDK restriction on definite singulars, as evidenced by low ratings of
definite singulars in the Generic category, where this restriction was violated;
the learners also correctly allowed indefinite singulars in generic sentences,
but disallowed them with kind-reference. Both of these findings can be traced
to L1-transfer, since English, Spanish and BrP all have the WDK restriction on
definite singular kind terms, as well as use indefinite singular NPs in generic
sentences.
Not traceable to L1-transfer, on the other hand, is learners success with plu-
ral generics. First, both Spanish-speaking and BrP-speaking learners of English
were successful at accepting bare plurals and rejecting definite plurals in
generic and kind-reference contexts, even though Spanish and BrP differ from
English with regard to plural generics. Second, English-speaking and Spanish-
speaking learners of BrP were equally successful at accepting definite plurals
in generic and kind-reference contexts, even though English, unlike Spanish,
uses bare plurals in such environments. We conclude that positive evidence
is sufficient to override L1-transfer in this domain, and to lead learners to
controlling for the learners native language (native BrP speakers were excluded from this
analysis). Given the small size of the sample, no correlations were found to be significant, but
there were marginal positive correlations between the proficiency score and the mean ratings
of bare singulars in the Generic category (r=.41, p=.053), as well as the ratings of indefinite
singulars in the Generic category (r=.35, p=.11). Interestingly, ratings of bare plurals (which
were quite low for the learner groups in comparison to the native group, as noted above) did
not improve with proficiency.
acquire the target semantics (namely, that the Down operation is not lexical-
ized on the definite article in English, but is lexicalized in BrP). We note that
similar success has been found for English-speaking learners of Spanish, who
also correctly assigned generic/kind readings to definite plurals (Ionin et al.
2013).
At the same time, learners were not target-like across the board. Clear dif-
ferences between learners and natives were found in three distinct places: (1)
in the English study, both learner groups rated definite singulars in the Kind-
reference category lower than native speakers did; (2) in the BrP study, bare
plurals were rated lower by the English group than by the native BrP group,
and lowest of all by the Spanish group; and (3) in the BrP study, bare singulars
in the Generic category were rated lower by the two learner groups than by the
native BrP group. We consider each of these findings in turn.
3.5.1 Low Ratings of Definite Singular Kind Terms in Learners English

