Crosslinguistic Studies on Noun Phrase Structure and Reference 900426082X, 9789004260825

Citation preview

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 Biographies viii

vii

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 Müller The Turkish NP 102 Željko Bošković and Serkan Şener

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 Patricia Cabredo Hofherr When Determiners Abound: Implications for the Encoding of Definiteness 212 Marika Lekakou and Kriszta Szendrői The Semantics and Syntax of Japanese Adnominal Demonstratives 239 Makoto Kaneko

vi

contents

From Noun to Name: On Definiteness Marking in Modern Martinikè 269 Anne Zribi-Hertz and Loïc 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 Hélade Santos Index

399

Acknowledgements We gratefully acknowledge the support of the project “Le calcul de la référence nominale dans les langues avec et sans articles”, sponsored by the Fédération 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 Université Paris-8, whose support enabled us to organise a series of workshops on the construal of nominal reference from a cross-linguistic perspective.

Biographies Željko Bošković (Ph.D. University of Connecticut, 1995) is Professor of Linguistics at the University of Connecticut. He is the author of The Syntax of Nonfinite Complementation: An Economy Approach (MIT Press, 1997), On the Nature of the SyntaxPhonology 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. Patricia Cabredo Hofherr 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 Linguistics at Ben Gurion University, Israel. Her publications include The Dynamics of Focus Structure (1997), Information Structure: The Syntax-Discourse Interface (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 extending 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, specialising in French Sign Language (LSF) linguistics. She is director of the research

biographies

ix

group Langues des signes et gestualité (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 Language, the development of a graphic representation for French Sign Language and the relationship between writing and deafness. She is particularly interested in the processes involved in the emergence of signs in French Sign Language. Elaine Grolla is an Assistant Professor at the University of São 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 pronouns and anaphors, the comprehension and production of relative clauses, wh-questions, and passive structures. Besides her work on first language acquisition, 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 UrbanaChampaign. 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 without 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 journals as Second Language Research, Bilingualism: Language and Cognition, Linguistic Inquiry, Natural Language Semantics, Language Acquisition and Language Learning. She is an associate editor of the journal Language Acquisition. Loic Jean-Louis is presently an MA student in linguistics at Université Paris-8. Working out the grammar of Martinikè (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 nominal exclamative clauses, free choice items, epistemic indefinite expressions in Japanese and in French. His current research focuses on the syntax and semantics 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 morphosyntactic 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 Acquisition 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 Language 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 retention in minority language-speaking bilinguals, or heritage speakers. Ana Müller is Assistant Professor of Linguistics at the University of São 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 research 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 University. Her specialty is in the syntax of Slavic languages, particularly Russian, and in cross-linguistic variation and typology worldwide. She is particularly interested 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 linguistics 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 nonconventional 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 different types of iconicity in sign languages and the typology of sign languages, comparing different sign languages as well as signed and vocal languages. Hélade Scutti Santos graduated in Spanish and Portuguese from the University of São Paulo, Brazil, in 2000. In 2005, she received an M.A. degree in Spanish from the same institution, where she conducted research on language attitudes/representations. From 2004 to 2007 she was an Assistant Professor of Spanish at the Universidade Federal de São Carlos, Brazil, where she conducted research on language attitudes/representations. She is currently a PhD candidate in Hispanic Linguistics at the University of Illinois at Urbana-Champaign. Her dissertation research focuses on the acquisition of Portuguese as a third language by learners who already know English and Spanish. Her work contributes to a better understanding of the acquisition of additional languages beyond the second language. Serkan Şener completed his PhD in the Department of Linguistics at the University of Connecticut 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 Szendrői 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 children, the acquisition of focus and intonation in autism. Anne Zribi-Hertz is Professor of Linguistics at Université Paris-8, and a member of the UMR “Structures Formelles du Langage” (a research centre supported by Université Paris-8 and the French CNRS). She has written two books (Découvrir la grammaire française, with L. Picabia, 1981; L’anaphore et les pronoms: une introduction à la syntaxe générative, 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 Français, to appear shortly.

Introduction Patricia Cabredo Hofherr and Anne Zribi-Hertz The papers collected in this volume explore the factors determining the referential interpretation of noun phrases across a wide array of typologically unrelated languages. The languages discussed include Armenian, Brazilian Portuguese, Catalan, Danish, French Sign Language (LSF), several West-Germanic languages, Modern Greek, Japanese, Karitiana, Martinique creole (Martinikè), Modern Hebrew, Norwegian, Russian, Spanish, Tatar and Turkish. The individual 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 linguistics, 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 perspective:

© 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 argumenthood? (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 referent 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., Löbner 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 maximality 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 descriptive 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 discourse referents and previously introduced referents (Kamp 1981, Heim 1982). In (2), for example, the indefinite noun phrases a man, a woman, a hat introduce new discourse referents, while the definite noun phrase the man refers

introduction

3

back to a familiar individual—familiar to the speaker and hearer, since previously 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 definites, 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, Enç 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 Kuroda’s (1965, 1979, a.o.) and Kuno’s (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) necessarily points to a uniquely identified referent pre-activated by the immediate discourse context:

4

cabredo hofherr and zribi-hertz

(4) a. Beoseu -ga o-goiss -da.1 bus -subj come-prog-dec (i) ‘There’s {a/the} bus coming.’ (ii) ‘It’s {a/the} bus that’s coming.’

(Korean)

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 assumed in works on Topicality—cf. Gundel’s (1988) Topic Familiarity Condition and Lambrecht’s (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 Information Structure, which leads her to assume that every utterance must contain a Topic—covert if not overt—instantiating 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.

5

introduction

entered on the card. Common ground information is thus ordered according to the ‘topics’ defined by each discourse referent. erteschik-shir 1997:17

Conversely, existential constructions are generally associated with new discourse 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. have some / two / many / few / ∅ dogs ‘There exist some / two / many / few dogs.’

(Spanish)

b. *Hay {él / 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 correlation 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 activation 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):

6

cabredo hofherr and zribi-hertz

(9) Givenness Hierarchy (Gundel, Hedberg and Zacharski. 1993 and references 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 hypothesis 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. Chierchia 1998, Cheng and Sybesma 1999, Coene and D’ Hulst eds, 2003, DobrovieSorin and Laca 2003, Bošković 2008, a.o.). Chierchia (1998) proposes that lexical 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 theory, 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). Bošković (2008, 2012) argues that argumental noun phrases are not universally DPs, proposes a list of systematic syntactic discrepancies between (determiner-less) NP-languages and DP-languages (cf. Bošković and Şener, 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 Abney’s (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: 33–35). A link between number and “definiteness” is suggested in some languages where plural marking may only occur in NPs otherwise marked as definite—a 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 subclasses of definite Noun Phrases observed cross-linguistically, for many languages turn out to have two paradigms of definite determiners (cf. Ebert’s 1970, 1971 seminal study of two definite articles in a variety of Frisian). In such languages with two definite articles, morphologically reduced definite determiners 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 köning kaam tu bischük. detred king came to visit ‘The king came for a visit.’ (Ebert 1971: 83, ex 30)

(Fehring Frisian)

b. Matje hee al wäler an näi bridj. Di gast kön M. has again a new bride detfull young-man can a nöös uk wel äi fol fu. detred nose also prt not full get ‘Matje has a new girl friend again. The young man can’t get enough, it seems.’ (Ebert 1971: 108, ex 12) This distinction can be overtly signalled by morphologically full definite determiners for “pragmatic (viz. discourse-linked) definiteness” (Löbner 1985), and by morphologically reduced definite determiners for “semantic definiteness” (ibid.). Similar oppositions can also be instantiated in languages that have a lexical determiner specialised in pragmatic definiteness, and bare nouns ambiguous between indefinite and semantically-definite readings (cf. Breu 2004 on

8

cabredo hofherr and zribi-hertz

Upper Sorbian; Zribi-Hertz and Jean-Louis, this volume on Martinikè). 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 Löbner (1985, 2011) semantic and pragmatic definiteness may be regarded as the two end-points of the Definiteness Scale in (11): (11) Scale of uniqueness (Löbner 2011:320): pragmatic >> semantic definiteness2 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 car—the 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 intermediate 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 33ʹ) 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

(13) Les pandas sont végétariens. def pandas are vegetarians.

9 (French)

The plural definite determiners in English and French are therefore not equivalent (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 distribution 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 definite 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 determiner. 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 Löbner (2011, pp. 280–282).

10

cabredo hofherr and zribi-hertz

(14) Marie est à l’hôpital, 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 determiner 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 (15bʹ) (Modern French): (15) a. i asimenia i pena vs. aʹ. i asimenia pena (Modern Greek) def silver def pen def silver pen ‘the silver pen’ (adapted from ex 2 Lekakou and Szendrői this volume) b. la niña más viva vs. bʹ. 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 Szendrői (this volume) argue that the distribution of the definite determiner in (15a/aʹ) corresponds to a syntactic difference in Modern Greek. The contrast between Spanish and French in (15b/bʹ) 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 & Müller this volume). 1.5 Articles and (In)definiteness from a Crosslinguistic Perspective In a nutshell, the conceptual issues which underly today’s 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 interact 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 grammar? Are there mechanisms of reference construal common to all natural grammars? Does the vocal or visual nature of the signifier have a crucial incidence 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 nonoccurrence 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 syntactic structure of noun phrases in Languages Without Articles. Nomi Erteschik-Shir (“Information Structure and (In)definiteness”) discusses the two most prominent examples of the interaction between Definiteness and Information Structure (abbreviated IS in what follows): Topicalization 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.

12

cabredo hofherr and zribi-hertz

Section 3 shows 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. Although the initial 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 existential clauses and gives some evidence that here again differences in canonical IS account for syntactic and morphological variation across languages. Asya Pereltsvaig (“Number and Numberlessness in Languages With and Without Articles”) is concerned with the representation of number in articleless languages, focusing on two distantly related languages—Russian 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 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 neither singular (‘one X’) nor plural (‘more than one X’). Pereltsvaig argues that the semantic number-neutrality of such nominals results from the lack the syntactic number feature, normally hosted in NumP, which she shows to be absent from such nominals. Depending on the language, such number-neutral nominals 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 Müller (“The cognitive basis for the Mass/Count distinction: 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 identification of stable units (in the sense of Chierchia 2010, who regards the cognitive 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 Müller 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 crucially involves unstable units.

introduction

13

Željko Bošković and Serkan Şener (“The Turkish NP”) argue for an analysis of the Noun Phrase in Turkish—a Language Without Articles—involving no Determiner Phrase above the NP node, an assumption in line with Bošković’s general theory of Noun Phrase structure developed in his previous works (cf. Bošković 2008), and which runs against the “DP Hypothesis” as developed by, e.g., Longobardi (2000). Empirical evidence in support of Bošković’s NP Hypothesis is provided for Turkish by the order of constituents within the Noun Phrase, and by some interesting constraints on interpretation. Bošković and Şener 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 constructions allow internal ellipsis within the Noun Phrase, with the elements located within the Classifier Phrase, hence outside of NP, surviving ellipsis. Bošković and Şener finally explore the possibility of a functional projection in predicate 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 markers—in five typologically different languages or language groups: West Germanic, Greek, Japanese, Martinikè 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 markers. In one language (LSF), the relevance of semantic “Definiteness” for linguistic description is overtly questioned. Rebekka Studler (“The morphology, syntax and semantics of definite determiners in Swiss German”) scrutinizes the three possible translations of English the in Swiss German: a strong article, a weak article, and a proximal demonstrative, 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 anaphorically 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 relative clauses”) further discusses the competition between full and reduced def-

14

cabredo hofherr and zribi-hertz

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 combine with reduced definite determiners in their Noun Phrase. In other WestGermanic 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 studies are not of the same type. She argues that while all the languages under consideration select the full definite determiner with contrastive restrictive relatives, 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 Szendrői (“When determiners abound: implications for the encoding of definiteness”) investigate the grammar of “definiteness” in Greek. Their point of departure is the so-called polydefinite construction, 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 parallel 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 definiteness 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 demonstratives”) 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 demonstrative 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 identified as adnominal adjuncts. Kaneko argues that these hypotheses further shed light on some data from L2 acquisition. Anne Zribi-Hertz and Loïc Jean-Louis (“From Noun to Name: on definiteness marking in Modern Martinikè”) 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 Löbner’s (1985) sense, viz. as crucially identified by means of anchoring to the discourse context or utterance situation. Two other morphemes, l(a)- and lé(-), are shown to form semantically definite DPs (in Löbner’s sense) which unambiguously identify individual terms independently of the discourse context and utterance situation: l(a)- is a word-level prefix, while lé(-) occurs either as a free morpheme (with common nouns) or as a prefix (with some country names). The distributional and semantic properties of l(a)- and lé(-) DPs make them similar to definite proper names. The authors argue that l(a)- and lé(-) form a subtype of definite DPs they call Names, characterised 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 semiological 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 corpus 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.

16

cabredo hofherr and zribi-hertz

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 Hélade Santos (“When articles 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 differ 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 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.

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 Proceedings of SALT 20, edited by Nan Li and David Lutz, 179–196. Ithaca, NY: Cornell University. Ariel, Mira. 1990. Accessing Noun-Phrase antecedents. London: Routledge. Bošković, Željko. 2008. “What will you have, DP or NP?” Proceedings of NELS 37, edited by Emily Elfner and Martin Walkow, 101–114. University of Illinois/Urbana-Champaign. Bošković, Željko. 2012. “On NPs and clauses.” In Discourse and grammar: From sentence types to lexical categories, edited by Günther Grewendorf and Thomas Ede Zimmermann, 179–245. Berlin: de Gruyter. Bošković, Željko, and Serkan Şener. This volume. “The Turkish NP”, 102–140. 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, 9–57. München: Otto Sagner. Cabredo Hofherr, Patricia. This volume. “Reduced definite articles with restrictive relative clauses”, 172–211.

introduction

17

Carlson, Greg. 1977. “A unified analysis of the English bare plural.” Linguistics and Philosophy 1:413–457. 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 30:509–542. Chierchia, Gennaro. 1998. “Reference to kinds across languages.” Natural Language Semantics 6:339–405. Chierchia, Gennaro. 2010. “Mass nouns, vagueness and semantic variation.” Synthese 174:99–149. Coene, Martine, and Yves D’Huls, 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. Indéfini, défini et démonstratif. Constructions linguistiques de la référence. Geneva: Droz. Corblin, Francis. 2001. “Défini et génitif: le cas des génitifs défectifs.” In Cahiers JeanClaude Milner, edited by Jean-Marie Marandin, 19–54. Paris: Verdier. Corblin, Francis. 2011. “Des définis para-intensionnels: être à l’hôpital, aller à l’école.” Langue française 171: 55–75. 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 déterminant dans les langues romanes.” In Les langues romanes: problèmes de la phrase simple, edited by Danièle Godard, 235–279. Paris: CNRS. Doetjes, Jenny. 1997. Quantifiers and selection, The Hague: HIL. Doron, Edit, and Ana Müller. This volume. “The cognitive basis of the mass-count distinction: evidence from bare nouns”, 73–101. Ebert, Karen H. 1970. “Zwei Formen des bestimmten Artikels.” In Probleme und Fortschritte der Transformationsgrammatik, edited by Dieter Wunderlich, 159–173. Tübingen: Max Hueber Verlag. Ebert, Karen H. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem nordfriesischen Dialekt (Fehring). Bräist/ Bredstedt: Nordfriisk Institut, 1971. Enç, Mürvet. 1991. “The semantics of specificity.” Linguistic Inquiry 22:1–26. Erteschik-Shir, Nomi. 1997. The dynamics of focus structure. Oxford: Oxford University Press. Erteschik-Shir, Nomi. This volume. “Information Structure and (in)definiteness”, 23–51. Fodor, Janet, and Ivan Sag. 1982. “Referential and Quantificational Indefinites.” Linguistics and Philosophy 5:355–398.

18

cabredo hofherr and zribi-hertz

Garcia, Brigitte, and Marie-Anne Sallandre. This volume. “Reference resolution in French Sign Language: the effects of the visuo-gestual modality”, 316–364. 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, 175–196. Oxford: Blackwell. Gundel, Jeannette, Nancy Hedberg and Ron Zacharski. 1993. “Cognitive status and the form of referring expressions in discourse.” Language 69:274–307. Hawkins, John. 1978. Definiteness and Indefiniteness: A study in reference and grammaticality 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 16:547–593. von Heusinger, Klaus. 2002. “Specificity and definiteness in sentence and discourse structure.” Journal of Semantics 19:245–274. 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”, 367–397. 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.), 277–322. Amsterdam Center. Amsterdam. Kaneko, Makoto. This volume. “The semantics and syntax of Japanese adnominal demonstratives”, 239–268. Keenan, Edward. 2003. “The Definiteness Effect: Semantics or Pragmatics?” Natural Language Semantics 11: 187–216, 2003. Kihm, Alain. 2003. “Qu’y a-t-il dans un nom? Genre, classes nominales, et nominalité.” In Typologie des langues d’Afrique et universaux de la grammaire, vol. I, edited by Patrick Sauzet and Anne Zribi-Hertz, 39–64. Paris: L’Harmattan. 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, 75–85. 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 Szendrői. This volume. “When determiners abound: implications for the encoding of definiteness”, 212–238. Leonetti, Manuel, 2008. “Definiteness effects and the role of the coda in existential constructions.” In Essays on Nominal Determination, edited by Henrik Høeg Muller and Alex Klinge, 131–162. Amsterdam: John Benjamins. Lewis, David. 1979. “Scorekeeping in a language game.” Journal of Philosophical Logic 8:339–359. Link, Godehard. 1983. The logical analysis of plurals and mass terms: A lattice theoretical approach. In Meaning, use, and the interpretation of language, edited by Reiner Bäuerle, Christoph Schwarze, and Arnim von Stechow, 302–323. Berlin: de Gruyter. Löbner, Sebastian. 1985. “Definites.” Journal of Semantics 4:279–326. Löbner, Sebastian. 2011. “Concept types and determination.” Journal of Semantics 28: 279–333. Longobardi, Giuseppe. 1994. “Reference and proper names: a theory of N movement in syntax and Logical Form.” Linguistic Inquiry 25:609–665. 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, 562–603. 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, 1–38. 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, 215–242. Amsterdam: John Benjamins. Pereltsvaig, Asya. 2007. “On the Universality of DP: A View from Russian.” Studia Linguistica 61: 59–94. Pereltsvaig, Asya. 2013. “Noun Phrase Structure in Article-less Slavic languages: DP or not DP?” Language and Linguistics Compass 7:201–219. Pereltsvaig, Asya. This volume. “On number and numberlessness in languages with and without articles”, 52–72.

20

cabredo hofherr and zribi-hertz

Poesio, Massimo. 1994. “Weak definites.” In Proceedings of SALT 4, edited by Mandy Harvey and Lynn Santelmann, 282–299. 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, 37–62. New York: Academic Press. Walker, Marilyn A., Aravind K. Joshi, and Ellen F. Prince, eds. 1998. Centering in Discourse. Oxford: Oxford University Press. Zribi-Hertz, Anne, and Loïc Jean-Louis. This volume. “From Noun to Name: on definiteness marking in Modern Martinikè”, 269–315.

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 idiosyncratic exceptions to this association. 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, in terms of their effect on a file system representing the discourse manipulation of referents in the common ground. It is shown how subordinate ISs afford an explanation of the fact that specific indefinites 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 determined 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, 86–101.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_003

24

erteschik-shir

may have more than one topic, the “main” topic (often the syntactically highest 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 sentence 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 constituent 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 several 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 topics, whereas the latter applies to continued topics. In some languages, however, topics selected from restrictive or contrastive sets are distinguished from continued topics. (Erteschik-Shir, Ibnbari, and Taube 2013) argue that Topicalization 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 Hinterhölzl 2007 distinguish Familiar Topics, Aboutness-shift Topics and Contrastive 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 intransitive individual-level predicates such as (5) is the subject. Intransitive individual-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. b. c. d. e.

#A little boy is intelligent. Dogs/a dog are intelligent. A student I know is intelligent. A DOG is intelligent, a CAT is not. TWO/SOME (of the) students are intelligent.

(only generic) (specific) (contrastive) (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 organized 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 topics in the discourse. In order to get to the top of the stack, the card (the referent 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.

information structure and (in)definiteness

27

The file system thus involves locating cards on top of a stack (topics) or positioning them there (foci). Additionally, each card is updated with the information 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. Participants in a discourse update their common ground according to the rules of IS outlined here. Topics and focus in this framework do not project syntactic structure à 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 syntactic 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 necessitates 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 sentences 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 sentence. The distinction between singular indefinites which do not provide valid topics and the specific, contrastive and partitive indefinites in (7c–e), 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 therefore 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, 41–42. Cases of unmodified specific 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., 61–67.

information structure and (in)definiteness

29

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 specific 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 specificity 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 compatible 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, 61–67. 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, 182–183 for an analysis of German data from Krifka 1994 including the scopal properties of similar topics. 9 E.g., Birner and Ward 2009, 1172.

information structure and (in)definiteness

31

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 elements, specific indefinites but not non-specific indefinites: (18) a. Hende mødte jeg i går. her met I yesterday ‘I met her yesterday.’ b. Pigen mødte jeg i går. girl-the met I yesterday ‘I met the girl yesterday.’ c. Blomster ser man om foråret. flowers sees one in spring-the ‘Flowers, one sees in the spring.’ d. Pigen mødte jeg i går, drengen mødte jeg først 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 mødte i går 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 mødte jeg i går a girl met I yesterday ‘I met a girl yesterday.’ g. *Hospitalet tog Peter på 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 topicalized whereas nonspecific, singular indefinites (18f) cannot. (18g) finally demonstrates that weak definites do not tropicalize.12 A surprising constraint on Danish topicalization is that Danish does not tolerate 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 interpretation, 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 mødte 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 mødte, 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 mødte Sara igår. 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.

information structure and (in)definiteness

33

b. ?Peter mødte jeg igår. Peter met I yesterday ‘I met Peter yesterday.’ c. Ham mødte hun igår. Him met she yesterday ‘She met him yesterday.’ These data show that topicalization in Danish is not only restricted to constituents 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 element 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 topicalization 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 disambiguating 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 igår. That drawer painted boy-the yesterday b. ?Peter forskrækkede 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 syntax and is most naturally construed as a parsing constraint on the identification of arguments at the interface with the articulatory-perceptual system, PF. I propose 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 informationstructural roles (topic, focus) which lays a heavy burden on the parser in a language such as Danish which lacks both agreement and case-marking morphology. 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 syntactically similar Norwegian and Danish, according to Raviv, is due to the morphological 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 så 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 readings. It follows that the two languages are distinct with respect to their tolerance 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 Norwegian were subject to the parsing constraint in (23), topicalization would be very restricted—an 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 unmarked 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 following interchange shows that subjects are indeed unmarked topics in Danish: (25) A. Fortæl mig om Peter/Hvad med Peter? Tell me about Peter/How about Peter?

information structure and (in)definiteness

35

B. Han er forelsket i Marie. He is in love with Marie. Bʹ. ??Marie er forelsket i ham. Marie is in love with him Bʺ. I ham er Marie forelsket. With him is Marie in love What is wrong with Bʹ 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 Bʺ 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, noncontrastive) 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 postverbally. (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 topicalized elements must be given, yet only “new” topics can be fronted. Vallduví 1992 refers to these as ‘Links’. Old topics are detached to the right (Vallduví’s ‘Tails’). The following examples are from Vallduví 1992:

15 In both languages topicalization is optional. Hebrew topicalization is in most cases contrastive. The two word orders (27a) and (27b) differ in that the subject is in focus in (27b), but not in (27a).

information structure and (in)definiteness

37

(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 wasn’t counting on anymore.’ b. Quant al Joan i la Isidora no t’ho sé 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 can’t 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 Vallduví, there is a difference. In Catalan only elements that are not topics in the previous sentence are topicalized. Such “new” topics are called shifted topics. Vallduví calls these fronted elements “Links” and views them as “address pointers” in a file system 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 languages. (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 don’t 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) Devušku Petja vstretil včera. girl.acc Peter met yesterday ‘Peter met the girl yesterday.’ (not: ‘Peter met a girl yesterday’) (34) Čvety my vidim tol’ko vesnoj. Flowers we see only in-the-spring ‘We see flowers only in the spring.’ (35) Devočke kotoruju ja vstretil včera ja dala xorošuju 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 doč’ Mother.nom/acc loves daughter.nom/acc ‘Mother loves (her) daughter.’

16 See Erteschik-Shir and Strahov 2004.

information structure and (in)definiteness

39

Generally, in Russian, case marking disambiguates the grammatical functions 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 interpretation 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 definiteness.17 This is probably why a specific reading of the topicalized element in (35) is excluded. Here the connection between the lack of definiteness morphology and the role of topicalization is transparent. In languages such as Danish, the role of topicalization is limited to marking IS. Indefinites can therefore topicalize as long as they are specific or contrastive. The properties of topicalization 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 topicalization with pronoun subjects is less pronounced and stops with the transition from Old English to Middle English Grammar. Since the decline affects pronoun subjects and full noun-phrase subjects differently, Speyer figures that

17 Whether or not Russian has DPs is debatable (e.g., Bošković 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 topicalization 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 sentence (37a) is unobtrusive. The English sentence (37b) is awkward. It requires a little break between the two accents. This looks as if—at least in English—a weak element between two accents is compulsory. Speyer calls this requirement 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 languages 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. Bʹ. ??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.

information structure and (in)definiteness

41

3.6 Fronting Non-Topics According to Frey 2006, not all initial elements are topics. Such cases are illustrated in (41). Here, according to Frey, the fronted elements are not referential and require contextualization: (41) a. Fast jeden Kollegen findet der berühmte 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 gefüttert 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 fühlt 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 context, 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 previously 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 Mädchen einen Hund gefüttert. 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 informationstructural 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. Glücklicherweise hat ein Mädchen 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. Desværre kom Hans/han ikke til selskabet. unfortunately came Hans/he not to the party ‘Unfortunately Hans/he didn’t come to the party.’

information structure and (in)definiteness

43

b. Hans/han kom desværre ikke til selskabet. Hans/he came unfortunately not to the party ‘Hans/he unfortunately didn’t 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, scenesetting adverbials and actual switch topics. If no switch topic is available, a continued 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. Svenonius’s description of the elements in initial position can be captured by the following 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 motivation fails. É. 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 É. 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 element which does not itself have information-structural properties, yet signals a particular information structure, namely one with a stage topic. These structures also have another property which may explain their informationstructural 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 therefore be to oust the subject from initial position in order to enable its interpretation as a non-topic, triggering, in this case, an out-of-the-blue interpretation. 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 element 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 question 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 canonical 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 reordering 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 É. 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 o’clock]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 hereand-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 contextual definition: either the place or the time are not contextually available

information structure and (in)definiteness

45

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 contextually 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 existentials.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. There’s city hall, the museum, and the park. b. There’s the meeting at 2 o’clock 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 day’s 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 what’s 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) (ii) (iii)

My souptop [has a/*the fly in it]foc Johntop [has a/*the hat on]foc 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 doesn’t, 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 introduces the inventory on this unpopulated but given stage. No coda is therefore required. In (49b), a coda is (optionally) present. This coda is however packaged differently with respect to IS. Compare (51) and (52): (51) sTopt [There’s [the meeting at 2 o’clock]] foc (52) sTop(t) [There is a meeting at 2 o’clock]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 o’clock’ is the element introduced on

information structure and (in)definiteness

47

stage. In the existential in (52), however, what is introduced on the new stage is ‘a meeting’, the coda ‘at 2 o’clock’ 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 2008’s inspiring paper. Leonetti demonstrates 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 particular 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 topicalization. Leonetti addresses the seeming non-adherence of Italian and Catalan to the definiteness effect illustrated in (53) and (54) respectively (p. 134). (53) C’è un cane. / C’è il cane. / C’è 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’è 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. Leonetti’s 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 idiosyncratic 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):407–444. 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, 19–41. Oxford: Oxford University Press. Birner, Betty J., and Gregory Ward. 2009. “Information Structure and Syntactic Structure.” Language and Linguistics Compass 3 (4):1167–1187. Brunetti, Lisa. 2009. “On links and tails in Italian.” Lingua 119 (5):756–781. doi: 10.1016 /j.lingua.2008.10.005. 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, 179–196. UMass/Amherst: GLSA.

information structure and (in)definiteness

49

Cohen, Ariel, and Nomi Erteschik-Shir. 2002. “Topic, Focus and the Interpretation of Bare Plurals.” Natural Language Semantics 10:125–165. Cresti, Diana. 1995. Indefinite topics. Ph.D. dissertation, Department of Linguistics & Philosophy, University of Massachusetts, Amherst. É. Kiss, Katalin. 2004. “The EPP in a Topic-Prominent Language.” In Subjects, Expletives, and the EPP, edited by Peter Svenonius, 107–124. 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, 119–150. Amsterdam: John Benjamins. Erteschik-Shir, Nomi. 2005b. “What is Syntax?” Theoretical Linguistics 31 (1–2):263–274. doi: doi:10.1515/thli.2005.31.1–2.263. Erteschik-Shir, Nomi. 2006a. “On the architecture of topic and focus.” In The architecture of focus, edited by Valéria Molnár and Susanne Winkler, 33–57. Berlin: Mouton de Gruyter. Erteschik-Shir, Nomi. 2006b. “What’s what?” In Gradience in Grammar, edited by Caroline Féry, Gisbert Fanselow, Matthias Schlesewsky, Ralf Vogel, 317–335. 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 explanation of island phenomena.” Theoretical Linguistics 6:41–85. Erteschik-Shir, Nomi, and Shalom Lappin. 1987. “Dominance and modularity.” Linguistics 25 (4):671–686. 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):301–323. Frascarelli, Mara, and Roland Hinterhölzl. 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 architecture of focus, edited by Valéria Molnár and Susanne Winkler, 235–264. Berlin: Mouton de Gruyter.

50

erteschik-shir

Gundel, Jeanette K. 1974. The role of topic and comment in linguistic theory. Ph.D. dissertation, University of Texas, Austin. Gundel, Jeanette K. 1988. “Universals of topic-comment structure.” In Studies in syntactic typology, edited by Michael Hammond, Edith A. Moravcsik and Jessica R. Wirth, 209–239. Amsterdam / Philadelphia: John Benjamins Publishing Co. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Ph.D. dissertation, University of Massachusetts, Amherst, Amherst. Holmberg, Anders. 2000. “Scandinavian Stylistic Fronting: How Any Category Can Become an Expletive.” Linguistic Inquiry 31 (3):445–483. Jakobson, Roman. 1971. “Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der Russischen Kasus.” In Selected Writings II, 23–71. The Hague: Mouton. Original edition, 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 Quantification. 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, 125–175. 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, 133–152. Heidelberg: IBM Deutscland Informationssysteme, GMBH Scientic Center. Krifka, Manfred. 1998. “Scope Inversion under the Rise-Fall Contour in German.” Linguistic Inquiry 29 (1):75–112. 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 Høeg Müller, 131–162. Amsterdam: John Benjamins. Milsark, Gary, L. 1974. Existential sentences in English. Ph.D. dissertation, M.I.T., Cambridge. Portner, Paul, and Katsuhiko Yabushita. 2001. “Specific Indefinites and the Information Structure Theory of Topics.” Journal of Semantics 18 (3):271–297. doi: 10.1093/jos /18.3.271. Rando, Emily, and Donna Jo Napoli. 1978. “Definites in ‘There’-sentences.” Language 54 (1):300–313. 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.” Philosophica no. 27:53–94. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery.” In Elements of Grammar, edited by Liliane Haegeman, 281–337. Dordrecht: Kluwer.

information structure and (in)definiteness

51

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 1:75–116. 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):243–256 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, 201–242. Oxford: Oxford University Press. Vallduví, Enric. 1992. The informational component. New York: Garland. von Fintel, Kai. 1994. Restrictions on Quantifier Domains. Ph.D. dissertation, Department of Linguistics, University of Massachusetts, Amherst. von Heusinger, Klaus. 2002. “Specificity and Definiteness in Sentence and Discourse Structure.” Journal of Semantics 19 (3):245–274. Ward, Gregory, and Ellen F. Prince. 1991. “On the topicalization of indefinite NPs.” Journal of Pragmatics 15 (8):167–178.

