Definiteness Effects : Bilingual, Typological and Diachronic Variation [1 ed.] 9781443898003, 9781443890571

Citation preview

Definiteness Effects

Definiteness Effects: Bilingual, Typological and Diachronic Variation Edited by

Susann Fischer, Tanja Kupisch and Esther Rinke

Definiteness Effects: Bilingual, Typological and Diachronic Variation Edited by Susann Fischer, Tanja Kupisch and Esther Rinke This book first published 2016 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2016 by Susann Fischer, Tanja Kupisch, Esther Rinke and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-9057-X ISBN (13): 978-1-4438-9057-1

CONTENTS

Preface ....................................................................................................... vii Introduction ................................................................................................. 1 Susann Fischer, Tanja Kupisch and Esther Rinke I Introductory Papers Chapter One ............................................................................................... 14 Definiteness Effect and Unaccusative Subjects: An Overview and Some New Thoughts Adriana Belletti and Valentina Bianchi Chapter Two .............................................................................................. 66 Definiteness Effects: The Interplay of Information Structure and Pragmatics Manuel Leonetti II Typological Perspective Chapter Three .......................................................................................... 120 Pseudo-Existentials and Definiteness Effects in Italian Silvio Cruschina Chapter Four ............................................................................................ 149 (In)definiteness Effects Once Again: The Case of Negated Existentials Nadia Varley Chapter Five ............................................................................................ 175 Definiteness Effect, Pronouns and Information Structure in Catalan Existentials Xavier Villalba Chapter Six .............................................................................................. 213 Existentials, Possessives and Definiteness in Samoyedic Languages Beáta Wagner Nagy

vi

Contents

III Diachronic Perspective Chapter Seven.......................................................................................... 244 Modal Verbs in German and Definiteness Effects on the Subject Argument – Focusing on Modern Standard German SOLLEN and Middle High German SULN “shall” Werner Abraham and Maiko Nishiwaki Chapter Eight ........................................................................................... 278 The Evolution of Definiteness Effects with French IL Y A from 1300 to Today Charlotte Coy Chapter Nine............................................................................................ 301 Existentials vs. Unaccusatives: The Definiteness Restriction in Romance Susann Fischer Chapter Ten ............................................................................................. 333 Verbal Aspect and Definiteness Effects in Catalan Absolute Small Clauses Jorge Vega Vilanova Chapter Eleven ........................................................................................ 374 Definiteness Effects in the History of French: The Case of Presentationals Michael Zimmermann IV Bilingual/Second Language Acquisition Perspective Chapter Twelve ....................................................................................... 404 Definiteness Effects in German-Turkish Bilinguals Acquiring English as Third language Tanja Kupisch Chapter Thirteen ...................................................................................... 424 Japanese Speakers’ L2 Acquisition of the English Definiteness Effect Neal Snape and Setsu Sekigami Chapter Fourteen ..................................................................................... 447 The Acquisition of Definiteness Effects in the L2 European Spanish of L1 German and L1 Turkish Speakers Marina Zielke

PREFACE

During the last four decades, especially with the rise of discussions on the syntax-semantics interface, definiteness restrictions have received increasing attention in theoretical and empirical research. Our idea has been to discuss and collect the different views in a single volume on “Definiteness effects: typological, diachronic and bilingual variation”. Most of the papers were presented at the DGfS (Deutsche Gesellschaft für Sprachwissenschaft) in 2012 in Frankfurt; additional papers were added by approaching leading experts on the topic. Now, some years later, we are proud to present you with 14 papers of cutting edge research on the topic. We hope that you will enjoy the articles as much as we have. There are quite a few people to whom we want to express our sincere gratitude for their help and support. Many thanks to the people who helped us reviewing the papers in the present volume: Artemis Alexiadou, Giuliano Bocci, Charlotte Coy, Barbara Egedi, Valentin Gusev, Wolfgang Meyer, Sabine Mohr, Svetlana Petrova, Oliver Schallert, Emanuela Sanfelici, Neal Snape, Renata Szczepaniak, Xavier Villalba, Jorge Vega Vilanova, Marina Zielke, and Michael Zimmermann, We would also like to thank Sarah Jobus and Kirsten Brock for their comments and corrections on our introductory chapter, and Luana D’Agosto, Miriam Geiss, Sarah Jobus, Julia Ritz and Sarah Zander for their help with the formatting and final proof reading. We also wish to thank Victoria Carruthers, Sophie Edminson, and Sean Howley for the support we received during the editing process. Last, but not least our families and friends deserve to be thanked for their understanding and support and the countless ways in which they brighten up our lives outside linguistics.

INTRODUCTION SUSANN FISCHER, TANJA KUPISCH AND ESTHER RINKE

Since Milsark (1977) and Perlmutter (1978) it has been well known that in certain sentence positions indefinite noun phrases are permitted (see 1a and 2a), while definite noun phrases are excluded (see 1b and 2b). Restrictions of this kind are termed “definiteness effects” and have been observed in existential (1) and unaccusative (2) constructions. (1)

a. b.

There is a cat in my garden. *There is the cat in my garden.

(2)

a. b.

There arrives a train. *There arrives the train.

The definiteness effect (DE; also definiteness restriction) has been discussed within different linguistic areas including syntax, semantics, and pragmatics, as well as combinations thereof. However, so far, no account has been able to explain all aspects pertaining to this phenomenon, including its motivation, exceptions to it, and variation across languages. Discussions have mostly been based on English, although analyses of languages other than English have revealed typological variation with respect to the DE. For example, based on Italian, a null-subject language in which definite subjects are licensed not only in preverbal (3a) but also in postverbal (3b) position, it has been discussed whether null-subject languages generally allow exceptions to the DE. (3)

a. b.

Il ragazzo è arrivato. the boy is arrived È arrivato il ragazzo. is arrived the boy ‘The boy has arrived.’ (Belletti 1988: 7)

2

Introduction

However, Turkish and Russian, despite also being null-subject languages, show definiteness effects in affirmative but not in negative existentials (White et al., 2012). Such observation call into question whether the DE is a universal phenomenon that can be traced to a common source, like for example the null-subject parameter. A thorough comparison of inter- and intra-linguistic variation shows that the DE correlates with a number of language- and constructionspecific properties. It turns out that, at least in some languages, existential sentences can be subdivided into different types, with each type behaving differently with respect to the DE. Furthermore, the information-structural interpretation of certain constructions and particular properties of strong pronouns, for example in Catalan, may have an impact on the DE. From a diachronic perspective, a relevant question is whether one particular language has shown the DE in all its diachronic stages and whether the restriction has always been equally strict. For example, based on data from Old and Middle French, some authors argue that the DE may surface in different ways across languages, and that the DE is construction dependent and correlates with changes in the clausal architecture of the language. By determining the factors relevant for the development of the DE in different constructions and languages, diachronic studies can contribute to answering central questions about the interplay between syntactic structure on the one hand, and semantic and pragmatic factors concerning the interpretation of sentences which do or do not abide by the DE on the other hand. Similarly, learner behaviour in language acquisition studies may help us to distinguish universal trends from language-specific patterns. Studies in second language acquisition show how learners acquire the DE and whether or not L2 production is dependent on DE effects in their native language(s). For example, if learners acquire the DE in a second language although it plays out differently in their first language(s), there are good reasons to argue that the DE is a semantic universal to which learners have access at all stages of language acquisition. In general, observations on definiteness restrictions have played an important role in the development and design of syntactic and semantic theories. They have influenced the way in which feature checking and the syntactic movement of objects and subjects are envisioned, and they have shaped our views on the syntax-semantics and semantics-pragmatics interfaces. For these reasons, the DE has received increasing attention over the past few years. However, most of the relevant studies are scattered in conference proceedings, working papers, and dissertation chapters, and are often hard to obtain. Our collection intends to fill this gap by offering a

Definiteness Effects: Bilingual, Typological and Diachronic Variation

3

compendium of up-to-date research. In particular, the book aims at contributing to the ongoing discussion about definiteness restrictions within a generative framework by presenting different perspectives –the typological, diachronic, and second language acquisition perspectivesí in a single volume. While primarily focussing on the key syntactic features of the DE, many of the papers also address its semantic and pragmatic aspects and the syntactic-semantic interface, using different methodological approaches, including, for instance, corpus searches, introspection, and grammaticality judgement tasks. The book consists of 14 chapters, which are representative of theoretically driven empirical research. The data covers Romance languages (Catalan, French, Italian, Spanish) and Germanic languages (English, German) as well as Bulgarian, Japanese, Turkish, and the Samoyedic languages. The introductory papers by ADRIANA BELLETTI & VALENTINA BIANCHI and by MANUEL LEONETTI explore the syntactic, semantic, and information-structural foundations of the DE in unaccusative and existential constructions and develop existing proposals by adding new empirical and theoretical insights. Belletti & Bianchi focus on syntactic and semantic aspects, while Leonetti is primarily concerned with the impact of information structure on the DE. Based on Belletti’s (1988) original proposal, BELLETTI & BIANCHI reconsider the status of VP-internal unaccusative subjects (i-subjects), proposing a unified analysis for the i-subjects of unaccusatives and the subjects of existential sentences which assumes that in these constructions no phi-complete probe is available which can check nominative Case via Agree. According to Belletti & Bianchi, i-subjects correspond to a defective nominal projection (NumP) which receives Partitive Case via Agree with a defective probe lower than T. This probe contains number and gender features but lacks a person feature. As a consequence of this defectiveness, i-subjects lack the D-layer. Therefore, they cannot saturate arguments and are interpreted as property-denoting elements which are incorporated into the predicate. Under this view, only weak indefinites can be i-subjects. Definite postverbal subjects, by contrast, are analysed as occupying a vP/VP-peripheral focus position. (4)

pro

è is

arrivata la ragazza. arrived.F.SG the girl

(Italian)

Overall, BELLETTI & BIANCHI see the DE as a deep and defining property of unaccusative verbs. Their view is supported by the early sensitiveness

4

Introduction

to the definiteness of the postverbal subject with unaccusative verbs in first language acquisition. MANUEL LEONETTI emphasizes the role of information structure (IS) in the analysis of the definiteness effect, arguing that the DE arises in two IS configurations: In constructions with wide focus the occurrence of anaphoric definite DPs is generally banned. In constructions with narrow focus exceptions are possible when the DP receives an availability/list reading. His account explains exceptions to the DE effect as the result of pragmatic inferences. In other words, pragmatic inferences can “repair” semantic mismatches by contextualizing the interpretation of the definite pivot in terms of narrow focus and an “availability reading”, as illustrated in (5). (5)

(answer to the question: Who can we leave the children with?) C’ è tua sorella. Loc is your sister ‘There’s your sister.’

Pragmatic inference is also held responsible for the resolution of cases of competing candidates in languages in which more than one form is available for a single meaning (e.g. Spanish and Sardinian haber ‘have’ vs. estar ‘be’) – the unmarked formal option is selected. Since estar is an alternative to haber in existential sentences, the DE is robust in Spanish/Sardinian haber constructions. In contrast, languages such as English and French may display definite pivots (with an ‫ދ‬availability‫ތ‬ reading) in existential constructions because no competing option is available. Although the two introductory papers focus on different aspects of the DE, the authors of these papers agree that the DE arises as a consequence of syntactic, semantic, and information-structural factors. Furthermore, they jointly assume that lexical factors play a role: Belletti & Bianchi propose that the DE is a defining property of unaccusative verbs; Leonetti attributes cross-linguistic differences to the availability of competing options in the lexicon. The authors also agree that exceptions to the DE are possible when postverbal definite subjects receive a narrow focus interpretation. The second part of this book discusses the DE from the perspective of typological variation, specifically, the question whether the DE figures across languages and under which conditions this is the case. A general theme is the discussion of exceptions to the DE in the languages under consideration and how to account for them.

Definiteness Effects: Bilingual, Typological and Diachronic Variation

5

SILVIO CRUSCHINA demonstrates on the basis of semantic and information-structure considerations that Italian existential sentences with a definite DP, e.g. C’è il cane ‘There’s the dog’ and C’è Gianni ‘There’s Gianni’, are not existentials proper. He analyses various types of cisentences (there-sentences) in Italian. Existentials proper (his type I) and ‘inverse locatives’ (his type II) look very similar to each other. However, as Cruschina argues, the latter result from syntactic operations related to information structure. Such operations transform a locative predication, which is morphologically similar to a type I existential. The other two types of ci-sentences can be characterized as deictic (his type III) and presentational (his type IV). Cruschina’s classification sheds new light on the apparent differences between Italian and other languages, such as English, with regard to definiteness effects. He argues that only type I cisentences are genuine existentials, while the other types, despite their superficial similarities, correspond to different constructions. The absence of definiteness effects in Italian and the relative differences with respect to English are therefore explained in terms of pseudo-existential constructions, which are independently available in Italian but not in English. At first sight, these pseudo-existential constructions resemble existential sentences proper, but a contrastive analysis of the pragmatic, semantic, and syntactic properties of the elements occurring in these sentences reveals that they are different from genuine existentials. The analysis thus suggests that, contrary to traditional assumptions, there are no exceptions to the DE in Italian. NADIA VARLEY focusses on negated existential constructions in Bulgarian. These allow arguments that are marked by a definite article and doubled by a clitic, as illustrated in (6). Varley argues that the occurrence of definite arguments in these constructions only constitutes an apparent exception to the DE. She assumes that definite arguments in negated existential constructions occur outside the domain of existential closure (Diesing 1992) and have either been left dislocated to CP or dislocated to a designated Topic-Position (TopP) within a split vP-domain. .

(6)

V gradinata in garden-the (deca-ta)]] (children-DEF)

[vP Top gi CL.ACC.PL

[䳭vP njama . NEG-have[-AGR].PRS (Bulgarian)

Both the obligatory occurrence of a coreferential clitic in such constructions and the incompatibility of the argument with a broad focus interpretation provide further evidence in support of Varley’s analysis. Since a decomposed vP-domain in the sense of Belletti (2005) offers an

Introduction

6

escape hatch for an otherwise illegitimate definite/specific DP in existential constructions, Varley argues that the DE as a grammatical constraint should be maintained. Apparent DE violations in Bulgarian existentials are due to independent discourse features in the syntax. XAVIER VILLALBA discusses existential constructions in Catalan. Like in Italian, existentials in Catalan allow definite DPs and proper nouns in the pivot position but show a strong restriction against personal pronouns. (7)

a.

Hi

havia el degà, a la reunió. had the dean at the.F meeting ‘At that meeting, the dean was present.’ *Hi ha ell. LOC has he ‘He is here.’ (Rigau, 1988: ex. 2d) LOC

b.

Villalba proposes that the DE in Catalan results jointly from pragmatic requirements on pivots concerning the existential predicate and the nature of the elements occurring in the pivot position. Pronouns are banned from pivot positions because the pragmatic and informational requirements of pivots exclude anaphoric elements. In addition, Catalan strong pronouns in pivot position cannot be contrastive because they cannot alternate with a clitic, or a weak or a null variant. However, when the contrastive reading is obtained by ‘external means’, such as focus particles (only) or reinforcers (oneself), pronouns receive a stronger referential interpretation and can be rescued as pivots. BEÁTA WAGNER NAGY discusses definiteness in several Samoyedic languages, e.g. Selkup and Nganasan, focussing on existential, locative, and possessive constructions in Nganasan. The Samoyedic languages have neither definite nor indefinite articles, which makes it more difficult to see what falls under the definiteness effect and what interacts in these constructions. The author shows that one way to express definiteness in Samoyedic languages is through the non-possessive use of possessive suffixes; furthermore, she demonstrates that possessive and existential constructions are closely related. (8)

mΩu-ðu sҨürü ƾilҨΩnu þii-mΩΩ earth-3SG snow.GEN under hide-PTCP.PASS.3SGVX ‘The earth is hidden under snow.’ [KNT, 1994]

Wagner Nagy proposes that the choice of the copula verb is decisive for the interpretation of a given sentence. All Northern Samoyedic languages have different copulas for locatives as compared to existential sentences.

Definiteness Effects: Bilingual, Typological and Diachronic Variation

7

In cases where no explicit morphological marking for the definiteness of the theme (neither a different copula nor a definiteness marker) is present, as in Selkup, locational and existential sentences only differ from each other in terms of word order. Overall, all authors agree that the DE is a universal phenomenon which arises from the complex interplay of syntactic, semantic, and informationstructural factors. They further agree that exceptions to the DE are only apparent and can be explained as alternative constructions which are only superficially identical to existential or unaccusative constructions, for example pseudo-existentials in Italian (Cruschina) or topicalization structures in Bulgarian (Varley). Similarly, there are certain means by which the NPs of an existential sentence can be modified so that the definiteness effect is avoided (Wagner Nagy). Like Belletti & Bianchi and Leonetti, the authors highlight the role of information structure in accounting for exceptional occurrences of definite noun phrases (cf. in particular Villalba’s analysis of Catalan). The third part of the book is concerned with definiteness effects from a diachronic perspective, focussing on different constructions, e.g. impersonal constructions and small clauses. A central question is whether the structures under investigation have remained diachronically stable with respect to the DE, and which factors have influenced the DE in various diachronic changes. WERNER ABRAHAM & MAIKO NISHIWAKI are concerned with modal verbs and definiteness interpretations of the subject argument of modal verbs; specifically sollen ‘shall’ in Modern Standard German and suln ‘shall’ in Middle High German are investigated. Their starting point is the generalization that certain features of past marking on the predicate elicit evidential readings of the German modal verb sollen. Abraham and Nishiwaki discuss to what extent contextual factors (aspectuality of the infinitival complement, grammatical person on the modal predicate) trigger definiteness effects on the embedded infinitivals. In other words, they seek to find how the polyfunctionality of German(ic) modal verbs is disambiguated through the (in)definiteness reading of the embedded construction. The contribution uses insights from the well-known relation between aspect and definiteness, which are claimed to be exponents of one and the same superordinate notion. CHARLOTTE COY investigates the diachronic development of different kinds of existential constructions in French. Based on a diachronic corpus study, she identifies exceptions to the DE and explores their diachronic evolution. The paper shows that the existential constructions under investigation have evolved differently: il y a, which is the oldest

8

Introduction

construction, shows a continuous increase in the types of definite determiners in the postverbal DP as well as in their frequency. Il existe, by contrast, is a more recent construction and co-occurs only rarely with a definite postverbal DP. Coy accounts for this discrepancy by arguing that il existe introduces a genuine existential construction, whereas il y a often fulfils other functions, appearing, for example, in ‘list readings’, as ‘reminders’, or to denote ‘locations’. Coy’s data provides evidence in favour of a higher degree of grammaticalization concerning the il y a construction. SUSANN FISCHER compares the distribution of definiteness effects in unaccusative and existential constructions across Romance. She argues that definiteness effects in unaccusatives have a syntactic explanation, while the DE in existential constructions has a semantic or pragmatic explanation. According to her analysis, the DE in unaccusatives is dependent on whether a language is a null-subject language or not, since in null-subject languages it is the verb that checks the EPP, and, as a consequence, postverbal subjects are allowed in positions where they escape a weak existential interpretation (cf. Fischer 2010). The DE in existential sentences, by contrast, is independent of whether a language is a null-subject language or not. In existential sentences, a number of other factors, such as word-order variation and its connection to information structure, play an important role. JORGE VEGA VILANOVA provides a new approach to absolute small clauses on the basis of new data from Modern and Old Catalan, e.g. Un cop acabats els deures, podeu sortir a jugar ‘When you finish your homework, you can go and play’ [literally, ‫ދ‬Once finished the homework‫]ތ‬. Based on a comparison of the properties of absolute small clauses across Romance, the author shows that some of their properties have not been properly explained in previous accounts. For instance, the nominal argument of small clauses shows definiteness restrictions that are not attested in Old Catalan. In order to explain these restrictions, Vega Vilanova proposes an extended syntactic structure for absolute small clauses. More specifically, he argues that the aspectual feature of the participle is decisive in explaining the definiteness restriction on the nominal argument. The author thus shows that the definiteness restriction on absolute small clauses is different in nature from the definiteness effect in existentials or unaccusatives, despite some similarities between these three types of constructions. MICHAEL ZIMMERMANN investigates the DE based on impersonal constructions (ICs) in medieval French and discusses the relation between the DE and the null-subject property. His starting point is the fact that

Definiteness Effects: Bilingual, Typological and Diachronic Variation

9

Modern Standard French generally shows Definiteness Effects, although there is a small set of well-defined exceptions. Since Modern Standard French is a non-null-subject language, this observation is consistent with the postulated link between the DE and the non-null-subject property. In its medieval stage, i.e. before the 17th century, French has often been analysed as a null-subject language because subject pronouns are not consistently overt. This view, however, has been debated, since the omission of subjects is syntactically constrained. Given this controversy, a relevant question is whether or not the DE can be observed in the medieval stages of French. Zimmermann discusses this question based on ICs in a large diachronic data corpus. He argues that, like in Modern Standard French, the DE is also found in medieval French, thus supporting his analysis of medieval French as a non-null-subject language. All the authors in this section contribute to the discussion on the defining properties of the DE on the basis of diachronic data. Coy shares with authors of previous chapters (e.g. Leonetti and Cruschina) the idea that languages may have different ways of realizing existential constructions and that existentials proper have to be differentiated from other (superficially identical) constructions. Fischer and Zimmermann, for example, argue against the view advocated by Belletti & Bianchi that the DE is a defining property of unaccusative verbs that holds universally. Rather, the availability of the DE depends on whether a language is a nullsubject language or not. Zimmermann argues with respect to French that the existence of the DE throughout the history of the language shows that the language has always been a non-null-subject language. Fischer distinguishes between unaccusatives and existentials. Only the latter universally exhibit the DE, whereas the former only show it when the language is a non-null-subject language. Vega Vilanova and Abraham & Nishiwaki focus on constructions other than existentials and unaccusatives, specifically absolute small clauses (Vega Vilanova) and modal constructions (Abraham & Nishiwaki). They provide evidence that the DE may be present but figures differently depending on the construction. The final part of the book combines three studies on the acquisition of the definiteness effect. All three studies investigate transfer effects in the acquisition of the DE in L2 learners of English or Spanish, and all three involve one language which does not show the DE in negative existentials, either Japanese or Turkish, so that transfer can be systematically predicted. While Snape & Sekigami and Zielke are concerned with L2 acquisition in the traditional sense, Kupisch addresses the question of the transfer source when two languages have been previously acquired.

10

Introduction

TANJA KUPISCH is concerned with the role of cross-linguistic influence in the acquisition of the Definiteness Effect in English as a third language (L3) by German-Turkish bilingual speakers. The central questions are whether the speakers transfer the knowledge of the DE from one of their early acquired languages to their third language English and, if so, whether transfer depends on language dominance or typological proximity. Three different groups of speakers were tested: Turkish dominant bilinguals, German dominant bilinguals, and a group of balanced bilinguals. Since the DE differs systematically between Turkish and English, but not between German and English, problems with the DE in English could point to transfer from Turkish. More specifically, transfer from Turkish should result in problems with negative existentials since this is where Turkish differs from English and German. The experimental data show better performance with acceptable sentences than with unacceptable ones, but no noticeable effects of transfer. The three groups of speakers did not differ in their accuracy rates and did not behave differently with respect to positive and negative existentials. This indicates that neither language dominance nor typological proximity foster transfer concerning the DE. NEAL SNAPE and SETSU SEKIGAMI investigate Japanese L2 learners of English regarding affirmative and negative existential sentences to see whether they accept violations of existential there constructions such as, for example, *There is the man in the room. As Snape and Sekigami argue, although Japanese L2 learners are known to struggle when using English articles, they may nevertheless know how definiteness functions. Similar to Turkish (cf. Kupisch, this volume; Zielke, this volume), Japanese shows the DE in affirmative sentences but not in negative sentences, so that cross-linguistic influence, if it occurs, should be visible with negative existentials. By testing 10 Japanese learners of English Snape & Sekigami show that Japanese L2 learners of English at advanced levels of proficiency are able to differentiate between grammatical and ungrammatical affirmative and negative existential sentences, while intermediate learners cannot detect the ungrammaticality of strong DPs in existentials. The results show that sensitiveness to the English DE is related to the learner’s level of English; the higher the proficiency level, the more likely it is that the learner will be able to detect ungrammaticality (in general and with respect to strong definites in existentials). The data suggest that problems with articles in second language acquisition may be unrelated to definiteness. MARINA ZIELKE investigates the acquisition of the DE in existential constructions in L2 European Spanish by monolingual L1 speakers of German and Turkish. The DE in European Spanish and German shows

Definiteness Effects: Bilingual, Typological and Diachronic Variation

11

similar effects, while Turkish differs from these two languages in that positive existentials show a DE, whereas negative existentials do not. Based on an acceptability judgement task, Zielke shows that all groups except the intermediate L1 Turkish group acquire the DE very well. The results thus show advantages in the acquisition of the DE if there are strong similarities between L1 and L2. At the same time, differences between L1 and L2 can be overcome at advanced levels of proficiency. Thus, in line with Snape & Sekigami’s results, proficiency plays a role. In summary, all three papers on foreign language acquisition show that the DE can be acquired at advanced levels of proficiency, independently of differences between the learners‫ ތ‬first and second languages. These results suggest that the DE is a semantic universal to which learners have access throughout their lifespan. All in all, the book brings together different approaches to the DE with fresh data from hitherto unexplored languages. Even from this short overview of the different papers, it seems clear that the DE can only be accounted for when different linguistic levels are included in the analysis of the phenomenon. Diachronic and acquisition data support the robustness of the effects in the evolution of language across centuries and within language learners. Thus, the contributions in this book provide new empirical insights and, we hope, carve new paths for future investigations.

References Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry, 19: 1-34. —. 2005. Extended doubling and the vP periphery. Probus 17, 1-35. Diesing, Molly. 1992. Indefinites. Cambrigde, Mass.: MIT Fischer, Susann. 2010. Word-order as a trigger for grammaticalisation. Amsterdam: Benjamins. Milsark, G.L. 1977. Toward an Explanation of Certain Peculiarities of the Existential Construction in English. In Linguistic Analysis 3: 1-29. Perlmutter, David M. 1978. Impersonal Passives & the unaccusative hypothesis, In Berkeley Lingusitic Society 4: 157-189. Rigau, Gemma. 1988a. Els predicats no verbals i l’efecte d'especificitat. Estudi General8, 51-64. White, Lydia, Alyona Belikova, Paul Hagstrom, Tanja Kupisch & Öner Özçelik. 2009. Restrictions on definiteness in second language acquisition. Affirmative and negative existentials in the L2 English of Turkish and Russian speakers. Linguistic Approaches to Bilingualism 2.1: 56-84.

I INTRODUCTORY PAPERS

CHAPTER ONE DEFINITENESS EFFECT AND UNACCUSATIVE SUBJECTS: AN OVERVIEW AND SOME NEW THOUGHTS* ADRIANA BELLETTI AND VALENTINA BIANCHI

1. Introductory Overview Milsark (1974, 1977) in his seminal work singled out a property which typically affects the post-verbal noun phrase of existential clauses in English there sentences like (1a): this noun phrase must be indefinite. He also noted that, in this language, the same property affects the post-verbal noun phrase of certain verb types (e.g. appearance verbs.) as in (1b), also possible in there sentences, although at a peculiar stylistic level in this case (e.g. fairy-tales etc.): (1)

a. There is a man/*the man in the garden. b. There arose a storm / *the storm here.

Before the unaccusative hypothesis made its way in the forefront of the theoretical debate in formal syntax (Perlmutter 1978, Burzio 1986), Milsark called these verbs the “inside verbals”, thus anticipating the fundamental insight of the unaccusative hypothesis according to which the

*

We thank David Beaver, Fréderique Berthélot, Cornelia Hamann, Günther Grewendorf, Petra Schulz, Ur Shlonsky for feedback on (often subtle) data. Although this article has been conceived and worked on jointly in all steps of its elaboration, Adriana Belletti takes direct responsibility of sections 2.3, 4.1, 5, 5.1, and Valentina Bianchi of sections 2.1, 2.2, 3, 4.2, 4.3. The introductory and concluding sections 1, 2 and 6 are in common. Adriana Belletti’s research was funded in part by the European Research Council/ERC Advanced Grant 340297 SynCart – “Syntactic cartography and locality in adult grammars and language acquisition”.

Definiteness Effect and Unaccusative Subjects

15

nominal argument of an unaccusative verb, which can appear in the postverbal position of there sentences and which in SV declaratives is the preverbal subject of the clause, is in fact its internal argument and never is nor was an external argument (following Williams’ 1981 influential terminology, widely adopted ever since). This property came to be known as the definiteness effect (DE). Much work has been devoted to there sentences and to the DE more generally in the eighties and later, also adding a crosslinguistic dimension by introducing descriptions in the same domain of languages different from English (Safir 1982, Stowell 1981, Williams 1984, Belletti 1988, Lasnik 1992, 1999, more recently Deal 2009, Fischer this volume and references cited therein; McNally 2011 for a review of the semantics literature on existential sentences; see also further references cited throughout, although no list can do justice to such a long lasting debate). DE characteristically emerges in structures featuring an expletive and a post-verbal subject in either existential sentences like (1a) or in sentences with an unaccusative verb as in (1b); similarly, the effect shows up in (2a) and in (2b) in French and in sentences with a transitive verb in the passive voice as in (2c), again illustrating with French. In all cases, only an indefinite noun phrase can occupy the postverbal position. (2)

a. Il

y a trois filles ici. there have.3SG three girls here ‘There are three girls here.’ b. Il est arrivé trois filles/*les filles. EXPL is arrived.MSG three girls/ the girls ‘There arrived three /*the girls.’ c. Il a été tué un homme/* l’homme. EXPL has been killed.MSG a man / the man ‘There was killed a man/*the man.’ EXPL

This type of sentences open up the issue of the licensing of the postverbal noun phrase; since this noun phrase corresponds to the preverbal subject of SV declaratives containing the same verb, it is often referred to as a postverbal subject. Within Government and Binding, one main issue was to account for the licensing of the postverbal subject w.r.t. Case. The proposed solution was to link the expletive and the associate noun phrase by means of a representational chain, whereby the expletive received Nominative Case in the preverbal subject position/Spec,IP and transmitted it to the associate; by hypothesis, only indefinite noun phrases could enter such a representational chain without violating any constraint, most notably the Binding Condition C (see Safir 1982: 172 ff., 239 ff.; Safir

Chapter One

16

1987; Rizzi 1982). Moreover, by means of this chain the features of the associate NP could be transmitted to I°, thus accounting for verb agreement in English (visible in e.g. 6 below). Later on Chomsky (1995 and references cited there) proposed instead a derivational relation, whereby the associate NP moves at LF and replaces the uninterpretable expletive in Spec,IP; this movement is triggered by the principle of Full Interpretation, whereby at the interface with semantics, the LF structure cannot contain any uninterpretable elements. At that stage of research, it was also assumed that null subject languages like Italian do not manifest the DE, rather, they display generalized “free” subject inversion, as exemplified in (3). It was assumed that in these languages, Spec,IP is filled by a phonologically null expletive pro; lack of DE was attributed to the hypothesis that Nominative Case could be directly assigned to the subject in its postverbal position (under government) in this type of language: (3)

pro

è arrivata la ragazza. is arrived.F.SG the girl ‘The girl has arrived.’

(Italian)

Belletti (1988) departed from this line of analysis in both respects. On the empirical side, she argued that even in a null subject language like Italian, the DE does arise when the postverbal subject fills the thematic internal argument position, rather than being in a ‘peripheral’, position external to the verb phrase (VP-adjoined, as in “outside verbals” in Milsark’s 1974 terms; Deal 2009 for recent rediscussion along these lines). In (4), for instance, the postverbal subject precedes a PP complement with no intonational break between the two, hence it cannot be external to the verb phrase, right adjoined to VP: in this configuration, the DE can actually be detected.1 (4)

a. All’improvviso è entrato un uomo /*l’uomo dalla finestra. suddenly is entered a man / * the man from-the window b. E’ stato messo un libro / *il libro sul tavolo. be.3SG been put a book / the book on-the table (Belletti 1988: 9, (17a-b), (18a-b))

1

If an intonational break precedes the PP, a different structural analysis is possible, whereby the postverbal subject is peripheral to the verb phrase and the following PP is right-dislocated. On this type of option for postverbal subjects, see the discussion in section 5 below.

Definiteness Effect and Unaccusative Subjects

17

This in fact naturally leads to the claim that there is no such thing as a generalized process of “free” subject inversion, but that postverbal subjects come in different types and are of different nature also in a null subject language like Italian, a view later developed in further detail in the frame of the cartographic approach (Belletti 2004 and section 5 below). Hence, the fact that in a language like Italian a postverbal subject is also possible with transitive and intransitive verbs as well as with unaccusatives independently of its definite (as in e.g. 3) or indefinite nature should be kept distinct from the phenomenon illustrated by the indefinite postverbal subjects of sentences like those in (4), which pattern with the English and French examples in (1) and (2). On the theoretical side, Belletti’s proposal completely dissociated the licensing of unaccusative subjects obeying the DE from the licensing via Nominative Case. In particular, she argued that the indefinite subject in (1), (2) and (4) does not inherit Nominative Case from the expletive in Spec,IP, due to the presence of the VP barrier and, more generally, to the assumed lack of Case transmission processes altogether; rather, the indefinite subject is directly Case-licensed by the unaccusative verb (a view later shared in Lasnik 1992, 1999). The latter assigns Partitive Case – an inherent-type Case associated with the internal argument position and which, because of its semantic import, is only compatible with indefinite noun phrases. According to Belletti’s (1988: 3) proposal, Partitive Case is not a Case lexically associated with a specific Th-role; rather it is associated with a structural position, the internal argument position to which the verb assigns a Th-role in connection with the first Merge operation. The fact that assignment of Partitive Case is contingent on assignment of a Th-role by the head V makes this Case more akin to an inherent Case than to a structural Case, assuming the bipartite distinction inherent vs structural Case operating in Government & Binding. However, as is discussed in Belletti (1988) and will be further addressed in the following discussion (section 4.1.), assignment of Partitive Case crucially involves the structural property of concerning the internal argument position.2 On the other hand, Nominative Case assignment (under government) is still invoked for postverbal subjects which are external to the verb phrase, 2

The fact that inherent Case is often typically associated with a particular Th-role does not imply that it always is/must be (Belletti 1988, section 2); the term lexical Case is sometimes used to refer to an inherent Case which is associated with a particular Th-role; sometimes Partitive Case can have this status as well, as mentioned in Belletti (1988: footnote 6). For critical views on the inherent nature of Partitive Case see Vainikka & Maling (1996).

18

Chapter One

hence in a different position than the internal argument position; no DE is manifested in these cases, which are possible with all verb classes, transitive, intransitives and also unaccusatives; the example in (3) illustrates the point with an unaccusative verb. Thus, this analysis sharply distinguishes VP-internal from vP/VP-peripheral “inverted/post-verbal subjects”3. In hindsight, Belletti’s proposal about the special licensing of VPinternal unaccusative subjects can be connected to a more general observation due to Diesing (1992). Noun phrases have different semantic properties according to whether they are interpreted inside or outside the verb phrase: indefinites, on their weak non-presuppositional interpretation (see section 2), appear inside VP, whereas presuppositional indefinites, as well as definite and strong noun phrases, seem to be licensed and interpreted in a position external the verb phrase (see de Hoop 1992, Ladusaw 1994, Diesing & Jelinek 1995 for further developments of this idea). This body of research has left us with an important insight that is worth rethinking in the light of current theoretical assumptions, also more closely considering the interface between syntax and semantics: (5)

VP-internal unaccusative subjects (henceforth: i-subjects, for internal subjects) involve a special licensing route which necessarily correlates with the DE.

In this article we reconsider the status of i-subjects and we propose a principled account of the DE that singles them out from other instances of postverbal subjects in inversion structures. In a nutshell: building on Belletti’s (1988) original insight, we propose in current terms that isubjects cannot be licensed via the usual Case/Agree route involving a phicomplete probe, i.e. T (for nominative) as in the minimalist reformulation of ‘structural Case’. We argue that, because of this licensing defectiveness, they cannot be interpreted as saturating arguments; rather, they must be interpreted at the interface as denoting a property; this crucially yields DE. The internal argument of unaccusatives corresponds to a defective nominal projection (NumP) and is syntactically licensed via Agree with a defective 3

The label vP incorporates the by now familiar notation inspired by the minimalist tradition according to which transitive and intransitive verbs are inserted in a vP shell containing further functional verbal elements as light verbs (“small v”) introducing the external argument and, possibly, other arguments of the verbal root. Unaccusatives are instead typically inserted in a VP, with no further small v (or anyway a somewhat reduced vP-shell; see the proposal on the existential verb phrase put forth in section 2.3). Section 4 and footnote 32 below for more.

Definiteness Effect and Unaccusative Subjects

19

M-probe lower than T containing number and gender features, but lacking the feature person. We locate the defective probe immediately above the projection of the lexical verb; the weak indefinite (NumP) is thus associated with Partitive Case through the Agree relation with the defective probe. This licensing route is typically unavailable for the external argument of (active) transitive and intransitive verbs. We assume that the defective nominal projection must be interpreted at the interface as denoting a property, and that it undergoes a form of ‘semantic incorporation’ (for which we will assume the implementation proposed by Chung & Ladusaw 2004). In this study we will also briefly reconsider some crucial features of existential sentences in which the DE is typically manifested in a standard way and reduce the fundamental aspects of this instance of DE to the type of computation that we assume to be generally at play with unaccusative verbs; we assume that the existential verb is indeed an unaccusative verb (as already first proposed in Burzio 1986, Stowell 1981 and then Moro 1997), with its own internal argument (following Williams’ 1984 original insight, in the terms already sketched out in Belletti 1988: footnote 5). The article is organized as follows. In section 2, after introducing the distinction weak vs. strong noun phrase, we give a more precise characterization of the DE: on the one hand, we show that i-subjects are not simply required to be indefinite, but more precisely weak (nonpresuppositional) indefinites, as they are unable to take wide scope w.r.t. negation (section 2.1); on the other hand, building on Vangsnes (2002), we distinguish the radical DE of i-subjects from a less tight restriction on definiteness that affects other types of postverbal subjects as in the so called Transitive Expletive Construction (TEC) (section 2.2). In section 2.3 we discuss existential sentences, and propose that the post-copular noun phrase is selected by the unaccusative existential verb (be). This discussion leads us to conclude that non-presuppositional indefiniteness is only required for internal arguments of unaccusative and passive verbs when they are licensed in situ, i.e. as i-subjects. In section 3 we show that the cluster of properties characterizing isubjects is shared by an apparently unrelated type of noun phrase, namely Maori he-indefinites as described in Chung & Ladusaw (2004, chapter 2): these too (a) are non-presuppositional, (b) necessarily take narrow scope w.r.t. negation, and (c) are restricted to the internal argument position. In section 4 we propose that this cluster of properties is a consequence of their syntactic licensing (section 4.1); this licensing triggers, at the interface, the application of the compositional rule of Predicate Restriction (section 4.2).

Chapter One

20

In section 5 we turn to the peripheral post-verbal subjects of Italian (exemplified in (3) above), which are exempt from the DE: we adopt and extend the analysis proposed in Belletti (2004), according to which these subjects are licensed in a Focus position in the periphery of vP/VP. In section 5.1 we show that the double licensing route for postverbal subjects in Italian (VP-internal vs. vP/VP-peripheral) can nicely account for some recent findings from L1 and L2 acquisition, which lend original new support to the unaccusative hypothesis through the manifestation of DE, which we view as a core property of unaccusatives. Finally, in section 6 we offer a summary and some concluding remarks.

2. A Closer Look at the Definiteness Effect The type of contrast exemplified in (1) above only involve indefinite vs. definite noun phrases; but the DE actually distinguishes two larger classes of noun phrases, which Milsark (1974) dubbed weak and strong, respectively: (6)

a. b.

There is a hole in my blanket. There are {three/some/many/no/a lot of} holes in my blanket. (weak noun phrases)

(7)

a. b.

*There is {every / each / neither} hole in my blanket. *There are {most / both / all} holes in my blanket. (strong noun phrases)

Following Milsark’s insights, the weak/strong opposition has been characterized in semantic terms. According to the relational view of quantification (Barwise & Cooper 1981, Keenan 1987), a quantificational determiner expresses a relation between the set of entities denoted by the noun it introduces – dubbed the restriction of the quantifier (R) – and the set denoted by the predicate, dubbed its nuclear scope (S). In (8), for instance, a relation is expressed between the set of snakes and the set of dangerous things:4

4

Note that a count common noun denotes a set of entities even in the singular: a noun like snake does not denote a single entity of a certain type, but the set of all entities which, intuitively, can be described as ‘snakes’. We are simplifying considerably for expository purposes; certain quantifiers involve a plural noun, but we skip over the interpretation of plurality for reasons of space.

Definiteness Effect and Unaccusative Subjects

(8)

21

{Every / some / no} snake is dangerous.

The determiner every expresses the subset relation: every entity that falls in the set of snakes also belongs in the set of dangerous things (though not necessarily vice versa); some expresses the relation of overlapping: the two sets have one or more element in common; finally, no expresses the relation of disjointness: the two sets have no element in common (i.e., nothing which belongs to the set of snakes also belongs to the set of dangerous things). In intuitive terms, the common feature of weak noun phrases is that the relation they express only makes reference to the intersection between the restriction set and the set denoted by the nuclear scope, but it does not require any presupposed knowledge about (the cardinality of) the restriction set. To exemplify, the truth conditions of Two snakes are dangerous require that the set of snakes has two members in common with the set of dangerous things: this can be verified independently of the actual cardinality of the (relevant) set of snakes (Reinhart 1987). By contrast, the truth conditions of Most snakes are dangerous cannot be verified independently of the cardinality of the restriction set: the sentence is true if and only if the set of dangerous snakes covers a large proportion of the set of snakes. The strong determiners of (7) are inherently presuppositional in that they can only be used felicitously when the restriction set is presupposed to be non-empty (for an accessible discussion of presuppositionality, see Heim & Kratzer 1998: 162 ff.). Since the quantificational relations expressed by weak determiners are essentially based on intersection, they are also known as intersective determiners. Empirically, weak determiners can be distinguished by the fact that they satisfy the following linguistic test (existentiality, Keenan 1987): a sentence of the form (9a) is true in exactly the same circumstances as (9b).5 (9)

a. b.

Det R is/are S Det R which is/are S exist(s).

To illustrate, consider (10):

5

This test is based on the property of existentiality (Keenan 1987): informally, quantificational determiner is existential if and only if the relation expressed between the two sets R and S holds in exactly the same circumstances in which the same relation holds between the intersection of R and S and the set of entities that constitutes the universe of discourse.

Chapter One

22

(10)

a. b.

No snakes are dangerous. No dangerous snakes exist.

(10a) and (10b) are intuitively equivalent– i.e. they are true in exactly the same circumstances, and false in exactly the same circumstances:6 hence, no qualifies as an existential determiner. The reader can easily verify that all the weak determiners exemplified in (6) (a, three, some, many) pass the test in (9). By contrast, consider the application of the test to the determiner every: (11)

a. b.

Every snake is dangerous. Every dangerous snake exists.

Suppose that there are 80 snakes, 45 of which are dangerous. In this case, sentence (11a) is false but sentence (11b) is true. Therefore, the two sentences are not equivalent: the determiner every does not pass the test for existentiality. The same holds for the other strong determiners listed in (7). The distinction between strong and weak noun phrases must be further refined because, as already noted by Milsark, the weak determiners listed in (6) also allow for a ‘proportional’ reading, in which the restriction set is already familiar in the discourse context. This reading emerges unambiguously when these determiners introduce a partitive of-PP embedding a definite description, as in (12). (12)

Two of the snakes are dangerous.

Diesing (1992) traced the difference to the fact that in this use, the restriction set is presupposed to be non-empty, and therefore labeled these noun phrases presuppositional indefinites. She also pointed out that even weak NPs that are not overtly partitive (e.g. two snakes) allow for a presuppositional interpretation in certain contexts. Enç (1991) characterized this reading of weak NPs as specific, arguing that the restriction set is not simply presupposed to be non-empty, but it is familiar, i.e. already introduced in the previous discourse context. We agree that in most actual contexts of utterance the restriction set will be 6

More explicitly: recall that no expresses the relation of set disjointness; the truth conditions of (10a) require that the set of snakes and that of dangerous things have no elements is common. This holds precisely in those circumstances in which the universe of discourse has no element in common with the set of dangerous snakes, which means that the set of dangerous snakes is empty.

Definiteness Effect and Unaccusative Subjects

23

familiar in Enç’s sense, or incorporate a contextually supplied covert domain restriction (in the sense of von Fintel 1994); however, both of these properties imply that the restriction set is non-empty, and therefore, Diesing’s presupposition is sufficient for our current purposes in order to distinguish the two relevant classes of noun phrases. Diesing (1992) – building on Kamp (1981), Heim (1982), Higginbotham (1987) – argued that noun phrases introduced by a weak determiner are interpreted as quantifiers only when they have a presuppositional interpretation; in contrast, when they are nonpresuppositional – as in the examples in (6) – they are not inherently quantified, but they undergo existential closure.7 Since presuppositional indefinites like that exemplified in (12) are not allowed as i-subjects, DE is sensitive to the nature of the indefinite noun phrase, which must be a nonpresuppositional indefinite.

2.1. The Narrowest Scope Constraint In this section, we refine the semantic characterization of i-subjects by also considering their scopal interactions. The literature on indefinites and on their unexpected scope properties is too rich to be summarized here (see, a.o., Abusch 1994, Kratzer 1998, Reinhart 1997, Schwarzschild 2002, Winter 1997). What is immediately relevant for our current purposes is the observation that i-subjects, contrary to weak noun phrases in preverbal subject position, obligatorily take narrowest scope with respect to other clause-mate operators, e.g. a modal or negation:8 (13)

a.

b.