The finding that definite singulars in the Kind-reference category were rated
low by both Spanish-speaking and BrP-speaking learners of English is quite
puzzling. As discussed earlier, both Spanish and BrP use definite singulars with
kind-reference. In fact, BrP speakers tested in their native language gave def-
inite singulars with Kind-reference the mean rating of 3.47 (Figure 3), while
BrP-speaking learners of English rated the same condition at only 2.81 (Fig-
ure 1). Similarly, as reported in Ionin et al. (2011a), Spanish speakers tested on
the Spanish version of the AJT rated definite singulars with Kind-reference at
3.77; in contrast, Spanish-speaking learners of English rated the same condition
at only 2.79 (Figure 1). Thus, the learners do not appear to be transferring the
availability of kind readings of definite singular NPs (in theoretical terms, the
ability of the definite article to combine with a taxonomic NP) from their L1
to their L2. It is possible that the learners are exhibiting hypercorrection: hav-
ing learned that definite plurals are disallowed with generic/kind readings in
English (unlike in Spanish or BrP), they overgeneralize and disallow definite
singular kind terms as well.
As reported in Ionin et al. (2011b), L2-English learners from article-less L1s,
Russian and Korean (who were of comparable English proficiency to the learn-
ers in the present study, and tested on the same English AJT as the learners in
the present study) exhibited a complete lack of knowledge that definite sin-
gular NPs are allowed with Kind-reference (while being fairly target-like on
other NP types). The learners in Ionin et al. (2011b) were less target-like than
those in the present study: they rated definite singulars very low both in the
Kind-reference category (where the WDK restriction holds) and in the Generic
category (where this restriction is violated), whereas the learners in the present
390 ionin et al.
study did make a distinction between the two (compare the ratings of definite
singulars in Figures 1 and 2). Thus, having articles, and definite singular kind
terms, in the native language (Spanish or BrP) does seem to confer an advan-
tage over having no articles and no definite singular kind terms in the native
language (Russian or Korean). However, English definite singular kind terms
appear to present difficulty even for learners coming from a language with arti-
cles.
In Ionin et al. (2011b), it was argued, based on the findings of Biber et al.
(1999) and Yoo (2009), that taxonomic definites are particularly rare in English
input and introduced late in English instruction, and that this is largely respon-
sible for learners difficulties. Interestingly, learners do not appear to gener-
alize from canonical definites to taxonomic definites: learning that definites
can denote unique/maximal individuals does not lead learners to generalize
that definites can also denote unique/maximal taxonomic entities.9 This is
inconsistent with Dayals (2004) view that taxonomic definites are derived by
the same mechanism as canonical definites: the same definite article (which
encodes the Iota operator) combining with a taxonomic NP rather than a com-
mon NP. It is possible Dayal is not entirely correct, and taxonomic definites
are derived by a different mechanism than canonical definites (cf. Ojeda 1991,
Chierchia 1998), which learners have not yet acquired. Alternatively, if taxo-
nomic and canonical definites are derived by the same mechanism, then we
need to conclude that learners do not have a full grasp of what definiteness is:
they learn based on the input, acquiring only those types of definites to which
they have been directly exposed, and not generalizing across definite environ-
ments. The relatively low frequency of definite singular kind terms means that
even learners from native languages which have definite singular kind terms
(Spanish and BrP) do not assume English to have them as well. Recall that
we found a strong and significant correlation between proficiency and rat-
ings of English definite singular kind terms (footnote 7). This suggests that
either more proficient learners have received more input in English (and hence
more input containing definite singular kind terms); or else that they have a
9 Evidence that the learners did acquire the basic use of definite singulars as denoting unique
individuals comes from learners target performance in a control category of anaphoric
definiteness (see Ionin et al. 2011a,b for discussion of this category). The Spanish-speaking
and BrP-speaking learners of English in the present study were fully target-like at accepting
definite singulars with anaphoric readings, giving them ratings of 3.73 and 3.84, respectively
(compared to 3.90 for native speakers), much higher than the ratings for definite singulars in
the Kind-reference category.
better grasp of the concept of definiteness, and (even without necessarily

getting much input with definite kind terms) have generalized from canonical
definites to taxonomic definites.
Note that English-speaking and Spanish-speaking learners of BrP did not
face the same problem with definite singular kind terms in BrP (see Figure 3).
We also have data (not reported here) on English-speaking learners of Spanish,
who rated Spanish definite singular kind terms nearly as high as native speak-
ers (ratings of 3.50 vs. 3.77). Thus, the problem of definite singular kind terms
appears to be exclusively a problem for learners of English. It is possible that
definite singular kind terms are less common in English input than in Spanish
and BrP input; however, this explanation requires corpus data to support it. An
alternative explanation is that learners of Spanish or BrP, having learned that
the target language uses definite plurals for generic/kind readings, simply gen-
eralize to definite singulars: they assume that any definite NP can be used for
generic/kind readings. However, we do not believe this can be the entire expla-
nation: as shown in Figures 3 vs. 4, learners of BrP accept definite singulars only
in reference to well-defined kinds (Figure 3) and not with non-well-defined
kinds (Figure 4), while definite plurals are accepted in both contexts. Simply
generalizing from definite plurals to definite singulars would not lead to this
result: the learners are clearly distinguishing between definite plurals (which
denote kinds of any type) and definite singulars (which denote only taxonomic
entities).
We leave the problem of English definite singular kind terms, and the ques-
tion of why Spanish and BrP definite singular kind terms do not present the
same problem, as issues requiring further investigation.
3.5.2 Low Ratings of Bare Plurals in Learners BrP