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 languages, focusing on two distantly related languages—Russian and Armenian— and an unrelated language, Tatar. It is argued that morphological number and semantic number are mediated by syntactic number, encoded even in articleless 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 (cf. Fukui 1986, 1988; Chierchia 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 Lyutikova’s 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, Željko Boškovič, 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_004

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 individuals. 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 number-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 plural individuals, but under normal circumstances the atomic interpretation is ruled out by pragmatic principles; however, in certain semantically or pragmatically 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 numberneutral 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 someone’s marital status.

54

pereltsvaig

counterparts in Russian or Armenian. In those languages, one also finds number-neutral nominals in contexts similar to those in the English examples:2 (2) a. U vas est’ deti? to you there-is children ‘Do you have (one or more) children?’ b. Bezdig unis? child have.2sg ‘Do you have (one or more) children?’ (cf. Bale et al. 2010)3

(Ru)

(W Ar)

Two things are noteworthy here. First, while in Russian number-neutral nominals are morphologically plural, much like in English (cf. (1a) above), in Armenian number-neutral nominals are morphologically singular (unmarked). Second, 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 nominals 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 constructions: 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,

on number and numberlessness

55

(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 doesn’t want to eat (one or more) cutlets anymore.’ (4) Medvedeva vybrali v prezidenty. Medvedev they-elected into presidents ‘Medvedev has been elected president.’

(Ru)

(5) klonirovanie životnyx / remont avtomobilej cloning animals / repairing cars ‘cloning of animals/an animal’; ‘repair of cars/a car’

(Ru)

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 aŕ-el. Ara-nom book aux.3sg.pr buy-perf ‘Ara has bought (one or more) books.’

(E. Ar.)

b. Dursek a pòłoc’ 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 kızıl alma aša-dı. Marat red apple eat-pst ‘Marat ate (one or more) red apple(s).’

(Ta)

Note that number-neutral nominals in Armenian and Tatar are uniformly morphologically singular (or rather unmarked). In fact, morphologically plural nominals in Armenian and Tatar can denote only plural individuals and cannot 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? child-indef.pl have.2sg ‘Do you have (two or more) children?’

(W. Ar.)

In the next section, we show that number-neutral plurals in Russian and number-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 Armenian—despite differing in morphological number—share 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 Russian and Armenian cannot be defined in purely semantic or pragmatic terms, as mentioned above, such nominals are clearly syntactically selected. For example, 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: typically, 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.

on number and numberlessness

57

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. Lena na-ate-sja from morning cutlets.gen ‘Lena ate her fill of cutlets first thing in the morning.’

(Ru)

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 (10a–c), or complements, as in (11). (Additional examples from Tatar are found in section 4 below.) (10) a. Lena na-jela-s’ rybnyx kotlet. Lena na-ate-sja fish cutlets.gen ‘Lena ate her fill of fish cakes.’

(Ru)

b. Lena na-jela-s’ kotlet po-požarski. Lena na-ate-sja cutlets.gen Pozharsky-style ‘Lena ate her fill of cakes Pozharsky-style.’ c. Lena na-jela-s’ kotlet s podžaristoj koročkoj. Lena na-ate-sja cutlets.gen with fried-up crust ‘Lena ate her fill of cakes with a fried-up crust.’ (11) Gorbačëva vybrali v prezidenty SSSR. Gorbachev they-elected into presidents USSR.gen ‘Gorbachev was elected president of the USSR.’

(Ru)

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 syntactic profile of a Small Nominal, more specifically of a bare NP. More specifically, number-neutral nominals lack the room for expressions of specificity. For instance, neither the complements of intensive reflexives nor the complements of v ‘into’ in Russian allow such modifiers as opredelënnye ‘certain, specific’, leading to the ungrammaticality of the following examples (other

58

pereltsvaig

adjectives, e.g. v počëtnye prezidenty lit. ‘into honorary presidents’, are grammatical): (12) a. *Ja najelas’ opredelënnyx kotlet. I na-ate-sja specific cutlets

(Ru)

b. *On izbiraetsja v opredelënnye prezidenty. He being-elected into specific presidents Similarly, in (Eastern) Armenian number-neutrality obtains only with nonspecific morphologically singular objects. If a specificity marker is added, the nominal is no longer number-neutral (example from Megedoomian 2011): (13) Ara-n girkh-ə aŕ-el a. Ara-nom book-sp buy-perf aux.3sg.pr ‘Ara has bought {the / a specific} book.’ (not ‘books’!)

(E. Ar.)

Furthermore, number-neutral nominals have no room for expressions of quantity. For example, a numeral or a measure noun cannot be added to such nominals in Russian, leading to the ungrammaticality of examples in (14) and (15); similarly, in Armenian a singular object with an overt numeral is not numberneutral (example in (16) is from Megerdoomian 2011). (14) a. *Ja najelas’ pjati kotlet. I na-ate-sja five cutlets

(Ru)

b. *Ja napilas’ stakana vody. I na-drank-sja glass water (15) *Obama i Medvedev byli izbrany v dva prezidenta. Obama & Medvedev were elected into two presidents ‘Obama & Medvedev were elected into presidents.’ (16) yerex-erk-ə me hat muk en č’ar-el. child-pl-nom one cl mouse aux.3pl.pr find-perf ‘The children have found a mouse.’ (not ‘mice’!)

(Ru)

(E. Ar.)

Moreover, number-neutral nominals have no room for higher adjectival modifiers, 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).

on number and numberlessness

(17) a. *Ja najelas’ otvratitel’nyx kotlet. I na-ate-sja terrible burgers

59 (Ru)

b. *On rešil ballotirovat’sja 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 movement either for scope or for focus. As shown by Kagan and Pereltsvaig (2011), complements of intensive reflexives can only have surface scope: for example, the sentence in (18) means that Lena has seen an eyeful of black-andwhite 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 context. (18) Lena nasmotrelas’ černo-belyx fil’mov. Lena na-watched-sja [black-and-white movies].gen ‘Lena has watched black-and-white movies to the limit.’

(Ru)

Pereltsvaig (2006) shows that complements of the preposition v ‘into’ are likewise 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 generalov ‘generals’. (19) a. Vsjakij soldat dolžen metit’ v generaly. every soldier.nom must aim into generals ‘Every soldier must aim to become a general.’ [∀soldier > ∃general] b. Vsjakij soldat dolžen 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 aŕ-el. (E. Ar.) all one child book aux.3sg.pr buy-perf ‘Every child has bought a book/books.’ (unambiguous: Karine Megedoomian, 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, numberneutral 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 same—S-O-Aux-V—Megedoomian argues that they have different structure: the vP is bracketed in both sentences below. (21) a. Ara-n [girkh a aŕ-el]. Ara-nom book aux.3sg.pr buy-perf ‘Ara has bought {a book/books}.’

(E. Ar.)

b. Aram-ə girkh-ə a [aŕ-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 immediately preceding it marking the left edge of the vP, or (b) in FocP. These two possibilities are instantiated by the two sentences in (21a–b), 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 sentences 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 aŕ-el]. Ara-nom book aux.3sg.pr fast buy-perf b. Aram-ə girkh-ə a [arag aŕ-el]. Aram-nom book-sp aux.3sg.pr fast buy-perf ‘It is the (specific) book that Aram bought fast.’

(E. Ar.)

on number and numberlessness

61

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 aŕ-el]. Ara-nom fast aux.3sg.pr book buy-perf ‘Ara has bought {a book/books} fast.’

(E. Ar.)

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 numberneutral 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), let’s 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 -nı 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 kızıl alma-lar aša-dı. Marat red apple-pl eat-pst ‘Marat ate red apples.’ (#a red apple) b. Marat kızıl alma-nı aša-dı. Marat red apple-acc eat-pst ‘Marat ate a/the red apple.’ (#red apples)

(Ta)

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 constructions places its limitations on the amount of functional architecture available 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 taš-(*lar) jırt stone-(*pl) house-pst ‘stone house’

(Ta)

b. attributivizer -le: [NP] N altın jezek(*-ler)-le kız gold *ring(*-pl)-atr girl ‘a girl with {a gold ring/gold rings}’ c. ezafe-2: [NumP] N kırsak-lı xatın-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. (*čın) altın jezek (*real) gold ring ‘a gold ring’ (impossible: ‘a ring of true gold’)

(Ta)

on number and numberlessness

63

b. taš(*-lar) jırt 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)): altın 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. kük čäčäk(*-lär)-le čaška blue flower(*-pl)-atr cup ‘a cup with a blue flower/flowers’

(Ta)

b. altın jezek-le kız gold ring-atr girl ‘a girl with a gold ring(s)’ c. *Tukaj-nıŋ räsem-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 number) 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 structure (as in (28b)). However, the complement of ezafe-2 cannot itself be ezafe3, 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 numberneutral. (28) a. kırsak-lı xatın-nar kijem-e belly-atr woman-pl clothing-3 ‘clothing (for) pregnant women’

(Ta)

64

pereltsvaig

b. bala-lar xastaxanä-se tabib-ı child-pl hospital-3 doctor-3 ‘a doctor from a children’s hospital’ (impossible: ‘a doctor from children’s hospitals’) c. *ukučı-nıŋ däftär-lär-e papka-sı pupil-gen notebook-pl-3 folder-3 intended: ‘a folder for a/the pupil’s notebooks’ The various nominal constructions are summarized below: (29) a. b. c. d.

ezafe-1: ATR -le: ezafe-2: ezafe-3:

[N] N [NP] N [NumP] N [DP] N

Crucially, only complements in the constructions in (29a–b), which are structurally 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 numberneutral nominals are found in many languages with articles, including Norwegian (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 år. Per has had dog in ten years ‘Per has had dog(s) for ten years.’ (Borthen 2003)

(No)

b. Eu acho que vi livro espalhado pelo chão. (BrP) I think that saw book spread on floor ‘I think that I saw book(s) spread on the floor.’ (Schmitt & Munn 2002)

on number and numberlessness

65

c. Busco pis. (Cat) look.1sg apartment ‘I’m 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 promoção. bought chair they were on sale ‘I bought (one or more) chairs. They were on sale.’

(BrP)

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 morphosyntactically 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 år. Per has had dog in ten years

(No)

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 vært svært 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 singular 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 ønsker seg kopp med bilde av Mikke Mus. Ola wants refl cup with picture of Mickey Mouse ‘Ola wants a cup with a picture of Mickey Mouse.’

(No)

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 prøvde jakke. all children-defsuff tried jacket ‘All the children tried on some jacket or other.’

(No)

Finally, Norwegian bare singulars are weakly referential in that they license the type-anaphor dét ‘that’ but not the token-anaphor den ‘it’ (cf. Borthen 2003: 39–41). (35) a. Per ønsker seg ny båt, Per wants refl new boat.masc, men dét får han nok aldri. but that-neut gets he probably never ‘Per wants a new boat, but he probably will never get that.’ b. Jeg ønsker meg sykkel til jul. I want refl bike to Christmas. ??Den er blå. ??it is blue ‘I want a bike for Christmas. It is blue.’

(No)

on number and numberlessness

67

To account for this clustering of properties, we assume that the central property 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 referentiality. 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 projection for number, NumP. This argument is fleshed out in the following section.

6

Proposal and Consequences

So far, we have argued that number-neutral nominals—including those in languages with and without articles—are 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 number 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 morphological 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 tantum nouns in Russian, such as nožnicy ‘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 tantum 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 compatible with paucal numerals, such as dva ‘two’, tri ‘three’ and četyre ‘four’. (36) *tri {sutki / sutok } three 24-hour.period.pl.nom / 24-hour.period.pl.gen ‘3 24-hour periods (i.e., 72 hours)’

(Ru)

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 specified 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 pluralia tantum noun itself results in the ungrammaticality of examples like (35). Note that there is no problem combining pluralia tantum nouns with nonpaucal 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 collective numerals (e.g., dvoe ‘two’, troe ‘three’, četvero ‘four’, pjatero ‘five’, etc.). Note

on number and numberlessness

69

that the set of collective numerals includes those that encode the same lower numerosities as paucal numerals: for example, semantically the collective numeral 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 tantum 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 devušek (OK: tri devuški) 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) languages: Russian, Armenian, Tatar. Instead, numberless nominals are found in various syntactic contexts where they are selected by certain lexical or functional 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 adjectival 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, 33–41. Somerville, MA: Cascadilla Proceedings Project. Bailyn, John Frederick. 2002. “Overt Predicators.” Journal of Slavic Linguistics 10:23–52. Bailyn, John Frederick. 2004. “The Case of Q.” In Proceedings of FASL 12, edited by Olga Arnaudova, Wayles Browne, María Luisa Rivero, and Danijela Stojanovic, 1–36. Ann Arbor, MI: University of Michigan Press. Bailyn, John Frederick and Andrew Nevins. 2008. “Russian Genitive Plurals are Impostors.” In Inflectional Identity, edited by Andrew Nevins and Asaf Bachrach, 237–270. Oxford: Oxford University Press. Baker, Mark C. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago, IL: University of Chicago Press. Baker, Mark C. 2003. Lexical Categories. Verbs, Nouns, and Adjectives. Cambridge, UK: Cambridge University Press. Bale, Alan, Michaël Gagnon, and Hrayr Khanjian. 2010. “Cross-linguistic Representations of Numerals and Number Marking.” Proceedings of SALT 20: 1–15. Beauseroy, Delphine and Marie-Laurence Knittel. 2008. Nombre et adjectifs: le cas des noms abstraits (= Number and adjectives: the case of abstracts nouns). Paper presented at the Workshop on nominal and verbal plurality, Paris, CNRS, 7–8 November 2008. Borthen, Kaja. 2003. Norwegian bare singulars. Ph.D. thesis, Trondheim: NTNU. Chierchia, Gennaro. 1998. “Reference to Kinds Across Languages.” Natural Language Semantics 6:339–405. Dobrovie-Sorin, Carmen, Tonia Bleam and M. Teresa Espinal. 2006. “Bare Nouns, Number and Types of Incorporation.” In Non-definiteness and plurality, edited by Svetlana Vogeleer and Liliane Tasmovski, 51–81. 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 47:87–128. Fukui, Naoki. 1986. A Theory of Category Projection and its Aplications. Ph.D. thesis, Massachusetts Institute of Technology.

on number and numberlessness

71

Fukui, Naoki. 1988. “Deriving the differences between English and Japanese: A case study in parametric syntax.” English Linguistics 5:249–270. 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, 226–240. 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, 221–238. 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: 433–500. Pereltsvaig, Asya. 2007. Copular Sentences in Russian. A Theory of Intra-Clausal Relations. Springer. Pereltsvaig, Asya. 2007. “On the Universality of DP: A View from Russian.” Studia Linguistica 61:59–94. Pereltsvaig, Asya. 2008. “Split Phrases in Colloquial Russian.” Studia Linguistica 62:5–38. Pereltsvaig, Asya. 2010. “As easy as two, three, four?” In Formal Approaches to Slavic Linguistics 18, edited by Wayles Browne, Adam Cooper, Alison Fisher, Esra Kesici, Nikola Predolac, and Draga Zec, 417–434. 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 7:201–219. 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 121:2153–2175. Rullmann, Hotze. 2011. “Scopal Deficiency and Number.” Paper presented at the Discussion Workshop on Weak Referentiality, Utrecht, March 2011.

72

pereltsvaig

Rutkowski, Paweł. 2002. “Noun/pronoun asymmetries: evidence in support of the DP hypothesis in Polish.” Jezikoslovlje 3:159–170. Rutkowski, Paweł. 2002. “Numerals as grammaticalised nouns: a generative approach.” Interlingüística 13:317–328. Rutkowski, Paweł. 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, 249–263. Seoul: Hankook/Korean Generative Grammar Circle. Rutkowski, Paweł. 2007. The Determiner Phrase hypothesis as a tool of syntactic analysis of Polish nominal phrases. Ph.D. thesis, Warsaw University. Rutkowski, Paweł. In press. The syntax of floating intensifiers in Polish and its implications 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 Toosarvandani, 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, 258–275. 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, 185–216. 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, 16–42. Oxford: Oxford University Press. Tatevosov, Sergei. 2006. “Measuring individuals, partitioning events: Semantics of cumulative verbs in Russian.” In Linguistic Investigations into Formal Description of Slavic Languages, edited by Peter Kosta and Lilia Schurcks, 529–544. 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 Müller

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 entities—such 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 entities—such 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 discrepancy between the mass-count linguistic contrast and the homogeneousatomic cognitive contrast, the distinction between mass and count nouns emerges in the work of these scholars as partly arbitrary and language specific. 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 Journées d’étude Langues avec et sans articles, 3–4 March 2011. We thank three anonymous referees for their comments, which were very helpful in the reformulation of some of our claims. Doron’s research was supported by the Israel Science Foundation grant #1157/10. Müller’s research was supported by the Conselho Nacional de Pesquisa grant #303407/2009-3, and by FAPESP grant #2011/51408-4.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_005

74

doron and müller

nouns are linguistically accessible after all, since one can for example predicate 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 classification 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 masscount distinction in the structure of nouns or noun phrases, yet semantically distinguishes nouns which can be counted from nouns which cannot. In section 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. Following 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 analysis 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 cognitive 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 existing 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-Arikém language spoken in Rondônia, in the western Brazilian Amazonic region. The language has around 400 speakers, most of them living in a demarcated reservation in Rondônia. 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 Dëne Suliné. 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 Müller 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 Müller during fieldwork. The examples are presented as follows—1st line: orthographic transcription of the Karitiana sentence; 2nd line: morphological segmentation; 3rd line: morpheme by morpheme gloss; 4th line: translation. 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 (1)

doron and müller

a. yn naka’yt pikom yn ∅-naka-’y-t pikom 1s 3-decl-eat-nft monkey ‘I ate (the/a/some) monkey(s).’

(Karitiana)

b. oho atakam’at Ora oho ∅-a-taka-m-’a-t Ora potato 3-inv-decl-caus-make-nft Ora ‘Potatoes, Ora created (them)’

(Karitiana)

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, Karitiana is not a classifier language, since, as the examples in (2) show numerals and common nouns combine directly. (2) a. yn naka’yt myhint pikom yn ∅-naka-’y-t myhin-t pikom 1s 3-decl-eat-nft one-obl monkey ‘I ate one monkey.’

(Karitiana)

b. yn naka’yt 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 nonvariable. On the other hand, 1st and 2nd person plural pronouns do not incorporate any morpheme with a plural meaning. They are formed by the suffixation 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:

77

the cognitive basis of the mass-count distinction

table 1

Personal pronouns in Karitiana (Müller et al 2006)

Pronoun Morphology Person

Meaning

yn na i yjxa yta ajxa i

I+ participant you + participant other (non-participant) I+other(s)+anaphor I+anaphor you+other(s)+anaphor other

y+n a+n i y+i+ta y+ta a+i+ta i

1sg 2sg 3 1pl (inclusive) 1pl (exclusive) 2pl 3

(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 Maria Conga pip Maria Conga postp ‘All jaguars are gone from Maria Conga.’

(Karitiana)

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 müller

Similarly, demonstratives too combine both with mass and count nouns: (5) a. tykat idjiera ahop aka a ese aka tykat i-djera-∅ ahop aka a ese aka impf nmz-cost-nft much cop deic water cop ‘How much does this water cost?’

(Karitiana)

b. dibm nakatari a õwã aka dibm naka-tar-i a õwã 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 contextual 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 i’orot 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 i’orot 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.’

the cognitive basis of the mass-count distinction

79

(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 sentence (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 cardinality presupposes individuation on both sides of the distributive relation—in our case, one boy per event of going to the river or two men per event of arriving. (10) a. myhint.myhint nakahori õwã se pip myhin-t.myhin-t naka-hot-i õwã se pip one-obl.one-obl decl-go.pl-ft child river postp ‘Boys will go to the river one at a time.’

(Karitiana)

b. sypomp.sypomp naotãm taso sypom-t.sypom-t na-otãm-∅ 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, distributive 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 i’orot 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 naotãm ouro sypom-t.sypom-t na-otãm-∅ oro two-obl.two-obl decl-arrive-nft gold ‘Gold arrived two at a time.’

(Karitiana)

80

doron and müller

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 interpreted 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 coercion 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 itipasagngãt João ’ep i-ti-pasag.pasag-t João tree 3-inv-count.rdpl-nft João ‘The trees, João is counting (them).’

(Karitiana)

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 João myhin-t.myhin-t ∅-na-mangat-∅ ’ep João one-obl.one-obl 3-decl-carry-nft tree João ‘João carried trees one by one.’

(Karitiana)

b. myhint namangat kandat ’ep João myhin-t ∅-na-mangat-∅ kandat ’ep João one-obl 3-decl-carry-nft much-obl wood João ‘João carried a lot of wood at once.’ (15) pyryhopiyn mejahygng amby parede sok pyry-hop-iyn mejahygng house parede sok ass-exist-nft rat house wall over ‘There was rat all over the wall.’

(Karitiana)

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 cognitive basis of the mass-count distinction

81

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 distinction 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 3

doron and müller

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 -áyim.6 These are shown in (18a–c) respectively:7 (18) a. Class I nouns (plural suffix -im) ħatul / ħatul-im cat.masc cat-pl b. Class II nouns (plural suffix -ot) tmun-a / tmun-ot picture-fem picture-pl

dbor-a / bee-fem

dbor-im bee- pl

ħalon / ħalon-ot window.masc window-pl

c. Class III nouns (plural suffix -áyim) magaꝑ / magaꝑ-áyim géreb / boot.masc boot-pl sock.fem

garb-áyim 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 -áyim 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, ꝑ, ḵ, marks the fricative allophones of the stops b, k, p, respectively. ħ 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,

the cognitive basis of the mass-count distinction

(19) órez rice.masc

*oraz-im rice-pl

ħacac *ħacac-im gravel.masc gravel-pl

avir air.masc

*avir-im air-pl

boc *boc-im mud.masc mud-pl

méši silk.masc

*meši-im silk-pl

ħem’-a *ħem’-ot butter-fem butter-pl

kutn-a *kutn-ot cotton-fem cotton-pl

ħalud-a rust-fem

83

*ħalud-ot 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 interpretation. 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 ruins

’atiq-ot *’atiq-a antiquities

šimur-im *šimur canned goods

šmar-im *šémer yeast-pl

m-áyim *ma water-pl

šam-áyim *šama sky-pl

qur-ey10 / *qur akabiš web-pl spider spider webs

kl-ey / *kli mita linen-pl bed bed linen

maca’-im maca’ bed-linen-pl bed

Second, there are also mass nouns which have a morphological contrast between singular and plural forms: (21) géšem / rain.masc

gšam-im rain-pl

šéleg / snow.masc

šlag-im 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’, šimur ‘preservation’. 10 -ey is the construct-state form of the plural suffixes -im and -áyim.

84

doron and müller

déše / grass.masc

dša’-im grass-pl

ed / ed-im steam.masc steam-pl

késeꝑ / ksaꝑ-im money.masc money-pl

dam / dam-im blood.masc blood-pl

adam-a / land-fem

adam-ot land-pl

ašp-a / ašp(-at)-ot rubbish-fem rubbish-pl

ħol / sand.masc

ħol-ot sand-pl

ruaħ / wind.fem

ruħ-ot wind-pl

merħab / merħab-im space.masc space-pl Semantically, the plural form of mass terms, when it contrasts with the singular, 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 cardinality modifiers. 3.2 The Cardinality Modifier Criterion Count nouns combine with cardinality modifiers such as one, two, several. (22) a. yéled eħad child.masc one.masc ‘one boy’ b. yald-a aħat child-fem one.fem ‘one girl’

yelad-im aħad-im child-pl one-pl ‘several boys’

šney yelad-im two.masc child-pl ‘two boys’

yelad-ot aħad-ot child-pl one-pl ‘several girls’

štey yelad-ot two.fem child-pl ‘two girls’

Cardinality modifiers do not combine with mass nouns, irrespective of whether these mass nouns are singularia tantum, pluralia tantum, or alternate in plurality. 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.

the cognitive basis of the mass-count distinction

(23) *órez eħad rice one *‘one rice’

85

*dam-im aħad-im blood-pl one-pl *‘several bloods’

*ħalud-a aħat rust-fem one.fem *‘one rust’

*štey ruħ-ot two.fem wind-pl *‘two winds’

*ħol-ot aħad-im sand-pl one-pl *‘several sands’

*šney šmar-im two.masc yeast-pl *‘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’, me’at ‘a little’, ódeꝑ/ yoter miday ‘too much’, kílo ‘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 a lot child-pl ‘many girls’ ódeꝑ botn-im too much peanut-pl ‘too many peanuts’

me’at ħatul-im a little cat-pl ‘few cats’

kílo tapuħ-im kilo apple-pl ‘a kilo apples’

86

doron and müller

b. *harbe a lot

yald-a child-fem

*me’at ħatul a little cat

*kílo tapúaħ kilo apple

*ódeꝑ bóten too much peanut When combining with mass nouns, measure quantifiers allow singular morphology (though plural morphology is also an option for mass nouns which have plural forms, preserving the abundance plural reading): (25) harbe ħol/ ħol-ot a lot sand/sand-pl ‘much sand’

me’at órez a little rice ‘a little rice’

kílo šum kilo garlic ‘a kilo garlic’

ódeꝑ géšem/gšam-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 (ħol ‘sand’, órez ‘rice’, šum ‘garlic’, géšem ‘rain’ etc.) and count nouns ( yald-a ‘girl’, ħatul ‘cat’, tapúaħ ‘apple’, bóten ‘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 crosslinguistically. First, nouns which are usually taken to basically be count nouns and are coerced to mass readings by the “universal grinder”, e.g. kébes ‘lamb’ also interpreted as meat, oꝑ ‘bird’ also interpreted as chicken meat, ec ‘tree’, also interpreted as wood: (26) a. šloša kbas-im three.masc lamb.masc-pl ‘three lambs’

ħamiša oꝑ-ot five.masc bird.masc-pl ‘five birds’

šney ec-im two.masc tree.masc-pl ‘two trees’ b. ódeꝑ kébes too much lamb ‘too much lamb’

kílo oꝑ kilo chicken ‘a kilo of chicken’

harbe ec a lot wood ‘a lot of wood’

the cognitive basis of the mass-count distinction

87

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 ében ton stone ‘a ton of stone’

harbe ħébel a lot rope ‘a lot of rope’

b. ħameš aban-im five.fem stone.fem-pl ‘five stones’

yoter miday bír-a too much beer-fem ‘too much beer’

šney ħabal-im two.masc rope.masc-pl ‘two ropes’

šaloš bír-ot three.fem beer.fem-pl ‘three beers’ or by the “subkind coercion” whereby šaloš bírot ‘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-occurrence 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. ħamiša gzar-im / šib’a bcal-im / štey ħás-ot / dla’-ot five carrots / seven onions / two lettuces / pumpkins šmona cnon-im / tíras-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 müller

b. kílo gézer / bacal / tut / ħás-a / léꝑet kilo carrot / onion / mulberry / lettuce / turnip / dlá’at / šumar ódeꝑ cnon / tíras / krub 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 individuated carrots which are measured.14 Similarly, though it is possible to interpret tíras ‘corn’ as corn grains in (28b), it is also interpretable as individuated corn cobs. Also, whereas ódeꝑ kébes ‘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), ódeꝑ gézer ‘too much carrot’ can be interpreted as ‘too many carrots’ despite of the singular form of gézer ‘carrot’. Moreover, if the mass interpretation 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. kílo tapuħ-im / agas-im / ħacil-im / qišu-im / eškoliy-ot kilo apples / pears / aubergines / courgettes / grapefruits harbe adaš-im / anab-im / zeyt-im / botn-im / dubdeban-im many lentils / grapes / olives / peanuts / cherries ódeꝑ agbaniy-ot / šeziꝑ-im / tapuz-im / pitriy-ot too many tomatoes / plums / oranges / mushrooms b. *kílo tapúaħ / agas / ħacil / qišu / 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’, abatíaħ ‘watermelon’. 15 We leave out borrowed nouns such as míšmiš (Arabic) ‘apricot’, ánanas (French) ‘pineapple’, anóna (Latin) ‘sweetsop’, aboqádo ‘avocado’, batáta ‘sweet potato’, fijóya ‘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. šney šeseq ‘two loquats’, kilo šeseq ‘a kilo loquats’.

the cognitive basis of the mass-count distinction

89

*harbe adaš-a / anab / záyit / bóten / dubdeban many lentil / grape / olive / peanuts / cherry *ódeꝑ eškolit / šeziꝑ / tapuz / pitriy-a too many grapefruit / plum / orange / mushroom Neither is the count interpretation in (28a) the result of coercion by the “universal 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 gézer / tut / šumar / bacal buy carrot / mulberry / fennel / onion ‘Buy carrots/ mulberries/ fennels/onions.’ b. #tiqni tapúaħ / agas / tapuz / agbaniya / šeziꝑ 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 gézer / cnon / bacal mi ma šeshe bought more carrot / radish / onion than what that biqašnu we-asked ‘She bought more carrots / radishes / onions than we asked.’

90

doron and müller

b. *hi qanta yoter tapúaħ / agbaniya / šeziꝑ mi ma šeshe bought more apple / tomato / plum than what that biqašnu we-asked In partitive and existential examples where there isn’t a contextual preference for plural readings and both singular and plural interpretations are in principle felicitous, the singular form of a count noun in (32b–33b) below only gives rise to a single unit interpretation, whereas the singular form of the flexible noun in (32a–33a) also refers to a plurality in addition to singular reference. (32) a. rob ha-gézer raquv most (of) the carrot is rotten ‘Most of the carrot is rotten.’ ‘Most of the carrots are rotten.’ b. rob ha-tapúaħ raquv most (of) the apple is rotten ‘Most of the apple is rotten.’ (33) a. yeš gézer b-a-tiq there (is) carrot in-the-bag ‘There is a carrot in the bag.’ ‘There are carrots in the bag.’ b. yeš tapúaħ 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-ħayim lo nacliaħ le-gadel et ha-gézer še hem in-the-life not we-succeed to-grow acc the-carrot that they criḵim bišbil miꝑ’al-ha-šimurim šel-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.’

the cognitive basis of the mass-count distinction

91

(35) b-a-ħayim lo nacliaħ le-gadel et ha-tapuħ-im/*tapúaħ še in-the-life not we-succeed to-grow acc the-apples/*apple that hem criḵim bišbil miꝑ’al-ha-šimurim šel-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 ohébet gézer / *tapúaħ / tapuħ-im I love carrot / apple / apples ‘I love carrots/apples.’ (37) a. mic gézer / *tapúaħ / tapuħ-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’, me’at ‘a little’, ódeꝑ/ yoter miday ‘too much’, kílo ‘a kilo of’ B. In the context of shopping C. In partitive and existential constructions

92

doron and müller

D. With amount relatives E. In reference to kinds On the other hand, flexible nouns are similar to mass nouns such as petruzílya ‘parsley’, šamir ‘dill’, téred ‘spinach’, šu’iyt ‘bean’, šum ‘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. kílo téred harbe petruzíly-a me’at šamir ódeꝑ šu’iy-t kilo spinach much parsley a little dill too much bean b. *téred eħad *štey petruzíly-ot *šamir eħad *šaloš šu’iy-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, mulberry, 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 recognizable 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 similarities, 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

the cognitive basis of the mass-count distinction

93

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 considered 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 normally 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 uniform 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 uniform 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 indiscriminate 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 müller

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 Chierchia’s (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 doesn’t, 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.

the cognitive basis of the mass-count distinction

95

(39) The singulative alternation (singulative is derived by fem suffix) Basic mass noun Singulative count noun Plural se’ar hair.masc

sa’ar-a a-hair-fem

sa’ar-ot hairs

sá’ar se’ar-a turbulence.masc storm-fem

se’ar-ot storms

ómer speech.masc

imr-a saying-fem

imr-ot sayings

síaħ discourse.masc

siħ-a conversation-fem

siħ-ot conversations

mávet death.masc

mit-a a-death-fem

mit-ot deaths

(40) The collective alternation (collective is derived by fem suffix) Basic count noun Plural

Collective mass noun19

a.

dag fish.masc

dag-im fish-pl

dag-a fish-fem

b.

ale leaf.masc

al-im leaves

alv-a 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 (40e–f) 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 müller

(40) The collective alternation, Continued Basic count noun Plural

Collective mass noun

c.

gole expatriate.masc

gol-im expatriates

gol-a diaspora-fem

d.

aꝑun a-pea.masc

aꝑun-im peas

aꝑun-a pea-fem

e.