7

Potrebbe entrare un uomo dalla finestra. might enter a man from-the window ‘It might be the case that a man (whatsoever) enters from the window.’ (* >might) Un uomo potrebbe entrare dalla finestra. a man might enter from-the window ‘A man might enter from the window.’ (—  >might)

See also Ladusaw (1994) for elaboration of Diesing’s proposal. The mechanism of existential closure in further discussed in section 4.2. 8 The scope judgements hold when the sentences are realized with no intonational break; see note 1 above.

Chapter One

24

(14)

a.

b.

? Non sono stati messi dei libri sul tavolo. not have been put.M.PL some books on-the table ‘It is not the case that any books were put on the table.’ (*some >not) Dei libri non sono stati messi sul tavolo. some books not have been put.M.PL on-the table ‘Some books were not put on the table.’ (—some >not)

In (13a), the indefinite noun phrase un uomo can only be interpreted in the scope of the modal potrebbe: the speaker does not presuppose the existence of a specific man, but merely asserts that some man or other could enter through the window. By contrast, in (13b) the unmarked interpretation is that a specific man is such that he could enter from the window.9 Similarly, (14a) can be paraphrased as ‘it is not the case that any books were put on the table’, whereas in (14b), there is a familiar (nonempty) set of books, some of which were not put on the table. A parallel difference emerges in French: (15)

a.

b.

(16)

a.

Il pourrait éclater un scandale demain. it might break-out a scandal tomorrow ‘It might be the case that some scandal or other breaks out tomorrow.’ (* >might) Un scandale pourrait éclater demain. a scandal might break-out tomorrow ‘A (certain) scandal might break out tomorrow. (—  >might) Il

n’est pas arrivé trois filles. NEG-is NEG arrived three girls ‘It is not the case that there arrived three girls.’ (*3 >not) Trois filles ne sont pas arrivées. three girls NEG-are NEG arrived.FPL ‘Three girls didn’t arrive.’ (3 >not) EXPL

b.

(15b), but not (15a), is about a specific scandal which might break out tomorrow; (15a) merely reports the possibility that some scandal or other 9

A non-specific interpretation akin to that of (13a) is marginally possible if the preverbal subject is under focus. Another possible interpretation, irrelevant here, is a generic interpretation of the weak preverbal subject; we refer to Diesing (1992) and Kratzer (1995) for discussion.

Definiteness Effect and Unaccusative Subjects

25

breaks out. (16b), contrary to (16a), is about three members of a set of girls whose existence is presupposed in the context; compare also the parallel data from English:10 (17)

a. b.

(18)

a. b.

There might break out some scandal tomorrow. (* >might) Some scandal might break out tomorrow. (—  >might) There hadn’t arrived many cars. (*many >not) Many cars hadn’t arrived. (many >not)

Thus, the generalization seems to be that i-subjects constrained by the DE cannot take scope over a higher operator (such as negation or a modal). Furthermore, note that indefinites are in general able to exceptionally take wide scope out of scope islands, like e.g. an if-clause. To exemplify, the sentences in (19) can be taken to mean that there exists one (specific) relative of mine such that, if s/he died, I would inherit a fortune: (19)

a. b.

If a relative of mine died, I would inherit a fortune. Se un mio parente morisse, erediterei if a my relative died.SBJV inherit.COND.1SG una fortuna. a fortune

By contrast, the wide scope interpretation is unavailable for an i-subject, as shown in (20): these examples can only be taken to mean that, by some awkward rule, I would inherit a fortune in case a relative of mine whatsoever died (independently of whether I actually have any living relatives):

10 There sentences in contemporary English are really productive in the existential construction (on which see section 2.3); with unaccusative verbs they are often felt to be a bit archaic and to belong to a high, somewhat literary style. Interestingly, however, speakers clearly detect contrasts and have judgements on them, as in the examples that we report here in (1b), (17a), (18a) and (20)a.

Chapter One

26

(20)

a. b.

If there died a relative of mine, I would inherit a fortune. Se morisse un mio parente in un incidente, if died.SBJV a my relative in an accident erediterei una fortuna. inherit.COND.1SG a fortune

This scopal evidence shows that the DE affecting i-subjects must be strengthened to (21): (21)

I-subjects are scopeless weak noun phrases.11

2.2. The Constraint on Definiteness of TEC Subjects Up to now, we have been assuming that the DE only concerns i-subjects of unaccusatives (and passives), as it is (mainly) the case in languages like English and French (on existentials, see section 2.3):12 (22)

11

a. b.

*There talked to John a man. /*There talked a man to John. *Il a parlé à Pierre trois filles. /*Il a parlé trois filles à Pierre. (Belletti 1988: 4, (6b); 5, (8b))

The relationship between scopelesness, non-presuppositionality, and a ‘thetic’ interpretation of the sentences is discussed in detail in Ladusaw (1994); we do not dwell of this point here for reasons of space, but we merely note that Ladusaw’s proposal is consistent with the semantic analysis of i-subjects that we develop in section 4.2 below. 12 In Italian, subject inversion with other verb classes is allowed, but it involves licensing in a vP-peripheral position; see section 5 for more discussion. It is sometimes noted in the literature that the French il construction may not be totally excluded with some intransitive/unergative verbs (e.g. Legendre & Sorace 2003). These structures may be amenable to the account hinted at in Belletti (1988) for similar cases in West Flemish, discussed in Haegeman (1986). Essentially, the proposal there was in terms of a possible extension to a higher argument of a VPinternal Case assignment such as partitive, as nominative is taken by the expletive er in preverbal subject position in these sentences, which is analyzed as a subject clitic by Haegeman (1986); note that also il in French has the status of a subject clitic. Alternatively, see also the discussion below on the slightly different type of constraint on definiteness found in so called TEC environments. It is not unconceivable that the French il construction with intransitives/unergatives gives rise to a different type of constraint on definiteness along the lines in (27) in the text from Icelandic. This is a rather subtle question, which we leave open here, generally focusing our attention on the core cases of DE with unaccusatives (and existentials).

Definiteness Effect and Unaccusative Subjects

27

However, as is well known Icelandic, West Flemish and German allow for the so called Transitive Expletive Construction (TEC), involving an expletive and a transitive verbs (see, a.o., Belletti 1988: 12-15; Bobalijk & Jonas 1996; Chomsky 1995: 342-348; 272-276); it is generally assumed that the external argument of the TEC is also subject to the DE: (23)

a.

Es

hat has geküsst. kissed Es hat EXPL has die Blumen the flowers EXPL

b.

(24)

a. b.

* Í gær yesterday Í gær yesterday

ein Mann/*?der Mann die Marie a man/ the.NOM man the.ACC Marie

ein Mann/*?der Mann dem Paul a man / the.NOM man the.DAT Paul gegeben. given (German; Belletti 1988: 14, (30b-c))

kláruðu sennilega þessar mys ostinn. finished probably these mice the.cheese klaruðu sennilega margar mys ostinn. finished probably many mice the.cheese (Icelandic; Bobalijk & Jonas 1996: 196, (3a-b))

These contrasts seem to undermine our starting hypothesis (5), according to which the special licensing property which correlates with the DE is restricted to internal arguments sitting in their thematic, first Merge position.13 As a matter of fact, the TEC subject is an external argument; as argued by Bobalijk & Jonas (1996), it even occupies a position external to the verbal projection (vP), since it necessarily precedes negation and a shifted direct object (25), as well as the low adverb completely (26): (25)

a.

borðuðu margir strákari bjúgunj ekki ate many boys the.sausages not ti (öll) tj . (all) ‘Many boys didn't eat (all of) the sausages.’ * það borðuðu bjúgunj ekki margir strákar EXPL ate the.sausages not many boys (öll) tj. (all) það

EXPL

b.

13 On the licensing of direct objects of transitive verbs, see notes 30,44 for some general remarks.

Chapter One

28

(26)

a.

það

luku sennilega einhverjir stúdentar finished probably some students alveg verkefninu. completely the assignment * það luku sennilega alveg einhverjir there finished probably completely some stúdentar verkefninu. students the assignment (Icelandic; Bobalijk & Jonas 1996: 212-214, (21), (23)) EXPL

b.

However, it can be shown that the ‘definiteness effect’ emerging on the subject external to the verb phrase in the TEC is actually weaker than the one emerging with unaccusative i-subjects. First, the translation of example (25a) clearly shows that the transitive subject is interpreted outside the scope of negation, contrary to what we observed in the case of unaccusative isubjects (see section 2.1). Second, while the definiteness restriction (21) excludes all presuppositional noun phrases, Vangsnes (2002) pointed out that in Icelandic, the TEC structure only excludes definite subjects, but allows for other strong/presuppositional noun phrases, like partitives embedding a definite description (cf. (12) above) or universal quantifiers: (27)

a.

það

hefur has það hefur EXPL has mýsnar. mice.the það hefur EXPL has mýs. mice það hafa EXPL have mýs. mice það hafa EXPL have mýs. mice EXPL

b.

c.

d.

e.

köttur/*kötturinn /*Kalli étið cat / cat.the / Kalli eaten *þessi köttur/*köttur Péturs this cat /cat Peter’s

mýsnar. mice.the étið eaten

sérhver köttur /einn af köttunum étið each cat /one of cats.the eaten

allir kettirnir /báðir kettirnir all cats.the /both cats.the

étið eaten

nokkrir kettir/ sumir kettir some cats /some.of.the cats

étið eaten

(Icelandic; Vangsnes 2002, (12))

Definiteness Effect and Unaccusative Subjects

29

Interestingly, Vangsnes (2002, (7)) showed that also in Icelandic, expletive structures with a low, VP-internal i-subject obey the radical definiteness restriction and exclude all presuppositional noun phrases: (28)

a.

það

hafa verið nokkrir kettir/* sumir kettir have been some cats / some.of.the cats í eldhúsinu. in kitchen.the það hefur verið * sérhver köttur /* einn af köttunum EXPL has been each cat / *one of cats.the /* kötturinn í eldhúsinu. / * cat.the in kitchen.the EXPL

b.

Vangsnes stressed that the same expletive, það, is involved in the TEC and in this structure; hence, the strength of the definiteness effect cannot be determined by properties of the expletive, but it is determined by the position of the subject, intermediate in (27) vs. VP-internal as a first merged internal argument in (28a, b). An initial survey of DE effects in the German TEC confirms that partitive presuppositional indefinites, i.e. partitives embedding a definite description, are indeed allowed, whereas definites are excluded; the status of universal quantifiers appears to be mixed.14 Pending further investigation,

14

According to five native speakers whom we consulted, the following similar pattern emerges in German (although judgments are subtle, especially on the borderline cases, such as e.; clear definites such as f. and g. or a noun phrase introduced by a definite determine are, however, excluded in this type of structure by all our informants): (i). a. Es hat jeder Gast die Braut geküsst. EXPL has each guest the bride kissed b. Es hat einer der Gäste die Braut geküsst. EXPL has one of-the guests the bride kissed c. Es haben alle Gäste die Braut geküsst. EXPL have all guests the bride kissed d. Es haben beide Gäste die Braut geküsst. EXPL have both guests the bride kissed e. ? Es haben einige Gäste die Braut geküsst. EXPL have some guests the bride kissed f. ?* Es hat dieser Mann die Braut geküsst. EXPL has this man the bride kissed g. ?* Es hat Peters Bruder die Braut geküsst. EXPL has Peter’s brother the bride kissed

Chapter One

30

we take these provisional results to establish the point that TEC subjects occupy a clause-intermediate position different from the internal argument position of i-subjects, and that in this position they do not obey the same radical definiteness restriction (21) as i-subjects, but rather a distinct and weaker restriction, which can be characterized as follows: (29)

The intermediate subject position occupied by TEC subjects 15 excludes definite noun phrases.

We leave for future research an explanation of this constraint; with respect to the current discussion, the relevant point is that TEC subjects do not undermine our initial hypothesis, with the DE a property of i-subjects formulated as in (21).

2.3 The Definiteness Effect of Existential Sentences One core case of DE is found in existential sentences like there be sentences in English of the type illustrated in (30); it is precisely through this type of sentences – systematically discussed for the first time in Milsark (1974, 1977) – that the whole phenomenon of DE was first given the descriptive and theoretical prominence that it still has: (30)

a. b.

There is a man (in the garden). There are cats (in the garden).

The DE in (30) has the same flavor as the cases reviewed in (1), (2) and (4) in the introduction and in section 2.1; the sole difference is that the verb of (30) is the copula be. 16 Similarly, in Italian, existential sentences contain the copula, accompanied by clitic ci: A systematic cross-linguistic investigation of definiteness effects in the TEC is beyond the scope of this article. 15 In slightly more technical terms, this position excludes quantifiers that are principal ultrafilters (in the sense of Barwise & Cooper 1981, 174). A principal ultrafilter is a family of sets each of which includes a given entity a ({XŽDe~aX}). A principal ultrafilter can thus be reduced to an entity-type denotation, cf. Partee (1984). 16 There are well known exceptions to DE. In existentials one exception is the possibility of a ‘list reading’, which is taken to be compatible with Partititive case. Similarly for the so called uniqueness intepretation with other unaccusatives. (i.e. È caduto il portafoglio, lit. fell the wallet), Belletti (1988:15-16); on uniqueness definites in existential there contexts, see Abbott (2001, examples (25)-(26)).

Definiteness Effect and Unaccusative Subjects

(31)

a. b.

31

C’è un uomo (in giardino). Ci sono gatti (in giardino.)

Ci of existential sentences is the same word as the locative clitic ci, meaning “in/to XP”, corresponding to a locative PP (32a,b). The same holds in English, where expletive there of existential sentences (30) (and of 1b), also corresponds to the word expressing a deictic locative PP, (32c,d, which translate 32a,b): (32)

a. b. c. d.

Vado/Vivo a Roma. Ci vado/vivo. I’ll go to/live in Rome. I’ll go/live there.

That existential sentences involve a locative-like element is a property found in a variety of languages. A thorough (re)visitation of the general properties of existential sentences, ultimately of the question of what makes sentences like (30) and (31) “existential”, would take our discussion too far afield. Here we focus on the relation between existential sentences like (30), (31) and sentences like (1), (2) and (4); our crucial assumption is that they both involve an unaccusative verb phrase. Building on some recent insights from McCloskey (forthcoming), based on properties of the Irish existential predicate ann, we can assume that an element of “Location” providing the necessary context for the existential meaning to be composed is present in existential clauses (see McCloskey’s “Instantiate” related to McNally 2009 and integrated with a definition of context à la Francez 2010).17 This is why it is not always easy to tease apart a real locative PP (often deictic, as in 33b) from the “Location” A further apparent exception to the DE is found in examples such as (i): (i)

All’improvviso suddenly

sono entrati i pompieri dalla finestra. are.3PL entered.MPL the firemen from-the window

Following Zamparelli (2002), we assume that in Italian a plural definite description denotes a kind, and in a context like (i) it can be shifted to the corresponding property and existentially bound, resulting in an indefinite interpretation: thus, the definite description in (i) denotes unspecified instances of the fireman kind. This interpretive route is not available for plural definite descriptions in English. 17 It is sometimes proposed that possibly all unaccusatives contain a locative, which may remain silent in the form of a covert PP in their structure (Moro 1997, Alexiadou & Shäfer 2010). The hypothesis in the text is different in that it singles out existentials as involving a particular Location voice in the verbal shell (different from a covert PP). See (35).

Chapter One

32

present in existential expressions (33a). Crucially, however, the DE is only manifested on the post-verbal noun phrase of the existential sentence. (33)

a. b.

There is a man/*the man (in the garden). A man/The man is there (, in the garden).

As recently discussed by Cruschina (this volume), Italian sentences involving clitic ci prima facie also give rise to the same uncertainty. However, much as in the English case in (33), only existential sentences manifest DE, whereas sentences involving locative ci and a post-verbal definite subject do not; in contrast with English (33a), a definite postverbal subject is possible in Italian as the focus of new information, as illustrated in (34b), contrasting with (34a) (on post-verbal new information subjects and the lack of DE constraining their distribution, see also section 5): (34)

a. b.

Ci sono gatti (in giardino). there are cats (in garden) C’è Maria/ tua sorella (, in giardino) there’s Mary/ your sister (in garden) (>> answer to ‘Who is there in the garden?’)

Granted that much, we are then led to assume that the verb of existential sentences (e.g. be) projects the same structure as any unaccusative verb, for relevant respects.18 In particular, the post-verbal noun phrase is the internal argument of the existential verb. The verb can be lexicalized as the copula, as in the English and Italian cases illustrated above, or as the related verb have (Kayne 1993), as is the case in various other languages, e.g. French (as in 2a) and Spanish.19 Let us refer to it with the general term be. Crucially, a location expression is also typically required (as explicitly revealed by Irish ann, according to McCloskey’s account). The following structure in (35) gives a concrete implementation of our proposal: the existential verb projects an unaccusative verb phrase, which does not contain any external argument; it contains instead an internal argument, as originally proposed by Williams (1984), and optionally also a PP, as it is typically the case with unaccusatives. Crucially, the functional 18

See Mc Closkey (forthcoming) for the observation that the existential meaning does not necessarily involve use of a verb as is the case for certain existential expressions solely containing the element ann – glossed as in it – in Irish; this possibility is represented crosslinguistically, Russian being another case in point. 19 It can also be lexicalized as a different verb, e.g. in the German es gibt construction.

Definiteness Effect and Unaccusative Subjects

33

structure of the existential unaccusative verb phrase is richer in that it also contains the expression of a Location voice, realized in a small v right above the phrase obtained after first Merge of be and its internal argument; by hypothesis, the Italian clitic ci is a possible realization of this voice. There present in existential sentences could have the same origin (see Moro 1997: 2.4): it could be analyzed as the specifier of the locative voice. The proposal is similar in this respect to Deal’s (2009) analysis of there as an expletive inserted low in the structure as the specifier of a small-v.20

20

The proposed analysis also shares with Deal’s (2009) approach the insight that the DE is derived as an effect of the noun phrase remaining in the internal argument position where it receives a weak interpretation. The analysis sketched in (35) is not immediately compatible with Moro’s (1997) proposal, according to which existential be selects a small clause complement on a par with copular be (e.g. John is nice/a doctor/in the room), with ci/there the predicate of the small clause raising into the subject position of the clause (or relevant clitic head in the case of ci). See also McCloskey (forthcoming) for critical discussion. Given the richer conception of the verbal functional structure we are working with (richer than the one assumed at the time of Moro 1997), since ci is the expression of the Location voice, it is in fact part of the predicate; in this respect the idea shares some of the insights of Moro’s approach. However, the crucial point for us is that existential be is an unaccusative verb, and the post-verbal noun phrase is its internal argument, along the lines of Williams’ (1984) (also Belletti 1988; fn.5; Higginbotham 1987, quoted in this respect in Moro 1997). In Lasnik (1999: 85-90) the small clause analysis of existential be is reconciled with Partitive Case under the idea that the predicate of the small clause raises into be to form a complex predicate, able to assign Th-trole and inherent Case. In Lasnik’s view, however, there is not a predicate but an expletive (as in traditional accounts, and more recently Deal 2009) and the responsible for Partite Case assignment/checking is be. This aspect of Lasnik’s proposal is very close to ours.

Chapter One

34

2 2 2 (PP) v 2 ci V IA21 be

(35)

T

A number of phenomena show that the indefinite post-verbal argument of existential clauses is a well-behaved internal argument; we exemplify from Italian. Firstly, bare plural noun phrases can be argued to be the overt manifestation of Partitive Case in a language like Italian (Belletti 1988: 29). As we discuss in section 4.1 (point vi.), Partitive Case (assigned through a functional head F: (48) below) can only access the internal argument position, giving rise to the DE. Since it is assigned to the internal argument of V, this implies that it is assigned in combination with Th-role assignment; this is a crucial property that Partitive Case shares with traditional inherent Case. A direct consequence of this analysis is then that Partitive Case cannot reach the subject position of a small clause complement. As Belletti (1988, (69)) discusses, this is indeed the case, since bare plurals are excluded as subjects of small clause complements of verbs like considerare, ritenere/consider: (36)

Ritengo/Considero [questi studenti / alcuni studenti reckon.1SG/consider.1SG these students / some students /*studenti intelligenti]. /*students intelligent.M.PL

By contrast, a bare plural is perfectly allowed in the post-verbal position of ci existential sentences: (37)

21

Ci sono there are

studenti (in corridoio). students in corridor

We postpone until section 4.1 a more precise proposal on the structural representation of the IA of unaccusatives. See also the following section 4.1 on the presence and role of a functional head that we label F right above the first projection V + IA, carrying number and gender ij-features. For the sake of clarity, at this stage of the discussion we have not integrated F in the representation (35). We do not further elaborate on the possible presence of a small v as a verbalizer in all unaccusatives (as in e.g. Deal 2009); cfr. footnote 31, which would be compatible with this option. Nothing crucial hinges on this decision in the present discussion.

Definiteness Effect and Unaccusative Subjects

35

Hence, the bare plural of the existential construction in (37) behaves indeed as an internal argument.22 Secondly, the post-verbal noun phrase can also be cliticized through ne-cliticization, as illustrated in (38a): this possibility is productively available only to internal arguments/direct objects (Burzio 1986, Belletti & Rizzi 1981). The same possibility is predictably completely excluded for the subject of a small clause (38b):23 (38)

a. b.

(Studenti) Ce ne sono___ (in corridoio). (students) there of-them are (in corridor) *(Studenti) Ne considero/ritengo___ (students) of-them consider.1SG/reckon.1SG intelligenti. intelligent.MPL

Furthermore, in Italian past participle agreement is obligatory with the postverbal noun phrase of ci existential sentences, also when it is a bare plural (39a; past participle agreement always is obligatory with the postverbal noun phrase of unaccusative verbs, 39b): (39)

22

a.

Non ci sono mai stati molti gatti/gatti (in giardino) Not there are ever been.MPL many cats/cats (in garden)

b.

Non sono not are

mai ever

arrivati molti gatti/gatti (in giardino) arrived.MPL many cats/cats (in garden)

Moro (1997, 116) claims that when ci incorporates into be, this endows the copula with selectional properties, so that the subject of the small clause of which ci is the predicate counts as being selected by be. This amounts to claiming that this noun phrase must be treated in fact as an internal argument of be, as in the account defended here. Kalulli (2008) and references cited therein on the difference between bare plurals (possible) and bare singulars (impossible) in English there existential sentences. 23 The possibility of ne-extraction in (i) is expected as a case of extraction from the accusative (not partitive), Case marked subject (under ECM/subject to object raising) of the complement of considerare/ritenere: (i) Ne considero/ritengo molti ___ intelligenti of-them consider.1SG/ reckon.1SG many intelligent.MPL

Chapter One

36

This is another property that the post-verbal noun phrase of ci existential sentences shares with internal arguments (unaccusative and passive postverbal subjects in particular).24 In conclusion, we assume with much classical literature (e.g. Williams 1984; see also footnote 20) that be of existential sentences is an unaccusative verb and that DE concerns its internal argument, in the same way and for the same reasons it does with unaccusative verbs in general.25

3. Intermezzo: Indefinites in Maori The cluster of properties that characterize English, French and Italian isubjects suggest that licensing an internal argument in situ gives rise to both syntactic and semantic defectiveness. From the syntactic viewpoint, the i-subject does not seem to Agree with a functional head endowed with a phi-complete probe (resulting in ‘structural Case’ licensing); from the semantic viewpoint, the argument cannot be a presuppositional quantifier. There is a long-standing intuition that these two aspects of defectiveness must be related to each other; in Belletti’s (1988) proposal, the relationship was captured by the inherent semantics of Partitive Case. Yet even the Partitive Case hypothesis does not address the deeper question of why licensing in the thematic position should correlate precisely with the observed semantic restrictions. As a step towards addressing this question, an important observation is that the same cluster of properties emerges in an apparently unrelated set of data, namely Maori he-indefinites as described by Chung & Ladusaw (2004: 21-73). In this section we summarize the relevant evidence from Maori. Maori has two types of indefinite noun phrases, introduced by the articles he and tƝtahi; the former, contrary to the latter, is not inflected for number: (40)

24

Kua riro T be.taken

[he pukapuka a Mere]. a book of Mere

In §4.1 we will argue that a functional head (F) which is involved through Agree in the assignment of partitive Case to i-subjects is also implicated in the obligatory past participle agreement of number and gender, thus explaining this correlation. 25 And also transitive verbs in the passive voice, as in the examples (2c), (4b) above.

Definiteness Effect and Unaccusative Subjects

(41)

37

Kua riro [tƝtahi pukapuka a Mere]. T be.taken a book of Mere ‘A book of Mere’s was taken.’ (Chung & Ladusaw 2004: 26-27, (10a-b))

Both types of noun phrase allow for a distributive interpretation in the scope of a universal quantifier, showing that they indeed receive an indefinite interpretation: (42)

a.

b.

I ia tau ka pǀti-tia [he kaiwhakahaere hou]. in each year T elect.PASS a chairperson new ‘Each year a new chairperson is elected.’ I ia tau, e ngaro ana [tƝtahi tangata] i te ngahere in each year T be.forgotten a person in the bush ‘Each year, someone gets lost in the bush.’ (Chung & Ladusaw 2004: 35, (21a-b))

However, tƝtahi-indefinites are also able to take wide scope with respect to other elements in the clause, in particular with respect to negation (which is realized by means of the negative copula kƗore), as shown in (43a). In contrast, he-indefinites necessarily take narrow scope w.r.t. negation, as shown in (44).26 (43)

a.

b.

26

KƗore [tƝtahi tangata] i waiata mai. ( >neg) T.not a person T sing to-here ‘A particular person didn’t sing.’ KƗore [tƝtahi tangata] e mahi mƗ-na (neg >) T.not a person T work T.of-him ‘No one would work for him.’ (Chung & Ladusaw 2004: 40, (27a), (23c))

The contrast between (43a) and (44) is reminiscent of the French contrast between the partitive article des in (i), which can be interpreted outside the scope of negation, and the article-less de-phrase in (ii), which cannot (see Kayne 1981, 95 ff.): (i) J’ai pas lu des livres 1sg.cl have.1sg not read of+the.pl books (ii) J’ai pas lu de livres 1sg.cl have.1sg not read of books

Chapter One

38

(44)

KƗore [he tangata] i waiata mai. T.not a person T sing to-here ‘No one at all sang.’ *‘A particular person didn’t sing.’ (neg > , *  > neg) (Chung & Ladusaw 2004: 40, (28a))

This shows that he-indefinites are just scopeless indefinites, displaying the behaviour that we identified as typical of our i-subjects. And indeed heindefinites, contrary to tƝtahi-indefinites, are restricted to occupying the internal argument position, as in (40) and (42a) above, and cannot occur as subjects of transitive or unergative verbs:27 (45)

a.

b.

(46)

a.

b.

27

*I whiu [he wahine] i tƗna mǀkai T throw a woman DO her pet ki te moana. into the ocean ‘A woman threw her youngest child into the ocean.’ (Chung & Ladusaw 2004: 57, (49)) ?* E mahi ana [he tangata]. T work a person ‘A man is working.’ (Chung & Ladusaw 2004: 57, (51b)) I whiu [tƝtahi wahine] i tƗna mǀkai T throw a woman DO her pet ki te moana. into the ocean ‘A woman threw her youngest child into the ocean.’ (Chung & Ladusaw 2004: 56, (48a)) E mahi ana [tƝtahi tangata]. T work a person ‘A man is working.’ (Chung & Ladusaw 2004: 57, (50b))

The reader may have noticed that example (44) contains a he-indefinite which is apparently the subject of an unergative verb. From the data discussed by Chung & Ladusaw, it seems that examples of this type always involve negation, which is a verb, giving rise to a bi-clausal structure; thus, it is likely that the negative copula plays a crucial role in licensing the he-indefinite here (cf. Chung & Ladusaw 2004, 43-44).

Definiteness Effect and Unaccusative Subjects

39

Thus, Maori shows an overt distinction between scopeless indefinites and other indefinites, confirming that the former constitute a morphosyntactically separate subclass. The crucial insight offered by Chung & Ladusaw is that scopeless indefinites are not saturating arguments, but rather, they are interpreted as further specifying the predicate. Building on their proposal, in the following section we will elaborate an analysis of i-subjects along the following lines: (i)

Unaccusative i-subjects remaining in the first Merge position are unaccessible to the T probe endowed with phi-features including person; hence, they cannot value an unvalued Case feature. (ii) Consequently, they can remain in situ only if they are syntactically defective, lacking the D projection, which is the introductor of an unvalued Case feature. (iii) Since i-subjects licensed in situ lack the D projection, they cannot be interpreted as saturating arguments. (iv) I-subjects, are interpreted at the interface via application of the rule of Predicate Restriction (Chung & Ladusaw 2004).

4. Licensing I-Subjects 4.1. Syntactic Licensing I-subjects of unaccusatives are merged in the internal argument (IA) position in the verb phrase; the argument structure of unaccusatives is such that no external argument (EA) is merged in a higher v*P28 spec position. This is the core of the classical unaccusative hypothesis (Perlmutter 1978, Burzio 1986), incorporating more recent insights and updating terminology. As mentioned above, under current minimalist assumptions an Agree type relation is normally established between a T-probe head endowed with ij-features and the external argument of transitive and intransitive vPs. The reason why Agree can only be established with the EA as the goal of the relation is straightforwardly locality, Relativized Minimality/ Minimal search (Rizzi 1990, 2004, Chomsky 2005): the IA of a transitive 28

The v* notation is meant to indicate a higher level of the verbal spine, without committing to specific assumptions on the exact layering of the verbal projection in the larsonian-type structure (Larson 1988). The lexical level may be identified with the root lexical level (R in Chomsky 2014) which then combines with the functional verbal structure v*.

Chapter One

40

verb phrase cannot be the goal of the Agree relation due to intervention of the EA, which is closer to the probing head T. This Agree relation can license an EA in a low position in the vP periphery (section 5 below), resulting in a linearly post-verbal subject due to further syntactic raising of the lexical verb into the T head(s). Note that such ‘peripheral’ post-verbal subjects are not constrained by the DE: (47)

a. b.

Ha parlato il presidente. has spoken the president (Il libro) l’ha recensito quel giornalista the book it has reviewed that journalist

As pointed out in connection with (3) above, a definite postverbal subject is also possible with unaccusatives, under conditions that we will make explicit in detail in section 5. Crucially, in these conditions the post-verbal noun phrase is not in the IA position, but in a position peripheral to the verb phrase, the same as in (47), which can be accessed independently of the class to which V belongs. We postpone until section 5 a detailed review of our assumptions on the syntax and (discourse) interpretation of ‘peripheral’ subjects in sentences containing transitive and intransitive verbs, as well as in sentences with unaccusative verbs and a definite postverbal subject. The question to be investigated now is how the i-subject of unaccusatives is licensed. Given the above standard assumptions on the argument structure of unaccusatives, absence of the EA in the unaccusative verb phrase could open up the possibility for the IA to become accessible to the probing T; to the extent that this relation also correlates with the availability of Nominative Case, one would not expect any special behavior of the i-subject of unaccusatives, in particular in terms of its definiteness along the lines discussed in the previous sections. The post-verbal i-subject of an unaccusative should behave exactly as the ‘peripheral’ post-verbal subjects of (47) (and (3)). However, it does not. We now address the licensing of the indefinite i-subject of unaccusatives and submit a proposal summarized in the following steps: i. Assume that the low part of the verb phrase is contained within a functional projection whose head we label F; presence of such head carrying number and gender features is made visible in agreeing past participles (e.g. with object clitics and transitive verbs in the active voice, and with the IA in passives and unaccusatives; see Belletti 2006, forthcoming for an overview on past participle

Definiteness Effect and Unaccusative Subjects

41

agreement; Collins and Thráinsson 1993 for a related proposal). F is a defective head, lacking the feature person. ii. Being a ȭ-related head, F counts as an intervener with respect to T and blocks the Agree relation from the higher ȭ-complete T probe, which is also endowed with the feature person; consequently, the IA cannot be probed as the goal of the relation with T;29 Nominative is thus not accessible to this position. iii. The IA/i-subject can only be probed by the defective head F; hence it can only be a defective noun phrase, not requiring to enter the Agree relation with ȭ-complete T. iv. Assume that the noun phrase is defective in that it lacks the high D (type) layer, containing the feature person; we label the defective noun phrase NumP for concreteness (see also Kalulli 2008 for related proposal on English). v. This amounts to claiming that the weak indefinite noun phrase lacking the D layer is in fact not a real argument of the verb, since D is the site of referentiality and argumenthood (cf. Longobardi 1994 and subsequent correlated literature, including Sheehan & Hinzen 2011). Assume that D contains the feature person. vi. As far as Case on the i-subject is concerned, we can continue to assume that Partitive Case is/can be carried by the defective weak indefinite, NumP. This is so since we assume Partitive Case to be the manifestation of an Agree relation with the defective F head, which is low in the structure (by assumption, immediately above VP). 30 vii. Partitive is thus not linked to a specific thematic role like traditional inherent(/lexical) Cases (e.g. Dative); it is however assigned to the IA, hence in conjunction with Th-role assignement, and in this respect, it shares a crucial property of traditional inherent Cases (cf. Belletti 1988, 27-31; footnote 2 above).31 As will be discussed 29 We assume that in languages like French, all the ȭ-features of T are valued under spec-head agreement with the pronominal expletive in spec/TP, whereas in languages like English and Italian, T agrees for number with F and for person with the expletive in Spec,TP (there in English; pro in Italian: cf. Rizzi 1982, Safir 1987 and more recently Cardinaletti 2004, Belletti 2005). 30 Partitive Case is then interpreted as the (only) Case compatible with weak indefinites. It is the Case of the IA, as is visible in languages which overtly realize it (e.g. Finnish, as discussed in the original work by Belletti 1988 summarized above). In contrast, definite objects are DPs which enter an Agree relation with a low probe complete of the person feature. 31 The idea shares the spirit of Lasnik’s (1999: 86) insight that Partitive Case, and possibly inherent Case in general, involves a (agreement-type) relation with a

Chapter One

42

in detail in §4.2, the semantic interpretation of the defective NumP is only possible when the NumP is the internal argument of V. The structure in (48) illustrates the proposal: (48)

TP 3 proNom 3 T[pers, num, Nom] FP 3 F[num,gen] VP

*

¥

3 V IA/NumP partitive

We assume that the unaccusative verb phrase does not generally contain a functional v head, the introducer of the external argument (but see footnotes 3, 21).32 Note that the i-subject, being the IA first merged with the lexical verb, meets the syntactic local condition for noun-incorporation, in the classical sense of Baker (1988). Hence, the i-subject is connected to the predicate in the strictest way. This is the syntactic condition tightly linked to the semantic interpretation to be discussed in the next section, according to which the i-subject is interpreted through Predicate Restriction following the core insight of Chung & Ladusaw (2004). Ultimately, it is part of the predicate, hence it is not a standard argument. We propose that defectiveness of the i-subject of an unaccusative verb is brought about by lack of (at least) the highest D portion in the weak functional head (AgrO, in the system assumed there). Hence, there is a functional component also with partitive/inherent Case. The difference with structural Case resides in the concomitant assignement/checking of a Th-role. In the system sketched out in the text F is the structural/functional component of partitive Case. 32 Should a small v head also be present in other unccusative verb phrases as hinted at in footnotes 3 and 21, the locality problem illustrated in (48) would anyway be created by the ȭ-related (gender and number) head F (not by the light verb-type head, which is not a ȭ-related head). We leave open her the question whether the unaccusative verb phrase is or is not a phasal domain, as our approach has no direct bearing on the issue (Deal 2009 for discussion).

Definiteness Effect and Unaccusative Subjects

43

indefinite noun phrase. We will therefore assume that weak determiners, on their non-presuppositional interpretation, do not fill the D head, but rather the lower projection, which we dub NumP (for relevant discussion see Higginbotham 1987, Zamparelli 1999 a.o.). Assuming that D is the functional domain expressing the person feature of a noun phrase (Longobardi 2008, a.o.), to the extent that D is usually taken to be the site of reference and argumenthood (Longobardi 1994, 2008 and related references), this assumption amounts to claiming that the person feature is the crucial feature to license a referential argument, as originally argued in Rizzi (1986: 543, where the proposal was phrased in a pre-DP model; we return to the notion of “referential argument” in section 4.2). In this respect, the person feature has a different status from number and gender; these features are contained in F and establish an Agree relation with the indefinite i-subject, as overtly witnessed by examples like the following in Italian (see also (39), section 2.3): (49)

Sono entrati (are)enteredpl masch

(alcuni) invitati some guestspl, masch

dalla finestra. from the window

The person feature is contained in the T head. We assume that in the lack of a value attributed through Agree, the unmarked/default 3rd person values the feature in T.33

4.2. Semantic Interpretation In the previous section we argued that i-subjects remaining in the thematic position can be syntactically licensed only as defective noun phrases (NumP), lacking the D-layer. The latter introduces an unvalued Case feature and requires Agree with a ȭ-complete probe, whereas i-subjects can only Agree with the F head, lacking the person feature. We have adopted the view that the person feature on D is a necessary condition for a nominal projection to qualify as a referential argument. The term “referential” – borrowed from the literature quoted above – although very intuitive, is not entirely accurate, because a full-fledged 33

A slightly different technical implementation would have the person feature in T valued through the spec-head relation with expletive pro. However, if the expletive has no person feature precisely because it is an expletive, i.e. it is not an argument, the only remaining value for T would be the default value as in the text proposal. We assume that the plural number feature may be ‘inherited’ by T via an Agree relation with the F head, as in Vangsnes (2002).

Chapter One

44

argument can not only be referential (denoting an entity of type e), but alternatively it can be a quantifier, denoting a family of sets (those sets that stand in the appropriate relation with the restriction set; cf. the beginning of section 2). Therefore, we adopt instead the notion of saturating argument as characterized in Chung & Ladusaw (2004: 6-10): a saturating argument is one that reduces the valence of the selecting predicate. For instance, in (50) the verb come has the valence for a single argument x; the referential argument John denotes an individual (j), and when it composes with the verb, it saturates the valence represented by x, yielding the set of events of coming by j:34 (50)

a. b.

John came. [Ox.Oe. comec(x, e)] (j) = [Oe. comec(j, e)]

In the technical notation of (50b), the first lambda prefix (Ox) defines a function which takes in input an individual (j) and returns another simpler function (([Oe. comec(j, e)]): the latter characterizes the set of events of coming by John. The saturation of the verb’s valence is made visible by the elimination of the most external lambda-prefix, and by the concomitant substitution of j in the place of x (lambda-conversion). The same valence reduction is observed in the case of a quantificational argument; for the sake of simplicity, we illustrate in (51) with a non-eventive predicate:35 (51)

a. b.

Every boy is lazy. [OX. {y: boyc(y)} Ž X}] ({z: lazyc(z)}) quantifier nuclear scope = {y: boyc(y)} Ž ({z: lazyc(z)})

The quantifier every boy is a function that takes in input a set (indicated by the capital variable X) and returns the value ‘true’ if and only if this set includes the restriction set of boys ({y: boyc(y)}). In our example, the quantifier takes in input the denotation of the predicate, which characterizes the set of lazy people: thus, the sentence is true if and only if 34

For the sake of simplicity we ignore the past tense here. The formula is also simplified in that the denotation of the noun and of the predicate are indicated as sets, rather than as the corresponding characteristic functions. Furthermore, for expository purposes we are also leaving out the intensional dimension. 35

Definiteness Effect and Unaccusative Subjects

45

the set of boys is included in the set of lazy people. Note that the valence of the predicate (indicated by the individual variable z) is “used up” to define the set that constitutes the nuclear scope of the quantifier: technically, the variable z is bound by set abstraction, and it is “no longer available to semantic composition” (Chung & Ladusaw 2004, 3, (4)). In a nutshell: a predicate is something that has an open valence; a saturating argument is something which reduces the valence of a predicate. Chung & Ladusaw propose a third category, that of non-saturating arguments. These are weak non-presuppositional noun phrases – in the sense explained in section 2, called ‘indefinites’ after Diesing (1992) – which syntactically fill an argument position, selected by a predicate, yet fail to reduce its valence. The crucial idea is that a non-saturating argument denotes a property. A special compositional rule, Predicate Restriction, combines it with the predicate in such a way that the property denoted by the non-saturating argument restricts the domain of the function denoted by the selecting predicate in the corresponding argument position. To illustrate, consider a simple VP like (52) (we leave out the external argument for the time being, and we return to it below): (52)

[VP [V meet] [a boy]]

Informally, the verb meet will characterize the set of events of meeting some individual x; the non-saturating indefinite a boy denotes the property of being a boy; the rule Restrict combines these two denotations so as to yield the set of meeting events whose Theme has the property of being a boy (the set of boy-meeting events). The compositional interpretation proceeds as in (53). As shown in (53a), the lexical V introduces a relation between an event e and an entity x which is the Theme of that event; the indefinite object, instead, denotes a property (53b). In (53c), the rule Restrict combines these two denotations by adding to the original function denoted by the V the restriction that its Theme must have the property of being a boy:

Chapter One

46

(53)

a. b. c.

[[ V ]] = Ox.Oe. meetc(x, e) [[ [a boy] ]] = boyc [[ [VP meet a boy] ]] = Restrict ([Ox.Oe. meetc(x, e)], boyc) = [Ox.Oe. meetc(x, e) & boyc(x)] (In prose: the function, whose domain is restricted to boys, which returns for each boy the set of events of meeting him.)

Note that, crucially, the restricted argument position, still marked by x, has remained unsaturated. 36 Since the restricted argument position is left unsaturated, it undergoes Existential Closure on top of the VP level: “...closure must be delayed until a point in the compositional process when further attempts at saturation are impossible in principle. We take this point to be the event level” (Chung & Ladusaw 2004: 125). Interestingly, the compositional mechanism proposed by Chung & Ladusaw yields the cluster of properties characteristic of our i-subjects. Firstly, only weak/indefinite noun phrases, which are neither inherently referential nor inherently quantified, can denote a property, so as to compose via Restrict. Secondly, this analysis accounts for the fact that i-subjects syntactically fill an argument position – specifically, the internal argument position – but nevertheless semantically qualify as ‘defective’ arguments. Finally, the hypothesis that Existential Closure applies at the VP-level yields a non-

36 This is the main difference between this proposal and ‘semantic incorporation’ à la van Geenhoven (1998), McNally & van Geenhoven (1998). For a comparison of the two approaches, we refer to Chung & Ladusaw (2004: 114-118). The most straightforward empirical justification for this analyical choice comes from languages like Chamorro, where an incorporated direct object can be doubled by a saturating ‘extra object’: (i) Si Carmen ga¨i-[ga’] i ga’lagu. Carmen AGR .have-pet the dog UNM ‘Carmen has the dog as pet.’ (Chung & Ladusaw 2004, 109, (70))

In (i), the extra object is the definite description i ga’lagu (‘the dog’): this is allowed precisely because the incorporated object ga’ ‘pet’ is a non-saturating argument, which restricts the internal argument position but fails to saturate it (Chung & Ladusaw 2004: 109-110). On the other hand, in languages like Maori a morphosyntactic licensing constraint conspires to rule out the occurrence of such an ‘extra object’.

Definiteness Effect and Unaccusative Subjects

47

presuppositional interpretation of the non-saturating argument and implies narrow scope w.r.t. any other operator, including negation (cf. section 2.1). We therefore propose the following hypothesis: (54)

I-subjects are interpreted at the interface via Predicate Restriction.

One question that remains to be addressed is why this licensing route is limited to the internal argument position (see Chung & Ladusaw 2004: 5960 for a sketch of a proposal). As mentioned in section 4.1, the configuration that allows for Predicate Restriction corresponds to the syntactic configuration for noun incorporation:37 this is the tightest possible relation with the lexical Verb under First Merge. From this perspective, we can hypothesize that external arguments are excluded from this licensing route because they are not directly related to V. Chung & Ladusaw actually assume that the external argument is selected by the V head, just like the internal arguments.38 Let us assume instead, following Chomsky (1995) and much subsequent work, that the external argument is introduced by the separate head v,, as mentioned in section 4.1 (see also footnotes 3, 21, 32). From a compositional viewpoint, v introduces a thematic relation between an individual and an event, and this relation is combined with the event description contributed by VP; for concreteness, we adopt the rule of Event Identification proposed by Kratzer (1996), which combines the VPs’ event description (Oe.P(e)) with the agent relation denoted by the sister head. (In this notation, e is the usual entity type and s is the type of events.).39 (55)

37

a. Event Identification (EI) f g ĺ h: Oxe. Oes. f(x)(e) š g(e) b. Oy.Oe.agentc(y, e) Oe.P(e) ĺ Oy.Oe.agentc(y,e) š P(e)

Within our approach, it would be possible in principle to stipulate that Predicate Restriction is contingent on incorporation into the verb, and that this option is only available for internal arguments (Baker 1988). However, Chung & Ladusaw (2004: 136-141) show that Maori applies Predicate Restriction both to incorporated direct objects and to he-indefinites of the type exemplified in section 3 above, which do not appear to undergo incorporation; thus, it seems that morphosyntactic incorporation is not a necessary condition for Predicate Restriction. 38 This assumption actually forces them to reorder the lambda prefixes in the predicate’s denotation, so that the restricted internal argument is the last one to be saturated by Existential Closure at the VP level (Chung & Ladusaw 2004: 10). 39 Actually, in Kratzer (1996), the head introducing the external argument is Voice. For our current purposes, Voice can be identified with Chomsky’s v, but see Harley (2012) for relevant discussion of the distinction between these two heads.