We now consider the fact that bare plurals were rated lower by learners of BrP
than by native speakers. English speakers were more target-like than Spanish
speakers, indicating a facilitating effect of L1-transfer for the English-speaking
learners of BrP. Nevertheless, even English speakers rated definite plurals in
BrP higher than bare plurals, the opposite of what transfer from English would
predict.
A possible explanation for the preference for definite plurals over bare plu-
rals in the BrP of English speakers is transfer from Spanish: as discussed above,
12 of the 14 English-speaking learners of BrP had learned Spanish as their sec-
ond language (and one more had learned French), so that BrP was really their
third language. It is possible that, faced with the task of learning BrP, learn-
ers draw on the other Romance language that they know; since Spanish (and
French) use definite plurals rather than bare plurals for generic/kind read-
392 ionin et al.
ings, the learners conclude that BrP must do the same.10 Note that nine of the
ten Spanish-speaking learners of BrP in the present study knew English, yet
they did not appear to be drawing on their knowledge of English in making
judgments about BrP, as evidenced by their very low ratings of bare plurals
with generic/kind readings. This suggests that learners of BrP who know both
English and Spanish draw primarily on Spanish, and not on English, in mak-
ing judgments about plural NP interpretation in BrPregardless of whether
Spanish is their L1 or their L2. This is consistent with the findings of Montrul
et al. (2009) as well as Montrul et al. (2011) in very different domains of BrP
(null and overt subjects in Montrul et al. 2009, object expression and clitics in
Montrul et al. 2011). Montrul et al. (2009, 2011) found that learners of BrP as a
third language who knew both Spanish and English transferred from Spanish
and not from English, regardless of which was their L1 and whichtheir L2.
Montrul et al. (2011) argue for a role of perceived structural similarity: learners
of BrP perceive BrP as being more like Spanish than like English, and transfer
accordingly (cf. Rothman 2011 for a similar claim). The same explanation can
potentially apply to the findings of the present study (see Ionin et al. 2011c for
more discussion of this possibility).
An alternative explanation is that a preference for definite plural gener-
ics over bare plural generics is a general feature of the acquisition of BrP,
not related to transfer from Spanish. Once againas with definite singular
kind terms in Englishinput would appear to play a role. Bare plurals in BrP
are quite formal, constrained largely to written, academic registers (cf. Mller
2002b); in contrast, definite plurals are quite natural in everyday speech as well
as in writing. If learners of BrP are exposed to definite plurals more than bare
plurals, this would naturally lead to the higher ratings of definite plurals. In
order to explore this question further, it is necessary both to consider corpus
data, and to test learners on a greater variety of environments where bare plu-
rals are allowed (such as existential contexts), in order to determine whether
learners have any knowledge of BrP bare plurals.
10 This explanation presupposes that the learners know that Spanish uses definite plurals rather
than bare plurals for generic/kind readings. Indirect evidence for this comes from Ionin
et al.s (2013) study of the acquisition of Spanish, which found English-speaking learners
of Spanish to be very accurate at accepting definite plurals rather than bare plurals with
generic/kind readings. However, more direct evidence would come from the same group
of English speakers being tested both on Spanish (their second language) and BrP (their
third language), and a comparison of their performance in the two languages. Such testing
is currently ongoing.
3.5.3 Low Ratings of Bare Singulars in Learners BrP