–

ħit-im ħit-a wheat plants wheat-fem

f.

–

se’or-im se’or-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 Basic count noun

Plural

Collective mass noun

a.

béged garment

bgad-im garments

bigud clothing-ASN

b.

ná’al shoe

na’al-áyim shoes

han’ala 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’, adaš-a/ adaš-im ‘lentils’, te’en-a/ te’en-im ‘figs’ etc. There are other languages in which nouns of this sort have singulative morphology (cf. Schwarzwald 1991).

the cognitive basis of the mass-count distinction

Basic count noun

Plural

97

Collective mass noun

c.

rehit rehit-im rihut piece of furniture20 pieces of furniture furniture-ASN

d.

péraħ flower

praħ-im flowers

priħa bloom-ASN

e.

mircéꝑet tile

mircaꝑ-ot tiles

ricuꝑ tiling-ASN

Both mass nouns with collective morphology and mass nouns related to count nouns with singulative morphology are fake mass nouns—mass nouns which nevertheless have natural units. What is special about the flexible nouns discussed 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 alternation:

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 niꝑtáħat mehava šney rehit-im be- rehit eħad 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 šney maḵšir-im be- maḵšir eħad these (are) two gadgets in gadget one ‘These are two gadgets in one.’

98

doron and müller

(42) The collective—singulative alternation (flexible nouns) Collective mass noun Singulative count noun Plural

4

gézer carrot

gézer carrot

gzar-im carrots

bacal onion

bacal onion

bcal-im onions

tut mulberry

tut mulberry

tut-im etc mulberries21

Conclusion

We have argued that fake mass nouns do not distort after all the correspondence between a clear cognitive distinction and the mass-count linguistic distinction. 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, mulberry, 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.

the cognitive basis of the mass-count distinction

99

ration. Chierchia’s (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.” Proceedings of the 28th West Coast Conference on Formal Linguistics, 33–41. Somerville, MA: Cascadilla Proceedings Project. Baerman, Matthew. 2007. “Morphological Reversals.” Journal of Linguistics 43:33–61. Barner, David and Jesse Snedeker. 2005. “Quantity judgments and individuation: evidence that mass nouns count.” Cognition 97:41–66. 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 53:520–542. Cheng, Lisa L., Jenny Doetjes and Rynt Sybesma 2008. “How universal is the Universal Grinder?” Linguistics in the Netherlands 2008, 50–62. Chierchia, Gennaro. 1998. “Reference to Kinds Across Languages.” Natural Language Semantics 6:339–405. Chierchia, Gennaro. 2010. “Mass nouns, vagueness and semantic variation.” Synthese 174: 99–149. Chung, Sandra. 2000. “On reference to kinds in Indonesian.” Natural Language Semantics 8:157–171 Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press. Doron, Edit. 2003. “Bare Singular Reference to Kinds.” Proceedings of Semantics and Linguistic Theory 13:73–90. Faust, Noam. 2011. Forme et fonction dans la morphologie nominale de l’hébreu moderne. PhD Thesis, Université Paris Diderot. Gillon, Brendan. 1992. “Toward a common semantics for English count and mass nouns.” Linguistics and Philosophy 15:597–639.

100

doron and müller

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 6–8, 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 5:133–158, edited by Olivier Bonami & Patricia Cabredo Hofherr. Landman, Fred. 2010. “Count nouns, mass nouns, neat nouns, mess nouns.” In Formal Semantics and Pragmatics: Discourse Context and Models, edited by Jurgis Skilters. Riga: The Baltic International Yearbook of Cognition, Logic and Communication. Link, Godehard. 1983. “The logical analysis of plural and mass terms.” In Meaning, Use and Interpretation of Language, edited by Reiner Bäuerle, Christoph Schwarze and Arnim von Stechow, 302–323. Berlin: de Gruyter. Reprinted in G. Link. 1998. Algebraic Semantics in Language and Philosophy, 11–33. Stanford: CSLI Publications. Mittwoch, Anita 1988. “Aspects of English aspect: on the interaction of perfect, progressive and durational phrases.” Linguistics and Philosophy 11:203–254. 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. Müller, 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, 122–135. Dept. Linguistics, UBC. Nicolas, David. 2002. “Do mass nouns constitute a semantically uniform class?” Kansas Working Papers in Linguistics 26:113–121. 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: 405–443. Rothstein, Susan. 2010. “Counting and the mass-count distinction.” Journal of Semantics 27:343–397. Schwarzwald, Ora. 1991. “Grammatical vs. Lexical Plural Formation in Hebrew.” Folia Linguistica 25:577–608. Tsoulas, George. 2006. “Plurality of mass nouns and the grammar of Number.” Paper presented at the 29th GLOW colloquium in Barcelona.

the cognitive basis of the mass-count distinction

101

Wilhelm, Andrea. 2008. “Bare nouns and number in Dëne Sųłiné.” Natural Language Semantics 16:39–68. Wright, William. 1859. A Grammar of the Arabic Language [1967. Third edition containing both volumes]. Cambridge: Cambridge University Press.

The Turkish NP* Željko Bošković and Serkan Şener This paper investigates the Turkish NP, with emphasis on the structural position of possessors (both overt and null possessors—they 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 ellipsis 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 elements, 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 Bošković’s (2008a, 2012a) generalizations regarding DP/NP languages. In principle, even if some articleless languages don’t have DP, it may not be out of the question that some could have it, hence we first test Turkish regarding Bošković’s generalizations. We show that Turkish patterns with NP, not DP languages, which has an important 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 interpretation 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

Bošković (2008a,2012a) argues for a no-DP analysis of languages without definite articles, based on a number of cross-linguistic generalizations acknowledging 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_006

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 summarized here; see Bošković 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 crosslinguistic 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 hasn’t/*has visited her in at least two years. (2) *John doesn’t claim [that Mary has visited her [NPI in at least two years]] (3) John doesn’t believe [that Mary has visited her [NPI in at least two years]] Bošković (op. cit.) notes that whether or not a language allows clause-mate NPIlicensing under NR out of indicative clauses depends on whether it has articles, 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

bošković and şener

Romanian, and Bulgarian, which have articles, allow it.) Note that the generalization 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 yıl-dır ziyaret et-me-di/*et-ti. P.-nom M.-acc at least two year-for visit do-neg-past/do-past ‘Pelin hasn’t/*has visited Mete in at least two years.’ (5) Mete [Pelin-∅/-i (*en az iki yıl-dır) Timbuktu-ya git-ti] M.-nom P.-nom/-acc at least two year-for T.-dat go-past san-m-ıyor. think-neg-pres ‘Mete doesn’t 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 argument 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 don’t 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. *Osmanlılar-ın İstanbul-un feth-i Ottomans-gen İstanbul-gen conquest-3sg.poss ‘Ottomans’ conquest of Istanbul.’ b. cf. İstanbul-un feth-i

1.3 Only Article-Less Languages May Allow Scrambling Bošković (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

105

the turkish np

1.4 Radical Pro-Drop is Possible Only in Article-Less Languages Bošković (2012a) defines radical pro-drop as the productive pro-drop of subjects 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 languages. Turkish has subject agreement and Spanish-style subject-drop. However, it also has productive object-drop although it lacks object agreement (see Şener 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 didn’t see anybody.’ (Negative Concord only) b. Nessuno studente ha letto nessun libro. no student has read no book

(Double Negation only)

Bošković (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) Hiçbir çocuk hiçbir kitab-ı oku-ma-dı. 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

bošković and şener

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 doesn’t have that exhaustivity presupposition, although it is definite: (9) Zhangsan’s three sweaters (10) Zhangsan de [san jian maoxianyi] Z deposs three CL sweater ‘Zhangsan’s three sweaters’ Bošković (2012a) shows that we are dealing here with a broader generalization, with the exhaustivity presupposition absent from NP languages (e.g. Russian, SC, Chinese, Japanese, Korean, Hindi, Bangla, Malayalam, and Magahi) in such contexts. Turkish again patterns with NP languages: (11) doesn’t have the exhaustivity presupposition:4 (11) Can-ın üç bisiklet-i John-gen three bicycle-3sg.poss ‘John’s three bicycles’ 1.7 Only Article-Less Languages May Allow LBE Bošković (2012a) shows that only article-less languages may allow AP leftbranch extraction (LBE). (Thus, such extraction is allowed in Russian, Polish, Czech, Ukrainian, Slovenian, Latin, Mohawk, Southern Tiwa, Gunwinjguan languages, 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 mašinu] *Expensivei he saw [ti car]

(Russian)

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 [kalın kitap] oku-du. P-nom thick book read-past ‘Pelin read a thick book.’ b. *Kalın1 Pelin [t1 kitap] oku-du. This still wouldn’t provide an argument for DP in Turkish, given the one-way correlation status of 1.7. In fact, Bošković (2012a) notes that LBE often requires A-N agreement, which Turkish doesn’t have. However, LBE is actually allowed in Turkish: an adjective can be postposed to the post-verbal field (Kornfilt 2003, Göksel and Kerslake 2005). (14) ?Pelin [t1 kitap] oku-du kalın1. Note that postposing in Turkish only involves non-contrastive elements that are given in the previous discourse, classified as [–contrastive,+discourse anaphoric] in Şener (2010). Fronting however typically involves [+contrastive] constituents, which may also be [+topic] or [+discourse anaphoric]. As for foci, they remain in-situ—they cannot undergo movement no matter whether they are + or -contrastive (Şener 2010). Kalın 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 anaphoric] and all movement of such elements in Turkish is to the right).5 1.9 Only Article Languages Allow the Majority Superlative Reading Živanovič (2008) reports that (15) in Slovenian only allows the plurality reading. 5 See Kural (1997), Kornfilt (2005), and Şener (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 (Bošković 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 yaşlı modifies the higher N.) (i)

*Pelin [[pro ti teyze-m]-in arkadaş-ı]-nı gör-müş P-nom aunt-1sg.poss-gen friend-3sg.poss-nom see-evidential.past yaşlıi. old ‘Pelin saw the friend of my old aunt.’

108

bošković and şener

(15) Največ 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. Živanovič (2008) notes that a broader generalization is at stake here, where the Majority Reading is allowed only in article languages (Živanovič 2008 and Bošković 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 concerns only traditional determiners, not the cases where the Majority Reading is expressed by nouns such majority or Turkish çoğu (as in insanlar-ın çoğ-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) İnsanlar en çok bira iç-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 Bošković (2012a) concerns inverse scope interpretation 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. Bošković (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 öğrenci her sandalye-yi kır-mış. 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. Susumu-nom book-acc read ‘Susumu read a/the book/books.’

(Japanese)

Bošković (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, Warrgamay, Indonesian, Vietnamese), all lack articles, establishing 1.11. Note however that 1.11 doesn’t conversely require all NP languages to lack number morphology. 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 morphology (its accusative counterpart cannot be so interpreted, see Aydemir 2004, Ketrez 2005, Öztürk 2004, Şener 2010).

110

bošković and şener

(22) Can (kalın) kitap oku-muş. 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 Bošković (2012a), which provides motivation for classifying 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 kız-lar]-la sinema-ya git-ti. school-abl M-nom girl-pl-with movies-dat go-past b. *Mete [ti kız-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 Bošković in press b for discussion of the phenomenon). A number of Bošković’s generalizations are irrelevant because Turkish doesn’t 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 Bošković 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, Şener (2010) argues this doesn’t 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 don’t undergo movement. This makes the focus-movement generalization irrelevant to Turkish.

the turkish np

111

rather straightforward test proposed in Despić (2011, 2013) that determines the position of possessors. In his NP analysis of Serbo-Croatian (SC), Bošković (2012a) treats SC possessors and demonstratives as NP-adjuncts. One of the arguments for this analysis, noted by Despić (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. Tarantinoi’s latest movie really disappointed himi. (24) a. *[NP Kusturicini [NP najnoviji film]] gai je zaista razočarao. Kusturica’s latest movie him is really disappointed ‘Kusturicai’s latest movie really disappointed himi.’ b. *[NP Njegovi [NP najnoviji film]] je zaista razočarao Kusturicui. his latest movie is really disappointed Kusturica Significantly, Turkish patterns with SC, not English. As (25) shows, the possessor apparently c-commands out of its TNP, as these sentences are clear violations of Conditions B/C. We therefore assume that possessors are also NPadjoined in Turkish, with the DP-layer missing in this language. (25) a. *[Özpeteki-in film]-i oi-nu hayal kırıklığına uğrat-tı. Ö.-gen movie-3sg.poss he-acc disappoint-past ‘Özpetek’s movie disappointed him.’ b. *[oi-nun film]-i Özpeteki-i hayal kırıklığına uğrat-tı. he-gen movie-3sg.poss Ö.-acc disappoint-past ‘His movie disappointed Özpetek.’

7 Japanese and Chinese pattern with SC (Bošković 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 wouldn’t 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 Bošković (2012a) (the co-indexed elements shouldn’t be focused). Note also that Condition A cannot be tested here due to interfering factors, see Despić (2011).

112

bošković and şener

Now consider word order within Turkish TNPs:8 (26) Poss»Dem»(A)»Num»(A)»N a. Can-ın şu (eski) üç (eski) bisiklet-i John-gen that old three bicycle-3sg.poss ‘those three old bicycles of John’s’ *Poss»Num»Dem»A»N b. *Can-ın üç şu eski bisiklet-i *Num»Poss»Dem»A»N c. *üç Can-ın şu eski bisiklet-i *Poss»A»Dem»Num»N d. *Can-ın eski şu üç bisiklet-i *A»Poss»Dem»Num»N e. *eski Can-ın şu üç bisiklet-i (27) Dem»Poss»(A)»Num»(A)»N a. şu Can-ın (eski) üç (eski) bisiklet-i *Dem»Num»Poss»A»N b. *şu üç Can-ın eski bisiklet-i *Num»Dem»Poss»A»N c. *üç şu Can-ın eski bisiklet-i *Dem»A»Poss»Num»N d. *şu eski Can-ın üç bisiklet-i *A»Dem»Poss»Num»N e. *eski şu Can-ın üç 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 Bošković (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 Bošković’s 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 Bošković (2012a). (However, note that we follow Bošković’s 2009, 2012b analysis of adjectives, where adjectives are located in multiple Specs of the same phrase, rather than Cinque’s 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 wouldn’t 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 (Bošković 2008b). Ketrez (2004), however, shows that the “two” birs have identical syntactic 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 general 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, Öztürk 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 BİR/İKİ bisiklet al-dı. A-nom blue one/two bicycle buy-past ‘Ali bought one/two blue bicycle(s).’ b. Ali bir/iki mavi bisiklet al-dı.

114

bošković and şener

(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 Despić 2011, Bošković 2012a.) (29) a. *[şu Özpeteki-in film]-i oi-nu hayal kırıklığına uğrat-tı. that Ö.-gen movie-3sg.poss he-acc disappoint-past ‘That movie of Özpetek’s disappointed him.’ b. *[şu oi-nun film]-i Özpeteki-i hayal kırıklığına uğrat-tı. that he-gen movie-3sg.poss Ö.-acc disappoint-past ‘That movie of him disappointed Özpetek.’ The violations of Conditions B/C in (29) provide clear evidence that the Dem that precedes Poss is not in a separate projection (it doesn’t 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)

?[Özpeteki-in

Ö.-gen

şu film]-i oi-nu hayal kırıklığına uğrat-tı. that movie-3sg.poss he-acc disappoint-past

Notice that (ii) is fully unacceptable (due to a Condition C violation). (ii)

*[oi-nun şu film]-i Özpeteki-i hayal kırıklığına uğrat-tı. he-gen that movie-3sg.poss Ö.-acc disappoint-past ‘That movie of his disappointed Özpetek.’

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. *[Özpeteki-in iki (tane) film]-i oi-nu Ö.-gen two CLL movie-3sg.poss he-acc hayal kırıklığına uğrat-tı. disappoint-past ‘Two movies of Özpetek’s disappointed him.’ b. *[Özpeteki-in eski film]-i oi-nu hayal kırıklığına uğrat-tı. Ö.-gen old movie-3sg.poss he-acc disappoint-past ‘Özpetek’s 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) don’t have the presupposition ‘Z has exactly three sweaters’: (31) Can-ın üç bisiklet-i John-gen three bicycle-3sg.poss ‘John’s three bicycles’ (32) Zhangsan de [san jian maoxianyi] Z deposs three cl sweater ‘Zhangsan’s 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 Zhangsan’s books’ Recall that Poss»Num»N 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

bošković and şener

(34) Can-ın üç bisiklet-i garaj-da. John-gen three bicycle-3sg.poss garage-loc ‘John’s 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 Num»Poss order, Poss is therefore interpreted in its reconstructed position. Bošković (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 bivše pantalone John’s former pants b. bivše Jovanove pantalone 12 This movement violates Bošković’s (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 doesn’t arise.

the turkish np

117

While Turkish only allows the Poss»A 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-ın eski pantolon-u John-gen former pants-3sg.poss ‘John’s 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 combined with Takahashi’s (1996, 2000, 2001) claim that pro doesn’t 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 wouldn’t move to the NP-adjoined position (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) [şu proi film]-i Özpeteki-i hayal kırıklığına uğrat-tı. that movie-3sg.poss Ö.-acc disappoint-past ‘That movie of his disappointed Özpetek.’ 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 doesn’t 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-doesn’t-move analysis is (i), where what appears to be an anaphor does not induce a Condition C effect. (i)

[kendii film]-i Özpeteki-i hayal kırıklığına uğrat-tı. own movie-3sg.poss Ö.-acc disappoint-past ‘His own movie disappointed Özpetek.’

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 rightadjoined to it (see Aygen 2002 and Şener 2008 for such adjunction analyses of emphatic kendi). Since pro doesn’t 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

bošković and şener

(39) [üç pro kitab-ım]-ı ev-de bırak-tı-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. Poss»Dem »Num»A»N b. Poss»Dem»A»Num»N c. Dem»Poss»A»Num»N d. Dem»Poss»Num»A»N

3

NP/Nʹ-Ellipsis

3.1 Impossible Cases We now turn to TNP-internal ellipsis, referred to below as NP/Nʹ-ellipsis. Notice first that ellipsis inside bare objects with numerals is disallowed. (41) *Pelin her gün [beş 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 NP’s number is irrelevant):15 (42) a. *Pelin [eski kitap] sat-tı, Suzan-sa [yeni kitap] sat-tı. 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 Özpeteki-i hayal kırıklığına uğrat-tı. he-gen own movie-3sg.poss Ö.-acc disappoint-past ‘His own movie disappointed Özpetek.’

b.

*[Özpeteki-in kendii film]-i oi-nu hayal kırıklığına uğrat-tı. Ö.-gen own movie-3sg.poss he-acc disappoint-past ‘Özpetek’s own movie disappointed him.’

15 See section 3.2.2 for accusative objects.

the turkish np

119

b. *Pelin [eski araba-lar] sat-tı, Suzan-sa [yeni araba P.-nom old car-pl sell-past S.-nom-however new -ler] sat-tı. -pl sell-past ‘Pelin sold old cars, while Susan sold new ones.’ NP/Nʹ-ellipsis also cannot strand possessors: (43) *[Pamuk-un kitab-ı-nı] oku-du-m, ama [Oe-nin kitab-ı-nı] P.-gen book-3sg.poss-acc read-past-1sg but O.-gen oku-ma-dı-m. read-neg-past-1sg ‘I read Pamuk’s book, but I didn’t read Oe’s.’ 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. Bütün gün kitap oku-du-m, *san-a da kitap oku-ma-n-ı 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. Bütün gün kitap oku-du-m, san-a da kitap oku-ma-n-ı tavsiye ed-er-im.

120

bošković and şener

We now turn to classifier constructions. Count nouns quantified by a numeral don’t require a Classifier-like element (CLL). However, CLL can be present, bringing in an individuating function (Öztürk 2004). (ii)

Dün bir kitap oku-du-m, san-a da {bir kitap yesterday one book read-past-1sg you-dat too oku-ma-n-ı/ bir kitap oku-ma-n-ı} 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 doesn’t (have to) incorporate. This test indicates numeral NPs don’t (have to) incorporate: (iii)

Dün iki kitap oku-du-m, san-a da iki kitap oku-ma-n-ı 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)

Dün iki kitap oku-du-m, san-a da iki kitap oku-ma-n-ı tavsiye ed-er-im.

Bare NPs with adjectives behave differently: (v)

a. Ben soğuk çay iç-eceğ-im, ??san-a da soğuk çay I-nom cold tea drink-fut-1sg you-dat also iç-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 soğuk çay iç-eceğ-im, san-a da soğuk çay iç-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.

*[Yaşlı bakıcı] çocuk-ları azarla-dı, ama [genç bakıcı] onlar-ı 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. üç kitap three book b. üç tane kitap three CLL book ‘three (items of) books’ Importantly, NP/Nʹ-ellipsis is possible if Num is accompanied by CLL, in contrast to (41), where there is no CLL and ellipsis is disallowed.17 (46) Pelin her gün [üç 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.’ Bošković (2012a, in press a,b) and Despić (2011) argue that non-adjectival numerals in SC have a distinct projection above NP, namely QP. (SC also has adjectival numerals which don’t 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). Consider (47). b. *Bu dönem [üç öğrenci] sınıfta kal-dı, ama geçen dönem [beş this semester three student flunk-past but last semester five öğrenci] sınıfta kal-mış-tı. flunk-past.perfect ‘This semester three students flunked, whereas in the past semester five students had flunked.’ c.

*[Pelin-in anne-si] dün gel-di, [Mete-nin anne-si] bugün P.-gen mother-3sg.poss yesterday come-past M.-gen today gel-di. come-past ‘Pelin’s mother came yesterday, Mete’s 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 gün üç tane ye-r. P.-nom every day three CLL eat-aor

122

bošković and şener

(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/Nʹ-ellipsis. Under a DP analysis of Turkish, possessors are analysed as occupying SpecDP, which naturally predicts that Poss-stranding ellipsis should be possible. The ungrammaticality 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-ın kitab-ı-nı] oku-du-m ama [Pamuk-un A.-gen book-3sg.poss-acc read-past-1sg but P.-gen kitab-ı-nı] oku-ma-dı-m. read-neg-past-1sg ‘I read Auster’s book, but I didn’t read Pamuk’s.’

18 Note that bare Num+CLL NPs don’t incorporate (see footnote 16). (i)

Dün üç tane kitap oku-du-m, san-a da üç tane kitap yesterday three CLL book read-past-1sg you-dat also oku-ma-n-ı 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 wouldn’t conform to Saito and Murasugi’s 1990 Spec-Head agreement requirement, but this requirement anyway has exceptions, see Bošković 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 doesn’t necessarily have to show the exact same properties as the clausal domain in terms of headedness). There are, however, 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-ın iki kitab-ı-nı] oku-du-m, ama [Pamuk-un A.-gen two book-3sg.poss-acc read-past-1sg but P.-gen iki kitab-ı-nı] oku-ma-dı-m. read-neg-past-1sg ‘I read two books of Auster’s, but I didn’t read two books of Pamuk’s.’ As discussed above, in contrast to the DP analysis, the impossibility of possessor-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 doesn’t change the judgment (see (50)). The same holds for demonstratives (51).20 (50) *[Auster-ın iki kitab-ı-nı] oku-du-m, ama [Pamuk-un A.-gen two book-3sg.poss-acc read-past-1sg but P.-gen üç kitab-ı-nı] oku-du-m. three read-past-1sg (51) a. *[Sartre-ın şu kitabı-nı] oku-du-k ama [Beckett’in S.-gen that book-acc read-past-1pl but B.-gen şu kitabı-nı] oku-ma-dı-k. read-neg-past-1pl ‘We read that book of Sartre’s but didn’t read that book/one of Beckett’s.’ b. *[Sartre-ın şu kitabı-nı] oku-du-k ama [Beckett’in şu kitab-ı-nı] oku-ma-dı-k. c. *[Şu Sartre-ın kitabı-nı] oku-du-k ama [şu Beckett’in kitab-ı-nı] oku-ma-dı-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 unacceptable if Acc is not elided (see below for case and demonstratives).

124

bošković and şener

Significantly, Poss-stranding ellipsis is not always disallowed; it is possible in the presence of CLL.21 (52) ?Pelin [Chomsky-nin üç tane kitab-ı-nı] P.-nom C.-gen three CLL book-3sg.poss-acc oku-muş, ama [Foucault-nun iki tane kitab-ı-nı] read-evidential.past but F.-gen two CLL oku-muş. read-evidential.past ‘Pelin read three books of Chomsky’s, but s/he read two books of Foucault’s.’ (53) cf. *… ama Foucault-nun iki oku-muş. 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 Foucault’s books, A says: (i)

#Pelin Foucault-nun iki tane oku-muş P-nom F-gen two CLL read-evidential.past

the turkish np

125

(55) *Pelin [Chomsky-nin üç tane kitab-ı-nı] P.-nom C.-gen three CLL book-3sg.poss-acc oku-muş, ama [Foucault-nun üç tane kitab-ı-nı] read-evidential.past but F.-gen oku-ma-mış. read-neg-evidential.past Regarding the binding properties of possessors, Condition-B/C effects are predicted to show up in this context too since CLLPs don’t change Poss’s c-command properties: (56) a. *[Özpeteki-in iki tane film]-i oi-nu Ö.-gen two CLL movie-3sg.poss he-acc hayal kırıklığına uğrat-tı. disappoint-past ‘Özpetek’s two movies disappointed him.’ b. *[oi-nun iki tane film]-i Özpeteki-i he-gen two CLL movie-3sg.poss Ö.-acc hayal kırıklığına uğrat-tı. disappoint-past ‘His two movies disappointed Özpetek.’ 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-mış ama Mete [yeni P.-nom old two book sell-evidential.past but M.-nom new (iki) kitap] sat-mış. 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 basegenerated 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

bošković and şener

(58) Pelin [kalın üç tane kitap] oku-du, Pınar-sa [ince P.-nom thick three CLL book read-past P.-nom-however thin üç tane kitap] oku-du. three CLL read-past ‘Pelin read three long books, but Pınar 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 [kalın üç tane kitap] oku-du, Pınar-sa [ince P.-nom thick three CLL book read-past P.-nom-however thin üç 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 üç tane masa-sı P.-gen old three CLL desk-3sg.poss b. *Eski Pelin-in üç tane masa-sı c. Pelin-in üç tane eski masa-sı d. *Pelin-in üç eski tane masa-sı If tane is CLL0, we need to assume that in (60c) we are dealing with a PF reordering (unless eski is NP-adjoined, see below). If tane and Num are in different 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, depending 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)

*Pınar ince üç 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 kalın+kitap under NP-ellipsis would require this option to also be available. However, it’s not easy to determine what is elided here, given that kitap-deletion doesn’t prevent the interpretation where Pınar read two long books (which is the only interpretation under kalın+kitap deletion, and is the most natural interpretation here). (61) Pelin [üç tane kalın kitap] oku-du, Pınar-sa [üç 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 Pınar read three short ones.’ (62) Pelin [üç tane kalın kitap] oku-du, Pınar-sa [iki tane {kalın 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 predicates: (63) a. Senin kuzeninin bu kazağı çok sevdiğini biliyorum. I know your cousin likes this sweater a lot. O yüzden, bu kazak artık [kuzen-i-nin because-of-that, this sweater-nom now cousin-3sg.poss-gen kazağ-ı]. sweater-3sg.poss ‘Because of that, the sweater is now your cousin’s.’ b. Bu benim kazağım. this is my sweater. Bu da [sen-in/Pelin-in/kuzen-in-in kazağ-ı] this too you-gen/Pelin-gen/cousin-3sg.poss-gen sweater-3sg.poss ‘And this is yours/Pelin’s/your cousin’s sweater.’ Following Bowers (1993) and Koster (1994), we assume that the relevant NPs here are dominated by a predicate projection, PredP, with the possessor generated 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

bošković and şener

(64)

However, a number of authors have argued that at least some cases of “stranded” 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 possessors allow only the possession reading,23 which is the one argued to involve no null element and, according to Tremblay, is confined to predicate positions; it is, e.g., disallowed with objects). Under this approach, (63) doesn’t involve ellipsis, hence there is no need to posit PredP and Poss-movement here. In fact, such constructions don’t require a linguistic antecedent, which suggests 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-malı. this (desk) P.-gen be-deontic.necessity ‘This desk should be Pelin’s.’ A linguist is doing an experiment. She turns to the colleague and says (turning the informant over to his colleague): b. Şu andan itibaren (o) sen-in. from-now-on he-nom you-gen ‘From now on, he’s 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).

129

the turkish np

(66) a. [Şu [iki kitab]]-ı oku-du-m ama bu-nu oku-ma-dı-m. that two book-acc read-past-1sg but this-acc read-neg-past-1sg Intended Reading: ‘I read these two books but didn’t read those two books.’ b.

… bu-nlar-ı … this-pl-acc ‘these’

c. Şu [iki kitab]]-ı oku-du-m ama [bu [iki kitap]]-nu that two book-acc read-past-1sg but this two book-acc oku-ma-dı-m. read-neg-past-1sg Intended Reading: ‘I read these two books but didn’t read those two books.’ Two forms, bu-nu and bu-nlar-ı, 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 shouldn’t 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 doesn’t 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]*(-ı) oku-ma-dı-m. this two book-acc read-neg-past-1sg b. Bu-*(nu) sev-er-im this-acc like-aor-1sg ‘I like this.’