Chapterr One

48

Suppose now w that the intternal argumeent undergoess Predicate Reestriction, as in (53) abbove: since thhis argument position p remaiins unsaturated d, the VP does not dennote a properrty of events (type ( ), bbut it is still a relation between inddividuals and events (typee ); thiss is a wrong type for Event Identiification to apply: (56) vP 3 v’ *EI([Oy.Oe.m meetc(y,e)šboyc((y)], [Ox.Oe.ageentc(x, e)]) 3 [[v]] = [Ox.O Oe.agentc(x, e)] VP v [[VP]] = Resstrict ([Oy.Oe. m meetc(y, e)], boy yc) = 3 [Oy y.Oe. meetc(y, e)) š boyc(y)] V DP 5 meet a boy It follows thhat, in order for the comp position to prooceed, the un nsaturated internal arguument must be b bound by Existential Cllosure (EC) at a the VP level, beforee the VP combbines with the v head: 40 (57)

[[VP ]]] = EC([Oy.Oee. meetc(y, e) 䴑 boyc(y)]) = [Oe. y. meettc(y, e) 䴑 boyc(y))] x,e)]) [[vc]] = EI ([Oe. y. meetc(y, e) 䴑 boyc(y)], [Oxx. Oe. agentc(x = [Ox. Oe. yy. meetc(y, e) 䴑 boyc(y) 䴑 aagentc(x,e)]

This impliess that in a traansitive structure, the defaault rule of Existential E Closure appplies at a comppositional level that does nnot include thee external argument. T This, in turn, gives us a reason why tthe external argument were, it woulld remain cannot be thhe target of Predicate P Resttriction: if it w unsaturated and the comppositional proccess could nott converge. 40

The same holds if we assume a a moree elaborate deccomposition like that in wer event Ramchand (22008). Here, the Init head existentially binds the low contributed bby its complement (Process Phrase) and introduces thee Initiator argument aboove it: once again, a the interrnal argument (Undergoer) introduced i within the Proocess Phrase must m be existentiially closed in oorder for the co omposition to go throughh: (i) [initP D DPEA [ init [ProcPP DPIA [ Proc...]]]]] (ii) [[init]]] = [Ȝf. Ȝx.ȜȜe. ec (f(ec) & causec(e, ec) & agentc(x, e) ] (iii) [[ inittc]] = [[ init]] ([[P ProcP]]) = [Ȝx. Ȝe. ec[[ProcP]]] (ec) & causeec(e, ec) & agenttc(x, e) ]

Definiteness Effect and Unaccusative Subjects

49

4.3. Intermediate Summary To sum up, we have proposed that unaccusative i-subjects licensed as IAs in the thematic position necessarily lack the D-layer, as this would introduce an unvalued Case feature inaccessible to any ȭ-complete probe. As a consequence, i-subjects denote a property and cannot be interpreted as saturating arguments, but they must undergo Predicate Restriction and subsequent Existential Closure at the VP level. This analysis accounts for the cluster of properties that characterize the ‘radical’ DE (21). Note also that the analysis presented in this section 4 embodies a particular view of the syntax-semantics interface, whereby Agree for ȭfeatures, far from being semantically inconsequential, constrains the possible interpretation of a noun phrase that is first merged in a thematic position: it is interpreted as a saturating argument if and only if it is probed by a ȭ-complete functional head; otherwise, it can only be interpreted as a restricting, but not saturating, argument. This view can be traced back to the spirit of Rizzi (1986), who proposed that (valued) ȭ-features are a prerequisite for argumenthood and for referentiality of a noun phrase; for more recent elaboration, see Longobardi (2008). Since in the standard Minimalist framework Agree for ȭ-feature yields valuation of unvalued Case, this theoretical stance subsumes the core of the Case Visibility Condition. Having proposed an account of i-subjects falling under the DE, in the next section we turn to definite post-verbal subjects in Italian. Building on Belletti (2001, 2004), we show that these exploit a completely different licensing route in the periphery of the verb phrase.

5. The Syntax and Interpretation of Post-Verbal Definite Subjects Consider a question answer pair like the following in (58): (58)

a. b.

Chi who Ha has

ha parlato? has spoken parlato Gianni. spoken John

In (58) an intransitive verb is present; the definite subject in the answer is post-verbal. This type of sentences shows very straightforwardly that there is no indefiniteness requirement on the post-verbal subject of an

Chapter One

50

intransitive verb. Indeed, not only can the definite subject be post-verbal, in fact it must be post-verbal in the discourse pragmatics of the exchange in (58). Similarly, in the exchange in (59) with a transitive verb, and in (60) with an unaccusative: (59)

a. b.

(60)

a. b.

(A proposito del libro) (concerning the book) Lo recensirà it.cl review.fut.3sg Chi who È is

Chi lo recensirà? who it.cl review.fut.3sg Gianni/ quel giornalista. John/ that journalist

è arrivato? is arrived arrivata Maria. arrived Mary

All the answers in (58)-(60)b contain a post-verbal subject which is the focus of new information, as it provides the new information requested in the question. Grounding the analysis in cartographic terms, Belletti (2004) has proposed that in cases of this type the subject fills a dedicated position in the low area of the clause, right in the periphery of the verb phrase, which is dedicated to host new information constituents.41 This peripheral position is a property of the clause structure, present in all clauses, independently of the verb class to which the verb of the actual sentence belongs. The fact that the post-verbal position of the subject is tightly linked to the focus interpretation has led to a qualification of the descriptive term “free inversion” often utilized in the literature on subject inversion: inversion is free as far as the verb class to which the verb of the sentence belongs, but it is not at all free as far as its discourse value is concerned.42 Hence, on the basis of exchanges like (58)-(60) the proposal can be made that a vP/VP-peripheral new information Focus position is present in the low part of the clause, along the lines in (61)a,-b; this position is exploited 41

E.g. also, typically a direct object in an object question like (i)a whose full answer diplays the SVO order, with the direct object following the verb: (i) a. Che cosa hai letto? What have.2SG read b. Ho letto un romanzo have.1SG read a novel. 42 The postverbal subject can also be a topic, with a totally different prosody than the one associated with (58)-(60). See Belletti (2004) for detailed discussion in terms of a vP periphery also containing a topic position.

Definiteness Effect and Unaccusative Subjects

51

by a post-verbal new information subject of the type in (58)-(60)b. We will assume with Belletti (2004) that this position is right above the lexical projection of the verb phrase: a vP for transitive and intransitives (61a), a VP in the smaller verbal projection corresponding to the argument structure of unaccusatives (61b).43 The external argument for transitive and intransitive verbs and the internal (definite) argument for unaccusatives moves to the Spec/FocP position, where it is interpreted as the focus of new information; subsequent movement of V to a higher inflectional head results in the VS order: (61)

a.

(transitive, intransitive)

V

3 3 FocP (new info) 3 3 Foc [vP] 3 EA 3 3

43 The VP is in fact contained within the functional projection FP expressing morphosyntactic features of Number and Gender, as assumed in (48); see also (66) below. We omit FP from the tree structures whenever irrelevant. More structure is to be postulated in the unaccusative VP to host a PP argument in case it is also present as the examples in (3), e.g. along the lines in (i): (i) VP 1 1PP V IA

Chapter One

52

b.

(unaccusatives) 3 V 3 FocP (new info) 3 3 Foc VP 3

IA

Notice that a crucial consequence of the proposed analysis so far concerns unaccusatives. With this verb class, the order VS can be derived in two different ways: either with the IA in its first Merge position, as in (48) above (reminded in (62) for its essential property), or with the IA in the Spec of low Focus, as in (61b) above. (62)

VP 3 V IA

It is only in the case of (48)/(62) that the noun phrase has to be a weak indefinite (NumP in our assumptions) and the DE is characteristically manifested. No definiteness requirement concerns the focalized postverbal subject of unaccusatives in the case of (61)b, as it doesn’t with the post-verbal subject of the other verb classes (cf. again (47)/(58b), (59b), (3)/(60b) above). The sentences in (58b)-(60b) are so called narrow focus sentences, in which a single constituent, the subject, is focalized. A post-verbal subject also occurs in so called all-new sentences; a typical instance is the answer to a what happened question with intransitives and unaccusatives: (63)

a. b. c. d.

Che cosa è successo? what is happened Ha parlato Maria. has spoken Mary È arrivata Maria. is arrived.fsg Mary E’ arrivato un ragazzo. is arrived.msg a boy

Definiteness Effect and Unaccusative Subjects

53

With transitives, the order in an all new-sentence is SVO, with O a lexical noun phrase (64b);44 however, if O is a topic, as in the question in (64c), and hence realized as a clitic pronoun in the answer, the subject can be post-verbal again, as illustrated in (64d): (64)

a. b.

c.

d.

Che cosa è successo? what is happened Gianni ha mangiato la torta./ Un camion has Gianni has eaten the cake / A truck has tamponato l’utilitaria. crashed-into the car Che cosa è successo alla torta? / Che cosa what is happened to-the cake? / what successo all’utilitaria? happened to-the car? L’ ha mangiata Gianni /L’ ha it.FSG.ACC has eaten John / it.FSG.ACC has tamponata un camion. crashed-into a truck

è is

We assume that in all new sentences it is the whole vP/VP which moves to Spec/FocP; the post-verbal occurrence of the subject – when it is the EA – is once again due to subsequent movement of V to a higher position. (65) illustrates the derivation (for the vP case):

44 This is probably due to Case requirements: the post-verbal subject would block correct assignment of accusative Case to the object, acting as an intervener with respect to the functional Case assigning head, external (in part) to the vP, as traditional AgrO. See Belletti (2004) for further discussion on the impossibility of VSO in Italian and on the marginal status of VOS (where O is a lexical object), which is only allowed with a topic interpretation of the VO chunk. See also Calabrese (1982) for a first discussion of the discourse value of this order.

Chapter One

54

3  V 3 FocP (new info) 3 3 Foc [VP] 3 EA 3 3

(IA)

(65)

When the subject is preverbal as in (64)b, we assume that it moves into a high subject position in TP (Cardinaletti 2004). The derived SVO order is also possible when the object is expressed through a clitic pronoun (e.g Gianni l’ha mangiata/Gianni it-CL has eaten). Hence, in the answer to a question like (64)c the subject is not necessarily post-verbal. The preverbal position of the subject in this case implies an interpretation whereby it is presupposed that the sentence is going to be about that particular subject (i.e. Gianni in the example above; see Rizzi (2005) on the so called aboutness interpretation of preverbal subjects). No such presupposition is involved in the answer in (64)d in which the subject is post-verbal. When the verb is unaccusative and the subject is a nonpresuppositional indefinite, as in (63d), the indefinite IA/NumP remains in its first merge position, and the whole VP(/FP) moves to the Focus position.45 One residual case must be discussed in detail: in (63c), the IA of the unaccusative verb is a definite noun phrase, hence it is a DP in our 45

It is a prediction of the proposed analysis that in this case the the post-verbal weak indefinite subject cannot be a narrow focus (it is not a saturating IA). Indeed, this is the case, as the following contrasts in French clearly illustrate (we thank Fréderique Berthelot for her feedback) i. Q: Qui est arrivé? who is arrived A: (*) Il est arrivé trois filles. EXPL is arrived three girls ii. Q: A:

Que What Il EXPL

c'est-il passé? there-is-EXPL happened est arrivé is arrived

trois filles. three girls

Definiteness Effect and Unaccusative Subjects

55

analysis. If the IA is a DP, by hypothesis it cannot remain in its first merge position, because it would be inaccessible to a ȭ-complete probe: hence, in (63c) the order VS is not a direct manifestation of the order derived through merge, as in cases corresponding to (62). Since the sentence is an all new sentence, the analysis cannot be the one in (61b), with the IA=DP in the specifier of VP-peripheral Focus, since this corresponds to the assignment of narrow focus to the sole DP. Our proposal for this type of cases capitalizes on the presence of the F projection which contains the unaccusative VP, as illustrated in (48). Given our assumptions, the definite DP argument of unaccusatives has to vacate the IA position, which is incompatible with real arguments; the definite DP thus moves to the Spec/FP position; in this position it becomes accessible to Agree with the T probe which is complete of the person feature, the crucial feature for argumenthood; Nominative is then accessible to this position for the DP. It will then be the whole FP to move to the Spec Focus position; the VS order is obtained through the subsequent movement of V to T in the familiar way. The derivation is illustrated in (66) for the relevant aspects.46 (66) TP 3 pro 3 T[pers, num, Nom] FocP (new info) V 3 FP 3 3 F[num,gen] VP 3

IA/DP

i.A. is clearly a grammatical sentence, but it is infelicitous as an answer to question i.Q. The same sentence is in fact not only grammatical but also felicitous as an all new answer to the what happened question in ii.Q. ii.A. is analyzed with the whole VP/FP in the specifier of the low Focus. 46 The geometry of this derivation is then close to that of verb classes with an external argument, with the argument DP in a position higher than the merge position as an IA, sister of V.

Chapter One

56

In conclusion, a post-verbal unaccusative subject can originate through three different routes: (i)

as the indefinite IA of the unaccusative verb in its first merge position, cf. (48)/(62), in all new sentences (with the whole verb phrase in the specifier of the low new information Focus position); (ii) as a narrow focus definite IA moved into the specifier of the low new information Focus position, with subsequent movement of V to T, as in (61b) (with IA first moved to Spec/F as in (iii) below); (iii) as the definite IA of the unaccusative verb in an all-new sentence, moved into the Spec/F position, with the whole FP subsequently moved into the specifier of the low new information Focus position, and subsequent movement of V to T (66). Only in the first case must the subject be a weak indefinite; according to our analysis, in this case it is actually not a real argument, but it is part of the predicate. Unaccusatives differ from other verb classes in having one peculiar and specific way to yield the order VS, namely (48)/(62); in this condition VS is the reflex of the first merge operation, and the familiar DE manifests itself.

5.1. DE is a Deep Property of Unaccusatives: Evidence from Acquisition One feature of our analysis is worth emphasizing here: the weak indefinite i-subject of unaccusatives is the IA of the verb, which is merged first, and is tightly connected to it. Semantically, it is part of the predicate; syntactically, it meets the strictest local conditions in the relation with V, those holding for noun incorporation. In essence, then, it is fair to say that the order VS – with S a (weak) indefinite – overtly manifests the core of the unaccusative verb class. Interesting converging evidence for this conclusion comes from recent results from acquisition studies, which we briefly review here; for further detail the reader is referred to Belletti & Guasti (2015 chapter 7). Cross-linguistic evidence has shown that children single out the unaccusative class from very early on in their development (Lorusso et al. 2005, Lorusso 2014; Friedmann 2007; Friedmann & Costa 2011). The evidence has characteristically been grounded on the fact that young children have an early mastery of the order VS with unaccusatives. For instance, the corpus study by Lorusso et al. (2005) has shown that Italian

Definiteness Effect and Unaccusative Subjects

57

speaking young children (age range 18 – 36 months, from the CHILDES database) have the subject in the post-verbal position more often with unaccusatives (66%) than with other verb classes (28% with transitives, 21% with intransitives). Similar conclusions are drawn by Friedmann & Costa (2011) on the basis of Hebrew and European Portuguese data. These results are relevant for our discussion since they suggest a privileged use of the post-verbal position of the subject with unaccusatives with respect to other verb classes. However, it is only until the recent work by Vernice & Guasti (2013) that appropriate use of the post-verbal subject with unaccusatives has been tested experimentally (through a repetition task) taking into consideration the definiteness of the post-verbal subject. In a similar vein, Lorusso (2014) has counted the occurrence of the order VS with unaccusatives according to definiteness in children’s spontaneous productions (same corpora as above from the CHILDES database; see Lorusso 2014 for further details). The results from these two studies are very interesting in the perspective of the present work as they both show, with distinct experimental methods (repetition and spontaneous production, respectively), early sensitiveness to the definiteness of the post-verbal subject with unaccusative verbs. In the experimental conditions created in Vernice & Guasti’s study, children were able to provide identical repetitions of sentences with a post-verbal indefinite subject significantly more often when the verb was an unaccusative (52% vs. 25% with intransitives). Lorusso (2014) has shown that indefinite subjects are rarely found in children’s spontaneous productions, and there is no difference according to verb class when the indefinite subject is pre-verbal, hence in the SV order (4% with unaccusative, 4% with intransitives; 1% with transitives); in contrast, when the indefinite subject is post-verbal, the difference is highly significant: 18% post-verbal indefinite subjects with unaccusatives vs 0 occurrences with both transitives and intransitives in the corpora analyzed. These results sharply indicate that not only do children treat the unaccusative verb class differently from early on, but also that they are aware of the DE from the earliest ages. This is quite a remarkable conclusion: pre-theoretically, one could have imagined that something like DE was a complex property, hence a property likely to be mastered late in acquisition. In contrast, given the theoretical analysis we have endorsed here, the DE is a deep property of unaccusatives, in fact one of the defining properties of the verb class itself. Hence, the acquisition of such property should not be demanding in principle, assuming that argument

58

Chapter One

structure and consequently verb-class distinctions are available from early on (Friedmann 2007, Gleitman 1990, a.o.). Indeed, this appears to be precisely the case. This in turn lend original new support in favor of the classical unaccusative hypothesis. As a final remark, we note that converging evidence also comes from a different mode of acquisition: adult L2 acquisition of Italian (see Belletti & Guasti 2015 for detailed presentation and discussion). Experimental studies from Belletti & Leonini (2004) and Belletti, Bennati, Sorace (2007) have shown that adult L2 speakers of Italian (whose L1 is a nonnull subject language like either German, or English or French) experience difficulty in the proper mastery of the VS order, with post-verbal S subject of new information (i.e. narrow focus as in examples (58)-(60), corresponding to structures (61)). In a spontaneous production task of Storytelling, however, the same L2 speakers were able to adequately use this same linear order. Indeed, in telling a similar story,47 the L2 speakers made use of an amount of post-verbal subjects which was remarkably similar to that of the native speakers acting as controls: 16% and 15% respectively. By looking at the actual occurrences of the VS structures the generalization is clear: the VS sentences of the L2 speakers were the same as those of the native controls as they were all describing the same scenes; in both groups the verbs used in telling the story were unaccusative verbs and the subject was indefinite (e.g. manca un cesto lit. ‘is missing a basket’). Once again, whatever the correct account for the difficulty with the order VS experienced in L2 when the subject bears narrow new information focus (see the references cited above), no such difficulty affects sentences displaying the order VS when they contain an unaccusative verb and an indefinite post-verbal subject. To get to this order with a (weak) indefinite subject the route is the first merge one, i.e. structure (62).48 This is a universal property, which need not be specifically learned in a second language.49 47 Description of the same silent movie. See Belletti, Bennati, Sorace (2007) for details. 48 With VP in the Spec of low focus, as the context is all new. 49 The same groups of L2 speakers in both studies by Belletti & Leonini (2004) and Belletti, Bennati, Sorace (2007) correctly answered with the VS order to all the questions containing an existential structure present in the videos (five questions as e.g.: Cosa c’era sul tavolo? /What was there on the table?). This strongly suggests that they answered by resorting to the existential structure, which in both the L1 and the L2 has an indefinite subject in postverbal position. The existential structure may also have an overtly expressed “location”, as in the case of there and y in

Definiteness Effect and Unaccusative Subjects

59

6. Concluding Remarks In this article we have revisited the well known definiteness effect which affects the postverbal noun phrase of existential sentences and more generally the internal argument of unaccusative verbs and transitive verbs in the passive voice when it remains in its First Merge position. We have argued that this position is unaccessible to a ȭ-complete probe, and therefore, the argument cannot be a full-fledged DP; it can only be a defective nominal projection, NumP, and it is syntactically licensed via Agree with a defective ȭ-probe which lacks the person feature, the crucial feature for argumenthood. The defective probe is located immediately above the projection of the lexical verb; the weak indefinite NumP is thus associated with Partitive Case through the Agree relation with the defective probe. We have shown that this licensing route is unavailable for the external argument of intransitive verbs and of active transitive verbs. Furthermore, we have argued that a NumP cannot be interpreted at the interface as a saturating argument; rather, it denotes a property and it is interpreted via Predicate Restriction. Whenever the post-verbal subject is a full-fledged DP, a different licensing route is at play. We have adopted the proposal in Belletti (2004) that the post-verbal subject is licensed either by new information focus, or by being part of a new information vp/VP: in both cases, licensing involves a low Focus position in the periphery of the verb phrase. The upshot of this analysis is that EA subjects can only be licensed via Focus, whereas IA subjects have an extra option, licensing in situ as a defective NumP; this asymmetry is supported in a new original way by evidence from first and second language acquisition, which suggests that the extra option is indeed a deep property of unaccusative verbs across languages.

References Abbott, Barbara. 2001. Definiteness and identification in English. Pragmatics in 2000: Selected Papers from the 7th International Pragmatics Conference, vol. 2, 1-15. Antwerp: International English and French, paralleling ci in Italian; notice that no overt location is present in German es gibt construction. Easiness with the existential sentences, irrespective of the precise realization of the construction in the L1, suggests that the existential structure shares deep properties across languages. We speculate that (a version of) structure (35) may reflect (some of ) these properties.

60

Chapter One

Pragmatics Association. Abusch, Dorit. 1994. The scope of indefinites. In Natural Language Semantics 2: 83-135. Alexiadou, Artemis & Florian Shäfer 2010. There-Insertion: An unaccusativity mismatch at the syntax-semantics interface. In WCCFL 28 Online Proceedings: https://sites.google.com/site/wccfl28pro/alexiadou-schaefer Baker, Mark. 1988. Incorporation: a theory of grammatical function changing. Chicago: University of Chicago Press. Barwise, John & Robin Cooper. 1981. Generalized quantifiers and natural language. In Linguistics and Philosophy 4: 159-219. Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry 19: 1-35. —. 2001. Agreement projections. In Mark Baltin & Chris Collins (eds.), The Syntax Companion. Blackwell. —. 2004. Aspects of the low IP area. In L. Rizzi (ed.). The Structure of CP and IP, 16-51. New York/Oxford: Oxford University Press. —. 2005. Extended doubling and the VP periphery. Probus 17. Reprinted in Structures and Strategies, 201-223. New York: Routledge, 2009. —. 2006. (Past) participle agreement: a case study. In Martin Everaert & Henk van Riemsdijk (eds.). The Blackwell Companion to Syntax, 493521. Blackwell Publishers. Updated version forthcoming. Belletti Adriana & Luigi Rizzi. 1981. The Syntax of ne: Some Theoretical Implications. In The Linguistic Review 1:117-154 Belletti, Adriana & Chiara Leonini. 2004. Subject inversion in L2 Italian. In Foster-Cohen, Susan H., Michael Sharwood Smith, Antonella Sorace & Mitsuhiko Ota (eds.). EUROSLA Yearbook: Volume 4: 95118. Belletti, Adriana, Elisa Bennati & Antonella Sorace. 2007. Theoretical and developmental issues in the syntax of subjects: evidence from nearnative Italian. In Natural Language and Linguistic Theory 25: 657689. Belletti, Adriana & Maria Teresa Guasti. 2015. The Acquisition of Italian. Morphosyntax and its interfaces in different modes of acquisition, John Benjamins. Bobalijk, Jonathan D. & Dianne Jonas. 1996. Subject positions and the role of TP. In Linguistic Inquiry 27: 195-236. Burzio, Luigi. 1986. Italian syntax: A Government-Binding approach. Dordrecht: Reidel.

Definiteness Effect and Unaccusative Subjects

61

Calabrese, Andrea. 1982. Alcune ipotesi sulla struttura informazionale della frase in italiano e sul suo rapporto con la struttura fonologica. In Rivista di Grammatica Generativa 7: 3-78. Cardinaletti, Anna. 2004. Toward a cartography of subject positions. In Luigi. Rizzi (ed.). The structure of CP and IP, 115-165. New York/Oxford: Oxford University Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: The MIT Press. —. 2005. On Phases. In Robert Freidin, Carlos Otero & Maria Luisa Zubizarreta, (eds.). Foundational Issues in Linguistic Theory: Essays in Honor of Jean Roger Vergnaud, 133–66. Cambridge, MA: MIT Press. —. 2014. Problems of Projection: Extensions. Ms.: MIT. Cruschina, Silvio. This volume. Pseudo-existentials and the definitness effect in Italian. Chung, Sandra & William Ladusaw. 2004. Restriction and saturation. Cambridge, Mass.: The MIT Press. Collins, Chris & Höskuldur Thráinsson. 1993. Object Shift in Double Object Constructions and the Theory of Case. MIT Working Papers in Linguistics 19: 131-147. Deal, Amy Rose. 2009. The origin and content of expletives: Evidence from selection. In Syntax, 12.4: 285-323. Diesing, Molly. 1992. Indefinites. Cambridge, Mass.: MIT Press. Diesing, Molly & Eloise Jelinek. 1995. Distributing arguments. In Natural Language Semantics 3: 123-176. Enç, Mürvet. 1991. The semantics of specificity. In Linguistic Inquiry 22: 125. Von Fintel, Kai. 1994. Restrictions on quantifier domains. PhD dissertation, University of Massachusetts at Amherts. Fischer, Susann. This volume. Existentials vs. unaccusatives: The definiteness restriction in Romance. Francez, Itamar. 2010. Context dependence and implicit arguments in existentials. In Linguistics and Philosophy 33: 11-30. Friedemann, Marc Ariel & Tal Siloni. 1997. AGRobject is not AGRparticiple. In The Linguistic Review 14: 69-96. Friedmann, Naama. 2007. Young children and A-chains: The acquisition of Hebrew unaccusatives. In Language Acquisition 14: 377-422. Friedmann, Naama & Joao Costa. 2011. Acquisition of SV and VS order in Hebrew, European Portuguese, Palestinian Arabic and Spanish. In Language Acquisition 18: 1-38. Van Geenhoven, Veerle. 1998. Semantic incorporation and indefinite descriptions. Stanford CA: CSLI

62

Chapter One

Gleitman, Lila. 1990. The structural sources of verb meanings. In Language Acquisition 1: 3-55 Haegeman, Liliane. 1986. Er sentences in West Flemish. Université de Genève de Hoop, Helen, 1992. Case configuration and Noun Phrase interpretation. PhD thesis, Rijksuniversiteit Groningen. (Published by Garland, 1997). Harley, Heidi. 2012. External arguments and the Mirror Principle: On the distinctness of Voice and v. ling.auf.net/lingbuzz/001526. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. PhD dissertation, University of Massachusetts at Amherts. Heim, Irene & Angelika Kratzer. 1998. Semantics in Generative Grammar. Oxford: Blackwell Publishers. Higginbotham, James. 1987. Indefiniteness and predication. In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness, 43-70. Cambridge, Mass.: The MIT Press. Kalulli, Dalina 2008. There is secondary predication in There-Existentials. In Charles B. Chang & Hannah J. Haynie (eds.). Proceedings of the 26th West Coast Conference on Formal Linguistics, 279-287. Sommerville, MA: Cascadilla Press Kayne, Richard S. 1981. ECP Extensions. In Linguistic Inquiry 12.1: 93133. —. 1993. Toward a modular theory of auxiliary selection. In Studia Linguistica 47: 3-31. Kamp, Hans. 1981. A theory of truth and semantic representation. In Jeroen Groenendijk, Theo Janssen & Martin Stokhof (eds.). Formal Methods in the Study of Language, 277-322. Amsterdam: Mathematical Centre Tracts 135. Keenan, Edward. 1987. A semantic definition of "Indefinite NP". In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness, 286-317. Cambridge, MA: MIT Press. Kratzer, Angelika. 1995. Stage-level and individual-level predicates. In Gregory N. Carlson & Fracis J. Pelletier (eds.). The Generic Book, 125-175. Chicago: The University of Chicago Press. —. 1996. Severing the external argument from its verb. In Johan Rooryck & Laurie Zaring (eds.). Phrase Structure and the Lexicon, 109-137. Dordrecht: Kluwer. —. 1998. Scope or pseudoscope? Are there wide-scope indefinites? In Susan Rothstein (ed.). Events and Grammar, 163-196. Dordrecht: Kluwer. Ladusaw, William. 1994. Thetic and categorical, stage and individual, weak and strong. Proceedings of SALT 4.

Definiteness Effect and Unaccusative Subjects

63

Larson, Richard. 1988. On the double object construction. In Linguistic Inquiry 19: 335-391. Lasnik, Howard. 1992. Case and expletive: Notes toward a parametric account. In Linguistic Inquiry 3.3: 381-405. —. 1999. Minimalist Analysis (chapter 4: Case and expletives revisited). Blackwell. Legendre, Géraldine & Antonella Sorace. 2003. Auxiliaries et intransitivité en français et dans les langues romanes. In Danièle Godard (ed.). Les langues romanes. Problèmes de la phrase simple. 185-233. Paris: CNRS Éditions Longobardi, Giuseppe. 1994. Reference and proper names: a theory of Nmovement in syntax and logical form. In Linguistic Inquiry 25: 609665. —. 2008. Reference to individuals, person, and the variety of mapping parameters. In A. Klinge & H. Müller (eds.). Essays on Nominal Determination, 189-211. Philadelphia/Amsterdam: John Benjamins. Lorusso, Paolo. 2014. Verbs in Child Grammar. The Acquisition of the Primitive Elements of the VP at the Syntax-Semantics Interface. PhD dissertation Universitat Antònoma de Barcelona Lorusso, Paolo, Claudia Caprin & Maria Theresa Guasti. 2005. Overt subject distribution in Early Italian children. In BUCLD 29 Online Proceedings, available at: www.bu.edu/bucld/proceedings/supplement/vol29/ McCloskey, James. Forthcoming. Irish existentials in context. To appear in Syntax. McNally, Louise. 2009. Properties, entity correlates of properties, and existentials. In Anastasia Giannakidou & Monika Rathert (eds.). Quantification, definiteness and nominalization, 163-187. Oxford/ New York: Oxford University Press —. 2011. Existential sentences cross-linguistically: Variations in form and meaning. Available at: http://www.upf.edu/pdi/louise-mcnally/_pdf/ publications/McNally_existentials.pdf McNally, Louise & Veerle van Geenhoven. 1998. Redefining the weak/strong distinction. Paper presented at the second Colloque de Syntaxe et Sémantique à Paris (CSSP 1997) October 16-18 1997, Université Paris 7. Available from: https://www.researchgate.net/publication/277297323_Redefining_the_ weakstrong_distinction Milsark, Gary. 1974. Existential Sentences in English. PhD Dissertation, MIT. —. 1977. Toward an explanation of certain peculiarities of the existential

64

Chapter One

construction in English. Linguistic Analysis 3: 1-29. Moro, Andrea. 1997. The Raising of Predicates. Predicative Noun Phrases and the Theory of Clause Structure. Cambridge University Press. Partee, Barbara. 1984. Noun phrase interpretation and type shifting principles. In Jeroen Groenendijk, Dick de Jongh & Martin Stokhof (eds.). Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, 115-143. Dordrecht: Foris. Perlmutter, David. 1978. Impersonal passives and the Unaccusative Hypothesis. In Proceedings of the 4th Annual Meeting of the Berkeley Linguistic Society, 157-190. Ramchand, Gillian. 2008. Verb Meaning and the Lexicon: A first Phase Syntax. Cambridge, Mass.: The MIT Press. Reinhart, Tanya. 1987. Specifier and operator binding. In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness. Dordrecht: Foris. —. 1997. Quantifier Scope: How labor is divided between QR and Choice Functions. In Linguistics and Philosophy 20: 335-397. Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris. —. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17: 501-557. —. 1990. Relativized Minimality. Cambridge, Mass.: The MIT Press. —. 2004. Locality and Left Periphery. In Adriana Belletti (ed.). Structures and Beyond. The Cartography of Syntactic Structures, vol.3, New York: Oxford University Press. —. 2005. On some properties of subjects and topics. In Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert & Giuseppina Turano (eds.). Contributions to the XXX Incontro di Grammatica Generativa, 203-224. Venezia, Cafoscarina. Safir, Ken. 1982. Syntactic Chains and the Definiteness Effect. PhD dissertation, MIT. —. 1987. What explain the definiteness effect? In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness, 71-97. Dordrecht: Foris. Schwarzschild, Roger. 2002. Singleton indefinites. Journal of Semantics 19: 289-314. Sheehan, Michelle & Wolfram Hinzen. 2011. Moving towards the edge. Linguistic Analysis 37: 405-458. Stowell, Timothy A. 1981. Origin of Phrase Structure. PhD dissertation, MIT Cambridge, MA Vainikka, Anne & Joan Maling. 1996. Is partitive Case inherent or structural? In Jack Hoeksema (ed.). Partitives, 179-208. Berlin:

Definiteness Effect and Unaccusative Subjects

65

Mouton de Gruyter. Vangsnes, Oystein. 2002. Icelandic Expletive Constructions and the Distribution of Subject Types. In Peter Svenonius (ed.). Subjects, Expletives, and the EPP. New York/Oxford: Oxford University Press. www.hum.uit.no/a/vangsnes/vangsnesepp.pdf Vernice, Mirta & Maria Teresa Guasti. 2013. The acquisition of SV order in unaccusatives: manipulating the definiteness of the NP argument. To appear in Journal of Child Language. Williams, Edwin. 1981. Argument structure and morphology. The Linguistic Review 1: 81-114. —. 1984 There-Insertion. Linguistic Inquiry 1.1: 131-153. Winter, Yoad. 1997. Choice functions and the scopal semantics of indefinites. Linguistics and Philosophy 20, 399-467. Zamparelli, Roberto. 1999. Layers in the Determiner Phrase. New York: Garland. —. 2002. Definite and bare kind-denoting Noun Phrases. In Frank Drijkoningen, Claire Beyssade, Paola Monachesi & Reineke BokBennema (eds.). Romance languages and linguistic theory 2000: Selected papers from Going Romance 2000, 305-342. Amsterdam: John Benjamins.

CHAPTER TWO DEFINITENESS EFFECTS: THE INTERPLAY OF INFORMATION STRUCTURE AND PRAGMATICS1 MANUEL LEONETTI

1. Introduction The intensive research that has been carried out over the past two decades on all aspects of Information Structure has had lots of positive results. The most salient one, in my opinion, has been providing us with new insights about how Information Structure (IS) shapes grammatical phenomena that were previously studied as completely unrelated to notions like Topic, Focus and Contrast (Erteschik-Shir 1997 is a major contribution to this perspective). With this situation in mind, the present paper aims to clarify the role of IS in the occurrence of definiteness effects (from now on, DEs) and the way it interacts with pragmatic inference. I will not attempt to present a full account of DEs, but rather to establish what the place of IS should be in a global approach to constraints on definiteness. It would be obviously wrong to claim that the role of IS has not been considered before in the literature on DEs: in fact, it is customary to mention the focal status of the DP in existential sentences, as well as the thetic (‘all-focus’, ‘all-new’) nature of these constructions, and there is 1

The investigation presented in this paper is included in the research project “Semántica procedimental y contenido explícito III” (SPYCE III), funded by the Spanish Ministerio de Economía y Competitividad (FFI2012-31785). It contains some developments of ideas originally presented in Leonetti (2008). I am grateful to Victoria Escandell-Vidal, Urtzi Etxeberria, Fátima Oliveira, Fabiola Stoian, Xavier Villalba and Michael Zimmermann for their help with data, to the editors of the volume for their infinite patience, and to two anonymous reviewers for useful comments and suggestions. Special thanks are due to Aoife Ahern for checking the English text.

Definiteness Effects

67

ample consensus on the importance of Topic / Focus articulation, not only for an account of DEs but, in a more general way, for the interpretation of indefinite DPs. However, there are still points in need of clarification, and I hope to show that a revision of data and proposals from an IS viewpoint may throw new light on old issues. My claim is that there are two ways in which IS is essential for understanding DEs, in particular when they affect anaphoric definite DPs. On the one hand, IS defines the basic grammatical environments where the DE can appear (existential contexts and unaccusative inversion): such environments are characterized by the presence of a stage topic and the interpretation of the relevant DP as a part of wide focus. On the other hand, IS defines the conditions for violations of the DE to occur: for the relevant DP to violate the constraint and still be acceptable, if it is an anaphoric definite DP, the crucial condition is that it receives a narrow focus reading, since narrow focus implies evoking a set of contextual alternatives, with the corresponding “availability readings”; this works as a sort of “survival strategy” for definiteness, i.e. a way to present given information as new. Here pragmatic inference plays the role of a last-resort mechanism for the resolution of interpretive mismatches. As in most studies of the Grammar / Pragmatics interaction, the ultimate goal of this paper is to specify what aspects of DEs are a part of the grammar of natural languages, and what aspects are better analyzed as derived from the interplay of grammatical and non-grammatical factors. The paper is organized as follows. Section 2 (‘Preliminary assumptions’) is an overview of basic assumptions and ideas that constitute the background of the ensuing discussion. Section 3 (‘The role of Information Structure’) is the central part of the paper, and is further divided into three subsections: the first one deals with the interaction between IS and (in)definiteness, the second one with the role of wide focus in specifying the conditions for the DE, and the third one with narrow focus in the counterexamples to the DE. Section 4 (‘Pragmatics: inferring relevant interpretations’) is devoted to a discussion of the two basic tasks of pragmatic inference in accounts of DEs: one is evaluating the options the grammatical system displays for the expression of one single content, and the other is elaborating the interpretive solutions for semantic mismatches. Finally, section 5 includes both conclusions and some speculations on the family resemblance among DEs and specificity effects.

Chapter Two

68

2. Preliminary Assumptions 2.1 DP-internal vs DP-external Factors for DEs If DEs are defined as constraints on the occurrence of definite DPs in certain contexts, the first point to be clarified is that there are two kinds of DEs, one due to DP-internal factors, and one due to DP-external factors. Only one type of DE is going to be discussed here, namely the type of DE where IS is involved and has a role to play. IS is obviously irrelevant for those DEs that depend only on factors internal to DP structure. A representative sample of such factors is shown in the examples in (1) and (2): in (1) nominal modifiers block the occurrence of indefinite determiners and force the insertion of the definite article, since the modifiers require uniqueness (cf. Lyons 1999: 246) –thus, these are actually “INdefiniteness effects”, instead of proper definiteness effects; in (2) the presence of modifiers –an infinitival relative clause in Spanish for (2a), an evaluative adjective in postnominal position in Spanish for (2b)excludes the definite article, in favor of indefinite determiners. (1)

a. b.

{*a / the} city of Madrid {*a / the} tallest girl in the class

(2)

a.

Tengo Have.PRS.1SG

b.

{una / *la} cosa importante que hacer a / the thing important that do.INF (Spanish) ‘I have an important thing to do.’ {un / #el} jugador asombroso a / the player amazing ( the amazing player) (cf. el asombroso jugador) (Spanish) ‘an amazing player’

The constraints illustrated in (1) and (2) result from the incompatibility of definiteness / indefiniteness with the semantic requirements of nominal modifiers. They are independent from factors external to DP structure and, therefore, will not be discussed here. By contrast, the best known cases of DE do in fact exhibit a close connection to IS: they involve existential sentences, as in (3), and certain kinds of unaccusative inversion patterns, as in (4), together with other less studied constructions. This is the kind of DE that this paper deals with. (3)

*There was the dog in the yard. (cf. There was a dog in the yard)

Definiteness Effects

(4)

*Il est It be.PRS.3SG

69

arrivé la fille. arrived the girl (cf. Il est arrivé une fille) (French)

‘The girl arrived.’

2.2 The DE as a Constraint on Interpretations All kinds of DE share a basic feature: the incompatibility of definiteness with a grammatical construction. Ideally, a global account of the DE should rely on a unified notion of definiteness. At the moment, there is no agreement about how this semantic notion has to be defined, and not even about whether one single notion should be established. I will assume that in fact a single notion of definiteness can be used, and I will rely on a characterization of definiteness based on the idea of uniqueness (or inclusiveness, as in Hawkins 1978): a definite determiner encodes the instruction to uniquely identify a referent by means of the descriptive content of the DP plus contextual or background information. Such an instruction is the basic linguistic meaning of definite determiners and pronouns; familiarity and giveness features are also frequently associated with the use of definite elements, but they are just the result of pragmatic inferences triggered by the encoded instruction. Now, whatever our view of definiteness may be, the point I wish to stress is that an account of the DE must be based on the meaning of definiteness. Obvious and self-evident as this assumption may seem, it is still worth mentioning, since in certain syntactic approaches to the DE it has been almost ignored. If the role of definiteness is taken as a starting point, it becomes natural to think of DEs as the manifestations of a semantic constraint. I will adopt this perspective, following Lyons’ (1999: 46) claim that the DE “is more likely to be a semantic or pragmatic constraint than a syntactic one”. There are good reasons to consider the DE as a semantic constraint, in two senses, i.e. as a constraint that has to be explained in semantic / pragmatic terms, and as a constraint against definite interpretations of DPs –or maybe even against a subset of definite interpretations-, more than as a formal constraint against the occurrence of definite determiners2. The following are, in my view, compelling arguments in favor of a semantic view of the DE: 2

Some authors (for instance, McNally 1992/1997, 2011) have pointed out that there is an asymmetry between definite determiners and universal quantifiers in existential sentences that any analysis of the DE should take into account. Despite its relevance, this is an issue I cannot deal with here. Therefore, I will consider only definite DPs headed by the definite article or by demonstratives or possessives, leaving universal quantifiers out of the scope of this paper.

Chapter Two

70

1. The literature on the DE offers extensive discussion of the wellknown counterexamples to the DE in different languages (cf. Suñer 1982, Lumsden 1988, Abbott 1993, 1997, Birner and Ward 1995, Francez 2007, Bentley 2013). It is striking that there is little cross-linguistic variation at this point: the counterexamples correspond to a limited set of cases that are relatively stable (see section 3.3 for discussion). This suggests that the explanation is more likely to be semantic than purely syntactic, both because of the fact that counterexamples are widespread and because they are quite systematic. A syntactic constraint would not admit such frequent and systematic violations. Thus, the facts are more easily accommodated in a semantic approach. 2. The DE is in force even in cases where there are no definite determiners involved. Consider the cases of Spanish free relatives in (5) and English bare plurals in (6). (5)

a.

b.

c.

(6)

Hay quien no lo cree. Have.PRS.3SG+LOC who not it believe.PRS.3SG ‘There are people who don’t believe it.’ Quien hizo esto sabía lo que hacía. Who do.PST.3SG this know.PST.3SG it that do.PSTIMP.3SG ‘The person who did this knew what (s)he was doing.’ No tengo a quien invitar. Not have.PRS.1SG to who invite.INF ‘I don’t have anyone to invite.’

There are beautiful places to visit.

Free relatives in Spanish usually display a definite reading, as can be observed in (5b), where the free relative is the sentential subject and is paraphrased as ‘The person who did this’; however, in contexts like the one in (5c), they receive an indefinite reading, with narrow scope with respect to negation. In (5a) a free relative appears inside an existential sentence, and its interpretation is obligatorily indefinite: it is not possible to assign it a definite reading. This is simply an instance of the DE in existential sentences with haber ‘have’ in Spanish, and shows that the constraint holds in the absence of explicit definite determiners. As for bare plurals in English, it is well known that they can receive both generic and existential readings, and this depends on the kind of predicate they are combined with (Individual-level or Stage-level). When they occur as the internal DP (the ‘pivot’, from now on) in a there-

Definiteness Effects

71

sentence, as in (6), only an existential reading is possible, i.e. the kind of reading that an indefinite DP gets. There are no determiners in bare plurals: it is the existential construction that blocks the possibility of a generic reading, and this is again a manifestation of the DE. 3. Finally, it is interesting that existential sentences, the most prominent context for DEs, constrain not only the occurrence of definite DPs, but the plausibility of specific interpretations of indefinite DPs as well. This cannot be unrelated to the facts concerning definiteness, and again suggests that it is some constraint on interpretations that is at stake. Evidence for this ‘anti-specificity’ effect comes from different sources, such as the anomaly of partitive DPs in there-sentences in English and Spanish, in (7) -I am aware of the fact that partitive indefinite DPs are not necessarily specific, but it is true that partitivity tends to favor specific readings; the impossibility of Differential Object Marking in existential pivots in Spanish and Hindi, in (8); the obligatory narrow scope reading of indefinite pivots, illustrated in (9); and the ungrammaticality of nonrestrictive relatives -obtained by relativizing the pivot DP- in Spanish and Italian, as in (10). (7)

(8)

a. b.

#There were two of them in the office. *¿Cuáles de tus amigos hay Which of your friends have.PRS.3SG

a.

Había (*a) una enfermera. (Spanish) Have.PST.3SG to a nurse ‘There was a nurse.’ *Jangal meN sher-ko hE. jungle in lion-ACC is ‘There is a lion in the jungle.’ (Hindi; from López 2012: 127)

b.

en el bar? in the bar (Spanish) ‘Which of your friends are [there] at the bar?’

(9)

There weren’t many girls. ( ‘Many girls were not there’)

(10)

a. b.