Finally, we consider the fact that learners rated bare singulars in the Generic
category lower than native speakers did. This is a difficult finding to evaluate,
since even native BrP speakers rated bare singulars lower than other target NP
types (bare plurals, indefinite singulars, definite plurals). The formal, written
nature of the AJT may have artificially lowered ratings of bare singulars, which
are used more in casual speech (the opposite of bare plurals); furthermore,
given that our study investigated a very limited set of syntactic environments, it
is not clear whether our findings with bare singulars would generalize to other
contexts (e.g., kind and generic readings in object position). In order to more
fully explore what learners know about BrP bare singulars, it is first necessary to
establish contexts where bare singulars are fully acceptable for native speakers.
Based on the current results, it is impossible to say whether learners consider
bare singulars in BrP to be entirely ungrammatical, or whether they would
allow them in a different syntactic configuration, and/or in spoken rather than
written register. We are currently preparing a study of bare singulars in BrP that
will address these issues.
4 Conclusion
The study reported here indicates that L1-transfer plays a limited role in the
expression of genericity for learners of English and BrP, and that L1-transfer is
overridden by considerations of frequency in the input and/or register. Both
definite singular kind terms in English, and bare plurals in BrP, are formal and
not overly frequent expressions, and both caused difficulty for learners regard-
less of their L1; bare singulars, which are more common in oral, casual registers,
also presented difficulty. In contrast, those NP types that are used freely (and
frequently) across registers, such as bare plurals in English, definite plurals in
BrP, and indefinite singulars in both languages, were successfully acquired.
The present findings suggest several avenues for future research. First, as
noted above, it is important to determine whether learners of BrP who know
both English and Spanish are transferring from one or both languages; to this
end, we are now testing learners both in their third language (BrP) and their
second language (English vs. Spanish). It is also important to test English-
speaking BrP learners who have not studied Spanish (or any other Romance
language), in order to determine whether the high ratings of definite plurals
are due to transfer from Spanish.
Second, we are planning to test a greater variety of syntactic configurations
for both generic and existential readings of bare singulars and bare plurals
394 ionin et al.
in BrP; the goal is to determine whether learners simply reject all bare NPs
in BrP as ungrammatical, or whether they are sensitive to differences among
contexts. Finally, in order to make the discussion of input effects on acquisition
less speculative, it is necessary to collect more information about both input
frequency and register effects for bare NPs in BrP. We hope that the present
study is the first step towards a more in-depth investigation of nominal and
article semantics in the acquisition of BrP as a second or third language.
References
Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan.
1999. Longman Grammar of Spoken and Written English. London: Longman.
Carlson, Gregory. 1977. Reference to Kinds in English. Ph.D. diss., University of Mas-
sachusetts at Amherst.
Carlson, Gregory, Rachel Sussman, Natalie Klein, and Michael Tanenhaus. 2006. Weak
Definite Noun Phrases. In Proceedings of NELS 36, edited by Christopher Davis,
Amy Rose Deal, and Youri Zabbal, 179196. Amherst, MA: GLSA, University of Mas-
sachusetts.
Chierchia, Gennaro. 1998. Reference to Kinds across Languages. Natural Language
Semantics 6:339405.