130

bošković and şener

(67) *[Şu [iki kitab]]-ı oku-du-m ama [bu iki [kitab]] -yi that two book-acc read-past-1sg but this two -acc oku-ma-dı-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) [Şu [iki (tane) kitab]]-ı oku-du-m ama [bu iki tane that two (CLL) book-acc read-past-1sg but this two CLL [kitap]] -yi oku-ma-dı-m -acc read-neg-past-1sg ‘I read those two books but didn’t 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 shouldn’t be analysed in this way. 3.2.1 Pronominal ki We have seen Turkish possessors cannot be stranded under NP/Nʹ-ellipsis: (69) *[Pamuk-un kitab-ı-nı] oku-du-m, ama [Oe-nin kitab-ı-nı] P.-gen book-3sg.poss-acc read-past-1sg but O.-gen oku-ma-dı-m. read-neg-past-1sg ‘I read Pamuk’s book, but didn’t read Oe’s.’ However, (69) becomes acceptable in the presence of ki, which attaches to Poss: (70) [Pamuk-un kitab-ı-nı] oku-du-m, ama [Oe-nin-ki-ni] P.-gen book-3sg.poss-acc read-past-1sg but O.-gen-ki-acc oku-ma-dı-m. read-neg-past-1sg ‘I read Pamuk’s book, but didn’t 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, Göksel and Kerslake 2005).26 Adopting this analysis, we argue that ki-NPs don’t 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-ı-nı] oku-du-m, ama [Oe-nin-ki P.-gen book-3sg.poss-acc read-past-1sg but O.-gen-ki kitab-ı]-nı oku-ma-dı-m. book-3sg.poss-acc read-neg-past-1sg

26 Ki also attaches to locative/temporal expressions, turning them into nominal modifiers: (i)

dün-kü hava yesterday-ki weather ‘yesterday’s weather’

132

bošković and şener

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-ın iki kitab-ı]-nı oku-du-m ama A.-gen two book-3sg.poss-acc read-past-1sg but [Pamuk-un-ki]-ni oku-ma-dı-m. P.-gen-ki-acc read-neg-past-1sg ‘I read two books by Auster, but didn’t read the one by Pamuk.’ (73) [Auster-ın iki kitab-ı]-nı oku-du-m ama A.-gen two book-3sg.poss-acc read-past-1sg but [Pamuk-un-ki-ler]-i oku-ma-dı-m. P.-gen-ki-pl-acc read-neg-past-1sg ‘I read two books by Auster, but didn’t 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 it’s 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 Beckett’s book.27 (74) [Sartre-ın şu kitab-ı-nı] oku-du-k ama S.-gen that book-3sg.poss-acc read-past-1pl but [Beckett-in-ki]-ni oku-ma-dı-k. B.-gen-ki-acc read-neg-past-1pl ‘We read that book by Sartre, but didn’t 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)

(ii)

*[Sartre-ın şu kitabı-nı] oku-du-k ama [Beckett-in {şu-nu-ki/şu-ki-ni}] S.-gen that book-acc read-past-1pl but B.-gen that-acc-ki/that-ki-acc oku-ma-dı-k. read-neg-past-1pl ‘We read that book by Sartre but we didn’t read that one by Beckett.’ [Şu Sartre-ın kitabı-nı] oku-du-k ama [şu Beckett-in-ki-ni] oku-ma-dı-k.

the turkish np

133

Finally, ki-NPs don’t require a linguistic antecedent, which confirms they don’t 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-malı. this (present) P.-gen-ki be-epistemic.modal ‘This (present) must be Pelin’s.’ 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 Taroo’s attitude is good, Hanako’s isn’t.’ 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]-ı aslında 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. Çantalarımız masada; benim*(-ki-ni) verir misin? Our bags are on the table; Can you pass me my one/mine? b. Borçları 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 suggests non-ki stranded possessors may be unambiguously possessive. Note also that ki is impossible in (79), which is possessive, but obligatory in (80).

134

bošković and şener

(79) Bu kitap kim-in? Ben-im(*-ki) this book who-gen I-gen-ki ‘Whose book is this?’ ‘Mine’ (80) Bu kitap-lar-ın hangi-si İngilizce? 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 don’t 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 contexts without a noun, but in such cases adjectives are essentially used as nouns, hence such cases shouldn’t be treated as involving ellipsis. Such cases often require overt manifestation of this special use of adjectives, such as special morphological marking on the adjective or use of an article, as in English ‘the rich’. Bošković (2005) observes that NP languages seem much more productive in this respect. (For an explanation of this fact, see Bošković 2013, who establishes a semantic condition—i.e., a type-shifting condition—on 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 However, as in other languages, it generally requires special morphological marking (see Bošković 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-ı sat-tı, Suzan-sa yeni-yi sat-tı. 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-ı sat-tı, Suzan-sa yeni-ler-i P.-nom old car-pl-acc sell-past S.-nom-however new-pl-acc sat-tı. sell-past ‘Pelin sold the old cars while Susan sold the new ones.’ 28 See Kornfilt (1997), Göksel 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 adjectives; they don’t involve ellipsis. That this is indeed the case is confirmed by the fact that such reduced NPs don’t 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)

Yaşlı bakıcı-(lar) çocuk-lar-ı azarla-dı, ama genç-*(ler) onlar-ı 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) çok işe yara-dı ama eski-?*(ler/si) epeyce sorun new car-nom-pl very useful-past but old-nom-pl-partitive quite trouble çıkar-dı. cause-past ‘The new car(s) was/were very useful but the old one(s) was/were very troublesome.’

(iii)

Hızlı araba-(lar) biz-i etkile-di, ama yavaş-?(lar) fast car-pl-nom we-acc impress-past but slow-pl-nom hayal kırıklığına uğrat-tı. disappoint-past ‘The fast car(s) impressed us, but the slow one(s) disappointed us.’

(iv)

Güçlü insan-(lar) biz-i etkile-r, ama zayıf-(lar) strong person-pl-nom we-acc impress-aor but weak-pl-nom hayal kırıklığına uğrat-ır. disappoint-aor ‘Strong people impress us, weak ones disappoint us.’

136

bošković and şener

(82) Eski-yi/Yuvarlağ-ı/Yeşil-i isti-yor-um. old-acc/round-acc/green-acc want-pres-1sg ‘I want the old/round/green one.’ (83) Hızlı-yı 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 gerçek soykırım iddiaları-nı tartış-tı, ama Senate-nom true genocide claims-acc discuss-past but sözde-ler-i tartış-ma-dı. alleged-acc discuss-neg-past Lit. ‘The senate discussed the true claims of genocide, but it didn’t discuss the alleged ones.’ (85) *Pelin sıradan tablo-lar-ı sev-me-z ama P-nom ordinary painting-pl-acc like-neg-aor but muhteşem-ler-i sev-er. magnificent-pl-acc like-aor Lit. ‘Pelin doesn’t like ordinary paintings, but she likes magnificent ones.’

(v)

Eski-*(ler) herşey-i bil-dik-leri-ni düşün-ür-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: Yeşil 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 unavailable, 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/Nʹ-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 possessorstranding ellipsis is disallowed in Turkish, unlike in English, and a strong argument 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 Takahashi’s (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 35:465–474. Aygen, Gülşat. 2002. “Finiteness, case, and clausal architecture.” PhD diss., Harvard University. Bošković, Željko. 2005. “On the locality of left branch extraction and the structure of NP.” Studia Linguistica 59:1–45.

138

bošković and şener

Bošković, Željko. 2007. “On the clausal and NP structure of Serbo-Croatian.” In Formal Approaches to Slavic Linguistics: The Toronto Meeting, 2006, edited by Richard Compton, Magdalena Goledzinowska, and Ulyana Savchenko, 42–75. Ann Arbor: Michigan Slavic Publications. Bošković, Željko. 2008a. “What will you have, DP or NP?” In Proceedings of NELS 37, edited by Emily Elfner and Martin Walkow, 101–114. Amherst: GSLA Publications. Bošković, Željko. 2008b. “The NP/DP analysis and Slovenian.” In Proceedings of the Novi Sad Generative Syntax Workshop 1, 53–73. Novi Sad: Filozofski fakultet u Novom Sadu. Bošković, Željko. 2009. “More on the No-DP analysis of article-less languages.” Studia Linguistica 63:187–203. Bošković, Željko. 2012a. “On NPs and clauses.” In Discourse and grammar: From sentence types to lexical categories, edited by Günther Grewendorf and Thomas Ede Zimmermann, 179–245. Berlin: de Gruyter. Bošković, Željko. 2012b. “On the edge.” Ms., University of Connecticut. Bošković, Željko. 2013. “Adjectival escapades.” In Formal Approaches to Slavic Linguistics: The Indiana Meeting, 2012, edited by Steven Franks, Markus Dickinson, George Fowler, Melissa Witcombe, and Ksenia Zanon, 1–25. Ann Arbor: Michigan Slavic Publications. Bošković, Željko. in press a. “Now I’m a phase, now I’m not a phase: On the variability of phases with extraction and ellipsis.” Linguistic Inquiry. Bošković, Željko. in press b. “Phases beyond clauses.” In Nominal constructions in Slavic and beyond, edited by Lilia Schürcks, Anastasia Giannakidou, Urtzi Etxeberria and Peter Kosta. Berlin: de Gruyter. Bowers, John. 1993. “The syntax of predication.” Linguistic Inquiry 24:591–656. 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, 85–110. Georgetown: Georgetown University Press. Despić, Miloje. 2011. “Syntax in the absence of determiner phrase.” PhD diss., University of Connecticut. Despić, Miloje. 2013. “Binding and the structure of NP in Serbo-Croatian.” Linguistic Inquiry 44:239–270. Elbourne, Paul. 2005. “The semantics of ellipsis.” Interdisciplinary Studies on Information Structure 3:63–109. Gajewski, Jon. 2011. “A Little More on MOST.” Talk given at CUNY. Göksel, Aslı 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 Linguistics 46:289–299. Hankamer, Jorge, and Ivan Sag. 1976. “Deep and surface anaphora.” Linguistic Inquiry 7:391–428. 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, 176–192. Chicago: University of Chicago. Ketrez, Nihan. 2005. “Children’s 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, 125–156. 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, 163–181. 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 Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, Raffaella Zanuttini, 255–277. Washington: Georgetown University Press. Kural, Murat. 1997. “Postverbal constituents in Turkish and the Linear Correspondence Axiom.” Linguistic Inquiry 28: 498–519. 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, 299–311. 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 2:61–81. Öztürk, Balkız. 2004. “Case, Referentiality and Phrase Structure.” PhD diss., Harvard University. Partee, Barbara. 2006. “A note on Mandarin possessives, demonstratives, and definiteness.” 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, 263–280. 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 István Kenesei and Robert M. Harnish, 91–117. Amsterdam: Benjamins. Saito, Mamoru, and Keiko Murasugi. 1990. “Nʹ-deletion in Japanese.” University of Connecticut Working Papers in Linguistics 3:87–107.

140

bošković and şener

Sezer, Engin. 1991. “Issues in Turkish syntax.” PhD diss., Harvard University. Şener, Serkan. 2008. “Non-canonical Case licensing is canonical: Accusative subjects of CPs in Turkish.” Ms., University of Connecticut. Şener, Serkan. 2010. “(Non-)Peripheral Matters in Turkish Syntax.” PhD diss., University of Connecticut. Şener, Serkan and Daiko Takahashi. 2010. “Argument Ellipsis in Japanese and Turkish.” MIT Working Papers in Linguistics 61:325–339. Takahashi, Daiko. 1996. “Move-F and Pro.” MIT Working Papers in Linguistics 29:255–265. 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, 297–317. Cambridge: MIT Press. Takahashi, Daiko. 2001. “Scrambling and empty categories.” MIT Working Papers in Linguistics 41:47–58. 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, 399–413. Stanford, CA: CSLI Publications. Živanovič, Sašo. 2008. “Varieties of most: On different readings of superlative determiners.” In Studies in Formal Slavic Linguistics: Contributions from Formal Description of Slavic Languages (FDSL) 6.5, edited by. Franc Marušič and Rok Zaucer, 337–354. Berlin: Frankfurt am Main. Zribi-Hertz, Anne. 1997. “On the dual nature of the ‘possessive’ marker in Modern English.” Journal of Linguistics 33:511–537.

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 morphological form, but also with respect to their semantic function and their syntactic structure. Based on several Swiss German data corpora there is evidence 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 analysis, 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 has—due to its particular feature structure—its own syntactic 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 Häussler, Uli Lutz, Albert Ortmann, Martin Prinzhorn, Manuela Schönenberger, Susanne Trissler, and Øystein 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_007

144

studler

section 3).1 However, Standard German has only one paradigm for the definite 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. dä Maa di Frou das Chend detred man detred woman detred child c. dää Maa this man

die Frou this woman

daas Chend 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 = demonstrative 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: 27–50). 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 with—starting with Russell’s 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 information, for example anaphorically accessible lexical material, or extralinguistic information like pointing gestures and deictic words, which indicate the referent. Löbner (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 between semantic definite reference and pragmatic definite reference is reflected in the use of two different article paradigms.5 While a strong (i.e. morphologically 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 paradigms pragmatic uniqueness has to be subdivided in anaphoric and in deictic uniqueness. Thus, I shall use a detailed categorization with inherent uniqueness, 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 information. Having settled the definite determiner’s morphology and semantics, one might ask how the correlation between the semantic concept of definiteness 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 Alexiadou 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 clearcut but rather that semantic and pragmatic uniqueness need to be thought of as the endpoints of a scale (for details see Löbner 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).

definite determiners in swiss german

147

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 deixis—which 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 definite 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 discuss 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 generative grammar, which should explain the distribution of the three paradigms. The morphological and semantic distribution is shown to have a syntactic correlation 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 fulfilled—an existential condition and a uniqueness condition. Since Russell’s 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 concepts have been proposed to explain the notion of definiteness—as inclusiveness (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 unique—in 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 referent. As Löbner (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 Löbner 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 II) pragmatic uniqueness a. textual information b. deictic information

inherent-unique reference IN or FN context-unique reference SN or RN anaphoric-unique reference SN/RN shift to IN/FN deictic-unique reference SN/RN shift to IN/FN

I will sketch these options while providing the relevant examples in the following 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 threefold distinction inherent-unique, anaphoric-unique, and deictic-unique. Inherent-unique reference: By using common knowledge for the process of singling 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-

definite determiners in swiss german

149

able. The semantic uniqueness of the reference can be reached in different ways. Based on Christophersen (1939), Ebert (1971), Hawkins (1978), Löbner (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 lexical material. Neither our common knowledge nor the nature of the nominal 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 reference (8a), where the referential anchor is found in an antecedent, and autophoric textual reference (8b), where the referential anchor is found in the nominal 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 nominal 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

definite determiners in swiss german

151

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 dä, di, das ‘detfull m./f./n.’. The proximal demonstrative originates from the same word stem as the stressed form dää, die, daas (‘this m./f./n.’). 3.1 Three Morphological Forms The three determiner paradigms in Swiss German arose from the morphological forms of the original demonstrative: (11) a. a stressed form (dää, die, daas) b. a weakened, but morphological full form (dä, 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 German (Weber 1987), Lucerne German (Fischer 1989), parts of the canton Valais (Bohnenberger 1913; Wipf 1910), and parts of the canton Grisons (Meinherz 1920; Hotzenköcherle 1934). For Bernese German, the three forms are also discussed in the studies of Nübling (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 use—even 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 morphological 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 demonstrative.

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. dat. sg. nom./acc. pl. dat. pl.

de em d(i) de

d(i) de d(i) de

s em d(i) 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 openmid and open depending on the dialect ([ə], [œ], [ɛ], [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 father’s.gen hut

definite determiners in swiss german

153

3.3 The Full Article The full article has the following forms (see Fischer 1989: 227). table 2

Full articles in Lucerne German Masculine Feminine Neuter

nom./acc. sg. dat. sg. nom./acc. pl. dat. pl.

dä dëm di(e) dëne

di(e) dër di(e) dëne

das dëm di(e) dëne

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 dër instead of reduced de c. vowel, e.g.: nom./acc. sg. m.: full dä 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 Masculine Feminine Neuter

nom./acc. sg. dat. sg. nom./acc. pl. dat. pl.

dää dëmm die dënne

die dër(e) die dënne

daas dëmm die dënne

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.: dää, die, daas nom./acc.pl.m./f./n.: die b. consonant doubling: dat.sg.m./n.: dëmm dat.pl.m./f./n.: dënne

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 function—special 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.

definite determiners in swiss german

155

(15) a. De Mond schiint hüt aber weder schöön. 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 grööscht Schtück Chueche gnoo. the Tim has detred biggest piece cake taken ‘Tim has taken the biggest piece of cake.’ d. De Köbi het d Nase broche. the Jack has detred nose broken ‘Jack has broken his nose.’ e. De Schlagzüger gseet besser uus as de Sänger. detred drummer looks better out than detred singer ‘The drummer looks better than the singer.’ f. I üsere Schtrooss stoot e Chile. I de Sakrischtei hets in our street stands a church in detred sacristy has=it letscht Joor brönnt. last year burnt ‘There is a church in our street. In the sacristy, it has burnt last year.’ g. De Löi esch de Kchönig under de Tiere. detred lion is the king among detred animals ‘The lion is the king of animals.’ h. De Tiimgeischt werd be üüs 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 abovedescribed 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 (Nübling 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 lääse. 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 glääse. 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 anaphoric-unique use of determiners. The full article is prototypically anaphoricdefinite. The anaphoric use of the full article is confirmed by the grammar books and monographs of Swiss German as well as by Nübling (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 constructions (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.

definite determiners in swiss german

157

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 proximal demonstrative dää, die, daas may be used for the reference with a pointing gesture. (17) Die Toorte han i sälber 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 sälber gmacht. this cake here have=n I self made ‘I made this cake here by myself.’ b. Die Toorte han i sälber gmacht, desi (em Chüeüschrangk) 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 confirmed by all grammar books and monographs as well as in the studies of Nübling (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 dèr 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 anaphoricunique 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 Kchönig gsi. De Kchönig 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 Lüüt, won i vo früener kenne, wone aui detred people, wo.rel.pron=n I from before know live all nömme 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 respective 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.

definite determiners in swiss german

159

(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 occurrences. 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 article. Moreover, the full article is preferred in every Swiss German dialect. In the interview transcript, the reduced article is almost always used with appositive relative clauses (which are not relevant for reference fixing), but it is only marginally used with restrictive relative clauses—that 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 Lüüt, wo jez i dem Gebiet woone detred people wo.rel.pron now in the area live ‘the people who are living now in this area’

BA

b. i de Phöngkt, wo s druf aa chunt in detred points wo.rel.pron it on at come ‘in the points which matter’

AG

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) or—as Gunkel 2007 names them defining restrictive relative clauses—defines 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 söu i di Böuder do heschtöue? where should I detfull painitings here put ‘Where should I put these paintings here?’

BE

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 article’s 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 contexts 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 DRC’s [descriptive RC’s]—like phrasal compounds—may serve to create new concepts.”

definite determiners in swiss german

161

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 gäge 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 dëmm Barcelona, i säg öich, esch’s so schön gsi. well in this Barcelona I tell you is=it so nice been ‘Well, in Barcelona, I tell you, it was so nice.’ b. Dää blööd Cheib het doch vou mis Outo zu Schrott this stupid moron has effectively prt my car to scrap gfaare. driven ‘This stupid moron has effectively smashed my car.’

AG

SO

These two functions remain unmentioned in the grammars of Swiss German. The standard literature on determiners in various dialects and languages, however, presents functions that seem similar to the above-described functions of problematized and/or emotional inherent-unique use—see 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-Müller (1991), Himmelmann (1997), Vangsnes (1999, 2001), and Breu (2004). In conclusion, there are—apart from the normal distribution of the definite determiners—some displacements to record, which indicate a unidirectional language change. First, the reduced article shows up in anaphoric-unique contexts, regularly in anaphoric contexts, and with limitation in autophoric contexts. 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 morphology and semantics has an analogue in syntax. A correlation between all three components—morphology, semantics, and syntax—would require three different syntactic structures for the three paradigms. Therefore, I will present a syntactic structure within the framework of generative grammar, which accommodates 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 grammaticalization from demonstrative to article: a deictic element [DEIKT], a definiteness 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):

163

definite determiners in swiss german

demonstrative pronoun anaphoric pronoun anamnestic pronoun

DEIKT DEM/DEF DET ANAPH DEM/DEF DET ∅ DEM/DEF DET

demonstrative determiner DEIKT ∅/DEF weakly demonstrative definite determiner ANAPH ∅/DEF

DET DET

definite article affixal article noun marker

DET DET DET

∅ ∅ ∅

∅/DEF ∅/DEF ∅/∅

These positions in the grammaticalization path mirror the three paradigms in their different uses. The first three positions are reserved for the demonstrative—the 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 demonstrative position with a deictic feature for the deictic-unique use of the full article 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 represented 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 German 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 features—in 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 demonstrative.

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 & Puskás 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 NP—a 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

definite determiners in swiss german

165

feature, a feature [ANAPH] for anaphoricity. Since the asserted TopP is problematic 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; Brugè 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 position—the 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 accordance 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. Conceptforming 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, parallel 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 relative 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 Wiltschko’s analysis, namely in D, I assume that the two articles are—according to their feature structure—in 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).

definite determiners in swiss german

6

167

Conclusion

The aim of this paper was to describe the three determiner paradigms—the reduced article, the full article, and the proximal demonstrative—in Swiss German and to evince their similarities and differences with respect to their morphological form, semantic function, and syntactic structure. For this purpose, I assumed a strong correlation between morphology, semantics, and syntax insofar 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 projections over the NP—a 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 leftmost periphery to check the deictic feature. The syntactic distribution with three functional projections is motivated by the feature structure of the determiners and by common assumptions about the noun phrase structure. The final landing sites of the determiners may explain the distribution with different 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., Cambridge, MA. Alexiadou, Artemis, Paul Law, André Meinunger, and Chris Wilder. 2000. Introduction to The syntax of relative clauses, edited by Artemis Alexiadou, Paul Law, André Meinunger, and Chris Wilder, 1–51. 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 indexikalitätsmarkierenden Funktion der demonstrativen Artikelform in deutschen Konversationen.” In Sprache. Verstehen und Handeln, edited by Götz Hindelang, and Werner Zillig, 301–311. Tübingen: Niemeyer. Bernstein, Judy. 1997. “Demonstratives and reinforcers in Romance and Germanic languages.” Lingua 102:87–113. Bisle-Müller, Hansjörg. 1991. Artikelwörter im Deutschen. Semantische und pragmatische Aspekte ihrer Verwendung. Tübingen: Niemeyer. Bohnenberger, Karl. 1913. Die Mundart der deutschen Walliser im Heimattal und in den Aussenorten. Frauenfeld: Huber. Bossard, Hans. 1962. Zuger Mundartbuch. Grammatik und Wörterverzeichnisse. Ein Wegweiser zur guten Mundart. Zürich: Schweizer Spiegel Verlag. Breu, Walter. 2004. “Der definite Artikel in der obersorbischen Umgangssprache.” In Slavistische Linguistik 2002. Referate des XXIII. Konstanzer Slavistischen Arbeitstreffens Bochum, 10.-12.9.2002, edited by Marion Krause, and Christian Sappok, 9–57. München: Sagner. Brugè, Laura. 2002. “The positions of demonstratives in the extended nominal projection.” In Functional structure in DP and IP. The cartography of syntactic structures, volume 1, edited by Guglielmo Cinque, 15–53. Oxford: Oxford University Press. Brugger, Gerhard, and Martin Prinzhorn. 1995. Some properties of German determiners. ms., Los Angeles/Vienna: University of South California/Universität Wien. Christen, Helen. 1998. Dialekt im Alltag. Eine empirische Untersuchung zur lokalen Komponente heutiger schweizerdeutscher Varietäten. Tübingen: 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 75:281–304. Ebert, Karen. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem Nordfriesischen Dialekt. PhD diss., Universität Kiel.

definite determiners in swiss german

169

Eroms, Hans-Werner. 1989. “Artikelparadigma und Artikelfunktionen im Dialekt und in der Standardsprache.” In Bayerisch-östereichische Dialektforschung. Würzburger Arbeitstagung 1986, edited by Erwin Koller, Werner Wegstein, and Norbert Richard Wolf, 305–328. Würzburg: Königshausen & Neumann. Felix, Sascha. 1988. “The structure of functional categories.” Groninger Arbeiten zur Germanistischen Linguistik (GAGL) 29:37–62. Fischer, Ludwig. 1989. Luzerndeutsche Grammatik. Ein Wegweiser zur guten Mundart. Zürich: Schweizer Spiegel Verlag. Giusti, Giuliana. 2002. “The functional structure of noun phrases. A bare phrase structure approach.” In Functional structure in DP and IP. The cartography of syntactic structures, edited by Giulielmo Cinque, 54–90. Oxford: Oxford University Press. Gunkel, Lutz. 2007. “Nominalphrasen des Typs derjenige (N) + Relativsatz in den europäischen Sprachen.” in Deutsche Grammatik im europäischen Dialog. Beiträge 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 14.2:137–175. 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 Jürgen Weissenborn and Wolfgang Klein, 187–207. Amsterdam/Philadelphia: Benjamins. Hawkins, John A. 1978. Definiteness and indefiniteness. London: Croom Helm. Heck, Fabian, Gereon Müller, 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, 226–233. Sommerville, MA: Cascadilla Press. Heinrichs, Heinrich Matthias. 1954. Studien zum bestimmten Artikel in den germanischen Sprachen. Giessen: Wilhelm Schitz. Heusinger, Klaus von. 1996. “Definite Kennzeichnungen, Anaphora und Salienz.” Linguistische Berichte 163:197–226. 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 structure.” Journal of Semantics 19:1–30. Himmelmann, Nikolaus. 1997. Deiktikon, Artikel, Nominalphrase. Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Hodler, Werner. 1969. Berndeutsche Syntax. Bern: Francke. Hotzenköcherle, Rudolf. 1934. Die Mundart von Mutten. Laut- und Flexionslehre. Frauenfeld: Huber. Ihsane, Tabea, and Genoveva Puskás. 2001. “Specific is not definite.” Generative Grammar in Geneva 2:39–54.

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: Benjamins. Kolde, Gottfried. 1996. Nominaldetermination. Eine systematische und kommentierte Bibliographie unter besonderer Berücksichtigung des Deutschen, Englischen und Französischen. Tübingen: 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. Tübingen: Narr. Lehmann, Christian. 2002. Thoughts on grammaticalization. Erfurt, Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9. Löbner, Sebastian. 1985. “Definites.” Journal of Semantics 4:279–326. Löbner, Sebastian. 2011. “Concept types and determination.” Journal of Semantics 28:279–333. Lewis, David. 1979. “Scorekeeping in a language game.” In Semantics of different points of view, edited by Rainer Bäuerle, Urs Egli, and Arnim von Stechow, 172–187. Berlin: Springer. Lyons, Christopher. 1999. Definiteness. Cambridge, MA: Cambridge University Press. Marti, Werner. 1985. Berndeutsch-Grammatik. Bern: Francke. Meinherz, Paul. 1920. Die Mundart der Bündner Herrschaft. Frauenfeld: Huber. Molnar, Valeria. 1993. “Zur Pragmatik und Grammatik des TOPIK-Begriffes.” In Wortstellung und Informationsstruktur, edited by Marga Reis, 155–202. Tübingen: Niemeyer. Nübling, Damaris. 1992. Klitika im Deutschen. Schriftsprache, Umgangssprache und alemannische Dialekte. Tübingen: Narr. Ortmann, Albert. in press. “Definite article asymmetries and concept types: semantic and pragmatic uniqueness.” In Proceedings Workshop 2009, Universität Düsseldorf. 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 für Sprachwissenschaft, Universität 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, 231–250. Cambridge, MA: Cambridge University Press. Rizzi, Luigi. 1997. “The fine structure of the left periphery.” In Elements of Grammar, edited by Liliane Haegeman, 282–337. Dordrecht: Kluwer. Russell, Bertrand. 1905. “On Denoting.” Mind 14:479–493. Scheutz, Hannes. 1988. “Determinantien und Definitheitsarten im Bairischen und Stan-

definite determiners in swiss german

171

darddeutschen.” In Festschrift für Ingo Reiffenstein zum 60. Geburtstag, edited by Peter K. Stein, Andreas Weiss, and Gerold Hayer, 231–258. Göppingen: Kümmerle. Scholze, Lenka. 2007. Das grammatische System der obersorbischen Umgangssprache. Unter besonderer Berücksichtigung des Sprachkontakts. PhD diss., Konstanz: Konstanzer 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 59:320–344. Studler, Rebekka. 2004. “Voller und reduzierter Artikel in der schweizerdeutschen DP.” In Linguistische Studien im Europäischen Jahr der Sprachen. Akten des 36. Linguistischen Kolloquiums in Ljublijana 2001, edited by Stojan Bračič, Darko Čuden, Saša Podgoršek, and Vladimir Pogačnik, 625–635. Frankfurt a.M.: Peter Lang. Studler, Rebekka. 2011. Artikelparadigmen. Form, Funktion und syntaktisch-semantische Analyse von definiten Determinierern im Schweizerdeutschen. Elektronische Hochschulschrift Universität Zürich. Suter, Rudolf. 1992. Baseldeutsch-Grammatik. Basel: Merian. Vangsnes, Øystein. 1999. The identification of functional architecture. PhD diss., University of Bergen. Vangsnes, Øystein. 2001. “On noun phrase architecture, referentiality, and article systems.” Studia Linguistica 55.3:249–299. Weber, Albert. 1987. Zürichdeutsche Grammatik. Zürich: 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 21.2:100–145. Wiltschko, Martina. to appear. “Descriptive relative clauses in Austria-Bavarian German.” Canadian Journal of Linguistics.

Reduced Definite Articles with Restrictive Relative Clauses* Patricia Cabredo Hofherr 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 combine 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: 95–108, 310 for a range of Swiss German varieties) (see 1b).1 (1)

a. I hob’s Buach, des (was) da Chomsky gschriem hot, I have-detred book that (which) the Chomsky written has nit glesn (ABav) not read ‘I didn’t read the book, which Chomsky wrote.’ (non-restrictive reading of the RC only, Brugger and Prinzhorn 1996:12–13)

* 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_008

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. nothing. ‘The first book he wrote is no good.’ (Ebert 1970:169, ex 33ʹ)

(FFr)

In what follows I show that these diverging observations are due to two interrelated factors. First, the incompatibility between reduced definite determiners 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 properties 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 Vorarlberger 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 restrictive 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 relativised 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 concern 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 determiners, 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 dön (see Ebert 1970, 1971). To avoid confusion with weak definite NPs as discussed in Carlson and Sussman (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, essentially following Ebert’s seminal study (Ebert, 1971). The morphological distinction 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 nondeictic definite readings. They mark entities which are generally unique (Ebert’s 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 discoursesituation uniqueness to inherent uniqueness as proposed, e.g., in Löbner 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 didn’t come. b. Compare: That man came and that man didn’t come. This man came and this man didn’t come. 3 As Schiering (2002) points out, this distinction co-incides with Löbner’s distinction between semantic and pragmatic definiteness (Löbner 1985, 2011).

reduced definite articles with restrictive relative clauses

175

(4) Scale of uniqueness (Löbner 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 car—the 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 Löbner (2011: 280–282).

176

patricia cabredo hofherr

Dea trottl hot scho widr min reagaschiam ipackt. detfull idiot has again my umbrella taken (Cabredo Hofherr and Schaden 2012b)

(VAr)

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 determiner, 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 1998:91–93

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 morphologically conditioned (see Weiss 1998:93 for Bavarian). When examining the choice of definite determiner in Bavarian below, we therefore have chosen combinations 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 restrictive RCs has been made for Standard German and for (Austro-)Bavarian German (for Standard German see Hartmann (1980), Raffelsiefen (1987), Nübling (2005: 112), Eisenberg et al. (2006: 624), Puig-Waldmüller (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.

reduced definite articles with restrictive relative clauses

177

However, there are counterexamples for this claim for Standard German. In this language, the reduced definite determiner is present in a number of grammaticalised 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 wasn’t a problem.’ (example due to an anonymous reviewer) (9) Es dürfe nicht sein, dass Anwohner, die eine Nachtparkgebühr 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 can’t 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 für 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 Morgen, 09.12.1998

178

patricia cabredo hofherr

These counterexamples cannot be explained by the fact that reduced determiners 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 frühr 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 frühr 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 converges with the observations in Studler (2008: 95–108, 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: 95–108, 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 identifying 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.’

reduced definite articles with restrictive relative clauses

3

179

The Classification of Relative Clauses

In the literature on RCs, a distinction is drawn between appositive and restrictive RCs. However, it has long been pointed out that this binary classification is not sufficient. In particular, different authors have provided convincing evidence 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/relative 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.