*tres latas, que había en el frigorífico, (Spanish) three cans that have.PST.3SG in the fridge *tre lattine, che c’erano in frigo, (Italian) three cans that LOC be.PST.3PL in fridge ‘Three cans, which there were in the fridge’

72

Chapter Two

At first sight, this cluster of facts may seem completely unrelated to the DE. However, it is reasonable to assume that it originates in the same basic condition that gives rise to the prototypical DE, because dissociating the data in (7)-(10) from the DE would lead us to miss important generalizations. All the data should be kept together if one aims to understand what kind of condition existential contexts impose on their internal DPs: the null hypothesis is that the whole set of phenomena brought together in (7)-(10), plus the classical DE, can be explained as effects of one single property of the construction; any alternative account would be, in principle, costlier and less natural. With this in mind, a brief comment on the facts is in order here. In (7) it is shown that certain indefinite DPs, namely those that have a partitive specific reading, tend to be excluded from the pivot position of existential sentences. This implies that not only definites are excluded, but also indefinites whose interpretation is based on some kind of definiteness component –i.e. the contextually given status of the set on which the indefinite operates in partitive structures (cf. Enç 1991 for a view of the DE along these lines). The DE, thus, is not just a ban against DPs headed by definite determiners, but rather a ban against a family of interpretations that includes D-linked or partitive readings of indefinites. The examples in (8) involve Differential Object Marking (DOM): as extensively discussed in López (2012), DOM is often incompatible with the pivot position of existential sentences –though there are counterexamples, at least in some Southern Italian dialects (cf. Bentley, Ciconte and Cruschina 2015). This is not necessarily to be interpreted as an ‘anti-specificity effect’, since DOM is not systematically related with specific readings of indefinites in all languages –Spanish is a good example-, but provides valuable evidence in favor of the existence of some abstract property of pivot positions in existentials that is responsible for the ban against definites, on the one hand, and (certain kinds of) specific indefinites, on the other, or even against any kind of discourse prominence that could be associated with the DP. Recall that DOM is in some sense the mirror image of the DE: DOM positions are usually characterized by definiteness and specificity, whereas DE positions exhibit exactly the opposite condition –they exclude definiteness and specificity. It is just logical to wonder whether there is some kind of connection between both phenomena, and this is why I include DOM in the discussion. What (9) indicates is a well-known property of existential pivots: they show narrow scope with respect to operators like negation. Again, scope is not to be directly identified with specificity, as specific readings of indefinite DPs are only in some cases a matter of scope. However, as with

Definiteness Effects

73

previous facts, the problem cannot be completely disjoint from the DE: it seems quite natural that one and the same feature of existential contexts may underlie both scope restrictions and DEs. In any case, if there is a link, which I do assume, it must be found in the semantics of the construction. Finally, the data in (10) have not received much attention in the literature, but I think they fit nicely in the argumentation. The problem concerns non-restrictive relatives, as already mentioned. This kind of relative clauses is incompatible with non-specific antecedents, as shown by the ungrammaticality of (11a), where the antecedent is a negative DP, and (11b), where the antecedent is a weak indefinite DP. (11)

a. b.

*no book, which I looked for, (cf. ‘no book that I looked for’) *few books, which I looked for, (cf. ‘few books that I looked for’)

Once this is assumed, the ungrammatical status of the examples in (10) must be explained as the result of the mismatch between the requirement of non-restrictive relatives and the weak, non-specific reading of the indefinite pivot –a hallmark of existential sentences-. Notice that the same examples are well formed when the relative clause is restrictive (cf. Sp. tres latas que había en el frigorífico ‘three cans that were in the fridge’): in such case there is no special requirement on the antecedent. If specific readings of indefinite pivots were available, the strings in (10) would be grammatical. Therefore, we have reasons to believe that existential contexts not only exclude definite DPs, but also specific readings of indefinite DPs. This confirms that the restriction imposed affects types of interpretations, rather than types of determiners. It seems difficult to capture the obvious connection between the two issues in a purely syntactic fashion. The most interesting consequence of this series of facts is that an adequate account of the DE should be able to capture the analogies with the restrictions on specificity (more precisely, on specific interpretations of indefinite DPs). I will try to bring together some ideas on this point in section 5. To sum up, I believe there is enough supporting evidence for a semantic approach to the DE. The DE, then, must be a constraint on possible interpretations, as rightly assumed in a large part of the literature. Syntax must play a major role in establishing the conditions for the DE, but the explanation of the restriction cannot be purely syntactic. What syntax does is set the conditions for the constraint to be in force, and a

74

Chapter Two

certain amount of cross-linguistic variation is to be found at that level (cf. section 3.2 for variation in information structure).

2.3 Syntactic Conditions for DEs From a syntactic point of view, the crucial generalization with respect to DEs is their association with internal arguments of predicates, i.e. with ‘object’ positions. No DE have been reported affecting external arguments or internal prepositional arguments. The DPs that obey the DE have to occur inside the VP, predominantly in an adjacent position to the verb, and inside the domain of focus. This syntactic condition is not new in the literature: it had already been noticed –and thoroughly investigated- in Belletti (1988), on the basis of data from Italian, and is to a certain extent covered by Diesing’s (1992) Mapping Hypothesis, with the assumption that material from VP is mapped into nuclear scope at Logical Form and makes a natural domain for existential closure. The condition holds not only in existential sentences, but in virtually all the syntactic environments associated with the DE, i.e., in unaccusative inversion, in extraposition of modifiers from DPs, and in have constructions. There must be some reason why constraints against definiteness and specificity affect only internal arguments3. This is one of the key questions we must address; and possible answers, in line with what has previously been said on the semantic nature of the DE, should ideally establish a link between certain positions and the semantics of definiteness. Needless to say, the DE is not found in all ‘object’ positions, but only in a small set of them (for instance, there are no restrictions in VP-internal positions of transitive clauses, as well as in locative inversion with fronted PPs). Thus, we need to take into account some additional syntactic 3 I assume that pivot DPs in existential contexts may be predicates in some languages and arguments of the existential verb in others. When they are arguments, they behave as semantically incorporated expressions, possibly composed by means of the ‘Restrict’ operation, in Chung and Ladusaw’s (2004) terms: thus, they are taken as predicate modifiers, which means that their status is quite close to that of truly predicative pivots (cf. McNally 1992/1997, Zamparelli 2000, Francez 2007, a. o., for the view that existential pivots are predicates). Independently of their predicative or argumental status, I maintain the basic, informal idea that they occupy an ‘internal argument’ slot, for two reasons: one is to give salience to the essential condition of being VP-internal, and the other one is to keep a parallelism with other DE contexts where the relevant DPs are in fact internal arguments (cf. unaccusative inversion and complements of havepredicates). Notice that this implies that being predicates is not the crucial condition for pivots to be affected by the DE, as proposed in Francez (2007).

Definiteness Effects

75

condition. A factor that is systematically invoked in the syntactic literature on the DE is the relationship with the presence of expletives in preverbal subject position. In languages like English and French the DE is in fact associated with the presence of the expletives there and il (see Sheehan 2006: ch. 3, Hartmann 2008, Fischer 2013, and Zimmermann this volume for discussion), as shown in (12) and (13). (12)

a. b.

There arrived {a man / *the man}. There is {a dog / *the dog} in the garden.

(English)

(13)

a. b.

Il est arrivé {un homme / *l’homme}. Il y a {un chien / *le chien} dans le jardin.

(French)

The presence of an expletive, be it overt or covert, is considered as an essential condition for DEs in Sheehan (2006): the idea is that, if no expletive is introduced, there will be no DE. Now, despite the clear correlation between expletives and DE that English and French –i.e. nonnull-subject languages- display, the idea that an expletive is required for the DE to occur raises some questions and difficulties. I would like to highlight the following points: a. Even assuming the correlation to be descriptively adequate, there seems to be no reason why the presence of an expletive in itself should constrain the appearance of definite DPs or definite / specific interpretations. Just stipulating that expletives are responsible for DEs does not help to explain why the two notions are connected. If expletives are analyzed as a sort of existential quantifiers, as in some classical proposals (Milsark 1977), then the incompatibility with definiteness is motivated, but I see no justification for adopting an analysis of expletives along these lines. The only way out, in my view, is assuming that the crucial factor for the DE is not the mere insertion of expletives, but some other significant property of the constructions that in some languages, under the appropriate conditions, manifests itself by means of expletives. b. The presence of expletives suggests that existential sentences (cf. 12b, 13b) and unaccusative inversions (cf. 12a, 13a) belong to the same kind of constructions, and in fact they share important features. However, existential and unaccusative inversions do not always pattern together with respect to the DE, as rightly pointed out in Fischer (2013). A case in point is Spanish: whereas it shows a robust DE in existential sentences, it allows for all kinds of definite postverbal subjects in unaccusative inversion, with no overt expletives in any case. This is an indication that there must be

76

Chapter Two

other relevant factors that possibly interact with the insertion of expletives and may be more decisive. c. At a purely descriptive level, the correlation does not seem to hold, if certain facts are considered: Italian, being a null subject language, has no expletives, but displays a DE in the two constructions under discussion, though it obeys different conditions than the ones relevant for French; according to Fischer (2013: 42), in Old French and Old English the DE applies independently of whether an expletive is introduced or not; finally, as Vangsnes (2002) has shown, in Icelandic the DE can be observed both with an expletive in initial position and without it, and moreover, its range varies with the precise position occupied by the subject. The facts indicate that the DE cannot be entirely dependent on the presence of expletives. d. The main role assigned to expletives obscures the existence of similarities between existential and unaccusative inversion, on the one hand, and other contexts where the DE shows up without expletives being involved, like have sentences (cf. I have two sisters vs *I have the sisters), on the other. This is not necessarily a shortcoming, as the two sets of contexts could well require different accounts. However, if possible, a unified account is desirable, and the role of expletives in it is by no means clear. To sum up, there are basically two problems with expletives as a syntactic factor for the DE: one is theoretical and has to do with the reason why expletives should be relevant, and the other one is empirical and relates to the cases in which the DE is not triggered by expletives. In my opinion, the most reasonable position with respect to this issue is trying to retain all the insights of analyses like Sheehan’s (2006) while at the same time giving a response to these problems. In doing this, IS enters the picture. As for the first problem, my proposal is that expletives, in non-nullsubject languages, are relevant because they either represent ‘stage topics’ or indicate the presence of a stage topic –and the corresponding absence of an aboutness topic. In ‘thetic sentences’ –also called ‘all-new’ and ‘allfocus’ sentences- the topic that any sentence is supposed to include as a basis for assessing predication is, according to Gundel (1974) and Erteschik-Shir (1997) among others, a ‘stage topic’, i.e., the situation given in the context. A stage topic indicates the spatio-temporal coordinates of the predication, and it is usually context-given. The central cases of DE contexts –existential sentences and unaccusative inversionsare typically thetic. In fact, recovering some explicit or implicit location is essential for an adequate contextualization of such sentence types, even in out-of-the-blue uses (cf. Francez 2007 and Bentley, Ciconte and Cruschina

Definiteness Effects

77

2015 for a similar view of existential sentences). Briefly, in languages that resort to expletives, it is the presence of a stage topic what underlies the insertion of such elements, according to rules that are subject to crosslinguistic variation. Thus, ‘stage topic’, instead of ‘expletive’, is the relevant notion for the DE. As stage topics are typically associated with thetic sentences, and both existential contexts and unaccusative inversions are clear examples of thetic sentences, the syntactic condition that seems to emerge as one of the key ingredients of DEs is that syntax must impose a wide focus interpretation to the sequence. Crucially, the condition is related to IS; it will be dealt with in section 3.2. As for the issue of the lack of a systematic correlation between expletives and DEs, it is no longer a problem once the alleged correlation is seen as an indirect link ultimately dependent on IS notions. The presence of expletives actually is due to factors that play no role in an account of DEs, such as the distinction between null subject languages and non-null-subject languages. There is a considerable amount of research on how the presence of initial locatives as stage topics, explicit or implicit, licenses wide focus inversion in many languages, often with consequences for the (in)definiteness of the postverbal subject, but this is a point that I cannot discuss here (see Rigau 1997, Borer 2005, Sheehan 2006, Lahousse 2011). The assumption I wish to stress here is simply that the DE seems to be associated with one particular kind of subject inversion, and this happens to be wide focus inversion with a stage topic (most commonly found with unaccusative predicates). Expletives are nothing more than grammatical markers associated with this construction in certain languages. To sum up, the role played by syntax in DEs amounts to the combination of the two conditions discussed so far: the DP must occur in some internal argument position, and the sentence structure must impose a thetic interpretation, with a stage topic. This ensures that the DE will affect only DPs in non-topical positions. A consequence of this view is that syntactic conditions for the DE are in fact motivated by IS (cf. Bentley 2004, 2013 for a similar view). The mediation of IS is required to understand why certain syntactic properties are related to a constraint on definiteness. So here we have a first reason for invoking IS in an account of the DE: IS is needed to define the type of grammatical environment where the DE can appear. A second reason for resorting to IS will be made clear in sections 3.2 and 3.3: I intend to show that IS is also needed to explain the conditions under which the DE is violated. This is enough to cover, in a broad sense, the classical cases of DEs in many different languages. However, it cannot include the conditions for

78

Chapter Two

the DE in those languages in which the constraint is mostly induced by lexical / aspectual factors, such as Hungarian (see Szabolcsi 1984 for a seminal study, and Kiss 1995 and Maleczki 2010 for further discussion) – the same applies to constructions that involve a lexical trigger for the DE, such as have predications4.

2.4 The Nature of the DE The next question is why definiteness is sensitive to the mentioned conditions. Looking for a satisfactory answer means investigating the real nature of the DE. Although it is not my aim to explore this issue in depth, some remarks are due on how the constraint is originated. I assume without further discussion that the DE results from a mismatch between the definiteness of the pivot and some semantic condition that DE contexts impose. Whatever such a condition may be, the DE results from a clash in semantic properties. There are basically two lines of thought in current approaches to the DE, and they are not necessarily incompatible: either there is a clash between the presuppositional nature of definite determiners and the felicity conditions of the construction, as the construction is intended to introduce new discourse referents that are not previously presupposed (Lumsden 1988, Zucchi 1995), or there is a clash between the referential properties of definite DPs and the requirement of the construction for the internal DP to be semantically incorporated, in some sense, or to be a property-denoting expression (McNally 1992/1997, 2011, Zamparelli 2000, Bende-Farkas and Kamp 2001). Both lines are well justified. An account inspired by presuppositionality captures in a natural way the fact that (most) definite DPs cannot comply with the requirement to introduce new discourse referents (the Novelty Condition; cf. 3.2). In such an account, the connection with IS should be established on the basis of this requirement. As for the second theoretical option, the main argument in favor of some version of semantic incorporation (for instance, the one defended in Chung and Ladusaw 2004) is that the properties which characterize DE contexts -a) narrow scope of the postverbal DP, b) specific interpretations excluded, c) only internal arguments affected, and d) only focal positions affected- are all typical properties of incorporated arguments. Very briefly, 4

I will not discuss this side of the problem here, as in this case IS does not display the major role it has in the central contexts. I assume that an account can be found for non-prototypical contexts of DEs that is compatible with the proposal in this paper.

Definiteness Effects

79

incorporated arguments are property-denoting expressions that are interpreted as predicate modifiers or qualifiers, thus forming a unit with the predicate they are arguments of. If existential sentences require some kind of semantic incorporation of the pivot, then its denotational properties should be provided by the predicate, via binding of the discourse referent or some other device, and at least a part of definite DPs –in particular, anaphoric definite DPs- would be banned from the pivot position because they are incompatible with a mode of semantic composition that treats them like predicate modifiers. IS conditions would simply make sure that semantic incorporation will proceed smoothly. In a thetic sentence, some process of ‘integration’ must take place between a predicate and its arguments and adjuncts which turn the sequence into a single informational unit. If integration succeeds, no informational partitions are imposed on the sentence: crucially, there will be no (Aboutness) Topic – Comment partition. This creates an optimal scenario for semantic incorporation. It is well known that definiteness is one of the conditions that tend to block integration, due to the salient discourse status of the referents of definite DPs: this means that definite DPs will resist integration and semantic incorporation, basically because they are good candidates for discourse topics, whereas indefinite DPs will show the opposite tendency. In what follows I will not argue in favor of any of the two mentioned lines of analysis, and I will remain neutral on this issue (although I suspect that an approach in terms of semantic incorporation could be more adequate to address the interplay with IS). The basic assumption I will rely on is that the DE originates in a semantic mismatch, and it is neither a pragmatic problem nor an IS problem itself. This, of course, does not imply that IS and pragmatics are irrelevant for a study of DEs: I intend to demonstrate that both IS and pragmatics must be taken into account, especially when trying to establish the conditions that govern the possible violations of the DE. IS and pragmatics have a prominent role when studying how speakers manage to circumvent the basic ban against definite DPs in certain environments. For clarity’s sake, it is worth emphasizing that IS is a component of the grammatical system, and not a pragmatic factor, as it is often claimed to be: the clearest evidence for such grammatical status comes from the existence of cross-linguistic variation in the syntax of topics and foci.

Chapter Two

80

3. The role of Information Structure 3.1. Topic, Focus and Definiteness That IS is relevant for the appearance of DEs in existential sentences and related constructions is hardly a novel insight. It has been repeatedly pointed out that pivot DPs are typically focal and existential sentences are central cases of thetic constructions, i.e. constructions lacking an aboutness topic, typically receiving an ‘all-focus’ / ‘all-new’ interpretation. I assume that such characterization is essentially correct, and that theticity has some role to play in a study of DEs: DEs are typically associated with focal domains, and no restrictions against definiteness have ever been attested in topical positions. This is in accordance with another widely accepted idea, namely, that aboutness topics are predominantly referential and definite: compared with existential sentences, they impose the opposite condition. The somewhat simplified picture of the relations between (in)definiteness and IS that results presents DE contexts as the mirror image of topic positions. Table 1 represents this view schematically. IS value of positions Topic positions Focal positions in DE contexts

Effects on (in)definiteness Definite DPs licensed; indefinite DPs often rejected Definite DPs mostly rejected ; indefinite DPs licensed

Table 1. (In)definiteness and IS. The way Table 1 describes the data deserves some comment. I believe it correctly captures the essence of the constraints, but a number of qualifications have to be added to get a more accurate picture of the situation. Well known evidence in favor of the dichotomy in Table 1 comes from languages without articles. Russian is a good example of how IS determines the interpretation of bare nouns in languages without articles. According to Geist (2012: ch. 4), bare NPs in Russian cannot be assigned an indefinite reading if they appear as aboutness topics: existential indefinite readings require the NP to be a part of comment (cf. Erteschik-Shir 2013). Mandarin and Cantonese are also languages without articles which systematically assign definite or generic interpretations to preverbal (topical) subjects. This confirms the existence of some kind of link between aboutness topics and definiteness (Lyons 1999: 232-236). Despite such correlation, Table 1 still misrepresents some important facts,

Definiteness Effects

81

since the notions ‘topic’ and ‘focus’ are too general. A more fine-grained typology of IS notions is needed. Let’s first review the case of topics. It is widely assumed that indefinites are typically excluded from detached or dislocated topic positions, except if they are generic or specific (cf. Reinhart 1981, Erteschik-Shir 1997, 2013, Cohen and Erteschik-Shir 2002, Gundel and Fretheim 2004, Endriss 2009, Dalrymple and Nikolaeva 2011). As topics must represent adequate addresses for information update, the only possibility for indefinites to count as aboutness topics seems to require a ‘strong’ reading, i.e. a specific or generic reading. According to such a perspective, only referential expressions can serve as aboutness topics. However, topics do not strictly impose a constraint against indefinite DPs: what topics quite often give rise to is, rather, a restriction against non-specificity, which is not what Table 1 predicts. Moreover, in many languages left-dislocated topics even allow for non-specific indefinites, in spite of received wisdom: the examples in (14) make a representative sample –notice that left dislocation introduces new, shifted topics (cf. Leonetti 2013: 130-141 for discussion). (14)

a.

b.

c.

Un gelat, me’ l menjaria amb An ice-cream CL.1SG-CL.MASC eat.COND.1SG with molt de gust. much of pleasure ‘An ice-cream, I would eat with much pleasure.’ (Catalan; Vallduví 2002: 1255) Un libro, l’ ho letto. A book, CL.MASC read.PRF.1SG ‘(At least) one book, I read.’ (Italian; Rizzi 2005) Alguna novela, el jurado (la) va a Some novel the jury CL.FEM go.PRS.3SG to descartar. discard ‘Some novel, the jury will discard.’ (Spanish; Leonetti 2013: 139)

Chapter Two

82

d.

Drei Three

Filme films

von by

Hitchcock(, die) Hitchcock them

kennt jeder. know.PRS.3SG everybody ‘Three of Hitchcock films everybody knows.’ (German; Endriss 2009: 46) e.

Brains you are born with. A great body you have to work at. (Ward and Prince 1991)

With data like these in mind, the situation with topics becomes really puzzling: on the one hand, they seem to host even non-specific indefinites on certain occasions; on the other hand, they are known to constrain the occurrence of indefinite DPs, and in particular non-specific DPs, in some cases. As argued in Leonetti (2013), some progress can be made by distinguishing between internal, unmarked topics (for instance, preverbal subjects, or shifted objects in certain languages) and external, marked topics (dislocated phrases). The constraints against non-specific DPs are typically found in internal topics, depending on the IS rules working in each language. External topics are more permissive, as shown in (14), possibly because they overtly encode an instruction to be interpreted as addresses for information update, and such rule activates certain accommodation processes when the dislocated DPs are not prototypical aboutness topics –for instance, with non-specific indefinites- that makes them acceptable with a contrastive reading. Contrast seems to be a necessary component of the interpretation of non-referential indefinite external topics: all examples in (14) require some kind of contrastive reading in their topics. Quite informally, contrast is the price a nonspecific indefinite has to pay to be accepted as a topic. More generally, as suggested in Leonetti (2014a), contrast provides the interpreter with a reward for adjusting the interpretation of a nominal constituent in a context where “something more” is required to access the intended referent / reading, basically because it forces the interpreter to consider a range of contextual alternatives that expands the context and triggers contextual effects, in Relevance-theoretic terms. The consequence for the simplified panorama in Table 1 is that at least two different kinds of topics must be distinguished in order to capture the correlations between topicality and definiteness / specificity; otherwise, a rather distorted view will result. Given this, it should not be surprising if the same need appears when investigating the correlations between Focus

Definiteness Effects

83

and definiteness. One of the central claims in this paper is that there is indeed a link between the DE and IS, but one that involves some notion which is more specific than the plain idea of Focus. The next two sections will be in fact devoted to explore the consequences of the distinction between wide focus and narrow (possibly restrictive or contrastive) focus. The naïve view underlying Table 1 should thus be revised and reorganized into a more articulated schema, which is well beyond the limits of this work. It is worth including here one final observation which is related to the examples in (14) and will become relevant later on, in section 3.2: in spite of their focal nature, pivots of existential sentences can be dislocated as external topics. Quite often this operation gives ill-formed results, as shown in the examples in (15). (15)

a. b.

#An owl, I think that there is in the garden. *Un livre, il y a sur la table. (French) a book it LOC have.PRS.3SG on the table

However, at least in Romance languages like Italian and Spanish, it is possible to obtain acceptable sentences with dislocated pivots under certain conditions: notice that in (16)-(17), indefinite topics are contrastive, as expected for pivots –in (16) the interpretation is equivalent to ‘at least some mistakes’, and in (17) it could be paraphrased as ‘As for white sharks (but I will not say anything about other species)’. (16)

Alcuni errori, ci sono, Some mistakes LOC be.PRS.3PL ‘Of course, there are some mistakes.’

certamente. (Italian) certainly

(17)

Tiburones blancos, (los) hay también en el Mediterráneo. sharks white them have.PRS.3SG also in the Mediterranean ‘White sharks, you can find in the Mediterranean too.’

In section 3.2 I intend to show that the conditions on the dislocation of pivots are actually the same as the conditions on definite pivots. This is why constructions like (16) and (17) will be relevant in the ensuing discussion.

84

Chapter Two

3.2 Definite Pivots inside Wide Focus There is general consensus about the focal status of pivots in existential sentences: it is widely assumed that (a) they can never be topics (leaving apart the cases of dislocation just mentioned in section 3.1), and (b) they must introduce referents that are new to the hearer, which suggests that they must be in focus. This is known as the ‘Novelty Condition’ (McNally 1992/1997, 2011). A confirmation of the focal status of pivots is the ban on generic readings of indefinite DPs in existential sentences, since generic readings are typically linked to topical positions. Then, if (a) pivots are focal, (b) Focus is related with novel information, and (c) indefiniteness is also related to the introduction of new referents in discourse, Focus must be a crucial ingredient for the DE. Nevertheless, Focus in itself is not enough to account for the DE. It is quite evident that definite expressions may be under focus, or a part of focus, without problems. Focus and definiteness are not incompatible at all: it is possible to mark definite DPs either as contrastive foci or as (parts of) informative foci. Thus, the situation is more or less the following: Focus must be related to the DE, if the constraint only affects focal DPs, but it is obvious that not all DPs in focal positions are affected by the DE. Something more is needed to explain the distribution of DEs. Focus alone cannot provide a full answer to the main questions we face. In order to get some clarification of the issue, two aspects of the role of Focus must be better specified: 1. Which kind of focus is relevant for the constraint? 2. How is Focus connected to the interpretive conditions imposed by existential sentences? This section is devoted to a discussion of such points. As for the first point, we can safely assume that informative focus, instead of contrastive focus, must be the relevant notion for the exclusion of definite DPs, simply because contrastive focus is perfectly compatible with definiteness (cf. contrastive focalization in Spanish: EL PAN te dije que trajeras ‘THE BREAD I told you to bring’). Once this has been established, we can look at informative focus in more detail, and try to ascertain whether the distinction between wide focus (as sentence focus or predicate focus) and narrow focus (so called “argument focus”) can throw some light on the behavior of pivots. The Spanish examples in (18) show that for indefinite pivots the distinction seems to be irrelevant: in (18a) the pivot is interpreted as a part of wide focus, whereas in (18b) it receives a narrow focus (argument focus) reading, and in both cases it is straightforwardly licensed.

Definiteness Effects

(18)

85

a.

A: - ¿Qué viste allí? What see.PST.2SG there? B: - De todo. Incluso había una leona Of everything even have.PST.3SG a lioness en el jardín. in the garden ‘A: - What did you see there? B: - All kinds of things. There was even a lioness in the garden.’

b.

A: - ¿Qué hay en el congelador? What have.PRS.3SG in the freezer? B: - Pues mira, hay salmón, verduras, y Then look have.PRS.3SG salmon, vegetables unos helados. and some ice-cream ‘A: - What is there in the freezer? B: - Look, there’s salmon, vegetables, and some ice-cream.’

However, things look different in the case of definite DPs, in particular with definites that cannot be interpreted as first mention DPs. There is some evidence that the DE is in force when the pivot occurs as a constituent inside wide focus. At first sight, at least, this should be related to the thetic nature of existential contexts. Notice that, if this is on the right track, Focus structure –in particular, the distinction between wide focus and narrow focus- enters the picture as a condition on the insertion of definite pivots, i.e. in cases in which a mismatch arises between the pivot and the existential context. The basic argument in favor of wide focus as a key condition for DEs – or, more precisely, for banning anaphoric definite DPs from certain contexts- is the fact that some languages show restrictions on definite pivots with one particular syntactic schema, namely ‘V + DP + Locative’. For the DE to hold, it is crucial that the pivot precedes the locative adjunct, and that both are inside the VP, integrated into wide focus. If the locative were 1. removed, 2. fronted or 3. dislocated, the DE would vanish or considerably weaken, which suggests that definiteness is in some sense incompatible with the presence of the locative only when it is explicit and occurs in the rightmost position inside the VP. As Sheehan (2006: 150) notices, these are exactly the conditions under which the expletive there occurs in English. Therefore, she argues, this parallelism strongly supports the existence of null expletives in VSPP order –verb + subject + prepositional phrase- in Romance null subject languages and their connection with DEs. Although I depart from such perspective, Sheehan’s insights must be seriously taken

86

Chapter Two

into account, since they highlight the importance of wide focus inversion as the locus of DEs. I will review examples from Italian, Catalan, Portuguese and English that bear on the role of wide focus. Italian provides us with clear examples of the effects of locatives both in unaccusative inversion and in existential sentences. Belletti (1988) showed that in unaccusative and passive structures the DE can be observed when an overt locative follows the subject, as in (19)-(20) (see also Benincà, Salvi and Frison 1988). (19)

È entrato {un ladro / *il ladro} dalla finestra. Be.PRS.3SG entered a thief / the thief from-the window ‘{A thief / The thief} entered through the window.’ (Cf. È entrato il ladro)

(20)

È stato messo { un libro / *il libro} sul tavolo. Be.PRS.3SG been put a book / the book on-the table ‘{A book / The book} has been put on the table.’ (Cf. È stato messo il libro, sul tavolo)

In Belletti’s analysis, the contrast between VSPP (with DE) and VS (without DE) has to do with the syntactic position of the inverted subject: it occurs in the pivot position –an internal argument slot- only in the first case. In VS, the subject could also appear in a higher position, adjoined to VP, as any other postverbal subject in a null subject language (in other versions of the proposal, this position is the specifier of a low Focus Phrase). According to Belletti, once the subject is not inside the VP, the DE vanishes. I will choose a slightly different interpretation of the facts: in my view, the DE holds when the DP is included in the domain of wide focus. According to Leonetti (2008), the same kind of contrast between final and non-final position for definite DPs is found in existential sentences in Italian (and Catalan). Again, the definite DP –the pivot- is acceptable only if the final locative PP is removed or dislocated (cf. the Coda Constraint in Leonetti 2008: 144), as shown in (21) and (22) –except in so-called ‘eventive existentials’5, where definite pivots are perfectly compatible 5

Eventive existentials are clear counterexamples to the DE in languages like French, Catalan, Italian and English: in this case, instead of asserting the presence of some entity somewhere, the sentence introduces a whole situation that is taking place at some moment and place, as in English examples like There is that man on the phone again or There is John waiting at the door for you, from Lyons (1999: 239). It is not completely clear what the syntactic analysis of the construction

Definiteness Effects

87

with the coda; the ill-formed sentences in (21) and (22) are marked with ‘??’, instead of an asterisk, because the anomaly is subtle and sometimes difficult to perceive for speakers, as it depends on Focus structure and intonational contour. (21)

a. b.

(22)

a. b.

??C’è quel problema nelle prime pagine6. LOC be.PRS.3SG that problem in.the first pages C’è quel problema, nelle prime pagine. / C’è quel problema. / Nelle prime pagine, c’è quel problema. ‘??There is that problem (in the first pages).’ ??Hi havia el degà a la reunió7. LOC have.PST.3SG the dean to the meeting Hi havia el degà, a la reunió. / A la reunió, hi havia el degà. / Hi havia el degà. ‘The dean was present at the meeting.’

So far, Italian and Catalan show that the DE is in force when the offending definite DP is followed by a locative phrase, whereas it is not when the DP occupies the rightmost position in the core sentence. Although this pattern cannot be found in every language, due to factors that I will discuss in section 4.1, it is certainly not a mere coincidence, and, moreover, it is significant for the connection between IS and the DE. Before discussing the implications of this fact, it is worth adding some cross-linguistic evidence of the presence of the pattern. Sheehan (2006: 149ff) points out that in European Portuguese postverbal definite subjects in unaccusative inversion are not licensed if followed by an overt locative; when the locative is not in the final position, the definite subject is acceptable, as shown in (23).

should be in each language (see Villalba 2013 for Catalan), but it is sure that eventive existentials should be distinguished from proper existentials, and the distinction reflects Sasse’s (1987) previous distinction between ‘entity-central’ and ‘event-central’ thetic sentences (valuable proposals can be found in Lambrecht 2002 and Dobrovie-Sorin and Beyssade 2004/2012 for French, and De Cesare 2006 and Cruschina this volume for Italian). I will not deal here with this kind of violations of the DE. 6 Needless to say, a reading of nelle prime pagine ‘in the first pages’ as a nominal modifier must be excluded. In such case, the PP would not be a coda, and the definite DP would be fully acceptable. 7 The Catalan data were first pointed out in Rigau (1994, 1997).

Chapter Two

88

(23)

Chegou {alguém / *o João} ao colégio. Arrive.PST.3SG someone / the João to-the school ‘{Someone / João} arrived at school.’ (Cf. Chegou o João / Ao colégio chegou o João)

Thus, the DE can be noticed only in VSPP wide focus inversion, but not in the VS pattern. This reproduces the situation already described in Italian subject inversion. As for existential sentences, in European Portuguese the DE is quite strong, due to the division of functions between sentences with haver –DE in force- and sentences with estar –no DE-, like in Spanish (cf. section 4.1); counterexamples are not so common. Nevertheless, they are not impossible to find, and it is significant that definite pivots in European Portuguese are, again, not followed by any coda: they appear in the final position. The examples in (24) were collected from a journalistic corpus (I am grateful to Fátima Oliveira for providing me with the data), and confirm the validity of the generalization: definite DPs circumvent the DE when they receive a narrow focus reading in final position (but the DE holds when they appear inside wide focus). Notice that the examples in (24) are analogous to the well-known ‘list’ readings of existential sentences in English (cf. section 3.3). (24)

a.

b.

c.

Para sobremesa, além de fruta, havia os doces For dessert, apart of fruit, have.PST.3SG the sweets seguintes: leite creme, following: custard, mousse e pudim de café. mousse and pudding of coffee ‘As dessert, apart from fruit, there were the following sweets: custard, mousse and coffee pudding.’ …nessa altura também havia os bares in that time also have.PST.3SG the bars gregos, onde… Greek, where… ‘At that time there were also the Greek bars, where…’ Do lado da Morais, havia os on the side of Morais have.PST.3SG the empregados e o caixeiro, … employees and the cashier ‘On the side of Morais, there were the employees and the cashier…’

Definiteness Effects

89

Finally, even English offers some support for a generalization about the non-final position of the offending DP in DE contexts. Safir (1982) points out that in presentative there-constructions certain contrasts arise that seem to reproduce the same pattern that is found in Italian and Portuguese; the relevant examples are in (25) –quoted in Belletti (1988)-. (25)

a. b.

*There hung the coat on the wall. There hung on the wall the flag of the country that John had fled.

Both Safir and Belletti suggest that the definite DP appears in different positions in (25a) and (25b): a position that is internal to VP in (25a), where the DE holds, and one that is adjoined to VP in (25b), where the DE is not in force. Moreover, the “heavy” nature of the DP in (25b) favors its rightmost position. I agree with the original intuitions, and it may well be true that different positions are involved in the contrast, but I believe that the crucial distinction is related to IS: only in (25a) is the definite DP inside wide focus. This applies to the whole variety of facts reviewed8. The surface generalization that emerges from the data is the following: definite DPs are excluded in DE contexts when they are followed by another constituent inside the VP, while they manage to “survive” when they occur in the rightmost position. It seems that definite pivots –or at least the most referential and discourse-linked among them- tend to avoid the internal position followed by the coda because they cannot be licensed in a DE context if they remain inside the domain of wide focus. When 8

An additional piece of evidence that is worth mentioning comes from Basque. Etxeberria (2012) confirms that the DE is found in Basque, as shown by the ungrammatical status of (i), but he crucially points out that in the varieties that create existentials by means of locative inversion, a definite pivot is allowed when the locative coda is topicalized, and a comma is introduced between the coda and the pivot, which gives rise to the contrast between (ii) and (iii). This strikingly reproduces the pattern already observed in Italian and Catalan. Though it seems that the facts can be again derived from the interaction of definiteness and wide focus, I leave this issue for future research. (i) *Zopan eulia dago. soup.DEF.in fly-Def be.PRS.3SG ‘*There is the fly in the soup.’ (ii) *Atean gizona dago. Door.DEF.in man.DEF be.PRS.3SG ‘*There is the man at the door.’ (iii) Atean, gizona dago Door.DEF.in man.DEF be.PRS.3SG

90

Chapter Two

definite pivots occur in the rightmost position, they are naturally interpreted as narrow foci, thus avoiding to be included in wide focus. In section 3.3 I show that this is in fact a systematic feature of a well defined class of DE violations. The fact that it is wide focus what really determines the exclusion of definite DPs fits quite naturally in an approach to the DE based on some variety of semantic incorporation. If the domain of Focus extends all over the sentence, a single informational unit is obtained, and some kind of integration between predicate, arguments and modifiers is needed. This results into an optimal context for semantic incorporation. As definiteness is a factor that hinders integration, in particular with anaphoric and discourse-linked definites, it will predictably produce a mismatch in contexts requiring semantic incorporation. Simply assuming that IS rules may vary from one language to another, certain apparent problems with cross-linguistic variation can be integrated under this approach. In the Romance domain, not all languages behave like Italian and Catalan in unaccusative inversion patterns. Whereas Italian, Catalan and Portuguese show a DE in VSPP order, as shown in (19)-(20) and (23), in Spanish and Romanian the restriction is absent, and definite postverbal subjects are allowed both in VS and in VSPP orders, as shown in (26)-(27). (26)

Ha entrado el ladrón por la ventana. Have.PRS.3SG entered the thief for the window ‘The thief entered through the window.’

(27)

A intrat Ion prin geam. Have.PRS.3SG entered Ion through window ‘Ion entered through the window.’

(Spanish)

(Romanian)

The absence of DE in the VSPP pattern in Spanish and Romanian represents a problem for the generalization on wide focus, since a definite DP is allowed in a grammatical context that is typically associated with a wide focus reading (see Leonetti 2014b for an analysis of VSX order in Romance). Nevertheless, there is a simple explanation for such contrast. In Romance, VSX is a marked pattern that needs a thetic / wide focus interpretation. Italian and Catalan, being languages that need to resort to explicit marking of informational partitions in marked word orders –by means of dislocation, focalization, clefts…, tend to exclude the formation of complex strings without partitions, and are strongly sensitive to factors that favor or hinder integration –like definiteness (recall the natural association

Definiteness Effects

91

of definiteness and topics). Spanish and Romanian represent, among the Romance languages, the opposite behavior: they are quite permissive with the formation of informational chunks in marked word orders, in the sense that they do not need to establish informational partitions by means of dislocations or cleft structures, so that VSX sentences are allowed with a wide focus reading, i.e. as single informational chunks, without problems. I assume that this is essentially the reason why Italian and Catalan lack VSX order, whereas Spanish and Romanian display it: Spanish and Romanian allow for wide focus readings of sentences that would need some Topic – Comment or Focus – Background partition in other Romance languages (for instance, VSX sentences). Due to this IS feature, they are not particularly sensitive to factors like definiteness in sentences requiring informational integration, and thus wide focus VSX inversions are acceptable even with definite postverbal subjects. Then, it is this feature of IS –tolerance to complex strings without informational partitions- what guarantees that VSPP sentences will show no DE in Spanish and Romanian. This type of cross-linguistic variation is perfectly compatible with what we know about DEs. And once again, some aspects of DEs appear to be dependent on certain features of IS. One last set of data must be introduced here, in order to enrich the empirical basis of the generalization on wide focus. The data concern the possibility of dislocating an indefinite pivot DP in an existential sentence in Spanish and Italian (cf. section 3.1). Left dislocation of the pivot implies introducing a resumptive element in the construction: the resumptive is an accusative clitic in Spanish, and a null pronoun (pro) in Italian. The contrasts in (28) and (29), where the dislocated contrastive topics are optional, suggest that the operation is subject to strict conditions. (28)

a. ??(Un problema,) lo hay A problem CL have.PRS.3SG

en la oficina. in the office (wide focus)

b. (Un problema,) lo hay, (en la oficina). ‘There is a problem at the office.’ (coda as given information) (29)

a. ??(Alcuni errori,) ci sono pro some mistakes LOC be.PRS.3PL

nelle prime pagine in.the first pages

b. (Alcuni errori,) ci sono pro, (nelle prime pagine). ‘Some mistakes, you can find in the first pages.’ (coda as given information)

Chapter Two

92

Resumptive elements associated to (explicit or implicit) dislocated topics seem to be acceptable either when the coda is topical and given –it may be implicit or detached- or when the coda is under narrow focus. In the first case, the sentence can be used as an answer to a question like ‘Are there problems at the office?’ or ‘Are there mistakes in the first pages?’; in the second, it can be an answer to a question like ‘Where do you think there are problems now?’ or ‘Where can I find some mistakes?’. In both situations, represented by (28b) and (29b), an informational partition is established between the coda and the rest of the construction, and the resulting sentence is acceptable. In (28a) and (29a), on the contrary, the coda is not separated as a topic or as a narrow focus, and a wide focus reading is assigned to the sentence; it could be an answer to a question like ‘Have you identified some problem?’ or ‘Did you find some mistakes?’. The result is an awkward utterance. This implies that the presence of the coda imposes the same conditions both to the insertion of definite pivots and to dislocation of the pivot. Both phenomena obey the Coda Constraint. Further confirmation of the constraint is found in existential sentences that accept only an ‘all-new’ reading, like the exclamations in (30): dislocation of the pivots and the corresponding insertion of a resumptive are forbidden, as shown in (31). (30)

a. b.

(31)

a. b.

9

¡Hay una serpiente Have.PRS.3SG a snake C’è un serpente LOC be.PRS.3SG a snake ‘There is a snake in the kitchen!’

en la cocina! in the kitchen in cucina! in kitchen

??La hay en la cocina. CL have.PRS.3SG in the kitchen ??C’è pro in cucina.9 LOC be.PRS.3SG in kitchen

Notice that in (29) and (31b) an interesting interaction between the DE and prodrop can be observed: as the null pronominal subject is interpreted as definite and anaphoric, it is excluded from the existential context, but if the partitive clitic ne is used, an indefinite interpretation of the pronominal pivot is guaranteed, and thus the sentence is perfectly acceptable (cf. Forchette? ??Ci sono in cucina, with a wide focus reading vs Forchette? Ce ne sono in cucina ‘Forks? There are some in the kitchen’). This confirms that there is in fact a DE in Italian existentials when the coda is inside wide focus. Pro-drop is allowed if there is no coda or if the coda is under narrow focus. See Burzio (1986: 129) for some observations on pro-drop in existentials.

Definiteness Effects

93

It seems that the only possibility for a definite resumptive to occur in an existential construction is to be kept apart from the locative coda in IS: whenever the resumptive and the coda happen to appear together inside the domain of wide focus, there is a clash. The clash is surely due to the association between the resumptive and the topic: if the resumptive links a topic with an internal position in the sentence, it cannot be taken as a part of wide focus at the same time, because a contradiction would arise. What is relevant here is that in existential sentences both the dislocation of the pivot as a topic and the insertion of definite pivots are constrained by IS conditions, and by the same IS conditions: both operations are incompatible with a wide focus interpretation in which focus encompasses the whole string, but compatible with an informational partition that isolates the coda as a topic or as a narrow focus. Moreover, the parallelism suggests that restrictions on topics and restrictions on definite DPs are just two manifestations of one single property of existential sentences, the one that lies behind the DE. Things are a bit more complicated than what I presented here (see Leonetti forthcoming for details), but this is enough to confirm the generalization on wide focus. If the wide focus / narrow focus distinction turns out to be so productive, then it is necessary to look at the effects of narrow focus as well.

3.3 Narrow Focus and Contrast This section aims at presenting evidence that a number of typical DE violations are associated with narrow focus on the pivot. Counterexamples to the DE actually fall into two main classes, which will turn out to be decisive for a better understanding of the problem (cf. Lumsden 1988, Abbott 1993, Birner and Ward 1995, Francez 2007). One group contains instances of so-called ‘self-establishing definites’, i.e. definite DPs that are not anaphoric and may establish new discourse referents, thus conforming to the basic requirement that pivots must be hearer-new. In this kind of definite DPs the uniqueness condition is satisfied by means of DP-internal material, since the descriptive content of the DP is rich enough to identify a unique referent, so that there is no need to look for uniquenesssupporting information in the speech situation or in the preceding discourse. Actually this is the crucial factor that allows them to be used as ‘first mention definite DPs’ and to introduce new discourse referents. This group includes a series of well-known cases: superlatives, cataphoric definites, definites and universal quantifiers with ‘kind’ or ‘type’ readings, and weak definites. The classical English examples reproduced in (32) are a representative sample of the first type of DE violations.

Chapter Two

94

(32)

a. b. c. d. e.

There is the strangest bird in that cage. There was the usual crowd at the beach last Sunday. There was the smell of pot all over the apartment. There is every reason to be suspicious. There is the outline of a human face hidden in this puzzle.

This kind of definite pivots can be found even in languages that display a robust DE, like Spanish (Leonetti 1999: 816-817): they represent a straightforward way for definites to fit in existential contexts avoiding the emergence of a serious mismatch between the construction and the pivot. In fact they don’t need any particular contextual condition to appear, and may be classified among “non-contextualized existentials”, following the generalization in Abbott (1993: 45): “…any NP which can be interpreted as introducing a new entity into the discourse should be possible in a noncontextualized existential.” A second group of counterexamples to the DE, equally well known, concerns anaphoric definite DPs, demonstratives, names and even pronouns, i.e. the kind of definites that cannot be used to introduce new referents and are thus excluded as ‘first mention definites’ –in a few words, discourse-linked definite DPs, i.e. the most reluctant expressions to be included in an existential context. Let’s assume, for simplicity’s sake, that a clear distinction can be drawn between ‘self-establishing definites’ and ‘anaphoric’ definites. This second group of cases shows two salient properties that have been widely discussed in the literature (in particular, in McNally 1992/1997: §5.2, Abbott 1993, Birner and Ward 1995, Hu and Pan 2007: 136, Bentley 2013: 700-702): one is the need for a specific context, which makes them “contextualized existentials”, in Abbott’s terms; the other is a restricted set of special readings that we can dub ‘enumerative’ or ‘availability’ readings (this comprises ‘list’ readings and reminders, just to mention different terms that have been used in the literature). The need for a special context means that contextualized existentials cannot initiate a discourse –in fact they typically occur as reactions or responses to other utterances-, and have the function of drawing the addressee’s attention to the existence of some entity or entities with respect to a set of contextual assumptions, either because the entities have been forgotten or overlooked, or because they are available and suitable for some previously established purpose or goal. The three classical English examples in (33), from Rando and Napoli (1978) and Abbott (1993), illustrate the major features of contextualized existential sentences or ‘list existentials’.

Definiteness Effects

(33)

a. b. c.

95

A: - Is there anything to eat? B: - Well, there’s the leftover chicken from last night. A: - I guess we’ve called everybody. B: - No, there’s still Mary and John. A: - How many people know about this? B: - There’s me and there’s you. That’s all.