Dayal, Veneeta. 2004. Number Marking and (In)definiteness in Kind Terms. Linguistics
and Philosophy 27:393450.
Dayal, Veneeta. 2011. Bare Noun Phrases. In Semantics: An International Handbook of
Natural Language Meaning, vol. 2, edited by Claudia Maienborn, Klaus von Heu-
singer, and Paul Portner, 1087108. Berlin: De Gruyter.
Dechert, Hans-Wilehlm, and Manfred Raupach, eds. 1989. Transfer in Language Produc-
tion. Norwood, NJ: Ablex.
Dobrovie-Sorin, Carmen, and Roberta Pires de Oliveira. 2008. Reference to Kinds in
Brazilian Portuguese: Definite Singulars vs. Bare Singulars. In Proceedings of SuB12,
edited by Atle Grnn, 107121. Oslo: ILOS.
Ellis, Nick. 2006. Selective Attention and Transfer Phenomena in L2 Acquisition: Con-
tingency, Cue Competition, Salience, Interference, Overshadowing, Blocking, and
Perceptual Learning. Applied Linguistics 27:164194.
Gabriele, Alison. 2009. Transfer and Transition in the SLA of Aspect. Studies in Second
Language Acquisition 31:371402.
Gass, Susan and Larry Selinker (eds.). 1992. Language Transfer in Language Learning.
Amsterdam: John Benjamins.
Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. diss.,
University of Massachusetts at Amherst.
Huebner, Thom. 1983. A Longitudinal Analysis of the Acquisition Of English. Ann Arbor:
Karoma.
Ionin, Tania, Heejeong Ko, and Kenneth Wexler. 2004. Article Semantics in L2-Acquisi-
tion: The Role of Specificity. Language Acquisition 12:369.
Ionin, Tania, Silvina Montrul, and Hlade Santos. 2011a. An Experimental Investigation
of the Expression of Genericity in English, Spanish and Brazilian Portuguese. Lingua
121:963985.
Ionin, Tania, Silvina Montrul, Ji-Hye Kim, and Vadim Philippov. 2011b. Genericity
Distinctions and the Interpretation of Determiners In L2 Acquisition. Language
Acquisition 18:242280.
Ionin, Tania, Silvina Montrul, and Hlade Santos. 2011c. Transfer in L2 and L3 Acquisi-
tion of Generic Interpretation. In Proceedings of the 35th Annual Boston University
Conference on Language Development, edited by Nick Danis, Kate Mesh, and Hyun-
suk Sung, 283295. Somerville, MA: Cascadilla Press.
Ionin, Tania, Silvina Montrul, and Mnica Crivos. 2013. A Bidirectional Study on the
Acquisition of Plural Noun Phrase Interpretation in English and Spanish. Applied
Psycholinguistics 34:483516.
Manfred Krifka, Francis Pelletier, Gregory Carlson, Alice ter Meulen, Gennaro Chier-
chia, and Godehard Link. 1995. Genericity: an Introduction. In The Generic Book,
edited by Gregory Carlson, and Francis Pelletier, 1125. Chicago: University of Chi-
cago Press.
Marsden, Heather. 2004. Quantifier Scope in Non-Native Japanese: A Comparative
Interlanguage Study of Chinese, English, and Korean-Speaking Learners. Ph.D. diss.,
University of Durham.
Montrul, Silvina, Rejane Prince, and Ana Thom-Williams. 2009. Subject Expression
in the L2 acquisition of Brazilian Portuguese. In Minimalist Inquiries into Child and
Adult Language Acquisition: Case Studies across Portuguese, edited by Acrisio Pires,
and Jason Rothman, 301325. Amsterdam: John Benjamins.
Montrul, Silvina, Rejane Dias and Hlade Santos. 2011. Clitics and Object Expression in
the L3 Acquisition of Brazilian Portuguese: Structural Similarity Matters for Trans-
fer. Second Language Research 27:2158.
Mller, Ana. 2002a. The Semantics of Generic Quantification in Brazilian Portuguese.
Probus 14:279298.
Mller, Ana. 2002b. Genericity and the Denotation of Common Nouns in Brazilian
Portuguese. Delta 18:287308.
Mller, Ana. 2003. Generic Sentences with Indefinite ad Bare Subjects in Brazilian Por-
tuguese. In The Proceedings of SULA 2, edited by Jan Anderssen, Paula Menndez-
Benito, and Adam Werle, 7186. Amherst, MA: GLSA, University of Massachusetts.
Munn, Alan, and Cristina Schmitt. 2005. Number and Indefinites. Lingua 115:821
855.
396 ionin et al.
Odlin, Terence. 1989. Language Transfer: Cross-Linguistic Influence in Language Learn-