180

patricia cabredo hofherr

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 description given in the RC is not true. bach, 1974

In many studies, (17) is taken to mean that a definite description with a restrictive 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.

reduced definite articles with restrictive relative clauses

181

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 relatives correspond to different information structures. The contrasting reading of restrictive relatives presented above is only one of at least two possible readings. 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 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”. 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 characterises 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). (Ebert’s 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.

182

patricia cabredo hofherr

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 identify 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 exemplified 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 weren’t). The examples in (18) are different: the house that Ina inherited is not necessarily contrasted with other houses that she didn’t inherit. Examples like (18) have an Information Structure corresponding to questionanswer 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

reduced definite articles with restrictive relative clauses

183

(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 necessary 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 identification 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 determiner 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 restriction takes the form of an additional property comparable to question-answerfocus, 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 property 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 restrictive. The inverse does not hold, however: restrictive modification need not be contrastive.

(ia–b), 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.

184

patricia cabredo hofherr

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. What’s wrong with Bill?—Oh, the woman he went out with last night was nasty to him. ((3.16) in Hawkins 1978:101) b. What’s 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 Löbner (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).

185

reduced definite articles with restrictive relative clauses

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 detred/ detfull centre of-detred universe ‘the centre of the universe’

(ABav)

b. a/ *dea Biagamasta von Hintertupf detred/ detfull major of Hintertupf ‘the mayor of Hintertupf’ c. Da Hans hot’n/ *den stärktsten 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 associated with a single entity by the function in question—the 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.

186

patricia cabredo hofherr

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 Landman (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 appositive nor restrictive, and which they call maximalising relatives. Maximalising 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 thereinsertion 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.

reduced definite articles with restrictive relative clauses

187

(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 contextual 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 verschüttet 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 reading is the object reading where the spilt champage is the one that will be

188

patricia cabredo hofherr

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) Carlson’s 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 räscht fo üseram leaba dafür brucha, zum dean / da we will the rest of our life for-it need to detfull / detred sekt zum sufa, wo dia gescht z’obad frschüttet hond. (VAr) champagne to drink rel they yesterday evening spilt have (I thank Gerhard Schaden for this example).

reduced definite articles with restrictive relative clauses

189

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) Carlson’s 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 appositives—unlike restrictives—do not stack (37).17

17 Vries (2002, 197) claims on the basis of example (i) that this is not true in the general case:

190

patricia cabredo hofherr

(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).

reduced definite articles with restrictive relative clauses

191

Notice that functional head nouns do not readily allow other determiners 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 restrictive RCs in what follows:18

18 The two types of restrictive RCs in (43a,bii) correspond to different patterns of definiteness marking in Haitian Creole as observed in Zribi-Hertz and Glaude (2007) (see Wespel (2008:131–136) 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.’ (Wespel’s ex. 84, taken from Damoiseau 2005:46) b. N + RC (functional head noun) Moun ki te envante òdinatè 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.’ (Wespel’s 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 dènye, manman m te kit pòmdetè. Pòmdetè yo, ke papa m Saturday last mother my pst cook potato potato pl rel father my te achte mache, te trè 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.’ (Wespel’s ex. 81)

192

patricia cabredo hofherr

(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 (1971:136–150) 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: 147–148):19 (44) a. i. a mensken wat suart hiar haa (detred) (FFr) ii. dön 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 construction.” [my translation] 20 Eberts original translations are (i) alle möglichen 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”

reduced definite articles with restrictive relative clauses

193

b. i. a lidj wat för fräihaaid stridj (detred) ii. dön lidj wat för fräihaaid 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) Ebert’s observation that the reduced definite article is used for intensional reference21 to a class is confirmed by the data discussed in Gunkel (2007). I will avoid Ebert’s 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/episodic, 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 Relativsätze) as follows (i) definitorische Relativsätze haben definitorischen Charakter; sie sorgen dafür, 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).

194

patricia cabredo hofherr

(Teilnehmer) die nächste 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 nächsten 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 nächsten 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 atemporal (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 måste sluta med mitt dem day-def I had-to.prs/pst close.inf with my pensionat. guesthouse ‘The day that I had to close my guesthouse’ (past, temporally anchored particular day)

(Sw)

reduced definite articles with restrictive relative clauses

195

b. Den dag jag måste 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 får hög-st-a siffra-n börjar. dem player-def rel gets highest number-det begins ‘The player that gets the highest number begins.’ (atemporal) (Gunkel 2007:233, ex 65a)

(Sw)

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 appositive: (49) a. d Bedängke, wo si händ detred doubts rel they have ‘the doubts that they have’

(Swiss German)

b. s Wüsse, wo d muesch haa detred knowledge that you have to.have ‘the knowledge that you have to have’ c. i de Phöngkt, wo s druf aa chunt in detred points where it on at comes ‘the point that is important’ (Studler 2008:105, exs 46a–c) 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))

196

patricia cabredo hofherr

Even though the RC in relativised V+N constructions is not appositive— since it is obligatory—we cannot simply classify all relativised V+N constructions 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 restrictive RCs, four types of non-contrastive restrictive RC must be distinguished, 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 fäscht 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.’

reduced definite articles with restrictive relative clauses

197

b. Da / Dea maa wo üsera bojler gflickt hot, hot gsejt, detred / detfull man rel our boiler repaired has has said dass mr d rohr usfürba lo söllan. 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 fulfill the description of the RC (contrastive restrictive reading of the RC). Functional restrictive RCs take the reduced definite determiner in the Vorarlberger dialect: (53) a. s/ ?des inschtitut wo i frühr 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):

198

patricia cabredo hofherr

(54) Amount relatives a. Da / dea tee wo i dr kanne isch isch kolt. detred / detfull tea rel in detred tea-pot is is cold ‘The tea in the tea-pot is cold.’ → no meaning difference

(VAr)

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 lüt wo rouchat sind mojschtens uzfrida. (VAr) detred/ detfull people rel smoke are mostly unhappy ‘People who smoke are mostly unhappy.’ b. d/ dia vögl wo im heabscht in süda zinad detred/ detfull birds rel in+detred autumn in south go kummad im früajor 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 näkschta 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

reduced definite articles with restrictive relative clauses

199

‘The driver who wins this race will become Olympic champion.’ (temporally anchored) Relativised V+N constructions are compatible with a reduced definite determiner even when modified by a RC: (57) a. es isch da punkt, uf dean as akut that is detred point on that it on-comes ‘that is the point that matters’

(VAr)

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 älteres 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 Frühstück. o’clock breakfast ‘In the hotel that they are staying in, breakfast is served from 7 o’clock.’

200

patricia cabredo hofherr

c. Functional head noun Karen wird im ersten Buch, das jemals über Britische Karen is in+detred first book, rel.pr ever over British Gospelmusik geschrieben wurde, erwähnt. 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 ändert 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. ‘The first book he wrote is no good.’ (Ebert 1970:169, ex 33ʹ) b. Intensional atemporal relativised DPs: reduced definite article i. a lidj wat för fräihaaid stridj (detred) ii. dön lidj wat för fräihaaid 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)

reduced definite articles with restrictive relative clauses

201

c. Establishing RCs: full definite article Anne hee det hüs 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 extraposition is not compatible with a reduced article. In the following examples, the particle ütj23 marks the right boundary of the main clause: (60) a. A köning nategt a/ dön mensken, wat detred king took-advantage detred/ detfull humans rel aarem wiar, ütj. (FFr) poor were prt b. A köning nategt *a/ dön mensken ütj, 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 reduced definite determiners on the head noun are based on a binary classification 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 ütj “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 (ütj) staying in clause-final position. Any clause that follows the particle ütj 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.

202

patricia cabredo hofherr

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 appear) that the reduced definite determiner is not compatible with focus in the RC: (62) #Da Brieaftroga wos bei UNS austrogt is imma pünktli. (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 manipulated 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

reduced definite articles with restrictive relative clauses

203

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 involve non-contrastive restrictive RCs with functional head nouns. Wiltschko proposes that RCs like the one in (66) are not restrictive, but belong to a different type, namely descriptive RCs. This assumption builds on the analysis proposed by Del Gobbo (2005) for Mandarin Chinese restrictive and descriptive 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 modifiers, while Chinese restrictive RCs can be classified as deictic or s-level modifiers. This means that Del Gobbo’s 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 property 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.

204

patricia cabredo hofherr

It has further been argued that reduced definite determiners are incompatible 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 Bach’s (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 variant 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 Partee’s 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 definite 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 Wiltschko’s proposal that the reduced determiner semantically 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.

reduced definite articles with restrictive relative clauses

205

difference between relativised DPs depending on the type of definite determiner. Wiltschko (2012) proposes the following structural difference between reduced and full determiners, where the context-variable is assumed to be associated with nP: (69) a. full definite determiners select an [nP[NP]] complement (Wiltschko’s 24a) b. reduced definite determiners select an [NP] complement (Wiltschko’s 24b) Wilschko’s 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 wänn i a nöje pfanna kouf, everytime when I a new pan buy,

(VAr)

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 ujnär 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).

206

patricia cabredo hofherr

a. lot mr noch ujnär 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 ujnär 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 definite 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) d’Maria renoviert grod des Haus, des d’Inna 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 is’n mit’m Maral los?) what is-prt with-detred Mary wrong? Dea man, mit dem’s 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.)

reduced definite articles with restrictive relative clauses

207

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 complementizer is realized only by a wh-pronoun”. Notice that Wiltschko’s 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 misclassified 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 influence 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 conclude (contra Wiltschko) that a uniform syntactic representation should be preferred. However, Wiltschko’s syntactic assumptions with respect to RCs are not crucial to her semantic analysis of reduced definite determiners. In particular, the present analysis adopts this author’s 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).

208

patricia cabredo hofherr

References Alexiadou, Artemis, Paul Law, André Meinunger, and Chris Wilder. 2000. “The syntax of relative clauses—Introduction.” In The Syntax of Relative Clauses, edited by Artemis Alexiadou, Paul Law, André Meinunger, and Chris Wilder, 1–52. Amsterdam/ 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.” Linguistic Inquiry 31:123–140. Breu, Walter. 2004. “Der definite Artikel in der obersorbischen Umgangssprache.” In Slavistische Linguistik 2002, edited by Marion Krause, and Christian Sappok, 9–57. München: Otto Sagner. Brugger, Gerhard, and Martin Prinzhorn. 1996. “Some properties of German determiners.” 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, 99–131. Amsterdam/Philadelphia: John Benjamins, Cabredo Hofherr, Patricia 2012b. “Verschmelzungsformen von Präposition und Artikel. Deutsch und Französisch kontrastiv.” In Deutsch im Sprachvergleich. Grammatische Kontraste und Konvergenzen. Jahrbuch des Instituts für deutsche Sprache 2011, edited by Lutz Gunkel, and Gisela Zifonun, 217–238. Berlin: de Gruyter. Cabredo Hofherr, Patricia, and Gerhard Schaden. 2012a. “Definite articles and restrictive 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 d’Automne de Linguistique Formelle (RALFe), Paris, November 2012. Carlson, Gregory. 1977. “Amount relatives.” Language 53:520–542. Carlson, Gregory, and Rachel Sussman. 2005. “Seemingly indefinite definites.” In Linguistic Evidence: Empirical, Theoretical, and Computational Perspectives, edited by Stephan Kepser, and Marga Reis, 71–86. Berlin: Mouton de Gruyter. Carlson, Gregory, Rachel Sussman, Natalie Klein, and Michael Tanenhaus. 2008. “Weak Definite Noun Phrases.” In Proceedings of NELS 36: 179–196. 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, 99–137. Del Gobbo, Francesca. 2005. “Chinese relative clauses: restrictive, descriptive or appos-

reduced definite articles with restrictive relative clauses

209

itive?” In Proceedings of the 30th Incontro di Gramàtica Generativa, edited by Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert, and Giuseppina Turano, 287–305. 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 Fortschritte der Transformationsgrammatik, edited by Dieter Wunderlich, 159–173. Tübingen: Max Hueber Verlag, Ebert, Karen H. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem nordfriesischen Dialekt (Fehring). Bräist/ Bredstedt: Nordfriisk Institut. Ebert, Karen. 1973. “Functions of relative clauses in reference acts.” Linguistische Berichte 23:1–11. Eisenberg, Peter, et al., editors. 2002. Duden. Die Grammatik. Mannheim, Wien, Zürich: 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 6:145–167. Grosu, Alex, and Fred Landman. 1998. “Strange relatives of the third kind.” Natural Language Semantics 6:125–170. Gunkel, Lutz. 2007. “Demonstrativa als Antezendentien von Relativsätzen.” Deutsche Sprache 35:213–238. Hartmann, Dietrich. 1980. “Über Verschmelzungen von Präposition und bestimmtem Artikel.” Zeitschrift für Dialektologie und Linguistik 47:160–183. Hawkins, John. 1978. Definiteness and Indefiniteness: A study in reference and grammaticality prediction. London: Croom Helm. Heinrichs, Heinrich Matthias. 1954. Studien zum bestimmten Artikel in den germanischen Sprachen. Giessen: Wilhelm Schmitz. von Heusinger, Klaus. 2002. “Specificity and definiteness in sentence and discourse structure.” Journal of Semantics 19:245–274. Holmberg, Anders. 1987. “The structure of NP in Swedish.” Working Papers in Scandinavian Syntax 33:1–23. Holmberg, Anders. 1993. “Two subject positions in IP in Mainland Scandinavian.” Working Papers in Scandinavian Syntax 53:29–41. 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.

210

patricia cabredo hofherr

Keenan, Ed, and Karen H. Ebert. 1973. “A note on marking transparency and opacity.” Linguistic Inquiry 4:421–424. Kempson, Ruth. 2003. “Nonrestrictive relatives and growth of Logical Form.” In Proceedings of WCCFL 22:301–314. Krifka, Manfred. “2007. Basic notions of information structure.” In Interdisciplinary Studies on Information Structure 6, edited by Caroline Féry, and Manfred Krifka, 13–56. Potsdam: Universitätsverlag. Löbner, Sebastian. 1985. “Definites.” Journal of Semantics 4:279–326. Löbner, Sebastian. 2011. “Concept Types and Determination.” Journal of Semantics 28: 279–333. Martin, Fabienne. to appear. “Decomposing (non-)restrictivity. Evaluative modifiers in post-head positions.” Lingua. Nübling, Damaris. 2005. “Von in die über in’n und ins bis im: Die Klitisierung von Präpositionen und Artikel als Grammatikbaustelle.” In Grammatikalisierung im Deutschen, edited by Torsten Leuschner, Tanja Mortelmans, and Sarah de Groodt, 103–131. Berlin: de Gruyter. Partee, Barbara. 1973. “Some transformational extensions of Montague Grammar.” Journal of Philosophical Logic 2:509–534. Puig-Waldmüller, 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.” Linguistic Analysis 17: 123–146. Schiering, René. 2002. Klitisierung von Pronomina und Artikelformen. Eine empirische Untersuchung am Beispiel des Ruhrdeutschen. Köln: Institut für Sprachwissenschaft, Universität zu Köln. 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., Universität Zürich. Vries, Mark de. 2002. The Syntax of Relativization. Ph.D. diss., U. Amsterdam. Vries, Mark de. 2006. “The syntax of appositive relativisation: On specifying coordination, false free relatives, and promotion.” Linguistic Inquiry 37:229–270. Weiss, Helmut. 1998. Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: 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 21:100–145.

reduced definite articles with restrictive relative clauses

211

Wiltschko, Martina. to appear. “Descriptive relative clauses in Austro-Bavarian German.” 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 Guéron, 265–299. Amsterdam/ Philadephia: John Benjamins.

When Determiners Abound: Implications for the Encoding of Definiteness* Marika Lekakou and Kriszta Szendrői

1

Introduction

The topic of this paper is the encoding of definiteness in Modern Greek. Modern 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 & Szendrői (2012).

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_009

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 polydefinite. 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 determiners make a semantic contribution. In other words, polydefinites are only polydefinite in the morphosyntax, not in the semantics. Despite existing differences 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 polydefinite construction and across polydefinites and monadic definites the shape of determiners is identical. This applies in all cases, i.e. across all case-numbergender 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= genitive, masc= masculine, neg= negative element, neut= neuter, nom= nominative, nonact= non-active, pl= plural, sg= singular, stan= standard.

214

lekakou and szendrői

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 usually associated with definite determiners (e.g. existence and uniqueness assertion/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 determiner 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 & Szendrői (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 & Szendrői (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 acceptable in those cases.

when determiners abound

215

(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 polydefinite 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 interpreted 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 otherwise 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

216

lekakou and szendrői

2.2 An Account in Terms of DP-Intersection In a series of papers (Lekakou & Szendrői 2007, 2009, 2012) we have highlighted a parallel between polydefinites and close appositives—a parallel also noted by Stavrou (1995); Kolliakou (2004); Panagiotidis & Marinis (2011)—and have proposed a unified account of both phenomena. Close appositives systematically 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 determiners (unlike close appositives in e.g. English). Note that here too this concerns 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 nominal 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 constructions, see Lekakou & Szendrői (2012) (especially section 3.1).

when determiners abound

(7) a. *i sikaminja i murja the bluberry treedial the blueberry treestand

217 (Stavrou 1995)

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 & Szendrői (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 external 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 attributive modification, in assuming that the interpretation of thematic identi-

7 See Lekakou & Szendrői (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 & Szendrői (2007, 2012), Szendrői (2010) for extensive discussion of this question from both a theoretical and an empirical perspective.

218

lekakou and szendrői

(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 interpretation involved within the larger constituent (polydefinite/close appositive). For detailed discussion of the operation and its restriction, see Lekakou & Szendrői (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 determiner 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.

when determiners abound

219

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 definiteness. 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.

220

lekakou and szendrői

If none of the overtly realized determiners is semantically contentful, where does definiteness reside in the polydefinite? We propose that semantic definiteness 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 identity 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 proposed 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 & Szendrői (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.’

when determiners abound

221

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 Modern Greek, and for the concomitant Def-D split? The answer is positive. Recall that proper names in Modern Greek obligatorily require the definite determiner. 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 Matushansky 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 semantic content, which is what we postulate for the Modern Greek determiner in general. Since the Modern Greek definite determiner evidently can be semantically 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.

222

lekakou and szendrői

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 polydefinites, namely with set intersection taking place among DP categories and with the fact that neither determiner is ‘real’ in the construction. We have proposed 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 semantically inert in Modern Greek. We extend our proposal to monadic definites, which also employ a covert source of definiteness and a semantically expletive D head. The three instances of the definite determiner—with monadic definites, with polydefinites, and with proper names—are 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 necessarily linked to the obligatory presence of determiners with proper names), but otherwise the definite determiner is generally semantically definite in Modern Greek. In brief, Modern Greek determiners are lexically ambiguous. This has been, implicitly or explicitly, the prevalent view in the literature on polydefinites: 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 analyses 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 treatment 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

when determiners abound

223

Take for instance the analysis of Alexiadou and Wilder (1998), where polydefinites underlyingly involve a reduced relative clause. One determiner is external 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 relative clause (which must be obligatory, as (15a) reflects an ungrammatical base order) delivers one order of the polydefinite, cf. (15b), and raising of the relativeclause-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 stipulated 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 proposed. 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’

224

lekakou and szendrői

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 regulating 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 pseudopartitives (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.

225

when determiners abound

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 illustrated 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 pseudopartitives in Modern Greek. This is indeed the case as Alexiadou et al. (2007) note: (20) To bukali *(to) aroma epese ke espase. the bottle the perfume fell and broke ‘The bottle of perfume fell down and broke.’

(container reading)

(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 genitives, (24)–(25).

226

lekakou and szendrői

(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 D’s.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 structure—we 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 discard the option that this kind of multiple definite marking is similar to that found in polydefinites/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 polydefinites, close appositives and PsP’s, 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 mat’im le-yeladim. the-movie about the-war neg suitable to-children ‘The movie about a/the war is not suitable for children.’

when determiners abound

4

227

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 doesn’t make it in polydefinites). 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 mat’im 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.

228

lekakou and szendrői

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.

when determiners abound

229

To account for the split and joint readings without postulating a crosslinguistically lexically ambiguous coordinator, Heycock & Zamparelli (op. cit) put forward an account, which is technically based on the idea that the coordinator gives rise to set product. Without going into the technicalities of the proposal, the direct consequence of this account is that when the coordinator applies to predicative categories, it will mimic the operation of set intersection. 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 coordinator applying to the predicative nominals delivers (in a technically nontrivial 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 Longobardi’s (1994:620) Italian examples for Modern Greek, Alexiadou et al. (2007: 67–68) 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 structure, whereas (18b) involves co-ordination of two DefPs. It comes as no surprise 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

230

lekakou and szendrői

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. Individuals 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 determiner, 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 definite 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 examples in (30), which contain such expressions, to the corresponding ungrammatical 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 projection, 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 Heycock & Zamparelli (2005: 244, ex 101) to account for the lack of a split reading of conjoined singular nouns in Italian.

when determiners abound

231

(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 John’s 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 matter 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 relevant data cannot be handled by semantic incorporation, along the lines of for instance Espinal & McNally (2011), who have discussed bare singulars in Catalan 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, situated higher in the extended nominal structure, must be responsible for individuation. This is compatible with our analysis, which maintains a predicate denotation of Modern Greek DPs. If, however, Num is responsible for individuation in the nominal domain (and delivers argumenthood for bare nouns), as argued by Alexopoulou & Folli, it can presumably also supply the plural individuals 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 individuals. 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.

232

lekakou and szendrői

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 predicative. 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’.

when determiners abound

233

As Alexopoulou points out, the predicative use of the proper name disallows 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 predicative use of proper names we propose that an operator is present, for concreteness, Partee’s (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 interpretation is the right one. In (32a) the meaning is that France is not Europelike. 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 opposite 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 denotation for proper names. One such proposal was put forward by Matushansky

234

lekakou and szendrői

(2009). The starting point of her analysis is the syntax and semantics of naming 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] = λx ∈ De. λR⟨e, ⟨n, t⟩⟩. R (x) (/ælis/) 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 article—the 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 collaborator; 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.

when determiners abound

235

(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 nominal phrases corresponding to a single referent. Finally, we put our analysis of definiteness in Modern Greek to the test by considering two potentially problematic data sets: one regarding the nature of the link between reference assignment 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.

236

lekakou and szendrői

References Androutsopoulou, Antonia. 1995. The licensing of adjectival modification. In Proceedings of WCCFL 14:17–31. Alexiadou, Artemis. 2006. On the cross-linguistic distribution of (in)definiteness spreading. Talk presented at the ÖLT Syntax Workshop, 8.-9.12.06, Universität Klagenfurt. Alexiadou, Artemis & Chris Wilder. 1998. Adjectival modification and multiple determiners. In Possessors, Predicates and Movement in the DP, edited by A. Alexiadou & C. Wilder, 303–332. 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 University 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, 137–173. Amsterdam: John Benjamins. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339–405. 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, 137–160. 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: 87–128. Gehrke, Berit & Marika Lekakou. 2013. How to miss your preposition. Studies in Greek Linguistics 33: 92–106. 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: Coordination, plurality, and the structure of DP. Natural Language Semantics 13: 201– 270. Higginbotham, James 1985. On Semantics. Linguistic Inquiry 16, 547–594. 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, 393–408. MIT Working Papers in Linguistics 57. Kariaeva, Natalia. 2004. Determiner spreading in Modern Greek: split-DP hypothesis. Ms. Rutgers University.

when determiners abound

237

King, Tracy Holloway and Mary Dalrymple 2004. Determiner Agreement and Noun Conjunction. Journal of Linguistics 40(1), 69–104. Kripke, Saul. 1980. Naming and necessity. Blackwell, Oxford. Kolliakou, Dimitra. 2004. Monadic definites and polydefinites: their form, meaning and use. Journal of Linguistics 40:263–333. Lazaridou-Chatzigoga, Dimitra. 2011. The distribution and interpretation of bare singular count nouns in Greek. Presented at the Weak Referentiality Workshop, Utrecht 9 September 2011. Lekakou, Marika & Kriszta Szendrői. 2007. Eliding the noun in close apposition, or Greek polydefinites revisited. UCL Working Papers 19:129–154. Lekakou, Marika & Kriszta Szendrői. 2009. Close apposition with and without noun ellipsis: an analysis of Greek polydefinites. In Proceedings of 29th Meeting of the Linguistics Department of Aristotle University of Thessaloniki, edited by Melita StavrouSifaki, Despina Papadopoulou & Maria Theodoropoulou, 151–166. Thessaloniki: Manolis Triantafyllidis Foundation. Lekakou, Marika & Kriszta Szendrői. 2012. Polydefinites in Greek: ellipsis, close apposition, and expletive determiners. Journal of Linguistics 48(1), 107–149. Longobardi, Giuseppe. 1994. Reference and proper names: a theory of N-movement in syntax and Logical Form. Linguistic Inquiry 25:609–665. Marmaridou, Sophia. 1989. Proper names in communication. Journal of Linguistics 25:355–372. Matushansky, Ora. 2009. On the linguistic complexity of proper names. Linguistics and Philosophy 31(5), 573–627. Panagiotidis, Phoevos. 2003. Empty nouns. Natural Language and Linguistic Theory 21: 381–432. Panagiotidis, Phoevos, & Thodoris Marinis. 2011. Determiner spreading as DP-predication. Studia Linguistica 65: 268–298. Partee, Barbara. 1986. Noun phrase interpretation and type-shifting principles. In Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, Jeroen Groenendijk & Martin Stockhof, 115–143. 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. D’Hulst, 329–354. 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 anniversary, Laura Brugè, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, Cecilia Poletto. Stavrou, Melita & Ianthi Tsimpli. 2009. Definite agreement in complex noun phrases.

238

lekakou and szendrői

In Proceedings of 29th Meeting of the Linguistics Department of Aristotle University of Thessaloniki, edited by Melita Stavrou-Sifaki, Despina Papadopoulou & Maria Theodoropoulou (eds.), 193–206. Thessaloniki: Manolis Triantafyllidis Foundation. Szendrői, Kriszta 2010. A flexible approach to discourse-related word order variations in the DP. Lingua 120(4): 864–878. Terzi, Arontho. 2010. “On null spatial Ps and their arguments”. Catalan Journal of Linguistics 9: 167–187. Velegrakis, Nikolaos 2011. The syntax of Greek polydefinites. Doctoral dissertation, UCL. Williams, Edwin. 1981. Argument structure and morphology. The Linguistic Review 1:81–114. 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 partitivity (without excluding maximality) in the deictic and anaphoric uses, or—in bridging uses—serves 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 “consistency 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 morphology 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 adnominal 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_010

240

kaneko

semantic and syntactic peculiarities, in comparison with demonstratives and definite determiners in other languages. Recently, Bošković (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 demonstratives 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 adnominal demonstratives include familiarity in their semantics, they lack a uniqueness 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 NPadjoined 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 remarks 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 Löbner (2011), and shows similar phenomena in Salish languages and in Malagasy. Section 4 introduces my own assumptions regarding the semantics and syntax of Japanese adnominal demonstratives, in the wake of some recent proposals on demonstratives, and assesses the validity of these assumptions with respect to some observations 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 speaker’s and hearer’s 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-company’s 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 subsidiaries.’

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.

the semantics and syntax of japanese adnominal demonstratives 243

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 Löbner’s (2011) theory of definiteness. Löbner (2011) discusses the interaction between basic noun types and determination types. He first distinguishes four basic noun types (sortal, individual, 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 uniqueness and inherent relationality respectively with the binary features [±U] and [±R], Löbner (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 enrichment to be compatible with the singular definite. To generalize such natural and forced associations between NP types and determination types, Löbner introduces a distinction between “congruent” and “incongruent” determination: 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. Löbner further points out a general 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 Löbner 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 characterized as incongruent definite markers, while the definite article the, which cannot be stressed, is essentially a congruent definite marker, although it covers 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 test”—Löbner 2011, 15), demonstratives, but not definite the, are acceptable, as in (5b,c). (5) a. deictic uses Don’t go there, the dog will bite you. (Yang and Ionin 2009, 4) b. This man is dumb and this man isn’t. (Löbner 2011, 18) c. *The man is dumb and the man isn’t. 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 [associated with discourse context reporting about a global discourse topic]; demonstrative determiners require that their descriptive content is interpreted relative to a non-default situation [immediate salient situation distinct from the discourse context] wolter 2006, 63–64

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

the semantics and syntax of japanese adnominal demonstratives 245

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) doesn’t 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 leak-prog ‘(#That/The) roof of our house is leaking.’ (ibid.)

[functional noun]

(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 determiners in English and by adnominal demonstratives and the zero form in Japanese. (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 demonstratives and English definite determiners display another crucial contrast, discussed 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 professor may publish more than one writing). In the latter case, so-no allows a non-maximal, partitive reading, as in (11b) where B’s reply meaning ‘which writing?’ 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, 597–600),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 Wolter’s (2006) generalization in (6), Schaden’s 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, 597–600) 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.

the semantics and syntax of japanese adnominal demonstratives 247

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 his—B: 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 Hata’s 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 reading 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 sono koinu(-tati) ‘so-no puppy (puppies)’, is not maximally identified with the antecedent, as confirmed by B’s 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 ⟨sorera-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 uniform, she makes use of so-no, whereas if she wants to emphasize their diversity, sore-ra-no is used.

248

kaneko

(13)

nana-hiki imasita. A: Pet shop-ni totemo kawaii koinu-ga 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 puppies—B: 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 puppies indicated by A’s gesture, as confirmed by B’s 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: I’ll 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 St’át’imcets (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).

the semantics and syntax of japanese adnominal demonstratives 249

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.] [Tákem i sqayqéqy’ecw-a] ats’x-en-ítas [ti spzúz7-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) Consultant’s comment: “they’re all seeing the same one.” But unlike Japanese demonstratives, ⟨ti … a⟩ does not necessarily convey familiarity, 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 further reidentified by the same demonstrative phrase: the first occurrence of ⟨ti smém’lhats-a⟩ is translated in English by ‘a girl’, and the second, by ‘the girl’. (16) Húy’-lhkan ptakwlh, ptákwkh-min lts7a going.to-1sg.subject tell.story tell.story-applicative here [ti smém’lhats-a] wa7 ku7 ítal láti7 [ti det girl-det imperfective report cry there det smém’lhats-a]. (idem. 108) girl-det ‘I’m 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 swúw’h-a⟩ which denote different referents, as shown by the translation. Such a reading is excluded for Japanese demonstratives. (17) Wa7 1ts7a pankúph-a [ti swúw’h-a] múta7 wa7 láku7 Be here Vancouver-det det cougar-det and be there líl’wat-a [ti swúw’h-a] t’it. (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.’ Consultant’s comment: “There are two different cougars.”

250

kaneko

Distributions more similar to Japanese adnominal demonstratives are exemplified by a Malagasy determiner ny. According to Paul (2009, 237), this determiner “uniformly encodes familiarity. […] If there is a relevant discourse referent 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 / maximal 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 akondro ‘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 demonstratives, 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 Elbourne’s (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 nonmaximal reading of ny acondro in (18) is confirmed by my Malagasy informant, Rasatranabo Razakanivony Aina Anthony.

the semantics and syntax of japanese adnominal demonstratives 251

demonstrative is computed;10 ii) Relation (signaled by R) constrains the relation 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 = ιx (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 = ιx (donkey’(x) & Localized-in(x)(g(i))) Elbourne (2008) also notes that the propositional contribution of a demonstrative 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 = ιx (books’(x) & Localized-in(x)(g(i))) Now, to make sense of the lack of uniqueness or maximality in Japanese adnominal 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 exposition.