It is clear that this group of counterexamples to the DE must be kept apart from the first group, as it represents a radical mismatch between the definiteness of the pivot and the requirements of the construction –the pivot does not conform to the novelty condition. The facts are particularly relevant for an analysis of DEs in connection with IS. Actually, most of the data discussed in this paper have to do with the behavior of ‘anaphoric’ definite DPs. First of all, it is important to recall that they reproduce a systematic pattern, since they can be detected in many different languages, and not only in existential sentences, but also in other DE contexts such as possessive constructions with the equivalents of have. The examples in (34)-(36), from Bentley (2013: 701-702), confirm that ‘list / availability’ readings are associated with discourse-linked definite DPs in DE environments in French, Brazilian Portuguese and Catalan; the examples in (37)-(38) show that the pattern is the same in Spanish and Italian, despite the different distribution of the DE in the two languages. The data are all from the Romance domain, but they suffice to show that the phenomenon appears every time that an anaphoric definite DP10 is used in a DE context (notice that the definite DP refers always to a set of children already established in discourse). (34)

a.

b.

10

Nous ne pouvons pas divorcer: We NEG can.PRS.1PL NEG divorce il y a les enfants. it LOC have.PRS.3SG the children ‘We cannot divorce: there’s the children.’ Nous ne pouvons pas divorcer: nous We NEG can.PRS.1PL NEG divorce we avons les enfants. have.PRS.1PL the children ‘We cannot divorce: we have the children.’

(French)

It is not impossible to obtain an ‘availability’ reading with an indefinite pivot, as pointed out in McNally (1992/1997: 193), especially if the indefinite DP is specific. However, in such cases the reading is much less salient than with anaphoric definites.

Chapter Two

96

(35)

a.

b.

(36)

a.

b.

(37)

(38)

Não podemos nos divorciar: tem nossos NEG can.PRS.1PL CL divorce have.PRS.3SG our filhos. (Brazilian Portuguese) children ‘We cannot divorce: there’re our children.’ Não podemos nos divorciar: temos as crianças. NEG can.PRS1PL CL divorce have.PRS.1PL the children ‘We cannot divorce: we have the children.’ No ens podem divorciar: hi ha els NEG CL can.PRS.1PL divorce LOC have.PRS.3SG the nens. (Catalan) children ‘We cannot divorce: there’re the children.’ No ens podem divorciar: tenim els NEG CL can.PRS.1PL divorce have.PRS.1PL the nens. children ‘We cannot divorce: we have the children.’

No podemos divorciarnos: tenemos NEG can.PRS.1PL divorce have.PRS.1PL los niños11. the children ‘We cannot divorce: we have the children.’

a to

a.

i bambini12 the children (Italian)

b.

Non possiamo divorziare: ci sono NEG can.PRS.1PL divorce Loc be.PRS.3PL

(Spanish)

‘We cannot divorce: there’re the children.’ Non possiamo divorziare: abbiamo i bambini. NEG can.PRS.1PL divorce have.PRS.1PL the children ‘We cannot divorce: we have the children.’

A second crucial fact, already noted in Rando and Napoli (1978), Lumsden (1988: 216-219), Abbott (1993) and Leonetti (2008: 144-145), is 11 In Spanish the existential construction with haber is incompatible with anaphoric definites (*Hay los niños), but the ‘list / availability’ reading shows up with tener ‘have’. This point will be discussed in section 4.1. 12 (38a) is ambiguous in Italian: it can receive an availability reading, but a locative reading is possible too.

Definiteness Effects

97

the absence of the locative coda in ‘contextualized’ existentials with definite pivots: the pivot is systematically the rightmost constituent, as can be observed in (33)-(38). In Abbott (1993) it is claimed that this is due to the given status of the coda in this kind of examples; in Leonetti (2008) the absence of the coda is analyzed as a particular case of the Coda Constraint, already discussed in section 3.2. This grammatical condition should ideally be derived from a global account that captures the whole cluster of properties of ‘contextualized’ existentials. The systematic occurrence of ‘list / availability’ readings suggests that they emerge as the interpretive solution to the semantic incompatibility between the pivot and the grammatical context –along the lines of the pragmatic accounts in Lumsden (1988), McNally (1992/1997) and Abbott (1993). I believe that this idea is worth developing to find an explanation for the two salient properties mentioned before: the ‘contextualized’ status of the sentences with definite DPs in (33)-(38), and the appearance of ‘availability’ readings. Some interesting questions arise at this point: why is a special contextualization required? And why does the context involve a particular set of assumptions instead of others? Why are notions like ‘list’ or ‘reminder’ relevant? Notice that ‘list’ and ‘reminder’ look like intuitive and superficial characterizations of something that still has to be integrated in the grammar of existential contexts (cf. Abbott 1993: 43 and Hu and Pan 2007: 141-142 for some criticism of such characterizations); in fact, it is possible to have lists in non-contextualized existentials as well. In Rando and Napoli’s (1978) proposal, examples like those in (33) escape the DE because it is not the (given) referent of the pivot that is introduced, but rather the list of entities where it belongs, counting the list as new information and thus obeying the Novelty Condition. In a few words, the members of the list may be familiar and given, but the list itself fits adequately in the existential environment. Now, such solution has some appeal, but it is still unsatisfactory: the notion of a list is not restrictive enough, it cannot capture the interpretations of all the relevant examples –as Hu and Pan (2007: 142) note on the basis of Chinese data, what is asserted is not the existence of a list, but the existence of a membership relation between an entity and a presupposed set, which seems a more efficient characterization of the problem-, and it is not clear how it emerges. Abbott (1993: 43) suggests that ‘list’ readings are the result of “the assertion of the existence of an item whose existence is presupposed”. My proposal is in line with Abbott’s, but gives a central role to IS. In a nutshell –as advanced in Leonetti (2008: 154-157)-, the idea is that the set of counterexamples to the DE known as ‘contextualized’ existentials can be defined as the set of existential sentences where an anaphoric definite pivot occurs under narrow

Chapter Two

98

focus. ‘List / availability’ readings are simply the result of placing a definite pivot under narrow focus. The need for narrow focus automatically explains the absence of a locative coda, since the pivot must occur in the rightmost position to be interpreted as narrow focus in a natural way, so that the coda has to be implicit or dislocated. What follows in this section is devoted to reviewing some empirical evidence supporting this idea and to elaborating some of its consequences (see section 4.2 for additional comments). 1. The first piece of evidence in favor of the role of narrow focus obviously comes from the data supporting the Coda Constraint in languages like Italian and Catalan. It is usual to assume that the DE is not present in these languages, given the well-formedness of examples like the following: (39)

a. b.

C’ è Gianni. LOC be.PRS.3SG Gianni Hi ha en Joan. LOC have.PRS.3SG the Joan ‘There is John.’

(Italian) (Catalan)

In Leonetti (2008) it is claimed that the absence of DE is only illusory, and depends on the presence of the locative coda, as already observed in examples (21)-(22). If we exclude so-called ‘eventive existentials’ and weak definites with ‘kind / type’ readings, the data confirm that all alleged cases of violation of the DE in Italian and Catalan display definite pivots that happen to occur as narrow foci, in final position. The generalization may be somewhat obscured by the intriguing status of the esserci construction in Italian, since it seems to conflate the properties of existentials and ‘pseudo-existential locatives’, as argued for in Zamparelli (2000) –see Cruschina this volume and section 4.1. for details; however, it is robust enough to take it as a manifestation of a general tendency. 2. The counterexamples to the DE in Hebrew analyzed in Ziv (1982) show that when anaphoric definite DPs are involved they must be under narrow focus. Ziv (1982: 75-77) claims that (40a) and (40b) are adequate as responses to questions like ‘Who is teaching linguistics at MIT?’ and ‘What’s in the brown closet?’, which implies that the sentences are equivalent to ‘list’ / ‘reminder’ there-sentences in English, and, moreover, that the definite pivots are narrow foci.

Definiteness Effects

(40)

a.

b.

99

yeš šam et xomski / yeš et xomski ‘be’ LOC ACC Chomsky / ‘be’ ACC Chomsky be am ay ti (Hebrew) at MIT ‘There is Chomsky, (at MIT).’ yeš ba’aron haze et hame’il haxadaš šelxa ‘be’ in the closet this ACC the coat the new your ‘In this closet there’s your new coat.’

Ziv considers other cases of definite pivots that do not obey the condition on narrow focus, like the DP hamexonit hazot ‘this car’ in (41), but notices that here the DP has a kind reading, equivalent to ‘this model’. The corresponding IS assigns narrow focus to the locative, and maintains the pivot inside the background constituent. This is possible because the pivot gets a kind reading, thus behaving like a weak definite, perfectly acceptable in the existential context. Such definite pivot needs no special licensing through narrow focus. Despite Ziv’s claim that sentences like (41) are actually more locative than existential, because of the DE violation, I believe we can still analyze them as common existentials: in fact, very similar data can be found in Spanish haber-sentences with accusative clitics as pivots. (41)

yeš et hamexonit hazot ecel kol soxen ba’arec. (Hebrew) ‘be’ ACC the car this at every dealer in the country ‘This car can be found at every dealer in the country.’

3. The next piece of evidence for narrow focus as a licensing mechanism for anaphoric definite DPs can be found in French. Zimmermann (this volume) carefully reviews all kinds of attested cases of violations of the DE in French unaccusative sentences, and finds out that the pivot “is freely allowed to be definite under (at least) one of the following conditions”: - use of restrictive ne…que ‘only’: (42)

Il n’ arrive It not arrive.PRS.3SG ‘Only the girls arrive.’

que les filles. that the girls

- use of the verbs rester ‘remain’ and manquer ‘lack’:

(French)

Chapter Two

100

(43)

Il reste / manque les filles. It remain.PRS.3SG / be-missing.PRS.3SG the girls ‘There remain the girls / The girls are missing.’

- ‘cataphoricity’ of the DP: (44)

Il a été vole à la bibliothèque les It have.PRS.3SG been stolen at the library the suivants13:… following ‘The following books were stolen at the library….’

livres books

- list item status of the DP (enumerative use): (45)

Il a dormi ici Jean, Paul et François. It have.PRS.3SG slept here Jean, Paul and François ‘Jean, Paul and François slept here.’

- restrictive complements in the DP (‘post-determination’): (46)

Il arrive les filles les plus belles du monde. It arrive.PRS.3SG the girls the most beautiful of-the earth ‘The most beautiful girls in the world arrive.’

- ‘genericity’ / non-referentiality of the DP / participation in idiomatic expressions: (47)

Il lui venait la sueur au front. It him/her come.PST.3SG the sweat to.the forehead ‘It brought the sweat to his/her brow.’

Descriptive grammars usually offer a list of contexts and factors like this, instead of a unified account of the availability of definite DPs. However, such account is not difficult to obtain, if one realizes that the factors are essentially the same ones already pointed out for the DE in English. More precisely, cataphoricity, restrictive complements and genericity / non referentiality belong to the series of factors characterizing the group of expressions we dubbed ‘self-establishing definites’ –a group that includes 13 In (44) the verb is not unaccusative itself, but the construction fits in the list adequately due to the unaccusative properties of passive sentences.

Definiteness Effects

101

weak definites, superlatives and kind readings, and all ‘first mention’ definite DPs. Such group is not particularly relevant here: it represents the kind of DE violations that is commonly found in different languages, with DPs that comply with the Novelty Condition. More interesting is noticing that the use of restrictive ne…que, rester and manquer, together with ‘list’ / enumerative uses, gives rise to a compact set of cases that share an important property: the postverbal definite DP is focused, and in particular it must be interpreted as narrow focus, mostly with a contrastive value (the same applies to simple examples like Il est arrivé l’ électricien ‘The electrician arrived’). Of course, the DP –a definite DP that typically receives an anaphoric interpretation- occurs in the rightmost position. Enumerative readings are the same phenomenon we just discussed for other languages, and ne…que has the same effects as the focal adverbs I will mention below: the facts are the same again, and definite DPs are licensed by narrow focus and contrast in a context where indefinite expressions are clearly preferred. The pattern is the same in enumerative existential sentences with avoir (cf. Dobrovie-Sorin and Beyssade 2004/2012: 110). Narrow focus reveals itself as a systematic factor in a well defined set of counterexamples to the DE. Moreover, postverbal definite DPs with enumerative readings display a number of common properties with inverted subjects appearing in the French construction Lahousse (2011) dubs as focus inversion: this kind of inversion pattern is characterized by narrow, contrastive focus on the subject. 4. An additional argument for the crucial role played by narrow focus is provided by the insertion of focus particles and focal adverbials with anaphoric definite DPs in certain languages. Catalan is a good example. As I already mentioned, Catalan allows for definite descriptions and names to occur as pivots in aver-hi constructions, but typically excludes personal pronouns, as shown in (48) –which is to be expected, if pronouns are the least prone elements to be inserted in an existential context. (48)

*Hi ha ell (al pati). LOC have.PRS.3SG he at-the courtyard ‘There’s him (in the courtyard).’

Interestingly, in spontaneous use and even in the written language it is possible to find cases of pronominal pivots, always in final position, without coda, and under the scope of a focal adverb like només ‘only’, as in (49) (I am grateful to Xavier Villalba for providing me the data; cf. Villalba this volume):

Chapter Two

102

(49)

a.

b.

Rafael Azcona representa la literatura Rafael Azcona represent.PRS.3SG the literature cinematográfica. A Itàlia els cinematographic at Italy the millors escriptors escrivien per al cinema, però best writers write.PST.3PL for the cinema, but a Espanya, gairebé només hi at Spain, almost only LOC ha ell. have.PRS.3SG he ‘Rafael Azcona represents the literature for cinema. In Italy the best writers used to write for cinema, but in Spain there was hardly anyone else apart from him.’ (El Punt, 23-12-11) Keane o Arctic Monkeys tenen bones Keane or Arctic Monkeys have.PRS.3PL good cançons, però pel que fa songs but for.the that do.PRS.3SG a personalitats, només hi ha ella. to personalities, only LOC have.PRS.3SG she ‘Keane or Arctic Monkeys have good songs, but as for personalities, there’s only her.’ (El Periódico, 24-6-11)

If narrow focus is essential for a strong pronoun to survive in an existential environment, here the focal adverb conspires with the rightmost position to make sure that the pronoun receives a narrow focus reading, enriched by contrast. The role of the focal element can be expected to be decisive in the licensing of strong pronouns as pivots (this was already pointed out in Beaver, Francez and Levinson 2005). Hu and Pan (2007) show that certain significant cases of violation of the DE in Chinese –excluding those related to eventive existentialsinclude a definite pivot not followed by a coda, and crucially associated to a focus particle like hai ‘in addition’, as in (50): (50)

Hai you neige ren / Zhangsan. In-addition have that man / Zhangsan ‘There is in addition that man / Zhangsan.’

Definiteness Effects

103

It seems clear that Chinese confirms the generalization: if an anaphoric / given definite pivot is acceptable in an existential sentence, it must be under narrow focus. 5. Finally, there is an observation made in Hartmann (2006) concerning ‘list’ readings in English that fits nicely into this discussion. Hartmann claims that the distinction between existential and ‘list’ readings in there-sentences parallels the distinction between predicational (John is a good friend) and specificational (My best friend is John) readings in copular sentences. Specification opens a list of which the referent of the postcopular DP is a member. The same interpretive operation is at work in ‘list’ readings. The simplest way to capture the parallelism is resorting to Focus structure: what the two constructions have in common is just narrow focus on the postverbal definite DP, and it is narrow focus that is responsible for the contextual emergence of something like a list –a point I deal with in what follows. Now that some cross-linguistic evidence for the role of narrow focus has been presented14, it is possible to show that narrow focus is the key notion to understand the pervasive phenomenon of ‘list’ / ‘availability’ readings in counterexamples to the DE. My claim is that all properties of such readings can be derived from narrow focus. The more salient syntactic feature is the absence of a locative coda in ‘contextualized’ existentials, and it has already been shown to be a consequence of Focus structure. As for the remaining properties, they will be discussed in section 4.2 and explained as pragmatic elaborations of narrow focus.

4. Pragmatics: Inferring Relevant Interpretations 4.1 Competition between Candidates This section aims at clarifying what the role of pragmatics is in the distribution of the DE. I assume that, under normal circumstances, pragmatic inference cannot override syntactic constraints and turn ungrammatical sentences into grammatical ones. However, it can still play a major role in adjusting interpretations in context and accounting for facts that are not completely determined by the grammatical system15. 14

Notice that even in a language like Hungarian, where constraints on definiteness are strongly dependent on aspectual and lexical factors, narrow focus is able to “neutralize” the DE under appropriate conditions (cf. Szabolcsi 1984, Kiss 1995, a. o.). 15 I assume that the role of pragmatic inference in utterance interpretation is essentially the one established in Relevance Theory (cf. Sperber and Wilson 1986 / 1995, Clark 2013).

Chapter Two

104

Simplifying a bit, pragmatic inference is crucial in two facets of interpretation: one is the selection of the relevant reading of a sentence on the basis of the set of possibilities offered by the grammar for the expression of a certain propositional content, and the other one is the resolution of interpretative mismatches by reaching an adequate interpretation that is compatible with the context and the semantics of a construction. I briefly deal with the first issue in what follows; the second aspect will be discussed in section 4.2. Consider the data in (51) and (52) from Spanish and Sardinian respectively: (51)

*Había {el rey / él / Felipe}. Have.PST.3SG the king / him / Felipe ‘There was {the king / him / Felipe}.’

(52)

*B’ ada Juanne. LOC have.PRS.3SG Juanne ‘There’s Juanne.’

(from Bentley 2004: 76)

The strongly deviant status of (51) and (52) shows that the DE is particularly robust in Spanish and Sardinian, if the two languages are compared with English and French, for instance, where the equivalent sentences with be and avoir are acceptable, with an ‘availability’ reading (cf. There was the king; Il y avait le roi). The question is why ‘availability’ readings for definite pivots cannot “save” existential clauses with haber ‘have’ in Spanish and with áere ’have’ in Sardinian, while being systematically acceptable in other languages. There is a simple and natural answer to such question: the constraint must be related to the presence of alternative ways of expressing the same content in Spanish and Sardinian, i.e. to the competition between alternative syntactic forms for one single meaning. In Spanish the competition involves the two verbs haber ‘have’ and estar ‘be’: while haber has specialized as an exclusively existential predicate, estar is a copular verb, and thus, is the straightforward way to express a locative predication. As expected, the DE shows up only in haber sentences. Estar is compatible with all kinds of definite DPs: in (53), ‘availability’ readings are among the possible interpretations of the sentence, together with locative readings. (53)

Está Be.PRS.3SG

{el rey / él the king / him

/ Felipe}. / Felipe

Definiteness Effects

105

Given this situation, it is reasonable to think that the strength of the DE with haber –i.e. the systematic exclusion of definite anaphoric DPs as pivots- is due to the impossibility of obtaining ‘availability’ readings that could be more easily obtained with estar: forcing the insertion of an anaphoric definite pivot with haber is a costlier option than using it with estar, since haber imposes conditions on pivots –typically, that they cannot be previously identifiable, in Bentley’s (2004) terms- that estar does not require. Such conditions imply that some kind of reinterpretation or interpretive adjustment is needed in the case of haber in order to license the definite pivot (cf. 4.2), whereas no extra cost is implied in combining a definite subject DP with estar. Thus, ‘availability’ readings of definite pivots are excluded with haber for economy reasons16. The reasoning is essentially the same as that which allows us to explain, for instance, why coreferential readings of strong pronouns in null subject languages –i.e. the readings that build topic continuity- are usually discarded in contexts where a null pronoun could be used instead of an overt one: in contexts where a null pronoun is a grammatical option, it will represent the optimal tool to convey a coreferential reading because it will be the simplest and most economical way to obtain such reading, with overt pronouns limited to the expression of non-coreferential readings –topic shift-, due to their marked status; if no additional factors intervene, a coreferential reading for overt pronouns will thus be highly dispreferred. In a few words, if more than one form is available to express a single meaning, the unmarked formal option –the most economical one compatible with that meaningshould be the optimal solution. The mechanism I am invoking may remind the reader of similar devices operating in Optimality-theoretic accounts of the syntax – semantics interface (cf. Blutner, Hendriks and de Hoop 2006). However, there is at least a significant difference with respect to such accounts: in my view, economy considerations involved in the competition between alternative forms interact with general pragmatic principles, and

16

A potential problem for this account comes from the behavior of the verbs aver ‘have’ and esser ‘be’ in Catalan: the use of esser in locative predications does not block the possibility of having ‘availability’ readings in definite pivots with aver (cf. Rigau 1994, 1997). The division of labor between the two verbs is possibly different from the one we find in Spanish with haber and estar: in fact, the typical ‘availability’ readings are not possible with esser, but they are with aver, so that a dividing line between the two options can still be drawn. European Portuguese represents another instance of the same basic problem. I leave this issue for future research.

106

Chapter Two

not directly with syntactic principles17. Notice that, if this is right, the evaluation of the competing options, and the consequential blocking of the costly interpretive solution in favor of a simpler one, are not a part of the grammatical system, but rather inferential operations performed on the output of grammatical rules. As is well known, under certain conditions pragmatic inferences can become conventionalized and integrate into the grammar. In Sardinian we find a parallel competition in presentational sentences between two options, one with ‘have’ –without agreement between the verb and the pivot- and one with ‘be’ –with agreement between the verb and the pivot. The DE appears with ‘have’, but not with ‘be’, as proved by the contrast between (52) and (54); as expected, definite pivots with ‘be’ can receive ‘availability’ readings. (54)

B’ este Juanne. LOC be.PRS.3SG Juanne (from Bentley 2004: 76)

As Bentley (2013) notes, Sardinian displays a system of differential encoding of the pivot, controlled by definiteness and specificity. Both agreement and the ‘be’ vs ‘have’ alternation, in fact, encode “the cognitive status in discourse” of the entities denoted by the pivot DP (Bentley 2004: 66). The construction with ‘have’ is associated to a requirement on pivots: they must denote “brand new unanchored” referents (Bentley 2004: 87); the construction with ‘be’ does not include any condition, and freely accepts definite DPs. Thus, the contrast between (52) and (54) in Sardinian can be explained along the same lines as the corresponding one in Spanish: the use of anaphoric definite pivots with ‘have’ is barred because the same content can be expressed at no extra cost with ‘be’. Languages like English and French, on the other hand, display ‘availability’ readings for definite pivots in existential sentences because there is no competition with alternative grammatical options that preclude the insertion of definite pivots. These facts can be treated by allowing pragmatic inference to choose certain grammatical options and discard others for economy reasons. A different problem is posed by the (partial) conflation of existential and locative syntactic patterns in languages like Italian (cf. Zamparelli 2000, Remberger 2009, Bentley 2013, Cruschina this volume). The fact 17 Of course, I do not intend to claim that economy principles are always irrelevant for syntax.

Definiteness Effects

107

that esserci sentences with definite pivots can be taken as proper existentials or as locatives has recently been considered as one of the reasons why the DE in Italian looks like a “soft” constraint that can only be detected under very restrictive conditions: if the same syntactic form is used to convey both kinds of interpretation, and locative predication does not exhibit any restriction against definite DPs, then it becomes difficult to ascertain under what circumstances the DE still holds. The central problem concerns the status of sentences like (55): (55)

C’ è tua sorella. LOC be.PRS.3SG your sister ‘There’s your sister. / Your sister is there.’

On the one hand, treating (55) as a sort of ‘inverse locative’, following Zamparelli (2000) and Cruschina (this volume), straightforwardly explains the presence of the definite DP as an inverted, focussed subject, and the locative reading. On the other hand, true existentials in Italian, according to Bentley (2013: 700), accept ‘availability’ readings, as indicated in the gloss for (55), which means that the sentence is actually ambiguous. The locative interpretation does not block the ‘availability’ one. A discussion of this particular issue is beyond the limits of this paper. Despite its interest, I leave it for future research.

4.2 Solving Interpretive Mismatches The second relevant task of pragmatic inference in the domain of DEs –as in any other grammatical domain, actually- is the resolution of interpretive mismatches. In this case mismatches result from the use of a definite DP, and in particular an anaphoric definite DP, in a DE context. As already pointed out, I assume that violations of the definiteness constraint do not correspond to syntactic ill-formedness, but rather to semantic anomaly, which allows for resolution mechanisms to operate under certain contextual conditions and give rise to acceptable interpretations, in the same way that coercion and accommodation operations produce acceptable readings of otherwise anomalous strings. I will consider only existential sentences, although the phenomenon extends to unaccusative inversion –as shown for French in section 3.3- and other constructions. ‘Availability’ readings in ‘contextualized’ existentials represent the central case of mismatch resolution, and the key factor is narrow focus on the pivot DP, as advanced in 3.3. The previous discussion has shown that ‘availability’ readings are a systematic phenomenon, and not just an

Chapter Two

108

isolated, language-specific fact. If they cannot be found in languages like Spanish, it is due, as pointed out in section 4.1, to the competition of some other construction that takes their discourse function (cf. Spanish *Hay tus padres vs Están tus padres ‘There are your parents’). Let’s take again example (55) in Italian as a typical case of existential sentence with an ‘availability’ reading –if it is used as an answer to something like Who can we leave the children with?, though it might have other possible readings as well, such as a presentative one ‘Your sister is {here / there}’. (55)

C’ è LOC be.PRS.3SG ‘There’s your sister.’

tua sorella. your sister

The definite DP tua sorella occupies the rightmost position in the sentence and is assigned narrow focus. What are the consequences of focus? As is well known, focus on a constituent introduces a set of contextual alternatives that are relevant for the interpretation of such constituent. In (55), the hearer’s sister must be taken as a member of the set of possible candidates to satisfy a condition in the context –for instance, possible candidates to take care of the hearer’s children. The set of alternatives must be already given in the context; otherwise, it has to be pragmatically accommodated, which is costlier. Once this is taken into account, it is not difficult to realize (a) why this particular use of existential sentences typically needs a special context, and (b) why it is precisely ‘availability’ readings that emerge. As for the second issue, the relevance of lists and enumerations –just to mention the notions traditionally invoked in the literature- is due to the emergence of contextual alternatives: lists are nothing more than sets of alternatives for previously established purposes, and ‘availability’ means that one alternative can be chosen among others. Recall that the crucial condition is not the existence of a list, but rather the possibility to infer a relation between an entity and a set of entities. The family of readings triggered is simply the kind of readings that can justify a narrow focus on a definite pivot. With respect to the need for a special context, it seems to be another effect of narrow focus and contrast. In order to get a felicitous use of the definite pivot, some clue must be provided about the reason why the set of alternatives is relevant for interpretation, and about the way it fits in the conversational exchange. Thus, the set of alternatives must be somehow salient in the context. This explains why ‘availability’ readings are so strongly context-dependent and cannot appear in out-of-the-blue uses of existential sentences. So far, the main properties of this kind of readings can be traced back to narrow

Definiteness Effects

109

focus. But one question still needs to be addressed: how is the mismatch between the given status of the definite pivot and the Novelty Condition solved? Narrow focus plays a central role here too. Though it is obviously true that the referent of an anaphoric definite DP is given, the relation between the focused pivot and the context is new information: in (55) the fact that the hearer’s sister could take care of the children is presented as hearer-new, in accordance with the Novelty Condition. Focus on the pivot forces the interpreter to infer an adequate background where the referent can be considered as a candidate to fulfill some role. In this way, the interpretation is pragmatically adjusted to comply with the condition imposed by the construction, thus avoiding an interpretive mismatch: the referent of the pivot is given, but its relation with the local context is (presented as) novel. Pragmatic inference, as in many other similar cases, provides the material that is needed to solve the mismatch. In this sense, ‘availability’ readings, far from representing a different kind of construction, are a particular case of a more general pragmatic phenomenon affecting regular existential sentences (cf. McNally 1992/1997: 194).

5. Conclusions / Speculations As I mentioned in the introduction, the goal of this paper is not to offer a full account of DEs, but rather to discuss the role of IS notions and of pragmatic inference in the occurrence of DEs. In order to set the basis of the discussion, in section 2 I presented a list of assumptions that I take as well established points. I reproduce them here, with some of their consequences: 1. The DE is not a syntactic constraint: it is a condition on interpretations, and should be explained in semantic / pragmatic terms. 2. It results from a mismatch between the definiteness of the pivot DP and some semantic condition which DE contexts impose, whatever it is (a Novelty condition, or a requirement for semantic incorporation, among other possibilities). A crucial consequence of this idea is that the DE will not show the typical robustness of syntactic constraints: being the result of an interpretive mismatch, it may be overridden if some “repair” mechanism is able to license the occurrence of a definite DP in a DE context, under the appropriate conditions. In fact, the counterexamples to the DE are widespread, as the literature on the topic has often pointed out.

110

Chapter Two

3. DEs obey certain syntactic conditions that are actually rooted in IS: they affect only DPs occurring in internal argument positions, thus typically focal positions, and in the most representative contexts – existential sentences and unaccusative inversion- such DPs must be part of wide focus in thetic, ‘all-focus’ sentences with a stage topic. DEs are never associated with topical positions. There must be a crucial connection between the semantic condition that DE contexts impose on their pivot DPs and the focal nature of those DPs –a connection that still has to be uncovered (one possibility is that IS paves the way for semantic incorporation of pivots). Once these assumptions are made explicit, it is possible to look at the role of IS in detail. This has been the aim of section 3. There are two ways in which IS is crucial for DEs: on the one hand, IS defines the basic grammatical environments where the DE can appear (cf. 3.1); on the other, it defines the conditions for violations of the DE to occur (cf. 3.2). My basic claim is that we need to resort to fine-grained distinctions among types of topics and types of foci to discover some correlation between IS and (in)definiteness. The distinction between wide and narrow focus turns out to be quite productive. As for the environments where DEs appear, it is well known that DEs are related to conditions on information focus. More precisely, anaphoric definite pivots are banned from existential contexts when they are inside wide focus, in particular when followed by the locative coda. Some facts from Romance languages illustrate this point. This is how IS sets the limits for the DE to arise. Intuitively, wide focus is relevant because it implies that some kind of integration process must take place between the verb and the rest of the constituents, and definiteness is a factor that may block integration into a single informational chunk. Conversely, the presence of some informational partition in the DE context may alter the acceptability of definite DPs: there is evidence from different languages supporting the idea that when anaphoric definite pivots are under narrow focus, typically in the rightmost position, the DE vanishes. The widely discussed phenomenon of ‘list’ / ‘availability’ readings is the most salient manifestation of this factor. And this is how IS sets the conditions for the DE to be overridden –in the case of anaphoric definite DPs, i.e. those definite DPs that cannot comply with the Novelty Condition. Needless to say, a full account of the interaction of IS and the semantics of DE environments is still to develop, but these observations on Focus structure may throw some light on the problem. The main question raised by the role of narrow focus is why and how it has the effects it has. Here is where pragmatics enters the picture. I do not

Definiteness Effects

111

consider IS as a pragmatic matter, as several linguists tend to do: it is rather a component of the grammatical system. Nevertheless, it interacts with pragmatic inference: pragmatic principles allow speakers and hearers to take IS marking as the input of the inferential system in the search for relevant interpretations. In section 4, I have suggested that pragmatic inference is responsible for two aspects of the distribution of DEs: one is the resolution of cases of competing candidates for the expression of a single meaning, and the other is the resolution of interpretative mismatches –more precisely, mismatches between a DE context and an anaphoric definite DP. In the first case, the inferential system selects one candidate as the optimal pairing of form and meaning, and discards its competitors on the basis of economy considerations: this is why anaphoric definite pivots are excluded in “differential pivot marking” systems like Spanish and Sardinian. In the second case, pragmatic inference contextualizes the interpretation of anaphoric definite pivots in DE contexts by elaborating the consequences of their occurring under narrow focus; the result is an ‘availability’ reading based on the salience of a set of contextual alternatives for the denotation of the definite DP. In a few words, ‘availability’ readings arise as the result of a ‘last-resort’ inferential mechanism that (under appropriate conditions) solves the incompatibility between the definite DP and its linguistic context. Such a survival strategy cannot work when the DP is under wide focus. Narrow focus and contrast are able to license the presence of overt DPs in environments that would otherwise exclude them (cf. Leonetti 2014a for an overview of the effects of contrast, and Sheehan 2006: 192 for a version of an original idea by Adriana Belletti according to which focus may assume the role of Case in rendering a DP visible). Pragmatics, thus, has a noticeable role in the licensing of definite pivots in existential contexts. It is quite obvious that this approach to DEs puts considerable emphasis on IS, on the one hand (along the lines of other recent proposals like Bentley 2013), and on the semantics – pragmatics interface, on the other hand. As a consequence, a large part of the sentences which were considered plainly ungrammatical, as violations of the DE, in the previous literature are analyzed as well formed strings that contain a semantic mismatch; such mismatches may in some cases be repaired by inferentially adjusting the interpretation. Another consequence of emphasizing the role of IS and the semantics – pragmatics interface, instead of placing the burden of explanation on syntax, is that of reinforcing the connections between DEs and so-called ‘specificity effects’, i.e. the correlations between certain grammatical contexts and the specific interpretation of the indefinite DPs appearing in them (treating DEs as special cases of

112

Chapter Two

specificity effects is by no means a novel proposal: cf. Enç 1991 and Kiss 1995, a. o. for elaborations of the idea). If we limit our attention to DPexternal factors, the expected connection is essentially the following one: DE contexts –contexts where definite (and specific) DPs tend to be excluded- are the mirror image of contexts triggering specificity effects – contexts where specific readings of indefinite DPs are forced. A case in point is provided by Differential Object Marking (DOM): marked objects in DOM languages are typically associated with definite / specific readings (cf. the brief comment in section 2.2). Another well-known case is represented by preverbal subjects in certain languages, especially with stative predicates: definite / specific subjects are usually preferred. The correlation between DEs and specificity effects is strengthened, in my view, by two important facts. The first one is that, in most cases of specificity effects, specificity is not encoded by the construction, but pragmatically inferred to adjust the interpretation of indefinite DPs to the requirements of the linguistic context; if specificity is thus a class of epiphenomena, it is triggered by conditions / features that are themselves unrelated to specificity (for instance, topicality, or discourse prominence). The second fact is that, at least in a significant number of cases, the basic conditions / features that trigger specificity effects are related to definiteness (in clitic doubling) and/or IS (in DOM, cf. Dalrymple and Nikolaeva 2011, and in object scrambling and internal topics in certain languages, typically in preverbal subjects18 in languages like Tagalog, cf. Sabbagh 2009: 706). Notice that specificity effects are typically forced in topical positions: marked objects and preverbal subjects –at least in SVO languages- are topical. This confirms that these contexts are the mirror image of DE contexts, and the underlying conditions that are responsible for the effects pertain to IS: a certain kind of focus domain sets the conditions for DEs, and a certain kind of (aboutness) topic positions sets the conditions for specificity. The conclusion is that IS is behind both families of interpretive effects. A unified account of all of them should then be possible, with an adequate level of abstraction. The central idea is that constraints on definiteness and specificity result from the necessity to make DP interpretation compatible with the requirements of the grammatical contexts. Determining the nature of such semantic 18

In Romance languages, specificity effects have been noticed in preverbal subjects, in particular with unaccusative predicates in contexts where the two grammatical options –preverbal and postverbal position- are in competition. The preverbal position is a marked option for unaccusatives, and thus deserves a marked reading that justifies it and triggers discourse-related effects. Here again competition between available options determines the preferred reading (cf. section 4.1).

Definiteness Effects

113

requirements still represents a thorny and intriguing problem. However, it seems clear that they are somehow related to IS, and that pragmatic inference is crucial to understanding how speakers adjust DP interpretation to comply with them.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of Pragmatics 19.1: 39-55. —. 1997. Definiteness and existentials. In Language 73.1: 103-108. Beaver, David, Itamar Francez and Dmitry Levinson. 2005. Bad subject: (Non-)canonicality and NP distribution in existentials. In Proceedings of Semantics and Linguistic Theory 15: 19-43. Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry 19.1: 1-34. Bende-Farkas, Agnes and Hans Kamp. 2001. Indefinites and Binding: from specificity to incorporation. Notes from a course given at the 13th European Summer School in Logic, Language and Information, Helsinski (Finland). Benincà, Paola, Giampaolo Salvi and Lorenza Frison. 1988. L’ ordine degli elementi nella frase e le costruzioni marcate. In Lorenzo Renzi (ed.). Grande grammatica di consultazione, vol. I, 115-225. Bologna: Il Mulino. Bentley, Delia. 2004. Definiteness effects: evidence from Sardinian. In Transactions from the Philological Society 102.1: 57-101. —. 2013. Subject markedness and definiteness effects in Romance theresentences. In Language 89.4: 675-712. Bentley, Delia, Francesco Ciconte and Silvio Cruschina. 2015. Existential Constructions: a Comparative Investigation into Romance Dialects of Italy. Oxford: Oxford University Press. Birner, Betty and Gregory Ward. 1995. Definiteness and the English existential. In Language 71: 722-742. Blutner, Reinhard, Petra Hendriks and Helen de Hoop. 2006. Optimal Communication. CSLI: Stanford. Borer, Hagit. 2005. The Normal Course of Events. Structuring Sense, vol. II. Oxford: Oxford University Press. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. Chung, Sandra and William Ladusaw. 2004. Restriction and Saturation. Cambridge (Mass.): MIT Press. Clark, Billy. 2013. Relevance Theory. Cambridge: Cambridge University Press.

114

Chapter Two

Cohen, Ariel and Nomi Erteschik-Shir 2002. Topic, Focus, and the interpretation of bare plurals. In Natural Language Semantics 10.2: 125-165. Cruschina, Silvio. this volume. Pseudo-existentials and Definiteness Effects in Italian. Dalrymple, Mary and Irina Nikolaeva. 2011. Objects and Information Structure. Cambridge: CUP. De Cesare, Anna Maria. 2006. Sul cosiddetto c’è presentativo. Forme e funzioni. In Anna Maria De Cesare and Angela Ferrari (eds.). Lessico, grammatica, testualità. Acta Romanica Basiliensia 18, University of Basel, 127-153. Diesing, Molly. 1992. Indefinites. Cambridge (Mass.): MIT Press. Dobrovie-Sorin, Carmen and Claire Beyssade. 2004/2012. Redefining Indefinites. Dordrecht: Springer. (English version of Définir les indéfinis, CNRS, 2004) Enç, Mürvet. 1991. The semantics of specificity. In Linguistic Inquiry 22.2: 1-25. Endriss, Cornelia. 2009. Quantificational Topics. A Scopal Treatment of exceptional Wide Scope Phenomena. Dordrecht: Springer. Erteschik-Shir, Nomi. 1997. The Dynamics of Focus Structure. Cambridge: Cambridge University Press. Erteschik-Shir, Nomi. 2013. Information structure and (in)definiteness. In Patricia Cabredo-Hofherr and Anne Zribi-Hertz (eds.). Crosslinguistic Studies on Noun Phrase Structure and Reference, Syntax and Semantics volume 39, 23-50. Leiden: Brill. Etxeberria, Urtzi. 2012. Quantification in Basque. In Edward Keenan and Dennis Paperno (eds.). Handbook of Quantifiers in Natural Languages, 83-164. Dordrecht: Springer. Fischer, Susann. 2013. Unaccusatives, existentials and the Definiteness Restriction: explorations of the syntax- semantics interface. In Sofiana Chiriacescu (ed.). Proceedings of the VI NEREUS International Workshop “Theoretical implications at the syntax / semantics interface in Romance”. 33-53, Arbeitspapier Nr. 127. Fachbereich Sprachwissenschaft Universität Konstanz. Francez, Itamar. 2007. Existential Propositions. PhD dissertation, Stanford University. Geist, Ljudmila. 2012. Indefiniteness and Specificity: Lexical Marking and Information-structural Conditions. Habilitationsschrift, University of Stuttgart. Gundel, Jeannette. 1974. The Role of Topic and Comment in Linguistic Theory. PhD dissertation, University of Texas at Austin.

Definiteness Effects

115

Gundel, Jeannette and Thorstein Fretheim. 2004. Topic and Focus. In Laurence Horn and Gregory Ward (eds.). The Handbook of Pragmatics, 175-196. Oxford: Blackwell. Hartmann, Jutta. 2006. Well, there’s the list reading. In Jakub Dotlacil and Berit Gehrke (eds.). In UiL OTS Working Papers in Linguistics. Utrecht Institute of Linguistics. 1-15. —. 2008. Expletives in Existentials. English there and German da. LOT Dissertation Series 181. Utrecht: LOT. Hawkins, John. 1978. Definiteness and Indefiniteness: a Study in Reference and Grammaticality Prediction. London: Croom Helm. Hu, Jianhua and Haihua Pan. 2007. Focus and the basic function of Chinese existential you-sentences. In Ileana Comorovski and Klaus von Heusinger (eds.). Existence. Semantics and Syntax, 133-145. Dordrecht: Springer. Kiss, Katalin E. 1995. The Definiteness Effect revisited. In István Kenesei (ed.). Approaches to Hungarian V, 63-88. Szeged: JATE. Lahousse, Karen. 2011. Quand passent les cigognes. Paris: Presses Universitaires de Vincennes. Lambrecht, Knud. 2002. Topic, focus and secondary predication. The French presentational relative construction. In C. Beyssade et al. (eds.). Romance Languages and Linguistic Theory 2000, 171-212. Amsterdam: John Benjamins. Leonetti, Manuel. 1999. El artículo. In Ignacio Bosque and Violeta Demonte (eds.). Gramática descriptiva de la lengua española, vol. I, 787-890. Madrid: Espasa. —. 2008. Definiteness effects and the role of the coda in existential constructions. In Alex Klinge and Henrik Høek Muller (eds.). Essays on Nominal Determination. From Morphology to Discourse Management, 131-162. Amsterdam: John Benjamins. —. 2013. Information structure and the distribution of Spanish bare plurals. In Johannes Kabatek and Albert Wall (eds.). New Perspectives on Bare Noun Phrases in Romance and beyond, 121-155, Amsterdam: John Benjamins. —. 2014a. On contrastive readings in the interpretation of NPs/DPs. In Sofiana Chiriacescu (ed.). In Proceedings of the VI NEREUS Workshop ‘Theoretical implications at the Syntax / Semantics interface in Romance’. 99-116. Arbeitspapier 127. Fachbereich Sprachwissenschaft, Universität Konstanz. —. 2014b. Spanish VSX. In Karen Lahousse and Stefania Marzo (eds.). Romance Languages and Linguistic Theory 2012. Amsterdam: John Benjamins.

116

Chapter Two

—. Forthcoming. Pronominal pivots in existential sentences. López, Luis. 2012. Indefinite Objects. Cambridge (Mass.): MIT Press. Lumsden, Michael. 1988. Existential Sentences. Their Structure and Meaning. London: Routledge. Lyons, Chris. 1999. Definiteness. Cambridge: Cambridge University Press. Maleczki, Márta. 2010. On the definiteness effect in existential sentences: data and theory. In Enikó Németh and Károly Bibok (eds.). The Role of Data at the Semantics-Pragmatics Interface, 25-56, Berlin: Mouton De Gruyter. McNally, Louise. 1992/1997. A Semantics for the English Existential Construction. New York: Garland. —. 2011. Existential sentences. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.). Semantics. An International Handbook of Natural Language Meaning, vol. 2. 1829-1848, Berlin: Mouton de Gruyter. Milsark, Gary. 1977. Towards an explanation of certain peculiarities of the existential construction in English. In Linguistic Analysis 3.1: 1-29. Rando, Emily and Donna Jo Napoli. 1978. Definites in there-sentences. In Language 54: 300-313. Reinhart, Tanya. 1981. Pragmatics and linguistics: an analysis of sentence topics. In Philosophica 27: 53-94. Remberger, Eva-Maria. 2009. Null subjects, expletives and locatives in Sardinian. In Georg Kaiser and Eva-Maria Remberger (eds.). In Proceedings of the workshop ‘Null subjects, expletives and locatives in Romance’, 231-261. Konstanz: Fachbereich Sprachwissenschaft der Universität Konstanz. Rigau, Gemma. 1994. Catalan presentational sentences and the properties of Agr nodes. In Guglielmo Cinque et al. (eds.). Paths towards Universal Grammar. Washington DC: Georgetown University Press, 343-359. Rigau, Gemma. 1997. Locative sentences and related constructions in Catalan: ésser / haver alternation. In Amaya Mendikoetxea & Miriam Uribe-Etxebarría (eds.). Theoretical Issues at the Morphology – Syntax Interface, 395-421. Bilbao: Universidad del País Vasco. Rizzi, Luigi. 2005. On some properties of subjects and topics. In L. Brugè et al. (eds.). In Proceedings of the XXX Incontro di Grammatica Generativa, 63-82. Venice: Ca’Foscarina. Sabbagh, Joseph. 2009. Existential sentences in Tagalog. In Natural Language and Linguistic Theory 27: 675-719.