ing. Cambridge, UK: Cambridge University Press.
Ojeda, Almerindo. 1991. Definite Descriptions and Definite Generics. Linguistics and
Philosophy 14:367398.
Oliveira, Roberta Pires de, Josa Coelho da Silva, and Mariana Rublescki Silveira Bres-
sane. 2010. O Singular Nu Denota Espcie: Uma Investigao Emprica (Bare Sin-
gulars are Kind Denoting Expressions: An Empirical Investigation). Delta 26:115
139.
Rothman, Jason. 2011. L3 Syntactic Transfer Selectivity and Typological Determinacy:
The Typological Primacy Model. Second Language Research 27:107127.
Schmitt, Cristina, and Alan Munn. 1999. Against the Nominal Mapping Parameter:
Bare Nouns in Brazilian Portuguese. In Proceedings of NELS 29, edited by Nancy
Hall, Masako Hirotani, and Pius Tamanji. Amherst, MA: University of Massachusetts
Occasional Papers in Linguistics.
Schmitt, Cristina, and Alan Munn. 2002. The Syntax and Semantics of Bare Arguments
in Brazilian Portuguese. Linguistic Variation Yearbook 2:185216.
Schwartz, Bonnie. 1998. The second language instinct. Lingua 106:133160.
Schwartz, Bonnie, and Rex Sprouse. 1994. Word Order and Nominative Case in Non-
Native Language Acquisition: A Longitudinal Study of (L1 Turkish) German Interlan-
guage. In Language Acquisition Studies in Generative Grammar: Papers in Honor of
Kenneth Wexler from the 1991 GLOW Workshops, edited by Teun Hoekstra and Bonnie
Schwartz, 317368. Amsterdam: John Benjamins.
Schwartz, Bonnie, and Rex Sprouse. 1996. L2 Cognitive States and the Full Transfer/
Full Access Model. Second Language Research 12:4072.
Slabakova, Roumyana. 2006. Learnability in the Second Language Acquisition of
Semantics: A Bidirectional Study of a Semantic Parameter. Second Language
Research 22:498523.
Slabakova, Roumyana. 2008. Meaning in the Second Language. Berlin: Mouton de Gruy-
ter.
Stvan, Laurel. 1998. The Semantics and Pragmatics of Bare Singular Noun Phrases.
Ph.D. diss., Northwestern University.
Thomas, Margaret. 1989. The Acquisition of English Articles by First- and Second-
Language Learners. Applied Psycholinguistics 10:335355.
Trenkic, Danijela. 2008. The Representation of English Articles in Second Language
Grammars: Determiners or Adjectives? Bilingualism: Language and Cognition 11:1
18.
Vergnaud, Jean-Roger, and Maria Luisa Zubizarreta. 1992. The Definite Determiner and
the Inalienable Constructions in French and English. Linguistic Inquiry 23:595652.
Wilkinson, Karina. 1991. Studies in the Semantics of Generic NPs. Ph.D. diss., Univer-
sity of Massachusetts at Amherst.
Yoo, Isaiah WonHo. 2009. The English Definite Article: What ESL/EFL Grammars
Say and What Corpus Findings Show. Journal of English for Academic Purposes
8:267278.
Zamparelli, Roberto. 2002. Definite and Bare Kind-Denoting Noun Phrases. In
Romance languages and Linguistic Theory 2000, edited by Claire Beyssade, Reineke
Bok-Bennema, Frank Drijkoningen, and Paola. Monachesi, 305343. Amsterdam:
John Benjamins.
Index Keywords
adjectives 5859, 107, 125127, 134137, 176, determiners