252

kaneko

further lack their inherent quantificational force, unlike their English counterparts, as represented in (23). (23) semantic hypothesis for Japanese adnominal demonstratives [[a-no / ko-no / so-no NP]]g =λPetλx [P(x) & Relation (x)(g(i))] (24) deictic and anaphoric uses a. [[so-no hon (in (1c))]]g=λx [book’(x) & x=g(i)] [g(i) is an atom] b. [[so-no koinu (in (13)/(14))]]g=λx [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 A’s 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 definite 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 languages [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.

the semantics and syntax of japanese adnominal demonstratives 253

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 implicitly includes the same NP as the one following -no. In this respect, this analysis is similar to Barker’s (1998) for possessive partitive constructions, as in (26a), whose semantics is represented by (26b).13 The partitive analysis of -no in deictic 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 John’s (Barker 1998, 701) b. λy [tools’(y) & y ≺ John’s 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 = λx [writing’(x)(g(i))] b. [[so-no gakusee-tati (in (12))]]g = λx[students’(x) & Supervise (g(i))(x)]

We can now understand why deictic and anaphoric uses of Japanese demonstratives 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 identity with the demonstratum or the denotation of the antecedent requires less effort from the hearer than partial identity. Therefore, although the demonstrative 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. propositional 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 propositional contribution of a demonstrative phrase only indirectly, by filling an argument 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”

the semantics and syntax of japanese adnominal demonstratives 255

(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 don’t 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 don’t 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 don’t 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 casemarked 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 classifier 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 Löbner 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.’

the semantics and syntax of japanese adnominal demonstratives 257

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 constructions”: in this language, partitivity may be expressed simply by the preposition de ‘of’ + DP denoting the superset, without any quantifying expression, as in de tes macarons ‘of your macaroons’ in (35a) (see Kupferman 2004, ZribiHertz 2006, among others). Such bare partitive constructions do not pass the consistency text, as shown in (35b). But when a numeral is inserted, acceptability is improved, as in (35c). The consistency test therefore supports, rather than refutes, the partitive analysis of Japanese demonstratives. (35) a. J’ai repris deux fois de tes macarons. ‘(lit) I re-took two times of your macaroons.’ (‘I took two extra helpings of your macaroons.’) b. *J’ai repris de tes macarons, mais pas de tes macarons. ‘(lit) I re-took of your macaroons, but not of your macaroons.’ c. J’ai 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 categories, but are decomposed into three morphological parts: definite, deictic and noun. Kayne and Pollock (2010) thus decompose the English demonstrative 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 ϕ [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 existence of NumP is not established for Japanese nominal structure and furthermore, Japanese adnominal demonstratives do not seem to be specified for the number feature (see footnote 7).15 On the other hand, Bošković (2009) claims that article-less languages, like Japanese, Korean, Serbo-Croatian, etc., lack the DP projection, and that determiner-like expressions (ex. possessives, demonstratives, etc.) in these languages are syntactically adjunctive modifiers. Leu (2008) similarly argues that demonstratives include a NP-modifier component, on the basis of the morphological 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 him-pl student-gen hope ‘the hope of them students’

(Yanagida 2011, 7)

b. [kare-ra][gakusee-tati]-no kiboo him-pl student-pl -gen hope ‘the hope of them students’

(ibid.)

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⟩.

the semantics and syntax of japanese adnominal demonstratives 259

component, suggesting that “the Japanese way of saying ‘this book’ is something like ‘book of here’”. Inspired by these previous works, I assume that Japanese adnominal demonstratives lack a DP projection responsible for uniqueness / maximality,17 are decomposed into two parts forming DemP, and should be analyzed as NPadjoined 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 rosä / di rot rosä / di- rosä the rose / the red rose / this rose b. det här / der där the here ‘this one’ / the there ‘that one’

(Leu 2008, 19) [Swiss German]

(idem. 21) [colloquial Swedish]

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 Yanagida’s 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 do- in pronominal, locative, directional and adverbial forms, as in (39a–e). These parallel properties suggest that demonstrative prefixes, 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 this / thatdistal / that / which} puppy b. {ko-re / a-re / so-re / do-re} this / thatdistal / that / which c. {ko-ko / aso-ko / so-ko / do-ko} {here / theredistal / there / where} d. {ko-tira / a-tira / so-tira / {this direction / thatdistal direction / that direction / do-tira} which direction}

[adnominal]

[pronominal]

[locative]

[directional]

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, demonstratives cannot be preceded by other modifiers, as illustrated in (40a,b). Contrastively, as noted by Kamio (1977) and Miyamoto (2009), among others, Japanese demonstratives may be preceded by a modifier, as witnessed by (41a). And although it is sometimes claimed that modifiers preceding a demonstrative 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 follow a restrictive modifier clearly indicates that so-no in (41b) is not a determiner closing a nominal projection.

the semantics and syntax of japanese adnominal demonstratives 261

(40) a. *expensive this car (Bošković 2009, 195) [English] b. ova skupa kola / *skupa ova kola [Serbo-Croatian] this expensive car / expensive this car (Bošković 2009, 194) (41) a. [hahaoya-o nakusita] a-no ko mother-acc lost thatdistal child ‘that child who lost his / her mother’

(Kamio 1977, 154)

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 ‘Hanako’s’ in (42a) (which allows ellipsis of the following NP, taido ‘attitude’) is in an argument position, while ame-no ‘rainy’ in (42b) (which doesn’t 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 ‘Taro’s attitude is good, but Hanako’s isn’t.’ (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 maximality 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 contexts, native English speakers choose the indefinite article, a, while Japanese

the semantics and syntax of japanese adnominal demonstratives 263

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-no’s 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 sono in Kenkyusya’s 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 investigation, 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 combination of existential presupposition + uniqueness/maximality presupposition, 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 demonstratives, 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 misuse 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 Yan’s comment is reproduced in (47). (46) a. Zhangsan de [na san ben shu] Zhangsan depossessive that three cl book ‘(lit) Zhangsan’s that three books’

(Partee 2006, 10)

b. na san ben [Zhangsan de shu] that three cl Zhangsan depossessive book ‘(lit) that three books of Zhangsan’s’

(ibid.)

(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 doesn’t 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 display 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 preceded 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)

the semantics and syntax of japanese adnominal demonstratives 265

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 demonstratives, 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, or—in bridging uses—serves 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 captured 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 morphosyntactic 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 Language and Linguistic Theory 16: 679–717. 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, 89–113. 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, 1109–1130. Berlin: de Gruyter Mouton. Bošković, Željko. 2009. “No-DP analysis of article-less languages.” Studia Linguistica 63.2: 187–203. Elbourne, Paul. 2008. “Demonstratives as Individual Concepts.” Linguistics and Philosophy 31: 409–466. Genius English-Japanese dictionary 2006. 4th printing. Tokyo: Taisyukan. Giusti, Giuliana and Rossela Iovino. 2012. “Latin as an Articleless DP Languages.” Paper presented at Journée organisée par le projet Calcul de la référence nominal: langue avec et sans articles,Paris,March15–16.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 Heusinger, Claudia Maienborn and Paul Portner, 996–1025. Berlin: de Gruyter Mouton. Hoji, Hajime, Satoshi Kinsui, Yukinori Takubo and Ayumi Ueyama. 2003. “The Demonstratives in Modern Japanese.” In Functional Structure(s), Form and Interpretation, edited by Yen-hui Audrey Li and Andrew Simpson, 97–128. 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: 147–168. Tokyo: International Christian University. Kaneko, Yumina 1996. “Knowledge of the English article system in second language learning: To “the” or not to “the”.” Undergraduate thesis. Smith College, Northampton, MA. Kaneko, Makoto 2012. “Japanese Demonstrative so-no as a Modifier Lacking Definiteness.” In Proceedings of Sinn und Bedeutung 2, edited by Ana Aguliar Guevara, Anna Chemilovskaya and Rick Nouwen, 335–348. Cambridge. MA: MIT Working Papers in Linguistics. Kayne, Richard and Jean-Yves Pollock 2010. “Notes on French and English demonstratives.” manuscript. New York University and Université Paris Est, EA4120. available from: http://as.nyu.edu/docs/IO/2652/KaynePollock1109NotesOnFrenchAnd EnglishDemonstratives.pdf Kenkyusya’s English-Japanese Dictionary for the General Reader. 1986. 5th printing. Tokyo: Kenkyusya.

the semantics and syntax of japanese adnominal demonstratives 267 Kinsui, Satoshi, Tomoko Okazaki & Mikyon Jo 2002. “Sizisi no rekisiteki / taisyoogengogakuteki kenkyuu (A Historical and contrastive study on demonstratives—Japanese/ Korean/ Turkish).” In Taisyoo Gengogaku (Contrastive Linguistics), edited by Naoki Ogoshi, 217–247. 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-hyoogen—sore-ra-no+meesi to byoogo-meesi (Plurality in Japanese—sore-ra-no+noun and nominal reduplication).” Mind and Language 3: 35–50. 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 University. Löbner, Sebastian. 2011. “Concept Types and Determination.” Journal of Semantics 28.3: 279–333. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Matthewson, Lisa. 1999. “On the Interpretation of Wide-Scope Indefinites.” Natural Language Semantics 7: 79–134. Miyamoto, Yoichi. 2009. “On the Nominal-Internal Distributive Interpretation in Japanese.” Journal of East Asian Linguistics 18: 233–251. Nakanishi, Kimiko and Satoshi Tomioka. 2004. “Japanese Plurals are Exceptional.” Journal of East Asian Linguistics 13: 113–140. 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), 608–594. Tokyo: Sanseido Pub. Okazaki, Tomoko. 2010. Nihongo sizisi-no rekisi-teki kenkyuu (Diachronic Study of Japanese Demonstratives). Tokyo: Hitsuzi Pub. Otsuki, Fumihiko. 1889–1996. Goho sinan. Tokyo: Benseisya Pub. Partee, Barbara H. 2006. “A Note on Mandarin Possessives, Demonstratives, and Definiteness.” 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, 263–280. 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, 215–241. Amsterdam: John Benjamins. Saito, Mamoru, T.-H. Jonah Lin, and Keiko Murasugi. 2008. “Nʹ-Ellipsis and the Structure of Noun Phrases in Chinese and Japanese.” Journal of East Asian Linguistics 17: 247–271. Wolter, Lynsey. 2006. “That’s 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, 1–16. Stanford: CSLI Publications. Yang, Mei and Tania Ionin. 2009. “L2 English Articles and the Computation of Uniqueness.” Proceedings of the 3rd Conference on Generative Approaches to Language. Acquisition North America (GALANA 2008), edited by Jean Crawford et al., 325–335. Somerville, MA: Cascadilla. Proceedings Project. Zribi-Hertz, Anne. 2006. “Pour une analyse unitaire de DE partitif.” In Indéfini et prédication, edited by Francis Corblin, Sylvie Ferrando and Lucien Kupferman, 141–154. Paris: Presses de l’Université Paris-Sorbonne.

From Noun to Name: On Definiteness Marking in Modern Martinikè Anne Zribi-Hertz and Loic Jean-Louis

1

Introduction1

This article bears on two functional morphemes written l(a)- and lé and pronounced [l(a)] and [le] which have developed in Modern Martinikè as definiteness markers of a sort, alongside the better known enclitic definite determiner la,2 which is common to all French-lexifier creoles (cf. Bernabé 1983, Gadelii 1997, Lefebvre 1998, Déprez 2007, Zribi-Hertz and Glaude 2007, Alleesaib 2012, a.o.). We shall argue that la conveys “pragmatic definiteness”, as defined by Löbner (1985, 2011), while l(a)- and lé form “semantically definite” DPs denoting individual terms in the manner of definite proper names. Since French— the lexifier language—ambiguously marks semantic and pragmatic definiteness by means of the same “definite article”, and since the definite determiner in Gbe languages—a plausible substratic influence on Caribbean creoles—is restricted to pragmatic definiteness (cf. Aboh 2001), the fact that the grammar of Martinikè should have developed three distinct overt markers of definiteness is, incidentally, evidence that creolisation cannot be viewed as a “simplification” process, as claimed by McWhorter (2001). 1 Previous stages of the research which led to this article were presented orally to various audiences—the International Conference on Bare Nouns and Genericity (Université 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 Thérèse Milcent and Loïsa Paulin for their judgements on the Martinikè 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 Martinikè, which crucially behaves as a phrasal enclitic.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_011

270

zribi-hertz and jean-louis

We first summarise (section 2) Löbner’s (1985) distinction between semantic and pragmatic definiteness (refined in Löbner 2011), which provides a theoretical background for the description to follow. Section 3 summarises some relevant information on DP syntax in Martinikè, and argues that the phrasal enclitic determiner -la must be semantically characterised as a pragmatic definiteness marker, in Löbner’s (1985, 2011) sense. The next two sections present the morphological, distributional and semantic properties of l(a)-N (section 4) and lé+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 lé+NP in Martinikè instantiate a class of DPs we propose to call Names, characterised semantically as a subtype of semantic definites denoting individual concepts, and syntactically, by the occurrence of a special functional (“Name”) projection distinct from nP.

2

Semantic vs. Pragmatic Definiteness

This distinction is developed by Löbner (1985, 2011), whose theory of Definiteness 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 descriptions 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, Löbner (1985) claims that definiteness involves nonambiguity (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 situation: “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). Uniqueness 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.” (Löbner 1985: 291). Thus, unlike some other theories of definiteness, Löbner’s 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 doesn’t. Non-ambiguity of reference may be established either independently of, or in relation to, the immediate situation or context of utterance. Löbner (1985, 2011) calls the first type semantic definiteness, and the second type, pragmatic definiteness, 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.” (Löbner 1985: 299). The most basic type of semantic definites denote one-place functional concepts (FC1s in Löbner 1985, individuals of type ⟨e⟩ in Löbner 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 Löbner 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 constitute constant functional concepts, as their value does not vary with their possible arguments.” (Löbner 1985: 299). Löbner (2011) further classifies lexical nouns into four lexical types (sortal: dog, flower; individual: moon, truth; relational: brother, friend; and functional: roof, back) and explores the articulation 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 individual or function. We find that Löbner’s theory of definiteness, which straightforwardly integrates proper names, provides us with a convenient background to describe the determiner system of Martinikè, a language which overtly distinguishes a marker of pragmatic definiteness and two markers of semantic definiteness 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, Löbner 2011, Cabredo Hofherr

272

zribi-hertz and jean-louis

this volume, Studler this volume), as well as in Upper Sorbian (Breu 2004). As regards French-related creoles, Wespel (2008) has observed that the occurrence or non-occurrence of the phrase-final definite determiner la (discussed below), correlates with the semantic contrast between what Löbner calls pragmatic and semantic definiteness. In what follows we shall show that alongside its enclitic marker of pragmatic definiteness, -la, Martinikè has two overt markers 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 pragmatic and semantic definite markers of Martinikè have developed from morphemes which are historically unrelated.

3

DP Syntax in Martinikè: A Quick Preliminary Survey

3.1 Generalised Bare Nouns Bare nominals are freely licensed in Martinikè in argument positions regardless of their Mass or Count denotation: (3) a. Roch té ka tonbé anlè tèt li.4 stone ant nonp fall on head 3sg ‘Stones were falling on his head.’ b. Lyon danjéré. lion dangerous ‘Lions are dangerous.’ c. Balenn sé mamifè. whale cop mammal ‘Whales are mammals.’ (4) a. Jan manjé 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.

from noun to name

273

b. Jan enmen poul. John like chicken ‘John likes chicken(s).’ c. Té ni chyen toupatou anlè 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 Martinikè may, context allowing, be construed as “existential” or “generic” (in Carlson’s 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 Martinikè 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)

Kòkú mòn távò cè bò ò émì ná xò távò ló. 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 -là and Gbe ló 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, Mauritian—cf. Alleesaib 2012), whose substrate languages are likely to have been different from those of Martinikè. 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 -là in one area but not in the other. An open issue.

274

zribi-hertz and jean-louis

and has been extensively discussed in the specialised linguistic literature (cf. Valdman 1978, Bernabé 1983, Germain 1983, Gadelii 1997, Pinalie and Bernabé 1999, Lefebvre 1998, Zribi-Hertz and Glaude 2007, Déprez 2007, Alleesaib 2012, Glaude 2012, a.o.). Martinikè: (5) a. Mari wè an chat/chyen (nwè). Mary see a cat dog black ‘Mary saw a (black) cat/dog.’ b. Mari wè chat -la / chat nwè-a / chyen-an. Mary see cat -det / cat black-det / dog -det ‘Mary saw the/this/that {cat/black cat/dog}.’ c. Mari wè chat ta’a / chat nwè ta’a / chyen ta’a. 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 locative morpheme -là which in Standard French may co-occur with the demonstrative 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- là. pass 1sg dm book loc ‘Pass me that book.’

(Standard/Dialectal French)

b. Passe- moi le livre- là. pass 1sg df book loc ‘Pass me that book.’

(Dialectal French)

The French phrase-final, noninflected, locative -là, 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 Löbner calls pragmatic definiteness. Thus the DP chat-la in (5b), like its proposed English translations, points to a cat creature whose nonambiguous identification crucially depends either on deixis (the referent is in sight of the speaker/hearer) or on the discourse context—the referent has been previously

from noun to name

275

mentioned, or is anchored to another, previously introduced, referent. Unlike English (and French) definite DPs, Martinikè DPs headed by the -la determiner 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 Martinikè to convey such intensional readings; and -la naturally occurs in associative contexts such as (8c), where the referent of the enhanced DP is d-linked to the sentence-initial locative: (7) a. Lyon danjéré. lion dangerous ‘Lions are dangerous.’ b. Lyon-an danjéré. lion-det dangerous ‘The lion is dangerous.’ (the aforementioned or visible lion) (8) a. Lè an moun malad, yo ka kriyé doktè. when a person sick 3pl nonp call doctor ‘When someone is sick, you call {doctors/the Doctor}.’ (the ‘doctor’ category). b. Lè an moun malad, yo ka kriyé doktè-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 trouvé doktè-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 Martinikè carries a locative feature and that the interpretation of -la DPs in Martinikè characteristically involves the spatial anchoring of their referent. Note in passing that the subject DP of (9a), adapted from Krifka (1995), is ambiguous in Martinikè 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:

276

zribi-hertz and jean-louis

(9) a. Lyon-an ka gwondé lè’y pran lodè manjé. 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 gwondé {lè’y /lè yo} pran lodè manjé. 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 licensed—and (7b)—where the Kind reading is barred—is syntactically correlated with the presence (9) vs. absence (7b) of a TMA specification. Summarising: the enclitic determiner -la, in Martinikè, only triggers pragmatic definite readings, in Löbner’s sense, a restriction we may correlate to the locative feature inherited from its French lexifier, possibly reinforced by the convergent semantics of Gbe ló (see fn. 4). -la, in Martinikè, indicates that the referent of its DP is unambiguously identified via its anchoring to the discourse situation or context. 3.3 Number Lexical categories are uninflected in Martinikè: 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, Martinikè 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 pòté gato épi lèt. Mary bring cake and milk ‘Mary brought cake(s) and milk.’ The Martinikè lexicon however contains a plural marker for the DP, sé ([se]), which occurs prenominally and only in combination with the specific determiner -la, as witnessed by the minimal pair in (11):

from noun to name

277

(11) a. Mari pòté sé (dé) gato-a. Mary bring pl two cake-det ‘Mary brought the (two) cakes (discourse-linked or visible).’ b. *Mari pòté sé (dé) 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, Déprez 2007) consistent with Kayne’s (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. lèt-la ‘the letter’; sé lèt-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

278

zribi-hertz and jean-louis

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 Martinikè 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 noun

French noun

English translation

MC noun

French noun

English translation

lachas lafèt lafwa lajòl lanmè lannuit lapòt lari larivyè latjüizin lavwa

chasse fête foi geôle mer nuit porte rue rivière cuisine voix

hunting party faith jail sea night door street river kitchen voice

labitid laj lajan lanmou lèd lègzamen lékòl lil lizin lonnè lous

habitude âge argent amour aide examen école île usine honneur ours

habit age money love help exam school island factory honour bear

6 This morpheme displays allomorphy characteristic of morphological attachment in Martinikè: la (la-plaj ‘(at/to/from) the beach’), lan (lan-mizè ‘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 IndianOcean 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-

from noun to name

279

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 jòl and ous alongside lajòl ‘jail’ and lous ‘bear’ in Martinikè: (14) a. Ni an (nouvo) lajòl/lous adan vil ta’a. have a new jail/bear in town dm-det ‘There is a (new) {jail/bear} in this town.’ b. *Ni an (nouvo) jòl/ous adan vil ta’a. 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 Löbner’s 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 separately. 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 morphological 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 concepts (identifying singular terms). Under the sortal reading they may occur as bare and may combine with any type of determiner (in particular with a cardinal, or with specific -la); under the individual reading these nouns occur as prefixed in Martinikè (l(a)-N) and cannot combine with any independent determiner. The following examples illustrate the complementary distributions and interpretations of simplex nouns and their l(a)-prefixed counterparts:

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 Löbner’s sense). 9 The subtypes exemplified in (15) through (18) are mentioned by Bernabé (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 Martinikè still needs to be thoroughly checked.

280

zribi-hertz and jean-louis

Types of Places10 (15) a. Ni dé (*la-)pisin/(*la)plaj/(*la)fak/(*la)montann adan péyi have two swimming-pool/beach/college/mountain in country ta’a. 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 rèté *(la-)montann. Mary nonp live la-mountain ‘Mary lives in the mountains.’ d. Man pa alé *(la-)pisin /*(la-)montann paske man té malad. 1sg neg go la-pool / la-mountain because 1sg ant sick ‘I didn’t go to the {swimming-pool/mountains} because I was sick.’ Abstract Properties (16) a. Pwof ta’a kòmèt 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 sé 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 Martinikè 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).

from noun to name

281

b. Sé (*la-)kouyonni Jan ki mété’y konsa. se stupidity John that put-3sg that.way ‘It is John’s stupidity that landed him where he is.’ c. Sé *(la-)kouyonni ki mété 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 voté an nouvo (*la-)lwa. 3pl vote a new law ‘They passed a new law.’ b. Mari ka rèspèkté *(la-)lwa. Mary nonp respect la-law ‘Mary respects the Law.’ c. Kité ’y bat (*lan-)mizè’y. let 3sg beat misery-3sg Lit. ‘Let him/her beat up his/her misery.’ (‘Let him/her manage on his/her own.’) d. Lapli tonbé anlè nou kon *(lan-)mizè sou lé pov. rain fall on 1pl like la-misery on lé poor ‘Rain poured over us like Misery on the poor.’ Some Temporal Nouns (19) a. Jan pasé trwa (*la-)jounen épi Mari. John spend three day with Mary ‘John spent three days with Mary.’ b. Jan ka dòmi *(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 pasé dé {lannuit /(*la)jounen} épi Mari. John spend two night day with Mary ‘John spent two {nights/days} with Mary.’

282

zribi-hertz and jean-louis

Nouns Available for Instrumental Marking (20) a. Fòk ni dé (*la-)ranm /(*la)vwèl adan kannòt necessary have two oar / sail in boat ta’a. dm-det ‘Two {oars/sails} are needed on this boat.’ b. Jan ka travèsé lanmè-a a*(la-){ranm/ *(la-)vwèl}. John nonp cross sea-det instr la- oar/ la-sail Lit. ‘John is crossing the sea à-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 dé ti plaj adan vil ta’a. 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 anlè 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 sinéma ‘movies’ does not; doulè ‘pain’, penn ‘sorrow’ and mizè ‘misery’ have la-forms, but bonè ‘happiness’, dézèspwa ‘despair’ and rimò ‘remorse’ do not. Non-alternating nouns crucially occur as bare under the targeted semantic-definite reading:

from noun to name

283

(la)plaj ‘beach’ [alternating N] vs. (*la-)sinéma ‘movies’ [non-alternating N] (22) a. Mari alé *(la-)plaj. Mari go la-beach ‘Mary went to the beach.’ b. Mari alé sinéma. Mari go movies ‘Mary went to the movies.’ (la)doulè, (la)penn ‘pain’ [alternating Ns] vs. (*la)rimò ‘remorse’ [non-alternating N] (23) a. Ni dé (*la-)doulè: doulè fizik épi doulè mantal. have two pain: pain physical and pain mental Lit. ‘There are two pains: physical pain and mental pain.’ b. *(La-)doulè pran’y. la- pain take -3sg Lit. ‘Pain took hold of him/her.’ (‘(S)he was suddenly in pain.’) (24) a. Ni dé rimò: rimò konsyan épi rimò 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-)rimò pran’y. 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 Martinikè, 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-dependent reading (Löbner’s “pragmatic” reading), and a context-free reading,

284

zribi-hertz and jean-louis

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 Löbner’s semantic definites, see below—do not necessarily contain the definite article—in languages which have one. Under Löbner’s 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 (‘John’s stupidity’) or for pragmatic definiteness) > lakouyonni ‘Stupidity’ (semantic definite DP: ‘the unique individual 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) (ii) (iii)

fimen ‘smoke’ (V) lafimen ‘smoke’ (N) souèf ‘to be thirsty’ lasouèf ‘thirst’ mò ‘dead’ (predicate) lanmò ‘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.). Valdman’s 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 lajòl ‘jail’, whose initial la is agglutinated, denotes a sortal concept, it may occur as an

285

from noun to name

to tell apart prefixal l(a)- from stable l(a): thus la is stable in lajòl ‘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 dé lajòl adan vil ta’a. have two jail in town dm-det ‘There are two jails in this town.’ b. Yo voté dé (*la-)lwa jodi-a. 3pl vote two law today-det ‘They voted two laws today.’

⇒ stable la

⇒ unstable la

L(a)-N is crucially incompatible with the enclitic specific determiner -la: (27) a. Mari anlè (*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 (i–c) below with, e.g., (27b)): (i)

a. Yo ka konstrwi lajòl toupatou. 3pl nonp build jail everywhere ‘They are building jails everywhere.’ b. Adan vil ta’a, (sé) lajol-la bò 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 lajòl. Mary jail ‘Mary is in jail.’

286

zribi-hertz and jean-louis

(28) a. Mari ka rèspèkté (*la-)lwa-a. Mary nonp respect law-det ‘Mary respects the (aforementioned) law.’ b. Mari ka rèspèkté la-lwa (*-a). Mary nonp respect la-law -det ‘Mary respects the Law.’ The complementary distribution of prefixal l(a)- and enclitic -la within a DP domain is expected under Löbner’s 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 (noncongruent 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 sé in Martinikè must always cooccur with the specific determiner in its DP (cf. section 3.3): (29) a. Mari ka rèspèkté sé lwa-a. Mary nonp respect pl law-det ‘Mary respects the(se) laws.’ b. *Mari ka rèspèkté sé la-lwa-a. Mary nonp respect pl la-law-det L(a)-N is incompatible with genitive modifiers (cf. (17b), (18c)) and with restrictive relativisation (cf. (30)): (30) a. Mwen pa konnèt (*la)lwa-a ou ka palé-a. 1sg neg know la-law-det 2sg nonp talk-det ‘I don’t know the law you are talking about.’ b. (*La)jistis-la yo ka pratiké-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 everywhere.’

from noun to name

287

c. (*L)esklavaj-la ou ka palé-a, i analizé adan liv slavery-det 2sg nonp talk-det 3sg analysed in book Dèlakanpann-an. Dèlakanpann-det ‘The slavery you are talking about, it is analysed in Delakanpann’s 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 nonprefixed counterpart, which may be construed as indefinite and/or nonsingular. 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 vérité. in politics neg have truth ‘In politics there is/are no truth(s).’ b. Tout vérité pa bon pou di. every truth neg good to tell Lit. ‘Every truth isn’t good to tell.’ (‘Some truths are better left untold.’) c. I di lavérité, tout lavérité, yen ki lavérité. 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 lé présizyon épi vérité. 1sg want precision and truth ‘I want (some) precision and (some) truth.’ (32) a. Jan pa enmen lenjistis. John neg like la-injustice ‘John doesn’t like Injustice/Unfairness.’ b. Jan pa enmen enjistis. John neg like injustice Lit. ‘John doesn’t like injustices/unfairnesses.’ (‘John doesn’t like unfair decisions/situations.’)

288

zribi-hertz and jean-louis

As witnessed by these latter sets of examples, the meaning of l(a)-N is conveyed 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 Löbner’s theory of congruence, which can easily accommodate bare nominals alongside proper names as a subclass of “semantic definites”, whose inherent conceptual uniqueness needs not be overtly marked in morphology. The properties of (l)a-N in Martinikè 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 definites”—a subtype of Löbner’s “semantic definites”, as recalled above (section 4.1.2.). Such definite DPs are characteristically incompatible with restrictive modifiers and associated with a “sloppy” (variable) 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 modifier 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. b. c. d.

Mary had to go to the store. Mary had to go to the store, and so did John. Mary had to go to the new store. Mary had to go to the new store, and so did John.

Similarly, l(a)-N in Martinikè 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), Furukawa (passim) on French.

from noun to name

289

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 anlè 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 kouté 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 kouté 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 Martinikè. We thus understand in (36a,b) that Mary and John love, and England and France abolished, the same preidentified 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. Langletè aboli lesklavaj, Lafrans osi. England abolish slavery France too ‘England abolished Slavery, so did France.’ c. Ladoulè pran Mari, ladoulè pran Jan. pain take Mary pain take John Lit. ‘Pain took hold of Mary, pain took hold of John.’

290

zribi-hertz and jean-louis

Löbner’s theory on the other hand allows us to propose a unified description of l(a)-Ns hosting common nouns, in Martinikè: 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 instantiations (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 situational contexts, while we construe (36b) as meaning that England and France abolished the same legal clause allowing slavery, rather than different situational 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 Martinikè 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 modifier, 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 lotèl 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 mennè anlo étidyan uni a small town province neg nonp attract many student étranjé. foreign ‘The university of a small provincial town does not attract many foreign students.’

291

from noun to name

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 supporting 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 kriyé la “laplaj”, men ou ka mandé 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 kriyé la “plaj”, men ou ka mandé kow poutchi. Lit. ‘They call this/that place “beaches” but one wonders why.’ This supports Löbner’s assumption that nouns denoting individual concepts are introduced in the syntax as default definite DPs, contra Matushansky’s (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 Martinikè 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}, tè legzil. oh France/India land exile ‘O France/India, land of exile!’ b. {Lafwans/Lend} sé an bèl péyi. France/India cop a beautiful country ‘{France/India} is a beautiful country.’

[vocative]

[definite subject]

15 Such nouns are historically derived from French monosyllabic, feminine nouns (France, Chine, Inde, etc.).

292

zribi-hertz and jean-louis

c. Mari ka vizité {Lafwans/Lend}. Marie nonp visit France/India ‘Mary is visiting France/India.’

[definite DO]

d. Espyon ta’a ka travay ba {Lafwans/Lend}. spy dm-det nonp work for France/India ‘This spy works (is working) for France/India.’

[definite PO]

e. Ni dé {Lafwans/Lend}: have two France/India ‘There are two France(s)/India(s).’

[sortal reading]

{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}. Mary go an-France/ann-India ‘Mary is in (went to) France/India.’ g. Mari soti {an-Fwans/ann-End}. Mary return an-France/ann-India ‘Mary (has) returned from France/India.’