Definiteness Effects

117

Safir, Ken. 1982. Syntactic Chains and the Definiteness Effect. PhD dissertation, MIT. Sasse, Hans-Jürgen. 1987. The thetic / categorical distinction revisited. In Linguistics 25: 511-580. Sheehan, Michelle. 2006. The EPP and Null Subjects in Romance. PhD Dissertation, University of Cambridge. Sperber, Dan and Deirdre Wilson. 1986 / 1995. Relevance. Oxford: Blackwell. Suñer, Margarita. 1982. The Syntax and Semantics of Spanish Presentational Sentence-Types. Washington, DC: Georgetown University Press. Szabolcsi, Anna. 1984. From the Definiteness Effect to lexical integrity. In Werner Abraham and Sjaak de Meij (eds.). Topic, Focus and Configurationality , 321-348. Amsterdam: John Benjamins. Vallduví, Enric. 2002. L’ oració com a unitat informativa. In J. Solá et al. (eds.). Gramàtica del català contemporani, vol. 2, 1221-1279. Barcelona: Empùries. Vangsnes, Øystein. 2002. Icelandic expletive constructions and the distribution of subject types. In Peter Svenonius (ed.). Subjects, Expletives, and the EPP, 43-70. Oxford: Oxford University Press. Villalba, Xavier. 2013. Eventive existentials in Catalan and the topic – focus articulation. In Italian Journal of Linguistics 25.1: 147-173. —. This volume. Definiteness Effect, pronouns, and information structure in Catalan existentials. Ward, Gregory and Ellen Prince. 1991. On the topicalization of indefinite NPs. Journal of Pragmatics 16. 167-177. Zamparelli, Roberto. 2000. Layers in the Determiner Phrase. New York: Garland. Zimmermann, Michael. This volume. Definiteness effects in French unaccusatives. Ziv, Yael. 1982. Another look at definites in existentials. In Journal of Linguistics 18.1: 73-88. Zucchi, Alessandro. 1995. The ingredients of definiteness and the definiteness effect. Natural Language Semantics 3: 33-78.

II TYPOLOGICAL PERSPECTIVE

CHAPTER THREE PSEUDO-EXISTENTIALS AND DEFINITENESS EFFECTS IN ITALIAN* SILVIO CRUSCHINA

1. Introduction Even though they share a number of similarities cross-linguistically, existential sentences vary noticeably from language to language in both structural and semantic properties. An interesting and well-known point of variation concerns definiteness: whereas languages like English and Spanish are subject to a Definiteness Effect, which limits the ability of definite and quantificational pivots to appear in the construction (cf. (1) and (2)), languages such as Italian and Catalan do not seem to be sensitive to such a restriction (cf. 3) (examples from Leonetti 2008: 132-134; see also McNally 1992, 2011, Moro 1997): (1) a. There are {some / two / many / few / no / Ø} dogs. (English) b. *There is {it / the dog / that dog / Fido}. (2) a.

Hay {algunos / dos / muchos / Ø } perros. (Spanish) has-PF {some / two / many / Ø } dogs b. *Hay {el / el perro / ese perro / Fido}. has-PF {it / the dog / that dog / Fido}

*

This chapter is an abridged and updated version of a previous working paper appeared as Cruschina (2012a). Part of the data and ideas presented in this chapter are based on research carried out within the project Existential constructions: An investigation into the Italo-Romance dialects, at the University of Manchester from 2010 to 2013 (Arts & Humanities Research Council grant AH/H032509/1). I wish to thank the other project members, Delia Bentley and Francesco Maria Ciconte, for helpful comments and fruitful discussion, and Lucia Molinu for the Sardinian data from Buddusò offered to our project. I also thank two anonymous reviewers and to the editors of this volume for useful suggestions and remarks.

Pseudo-Existentials and Definiteness Effects in Italian

(3) a. b.

C’è un cane / C’è il PF-is a dog / PF-is the Hi ha un gos / Hi ha el PF has a dog / PF has the

121

cane / C’è Gianni. (Italian) dog / PF-is John gos / Hi ha en Joan. (Catalan) dog / PF has the John

The primary aim of this chapter is to demonstrate, on the basis of both semantic and information-structure considerations, that Italian existential sentences with a definite DP (cf., e.g. (3a)) are not existentials proper, but rather locative sentences. This amount to saying that, contrary to traditional assumptions, in Italian there are no true exceptions to the Definitness Effect. I will thus draw up a classification of existential sentences  or ci-sentences (there-sentences), more appropriately  in Italian, concluding that ci-sentences that apparently violate the Definiteness Effect are in fact pseudo-existentials. A distinction will first be made between existential sentences proper (type I) and locative sentences that I labelled ‘inverse locatives’ (type II) (cf. §2 and §3). Inverse locatives are the result of information-structure syntactic operations that are typical of Italian, but that are independtly unavailable in languages like English, i.e. postverbal focalization of the subject (the existential DP) and clitic resumption of a dislocated or implicit topic phrase (the locative constituent): these operations transform a locative predication into a structure which is morphosyntactically very similar, if not entirely identical, to existentials, contributing to the confusion between the two types. Section 4 will deal with some problems related to the suggested analysis of inverse locatives, in particular to the information and syntactic status of the locative phrase, and to the nature of the pronominal element ci in this sentence type. A third ci-sentence type will be identified in section 5, with reference to those locative sentences that are characterized by a particular deictic interpretation (type III). In this section, the exact role of the clitic ci in existential sentences will be compared to locative and deictic ci. Finally, a further type of Italian cisentences will be discussed in section 6, namely, presentational sentences, whose complexity and hardly studied properties can create problems of ambiguity with other sentence types. The classification proposed here sheds new light on the apparent differences between Italian and other languages, such as English, with regard to definiteness effects. Only type I ci-sentences are genuine existentials, while the other types, despite the superficial similarities, correspond to different constructions. The absence of definiteness effects in Italian and the relative differences with respect to English are therefore explained in terms of pseudo-existential constructions which are

Chapter Three

122

independently available in Italian but not in English. At first sight these pseudo-existential constructions resemble existential sentences proper, but a contrastive analysis of the pragmatic, semantic, and syntactic properties of the elements occurring in these sentences reveals that they should not be misteken for genuine existentials.

2. Definiteness Effects and Types of ci-sentences in Italian Semantically, an existential sentence asserts the existence or the presence of some entity (cf. McNally 2011). However, the pragmatic function of existential sentences is normally not to assert the existence of some entity but “to introduce the NP referent into the discourse world of the interlocutors by asserting its presence in a given location” (Lambrecht 1994: 179). This function has an evident connection to focalization and to the speaker’s intention to direct the hearer’s attention to a specific constituent within the utterance. The focal nature of the DP included in the existential sentence is widely acknowledged in the literature, and is generally considered to be a necessary feature for the well-formedness of the existential sentence, especially within the pragmatic accounts of the definiteness effect (cf. McNally 1992, Abbott 1993, 1997, Ward & Birner 1995, Zucchi 1995). The DP must be hearer-new and is hence predominantly, although not exclusively, indefinite. In Italian different conditions on topicalization will require dislocation of the locative phrase more often than in English, whenever the locative expression is presupposed or somehow related to the previous discourse (cf. Bentley, Ciconte & Cruschina 2015: Ch. 2), but the general properties of existential sentences with indefinite (or hear-new) DPs are the same in the two languages. If no significant differences are detected between English and Italian in regard to existentials with indefinite DPs (cf. (4) and (5)), no direct comparison is possible when definite DPs are considered, simply because they are not acceptable in English ((cf. (6) and (7)):1 (4) a.

1

C’è ci-is

un a

cane in giardino. dog in garden

I am here abstracting away from the well-known exceptions discussed in the literature on English existentials, including definite DPs with a list or a purpose interpretation (cf. Milsark 1974, Rando & Napoli 1978, Abbott 1993, 1997, Ward & Birner 1995). I believe that these exceptions also hold in Italian (see Bentley, Ciconte & Cruschina 2015), showing that definite DPs are not indiscriminately excluded from genuine existential sentences.

Pseudo-Existentials and Definiteness Effects in Italian

b.

Ci sono due bambini in cucina. ci are two children in kitchen

(5)

a. b.

There’s a dog in the garden. There are two children in the kitchen.

(6)

a.

C’è Gianni in giardino. ci-is John in garden C’era tua sorella in cucina. ci-was your sister in kitchen

b.

(7)

a. b.

*There’s John in the garden. *There was your sister in the kitchen.

(8)

a. b.

John is in the garden. Your sister is in the kitchen.

123

The locative predications in (8) are needed in order to translate the Italian sentences in (6) into English. Before reaching any final conclusion with regard to the differences between (6) and (7), let us take a close look at the alleged existential sentences with a definite DP in Italian, starting from some observations on their meaning and information structure. First of all, it has been pointed out that Italian existential sentences with a definite DP have a strong locative flavour (cf. Moro 1997, Zucchi 1995, Zamparelli 2000). Consequently, definite descriptions of entities that cannot normally have a physical realization and cannot be placed in a concrete location (cf. 9) are marginal: (9)

a. ?? Ci sono le due soluzioni di questa equazione. ci are the two solutions of this equation b. ?? Ci sono i problemi complessi che conosci nella ci are the problems complex that know.2SG in-the mia teoria. (Zamparelli 2000: 69) my theory

(10)

a. b.

Non not non not

ci sono molti ci are many c’è Gianni ci-is John

numeri primi numbers prime in giardino. in garden

in questo insieme. in this set (Moro 1997: 138)

Chapter Three

124

Moro (1997: 138) argues that while (10a) is a genuine existential sentence, (10b) is “truly locative” because it contains a locative predicate (i.e. in giardino) that is (negatively) predicated of the subject of the proposition (i.e. Gianni). This distinction is confirmed by the comparison with the corresponding locative predications:2 (11)

a. *Molti numeri primi non sono in questo insieme. many numbers prime not are in this set [*‘There are not many prime numbers in this set’] b. Gianni non è in giardino. John not is in garden ‘John is not in the garden.’

The existential sentence (10a) does not have a locative predication counterpart (cf. 11a), while the propositions expressed by (10b) and (11b) are semantically equivalent. The distinction becomes sharper when sentences involving abstract entities are contrasted – as in (10a)/(11a) –, but it also emerges in other contexts: (12)

a.

b.

(13)

a. b.

C’è un gatto in giardino. ci-is a cat in garden ‘There’s a cat in the garden.’ ?Un gatto è in giardino. a cat is in garden ?‘A cat is in the garden.’ C’è Gianni in giardino. ci-is John in garden Gianni è in giardino. John is in garden ‘John is in the garden.’

The (b) sentences in (12) and (13) are ordinary subject-predicate sentences, with a subject DP in a canonical subject position. An antidefiniteness, or indefiniteness, effect makes indefinite DPs bad topics (cf. Beaver, Francez & Levinson 2005, Bentley 2010, 2013), and explains why sentences like (12b) are somewhat marginal. The indefinite DP in (12a), instead, is a non-canonical subject in a non-standard structure. For these 2

It has independently been shown that not all existential sentences have a locative predication counterpart (Francez 2007). This fact goes against those theories that try to reduce existential sentences to locative structures (cf. Freeze 1992, 2001).

Pseudo-Existentials and Definiteness Effects in Italian

125

reasons, it would be difficult to consider (12b) as fully equivalent to (12a). On the contrary, the two sentences in (13) have indeed different pragmatic contexts of use but an equivalent meaning and underlying structure. A context requiring a structure like that in (13a) would be the question Who is in the garden?, to which (13a) would be a perfectly felicitous reply. As will be discussed below, the different position of the subject reflects specific information-structure conditions. These considerations lead us to hypothesize that of the sentences in (12) and in (13), only (12a) is an existential proper, whereas the others are locative, including (13a) and their English equivalents. Let us now move on to the information structure of the sentences under examination. Locative copular predications are standard predicate-focus structures in which the grammatical subject is also the topic or subject of the predication. The subject is the topic of the locative predication (13b), but is (part of) the focus in the existential sentence (12a), as well as in the pseudo-existential locative sentence (13a). What about the locative phrase? Does it have the same status in (12a) and (13a)? I will argue that it does not. The locative coda of an existential sentence is generally part of the focus in a sentence-focus structure (Lambrecht 1994). However, it has been observed that in languages that are apparently insensitive to the definiteness restriction, the definite DP of the (pseudo-) existential sentence attracts narrow focus, so that if a locative coda is present, this is either right- or left-dislocated, as indicated by a comma in the following examples: (14) a. ?? C’è la statua di Michelangelo in Piazza della Signoria. b. C’è la statua di Michelangelo, in Piazza della Signoria. ci-is the statue by Michelangelo in Piazza della Signoria ‘Michelangelo’s statue is in Piazza della Signoria.’ (Leonetti 2008: 140) Leonetti (2008) shows that an existential sentence with a definite DP cannot be uttered with the neutral intonation typical of sentence-focus structures (cf. Rigau 1994, 1997 for similar data in Catalan). The intonation contour of this kind of sentence is in fact characterized by a primary pitch accent on the postverbal subject, signalling its status as the sole focus of the clause, namely, as an argument-focus bearing a [focus] feature. The locative coda is instead dislocated, as expressed by the following constraint:

126

(15)

Chapter Three

Coda Constraint (Leonetti 2008: 142) 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).

Summing up, both semantic and information-structure considerations lead us to the conclusion that Italian existential sentences with a definite DP are not existentials proper, but rather locative sentences. To distinguish them from genuine existential sentences, I call this type of sentence ‘inverse locatives’. Two major types of ci-sentences have therefore been identified in Italian: (i) existential sentences, which typically involve an indefinite DP, and (ii) inverse locatives, with a definite DP. Inverse locatives are semantically equivalent to the corresponding canonical locative predications, but display the reverse information structure: the DP is a topic in the canonical locative predication but a focus in the inverse locative. On the contrary, the locative PP is part of the focus in the canonical locative predication but a topical, dislocated constituent in the inverse locative.

3. Existential Sentences and Inverse Locatives The term ‘inverse locative’ is certainly reminiscent of Moro’s (1997) analysis of existentials as inverse copular sentences. The current analysis is undoubtedly inspired by Moro’s work, but it differs from it in a crucial way: Moro applies his theory to both types of ci-sentences (i.e. locatives and existentials) regardless of the definiteness of the DP. In contrast, I follow his theory only for the second type of ci-sentences, namely, for inverse locatives. Building on the assumption that information-structure differences reflect differences in the underlying syntactic properties, I would like to claim that the following structures correspond to the two types of ci-sentence identified:3 3 In line with many analyses of existential and copular sentences, I am adopting a small clause structure to characterize the relation between the elements of the predication (cf. e.g. Stowell 1978, Moro 1997), but the same concepts could as well be described with a Predicative Phrase structure (Bowers 1993; cf. Hazout 2004, Remberger 2009). The special properties of small clauses have raised a number of controversial issues in the literature. I adopt a small clause structure mainly for illustrative purposes and because of the straightforward comparison that allows me to draw between the two constructions under investigation. The main difference between these two structures lies in the order of the elements within the small clause. The proform is related to the subject in (16) but is generated in the

Pseudo-Existentials and Definiteness Effects in Italian

(16)

127

TYPE I: Existential Sentences (e.g. C’è un gatto in giardino) SC

DP ci (17)

DP un gatto

TYPE II: Inverse Locatives (e.g. C’è Gianni in giardino) SC

DP Gianni

DP ci

We will discuss the status of the locative coda at the end of this section and, in greater depth, in the next section. For the moment, let us concentrate on the implications of the different configurations in (16) and (17). In line with several studies, the structure in (16) analyses the DP of the existential sentence as a predicate nominal (cf. Williams 1984, 1994, Hazout 2004, Francez 2007). According to this analysis, English there is an expletive syntactic subject that originates in the subject position of the existential small clause. Italian ci is first merged in the same position within the small clause and then attaches to Infl due to its clitic status. In (17), it is the definite DP that is the subject of the small clause, while the locative pronoun (English there, Italian ci) is the predicate (cf. Moro 1997). This distinction accounts for the criticism that English there can never occupy a true predicate position, but can only occur as what seems to be a syntactic subject (cf. Williams 1994: 135, Hazout 2004: 397). It also explains the different relationships between ci-sentences and canonical locative predications. In the previous section, we saw that existential sentences do not have a direct locative predication counterpart. This follows from the fact that in

predicate position in (17), presumably together with a predicate locative phrase that is subsequently dislocated (cf. fn. 6). The DP functions as predicate in (16), while it serves as the subject of the locative construction in (17). As for the correlation between definiteness and sentence type, it must be noted that definite DPs are found not only in inverse locatives but also, and probably more commonly, in another type of ci-sentence, that is, in presentational sentences (cf. §6).

128

Chapter Three

existential sentences only the subject (i.e. the locative pronoun) can raise out of the small clause,4 and complies with the observation that the structure in (16) does not have an ‘inverse’ equivalent that could feature the raising of the predicate DP.5 The raising of the subject DP in (17) would give rise to a canonical locative predication with a locative PP in predicate position: (18)

a. b.

Gianni è in giardino. John is in the garden.

Italian inverse locatives (cf. 19a), instead, are derived by the raising of the pro-predicate ci to Infl. In addition, the subject DP moves to the clauseinternal FocP, in order to have its [focus] feature valued and checked, while the locative coda moves to a topic projection in the left periphery of vP (cf. Belletti 2004, Moro 2009): (19)

a. b.

C’è Gianni in giardino. *There’s John in the garden.

(20) [IP Infl ... [FOCP [TopP [vP è [SC [DP Gianni] [DP ci ]] [PP in giardino] ]]]]

In other words, as illustrated in (20), type II ci-sentences involve a focalization strategy that turns a locative predication into an argument-focus structure. In fact, the main property of inverse locatives is not the raising of the pro-predicate out of the small clause, which is independently motivated by its morphological status, but rather the movement of the subject to FocP. This movement targets a functional projection that is active in Italian (cf. Belletti 2004), but not in English; this would offer a syntactic explanation for the unavailability of inverse locatives in the latter language, as well as for the apparent violation of the definiteness effect in Italian. The locative phrase generally occurs at the end of the sentence as part of the focus domain. However, it is not difficult to imagine sentence-focus contexts in which the locative phrase occurs sentence-initially constituting 4

The function of ci in this sentence type will be further discussed in section 5. To a certain extent, it is possible to have a topicalized DP in English existential sentences, as in the following example (from McNally 2011: 1834): (i) They told us there was a solution, and indeed a solution, there was. 5

Pseudo-Existentials and Definiteness Effects in Italian

129

what the sentence is about (i.e. aboutness topic), or predicate-focus contexts in which it is reintroduced into the discourse as presupposed, after having already been mentioned in the previous discourse (i.e. referential topic) (cf. Cruschina 2012b, and Bentley, Ciconte & Cruschina 2015: Ch. 2). In genuine existential sentences, the left-dislocated locative constituent generally represents the aboutness topic of the clause in which it occurs (cf. 21a). What type of topic are the dislocated constituents appearing in inverse locatives? Consider the following two sentences containing a leftdislocated locative phrase and differing with respect to the definiteness of the DP: (21)

a.

b.

Nel sistema solare ci sono otto pianeti. in-the system solar ci are eight planets ‘In the solar system there are eight planets.’ Nel sistema solare, c’è Venere. in-the system solar ci-is Venus ‘Venus is in the solar system.’ [*‘In the solar system, there’s Venus’]

Unlike the existential sentence (21a), example (21b) would prove pragmatically infelicitous in a context requiring a new topic or in which that given topic is not already active in the discourse. For instance, only (21a), but not (21b), can be conceived of as an appropriate start of a science book chapter. This is due to the fact that (21b) is an inverse locative that involves a focalization strategy that is not compatible with an aboutness topic or, more precisely, with those contexts that license an aboutness topic. From a referential viewpoint, an aboutness topic can introduce a new topic, whereas referential topics simply restore into the discourse a referentially given or old-information constituent. The dislocated locative phrase of inverse locatives can therefore only be a referential topic. Sentence (21b) would in fact prove perfectly felicitous if the additive particle anche ‘also’ was inserted before the subject DP (Nel sistema solare, c’è anche Venere; cf. 23a), given that it would presuppose that the locative phrase has already been introduced in the discourse. More specifically, it would presuppose that the predication holds for at least one other individual alternative to the constituent in focus. Similarly, (21b) would be pragmatically felicitous within a reply to the question in (22a): (22) a. Quali sono i pianeti del sistema solare oltre alla Terra? ‘Which are the planets in the solar system other than the Earth?’

130

Chapter Three

b. Non li ricordo tutti. Sicuramente, nel sistema solare, c’è Venere. ‘I don’t remember them all. Surely, Venus is in the solar system.’ (23) a. (Nel sistema solare,) c’è anche Venere (, nel sistema solare). b. Sicuramente (, nel sistema solare) c’è Venere (, nel sistema solare). In these contexts, as is typical of referential topics, the locative phrase could be dislocated either to the left or to the right of the core sentence; alternatively, it could be omitted altogether, as illustrated in (23). On the contrary, aboutness topics cannot be omitted and can only appear at the beginning of the clause. These pragmatic observations support the proposed analysis of these sentences as inverse locatives.

4. The Locative Phrase and the Nature of the Proform ci Further evidence in favour of our classification comes from the relationship between the different discourse status of the locative phrase and the role of the proform in type I and II ci-sentences. The need for the two separate structures (16) and (17) is motivated by the conceptually distinct roles that the DP plays in existentials and in locatives: the DP is a predicate in existential sentences, but it is a subject in locative predications, including inverse locatives. The structures under discussion have important implications for the status of the locative phrase and the proform ci present in both sentence types. The question of whether the coda should be treated as a post-nominal predicative modifier (cf. Williams 1984) or as a separate constituent, both syntactically and semantically (cf. Keenan 1987, 2003) has long been controversial. In most recent analyses, the locative expression included in the existential construction is considered to be a VP-external adjunct (cf. McNally 1992, Zucchi 1995, Moro 1997, Leonetti 2008). The status of the locative phrase as a predicate in locative predications is instead undisputed. The crucial question that inverse locatives raise is the following: in what sense does the proform ci represent the predicate of the clause? The answer I propose to this question relies on the hypothesis that ci and locative coda are part of a clitic-dislocation structure. In other words, ci is a resumptive clitic.6 6

Under the big-DP analysis of clitic-dislocation, the locative PP should be analyzed as base-generated in the complement position of a complex DP with a clitic head (cf. Cecchetto 2000, Belletti 2005), rather than as a VP-external adjunct as defended in several studies (cf. Moro 1997, Leonetti 2008, among others). The base-generated position and the derivation of the locative PP is certainly an important issue; however, I prefer to leave this question open for the moment.

Pseudo-Existentials and Definiteness Effects in Italian

131

4.1 Dislocation and Clitic Resumption One important fact that supports the clitic-dislocation analysis is that, as we have already observed (cf. 15), in inverse locatives the coda is prosodically and syntactically dislocated. The recognized locative nature of our type II ci-sentence in Italian is thus accounted for by assuming that ci plays the role of a locative clitic that is co-indexed with the dislocated locative PP. This hypothesis relies on several pieces of empirical evidence. Firstly, ci can in fact replace the locative PP of a previous locative predication sentence, thus acting as a pro-predicate. In the set of sentences in (24), the locative expression introduced in the question (24a) must be represented in the possible answers next to the copula – being semantically empty, the copula alone is not a complete predicate. It could simply be repeated (24b) or, alternatively, it can be substituted with either the clitic ci or a locative adverb (24c), but not with both (24d): (24)

a. Gli asciugamani sono nel cassetto? the towels are in-the drawer ‘Are the towels in the drawer?’ b. Sì, sono *(nel cassetto). yes are in-the drawer ‘Yes, they are in the drawer.’ c. Sì, ci sono / sono là. yes ci are are there ‘Yes, they are there.’ d. *Sì, ci sono là / ci sono nel cassetto. yes ci are there ci are in-the drawer

Secondly, this analysis of the relationship between ci and the dislocated locative coda makes a strong prediction: when the coda is not dislocated but is part of the focus, the sentence should become awkward or entirely ungrammatical since it would constitute a case of clitic doubling that is independently ruled out in Italian. It is well-known that in Italian resumptive clitics resume topics, but they are incompatible with foci (cf. Rizzi 1997). An information-structure articulation that violates such constraint is thus expected to give rise to a noticeable degradation of the ci-sentences in question. This prediction is indeed born-out: a focal coda, either preverbal or postverbal, is incompatible with a definite DP in an

More crucial for the present analysis is the claim that the locative phrase targets a topic position within the left periphery of the vP.

132

Chapter Three

inverse locative.7 The locative coda constitutes the focus of the clause in the following cases: (i) when it corresponds to a wh-phrase, (ii) when it bears contrastive focus, and (iii) when it is the focus of an answer to a whquestion. Let us consider the first case: (25)

a. Dove (*ci) sei tu? where ci are you ‘Where are you?’ b. Dove (*ci) sono i fiori bianchi? where ci are the flowers white ‘Where are the white flowers?’ c. Dove pensi che (*ci) sia Gianni? where think.2SG that ci be.SUBJ.3SG John ‘Where do you think John is?’

(26)

a. In quale stanza (??c’) è tua sorella? in which room ci is your sister ‘Which room is your sister in?’ b. In quale cassetto (??ci) sono i piatti d’ argento? in which drawer ci are the plates of silver ‘Which drawer are the silver plates in?’

Care is needed with the judgments on these wh-questions, especially those in (26). An echo-question reading, as well as a D-linked interpretation of the locative wh-phrase should be disregarded.8 The grammaticality of the 7

Given the general restriction that imposes one focus per sentence in Italian, the definite DP cannot be a focus in this context, thus failing to satisfy another condition for the pragmatic felicity of inverse locatives. However, this condition appears to be secondary with respect to the coda constraint, in the sense that it can be violated when the definite DP is topical and the focus is on the predicate made up of the clitic ci and the copula: (i) a. [Gianni]TOP [c’è]FOC, [in giardino]TOP b. *[Gianni]TOP [c’è in giardino]FOC As shown in (ia), irrespective of the informational status of the subject, the locative PP must still be dislocated and cannot be part of the focus of the sentence together with the predicate (cf. also (24) above). Given the locative value of the clitic ci, sentences like (iib) must be ruled out because they would create a clitic doubling structure (cf. Moro 1997: 154). 8 For many speakers, the clitic ci becomes more, or even totally, acceptable with a D-linked reading of the wh-phrase. Although I am not able to provide a full explanation for this contrast, it is worth noting that D-linked wh-phrases have been described as bearing a [topic] feature: they are similar to topics not only with

Pseudo-Existentials and Definiteness Effects in Italian

133

sentences above can be rescued by omitting the clitic ci, which would basically yield a canonical locative predication. Crucially, the same sentences are not as degraded with an indefinite DP, namely, as in existential sentences (type I), as shown in (27): (27)

a. Dove pensi che ci siano più problemi? where think.2SG that ci be.SUBJ.3PL more problems ‘Where do you think there are more problems? b. Dove ci sono tanti fiori? where ci are many flowers ‘Where are there (so) many flowers?’ c. In quale giardino ci sono alberi di limone? in which garden ci are trees of lemon ‘In which garden are there lemon trees?’

4.2 Evidence from Sardinian The contrast between indefinite and definite DPs with respect to the presence of the proform ci in wh-questions is even more evident in Sardinian, where the type of structure is signalled by a change in the copula and in the agreement pattern. Copula alternation in Sardinian existential sentences has long attracted the attention of linguists. It is generally related to the definiteness of the DP, so that copula HAVE appears with indefinite DPs, while BE is used with definite DPs (cf. Jones 1993: 113, La Fauci & Loporcaro 1997, Bentley 2004, 2011, and Remberger 2009). It was observed by Jones (1993: 114) that the clitic bi – corresponding to Italian ci – has a tangible locative value when it occurs with the BE copula. This idea is further developed in Remberger (2009), in which a claim similar to the present proposal for Italian is put forward: Sardinian sentences with copula BE are locative structures. The ungrammaticality of (28b) can therefore be attributed to the creation of a clitic doubling configuration between the locative clitic and the interrogative locative phrase: (28)

a.

In ube b’at metas frores? (Sardinian) where cl-has many flowers ‘Where are there many flowers?’

respect to their interpretation but also with respect to extraction phenomena (cf. Rizzi 2001).

Chapter Three

134

b. *In ube bi sun sos prattos? where cl are the plates ‘Where are the plates?’

(Jones 1993: 114)

Further data from the Logudorese variety of Sardinian spoken in Buddusò illustrate this point: an interrogative focal locative constituent and the clitic bi are only compatible in existential sentences, which in Sardinian are characterized by the use of the copula HAVE, the presence of an indefinite DP, and the lack of agreement between them (29a). The same pronominal form is instead ruled out in those sentences that we have identified as inverse locatives, comprising a definite DP that agrees with the copula BE (29b, 30): (29)

a.

b.

(30)

a.

b.

c.

Inue b’at duos sindigos? (Buddusò, Sardinian) where cl-has two mayors ‘Where are there two mayors?’ Inue (*bi) son sos duos sindigos? where cl are the two mayors ‘Where are the two mayors?’ Inue (*bi) ses tue? (Buddusò, Sardinian) where cl are you ‘Where are you?’ Inue (*bi) son sos politicos onestos? where cl are the politicians honest ‘Where are the honest politicians?’ In cale calasciu (*bi) son sos piattos de pratta. in which drawer cl are the plates of silver ‘In which drawer are the silver plates?’

Moreover, the speakers consulted confirm the (prosodic) dislocation of the locative coda as a necessary condition for the naturalness of inverse locatives with copula be and a definite DP: (31)

B’est sorre tua, in uffisciu. (Buddusò, Sardinian) cl-is your sister in office ‘Your sister is in the office.’

In Italian, a similar behaviour is observed for the pronominal clitic in sentences with a contrastively focalized fronted coda (32) and in replies to wh-questions that require a focal locative phrase in the answer (33):

Pseudo-Existentials and Definiteness Effects in Italian

135

(32) a. IN GIARDINO (*c’) è il gatto bianco, non al balcone! in garden ci is the cat white not at-the balcony ‘There is a white cat in the garden, not on the balcony.’ b. IN GIARDINO c’è un gatto bianco, non al balcone! in garden ci-is a cat white not at-the balcony ‘The white cat is in the garden, not on the balcony.’ (33)

A:

Dov’è mia sorella? where-is my sister ‘Where is my sister?’ B: Tua sorella è in cucina. your sister is in kitchen ‘Your sister is in the kitchen.’ B': * C’è tua sorella in cucina. ci-is your sister in kitchen ‘Your sister is in the kitchen.’ [*‘There’s your sister in the kitchen.]

In this section, I provided arguments and evidence in support of the claim that, in inverse locatives, ci is a resumptive clitic referring to a dislocated or implicit locative phrase. This is clearly different from existential sentences where the locative coda need not be dislocated and where the proform ci does not refer to an identifiable location. The distinctive behaviour of existential sentences and inverse locatives with respect to the locative coda, and the nature of the clitic proform, raise two issues. First, one may wonder what happens in the absence of a locative coda, which is in fact optional both in existential sentences and in inverse locatives. Second, a strongly locative characterization of the clitic ci when it cooccurs with a definite DP (i.e. in inverse locatives) raises questions concerning the status of ci in existential sentences, where the same locative function surely cannot be attributed. These questions will be addressed in the next section.

5. Locative and Deictic ci vs. Pro-argument ci We have so far seen that when it occurs in an inverse locative structure, the pronoun ci functions as a locative pro-predicate referring to a locative phrase. The locative constituent can be explicitly present within the same sentence, in the form of a dislocated locative PP, or can be implicit in the context. If no locative is implicit or salient in the discourse, the clitic ci

Chapter Three

136

assumes a strong deictic value, that is, a default interpretation of ‘here and now’ (cf. 34):9 (34)

a.

b.

c.

C’è Gianni. ci-is John ‘John is here.’ Guarda: c’è tua sorella! look ci-is your sister ‘Look: your sister is here!’ C’è l’aereo, finalmente. ci-is the-airplane eventually ‘The airplane is here, eventually.’

Let us call these deictic clauses type III ci-sentences, to be distinguished from the other two types that we have previously identified: type I (i.e. existential sentences) and type II (i.e. inverse locatives). Locative-deictic ci designates the perceptual identification of a referent in the speaker’s proximal physical space. This interpretation requires that the DP be specific and referential. It is not implausible to assume that when it conveys this deictic meaning, the clitic ci still serves the function of a locative predicate, as in type II ci-sentences. The focus is obviously on the DP, which therefore has to move out of the small clause and target the clause-internal FocP. As we can see in the translations of the examples in (34), type III ci-sentences are rendered into English with the deictic locative adverb here, and a focal pitch accent on the DP subject is necessary to convey the equivalent meaning. This sentential category shares many properties with inverse locatives, to the extent that it is reasonable to consider it as a subcategory of type II. Let us turn to the question of the role of ci in existential sentences (type I). Following Williams (1984, 1994), Hazout (2004), and Francez (2007), I have assumed that the nominal DP is the main predicate of the existential sentence. What then constitutes the subject of the existential predication? Several scholars have hypothesized that the topic or subject of the predication is a location, which can be either explicit or implicit (cf. Partee & Borschev 2002, 2007, Leonetti 2008, Parry 2010). In many languages, the distinction between locative predications and existential sentences is only marked by a different word order, which may be interpreted as a signal of the distinctive Topic-Comment articulation: the 9

An alternative interpretation of (34a) is possible when this sentence is used to express availability for a specific purpose. This reading constitutes an exception to the definiteness effect (cf. Abbott 1993).

Pseudo-Existentials and Definiteness Effects in Italian

137

topic is the DP is in a locative predication but the locative phrase in the existential construction (cf. e.g. Freeze 1992, 2001; see also Partee & Borschev 2002, 2007). Francez (2007) argues that, independently of the presence of a locative phrase within the sentence, the argument of the property denoted by the existential DP is always an implicit contextual domain (something intuitively similar to a location). An overt locative coda contributes to the restriction of its identity.10 Irrespective of the precise semantic analysis, all the studies mentioned above endorse the view that a location, and not the referent denoted by the DP, is the subject of the predication in the existential construction. This hypothesis thus offers an explanation of the locative nature of the proform in Italian and in other Romance languages; as a matter of fact, the proform has often been attributed a subject function (cf. Burzio 1986, Tortora 1997). Accordingly, as illustrated in the structure (16) above, in Italian existential sentences the clitic ci is to be interpreted as a pro-argument, namely, a pronominal form that is linked to or stands for an abstract locative subject of predication. It is reasonable to argue that although it is morphologically locative both in existential sentences and in inverse locatives, the clitic ci is more grammaticalized in the former construction, having lost its anaphoric and referential connotation; this explains why an etymologically identical element has now two completely different functions in the two constructions under discussion.11 Even though in existential sentences ci does not replace a locative complement and, consequently, no longer functions as a real locative pronoun, it has not been completely stripped of its spatial reference, in that it expresses a contextual location or, perhaps more appropriately, it is bound to an abstract or null locative topic which provides the spatial parameters of the predication that asserts the presence and/or existence of a given entity. Summing up, in this section I propose that one more property has to be included among the differences between type I, on the one hand, and type II and III ci-sentences, on ther other, namely, the role of the proform ci. This pronominal clitic refers to a precise location in type II, and to the deictic sphere of the speaker in type III, but it functions as a pro-argument in type I, that is, as a pronominal form standing for an abstract or null locative subject of predication. 10

For a more detailed contrastive comparison between existentials and locatives, see Cruschina 2014, 2015, Bentley & Cruschina in press, Bentley, Ciconte & Cruschina 2015). 11 Recent diachronic findings on the emergence of the proform in Italo-Romance existential sentences may provide an account of its development (cf. Ciconte 2009, 2010, Parry 2010; see also Bentley, Ciconte & Cruschina 2015: Ch. 5).

138

Chapter Three

6. Presentational ci-sentences According to the analysis carried out so far, the definiteness of the DP correlates with the type of ci-sentences identified: indefinite DPs typically occur in existential sentences (type I), while definite DPs are generally involved in inverse locatives (type II). This correlation, however, is not rigid and fully systematic, in that it is possible for a definite DP to appear in an existential sentence under special interpretations such as a list or reminder reading. These exceptions have been widely described for English (cf. e.g. Milsark 1974, Rando & Napoli 1978, Abbott 1993, 1997, Ward & Birner 1995) and presumably hold across languages, also including Italian. Another problem that may contribute to the lack of clearcut boundaries between ci-sentence types is the existence of a further construction, typical of spoken Italian, which exhibits similar morphosyntactic properties: presentational ci-sentences, which we will call type IV ci-sentences. The DP occurring in this sentence type is not subject to a strict definiteness restriction, although definites and specific indefinites, as opposed to non-specific indefinites, are more commonly found:12 (35)

a. C’è mio fratello malato / che sta male. ci-is my brother ill who stays badly ‘My brother is ill. / My brother feels sick.’ b. C’è Maria che canta in un teatro di ci-is Mary who sings in a theatre of ‘Mary sings in a theatre in Berlin.’ c. C’è Gianni infuriato / nei guai. ci-is John furious in-the trouble ‘John is furious / in trouble.’

Berlino. Berlin

From a pragmatic perspective, the main property of presentational sentences is that they introduce a new proposition into the discourse. The whole clause is thus presented as sentence-focus (Lambrecht 1994, 2002). The DP referent is first introduced in a manner similar, if not entirely identical, to ci-sentence types I and II. This postverbal DP then functions as the topic of the predication expressed either by an adjectival predicate (35a, 35c) or, more frequently, by a pseudo-relative clause (35b). 12

This sentence type is called ‘eventive’ in Leonetti (2008) and in Villalba (2013) to indicate those presentational sentences that describe the occurrence of an event or something that happens at a given time and place (Leonetti 2008: 141). See also Marten (2013) for a similar construction in Swahili.

Pseudo-Existentials and Definiteness Effects in Italian

139

Semantically, the sentences in (35) can be paraphrased with the corresponding canonical Topic-Comment predications (e.g. Mio fratello è malato/sta male, Maria canta in un teatro di Berlino, etc.) without determining any alteration of their truth conditions (cf. the English translations). Presentational sentences have some features in common with both existential sentences and inverse locatives, giving rise to a certain degree of ambiguity. The basic function shared by presentational and existential sentences is that of introducing a new referent into the discourse. However, it has been argued that existential sentences proper presuppose a location (cf. Partee & Borschev 2002, 2007, Koontz-Garboden 2009) which, as discussed in the previous section, functions as the subject of predication. On the contrary, presentational sentences may lack any locative anchoring (cf. 35a, 35c), and the DP referent is generally introduced not to predicate a property of an argument but with the purpose of making it available as the subject of subsequent predication (Lambrecht 1994). One further difference is that the DP is generally indefinite in existential sentences – presumably because only this type of DPs can function as a predicate (cf. Higginbotham 1987) – but is typically definite in presentational sentences. The fact that the DP of a presentational sentence can be definite may create a certain ambiguity between type IV, on the one hand, and types II or III, on the other, whenever a locative phrase is implicit or salient in the context, or when ci has a spatial deictic function. The following example may better be analysed as an occurrence of type IV ci-sentence (from Berruto 1986: 71, n.2): (36)

C’è il Signor Ponza che chiede di essere ricevuto. ci-is the Mr Ponza who asks of be.INF received ‘Mr Ponza here asks to be received.’

However, in this sentence the clitic ci seems to have the default interpretation of ‘here and now’ typical of type III, as argued in Berruto (1986) and as evident from its most appropriate English translation. It could also be the case that a locative constituent is already present and active in the discourse, and that ci serves as a clause-internal link to that constituent, which is represented as right-dislocated in (37). In such a context the subject DP would be argument-focus, as is peculiar to sentences of type II:

140

(37)

Chapter Three

C’è il Signor Ponza, in salotto, che chiede ci-is the Mr Ponza in living-room who asks di essere ricevuto. of be.INF received ‘Mr Ponza is in the living room, asking to be received.’

In this example, the relative clause behaves as an ordinary appositive relative clause, rather than a pseudo-relative. Judgements on the possible semantic differences between these sentences are difficult. At first sight, similar pairs sound fully equivalent to most native speakers. However, a careful examination reveals that while (36) may be taken as an instance of presentational sentence, sentence (37) might be better interpreted as an inverse locative. From an information viewpoint, (36) is in fact most naturally uttered as a single unit, whereas in (37) a pause between the constituents is normally produced, which may well signal that the locative phrase is right-dislocated and that the relative is an appositive relative clause. Even more complex are the cases in which the predicate is an adjective or a non-locative PP. It is well known that existentials admit an adjectival coda, which is required to convey a stage-level predicate (see e.g. Milsark 1974, McNally 1992). In the following example, ambiguity emerges between an existential and a presentational reading: (38)

Ci sono molti ragazzi malati. ci are many guys will ‘There are a lot of guys ill. / A lot of guys are ill.’

Although it is not always easy to differentiate between the two structures, it must be noted that the most natural interpretation with specific DPs is the presentational one: (39)

Ci sono molti dei ragazzi di questa ci are many of-the guys of this ‘Lots of the guys in this classroom are ill.’

classe malati. class ill

Less problematic are similar cases involving a definite DP. Given that the predicate is non-locative, the pronoun ci cannot be analysed as a locative resumptive clitic, and the locative inverse interpretation is thus ruled out:

Pseudo-Existentials and Definiteness Effects in Italian

(40)

a.

C’è Gianni ci-is John ‘John is ill.’

b.

C’è tua sorella in pericolo. ci-is your sister in danger ‘Your sister is in danger.’

141

infuriato / malato. furious ill

Note that if the information-structure unity typical of presentational sentences is disrupted, for instance by forcing an argument-focus interpretation of the DP through focus fronting, the clitic ci must be left out: (41)

a.

b.

Anche Gianni (*c’) è infuriato / malato. also John ci is furious ill ‘John too is furious / ill .’ Soltanto tua sorella (*c’) è in pericolo. only your sister ci is in danger ‘Only your sister is in danger.’

In these examples, the presence of the focalizing adverbs anche ‘also/too’ and soltanto ‘only’ require narrow-focus on the associated DP, and allows for focus fronting under a contrastive or surprise interpretation (cf. Cruschina 2012b). Recall that this is the information structure that characterizes inverse locatives. In fact, if the predicate is a locative PP, the clitic ci is perfectly acceptable in the fronting structure: (42)

a. b.

C’è Gianni in giardino. ci-is Gianni in garden Anche Gianni c’è in giardino. also Gianni ci-is in garden ‘John too is in the garden.’

Presentational sentences share significant similarities with other constructions. In particular, the pseudo-relative in the sentence type under discussion resembles the pseudo-relative that occurs with perception verbs. For this construction, it has been independently argued that the pseudo-relative corresponds to a tensed CP that functions as the predicate of a small clause (cf. Guasti 1993, Cinque 1995, Casalicchio 2013; cf. also Belletti 2008 on cleft-sentences):

142

(43)

Chapter Three

a. Ho visto [SC Gianni arrabbiato / in difficolta ] have.1SG seen John angry in difficulty ‘I saw John angry / in trouble.’ b. Ho visto [SC Gianni [CP che correva verso casa ] have.1SG seen John who ran.3SG towards home ‘I saw John running home.’ (Cinque 1995: 250, 251)

It is reasonable to think that the pseudo-relative of presentational sentences has the same function as with perception verbs, acting as the predicate of the small clause in which the predication obtains. It must be observed that this analysis has an important implication for the status of the clitic ci: if the DP and the adjectival phrase are the argument and the predicate of the small clause, respectively, the proform ci cannot be regarded as an element generated within the same small clause. Hence, it cannot be attributed any role in the predication. If the syntactic analysis suggested here is on the right track, the only possibility left is that in this structure the pronoun ci does not undergo any syntactic derivation, but is lexicalized together with the copula BE as a marker of the presentational construction (see Russi 2008).13 From these considerations, it follows that the main pragmatic difference between presentational sentences and the other ci-sentence types is that in this construction ci does not point to a referent in the external world and is not connected to the contextual spatio-temporal parameters or settings of the predication. It rather expresses a discourseinternal function, marking a new proposition as relevant and pertinent to the discourse: a function that is indeed characteristic of the narrative and spoken language (cf. Berruto 1986; see also Berretta 1995, De Cesare 2007, Casalicchio 2013 and Bentley, Ciconte & Cruschina 2015: Ch. 2). In this respect, a major property distinguishes type IV from the other types, which may be viewed as reflecting the special grammaticalized status of ci in the lexicalized expression ci+copula: presentational sentences cannot be negated (cf. 45):14

13

Russi (2008) does not distinguish between ci-sentence types, and suggests that when it joins with the copula BE, the element ci is always a grammaticalized morpheme specialized in the marking of the existential/presentational construction. 14 Type II can only be negated if the sole predicate, but not the subject DP, is the focus of the sentence, falling within the scope of negation: (i) a. [Non c’è]FOCUS [Gianni], [in giardino] b. *[Non c’è Gianni]FOCUS [in giardino] c. ??/* Non c’è [Gianni]FOCUS [in giardino]

Pseudo-Existentials and Definiteness Effects in Italian

(44)

a.

b.

(45)

143

Non ci sono orsi bianchi al Polo Sud. (type I) not ci are bears white at-the Pole South ‘There are no polar bears in the South Pole.’ Non c’è Gianni in giardino. (type II) not ci-is John in garden ‘John is not in the garden.’

a. *Non c’è Gianni infuriato. b. *Non c’è Gianni nei guai.

(type IV) (type IV)

This difference can be ascribed to the fully grammaticalized status of ci in esserci ‘be there’ in presentational sentences, which is now an independent lexicalized verb. It is obviously possible to negate the predication that it introduces (e.g. Gianni non è infuriato ‘John is not furious’), but not the semantically empty locution made up of ci and the copula.

7. Conclusions Starting from specific observations on information structure, in this paper I have identified different types of Italian existential sentences. To avoid terminological confusion, I have assigned a distinctive label to each type, referring to all of them as ci-sentences because they all share the property of being introduced by the element ci attached to the copula. Several pieces of evidence in favour of the distinctions proposed have been reviewed. The properties of the four types of Italian ci-sentences are summed up in the following table: TYPE

STRUCTURE

FOCUS

I II III IV

existential inverse locative deictic locative presentational

sentence/predicate argument argument sentence

ci pro-argument pro-predicate pro-predicate lexicalized

DP predicate (indef.) argument (definite) argument (definite) argument (specific)

Table 1. Italian ci-sentences. This is consistent with the proposed analysis of this sentence as an inverse locative predication. While negative existential sentences must presuppose a location (44a), negative locative predications must presuppose the existence of the entity denoted by the DP (cf. Partee & Borschev 2007, and references therein), which is therefore dislocated as a referential topic in (44b) (cf. (i) above). Unless they are interpreted as mere instances of type II, with ci referring to a contextually implicit location, type III ci-sentences (e.g. C’è Gianni! ‘John is here’) cannot be negated.