204, 213215 full/ reduced definite determiners
anaphoricity 59, 6566, 107, 117, 146150, 172207
154167, 174175, 202, 239265 weak/ strong definite determiners (see
appositives 165167, 216218, 222, 224, 226, also full/ reduced definite determiners)
230 143166
articles (see determiners)
ellipsis 118137, 217, 261267, 288289
bare existentials 23, 4448
bare noun 57, 6067, 70, 7398, 231, externalization system 48
234, 250, 272273, 275276, 282285, ezafe 6264
291296, 299300, 306, 308311, 345,
350, 358, 367394 gesture / gestural 319, 323, 325, 355
bare NP 120, 122, 258 genericity 340341, 343, 367393
bare object 118119
bare plural 3132, 369394 Information Structure (IS) 2348, 181183
bare singular 6067, 75, 91, 227, 231, 284, canonical IS 23, 3537, 40, 44, 48
288, 369394 intensive reflexives 5459
classifier joint and split readings under coordination

noun classifier 110, 120122, 130 227232
numeral classifier 253257, 264
classifier handshapes 320331, 341 kind-reference 371375, 378382, 384385,
342 388390
collective
collective nouns 9498 L1-L2-Transfer 368, 375377, 380, 384,
collective numerals 6869 387389, 391393
collective predeterminer 296297, left-branch extraction 106107
300304, 306307, 310311 lexematic unit 329, 244248, 357
count nouns 7398, 120, 188
countability 7398 mass-count distinction 7398
mass nouns (see mass-count distinction)
definiteness fake mass nouns 74, 81, 9294, 9798
definiteness concord/agreement 224 modality (gestural vs. vocal) 317, 320, 325,
226 328, 347
definiteness effect 45, 23, 4548
semantic vs. pragmatic definiteness 146, number 5270, 7576, 8788, 109110,
271272, 294295, 308, 311312 257258, 276277, 310
polydefinites 212235 number neutral 12, 5268, 7576, 276,
demonstratives 65, 76, 78, 111113, 123, 375
128130, 132 137, 144146, 150154, 156158, numerals 6770, 76, 7879, 113, 118122,
160167, 174, 239265, 274, 296 230232

400 index keywords
pluralia tantum 53, 6869, 8384 signing space 317319, 323, 334, 336337,
polydefinites, see definites 342, 346347, 351, 354
possessor 63, 106, 110134, 137, 243, 253254, locus 318320, 331332, 335338, 342,
258259, 264 348, 352, 354
Predicate Projection (PredP) 127130 sign languages 316357
proper names 8, 1315, 34, 149, 154, 175, singulative nouns 9498
180, 189, 212, 218, 221222, 227, 232235, specificity 5860, 192193, 277279,
269271, 279, 288, 290291, 294295, 305, 285289, 297298, 303306, 310, 316, 334,
306, 308, 310311 336341, 343, 345346, 348, 354, 357
pro-drop 105, 117118, 132, 137 specific indefinites 23, 2831, 3839, 43,
prosody 3940, 113, 183 113
pseudopartitives 224225
topic
restrictive 24, 27, 158160, 165167, 172207, continued topic 2425, 43, 47
215219, 260261 shifted topic 24, 3738, 47
relative clauses stage topic 25, 4047
restrictive RCs 158160, 16567, 172207 topic drop 2425
non-restrictive RCs 172, 179180, 194, topicalization 35, 2425, 3040, 43, 4748
207 transfer structures
appositive RCs 158, 165, 167, 179, 182, 186, transfer unit 327, 331, 348, 351353, 357
189191, 194196, 201, 203 transfer of size and shape 328329, 333,
role shift 323324, 329, 331, 333, 341 351
situational transfer 327329, 333, 341,
second language acquisition (L2-acquisition) 353
262263, 367393 personal transfer 316, 328331, 348353
semiological model 325329, 332333,
338356 universality of DP 12, 5270, 102137
Index Languages
Austro-Bavarian (West Germanic) 166, 172, Martinique creole (Martinik) 269315

176, 181, 185, 207 Modern Hebrew 24, 3536, 8298
Armenian 5461, 6869 Modern Greek 212239
Bavarian (West Germanic) 172, 176, Norwegian 3435, 39, 6466

206207
Brazilian Portuguese 367376 Russian 24, 3839, 5458, 6769, 106,
389390
Catalan 24, 3637, 47
Spanish 5, 10, 47, 64, 367393
Danish 3142 Standard German (West Germanic) 4043,
108, 172, 177, 187, 199200
Fehring Frisian (West Germanic) 172174, Swiss German dialects (West Germanic)
178, 192, 198, 200201, 207 143171
French Sign Language (Langue des signes
franaise, LSF) 316357 Tatar 5557, 6164
Turkish 102139
Japanese 239268
Vorarlberg dialect (West Germanic) 176, 188,
Karitiana 7485, 99 196199, 205

(Syntax and Semantics) Patricia Cabredo Hofherr, Anne Zribi-Hertz (Eds.) - Crosslinguistic Studies On Noun Phrase Structure and Reference-Brill (2013)

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

(Syntax and Semantics) Patricia Cabredo Hofherr, Anne Zribi-Hertz (Eds.) - Crosslinguistic Studies On Noun Phrase Structure and Reference-Brill (2013)

Uploaded by

Copyright:

Available Formats

Crosslinguistic Studies on Noun Phrase Structure and Reference

Syntax & Semantics

Jeffrey T. Runner (University of Rochester)

The titles published in this series are listed at brill.com/sas

Patricia Cabredo Hofherr

Copyright 2014 by Koninklijke Brill nv, Leiden, The Netherlands.