[locative/ Place or Goal]

[locative/ Source]

h. Yo ka kriyé péyi ta’a {“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 Martinikè 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 (Almàn ‘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 unprefixed by all Martinikè speakers, as illustrated below in (46).

from noun to name

293

‘Sicily’), a l(a)-prefixed form (Lalmàn, Lasisil), and a locative form prefixed by an(n)- (an-Sisil, ann-Almàn): – 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/ Almàn}, tè legzil! oh Sicily/ Germany land exile ‘Oh {Sicily/Germany}, land of exile!’ b. *O {Lasisil/Lalmàn}, tè legzil! (41) a. {Sisil/Almàn} sé an bèl péyi. Sicily/Germany cop a beautiful country ‘Sicily/Germany is a beautiful country.’ b. *{Lasisil/Lalmàn} sé an bèl péyi. (42) a. Ni dé {Sisil/Almàn}(…) have two Sicily/Germany ‘There are two Sicilies/Germanies.’ b. *Ni dé {Lasisil/Lalmàn} (…) – The l(a)-form occurs in governed positions—V+O (43a), P+O (43b): (43) a. Mari pa enmen {Lasisil/Lalmàn} Mary neg like la-Sicily/l-Germany ‘Mary doesn’t like {Sicily/ Germany }.’ b. Espyon ta’a ka travay ba {Lasisil/ Lalmàn}. 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-Almàn}. Mary an-{Sicily/ Germany }. ‘Mary is in {Sicily/ Germany }.’

294

zribi-hertz and jean-louis

b. Mari ay {an-Sisil/ann-Almàn}. Mary go an-{Sicily/ Germany}. ‘Mary went to {Sicily/Germany}.’ c. Mari soti {an-Sisil/ann-Almàn}. 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 kriyé péyi ta’a {“Lasisil”/“Lalmàn”}. 3pl nonp call country dm-det la-sicily/la-germany b. *Yo ka kriyé péryi ta’a {“Sisil”/“Almàn”}. Lit. ‘This country is called “Sicilies/Germanies”.’ This contrast again supports Löbner’s (2011), rather than Matushansky’s (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 Endonézi ‘Indonesia’, as illustrated in (46): (46) a. Wo {Patagoni/Endonézi}, tè legzil. oh Patagonia/Indonesia land exile ‘O Patagonia/Indonesia, land of exile!’ b. {Patagoni/Endonézi} sé an bèl péyi. Patagonia/Indonesia cop a beautiful country ‘Patagonia/Indonesia is a beautiful country.’ c. Mari ka vizité {Patagoni/Endonézi}. Marie nonp visit Patagonia/Indonesia ‘Mary is visiting Patagonia/Indonesia.’

[vocative]

[definite subject]

[definite DO]

295

from noun to name

d. Espyon ta’a ka travay ba {Patagoni/Endonézi}. [definite PO] spy dm-det nonp work for Patagonia/Indonesia ‘This spy works (is working) for Patagonia/Indonesia.’ e. Ni dé {Patagoni/Endonézi}: have two Patagonia/Indonesia ‘There are two Patagonia(s)/Indonesia(s).’

[sortal reading]

{Patagoni/Endonézi} moun rich, Patagoni/Endonézi 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/Endonézi}. Mary go Patagonia/Indonesia ‘Mary is in (went to) Patagonia/Indonesia.’

[locative: Place/Goal]

b. Mari soti {Patagoni/Endonézi}. Mary return Patagonia/Indonesia. ‘Mary (has) returned from Patagonia/Indonesia.’

[locative/ Source]

The alternation of zero morphology and l(a)- marking on definite country names echoes the alternation of zero and l(a)- marking on Martinikè common nouns discussed in section 4.1 ((La)doulè/Rimò), 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 Martinikè as a semantic definiteness marker selecting for its complement an individual term, in Löbner’s (1985, 2011) sense—either 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 definite 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

296

zribi-hertz and jean-louis

DP instantiating what we propose to call a Name—a DP denoting an individual concept of type ⟨e⟩. Correlatively, l(a)- is incompatible in its DP with the pragmatic 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

Lé-NP

5.1 The Collective Predeterminer lé We found the Martinikè determiner morpheme lé17 mentioned in two textbooks bearing on French-lexifier grammars: Bernabé (1983) and Germain (1983). Lé shares with l(a)- some properties characteristic of semantic definites, but differs from l(a)- as regards morphology (lé with common nouns is not a prefix, but a free morpheme) and interpretation (lé has a “collective” flavour while l(a)- serves to denote an atomic concept). The triplet of examples in (47), where lé+NP is contrasted with sé+NP-la and bare nouns in the argument of an existential predicate, gives a first hint of the semantics of lé: (47) a. An défilé-a, té ni ponpyé, majorèt, konséyé-minisipo. in parade-det ant have fireman majorette councillor-town ‘In the parade there were firemen, majorettes, town councillors.’ b. An défilé-a, té ni sé ponpyé-a, sé majorèt -la, sé in parade-det ant have pl fireman-det pl majorette det pl konséyé-minisipo-a … councillor town-det ‘In the parade there was: the(se) firemen, the(se) majorettes, the(se) town councillors.’ c. An défilé-a, té ni lé ponpyé, lé majorèt, lé konséyé-minisipo … ‘In the parade there was: the Firemen, the Majorettes, the Town Councillors.’ In (47a) the bare nouns receive an existential Kind reading (in Carlson’s 1977 sense), as their analogues in the English translations: each enhanced DP denotes an undefined quantity of entities respectively assigned to the ‘fire17 The grammar of lé exhibits some variation among Martinikè speakers: the lé we describe is regarded as typical of the Northern variety.

from noun to name

297

man’, ‘majorette’ and ‘town councillor’ categories. In (47b), the DPs overtly specified as plural (sé) and specific (-la) are construed as pragmatic definites: they refer to three sets of entities crucially identified via anchoring to the discourse or situation context: the aforementioned firemen/town councillors/majorettes, those of the town where the parade took place, or those in sight of the speaker and hearer. In (47c) the enhanced lé+NPs denote three collective entities (“groups”, in Landman’s 1989 terminology) whose unambiguous identification 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 instantiated in any human collectivity. The “group” intuition is consistent with the collective label Bernabé (1983) attaches to the lé determiner. Calling lé a collective determiner on the other hand fails to capture the semantically definite reading of its including DP, a property common to lé and l(a)-, which we signal above and below in our translations of lé+NP by capitalising the initial of the N head. We now present in further detail the morphosyntactic properties of the semantic definite determiner lé and of the DPs it occurs in. 5.2 Morphology Unlike instable l(a)-, lé with common nouns is not a prefix, since lexical material may be inserted between it and the following noun: (48) a. An défilé-a, lé vyé ponpyé té douvan, lé jenn ponpyé in parade-det lé old fireman- ant in-front lé young fireman té dèyè. ant in-back ‘In the parade, the Old Firemen were in front, the Young Firemen were behind.’ b. Sinéma, lé dézyèm wòl ka genyen mwens lajan ki lé movies lé minor role nonp earn less money than lé prèmyé wòl. leading role ‘In the movies, the Minor Roles earn less money than the Leading Roles.’ Further evidence of the nonprefixal status of lé in such examples is the fact that unlike l(a)-, lé 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):

298

zribi-hertz and jean-louis

(49) a. la + jistis > lajistis ‘justice’ ‘Justice’ b. la + enjistis > *la enjistis ‘injustice, unfairness’ c. la + mizè > *lamizè ‘misery’ (50) a. lé + ponpyé ‘fireman’ b. lé + étidyan ‘student’ c. lé + mèb ‘piece of furniture’

>lenjistis ‘Injustice’ >lanmizè ‘Misery’, ‘Poverty’

> lé ponpyé ‘the Fire Brigade’ > lé étidyan *>létidyan/lézétidyan ‘the Students’ > lé mèb *>len mèb ‘the Furniture’18

5.3 Distribution Lé+NP denotes an animate group in several of the above examples, but it may also freely denote inanimate groups: (51)

Adan an jaden potajé, an pwensip lé tomat kay an solèy in a garden vegetable in general lé tomato go in sun épi lé jiromon alonm. and lé 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 Bernabé (1983), lé readily combines with patronyms, as in (52a); but patronyms may also combine with the plural marker sé (and the specific definite determiner -la): lé Lakwa ‘the Lakwas’ in (52a) is construed as denoting a family unambiguously identified by the Lakwa patronym independently of the discourse context, while sé Lakwa-a in (52b) denotes a specific set of spatiotemporally anchored members of the Lakwa family (‘the aforementioned Lakwas’, or ‘the Lakwas of our town or neighbourhood’): (52) a. Lé Lakwa jadinyé dépèranfis. lé Lakwa gardener for generations ‘The Lakwas have been gardeners for generations.’

18 Mèb ‘furniture’ is a Count noun in Martinikè, as its French lexifier meuble.

from noun to name

299

b. Sé Lakwa-a jadinyé dépèranfis. pl Lakwa-det gardener for generations ‘The(se) Lakwas have been gardeners for generations.’ Lé+NP may also naturally adjoin to a plural deictic pronoun, as in (53c), in which case it contrasts with both bare nouns and sé N-la: (53) a. Zòt, étidyan, zòt té pou konprann. 2pl student 2pl ant mod understand ‘You students should have understood.’ b. Zòt, sé étidyan-an, zòt té pou konprann. 2pl pl student-det 2pl ant mod understand ‘You the students (of this place/in question), you should have understood.’ c. Zòt, lé étidyan, zòt té pou konprann. 2pl lé+étudiant 2pl ant mod understand ‘You the Students (as opposed to, e.g., the Faculty), you should have understood.’ Last but not least, lé 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. Sé gran -an mwen rapid ki sé 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, lé gran mwen rapid ki lé piti. in principle lé big less fast than lé small ‘As a rule, big ones are slower than small ones.’19

19 Interestingly, Martinikè elliptical DPs such as lé gran translate in English as bare nominals (big ones), while the article triggers a pragmatic (d-linking) effect in English (the big ones). In

300

zribi-hertz and jean-louis

5.4 Interpretation The interpretive properties of lé+NP may—for description’s sake—be decomposed into four ingredients: (i) Definite; (ii) Plural; (iii) Group; (iv) Antispecific. Below we bring out each property separately. 5.4.1 Lé+NP Is Construed as Definite The definite effect of lé is brought out by such minimal pairs as (55) (a shortened variant of (47)), where lé-NP is contrasted with a bare noun: (55) a. An défilé-a, té ni ponpyé épi majorèt. in parade-det ant have fireman and majorette ‘In the parade there were firemen and majorettes.’ b. An défilé-a, té ni lé Ponpyé épi lé Majorèt. in parade-det ant have lé fireman and lé majorette ‘In the parade there was: the Firemen and the Majorettes.’ In (55a), the bare nominals are construed as existential Kinds, in Carlson’s (1977) terms. In (55b) lé 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 Lé+NP Is Construed as Plural The fact that the pronouns coreferring with lé Rigobè ‘the Rigoberts’ in (56) surface as yo (3pl) rather than i (3sg) is evidence that lé+NP is construed as a set of atomic entities: (56) Chèrchè-a fè an étid anlè lé Rigobèz. scientist-det make a study on lé Rigobert I touvé kè {yo/*i}z ka pòté jenn-nan. 3sg find that 3pl/3sg nonp carry gene-det ‘The scientist made a study on the Rigobèz(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 sé gran-an in Martinikè, and the non-d-linked reading conveyed by lé gran.

from noun to name

301

Men lé Rigobèz pa sav ki {yo/*i}z ka pòté jenn ta’a. but lé Rigobert neg know that 3pl/3sg nonp carry gene dm-det ‘But the Rigobèz (family) are not aware that theyz carry this gene.’ 5.4.3 Lé +NP is Construed as a Group (Collective Reading) Lé 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 Marx’s classification (58): (57) a. Lé Léopar sé {mizisyen/an gwoup} Senpyè. lé Leopard it(is) musician/a group St-Pierre ‘The Leopards are {musicians/a group} from Senpyè.’ b. #Sé Léopar-la sé {mizisyen/an gwoup} Senpyè. pl Leopard-det it(is) musician/a group St-Pierre Lit. ‘These Leopards are {musicians/a group} from St-Pierre.’ (58) a. Léta ni klas laboryè épi klas profitè. State have class working and class exploiting ‘The State comprises working classes and exploiting classes.’ b. #Léta ni sé klas laboryè-a épi sé klas profitè-a. State have pl class working-det and pl class exploiting-det Lit. ‘The State comprises these working classes and these exploiting classes.’ c. Léta ni lé klas laboryè épi lé klas profitè. State have lé class working and lé class exploiting ‘The State comprises: the Working Classes and the Exploiting Classes.’ (ex. inspired by Landman’s 1989 ‘Das Kapital series’) Lé is conversely barred from the argument of a Kind-selecting predicate such as the one translating ‘to become extinct’: (59) a. Konsyèj ka disparèt. janitor nonp disappear ‘Concierges [French janitors] are about to be extinct.’ b. *Lé Konsyèj ka disparèt.

302

zribi-hertz and jean-louis

In DP positions contextually consistent with either collective or distributive readings, lé+NP unambiguously selects the collective (Group) reading. Plural sé … -la DPs are, contrastively, ambiguous between the collective and the distributive readings: (60) a. Sé lengwis -la ni plis ki 8,000 liv an-près. 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. Lé lengwis ni plis ki 8,000 liv an-près. lé 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. Sé Lakwa-a épi sé Rigobè -a pa ka antann. pl Lakwa-det and pl Rigobè-det neg nonp get.along ‘The(se) Lakwas and the(se) Rigobès don’t get along.’ [amgiguous: individuals or groups] b. Lé Lakwa épi lé Rigobè pa ka antann. ‘The Lakwa family and the Rigobè family don’t get along.’ (group reading) Lé+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. Sé Rigobè -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. *Lé Rigobè pa ka antann ant yo. – pyès (negative-polarity distributive quantifier: ‘none of the X’) in (63b): (63) a. Pyès sé Rigobè -a pa ka antann épi nouvo mè-a. none pl Rigobert-det neg nonp get.along with new mayor-det ‘None of the Rigobès gets along with the new mayor.’

from noun to name

303

b. *Pyès lé Rigobè pa ka antann épi nouvo mè-a. – chak (positive-polarity distributive quantifier: ‘each’) in (64b): (64) a. Adan défilé-a, sé ponpyé-a, yo chak té ka tchenbé an in parade-det pl fireman-det 3pl each ant nonp hold a èkstenktè. extinguisher Lit. ‘In the parade, the firemen, they were each holding a fire extinguisher.’ b. *Adan défilé-a, lé ponpyé, yo chak té ka tchenbé an èkstenktè. 5.4.4 Lé+NP Is Antispecific Lé+NP is infelicitous if the identification of the referent is dependent on the discourse context or situation. Thus, lé 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. Sé ponpyé-a ja tiré nich mouchamyèl -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. *Lé Ponpyé ja tiré nich mouchamyèl -la an fitay kay -la. Lé is incompatible in its DP with the specific determiner -la:20 20 In some idiolectal varieties of Martinikè, lé is but a morphological variant of the plural marker spelt out as sé in mainstream Martinikè: (i)

a. Lé boug-la vini. guy -det come (Sainte-Marie speaker; quoted by Bernabé 1983: 648) b. Sé boug-la vini. pl guy -det come ‘The(se) guys came/arrived.’ (mainstream Martinikè; our own example)

304

zribi-hertz and jean-louis

(66) a. Sé lengwis*(-la) ni plis ki 8,000 liv an près. pl linguist-det have more than 8,000 book in press ‘The(se) linguists have over 8,000 books in press.’ b. Lé lengwis (*-la) ni plis ki 8,000 liv an près. ‘The Linguists (collectively) have over 8,000 books in press.’ Lé+NP cannot host a cardinality marker: (67) a. Sé senk ponpyé-a ka maché adan défilé-a. pl five fireman-det nonp march in parade-det ‘The five firemen are marching in the parade.’ b. *Lé senk ponpyé ka maché adan défilé -a. With elliptical Ns, lé+NP is construed as discourse-free (antispecific), contrasting with specific sé … la, cf. (53) above. Lé+NP cannot host a restrictive relative clause, as illustrated below by (69c), contrasting with (68b) and (69b): (68) a. Pou jwé jé ta’a fòk séparé sé chif -la épi sé to play game dm-det must separate pl number -det and pl koulè -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 jwé jé ta’a fòk séparé lé chif épi lé to play game dm-det must separate lé numbers and lé koulè. face-cards ‘To play this game you must separate the Numbers from the Face Cards.’

In the grammar we are describing, sé and lé are not free variants of the plural marker, as in (i) above, but in strict complementary distribution.

from noun to name

305

(69) a. Pou jwé jé ta’a fòk séparé 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 jwé jé ta’a fòk séparé sé kat-la ki pi to play game dm-det must separate pl card-det which more piti ki 7-la di sé lézòt-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 jwé jé ta’a fòk séparé lé kat ki pi to play game dm-det must separate lé card which(are) more piti ki 7 di lé kat apati di 7. small than 7 from lé card from-7-up Lé+NP triggers what looks like narrow-scope effects in such examples as (70): (70) a. Lè difé pri, fók kriyé lé ponpyé. when fire break-out must call lé fireman ‘When (a) fire has broken out, one must call the Fire Brigade.’ b. Lè difé pri, fók kriyé sé ponpyé-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 individuals 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 Zwarts’s (2010) and Beyssade’s (2013) analysis of “weak definites” as Kind-denoting, and with Löbner’s theory of definiteness predicting the congruent compatibility of definiteness markers with individual concepts. However, all instances of lé are not associated with the seemingly “narrowscope effect” we get in (70a): thus, lé Léopar ‘the Leopards’ (a music band from

306

zribi-hertz and jean-louis

Martinique) and lé Lakwa ‘the Lacroix’s’, as definite proper names, have a constant value (wide-scope effect) in a range of situations. 5.5 Lé+NP in Denomination Predicates Lé+NP is selected over bare nouns in denomination predicates whose argument denotes a group: (71) a. Yo ka aplé group ta’a “Lé Ponpyé”. 3pl nonp call group dm-det lé fireman ‘This group is called “the Firemen”.’ b. *Yo ka aplé group ta’a “ponpyé”. Lit. ‘This group is called “firemen”.’ (72) a. *Yo ka aplé sé moun ta’a “Lé Ponpyé”. b. Yo ka aplé sé moun ta’a “ponpyé”. 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 Lé+NP: Recap The above data show that lé is a free morpheme which signals its DP as definite but antispecific (semantically definite in Löbner’s sense), and plural but group-denoting. The “collective” effect may be derived from the fact that the head noun of lé+NP (as the head noun of la-N) denotes an individual concept, rather than a sortal concept; lé+NP however contrasts with la-N in that the group-denoting individual concept contains a plurality of members. The properties of lé+NP in Martinikè echo those of French definite plural proper names such as les Etats Unis (‘the United States’):21

21 Lé actually also occurs in some Martinikè translations of plural country names, in which case it exhibits a prefixal morphology, signalled by external sandhi: (i)

Lé-zétazini sé an bèl péyi. lé-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 lé.

from noun to name

307

(73) a. *L’/Cet Etat Uni a besoin d’argent. 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 d’argent. 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 compétition 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 préfère 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-là 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/.

308 6

zribi-hertz and jean-louis

Conclusion: L(a)- and lé as Name Markers

The properties brought out for la-N and lé+NP in Martinikè lead us to identify l(a)- and lé as markers of semantic definiteness, in Löbner’s (1985, 2011) sense: they both signal the unambiguous identification of the referent of their DP as an individual concept. L(a)- and lé differ morphologically in that l(a)- is a lexically constrained prefix, while lé occurs as a free morpheme, except with country names. And they differ semantically as to number specification: l(a)-N is straightforwardly singular, while lé+NP is both plural and singular: plural since it triggers plural anaphora (licensed by the plurality of the group’s 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 lé+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 lé actually license variable readings, as especially witnessed by their compatibility with proper nouns (Lasisil, Lé Léopar). We propose that l(a)- and lé are best characterised as markers of semantic definiteness identifying individual concepts, as opposed to sortal, functional and relational concepts. In this respect, l(a)- and lé together contrast with the phrase-final enclitic determiner -la, which signals pragmatic definiteness; and la-N and lé+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 characterise l(a)- and lé as Name markers: (74) a. Mari ka respekté Lalwa. ‘Mary respects the Law.’ b. Mari pa enmen Lasisil. ‘Mary does not like Sicily.’ c. (…) fok kriyé lé Ponpyé. ‘One must call the Fire Brigade.’ d. Mari enmen lé Léopar. Marie likes the Leopards.

[la+ common n]23 [la+ proper n] [lé+ common n] [lé+ proper n]

As shown in section 4, singular Names in Martinikè may surface as l(a)-N or as bare nouns—a 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 lé+NP varies in the textbooks we consulted, but we take it upon ourselves to capitalise their initial to signal their “Name” status.

309

from noun to name

(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 marking—by lé in Martinikè (lé Ponpyé ‘the Firemen’, lé Léopar ‘the Leopards’, lé Zétazini ‘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 Martinikè—-la, l(a)- and lé—calls 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 lé+NP (Names)24

df = definiteness NbP = Number Phrase

NmP = Name Phrase nP = noun Phrase (categorial phrase)

24 The diagram in (75) leaves out country names (e.g. lé Zétazini), where lé must be anaysed as a prefix, on a par with l(a)-.

310

zribi-hertz and jean-louis

(76) Deriving (sé) NP-la (pragmatic definite DPs)

df+loc = pragmatic definiteness (spatiotemporal anchoring) Under the proposed analysis, Names are DPs characterised by (i) the occurrence 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, lé in the plural. L(a)-, being a prefix, may only take designated lexical nouns as its complement, while lé, 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 Longobardi’s (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 Rimò and Pagatoni, syntactic restriction for Sisil), the Name ends up in D° with no prefix (e.g. Rimò, Sisil, Patagoni). Lé is identified in (75) as a Name marker (like l(a)-);25 however, as a free morpheme, lé raises up to D and leaves the noun below.26 Under the analysis proposed in (75), l(a)- and lé are restricted not only

25 This analysis sheds light on the dialectal variation concerning lé mentioned in fn. 17 and 20. In the Northern variety of Martinikè we describe, lé 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 Martinikè where lé combines with -LA, it is merged in Nb°, hence stands as an individual or dialectal variant of sé (cf. section 3.3) in pragmatic definite DPs. 26 We could alternatively assume that lé, 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

from noun to name

311

to semantically definite DPs (hosting no Locative feature in D), but more precisely 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 lé+NP, and on distributivity for lé+NP, since Quantity, restrictive modification and distibutivity correlate with a sortal or relational construal of the noun. Assuming that pragmatic definiteness in Martinikè 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 movement therefore takes place in (76) but not in (75). The Martinikè data presented in this study bring empirical support to the conceptual distinction drawn by Löbner (1985, 2011) between “semantic” and “pragmatic” definiteness. Like the West-Germanic dialects reported to distinguish “strong” (morphologically full) from “weak” (morphologically reduced) definite articles respectively conveying these two types of definite interpretations (cf. Ebert 1970, Löbner 1985, 2011, Schiering 2002, Studler 2008, Cabredo Hofherr this vol), the morphology of Martinikè distinguishes pragmatic from semantic definiteness. However, the morphemes l(a)- and lé of Martinikè 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 lé+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 characterised 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 visible—as by l(a)- or lé in Martinikè— but needs not be, as witnessed by the many examples of bare Names in various languages—including English, French and Martinikè itself. While bare Names always seem construed as semantically singular, at least in the languages under discussion, plural Names seem to require some overt functional marking (English the, French les, Martinikè lé), a restriction reflecting their marked nature correlating with a “collective” effect. The fact that both l(a)-N and lé+NP are selected over their bare counterparts in denomination contexts (cf. (38), (71)) supports Löbner’s (2011) analysis of proper names, once merged in syntax, as default definite DPs, rather than Matushansky’s (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 lé, beyond their different surface positions in the structure.

312

zribi-hertz and jean-louis

The data presented in this study show that the morphosyntax of definiteness is radically different in Martinikè and in French, the lexifier language. While French uses a single “definite article” to convey pragmatic and semantic definiteness—a reminder of this morpheme’s pronominal origin—,27 Martinikè has developed a marker of pragmatic definiteness—-la—historically derived from a French deictic locative, and two Name markers—l(a)- and lé— which have retained the uniqueness presupposition of their lexifier (the French definite article) but not its pronoun-inherited “anaphoric” feature. Should French and Martinikè be equally listed in typological charts as “Languages With Definite Determiners”? We let typologists ponder over this practical issue.

References Abbott, Barbara. 1999. “Support for a unique theory of definite descriptions.” In Proceedings of SALT 9, edited by Tanya Matthews and Devon Strolovitch, 1–15. Amherst University: GLSA. Aboh, Enoch Oladé. 2001. “La morphosyntaxe de la périphérie gauche nominale.” Recherches Linguistiques de Vincennes 31: 9–26. Aguilar, Ana, and Joost Zwarts, 2010. “Weak definites and reference to kinds.” In Proceedings of SALT 20, edited by Nan Li and David Lutz, 179–196. Ithaca, NY: Cornell University. Alleesaib, Muhsina. 2012. “DP syntax in Mauritian.” Doctoral dissertation, Université Paris-8. Baker, Philip. 1984. “Agglutinated French articles in Creole French: their evolutionary significance.” Te Reo 27: 89–129 Barker, Chris. 2005. “Possessive weak definites.” In Possessives and beyond: semantics and syntax, edited by Ji-Yung Kim, Yury Lander and Barbara Partee, 89–113. University of Massachusetts: BookSurge Publishing. Bernabé, Jean. 1983. Fondal-natal. Paris: L’Harmattan 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: 93–102.

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.

from noun to name

313

Breu, Walter, 2004. “Der definite Artikel in der obersorbischen Umgangssprache.” In Slavistische Linguistik 2002, edited by Marion Krause and Christian Sappok, 9–57. München: Otto Sagner. Cabredo Hofherr, Patricia. This volume. “Reduced definite articles with restrictive relative clauses”, 172–211. Carlson, Greg. 1977. “A unified analysis of the English bare plural.” Linguistics and Philosophy 1: 413–457. 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. Cervinka, Bernadette. 1990. “Quelques remarques à propos du morphème la- dans les couples du type tè/latè, lin/lalin, mizè/lanmizè, etc. (créole martiniquais).” Master’s thesis, Université de Pointe-à-Pitre, Guadeloupe. (Published in Espace Créole 9 http://www.potomitan.info/travaux/espacecreole/morpheme.htm) Chaudenson, Robert. 2007. “Le substrat dans la créolisation: mythes et réalités.” In Grammaires créoles et grammaire comparative, edited by Karl Gadelii and Anne Zribi-Hertz, 27–48. Saint-Denis, France: Presses Universitaires de Vincennes. Confiant, Raphaël. 2007. Dictionnaire du créole martiniquais. http://www.potomitan .info/dictionnaire/ Corblin, Francis. 1987. Indéfini, défini et démonstratif. Constructions linguistiques de la référence. Geneva: Droz. Corblin, Francis. 2001. “Défini et génitif: le cas des génitifs défectifs.” In Cahiers JeanClaude Milner, edited by Jean-Marie Marandin, 19–54. Paris: Verdier. Corblin, Francis. 2011. “Des définis para-intensionnels: être à l’hôpital, aller à l’école.” Langue française 171: 55–75. Corblin, Francis. 2013. “Locus et telos: aller à l’école, être à 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. Université de Poitiers. Déprez, Viviane. 2007. “Probing the structuring role of grammaticalization: nominal constituents in French-lexifier creoles.” Journal of Pidgin and Creole Languages 22-2: 263–307. Ebert, Karen. 1970. “Zwei Formen des bestimmten Artikels.” In Probleme und Fortschritte der Transformationsgrammatik, edited by Dieter Wunderlich, 159–173. Tübingen: Max Hueber Verlag. Ebert, Karen. 1971. Referenz, Sprechsituation und die bestimmten Artikel in einem nordfriesischen Dialekt (Fehring). Bräist/ Bredstedt: Nordfriisk Institut. Flaux, Nelly. 1992. “Les syntagmes nominaux du type le fils d’un paysan: référence définie ou indéfinie?” Le français moderne 1: 113–140. Flaux, Nelly. 1993. “Les syntagmes nominaux du type le fils d’un paysan: référence définie ou indéfinie?” Le français moderne 2: 23–45.

314

zribi-hertz and jean-louis

Furukawa, Naoyo. 1986. L’article et le problème de la référence en français. Tokyo: France Tosho. Furukawa, Naoyo. 2010a. “L’article défini et le problème dit de l’unicité: quantité ou qualité?” Bulletin d’Études de Linguistique Française 44: 65–82. Furukawa, Naoyo. 2010b. “Article défini, son emploi ‘intensionnel’ et énoncé tautologique.” Bulletin d’études françaises 41: 51–71. Gadelii, Karl. 1997. Lesser Antillean French creole and Universal Grammar. Göteborg University: Gothenburg Monographs in Linguistics 15. Germain, Robert. 1983. Grammaire créole. Paris: L’Harmattan. Glaude, Herby. 2012. “Aspects de la syntaxe de l’haïtien.” Doctoral dissertation, Université Paris-8. Hawkins, John. 1978. Definiteness and indefiniteness: a study in reference and grammaticality. 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: 273–324. 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, 1–124. Chicago: The University of Chicago Press. Landman, Fred. 1989. “Groups, I.” Linguistics and Philosophy 12: 559–605 Lefebvre, Claire. 1998. Creole genesis and the acquisition of grammar. Cambridge UK: Cambridge University Press. Löbner, Sebastian. 1985. “Definites.” Journal of Semantics 4: 279–326. Löbner, Sebastian. 2011. “Concept types and determination.” Journal of Semantics 28: 279–333. Longobardi, Giuseppe. 1994. “Reference and proper names. A theory of N-Movement in syntax and Logical Form.” Linguistic Inquiry 25: 609–655. Lyons, Chris. 2000. Definiteness. Cambridge: CUP. Matushansky, Ora. 2008. “On the linguistic complexity of proper names.” Linguistics and Philosophy 31: 573–627. McWhorter, John. 2001. “The world’s simplest grammars are creole grammars.” Linguistic Typology 5–3/4: 125–156. Milner Jean-Claude. 1982. Ordres et raisons de langue. Paris: Seuil. Ndayiragidje, Juvénal. 1984. “La source du déterminant agglutiné en créole haïtien.” The Canadian Journal of Linguistics 34: 313–317. Pinalie, Pierre and Jean Bernabé. 1999. Grammaire du créole martiniquais en 50 leçons. Paris: L’Harmattan.

from noun to name

315

Poesio, Massimo. 1994. “Weak definites.” In Proceedings of SALT 4, edited by Mandy Harvey and Lynn Santelmann, 282–299. Ithaca, NY: Cornell: University. Roberts, Craige. 2003. “Uniqueness in definite noun phrases.” Linguistics and Philosophy 26: 287–350. Russell, Bertrand. 1905. “On denoting.” Mind 14: 479–493. Strawson, Paul. 1950. “On referring.” Mind 59: 320–344. Studler, Rebekka, 2008. “Artikelparadigmen Form, Funktion und syntaktisch-semantische Analyse von definiten Determinierern im Schweizerdeutschen.” Ph.D. diss., Universität Zürich. Studler, Rebekka. This volume. “The morphology, syntax and semantics of definite determiners in Swiss German”, 143–171. Valdman, Albert. 1978. Le créole: 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 Guéron, 265–298. 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 languages (SL) has to take into consideration not only lexical signs but also the constructions currently referred to as classifier constructions in the literature (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 reference—and particularly the expression of definiteness and specificity—crucially 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 nonconventional 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 discourse 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_012

reference resolution in french sign language (lsf)

2

317

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 formal 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 fundamental 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 reception 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 relations. This contrasts with other functions labelled topographical or descriptive. 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 arbitrarily assigning linguistic entities to a location in the signing space, called the locus, allowing, for example, the identification of verbal arguments. The locations can then serve as pronouns, allowing renewed reference later in the discourse. 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 nonarbitrary 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 advocated 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 person 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 dominant 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 (situational 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 signer’s 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

reference resolution in french sign language (lsf)

319

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 “conceptual 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. Liddell’s analysis relies on Fauconnier’s (1985) theory of mental spaces, which gradually developed into the conceptual 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 generally, the set of directionality phenomena must be understood as pure pointing gestures towards spatially-grounded conceptual entities that are always “referents”, 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. Barberà (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 view”5 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 character 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 representation 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 location.” (Winston 1995: 90). 5 For “referential locus”, following Lillo-Martin & Klima (1991).