144

Chapter Three

The extension of the focus varies from type to type, and only types II and III are clearly characterized by an argument-focus structure. The lack of distinctive syntactic marking for the other types supports the idea that they involve cases of broad focus including the whole sentence (type I and type IV) or the predicate (arguably, type I). These considerations have proved fundamental for the classification proposed, clarifying the generally acknowledged intuition that existential sentences serve primarily to introduce a new referent into the discourse, and shedding some light on the controversy as to whether they correspond to sentence- or predicate-focus structures. Moreover, types II and III are derived by two syntactic information-structure related operations that are not available in English, namely, postverbal focalization of the subject and, limited to type II, cliticdislocation of the locative coda. Type IV, instead, is an independent structure specific to spoken Italian and only partially related to the other categories: it is the result of the grammaticalization of the proform ci and its lexicalization together with the copula as an independent verb. A similar development has been described for other languages (e.g. in Catalan, cf. Leonetti 2008 and Villalba 2013, and – presumably – in Swahili, cf. Marten 2013), but is not found in all languages. This explains why type I, II and III are not available in English.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of Pragmatics 19.1: 39-55. —. 1997. Definiteness and existentials. In Language 73: 103-108. Beaver, David, Itamar Francez & Dmitry Levinson. 2005. Bad subject: (non-) canonicality and NP distribution in existentials. In Effi Georgala & Jonathan Howell (eds.). Proceedings of Semantic and Linguistic Theory XV, 19-43. Ithaca, N.Y.: Cornell University Press. Belletti, Adriana. 2004. Aspects of the low IP area. In Luigi Rizzi (ed.). The Structure of IP and CP. The Cartography of Syntactic Structures, Vol. 2, 16-51. Oxford/New York: Oxford University Press. —. 2005. Extended doubling and the VP periphery. In Probus 17.1: 1-35. —. 2008. The CP of clefts. In Rivista di Grammatica Generativa 33: 191204. Bentley, Delia. 2004. Definiteness effects: evidence from Sardinian. In Transactions of the Philological Society 102.1: 57-101. —. 2010. Principles of subject markedness in Romance. In Archivio Glottologico Italiano 95.2: 152-189.

Pseudo-Existentials and Definiteness Effects in Italian

145

—. 2011. Sui costrutti esistenziali sardi. Effetti di definitezza, deissi, evidenzialita. In Zeitschrift für Romanische Philologie 127.1: 111-140. —. 2013. Subject markedness and definiteness effects in Romance there sentences. In Language 89.4: 675-712. Bentley, Delia & Silvio Cruschina. in press. Existential constructions. In Susann Fisher & Christoph Gabriel (eds.). Manual of Grammatical Interfaces in Romance. Berlin: Mouton de Gruyter. Bentley, Delia, Francesco Maria Ciconte & Silvio Cruschina. 2015. Existentials and Locatives in Romance Dialects of Italy. Oxford: Oxford University Press. Berretta, Monica. 1995. Come inseriamo elementi nuovi nel discorso/1: ‘C’e il gatto che ha fame’. In Italiano e Oltre 10: 212-217. Berruto, Gaetano. 1986. Un tratto sintattico dell’italiano parlato: il c’e presentativo. In Lichem, Klaus, Edith Mara & Susanne Knaller (eds.). Parallela 2. Aspetti della sintassi dell’italiano contemporaneo, 61-73. Tubingen: Narr. Bowers, John. 1993. The syntax of predication. In Linguistic Inquiry 24.4: 591-656. Burzio, Luigi. 1986. Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Casalicchio, Jan. 2013. Pseudorelative, gerundi e infiniti nelle varietà romanze: somiglianze (solo) superficiali e corrispondenze strutturali. Ph.D. Thesis, University of Padua. Published by LINCOM, München. Cecchetto, Carlo. 2000. Doubling structures and reconstruction. In Probus 12: 93-126. Ciconte, Francesco Maria. 2009. Pro-forms in existential constructions of early Italo-Romance vernaculars. In Georg Kaiser & Eva-Maria Remberger (eds.). Null subjects, expletives and locatives in Romance, 183-198. Konstanz: Universität Konstanz, Konstanzer Arbeitspapiere des Fachbereichs Sprachwissenschaft (No. 123). Ciconte, Francesco Maria. 2010. Existential constructions of Early ItaloRomance Vernaculars. Ph.D. Thesis, University of Manchester. Cinque, Guglielmo. 1995. The pseudo-relative and ACC-ing constructions after verbs of perception. In Italian Syntax and Universal Grammar, 244-275. Cambridge: Cambridge University Press. Cruschina, Silvio. 2012a. Focus in existential sentences. In Valentina Bianchi & Cristiano Chesi (eds.). Enjoy Linguistics! Papers offered to Luigi Rizzi on the occasion of his 60th birthday, 77-107. Siena: CISCL Press. —. 2012b. Discourse-Related Features and Functional Projections. Oxford/New York: Oxford University Press.

146

Chapter Three

—. 2014. Existential and locative constructions in Italo-Romance. In L’Italia Dialettale 75: 55-80. —. 2015. Patterns of variation in existential constructions. In Isogloss 1: 33-65. De Cesare, Anna-Maria. 2007. Sul cosiddetto ‘c’e presentativo’. Forme e funzioni. In Anna-Maria De Cesare & Angela Ferrari (eds.). Lessico, grammatica e testualita, tra italiano scritto e parlato. University of Basilea: Acta Romanica Basiliensia 18: 127-153. Francez, Itamar. 2007. Existential Propositions. Ph.D. Thesis, Stanford University. Freeze, Ray. 1992. Existentials and other locatives. In Language 68.3: 553-595. —. 2001. Existential constructions. In Haspelmath, Martin, Ekkehard König, Wulf Oesterreicher & Wolfang Raible (eds.). Language Typology and Language Universals: An International Handbook, Vol. 2, 941-953. Berlin: Walter de Gruyter. Guasti, Maria Teresa. 1993. Causative and Perception Verbs. A Comparative Study. Turin: Rosenberg & Sellier. Hazout, Ilan. 2004. The syntax of existential constructions. In Linguistic Inquiry 35: 393-430. Higginbotham, James. 1987. Indefiniteness and predication. In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness, 43-70. Cambridge, MA: The MIT Press. Jones, Michael A. 1993. Sardinian Syntax. London: Routledge. Keenan, Edward. 1987. A semantic definition of indefinite NP. In Eric Reuland & Alice ter Meulen (eds.). The Representation of (In)definiteness, 286-317. Cambridge, MA: The MIT Press. —. 2003. The definiteness effect: Semantics or pragmatics?. In Natural Language Semantics 11: 187-216. Koontz-Garboden, Andrew. 2009. Locative and existential constructions in Ulwa. In Anthropological Linguistics 51.3/4: 244-268. La Fauci, Nunzio & Michele Loporcaro. 1997. Outline of a theory of existentials on evidence from Romance. In Studi italiani di linguistica teorica e applicata 26: 5-55. Lambrecht, Knud. 1994. Information Structure and Sentence Form. Cambridge: Cambridge University Press. —. 2002. Topic, focus, and secondary predication. The French presentational relative construction. In Beyssade, Claire, Reineke BokBennema, Frank Drijkoningen & Paola Monachesi (eds.). Proceedings of Going Romance 2000, 171-212. Amsterdam: John Benjamins.

Pseudo-Existentials and Definiteness Effects in Italian

147

Leonetti, Manuel. 2008. Definiteness effects and the role of the coda in existential constructions. In Henrik Høeg Müller & Alex Klinge (eds.). Essays on Nominal Determination, 131-162. Amsterdam: John Benjamins. Marten, Luz. 2013. Structure and interpretation in Swahili existential constructions. In Italian Journal of Linguistics 25: 45-73. McNally, Louise. 1992. An Interpretation for the English Existential Construction. Ph.D. Thesis, University of California, Santa Cruz. Published by Garland Press, New York, 1997. —. 2011. Existential sentences. In Maienborn, Claudia, Klaus von Heusinger & Paul Portner (eds.). Semantics: An International Handbook of Natural Language Meaning, Vol. 2, 1829-1848. Berlin: Mouton de Gruyter. Milsark, Gary. 1974. Existential Sentences in English. Ph.D. Thesis, MIT. Published by Garland Press, New York, 1979. Moro, Andrea. 1997. The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure. Cambridge: Cambridge University Press. —. 2009. Rethinking symmetry: a note on labeling and the EPP. In Snippets 19: 17-18. Parry, Mair. 2010. Non-canonical subjects in the early Italian vernaculars. In Archivio Glottologico Italiano 95: 190-226. Partee, Barbara & Vladimir Borschev. 2002. Genitive of negation and scope of negation in Russian existential sentences. In Jindrich Toman (ed.). Annual Workshop on Formal Approaches to Slavic Linguistics: The Second Ann Arbor Meeting 2001 (FASL 10), 181-200. Ann Arbor: Michigan Slavic Publications. —. 2007. Existential sentences, BE, and the genitive of negation in Russian. In Ileana Comorovski & Klaus von Heusinger (eds.). Existence: Semantics and Syntax, 147-190. Dordrecht: Springer. Rando, Emily & Donna Jo Napoli. 1978. Definites in there-sentences. In Language 54: 300-313. Remberger, Eva-Maria. 2009. Null subjects, expletives and locatives in Sardinian. In Georg Kaiser & Eva-Maria Remberger (eds.). Null subjects, expletives and locatives in Romance, 231-261. Konstanz: Universität Konstanz, Konstanzer Arbeitspapiere des Fachbereichs Sprachwissenschaft (No. 123). Rigau, Gemma. 1994. Catalan presentational sentences and the properties of Agr nodes. In Cinque, Guglielmo, Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaella Zanuttini (eds.). Paths towards Universal

148

Chapter Three

Grammar: Studies in Honor of Richard S. Kayne, 343-359. Washington: Georgetown University Press. —. 1997. Locative sentences and related constructions in Catalan: ésser / haver alternation. In Amaya Mendikoetxea & Myriam UribeEtxebarría (eds.). Theoretical issues at the morphology-syntax interface, 395-421. Bilbao: Universidad del País Vasco. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.). Elements of Grammar, 281-337. Dordrecht: Kluwer. —. 2001. Reconstruction, weak island sensitivity, and agreement. In Cecchetto, Carlo, Gennaro Chierchia & Maria Teresa Guasti (eds.). Semantic Interfaces, 145-176. Stanford, CA: CSLI. Russi, Cinzia. 2008. Italian Clitics: An Empirical Study. Berlin/New York: Mouton de Gruyter. Stowell, Tim. 1978. What was there before there was there?. In Farkas, Donna, Wesley M. Jacobson, & Karol W. Todrys, (eds.). Proceedings of the 14th Regional Meeting of the Chicago Linguistics Society, 458471. Chicago: Chicago Linguistics Society. Tortora, Cristina. 1997. The Syntax and Semantics of the Weak Locative. Ph.D. Thesis, University of Delaware. Villalba, Xavier. 2013. Eventive existentials in Catalan and the topic-focus articulation. In Italian Journal of Linguistics 25: 147-173. Ward, Gregory & Betty J. Birner (1995). Definiteness and the English existential. In Language 71: 722-742. Williams, Edwin. 1984. There-insertion. In Linguistic Inquiry 15: 131153. —. 1994. Thematic Structure in Syntax. Cambridge, MA: MIT Press. Zamparelli, Roberto. 2000. Layers in the Determiner Phrase. New York: Garland. Zucchi, Alessandro. 1995. The ingredients of definiteness and the definiteness effect. In Natural Language Semantics 3.1: 33-78.

CHAPTER FOUR (IN)DEFINITENESS EFFECTS ONCE AGAIN: THE CASE OF NEGATED EXISTENTIALS NADIA VARLEY

Abstract. This paper explores the nature of the Definiteness Effect (DE) while testing its applicability with respect to Bulgarian (negated) existentials. The DE universality is first questioned and then settled within a theoretical framework, which takes grammar to consist of distinct discourse-anchored functional heads serving as an escape hatch for elements that are otherwise constrained by universal formal properties operative in natural languages. In a nutshell, the general position of this paper boils down to the claim that the idea of the DE as a formal constraint of natural grammars should be maintained till a better explanatory structural account has been offered.

1. Introduction: Approaching the Object of Enquiry At least since Milsark’s (1974) analysis of English ‘there’ in existential sentences, most researchers agree that existential constructions are incompatible with presuppositional (definite/specific/anaphoric) DPs (list readings aside)1. Thus, in neutral contexts, existentials of the kind “There are *the children in the garden” are considered infelicitous. This ban seems to be valid not only in English:

1

As Rando & Napoli (1978) observe, ‘list sentences’ allow for definite NPs, because it is the list which should obey the restriction of non-anaphoricity (new information): A. I don’t have any friends. B. Oh, don’t be silly! There’s John and me and Susann and Peggy […] (Rando & Napoli 1978: 308, fn. 9)

Chapter Four

150

(1)

a. V sadu est’ (*éti/ *moi) deti. in garden be/have[-AGR].PRS *these/*my children ‘There are *the/*these/*my children in the garden.’ b. V in

gradinata garden-the

c. Im Garten gibt in-DAT garden gives

(Russian) ima deca(*ta). have[-AGR].PRS children-*DEF (Bulgarian) es (*die) Kinder. EXPL *the children (German)

Yet, there are apparent contexts when the internal arguments of existential predicates can escape the restricting effects posited by the domain of existential closure ‫׌‬vP (which I take to be below the phase edge, the lowest [Spec,vP], or VP in the sense of Diesing 1992). Thus, there is cross-linguistic empirical evidence that the observed Definiteness Effect (DE)2 can be relaxed under negation: (2)

a. Detej net (v sadu)3. children-GEN.PL NEG-be/have[-AGR].PRS (in garden) (Russian) b. Deca-ta gi njama (v gradinata). children-DEF CL.ACC.PL NEG-have[-AGR].PRS in garden-the ~‘The kids are not in the garden.’ (Bulgarian)

In Russian, the arguments of negated existentials are obligatorily licensed Genitive Case (the so-called Genitive-of-Negation). Furthermore, in an articleless language like Russian, the meaning of bare nouns (NPs not modified by a possessive, demonstrative or another definiteness-marking adjective) is decoded by means of Information Structure (IS). Thus, in out2

The following abbreviations are used throughout: ACC=accusative; AGR= agreement; AOR=aorist; CL=clitic; ClDbl=clitic doubling; DAT=dative; DE= Definiteness Effect; EC=existential closure; EXPL=expletive; F=feminine; GEN= genitive; IMP=impersonal; IMPRF=imperfect; (in)def=(in)definite; ILP/SLP= individual-level/stage-level predicate; IS=Information Structure; Loc=locative; M=masculine; MOD.PRT=modal particle; N=neuter; NEG=negation; NOM= nominative; NPI=negative polarity items; PL=plural; PRF=perfect; PRS=present; SG=singular; SPCF=specific. 3 Importantly, the locative ‘anchor’ is optional in these negated existentials.

(In)definiteness Effects Once Again

151

of-the-blue contexts, the topicalised NP in example (2a) is most likely to be interpreted as a definite/specific DP by the native speakers of Russian. In the reverse constellation “LocP-NEG V-NP”, the neutral interpretation of the NP would be one of an indefinite noun. In contrast, in Bulgarian, the arguments of negated existentials obviating the DE are unambiguously definite – both marked by a definite article and clitic-doubled. It is worth noting that the DE obviation under negation is not selfevident, given the unacceptability of definite expressions in negated existentials in other languages: (3)

a. *There wasn’t this beautiful girl at the party. (English) b. *Det var ikke den dama på festen i går. EXPL was not that lady-DEF on party-DEF yesterday (Norwegian) (Vangsnes 1999: 75, (115a), (116))

Vangsnes argues that “negated existentials predicate the non-existence of a referent denoted by the postverbal NP” (1999: 75), which according to him explains the ungrammaticality of (3a-b) above. Moreover, this author assumes that negation can be used in order to test specificity, so that negating a well-formed existential with a specific/referential NP yields ungrammatical results, (4b): (4)

a. There is a certain psychiatrist waiting in this office. b. *There isn’t a certain psychiatrist waiting in this office. (Vangsnes 1999: 36, (47a-b))

In Bulgarian, however, there is no such restriction. On the contrary, this language freely allows for definite/specific DPs in negated existentials; consider the Bulgarian version of (3a): (5)

Na partito vþera go njamaše on party-the yesterday CL.ACC.N/M NEG-have[-AGR].IMPRF xubavoto momiþe. beautiful-DEF.N girl (Bulgarian)

Bulgarian comes in handy because in this language the definite/specific arguments of negated existentials are both unambiguously marked by a definite marker and obligatorily clitic-doubled. Thus, in the scope of negation, a property can turn into a definite/specific expression, which is

152

Chapter Four

then – as I will be arguing – forced to move out of the domain of existential closure (EC) to a designated Topic position. What’s more, pace Vangsnes’ claim above, I will argue that negated existentials are not about “the non-existence of a referent” but about its absence relative to some spatio-temporal anchor (cf. §2.4). To the best of my knowledge, the facts I am going to present and tie together have not received any attention in the literature, at least not in connection to the DE. Given (2b) and (5), the main question posited here is whether Bulgarian (negated) existentials with definite DPs provide feasible counterevidence against the universality of the DE. Anticipating the conclusion, it is the aim of this paper to show that this apparent DE violation is none, once we decompose vP and CP. The organisation of this paper is as follows: In the next section I briefly sketch the theoretical background this paper builds on as well as discuss relevant notions such as existential predication, definiteness restriction, and existential negation. Then, in §3 I proceed with the issuerelated empirical facts from Bulgarian. Section 4 accounts for the apparent DE obviation in Bulgarian (negated) existentials, while implementing the idea that there are dedicated functional heads enabling the licensing of definite/specific arguments in existentials, with the latter being typically considered to shun definite expressions. The paper concludes with a brief summary of the questions under investigation.

2. Theoretical Background First, let me outline the relevant theoretical background in terms of which the discussion and argumentation of my proposal is couched. The theoretical scaffold of this paper is a clause partition into nuclear scope and restrictive clause as initially proposed in Diesing (1992) augmented with the cartographic approach, according to which CP, TP, and (upper) vP consist of multiple designated functional heads.

2.1. Diesing’s (1992) Mapping Hypothesis In an influential work, Diesing (1992) proposes that clauses be split into (i) nuclear scope or existential closure (VP), and (ii) restrictive clause (IP). Thus, only NPs remaining in the scope of existential closure can be interpreted as existential arguments/indefinites:

(In)definiteness Effects Once Again

153

Figure 1. The Mapping Hypothesis (tree splitting; Diesing 1992: 9, (13)).

Briefly, the Mapping Hypothesis predicts that a VP-internal NP gets bound by EC hence receiving a non-specific/existential reading, whereas material outside VP gets bound by a universal quantifier or generic operator, thus yielding a specific/generic/definite interpretation. I take Diesing’s claim that semantic partition has syntactic roots as the basis of my analysis. Intuitively, definite/specific NPs are infelicitous within the domain of EC. It is therefore all the more interesting to explain the apparent definite DP appearances in Bulgarian existentials. Following Diesing’s assumption that Topic and Focus are derived from the syntactic structure, let me now sketch the distribution of these discourse-anchored functional heads within the clausal architecture.

2.2. On the Distribution of Discourse Functional Heads Motivated by the principles of structural locality and local simplicity, the cartographic approach excludes the postulation of any complex X0/XPs as atoms of syntactic computation, which renders C, T and v-domains radically decomposed. Since Rizzi’s (1997) seminal paper, there has been an abundant line of theoretical and empirical investigation into both propositional and functional domains. Accordingly, the propositional domain CP has been traditionally assumed to consist of Force and Finiteness, with Focus and iterated Topic positions sandwiched in between (Rizzi 1997): (6)

[CP Force … Top … Foc … Top … Fin [TP ... [vP … ]]]

The logic of this clausal partition will be used in this paper throughout as to express aspects of the relationship between the discourse-interpretive component and syntactic configuration. For the purposes of argumentation,

Chapter Four

154

and much in line with Belletti (2005), I will mainly apply this conceptual thread to the first phase vP, while corroborating it with empirical evidence from Bulgarian existentials: (7)

[CP…[TP…[vP Top…Foc...Top…[vP ‫ ׌‬closure (NP)[VP (NP) ]]]]] NP move out of existential closure

Thus, a designated Top position within a split vP can best explain the appearance of a definite DP in the existential construction in (8): (8)

V gradinata in garden-the

[vP Top gi [‫׌‬vP njama (deca-ta)]. CL.ACC.PL NEG-have[-AGR].PRS (children-DEF) (Bulgarian)

The definite arguments of Bulgarian existentials are then to be analysed as definite/specific expressions outside the domain of EC, having escaped either by means of left dislocation (to CP) or via vP-internal topicalisation. As soon as an internal argument NP is matched for a specificity feature, it is forced to move out of vP/VP via an escape hatch to a higher domain in order to be interpreted.

2.3. Existential Predicates Importantly, as observed by Milsark (1974), there is a class of predicates which cannot co-occur with existential “there”: (9)

a. There was a man sick/*tall. b. There are people sick/*tall. (Milsark 1974: 214, (180 a-b;g-h))

In subsequent literature, the asymmetry between predicates such as ‘sick’ vs. ‘tall’ has come to be known as a distinction between stage-level (SLP) and individual-level predicates (ILP), with former allowing for both existential and generic meaning, and latter yielding only generic interpretation (Kratzer 1995). Most crucially, Kratzer argues that there is a syntactic distinction between SLPs and ILPs in terms of argument structure. Thus, SLPs have an extra argument position (Davidsonian argument), which ILPs do not license. This is evidenced by the fact that SLPs can be modified by spatial and temporal adjuncts, while ILPs cannot:

(In)definiteness Effects Once Again

155

(10)

a. Manon is dancing on the lawn. [dancing(Manon,l) & on-the-lawn(l)] b. Manon is dancing this morning. [dancing(Manon,l) & on-the-lawn(l)]

SLP

(11)

Manon is a dancer. dancer(Manon)

ILP (Kratzer 1995: 128, (12-14))

Thus, the observation has been that only SLPs are compatible with existential sentences, (12a) vs. (12b): (12)

a. There are firemen available. b.*There are firemen altruistic. (Kratzer 1995: 125, (1a-b))

Given that only SLPs can yield existential interpretation, it logically follows that only this sort of predicates can generate an extra argument position, which is a welcome result towards the explanation of certain ‘non-canonical’ properties of existential constructions as well as their syntax.

2.4. ‘Existential’ Syntax There has been this observation in the literature that cross-linguistically existentials often overlap with or are at least related to locative and possessive constructions (cf. e.g. Freeze 1992 for a unified locative analysis of these three construction types)4. Although existentials may 4

However, as Freeze (1992) argues, the difference between locatives and existentials lies in the availability of the DE in the latter but not in the former. This asymmetry is often evidenced by different constituent orderings across languages. Cf. the following Bulgarian examples: (i) Deca*(ta) sa v gradinata. (locative) children-DEF are-3PL in garden-the ‘The children are in the garden.’ (ii) V gradinata ima deca(*ta). (existential) in garden-the have[-AGR].PRS children-INDEF ‘There are children in the garden.’ (iii) Vaniya i Petâr imat deca(*ta). (possessive) Vanya and Peter have-3PL.PRS children-INDEF ‘Vanya and Peter have children.’ (Bulgarian)

156

Chapter Four

share certain properties with locatives and possessives, I follow the view that it is not necessary that a locative phrase specifies an existential construction (cf. also Allan 1971). In fact, the arguably most exploited “there-be” existentials in the literature (Milsark 1974; Ward & Birner 1995; Deal 2009, among many-many others) are not about locative predication at all (the upper-case letters mark the sentential Focus): (13)

a. THEREexist ARE cannibals in New Guinea. b. Thereexist ARE cannibals in New Guinea. c.*THEREexist are cannibals in New Guinea. (Allan 1971: 3, (4a-c))

Importantly, as Allan (1971) argues, ‘there’ in existentials cannot be empathically stressed without the auxiliary ‘be’ being stressed as well, hence the unacceptability of (13c). Building on this insight and in accord with Kratzer’s (1995) “extra argument in SLPs”-proposal, I will suggest that ‘there’ is generated in [Spec,v‫׌‬P] in a local configuration with the existential predicate:

Figure 2. The syntax of existential predication.

As such, ‘there’ is not the proform of LocP, the latter being a locative modifier under my account5. This is in line with Allan’s (1971) approach towards I assume that these three construction types share the same property with respect to the SLP selection (cf. §2.3). Nevertheless, only the arguments of existentials and possessives yield existential interpretation. 5 Cf. also Leonetti (2008), according to whom the coda (or postverbal locative XP in existential sentences) is generated as a VP adjunct: (i) [IP … [VP V [DP] [XP]] (Leonetti 2008: 139, (12))

(In)definiteness Effects Once Again

157

‘there’ in English existentials, with the minor – but crucial – difference that I assume that ‘there’ is not an existential operator but the spell-out of the Kratzerian spatio-temporal argument (Kratzer 1995). I take this ‘special’ argument to start off in [Spec,v‫׌‬P]. In many languages the spatio-temporal argument is spelled-out as Ø (e.g. Bulgarian, Russian, Spanish, a.o.). The ‘there’-problem has always been rooted in the fact that many accounts of existentials cross-linguistically have built on the model of English, which has led to many unwelcome stipulations (though cf. Freeze for arguing that “the English existential is unique”, 1992: 555). Assuming that ‘there’ is generated in [Spec,v‫׌‬P], where it copies the phi-features of the existential predicate ‘be’ and then raises to [Spec,TP] in order to satisfy the EPP condition6 can explain the DE observed within the domain of existential closure. With ‘there’ raised to the ‘EPP-position’, ‘there’associates can (in fact, are forced to) remain in situ, thus yielding a weak, existential interpretation. These succinct thoughts about the ‘there-be-NP’ complex can explain the general infelicity of definite DPs in postverbal positions in English-type languages. Therefore, there should be an alternative explanation of the much discussed cases of DE violation in ‘there’-existentials, to which I will briefly return in §4.1. Moreover, the empirical facts from languages lacking the equivalent of the expletive ‘there’ but still adhering to the DE, while obviating it in certain ‘special’ structural environments, invoke a more fine-grained approach towards the architecture of the vP-domain. In terms of the typological distribution of definite expressions within existentials, the question under investigation will be to explore why some languages do seemingly allow for definite arguments within existentials, while others appear to be more restrictive in this regard. A hint to the answer is the assumption that languages encode IS in different ways (cf. §4).

2.5. Some Words on Existential Negation Since the main object of exploration is the nature of apparent DE violation in negated existential constructions, some words should be spent on the

Under this account, the coda is crucial for the interpretation of the VP-internal DP. As Leonetti shows, the apparent DE violation in Italian and Catalan is none in cases when the DP and the locative XP co-exist in the VP-domain. 6 Here I adopt the EPP-account of Alexiadou & Anagnostopolou (1998), according to which the EPP can be satisfied either by agreement or by an XP (cf. also Deal 2009 for an approach towards ‘there’ generated vP-internally and raised to [Spec,TP] which builds on locality and cyclicity in the sense of Chomsky 2001).

158

Chapter Four

status of sentential negation. Intuitively, sentential negation is about denying a proposition (NegP = ¬ proposition; cf. Lyons 1977: 768-77)7. Crucially, I assume that NegP happens on a parallel plane converging to the core clausal architecture at the juncture when syntactic material gets mapped onto the interfaces. This is compatible with the insight that NegP is ‘located’ above vP, at the Transfer point after the phase has been handed over to the interfaces (I assume that unaccusative vPs (propositions) are also phases, cf. Chomsky 2001 on vP as a ‘weak phase’). When merged, propositional/sentential negation binds the arguments within EC, e.g. by introducing an existential quantifier or negative polarity items (NPI) binding the indefinite NP: (14)

a. V gradinata [NegP njama [vP ¬‫׌‬xሺnikakvi) deca]]. in garden-the NEG-have[-AGR] any/none-PL children-PL (Standard Bulgarian) b. V gradinata [NegP njama [vP ¬‫׌‬xpuknato dete]]. in garden-the NEG-have[-AGR] cracked-SG.N child-SG.N ‘There are no kids in the garden (whatsoever).’ (colloquial Bulgarian)

The fact that NPIs are indefinites licensed by negation under Agree makes them incompatible with definite/specific DPs8:

7

That negated NPs are not empty sets is evidenced by the following example: (i) There were no students at my lecture. I had forgotten that I had given them a day off. (Vangsnes 1999: 38, fn. 16) As this example shows, the pronoun in the second sentence refers to a set of students, which cannot be empty. As such, the pronoun establishes a discourse referent, whose content is identifiable via the context. I thus assume that negated existentials with definite objects imply presupposition that there exists a nonempty set of individuals, which under certain circumstances could have been there. Or simply, in order to negate something you first have to be aware of its existence. 8 Importantly, the NPI ‘nikakvi’/none (any) is compatible with definite expressions when used as a secondary predicate: (i) Decata (*gi) njama nikakvi (v gradinata). children-DEF CL.ACC.PL NEG-have[-AGR] none-PL in garden-the ~ ‘There is no sign of the children (in the garden).’ (Bulgarian) This fact, however, is beyond the scope of investigation of the present paper.

(In)definiteness Effects Once Again

(15)

159

V gradinata (*gi) njama nikakvi(-*te) deca. in garden-the *CL.ACC.PL NEG-have[-AGR] none-*DEF.PL children Intended: ‘There are no (specific) children in the garden.’ (Bulgarian)

Formally, I assume the following structure for the domain of negative EC, where Neg0 enters into a local relationship with ‫׌‬vP and ‫ ׌‬predicate:

Figure 3. The syntax of existential negation.

Nevertheless, my intuition is that negation – even if supportive –, is possibly not the most crucial factor for licensing definite objects in Bulgarian existentials9.

9

In fact, in certain contexts, the DE can be obviated in assertive existential constructions just as well, cf. §3.3. This would be my partial answer to the question posited by a reviewer, namely “[…] what prevents definite noun phrases from appearing in assertive/positive existentials by making use of the same escape hatch – just the same way as it has been suggested for DPs in negative existentials”. The restriction on definite DPs in assertive existentials is rather bound to the (lexical) semantics of the utterance, a discussion beyond the limitations of this paper. In fact, definite expressions often go well with assertive existentials: (i) Blagodarna sâm, þe gi ima decata, thankful-FEM be-1SG that CL-ACC.3PL have[-AGR] kids-DEF þe te ima teb. That CL.ACC-2SG have[-AGR] you-ACC ~‘I am thankful that there are the children/that there is you.’ (Bulgarian; Franks & King 2000: 54, (75f))

160

Chapter Four

3. Definiteness, Specificity, and Information Structure in Bulgarian Citing previous work, Milsark observes that “[p]erhaps the most frequently noted fact […] is that the NP of existential sentences can never be interpreted as topic [italics mine] [because] non-quantified NP cannot be topic” (Milsark 1974: 218). Nevertheless, the Bulgarian facts which I am going to discuss provide counterevidence to the first part of Milsark’s claim and show that the definite arguments of negated existentials can í in fact, must í be topics.

3.1. Bulgarian Bulgarian is a South Slavic language, which unlike the majority of other Slavic varieties has definite articles and indefinite determiners10 in the lack of operative case system11 as well as exhibits the so-called clitic doubling (ClDbl) of internal arguments (e.g. Mišeska Tomiü 2006: 86-108; 259270; also Guentchéva 2008). Importantly, obligatory clitic doubling in Bulgarian occurs with definite/specific and topicalised internal arguments (cf. Franks & Rudin 2005). The clitic integrates the phi-features of its DPassociate and as such very much approximates the characteristics of object agreement in this language. The more striking is this observation given that the (definite) internal argument is obligatorily clitic-marked in impersonal, non-agreeing constructions in Bulgarian such as haveexistentials (cf. Guentchéva 2008; Krapova & Cinque 2008). Let me first sketch the main features of Bulgarian (obligatory) clitic doubling and then proceed with the (in)definiteness effects observed in Bulgarian have-existentials.

3.2. On Clitic Doubling in Bulgarian As mentioned, pronominal clitics phi-, case- and number-agree with their associate DPs. Given that the associate DP is the actual argument, a logical assumption would be that the clitic has the role of (object) agreement or is 10

To be on the safe side, I should stress that contemporary Bulgarian is (possibly) on the way to develop ‘edin’/one into indefinite article. For the time being, it is more accurate to refer to ‘edin’ as ‘indefinite determiner’ (cf. e.g. Geist 2013). 11 Remnant case is preserved in the pronominal system of Bulgarian. A reviewer asks whether I assume any logical relationship between the availability of articles vs. operative case system: The answer is ‘no’. This observation only serves to the typological classification of Bulgarian compared to other Slavic varieties.

(In)definiteness Effects Once Again

161

the instantiation of some formal feature. Nevertheless, this cannot be the whole truth since not all objects are subject to obligatory ClDbl in Bulgarian12: (16)

Cvetjata (gi) kupix, a vinoto (go) zabravix. flowers-the (CL) bought-1SG and wine-the (CL) forgot-1SG ‘I bought the flowers (I meant to buy), but forgot the bottle of wine (which I also meant to buy).’ (Bulgarian)

The observation that obligatory ClDbl in Bulgarian occurs with definite/specific and topicalised internal arguments (cf. Franks & Rudin 2005; Guentchéva 2008) is also in line with Kallulli (2008) who argues that ClDbl produces and encodes IS in a systematic way, which renders ClDbl a “topic-licensing operation”: (17)

a. Strandžata izvednâž *(go) svali Strandzha-the(nickname) suddenly CL.ACC.M overthrew-2/3SG.PST bolestta na legloto. illness-the on bed-the ‘Strandzha was knocked down by a sudden illness.’ b. Dvete zaptieta *(gi) izjali vâlci taja zaran. two-the gendarmes CL.ACC.3PL ate-PRF wolves this morning ‘Both gendarmes were eaten by wolves this morning.’ (Bulgarian; in Haralampiev 2001: 206)

Without ClDbl, ‘Strandzha’ in (17a) is interpreted as “suddenly knocking down the illness onto the bed”, and both gendarmes in (17b) as “having eaten wolves this morning”. Importantly, while the topicalised associate in ClDbl constructions can be omitted, the clitic can never be: (18)

12

(Tebe) ne mogat da *(te) you-ACC NEG can-3PL MOD.PRT you-CL.ACC kâm nikogo. to no one-ACC. ‘They can’t attach you to anybody.’

prikrepjat attach-3PL

Both focused objects in (16) can be optionally clitic-doubled. This is a matter of regional and interlinguistic variation.

162

Chapter Four

Only, when ‘tebe’ is contrastively focused, can the clitic be dropped: (19)

Samo tebe ne mogat da (te) prikrepjat only you-ACC NEG can-3PL MOD.PRT you-CL.ACC attach-3PL kâm nikogo. to no one-ACC ‘It is only you whom they can’t attach to anybody.’ (Bulgarian; modified from Krapova & Cinque 2008: 260, (8a))

These facts demonstrate that obligatory clitic-doubling is topic-sensitive. The pronominal clitic is introduced as a special functional head which has a DP associate within its domain. In the spirit of the ‘big DP’ accounts offered in the literature, Franks & Rudin (2005) offer an approach of clitic doubling in Bulgarian, which I adopt for the purposes of my analysis. These authors claim that clitics in Bulgarian, being K0 heads, head their own projection, KP, which incorporates a DP shell: (20)

[KP clitic [DP D [NP/QP/AP ]]] (Franks & Rudin 2005: 113, (13))

Thus, only when the DP vacates the KP boundary does clitic doubling become obligatory in Bulgarian. This can explain the asymmetry between the obligatorily clitic-doubled sentence in (21a) and the optionality of the clitic in (21b): (21)

a. Statijata naj-posle *(ja) napisax. article-the at last CL-ACC.F wrote-1SG.PST b. Naj-posle (ja) napisax statijata. at last CL-ACC.F wrote-1SG.PST article-the ‘I wrote the article at last.’

Importantly, the clitic always raises above the verb in non-V1 constructions13. 13

Clitics in Bulgarian strictly obey the Tobler-Mussafia Law, which means that V1 constructions in this language are always verb-enclitic: (i) Njama ja Marija. (Bulgarian) NEG-have CL.ACC.F Maria ‘Maria is not here.’ (ii) Razxodix go kuþeto, ne se pritesnjavaj. walked-AOR.1SG CL.ACC.N dog-DEF.N NEG REFL worry-2SG ‘I already walked the dog, don’t worry.’

(In)definiteness Effects Once Again

163

Hence, ClDbl is the reflex of a DP (with a [+TOP] feature) vacating the KP domain to a higher Topic position. Importantly, the clitic-doubled version of (21b) yields a different interpretation than its clitic-less counterpart. This, as I will argue in §4, is due to the availability of TopP within the vPdomain. Let me sum up the main features of ClDbl in Bulgarian which are relevant for the analysis to-come: (i) the relationship between the pronominal clitic and its associate is a form of (object) agreement14, which explains their sharing phi-, case- and number-features15; (ii) ClDbl is a grammatical device with direct impact on Information Structure (Topic/Focus) in this language; and (iii) is tied to the syntactic mapping of arguments within EC vs. that within restrictive clause in the sense of Diesing (1992). In sum, the role of the clitic is to remove the definite/specific expression (the topic) from the focused domain much in the same way scrambling in German effects the semantic interpretation of the clause: (22)

[CP daß [IP Otto immer [VP Bücher über Wombats liest]]] that Otto always books about wombats reads ‘[…] that Otto always reads books about wombats.’ existential closure interpretation of ‘Bücher’/books

This is possibly due to the fact that bare heads (clitics) cannot be topicalised in Bulgarian, an observation I have yet no straightforward explanation for. 14 A long-standing insight in the literature is that the definite article instantiates an agreement between D and NP, which is reproduced by the structure in (20), repeated here as (i): (i) [KP clitic [DP D [NP/QP/AP ]]] In a sense, adopting the idea that pronominal clitics are agreement markers is the most logical outcome of the line of reasoning pursued here. Furthermore, it will be not counterintuitive to assume that clitics grammatically block semantic incorporation (cf. also Leonetti 2008 for some considerations along these lines). This can also explain the optionality (and in certain cases the ban) of clitic doubling with focused DPs/NPs in Bulgarian (cf. Suñer’s 2000 discussion of the focus incompatibility with clitic doubling in Spanish). 15 A potential complication arises when accounting for such cases when the clitic agrees only with the first DP-associate: (i) Njama go /*?gi [&P Petâr i decata]. NEG-have[-agr] CL.ACC.M /*CL.ACC.PL Peter and kids-the ‘Neither Peter nor the kids are there.’ This ‘non-agreement’ is evidence that the obligatory clitic doubling in Bulgarian existentials is formally licensed.

164

(23)

Chapter Four

[CP daß [IP Otto Bücher über Wombats immer [VP liest]]] that Otto books about wombats always reads QR reading bound by the adverbial ‘immer’/always (German; Diesing 1992: 107, (31); 108, (32))

Once the NP has escaped the domain of existential closure in German, it is interpreted as specific or generic. As observed in previous research (Franks & Rudin 2005; Guentchéva 2008; Krapova & Cinque 2008), obligatory clitic doubling is very much dependent on the nature of the predicate involved. Thus, psych and psychical perception/experiencer and possessor impersonal predicates in Bulgarian require an obligatory clitic. Importantly, the fact that the arguments of impersonals are typically interpreted as the sentential focus favours indefinite NPs in these constructions. This requires a thorough investigation into the cases of apparent DE violation in Bulgarian haveexistentials, which is done in the next section.

3.3. Have-Existentials in Bulgarian Existentials in Bulgarian are formed with the non-agreeing form of the relational predicate ‘have’, cf. the agreement patterns in (24a) vs. (24b): (24)

a. V muzeja ima skâpi kartini in museum-the have[-AGR].PRS expensive paintings-INDEF.PL /*skâpite kartini. /*expensive-DEF paintings ‘There are valuable paintings in the museum.’ (existential; Bulgarian) b. Mnogo mecenati imat skâpi many patrons of the arts have-3PL.PRS expensive kartini /*skâpite kartini. paintings-INDEF/*expensive-DEF paintings ‘Many patrons of the arts possess valuable paintings.’ (possessive; Bulgarian)

While apparently operative in both existentials and possessives, the DE seems to relax under negation in Bulgarian existentials, cf. (25b):

(In)definiteness Effects Once Again

(25)

165

a. V muzeja njama skâpi kartini. in museum-the NEG-have[-AGR] expensive paintings-INDEF.PL ‘There are no valuable paintings in the museum.’ b. V muzeja gi njama in museum-the CL.ACC.3PL NEG-have[-AGR] skâpite kartini. expensive-DEF.PL paintings ~‘The valuable paintings are not to be found in the museum.’ (negated existential; Bulgarian)

As already discussed, clitic doubling is obligatory in impersonal ima/njama-existentials with definite DPs: (26)

Men *(me) ima/ njama v grafika me-ACC me-CL.ACC have/NEG-have[-AGR] in schedule-the za utre. for tomorrow ‘I am/I am not in the (working) schedule for tomorrow’ (lit. ‘There is (no) me in […]’) (Bulgarian)

Nevertheless, it is worth noting that such DE obviation does not go without certain semantic repercussions. Thus, whereas an abstract LocP such as that in (26) above, ‘v grafika’/in the schedule, ‘v spisâka’/on the list, ‘v dogovora’/in the contract, etc. is well-formed in both ‘assertive’ and ‘absence’ existentials, factual LocPs like ‘v gradinata’/in the garden, ‘vkâšti’/at home, etc. are not necessarily acceptable: (27)

Men *(me) *ima/ njama vkâšti. me-ACC me-CL.ACC *have/NEG-have[-AGR] at home ‘I am *(not) at home’ (lit. ‘There is (no) me at home’). (Bulgarian)

Asymmetries like those in (26) vs. (27) lead me to assume that the DE violation cannot be merely explained by purely pragmatic approaches (pace Abbott 1993; Ward & Birner 1995, a.o.; cf. §4.1). Furthermore, the semantic nature of LocP in these existentials should be explored in more depth in order to formally account for the effects observed in (26) vs. (27).

166

Chapter Four

Intuitively, this difference has to do with the event structure of the proposition16. However, I leave this complication for future research.

4. Elaborating on the Relationship between Topic-Focus Partition and DE Languages encode IS in different ways. In this paper, I pursue an approach which bears a distant resemblance to Leonetti (2008) who states that “[t]aking into account the principles of information structure in each language is essential for our understanding of the link between syntactic positions and definiteness” (Leonetti 2008: 139). I assume with Leonetti that definiteness/specificity is intrinsically bound to Topic syntactic positions, while indefiniteness/non-specificity is related to Focus. Nevertheless, the proposal I advance here formally diverges from Leonetti’s account. Most notably, it is a structural approach towards DE which takes the decomposition of the clause into nuclear scope and restrictive clause in the sense of Diesing (1992), combined with the workings of Information Structure (IS), more precisely with information partition of the clause (both vP and CP) into Topic and Focus functional heads. Along these lines, existential closure is focus sensitive, whereas restrictive clause is bound to referentiality and definiteness/specificity.

4.1. Are Definiteness Effects merely Pragmatic? Some authors have argued that pragmatic constraints are the most crucial factor for the explanation of the DE violation in there-existentials (Abbott 1993; Ward & Birner 1995, a.o.):

16

Preposing the LocP in example (27) does not make the sentence sound better either (in contrast to the Catalan and Italian data presented in Leonetti (2008) who argues that it is VP-internal codas that posit constraints on definite expressions within VP, his ‘Coda Constraint’, p. 142, (22)), cf.: (i) ?Vkâšti me *ima/ njama (Bulgarian) at home me-CL.ACC *have/NEG-have[-AGR] ‘I am *(not) at home’ (lit. ‘There is (no) me at home’).

(In)definiteness Effects Once Again

(28)

167

a. There weren’t the funds necessary for the project we had in mind. b. There are those who would claim that definites are impossible in existentials. c. There was the smell of pot all over the apartment. d. There was the usual crowd at the beach last Sunday. (Abbott 1993: 44, (10a-d))

Abbott argues against the proposal that such definite DPs be interpreted as list readings (pace Milsark 1974, a.o.). Instead, she claims that these “‘cataphoric’ definites, which do not denote an entity already introduced into the discourse, but rather are licensed by their ability to specify a uniquely identifiable entity which is being newly introduced, are also acceptable without special contextuali[s]ation, and thus can occur in NEs [non-contextualised existentials]” (Abbott 1993: 44). Whereas I agree that notions such as common ground and truth conditions on propositions are relevant for the interpretation of the definite expressions in there-existentials, it is not clear to me how a purely pragmatic approach can explain the Bulgarian facts discussed in §3.3. Furthermore, Abbott proposes that the unacceptability of utterances like: (29)

*There is my sister in the room.

is derived from pragmatics (e.g. focus incompatibility of the NP in question) rather than from syntactic/semantic constraints. Abbott further argues that due to the fact that the NP ‘my sister’ is unique, it cannot be modified by a LocP ‘in the room’. Whereas, this observation might hold for ‘assertive’ existentials in Bulgarian, cf. example (30a)17, the ‘absence’ existentials do allow for this unique referent in combination with LocP, (30b):

17 But again (cf. example (26) above), when modified by an abstract LocP (e.g. “in the testament”), the sentence is fully acceptable: (i) Sestra mi ja ima v zaveštanieto. CL-ACC.F have[-AGR] in testament-the sister me-CL ~‘My sister is in the testament.’ This is yet further evidence that semantics is an important factor to be taken into consideration when assessing the full range of the DE.