This book is printed on acid-free paper.

Information Structure and (In)definiteness 23

The Morphology, Syntax and Semantics of Definite Determiners in Swiss

From Noun to Name: On Definiteness Marking in Modern

When Articles Have Different Meanings: Acquiring the Expression of

We gratefully acknowledge the support of the project Le calcul de la rfrence

Patricia Cabredo Hofherr

Hlade Scutti Santos

ment of Turkish Language and Literature at Yeditepe University, Istanbul. His

1 Issues in the Analysis of Definiteness

Determiners and definiteness have been extensively studied in formal linguis-

koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_002

(i) To what extent are (in)definite determiners necessary for argument-

1.1 Nominal Reference and Definiteness

back to a familiar individualfamiliar to the speaker and hearer, since previ-

This dichotomy between previously introduced discourse referents for def-

Specificity has been linked to an existence presupposition: while the specific

(3) a. He is looking for a (certain) secretary.

(4) a. Beoseu -ga o-goiss -da.1 (Korean)

b. Beoseu -neun o -goiss-da.

(5) a. Topic-familiarity condition (Gundel 1988)

(6) Utterances are conceived of as a set of instructions by a speaker to a

entered on the card. Common ground information is thus ordered accord-

Conversely, existential constructions are generally associated with new dis-

(7) a. There was a child in the garden.

(8) a. Hay {algunos / dos / muchos / pocos / } perros. (Spanish)

b. *Hay {l / el perro / ese perro / Fido}.

As Erteschik-Shir (this volume) shows, apparent counterexamples to the cor-

Accessibility (cf. Ariel 1990) restricts the availability of discourse referents as

1.2 (In)definite Articles and Argumenthood

proposed in particular that NPs introduced by the indefinite determiner a(n)

1.3 Semantic Subtypes of Definite Noun Phrases

(10) a. A kning kaam tu bischk. (Fehring Frisian)

b. Matje hee al wler an ni bridj. Di gast kn

This distinction can be overtly signalled by morphologically full definite deter-

Upper Sorbian; Zribi-Hertz and Jean-Louis, this volume on Martinik). In some

(11) Scale of uniqueness (Lbner 2011:320): pragmatic >> semantic defi-

(12) Det / At iast buk, wat hi skrewen hee, docht

Differences in the distribution of definite determiners can also be observed

(13) Les pandas sont vgtariens. (French)

1.4 Definite Determiners and Definiteness

(13) a. He took the train to come here, and so did Mary.

(14) Marie est lhpital, et Jean aussi.

(15) a. i asimenia i pena vs. a. i asimenia pena (Modern Greek)

b. la nia ms viva vs. b. la fille la plus intelligente

Well-known examples of missing determiners are provided by mass nouns

(16) a. He bought oil for the car

(17) a. Oil is expensive.

1.5 Articles and (In)definiteness from a Crosslinguistic Perspective

Section 3 shows how topicalization is constrained differently in Danish, Nor-

inite articles in West-Germanic languages (Austro-Bavarian German, Fehring

of the WH-prefix do-; (ii) the Japanese demonstratives may be preceded by a

Garcia, Brigitte, and Marie-Anne Sallandre. This volume. Reference resolution in

Poesio, Massimo. 1994. Weak definites. In Proceedings of SALT 4, edited by Mandy

Noun Phrase Syntax and Interpretation:

koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_003

(1) Dani hevi xalav me-ha-super ve-sam ba-mekarer