320

garcia and sallandre

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 Barberà: “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” (Barberà 2012: 61–62). 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 dispute—the 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 description of nominal referents. The characteristic property of these constructions is that each component (of the same parametric type as those used for lexical 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).

reference resolution in french sign language (lsf)

321

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, conventional and lexically specified components. Notice that these analyses are primarily 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 constructions using them, they quickly became known as classifiers or classifier 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 contructions, 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 Matteo’s 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).

322

figure 2

garcia and sallandre

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 presented, 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.

reference resolution in french sign language (lsf)

323

Liddell’s proposal (1995, 2003) mirrors his analysis of what he calls “indicating verbs” (commonly known as “directional”): he proposes to acknowledge 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 Liddell’s view these two types of constructions offer the possibility for the different articulators 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 formal 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 gesture—consequently beyond the scope of linguistic analysis. The same appeal to so-called gesture is Liddell’s solution to the other disputed 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 characters 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

324

garcia and sallandre

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 phenomena. 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 narrator’s selective reproduction, of the reported action. Winston’s (1991) and Metzger’s (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 categories: surrogates combined with “linguistic signs”, and pure surrogates (“without 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.” Liddell 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 combination 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 constructions 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.

reference resolution in french sign language (lsf)

325

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 referent 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 descriptions 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 character’s 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 linguistically. 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 discourse in LSF, and 30% of prescriptive discourse. According to Antinoro Pizzuto et al (2008), units of non-conventional type constitute the major mode of anaphoric expression in LSF, LIS and ASL (80–95 %). 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.

326 3

garcia and sallandre

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 l’énonciation (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 (énonciation, 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 produced. 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 Jakobson’s well-known notion of “shifter.” The Situation Domain both links and linguistically co-determines the signer/speaker and his addressee: the very act of uttering simultaneously establishes them as 1st and 2nd person. This relationship 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 linguistic, cannot be reduced to the real (physical) participants in the interaction. A secondary Situation Domain can be opened up in an utterance, for example through the various mechanisms of reported discourse in which a character 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 énonciativistes call: plan de l’énonciation (> Situation Domain) and plan de l’ énoncé (< 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.

reference resolution in french sign language (lsf)

327

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 understanding 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 interrelational 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, unités 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.

328

garcia and sallandre

a distance; personal transfer (PT, transfert personnel) involves the signer literally becoming the denoted entity, thereby showing this entity’s 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 horse’s tail), situational transfer (the horse jumps over the fence) and personal transfer (galloping horse) (corpus ls-colin, cuxac et al 2002)

These three structures are considered as the structural result (on a phylogenetic 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 (iconicité d’image) 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 communication—whereas hearing communication is bi-modal—SLs 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 semiology 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 argumentation 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 indispensable for our purposes.

reference resolution in french sign language (lsf)

329

A condition for initiating a linguistic interaction in SL is setting the signer’s gaze, which is also a clue to his/her semiological intent (saying vs. saying while showing). The intent to say while showing requires the signer’s gaze to be detached from the addressee (the co-enunciator), thereby signalling the temporary 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 emergence (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 embodying the young teacher he used to be, he uses the LU [TEACH]. In its conventional 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.

330

figure 6

garcia and sallandre

Semi-transfer ‘TEACH awkwardly’ (corpus creagest, garcia & l’huillier 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 teaching’s beneficiary. However, in this case, the LU is identified solely by the manual handshapes (the flattened O), while movement and orientation are disorganized and non-symmetrical, non-conventional, thereby conveying 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 personnel, 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 illustrative 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.

reference resolution in french sign language (lsf)

331

interacts with the addressee). In fact, during the personal transfer of himself as a young teacher (LU handshape [TEACH]), teaching awkwardly (hand movement and orientation), the signer’s 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-deprecation—his 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 phenomena 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 handshapes” are only one component of transfer structures. Far from “categorising 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: 97–130).

332

garcia and sallandre

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 signer’s 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 constructions 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 discrepancies 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) Combination of gestural and linguistic or pure gestural

The Semiological Model

Outside the Illustrative intent

Under the Illustrative intent

Personal pronouns Issue: 1st /2nd /3rd person

Personal pronouns 1st / non-1st person

Personal pronouns

[Personal pronouns in PT in reported discourse]

Directional verbs

Indicating verbs

Directional verbs

[Directional verbs in PT in reported discourse]

20 We employ here Peirce’s (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).

reference resolution in french sign language (lsf)

Literature

Loci

Liddell (2003) Combination of gestural and linguistic or pure gestural Tokens

Classifier Handshapes Depicting Handshapes

333

The Semiological Model

Outside the Illustrative intent

Under the Illustrative intent

Activation of points in space to indicate referents = diagrammatical iconicity Proforms (handshapes) (present in all types of transfer) TSS (transfer of size and shape) ST (situational transfer)

Classifier constructions (and diverse terminologies, see Emmorey 2003)

Depicting verbs (topographical “diagrammatical space”)

Role shifts Constructed actions (including more or less “constructed dialogues”)

Surrogates with signing

Semi-transfer

Surrogates without signing

PT (personal transfer) (including PTs in reported discourse) Personal transfer Double transfer (PT + ST)

Depicting verbs in a “viewer space”: a mix of depicting verb and surrogate blend (Dudis 2004: “multiple references”)

334

garcia and sallandre

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 typically 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 conflicting 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.

reference resolution in french sign language (lsf)

335

4.1 Formal Approaches In accordance with the underlying conceptualisation of space (2.1), descriptions conducted from a formal perspective centre on the linear distribution of constituents. The first study, focusing on nominal determination in ASL (Zimmer 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, characterised 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” structure, 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 locative adverbial. Indefiniteness, according to this author, is formally marked by very similar pointing, also prenominal, but characterised by upward orientation as opposed to orientation to a reference point in the space. According to MacLaughlin, the DET signals the status of the referent as definite/indefinite and as specific/non-specific. MacLaughlin relies of the following definitions: 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 referent is identifiable by the adressee, which implies that it is also specific. More generally, in MacLaughlin’s 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 movement—the more marked this movement, the less identifiable the referent ([SOMETHING/ ONE]). Note that, although she argues on this basis that ASL has a dual definite/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).

336

garcia and sallandre

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 associated with movement, as in ASL, and it is not linked to a degree of referential 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 discourse 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 MacLaughlin’s analysis of ASL, in HKSL the various markers are described as optional. Barberà (2012), a more recent study of Catalan SL (LSC) from a formal semantics 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 formal analyses on this issue, but considers the discourse level, adopting for this purpose the framework of Discourse Representation Theory (Kamp and Reyle 1993). Barberà finds no formal binary definite/indefinite marking in LSC, and therefore calls into question MacLaughlin’s and Tang and Sze’s 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 opposition 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 specificity—the lower frontal space is specific while the upper front space is non-specific. According to Barberà, 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 (Barberà 2012: 243). Finally, Barberà highlights the association between “strong spatial marking” and specificity in the lower zone of the frontal space, and “weak spatial marking” 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

reference resolution in french sign language (lsf)

337

and no directed gaze. For Barberà, 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 specifiable 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 reference—animate being, object, place, time, or abstract concept—can 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 referents, 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 definiteness, corresponding to Lyons’s (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 transition movement, the direction in which the index finger ‘points’ being irrelevant” (EngbergPedersen 1993: 119).

338

garcia and sallandre

being referred to without however giving him any locative (or qualitative) information about it (Lyons 1977: 654)” (Engberg-Pedersenn 1993: 119). Nevertheless, Engberg-Pedersen stresses that this encoding is not obligatory for the expression 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 Liddell’s, there is a deictic basis for locus-marked determiners (as for all directed pointing signs). Thus, it is due to its semiological 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 universe (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 definiteness 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 referent’s 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

Rinfret’s (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 question 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 systematic study of the various strategies of spatial association of a noun in LSQ, covering manual strategies (pointing, direct localisation of a sign) and nonmanual 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 marking. 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.

reference resolution in french sign language (lsf)

339

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 contextual interpretation of grammatical markers.”23 Her choice is supported by a major weakness of formal approaches to reference resolution in SLs—the 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” without 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 provides a better explanatory scope compared to exclusively syntactic or semantic theories of definiteness. Ariel’s notion of accessibility is a gradient concept, which refers to the degree of activation of information about a referent in shortterm and long-term memory. This activation, according to Ariel, depends on a variety of factors, including “salience.” Rinfret adopts Epstein’s (2002) theoretical framework for two main reasons. First, Epsteins’s 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, Epstein’s (2002) proposal is interesting in two respects. First, Epstein extends Ariel’s 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 encyclopaedic 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 Fauconnier’s (1985) theory of mental spaces as a discourse model in which she elaborates her accessibility analysis. She refers to Liddell’s application of Fauconnier’s theory in the

23 “une interprétation contextuelle des marques grammaticales” (Rinfret 2008: 123). 24 This idea is already present in Ariel’s notion of discourse salience (see also EngbergPedersen’s discussion of higher relevance of a referent in discourse). Epstein’s contribution is in the systematic consideration of this aspect and the fact that (in)definiteness cannot be treated solely in relation to referential function.

340

garcia and sallandre

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 accessibility, is that spatial marking, which is unique to SLs, facilitates accessibility of discourse referents, as the relative permanence of the spatial association facilitates the recoverability of these referents. Rinfret’s results overall confirm her hypotheses. Aside from her conclusion that definiteness is not grammatically encoded in LSQ, and that any type of referent 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, Rinfret’s conclusions confirm the important role of spatial marking as a signal of referential accessibility, while the absence of spatial association indicates a highly accessible referent. Among the various strategies for spatial association, manual 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 partially 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 referent, just as spatial association through pointing is not the sole condition for a specific reading, given that contextual information may suffice. Thus, according to Rinfret’s 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 phenomena. 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 accordance with Barberà, 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-

reference resolution in french sign language (lsf)

341

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 disputed in the literature, despite their frequency in discourse, and their constant intertwining with LUs. Several authors have mentioned this choice explicitly. Barberà (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 established or not […]. However, role shift constructions are outside the scope of this dissertation. barberà 2012: 220, note 5

Engberg-Pedersen considers at least some of these constructions.26 Nevertheless, 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 Epstein’s 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 Jakobson’s 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.

342

garcia and sallandre

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-Pedersen’s position seems to be that, like Liddell, she considers “depicting verbs”, manual pointing and “indicating verbs” as only partially grammatical. For Engberg-Pedersen, therefore, only non-locus-modified manual pointing is “fully grammatical.” In this perspective, anything that is considered “gestural” is not seen as referential. On the basis of different justifications simply given in passing, Rinfret similarly 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 Winston’s (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 relation 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: 294–306) 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 description 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 “L’association spatiale (…) est faite de façon complexe lorsque le signeur situe un élément dans l’ espace à l’ aide d’ un verbe ou d’ un classificateur, c’est-à-dire lorsqu’il établit, dans l’ espace, une relation ou une action avec une entité avant même de l’avoir nommée” (Rinfret 2008:90–91). [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.]

reference resolution in french sign language (lsf)

343

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 signer’s 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 dialogues between deaf adults on various topics of daily life (career, educational background, areas of expertise in science, technology, art, and others). Consequently, 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 concentrate 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 referents (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.

344

garcia and sallandre

To facilitate the understanding of some examples, we include below (Figure 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 nº1 & 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.

reference resolution in french sign language (lsf)

345

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-Perdersen’s analysis of DSL, Rinfret’s analysis of LSQ, but also Barberà’s analysis of LSC (see section 3).

346

garcia and sallandre

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 similar 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 transfer 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 labialisation 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 primarily 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).

reference resolution in french sign language (lsf)

figure 9

347

Pointing (left) then LU [HORSE] (right) (corpus ls-colin, cuxac et al 2002)

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, it’s 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 là (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 signer’s 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 LU’s latent iconicity, which is irrelevant so long as the UL is used as such, that is without any “showing” intent.

348

garcia and sallandre

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 pointing—to 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

5.2.3

Spatialised LU [HOME] (left) followed by pointing (right) (corpus ls-colin, cuxac et al 2002)

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

reference resolution in french sign language (lsf)

349

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) (corpus ls-colin, cuxac et al 2002)

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”.

350

garcia and sallandre

The signer opens the story with a very detailed description of a long wooden fence. As is the rule in this type of construction—characteristic of the sayingwithout-showing mode of telling (see section 3.)—the LU WOOD is accompanied 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 specifying (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) (corpus ls-colin, cuxac et al 2002)

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.

reference resolution in french sign language (lsf)

351

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 (corpus ls-colin, cuxac et al 2002)

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.

352

garcia and sallandre

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 ruminate 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 signer’s 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). (corpus ls-colin, cuxac et al 2002)

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.

reference resolution in french sign language (lsf)

353

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 combination 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 viewpoints 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, it’s a horse.” In this case, the TUs are topical while the LU HORSE takes on the focus function. This inversion, compared to the more common combination discussed above, produces a rhetorical “zoom” effect, creating a little suspense (“whatever is that thing which is moving like this? Ah, it’s a horse!”).

figure 15

Situational transfer of the galloping horse in the field (left), personal transfer of the galloping horse (middle), LU [HORSE] (right) (corpus ls-colin, cuxac et al 2002)

354

garcia and sallandre

This sharp stylistic effect contrastively confirms the prototypical, unmarked character of the topical-LU/focal-TU combination. At the same time, this example 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)definiteness (see Engberg-Pedersen 1993; MacLaughlin 1997) or specificity/nonspecificity (Zimmer & Patschke 1990; Barberà 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 section 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 irrelevant—even if the motion is directed upwards (see figure 16), it does not serve to create a locus. The signer’s 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).

reference resolution in french sign language (lsf)

figure 16

355

Modal-enunciative pointing, by signer 1 (left photo) (corpus degels, boutora & braffort 2011)

In the second example—the response to the preceding utterance—signer 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 addressee’s 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).

356

garcia and sallandre

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

6

Modal-enunciative pointing by signer 2 (right photo) (corpus degels, boutora & braffort 2011)

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

reference resolution in french sign language (lsf)

357

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 referential (and in our view, fully controlled by Grammar, since they may introduce or instantiate referents on their own). Moreover, the frequent and apparently prototypical (unmarked) close association between LUs and TUs in introducing new discourse referents is of great interest. In the sequential LU-TU combination, 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 relevant in LSF for the expression of either definiteness or specificity. In contrast, we suggest that in further research the distribution of manual pointing in relation to gaze should be examined more closely. Furthermore, given our intention to expand reflection on nominal determination beyond the expression of reference 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, 167–174. Chicago, IL: University of Chicago Press.

358

garcia and sallandre

Antinoro Pizzuto, Elena, and Micaela Capobianco. 2008. “Is pointing “just” pointing? Unravelling the complexity of indexes in spoken and signed discourse.” Gesture 8:82–103. 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 Müller de Quadros, 475–495. Petrópolis/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, 59–90. New York: Academic Press. Barberà, Gemma. 2012. “The meaning of space in Catalan Sign Language (LSC): Reference, specificity and structure in signed discourse.” PhD dissertation, Universitat Pompeu Fabra, Barcelona. Battison, Robin. 1973. “Phonological deletion in American Sign Language.” Sign Language Studies 5:1–19. Benveniste, Emile. 1966. Problèmes de linguistique générale, volume 2. Paris: Gallimard. Boutora, Leïla and Annelies Braffort. 2011. Corpus DEGELS1, ⟨oai:sldr.fr:sldr000767⟩ Bouvet, Danielle. 1996. Approche polyphonique d’un récit 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 Vollhaber, 205–228. Hamburg: Signum Press. Brennan, Mary. 2001. “Encoding and capturing productive morphology.” In Sign Transcription and Databases Storage of Sign Information, Sign Language and Linguistics, volume 4, number 1/2, edited by Ronnie Wilbur, 47–62. Amsterdam/Philadelphia: John Benjamins. Cormier, Kearsy, David Quinto-Pozos, Zed Sevcikova and Adam Schembri. 2012. “Lexicalisation and de-lexicalisation processes.” In “Sign languages: Comparing depicting constructions and viewpoint gestures”, Language and Communication 32:329–348. Culioli, Antoine. 1990. Pour une linguistique de l’énonciation. Paris: Ophrys. Cuxac, Christian. 1985. “Esquisse d’une typologie des Langues des Signes.” Journée d’études 10 (special issue: Autour de la Langue des Signes), 35–60. Paris: Université René Descartes. Cuxac, Christian. 1996. “Fonctions et structures de l’iconicité des langues des signes.” Thèse de doctorat d’État. Paris: Université Paris 5.

reference resolution in french sign language (lsf)

359

Cuxac, Christian. 2000. La Langue des Signes Française (LSF). Les voies de l’iconicité. Faits de Langue 15–16. Paris: Ophrys. Cuxac, Christian. 2004. “‘Phonétique’ de la LSF: une formalisation problématique.” Silexicales 4 (special issue) Linguistique de la LSF: recherches actuelles: 93–113. Cuxac, Christian; Annelies Braffort, Annick Choisier, Christophe Collet, Patrice Dalle, Ivani Fusellier, Gwenaëlle 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 diagrammatic iconicity.” In Verbal and Signed Languages: Comparing Structures, Constructs and Methodologies, edited by Elena Pizzuto, Paola Pietrandrea and Raffaele Simone, 13–33. Berlin: Mouton de Gruyter. Cuxac, Christian and Elena Antinoro Pizzuto. 2010. “Emergence, norme et variation dans les langues des signes: vers une redéfinition conceptuelle.” In Langage et Société 131 (special issue Sourds et langue des signes. Norme et variations), edited by Brigitte Garcia and Marc Derycke, 37–53. DeMatteo, Asa. 1977. “Visual Imagery and visual Analogues in American Sign Language.” In On the other hand: New Perspectives on American Sign Language, edited by Lynn A. Friedmann, 109–136. 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: 223–238. 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: SIGNUMVerlag. Engberg-Pedersen, Elisabeth. 2003. “From Pointing to Reference and Predication: Pointing 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. “L’article défini en ancien français: l’expression de la subjectivité.” Langue française 107: 58–71. Epstein, Richard. 2002. “The Definite Article, Accessibility, and the Construction of Discourse Referent.” Cognitive Linguistics 12.4: 333–378. 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.”

360

garcia and sallandre

In Conceptual Structure, Discourse and Language, edited by Adele Goldberg, 113–130. 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 51:676–710. 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, surdité, langue(s) des signes et épistémologie des sciences du langage. Problématiques de la scripturisation et modélisation des bas niveaux en Langue des Signes Française (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 Société, nº 131, Fondation Maison des Sciences de l’Homme, Paris. Garcia, Brigitte and Marie-Thérèse L’Huillier. 2011. Corpus Creagest, Dialogues en LSF adulte. Garcia, Brigitte, Marie-Anne Sallandre, Camille Schoder and Marie-Thérèse L’Huillier. 2011. “Typologie des pointages en Langue des Signes Française (LSF) et problématiques de leur annotation.” In Proceedings of Défi Geste Langues des Signes (DEGELS 2011), 107–119. TALN 2011 Montpellier, 1 July 2011. Hickmann, Maya, 2003. Children’s discourse: person, space and time across languages. Cambridge: Cambridge University Press. Jakobson, Roman. 1963 [1957]. Essais de linguistique générale. Paris: Editions de Minuit. Johnston, Trevor and Adam Schembri. 1999. “On defining lexeme in a sign language.” Sign Language and Linguistics 2:115–185. 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 Française. Edited by Brigitte Garcia. L’Harmattan: Paris. Kamp, Hans and Uwe Reyle. 1993. From discourse to logic. Introduction to model theoretic 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, 376–396. Chicago: University of Chicago Press. Klima, Edward and Bellugi, Ursula (eds.) 1979. The Signs of Language. Cambridge, MA: Harvard University Press.

reference resolution in french sign language (lsf)

361

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 Application. 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, 176–198. 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, 19–41. Hillsdale, NJ: Lawrence Erlbaum Associates. Liddell, Scott K. 2003. Grammar, Gesture, and Meaning in American Sign Language. Cambridge: Cambridge University Press. Liddell, Scott K. and Metzger, Melanie. 1998. Gesture in sign language discourse. Journal of Pragmatics 30:657–697. 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, 155–170. 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, 365–387. 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, 191–210. Chicago: University of Chicago Press. Lyons, John. 1977. Semantics, Volumes 1–2. 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, 57–107. 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, 175–190. 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

362

garcia and sallandre

française de Belgique (LSFB): morphologie, syntaxe, énonciation. Rivages Linguistiques. 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 Morphology.” In Sign Language Research: Theoretical Issues, edited by Ceil Lucas, 118–132. Washington D.C.: Gallaudet University Press. Perniss, Pamela, M., Roland Pfau, and Markus Steinbach. 2007. Visible Variation. Comparative Studies on Sign Language Structure. Trends in Linguistics. Berlin: Mouton de Gruyter. Peirce, Charles Sanders. 1978. Écrits sur le signe. Translation by Gérard 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 Proceedings of SALT XV, edited by E. Georgala and J. Howell, 152–168. Ithaca, NY: Cornell University. Quer, Josep. 2011. “Reporting and quoting in signed discourse.” In Understanding Quotation, edited by Elke Brendel, Jörg Meibauer and Markus Steinbach, 277–302. Berlin: Mouton de Gruyter. Quer, Josep and Santiago Frigola. 2006. “The workings of indexicals in role shift structures in Catalan Sign Language (LSC).” Actes del VII Congrés de Lingüística General, Universitat Barcelona. 18 al 21 d’abril de 2006. CD ROM. Quinto-Pozos, David. 2007. “Can constructed action be considered obligatory?” Lingua 117:1285–1314. Riegel, Martin, Jean-Christophe Pellat and René Rioul. 2002. Grammaire méthodique du français. Paris: Quadrige/PUF. Rinfret, Julie. 2008. L’association spatiale du nom en Langue des Signes Québecoise: Formes, fonctions et sens. PhD Thesis, University of Québec Montréal (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 unités du discours en Langue des Signes Française. Tentative de catégorisation dans le cadre d’une grammaire de l’iconicité. 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-

reference resolution in french sign language (lsf)

363

bergen, Lorraine Leeson, and Onno Crasborn, 103–125. Amsterdam/ Philadelphia: John Benjamins. Sallandre, Marie-Anne and Brigitte Garcia. (2013). “Epistemological issues in the semiological 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, Aurélie Sinte, Mieke Van Herreweghe and Myriam Vermeerbergen. Berlin / New York: Mouton De Gruyter, 159–177. Sandler, Wendy and Diane Lillo-Martin. 2006. Sign Language and Linguistic Universals. Cambridge: Cambridge University Press. Schembri, Adam. 2003. “Rethinking ‘Classifiers’ in Signed Languages.” In Perspectives on classifier constructions in sign languages, edited by Karen D. Emmorey, 3–34. 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, 271–296. Mahwah, NJ: Lawrence Erlbaum Associates. Smith, Cheri, Ella Mae Lentz, and Ken Mikos. 1988. Signing naturally: teacher’s curriculum guide, Level 1. Berkeley CA: DawnSign Press. Stokoe, William C. 1960. “Sign Language Structure.” Studies in Linguistics—Occasional Paper, 8 (rev. ed. Linstok Press, Silver Spring, MD, 1978). Supalla, Ted. 1978. “Morphology of Verbs of Motion and Location in American Sign Language.” 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), 27–46. 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 26:168–192. Wilbur, Ronnie. 1979. American Sign Language and sign systems: research and application. Baltimore: University Park Press.

364

garcia and sallandre

Winston, Elisabeth. 1991. “Spatial referencing and cohesion in an American Sign Language text.” Sign Language Studies 73:397–410. Winston, Elisabeth. 1995. “Spatial mapping in comparative discourse frames.” In Language, Gesture, and Space, edited by Karen Emmorey and Judy S. Reilly, 87–114. Hillsdale, 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, 17–30. Hamburg: Signum Press. Zimmer, June and Patschke Cynthia, 1990. “A Class of Determiners in ASL”. In Sign Language Research: Theoretical Issues, edited by Ceil Lucas, 201–210. Washington DC: Gallaudet University Press. Zucchi, Sandro. 2004. Monsters in the Visual Mode? Ms., Università 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 Hélade 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 Portuguese, and in the acquisition of Brazilian Portuguese by native speakers of English and Spanish. English, Spanish and Brazilian Portuguese exhibit a threeway 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 Mónica 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.

© koninklijke brill nv, leiden, 2014 | doi: 10.1163/9789004260825_013

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 crosslinguistic 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 contexts 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 speakers (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 section, English, Spanish, and BrP differ from one another in how they express generic interpretation for NPs with and without articles; the three-way difference 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 discussing the findings, we pose questions for further research.

369

when articles have different meanings

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 interpreted as being about a specific hummingbird, but this reading is irrelevant for our purposes). In contrast, English definite plurals normally lack generic readings, 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 hummingbirds 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. b. c. d. e.

The hummingbird is a bird. A hummingbird is a bird. *Hummingbird is a bird. The hummingbirds are birds. Hummingbirds are birds.

[definite singular, √generic] [indefinite singular, √generic] [*bare singular] [definite plural, #generic] [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 pájaro. the hummingbird is a bird

[definite singular, √generic]

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 pájaro. a hummingbird is a bird c. *Picaflor es un pájaro. hummingbird is a bird d. Los picaflores son pájaros. the hummingbirds are birds e. *Picaflores son pájaros. hummingbirds are birds

[indefinite singular, √generic]

[*bare singular]

[definite plural, √generic]

[*bare plural]

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 Müller 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 é uma ave. the hummingbird is a bird

[definite singular, √generic]

b. Um beija-flor é uma ave. a hummingbird is a bird

[indefinite singular, √generic]

c. Beija-flor é ave. hummingbird is bird d. Os beija-flores são aves. the hummingbirds are birds e. Beija-flores são aves. hummingbirds are birds

[bare singular, √generic]

[definite plural, √generic]

[bare plural, √generic]

The judgments reported in (3) through (5) above, about which NP types are allowed in generic environments in English, Spanish and BrP, received experimental support from a study with native speakers by Ionin et al. (2011a). We next consider the sources of these generic readings.

when articles have different meanings

371

2.1 Sentence-Level vs. NP-Level Genericity As discussed by Krifka et al. (1995), genericity can come either from the sentence level or from the NP-level. At the sentence level, generic sentences are sentences that describe habitual or characteristic events/behaviours, in contrast 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 describe 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 history 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 Polly’s 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. b. c. d. e.

The hummingbird is rare in the United States. [definite singular] #A hummingbird is rare in the United States. [#indefinite singular] *Hummingbird is rare in the United States. [*bare singular] #The hummingbirds are rare in the United States. [#definite plural] 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 Müller (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 Munn’s analysis. Ionin et al.’s findings on bare singulars (7c) were inconclusive: 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 é raro em São Paulo. the hummingbird is rare in Sao Paulo b. #Um beija-flor é raro em São Paulo. a hummingbird is rare in Sao Paulo

[definite singular]

[#indefinite singular]

c. ?Beija-flor é raro em São Paulo. hummingbird is rare in Sao Paulo

[?bare singular]

d. Os beija-flores são raros em São Paulo. the hummingbirds are rare in Sao Paulo

[definite plural]

e. Beija-flores são raros em São Paulo. hummingbirds are rare in Sao Paulo

[bare plural]

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 subjects. 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 singulars were to some extent allowed with kind readings, but not as fully as definite 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. Müller 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 written 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 corpora, including the example in (8). Thus, the status of bare singulars in BrP

373

when articles have different meanings

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 infrequently) appear with kind predicates suggests the opposite. Furthermore, the fact that bare singulars were rated relatively low in subject position, in a written 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 readings, and/or that register rather than grammar is responsible for the low ratings. (8) Boto chinês está 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 (9c–d). (9) a. b. c. d.

The Coke bottle has a narrow neck. #The green bottle has a narrow neck. Coke bottles have narrow necks. Green bottles have narrow necks.

[non-generic reading ok]

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 sentences of the form in (9b), with definite singular NPs that denoted non-welldefined 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 sentences, 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 individual 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 lexicalized, 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 environments, 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 environments, bare plurals are indefinite: e.g., in generic sentences like (3e), bare plurals, like indefinite singulars, are indefinite terms bound a generic operator (on Chierchia’s 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 ambiguity between canonical definite readings, kind readings, and indefinite readings (cf. Zamparelli 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 discussion.

when articles have different meanings

375

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 context, 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 Müller (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 DobrovieSorin 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 definite 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 others). 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 categories 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 L2acquisition 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 interpretation (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 distinctions lexicalized in the L1 (such as perfective aspect, or definiteness, or kindformation) 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 nominals, we focus on four distinct populations: adult Spanish-speaking and BrPspeaking learners of English, as well as adult English-speaking and Spanishspeaking learners of BrP. Comparing two learner groups for each target language 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.

when articles have different meanings

3

377

Experimental Study

3.1 Participants In order to test the predictions of L1-transfer outlined above, we have conducted 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 multiplechoice 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 studying 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 difference 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 country. 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 four—in Brazil (mean proficiency test score 80%, range 60% to 90 %). An independentsamples t-test found no difference in proficiency test scores between the L1English 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. university).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, respectively. 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).

when articles have different meanings

379

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. Unfortunately, there are many animals that can’t be found in a zoo, or anywhere else. It’s very sad. For example … a. b. c. d. e.

The dodo bird is extinct. A dodo bird is extinct. Dodo bird is extinct. The dodo birds are extinct. Dodo birds are extinct.

[definite singular] [indefinite singular] [bare singular] [definite plural] [bare plural]

(11) Kind-reference category (BrP): Eu gosto muito de ir ao zoológico. Infelizmente, há vários animais que nós não vemos mais no zoológico ou em nenhum outro lugar. É muito triste! Por exemplo … a. b. c. d. e.

O pássaro dodô está extinto. Um pássaro dodô está extinto. Pássaro dodô está extinto. Os pássaros dodô estão extintos. Pássaros dodô estão extintos.

[definite singular] [indefinite singular] [bare singular] [definite plural] [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. b. c. d. e.

The green lamp is very relaxing. A green lamp is very relaxing. Green lamp is very relaxing. The green lamps are very relaxing. Green lamps are very relaxing.

[definite singular] [indefinite singular] [bare singular] [definite plural] [bare plural]

(13) Generic category (BrP): O meu irmão tem estado de mal humor ultimamente. Não sem motivo: o apartamento dele é muito desconfortável e deve ser muito deprimente morar lá. E ele tem um lustre com uma luz muito fraca e desagradável. Eu disse a ele que ele deveria comprar uma luminária nova: alguma coisa agradável. Por exemplo, eu sei que …

380

ionin et al.

a. b. c. d. e.

A luminária verde é muito relaxante. Uma luminária verde é muito relaxante. Luminária verde é muito relaxante. As luminárias verdes são muito relaxantes. Luminárias verdes são muito relaxantes.

[definite singular] [indefinite singular] [bare singular] [definite plural] [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 indefinite singular generics ((11a,b) and (13a,b)), while giving lower ratings than native BrP speakers to bare singulars ((11c) and (13c)). With respect to plural 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 L1Spanish 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

when articles have different meanings

figure 1

381

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