168

(30)

Chapter Four

a.*?V stajata ja ima sestra mi. in room-the CL.ACC.F have[-AGR] sister CL.DAT.1SG Intended: ‘My sister is present in the room.’ (assertive existential; Bulgarian) b. V stajata ja njama sestra mi. in room-the CL.ACC.F NEG-have[-AGR] sister CL.DAT.1SG ~‘My sister is not present in the room.’ (negated existential; Bulgarian)

However, the indefinite version of (30a) fails completely on lexicalsemantic grounds, cf. the asymmetry between (31) and (32): (31)

*#V stajata ima sestra. in room-the have[-AGR] sister-INDEF Intended: ‘(A) sister is present in the room.’18

(32)

V stajata ima piano. in room-the have[-AGR] piano-INDEF ‘There is a piano in the room.’

(Bulgarian)

Such asymmetries can counter any approach that considers only one module of grammar to be responsible for the DE observed.

4.2. The DE: An IS and yet Syntactic Constraint Functionalist analyses (e.g. Ward & Birner 1995) cannot fully account for the occurrence of definite DPs in existentials. Given the approach undertaken here, the apparent instances of DE violation can be explained by positing specialised functional heads in the syntax. I am going to argue that the formal analysis of the DE in existentials can be preserved by showing that existentials with definite objects provide a resort for the definite expression to a topic position either within a split CP in the sense of Rizzi (1997), cf. (33a), or split vP (Belletti 2005), (33b). This is clearly manifested in Bulgarian negated existentials:

18 Of course, this holds only for those readings when ‘sestra’/sister is used as a kinship term, and not as ‘nun’ or ‘nurse’.

(In)definiteness Effects Once Again

(33)

169

a.[TOP Decata *(gi) [vP njama. children-DEF CL.ACC.PL NEG-have[-AGR] b. Njama [TOP *(gi) decata. NEG-have[-AGR] CL.ACC.PL children-DEF ~’The children are absent.’ (Bulgarian)

As soon as the definite/specific object in Bulgarian existentials is topicalised, it is obligatorily clitic-doubled (cf. §3.3)19. What is more important, the evidence provided above suggests that there is no option for the definite argument to get a broad focus interpretation in negated existentials (recall the by far indisputable claim in the literature that negation is an operator closely related to Focus). A further argument in favour of vP-internal designated Topic positions in Bulgarian is the fact that in declaratives the existential have-predicate ‘ima’/‘njama’ does not raise to T20. The discussion so far lead me to conclude that decomposing vP provides ‘offshore’ slots for the otherwise illegitimate definite/specific DPs in existentials. As such, my modification of Diesing’s structure (Fig.1) is a partitioned vP with allocated Foc and Top heads in the spirit of Rizzi’s split CP and Belletti’s split VP accounts:

19

Assuming that definites must “check” Case either via agreement or by means of a preposition can account for the obligatory ClDbl in Bulgarian impersonals (this holds if the claim that ClDbl is a kind of object agreement is on the right track). 20 Following Alexiadou & Anagnostopolou’s EPP parametric approach, the motivation behind the v-to-T movement is EPP-checking. Since there is no need of checking agreement features in Bulgarian existentials, v-raising becomes redundant. In (33b), on the other hand, the movement of the negated existential verb ‘njama’ is discourse-driven (cf. Lambova 2004 on discourse-motivated VSO derivations in Bulgarian).

Chapter Four

170

Figure 4. Split vP with TopP and FocP; existential closure in low vP/VP.

Further cross-linguistic evidence for the relevance of IS and Topic respectively as regards the DE phenomenon comes from German: (34)

gibt den Weihnachtsmann nicht. gives the-ACC Santa Claus not b. Den Weihnachtsmann gibt es nicht. the-ACC Santa Claus gives EXPL not ‘Santa Clause does not exist.’ a. Es

EXPL

(35)

Gott sei Dank, es gibt den Sonntag. God be thank, EXPL gives the-ACC Sunday ~‘Thank God, there is (at least) Sunday.’ (German; via google)

Arguably, the DE can be violated in German ‘es gibt’ existentials. This can be due to the existential predicate, which is transitive in this language. I would like, however, to explore the possiblity that the DE violation is due to the availability of specialised functional heads immediately above the domain of ‫׌‬vP (Fig. 4).

(In)definiteness Effects Once Again

171

In German, definite expressions in existentials are possible because they are licensed in the restrictive clause, above vP, (34a; 35) or in the left periphery, in [Spec,TopP], cf. (34b). Existential closure is thus evacuated. Only indefinites may stay in situ: (36)

Heute gibt es [vP einen interessanten Film auf Arte]. today gives EXPL [vP a-ACC.M interesting movie on Arte] ‘There is an interesting movie at Arte today/tonight.’ (German)

As Diesing (1992) has shown, indefinites and definites occupy different structural positions. This is also evidenced by the following asymmetry as regards the negation in German: (37)

Heute gibt es leider [keinen today gives EXPL unfortunately none-ACC.M.INDEF interessanten Film]. interesting ACC.M movie ‘There is no interesting movie today/tonight.’

(38)

Heute gibt es [den gestrigen interessanten Film] today gives EXPL the-ACC.M yesterday’s interesting movie leider nicht. unfortunately not ‘Unfortunately, the yesterday’s interesting movie is not on today.’ (German)

Thus, only definites raise out of the domain of EC past the scope of negation. The presence of the quasi argument “es” in [Spec,TP] blocks the movement of the definite expression to the site, which under Diesing’s Mapping Hypothesis has been known as “restrictive clause”. Accordingly, the DP ‘den gestrigen interessanten Film’ in (38) is moved to a dedicated [Spec,TopP] position above ‫׌‬vP. In this section, I provided evidence that there are designated Topic positions within split CP and split vP domains. As a result, the answer to the question whether negated existentials in Bulgarian supply counterevidence to the universality of the DE as a formal constraint is ‘no’. The apparent definite and clitic-doubled DPs in Bulgarian existentials are due to IS principles operative in this language. As such, the interplay between syntax and IS (Topic-Focus functional heads) can – at least partially – unravel the DE “puzzle”.

172

Chapter Four

5. Concluding Remarks In this paper I discussed cases of apparent DE obviation in existentials and proposed a solution couched in terms of the cartographic method, which allocates discourse features such as Topic and Focus within syntactic phase domains (Rizzi 1997; Belletti 2005). I argued that the DE as a grammatical constraint should be maintained. As such, the aim has been to restate the syntactic validity of the DE in terms of cartographically split vP/CP and syntax encoding Information Structure. Building mainly on evidence from Bulgarian negated existentials I have shown that the apparent DE violation in Bulgarian existentials is due to independently motivated discourse features in the syntax. This fact sustains the validity of the DE within the ‫׌‬vP/VP domain (Diesing 1992). To sum up, the Bulgarian facts presented in this paper support the idea that the DE is to be maintained as a universal constraint that natural grammars resort to. Importantly, IS principles are essential for deriving the (In)Definiteness Effects observed across languages. In this regard, Information Structure is a crucial negotiator between the semantics of (in)definiteness and syntactic structure. As such, the question posited at the beginning of the paper, namely whether Bulgarian negated existentials provide counterevidence against the DE universality, must be negated. Last but not least, the novelty of my approach consists in relating certain (well-known) empirical facts about clitic doubling and topicality/ specificity in Bulgarian to the DE in (negated) existentials. To my knowledge, this particular connection has gone unnoticed in the literature, which provides a fertile ground for further investigation of the relationship between information structure, (in)definiteness effects, and negation.

Acknowledgements I am grateful to Esther, Susann, and Tanja for organising the DE workshop at the DGfS 2012 in Frankfurt. I also owe them a debt of gratitude for their patience and good will as regards the final submission of the paper. A big ‘thank you’ goes to two anonymous reviewers whose remarks, constructive criticism, and helpful sugestions have only been beneficial to the improvement of this paper. The usual disclaimers apply.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of Pragmatics 19: 39-55.

(In)definiteness Effects Once Again

173

Alexiadou, Artemis & Elena Anagnostopolou. 1998. Parametrizing AGR: word order, V-Movement and EPP-Checking. In Natural Language and Linguistic Theory 16: 491-539. Allan, Keith. 1971. A note on the source of There in existential sentences. In Language 7.1: 1-18. Belletti, Adriana. 2005. Extended doubling and the VP periphery. In Probus 17.1: 1-35. Chomsky, Noam. 2001. Derivation by phase. In K. Hale & M. Kenstovicz (eds.). A Life in Language, 1-54. Cambridge: MIT Press. Deal, Amy Rose. 2009. The origin and content of expletives: Evidence from ‘‘Selection’’. In Syntax 12.4: 285-323. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Franks, Steven & Tracy Holloway King. 2000. A Handbook of Slavic Clitics. Oxford: Oxford UP. Franks, Steven & Catherine Rudin. 2005. Bulgarian clitics as Kº heads. In Steven Franks, Frank Y. Gladney, and Mila TassevaǦKurktchieva (eds.). Formal Approaches to Slavic Linguistics: The South Carolina Meeting, 104–116. Ann Arbor, MI: Michigan Slavic Publications. Freeze, Ray. 1992. Existentials and other locatives. In Language 68.3: 553-95. Geist, Ljudmila. 2013. Bulgarian edin: the rise of an indefinite article. In Uwe Junghanns et al. (eds.). Proceedings of FDSL 9 (Formal Description of Slavic Languages), 125-48. Frankfurt a.M: Peter Lang. Guentchéva, Zlatka. 2008. Object clitic-doubling constructions and topicality in Bulgarian. In D. Kallulli & L. Tasmowski (eds.). CliticDoubling in the Balkan Languages, 203-23. Amsterdam: J. Benjamins. Haralampiev, Ivan. 2001. Istoriþeska Gramatika na Bylgarskia Ezik (A Historical Grammar of the Bulgarian Language). V.Tarnovo: Faber. Kallulli, Dalina. 2008. Clitic doubling, agreement and information structure: The case of Albanian. In D. Kallulli & L. Tasmowski (eds.). Clitic-Doubling in the Balkan Languages, 227-55. Amsterdam: J. Benjamins. Krapova, Iliyana & Guglielmo Cinque. 2008. Clitic reduplication constructions in Bulgarian. In D. Kallulli & L. Tasmowski (eds.). Clitic-Doubling in the Balkan Languages, 257-87. Amsterdam: J. Benjamins. Kratzer, Angelika. 1995. Stage-Level and Individual-Level Predicates. In Carlson, G. & F. J. Pelletier (eds.). The Generic Book, 125-75. University of Chicago Press.

174

Chapter Four

Lambova, Mariana. 2004. On triggers of movement and effects at the interfaces. In A. Breitbarth & H. C. van Reimsdijk (eds.). Triggers, 231-58. Berlin: Walter de Gruyter. Leonetti, Manuel. 2008. Definiteness effects and the role of the coda in existential constructions. In H. Hoeg-Mueller & A. Klinge (eds.). Essays on Nominal Determination, 131-62. Amsterdam: J. Benjamins. Lyons, John. 1977. Semantics 2. Cambridge: Cambridge UP. Milsark, Gary. 1974. Existential Sentences in English. Unpublished Ph.D. Dissertation, MIT Cambridge, MA. Mišeska Tomiü, Olga. 2006. Balkan Sprachbund Morpho-syntactic Features. Dordrecht: Springer. Rando, Emily & Donna Jo Napoli. 1978. Definites in There-sentences. In Language 54.2, 300-13. Rizzi, Luigi. 1997. The fine structure of the left periphery. In L. Haegeman (ed.). Elements of Grammar, 281-337. Dordrecht: Kluwer. Suñer, Margarita. 2000. Object-shift: comparing a Romance language to Germanic. In Probus 12: 261-89. Vangsnes, Øystein Alexander. 1999. The Identification of Functional Architecture. Unpublished Ph.D. Dissertation, University of Bergen. Ward, Gregory & Betty Birner. 1995. Definiteness and the English existential. In: Language 71.4: 722-42.

CHAPTER FIVE DEFINITENESS EFFECT, PRONOUNS AND INFORMATION STRUCTURE IN CATALAN EXISTENTIALS1 XAVIER VILLALBA

1. Introduction It is a well-attested fact that pronouns cannot be pivots of existential sentences generally: (1)

1

a.

*There is him/he in the room.

b.

(Milsark 1974: ex. 64) #There were them/those waiting outside. (McNally 1992: ex. 6b)

I am very grateful to Susann Fischer, Tanja Kupisch and Esther Rinke for their kind invitation to participate in this volume, even though the flu didn’t allow me to attend the DGfS 2012 workshop “Definiteness Effects” in Frankfurt. Since I recovered from the flu, I have presented parts of this work at the Universitat Autònoma de Barcelona, the Universitat Pompeu Fabra, and at Radboud University at Nimegen. In all the places, I received intelligent feedback from the audience, particularly from Bart Geurts, Josep Fontana, Klaus von Heusinger, M. Teresa Espinal, Laia Mayol, Louise McNally, and Rob van der Sandt. Moreover, two anonymous reviewers provided many insightful comments and suggestions that contributed to increase the quality of the paper. Finally, I must acknowledge Susann Fischer and Manuel Leonetti’s constant inspiration and advice in my expedition to Romance existential sentences. This research was funded by research projects FFI2011-23356 and FFI2014-52015 (Ministerio de Ciencia e Innovación, Spain), and 2009SGR1079 (Generalitat de Catalunya), awarded to the UAB.

Chapter Five

176

When this evidence is considered, the usual answer has been subsuming it under the Definiteness Effect (DE), namely the ban against definite DPs in pivot position: (2)

a. b.

#There is Mary at the front door. #There is the solution.

Evidence for this unification, comes from the fact that pronouns pattern with definite NPs and proper names in respecting the DE, but also in violating it in certain contexts, as in list readings (Milsark 1974, Rando & Napoli 1978, Abbott 1997): (3)

a.

b.

(4)

A. My God! How many people know about this? B. There’s me and there’s you. That’s all. (Rando and Napoli 1978: ex. 37) OK, let’s finish up this guest list. There’s you and me. Who else is coming? (Abbott 1997: 6)

A. I don’t have any friends. B. Oh, don’t be silly! There’s John and me and Susan and Peggy… (Rando and Napoli 1978: fn. 9)

So it seems that the DE makes the correct generalization regarding the behaviour of pronouns as pivots of existentials. Yet, when we move from English to Catalan, things change radically. Catalan is a language that easily allows definite DPs and proper nouns in pivot position (Rigau 1988a, Ramos 1998, Brucart & Rigau 2002, Leonetti 2008):2 (5)

a.

Hi

havia el degà, a la had the dean at the.F ‘At that meeting, the dean was present.’ En aquella assemblea, hi havia at that assembly LOC had ‘At that meeting, Mary was present.’ LOC

b.

2

reunió. meeting la Maria. the.F Mary

Throughout the text I follow the Leipzig Glossing Rules. I will use the following abbreviations: F=feminine, LOC=locative, PL=plural, PST=past, REFL=reflexive, SG=singular.

Catalan Existentials

177

Yet, regardless of the evidence against the DE in (5), there is consensus among Catalan grammarians in forbidding personal pronouns as pivots of the existential construction: (6)

a.

*Hi LOC

b.

ha has

ell. he (Rigau 1988a: ex. 2d) hi havia LOC had

‘He is here.’ *En aquella assemblea, at that meeting ella, tu i jo. she you and I ‘At that meeting it was she, you and I who were present.’ (Ramos 1998: ex. 7b)

There is no clear answer to this fact in Catalan: if you allow proper nouns in existentials, why should not a personal pronoun? So, if we take the DE as the explanatory factor underlying the ban against pronominal pivots, Catalan data remain mysterious, for Catalan violates the DE, while it shows a strong restriction against pronouns in pivot position. Moreover, to make the picture even more complex, one should note a fact neglected in the literature, namely that pronominal pivots are possible in Catalan existentials under special circumstances (examples from a Google search):

(7)

a.

b.

unes quantes dones, entre les a.F.PL some.F.PL women among the.F.PL quals hi havia ella i which.PL LOC had she and la Victòria. the.F Victory (M. Carme Rodríguez Virgili Benvinguda, Marta, p. 180) ‘few women, which she and Victoria were among.’ Al taller d'en Grau només hi at-the studio of-the Grau only LOC havia ell i en Xuriguera. had he and the Xuriguera (Ferran Canyameres, Diari íntim, p. 546) ‘Only he and Xuriguera were at Grau’s studio.’

Chapter Five

178

c.

d.

eren were hi

les the.PL havia had el the

sis, i allà només six and there only ell, la Infermera LOC he the.F nurse i Nen. and child (Jordi Dausà, La caixa de les joies, p. 47) ‘It was six o’clock and only he, the Nurse and the Child were there.’ pel que fa a personalitats, by-the that makes at personalities només hi ha ella. only LOC has she (El Periódico. 24/06/2010.) ‘as for personality, she is the only one.’

Interestingly, Italian, which also allows definites and proper names as pivots, parallels Catalan nicely, as the following examples picked up from the Internet demonstrate (see also Cruschina 2012, this volume):3 (8)

a.

b.

c.

3

Quando in cucina c’è solo lui when in kitchen LOC-is only he ‘When he is the only one in the kitchen.’ (book title by Nico Conti) «Emiliano? In Puglia non c’è solo lui» Emiliano in Puglia not LOC-is only he ‘Emiliano, he is not the only one in Puglia.’

Matteo Salvini: ormai a destra c’è solo lui Matteo Salvini now at right LOC-is only he ‘Matteo Salvini: now he is the only one at the right wing.’

The very same behaviour is found in Sardinian, as the detailed work by Bentley (2004, 2011) shows.

Catalan Existentials

d.

179

Case ai poveri, ma non ci houses to-the.M.PL poor but not LOC sono solo loro. are only they ‘House for the poor, but they are not the only ones.’

Furthermore, English also allows pronominal pivots in exactly the same circumstances, as the following examples from the Internet attest: (9)

a.

b.

c.

Behind him, they finished by shooting the already-dead Stebbins, and now there was only him, alone on the road, walking toward where the Major's jeep had stopped diagonally across the white line... (Stephen King, The long walk) Only Mikhail was sanity. He was on her mind continually, invading her thoughts, pushing out everything insane until there was only him. (Christine Feehan, Dark prince) “So I saw that there was only me. There was only me who could worry about what was happening here, inside these walls of my life. (Markus Zusak, Underdog)

To sum up, the issue raised by pronouns in Catalan and Italian existentials is the following: since they are languages systematically violating the DE, a solution based on the DE, as the one suggested for English, is difficult to defend. Rather, following McNally (1992) and Leonetti (2008), I will argue for a different line of attack, arguing that the DE is not a unitary phenomenon, but the conjunction of several pragmatic requirements involving the informational conditions imposed on pivots and the particular nature of the elements appearing in this position.4 This approach will receive support from the behaviour of pronominal pivots in Romance and Germanic languages, which can override the general cross-linguistic ban against pronouns in existentials through a noteworthiness reading on pronouns. Since this ‘rescuing mechanism’ is available across the board, regardless of the fact that the languages respects or violates the DE, the 4

After the completion of this chapter Manuel Leonetti made me aware of Leonetti (2014), which deals with similar problems in Spanish existentials, so I could not consider his insightful approach here.

180

Chapter Five

conclusion will be that the behaviour of pronouns in existentials cannot be subsumed under such a principle unless we rethink its nature. The structure of the article will be the following. First, I will consider the different approaches to the DE, showing they haven’t accounted for pronominal pivots. Then, in section 3, I will concentrate on the paradigm of pronominal pivots to determine their exact distribution. Afterwards, in section 4, I will offer a solution based on the nature of strong pronouns in Catalan (and Italian) and the creation of a set of contextually salient alternatives capable of generating a noteworthiness reading compatible with the referential restrictions imposed on existential pivots. Finally, the main conclusions will be presented.

2. The Definiteness Effect in Existential Constructions Since the theoretical literature on the DE in existential constructions has grown steadily since the first exhaustive description of the phenomena by Milsark (1974), it would be impossible to summarize all the proposed solutions. Hence, due to space limitations, I will commit myself to offer a very quick overview of the main approaches, concentrating on the predictions they raise for pronouns.

2.1 Syntactic Solutions The first syntactic solution to the DE was proposed by Safir (1985, 1987), who resorted to the notion of case chains. In a nutshell, the expletive and the pivot form a chain, so that the case assigned to the expletive is transferred to the pivot. Yet, this amounts to a violation of principle C of binding theory, for the pivot is bound by the expletive. Safir argues that this is precisely what bans definite pivots. What about indefinite pivots? He claims that they are interpreted predicatively, not referentially, bypassing binding requirements. Hence, the DE is the result of a clash between case requirements and binding principle C (or B, in the case of pronouns, we should add). Leaving aside technical and theoretical problems, Safir’s solution cannot be correct on empirical grounds, as long as the full gamut of definite pivots is considered:5

5

The % symbol in McNally’s example indicates that the solution with even wasn’t admitted by all the speakers.

Catalan Existentials

(10)

a. b. c. d.

181

There were the same people at both conferences. (Prince 1992) There was the usual crowd at the beach. (Prince 1992) There was only/%even Kent available. (McNally 1992) Instead of a couple hundred other passengers, there was her, two flight attendants, a pilot, copilot and enough food to feed Rhode Island. (Francez 2007: 96)

All these examples also pose a hard problem to the syntactic solution in Moro (1997: 155-156), who builds on the assumption that the existential is a species of the locative construction and reduces the DE to the impossibility of strong determiners to appear isolated in the postcopular position at LF: (11)

a. a’. b. b’.

there are many/few/three girls girls are many/few/three *there are every/the/most girls * girls are every/the/most

For Moro, the noun girls moves to the preverbal subject position in both cases, stranding the determiner, but the movement is overt in the case of locatives (11a’-b’) and covert in the case of existentials (11a-b). Since strong determiners cannot have an adjectival (predicative) reading, they are excluded in both cases. When moving to Italian, who violates the DE, but still cannot strand strong determiners with locatives (12), Moro assumes that the difference has to do with the pro-drop parameter: Italian definite pivots appear right-adjoined to the small clause, just as inverted subjects in general. (12)

a. b.

le ragazze the.F.PL girls *le ragazze the.F.PL girls maggior biggest

sono are sono are parte part

molte/poche/tre many.F.PL/few.F.PL/three) ogni/le/la every/the.F.PL/the.F

Leaving aside the fact that (12a) is a case of determiner doubling, which Moro leaves uncommented, it is unclear that this solution works in general, not even for the examples that he examines. Moro doesn’t consider the

182

Chapter Five

existential alternative to (11b), which we must assume that it is fine, but rather a case with a proper name: C’è Gianni (in questo giardino) lit. ‘there is John (in this garden)’. It is unclear to me how the adjectival nature of the determiner may play any role in such cases… without any determiner! In any event, just as Safir’s, Moro’s analysis is too restrictive, and, for our purposes, gives no answer to the impossibility of pronominal pivots in languages like Italian: if the syntactic strategy is available for definite DPs and proper nouns, why should it not be for pronouns? The bottom line is that purely syntactic solutions are at pains with the subtle intra- and crosslinguistical variation, which is not amenable to mainstream all-or-nothing syntactic principles.

2.2 Universal Definiteness Hierarchies In part as reaction to the cross-linguistic variation involved in the phenomenon of DE, several authors have attempted to derive it from a combination of a universal hierarchy of definiteness and/or animacy and the particular typological features of each language and/or construction. (13)

Definiteness scale (Aissen 2003) Personal pronoun > Proper name > Definite NP > Indefinite specific NP > Non-specific NP

(14)

Extended animacy Hierarchy (Dixon 1979: 85 in Croft 1990: 130) first/second person pronouns > third person pronoun > proper names > human common noun > nonhuman animate common noun > inanimate common noun

These hierarchies have proved very insightful for treating a wide range of phenomena (differential object marking, clitic doubling or object agreement, among others), including the DE in existentials, as in Ramos (1998: 43), who offers the following version: (15)

Definiteness Hierarchy (Ramos 1998: 43) 1, 2 > 3 > tots (‘all’) > proper noun > common definite animate noun > common definite inanimate noun > common indefinite animate noun > common indefinite inanimate noun > common indefinite animate noun with no determiner > common indefinite inanimate noun with no determiner > abstract noun with no determiner.

Catalan Existentials

183

From this hierarchy, Ramos argues, languages fix different ‘tolerance thresholds’ for each construction, which would explain cross-linguistic variation. For example, languages like Catalan or Italian would have a more tolerant threshold for existential verbs (proper nouns) than languages like English or Spanish (between definites and indefinites). This proposal, however, is problematic on theoretical and empirical grounds. The first clear shortcoming is the very nature of the hierarchy, which is formed on a very polysemic concept as ‘definiteness’ (the same critique is valid for the syntactic approaches discussed in 2.1). Are we talking about grammatical definiteness, as one would primarily expect, or rather about semantic definiteness? Moreover, if some notion of ‘semantic definiteness’ is intended, it is certainly surprising on purely conceptual grounds to consider pronouns, which are entities corresponding to logical variables, as more definite than proper nouns, which are rigid designators, as the semantic and philosophic literature remarks (Kripke 1980, Soames 2002). Besides the theoretical doubts just expressed, Ramos’ proposal does not seem to be rooted on solid empirical ground, neither. As I will discuss in some detail in section 3, the threshold seems movable not only for different constructions in different languages, but even for the same construction in the same language. However, one insurmountable shortcoming is that this hierarchy leaves without explanation the robust contrast between weak and strong determiners in existentials (Milsark 1977, Barwise & Cooper 1981, Keenan 1987, 2003), for this distinction is not based on definiteness/animacy, but rather on logical properties of determiners. Here is the list offered by Keenan (2003: 188-189): (16)

a. b.

e. f. g. h.

Aren’t there at least three students in your logic class? There is at most one undergraduate student who objects to that. Are there more than five students in the course? Aren’t there nearly fifty students giving talks at the conference? Were there no students at the party? Just how many students were there at the demonstration? There are exactly two students enrolled in the course. There are fewer than five fountains in the garden.

a. b. c.

*There are most students in my logic class. *Isn’t there the student who objects to that? ??Are there two out of three students in the course?

c. d.

(17)

Chapter Five

184

d. e. f. g. h.

*Isn’t there every student giving a talk at the conference? *Was there neither student at the party? *Which of the two students were there at the demonstration? *There are both students enrolled in the course. *There are all but two fountains in the garden.

One can easily appreciate that the definiteness hierarchy cannot offer a rationale for the contrast between the weak determiners in (16) and the strong ones in (17). A more promising approach is the one advanced in Beaver et al. (2006) (see also Aissen 1999 and Bentley 2013). From a cross-linguistic study involving English, Hebrew, Russian and Dutch, these authors defend that the DE depends on the subjecthood semantic and pragmatic properties of pivots: there is an inverse correlation between the presence of an NP as a pivot and its frequency as the subject of its copular version. This leads them to propose the following hierarchy: (18)

Local pro > non-local pro > proportional NPs > definite descriptions, demonstratives > prototypical indefinites > Ļmon NPs

The semantic distinction between NPs solves the problem just noted regarding Ramos’ (1998) proposal, but the proposal still suffers from two main drawbacks. On the one hand, as the authors acknowledge, the universality of the ranking is grounded on the assumption that existentials are non-canonical correlates of locative copular sentences (Safir 1985, Hoekstra & Mulder 1990, Freeze 1992, Moro 1997), which has been severely criticized by many authors (Barwise & Cooper 1981, Williams 1984, McNally 1992, Francez 2007), as empirically and theoretically inadequate. On the other hand, none of the languages considered is a wholesale null subject language, a fact that will prove important in section 4: English and Dutch are not, and Hebrew and Russian are at most partial pro-drop languages. In any event, the ideas in Beaver et al. (2006) will be reconsidered under a different light in section 4, when the features of pronouns are described in more detail.

Catalan Existentials

185

2.3 Semantic Solutions Milsark (1974) is the original source of the classical classification of determiners allowed in the pivot into weak and strong determiners. Yet, his classification, based on the behaviour of DPs in existential sentences wasn’t grounded on independent properties, which yield a circularity issue when dealing with the DE. The coherent move was considering the logical properties of determiners with the theory of Generalized Quantifiers placed in the forefront (see Barwise & Cooper 1981, Keenan & Stavi 1986, Keenan 1987 and Zucchi 1995).6 Several logical properties have been proposed, but symmetry has been a good candidate for grounding Milsark’s generalization:7 (19)

a. b.

A determiner D is symmetric if DAB = DBA. Definiteness effect: the pivot of an existential sentence must be headed by a symmetric determiner.

The basic idea is that the two properties related by the determiner, the noun (A in the definition) and the VP (B in the definition) can be reordered without altering the truth conditions of the sentence. For instance, the indefinite determiner a is symmetric: (20)

A man smokes. = A smoker is a man.

In contrast, the determiner every is asymmetric: (21)

Every man smokes.  Every smoker is a man.

Crucially, this property derives the set of determiners allowed as pivots of existentials (for space limitations, I leave the application of the test to the reader): (22)

6

Symmetric determiners: a. Aren’t there at least three students in your logic class? (Keenan 2003) b. Were there no students at the party? (Keenan, 2003)

Even though formally semantic, Barwise & Cooper (1981) and Zucchi (1995) crucially rely on pragmatic conditions to account for the DE. See McNally (2011) for discussion. 7 This is a very simplified exposition of the rather technical analyses developed by Barwise & Cooper (1981), Keenan & Stavi (1986) and Keenan (2003).

Chapter Five

186

c. d.

(23)

Just how many students were there at the demonstration? (Keenan 2003) There is some drummer I know in every rock band. (Francez 2007)

Asymmetric determiners: a. *There are most students in my logic class. (Keenan 2003) b. *Isn’t there the student who objects to that? (Keenan 2003) c. *Isn’t there every student giving a talk at the conference? (Keenan, 2003) d. *There are both students enrolled in the course. (Keenan 2003)

Even though this is a huge improvement over the initial ideas by Milsark (1974) and Barwise & Cooper (1981), this approach is doomed to fail as a full encompassing principle. For instance, it must leave out the following cases from Keenan (2003: fn. 1): List contexts: (24)

A: How do I get to UCLA from here? B: Well there’s always the bus, but it doesn’t run very often.

“Presentative” superlatives: (25)

Wow! There’s the biggest dog I’ve ever seen in the yard!

“Reminder contexts”: (26)

Well, yes, there’s always that.

Then, semantic analyses are just as restrictive as syntactic ones, asking for a typology of existential constructions with different requirements regarding the DE, and/or weakening its effects by additional pragmatic principles.

Catalan Existentials

187

2.4 Pragmatic Solutions Given the inability of syntactic (2.1) and semantic solutions (2.3) to offer a unified solution to the existent intra- and cross-linguistic variation regarding the DE, and given the quite sketchy answers suggested by accounts based on universal scales (2.2), several pragmatic approaches have flourished in the literature, offering the best characterization of the phenomenon to date. Most of them are based on Prince (1992) original insight that pivots of existential sentences are hearer-new, namely that the referent introduced by the pivot is assumed by the speaker not to belong in the hearer’s knowledge store (see Ward & Birner 1995: 728 for a reformulation and detailed analysis). Indeed, Prince noted that the following definite pivots are fine because they are hearer-new: (27)

a. b. c.

There were the same people at both conferences. There was the usual crowd at the beach. There was the stupidest article on the reading list.

Yet, two basic sorts of empirical problems are found in the literature against the hearer-new approach. On the one hand, McNally (1992) argues that some pivots don’t even introduce an entity, as the impossibility of cross-sentential anaphora shows: (28)

There is no assistant available. #He is from Hamburg.

On the other hand, as pointed out by Francez (2007: 111), pivots in Hebrew existentials can be clearly hearer-old pronouns (several cases of hearer-old pivots are also described in Abbott 1992, 1997, and criticised in Ward & Birner 1997):8 (29)

8

A: B:

‘Why did you go to MIT if you don’t like Chomsky?’ lo yadati Se-yeS oto Sam. NEG know[1.S] that-EX him there ‘I didn’t know he was there.’ (Lit: I didn’t know there is him there.)

Ward & Birner (1995, 1997) argue that these cases are hearer-old, but they are used as if they were hearer-new by the speaker. However, such a solution is highly problematic, for it introduces a high degree of indeterminacy in the interpretation of these sentences, which should be testable experimentally.

Chapter Five

188

These two pieces of evidence have pushed pragmatic approaches into two directions. On the one hand, some scholars have refined the original condition on hearer-newness to cover the conflictive cases. This is the case of McNally’s (1992) Felicity Condition, which is supplemented with certain ad hoc amendments (nf stands for ‘nominalized function’): (30)

Felicity condition (McNally 1992: 120) The discourse referent corresponding to the instantiation of the nf-argument of the existential predicate must be novel.

Crucially, as she discusses, this pragmatic condition can be defeated in certain circumstances, for instance when the truth of the whole existential sentence is presupposed, as in (31)

a. b.

There was *(only) Kent available. Among those in the cage, there was *(only) the brown dog barking.

Why should only make a difference regarding the Felicity Condition? As remarked by Rooth (1985: 120), only contributes with an assertion of uniqueness (“Kent, but nobody else, was available”) and a conventional implicature (many would say ‘a presupposition’, see Roberts 2011 for discussion) that the proposition “Kent was available” is true. Hence, we are in one of those cases that, according to McNally, the novelty requirement on the pivot is dispensed with. McNally compares only with even, which contributes with the assertion that “Ken was available” and with a conventional implicature that “everybody was available and Ken was the least unlikely to be available” (Rooth 1985: 120): (32)

a. b.

*There was even Kent available. *Among those in the cage, there was even the brown dog barking.

Leaving aside its ad hoc flavour, this solution does not seem empirically adequate, for we do find examples with definites (and pronominal!) pivots with even or also, which do not presuppose the truth of the existential proposition (I will turn back to these examples in 4.3):

Catalan Existentials

(33)

a.

b. c.

189

He was doing that at the expense of String Theory and Lego fun time. And there was even him tolerating harp music in the midst of it all. Hell, there was even him to consider! Besides me being a coward there is also HIM stopping me.

The second development of the pragmatic approach involves a finergrained analysis of the topic-focus articulation of existential sentences (Abbott 1993, 1997, Kim 1997, Lambrecht 2000, Francez 2007, Leonetti 2008, Cruschina 2012). The idea is clearly expressed by Abbott (1993: 41), who considers that the main task of existential constructions is “to draw the addressee’s attention to the existence and/or location of the entity or entities denoted by the focus NP”. Then, definites, proper nouns or pronouns, which are highly presuppositional, will be particularly ill-suited for this task, resulting in the DE. This idea is pursued by Leonetti (2008), who derives the DE from the fact that existentials impose very specific pragmatic and informational restrictions on the pivot, which result in the mixed and complex patterns identified as the DE. Namely it is within the limits of the existential construction that the referential properties of the DP are to be established, which is in line with the Novelty Condition –moreover, I would claim that the Novelty Condition actually is a result of this requirement, since it bans access to given information and thus to anaphoric readings. […] When definite expressions rely on contextual information for their interpretation, thus being assigned a deictic or an anaphoric reading, they are typically excluded from proper existentials (for instance, pronouns); in case they satisfy the uniqueness condition by means of linguistic information that is accessible in the sentence, in principle they should not be excluded from existentials (it is the case of “weak” definites and “cataphoric” definites, where definiteness is solved inside the DP). Leonetti 2008: 24 [my emphasis, XV]

The idea is that the DE is the sum of several factors involving the particular requirements of the predicate and the pragmatic and informational properties of the pivot. This would allow us to offer a better account of the very complex pattern of the DE intra- and crosslinguistically. Now, to deal with the well-known counterexamples to the DE, while maintaining it as a basic feature of existential constructions, two main solutions are found in the literature. On the one hand, one can argue that the definite pivot is not a wholesale semantic definite (for instance, a “weak definite”; see Abbott 1999, Lyons 1999). On the other hand, one can deny that the cases violating the DE are proper existential

Chapter Five

190

constructions (Leonetti 2008, Cruschina 2012, this volume). Indeed, Leonetti restricts the DE violations to enumerative constructions, which involve a list-reading (34), or to eventive existentials (35) (I adapt the glosses to the Leipzig Glossing Rules): (34)

a.

b.

(35)

a.

C’è Gianni. loc-is John Italian: Cruschina (2012) ‘John is here.’ Il y a Jean, Marie et un ami. There have Jean Marie and a friend French: Dobrovie-Sorin & Beyssade (2012) ‘There are Jean, Marie and a friend.’ C’è

Gianni infuriato / nei guai. John furious in-the trouble Italian: Cruschina (2012) ‘John is furious / in trouble.’ Il y a le facteur it LOC has the mailman qui arrive. who arrives French: Dobrovie-Sorin & Beyssade (2012) ‘There is the mailman coming.’ Hi ha la Maria passejant-se LOC has the.F Mary walking-REFL pel jardí. by-the garden Catalan: Villalba (2013) ‘Mary is walking by the garden.’ LOC-is

b.

c.

For Leonetti (2008), the sentences in (34) are not proper existentials; hence the DE is not applicable here. As for the eventive existentials in (35), he argues that they are associated with a small clause as part of a broad focus, but crucially the pivot and the coda within the small clause display an internal topic-focus articulation, where the pivot is interpreted as a topic of the focus coda predicate (see Villalba 2013 for a critique of Leonetti’s 2008 analysis). Accordingly, he argues, since for the DE to be

Catalan Existentials

191

operative the pivot must be part of a wide focus, it is not operative in this case.9 To sum up, it seems quite reasonable that only a pragmatic approach along the lines defended by Leonetti (2008) can handle the wide range of intra- and cross-linguistic variation regarding the DE. Yet, we must still consider how pronominal pivots fit into this picture.

3. Pronominal Pivots in Catalan and Beyond We have just seen that none of the approaches to the DE in existentials fully took into account pronominal pivots: pronouns were simply neglected or they were subsumed without discussion under the treatment for definite DPs. In order to see the real impact of pronominal pivots on DE accounts, let us review in some detail their distribution in existential sentences. The first important fact that must be considered is the general ban against pronominal pivots both in languages respecting the DE, like English or Spanish, and in languages where the DE is violated generally (but see Leonetti 2008, Cruschina 2012, this volume, and Villalba 2013 for qualifications): (36)

(37)

a. b.

*There is the boy/Peter. Hi ha el noi/ LOC has the boy ‘The boy/Peter is there.’ (Catalan: Rigau 1988a: ex. 2b-c)

a. b.

*There is him. *Hi ha ell. LOC has him ‘He is here.’ (Catalan: Rigau 1988a: ex. 2d)

en the

Pere. Peter

One could argue that this asymmetry follows from the higher position of pronouns in the definiteness hierarchy (see 2.2): pronouns are banned across the board in pivot position, but Catalan has a more permissive threshold regarding definite DPs and proper names than English. However, on closer inspection, one finds that pronominal pivots are possible in Catalan existentials under special circumstances, as I have shown in section 1. This evidence, which has been neglected in the relevant literature, can be classified into at least three main groups. First, 9

I am thankful to Manuel Leonetti for helping me clarifying this point.

Chapter Five

192

one can find a big number of examples where the pronoun is associated to the focal particle només ‘only’ (see also Beaver et al. 2006: 32 for English only): (38)

a.

b.

c.

en aquella central només hi havia jo in that.F store only LOC had I ‘in that store it was only me’ Les aventures i desventures d'un carregador de mòbil

I només hi havia ella que el and only LOC had she that him pogués aturar. could stop ‘And only her was there to stop him.’ (Sebastià Juan Arbó, Hores en blanc. p. 64) d'empleats només hi havia ella i of-employees only LOC had she and una auxiliar. a.F assistant ‘as for employees, it was her and an assistant’ (Rodalies de Catalunya: molta feina per fer. El Punt Avui, 10/11/2010)

Second, many examples are found where the pronoun is reinforced by mateix ‘oneself’ (lit. ‘same’): (39)

a.

b.

al president li han to-the president to.him have.3PL començat a créixer enemics began to grow enemies –entre els quals hi among the.PL which.PL LOC ha ell mateix–. has he same (Josep M. Espinàs, Una història poc exemplar. El Periódico. 9/07/2008) uns quants, entre els quals a.PL some.PL among the.PL which.PL hi havia, jo mateix, […] LOC had I same (No volem tombes. Revista de Cambrils. 07/1997)

Catalan Existentials

c.

193

els disset consellers, entre the.PL seventeen delegates among els quals hi ha ell mateix the.PL which.PL LOC has he same (Associació de promotors de Barcelona. Caixa Girona condiciona la fusió a un consell territorial http://apcebcn.cat/content/index.php?option=com_conte nt&task=view&id=18629&Itemid=9)

The third group includes a pronoun coordinated with a definite DP or a proper noun: (40)

a.

b.

c.

pels funcionaris en vaga, entre by-the.PL civil.servants on strike among els quals hi havia jo the.F.PL which.PL LOC had I i alguns dels que sou aquí. and some.PL of-the.PLthat are.2PL here ‘by the civil servants on strike, among which I was and some of those you are here.’ (Pasqual Maragall. Solemne signatura de l’acord per a un govern catalanista i d’esquerres a la Generalitat de Catalunya. http://www.unitatdaran.org/2003/12/solemne-signaturadelacord-per-a-un-govern-catalanista-idesquerres-a-la-generalitatde-catalunya/) unes quantes dones, entre les a.F.PL some.F.PL women among the.F.PL quals hi havia ella i which.PL LOC had she and la Victòria. the.F Victory ‘few women, which she and Victoria were among.’ (M. Carme Rodríguez Virgili Benvinguda, Marta, p. 180) un sopar on hi havia ell a dinner on LOC had he i la seua dona, […] and the.F his.F wife ‘a dinner which attended he and his wife’ (Ramon Lapiedra, La raó cívica.p. 49)

Chapter Five

194

Interestingly, these three patterns are also found in a language respecting the DE, like English, as in the examples already mentioned in section 1: (41)

a.

b.

c.

Behind him, they finished by shooting the already-dead Stebbins, and now there was only him, alone on the road, walking toward where the Major's jeep had stopped diagonally across the white line... (Stephen King, The long walk) Only Mikhail was sanity. He was on her mind continually, invading her thoughts, pushing out everything insane until there was only him. (Christine Feehan, Dark prince) “So I saw that there was only me. There was only me who could worry about what was happening here, inside these walls of my life. (Markus Zusak, Underdog)

Leaving aside the inability of syntactic or semantic solutions to offer an explanation for the complex distribution of pronominal pivots, the definiteness hierarchy is also at pains with these data, unless a clear pragmatic notion of definiteness is developed. Note as well that even the constructional differentiation developed by Leonetti (2008) and Cruschina (2012) runs into problems. For instance, if we follow the rationale of Leonetti’s proposal that the DE is not operating in presentational and eventive existentials, one would expect pronominal pivots to be fairly common in both cases. Yet, as I will discuss at length in the next section, this is quite dubious on empirical grounds. On the one hand, even though pronouns are quite normal in list readings, they typically appear in coordination structures in all the works dealing with the phenomenon (my italics): (42)

a. B. b.

(43)

A: B:

A. My God! How many people know about this? There’s me and there’s you. That’s all. (Rando & Napoli 1978: ex. 37) OK, let’s finish up this guest list. There’s you and me. Who else is coming? (Abbott 1997: 6) I don’t have any friends. Oh, don’t be silly! There’s John and me and Susan and Peggy… (Rando & Napoli 1978: fn. 9)

This is also true of Catalan:

Catalan Existentials

(44)

a.

Hi

havia had company mate

LOC

ella *( she de of

i and pis). house

195

el the

However, proper names do not abide to this restriction, as we have just seen. On the other hand, pronominal pivots are clearly ungrammatical in eventive existentials, against Leonetti’s prediction: (45)

*Hi LOC

ha ella passejant-se has her walking-REFL Catalan; Villalba (2013: 39b)

pel by-the

jardí. garden ‘She is walking by the garden.’

Since the pronoun is excluded in principle by the DE, we are dealing with an existential sentence. Hence, the addition of the coordinate or of the focal adverb només ‘only’ could not alter this situation, against fact. In conclusion, the defendants of the constructional approach must accept that some instances of violation of the DE do affect existential sentences, calling for a specific solution, and weakening the generality of their approach. To sum up, the issue raised by pronouns in Catalan and Italian existentials is the following: since they are languages systematically violating the DE, a solution based on a “general-purpose DE” is untenable, be it syntactic, typological, semantic or pragmatic. Rather, I will argue for a different line of attack, arguing that the DE is not a unitary phenomenon, but the conjunction of several pragmatic requirements involving the informational conditions imposed on pivots by the existential predicate and the particular nature of the elements appearing in this position. This pragmatic approach will receive support from the behaviour of pronominal pivots in Germanic and Romance languages, which can override the general ban against pronouns in existentials when a contrast reading is made salient. Since this ‘rescuing mechanism’ is available in languages like English or Spanish, which respect the DE, and in languages like Catalan or Italian, which violate the DE, the conclusion will be that the behaviour of pronouns in existentials cannot be subsumed under such a generalization.

196

Chapter Five

4. Pronouns and the DE The solution I will defend for the special behaviour of pronominal pivots is grounded on two pillars. On the one hand, I will take as essentially correct the pragmatic view as defended in Leonetti (2008) that existentials impose very specific pragmatic and informational restrictions on the pivot, which result in the mixed and complex patterns identified as the DE. Namely it is within the limits of the existential construction that the referential properties of the DP are to be established, which is in line with the Novelty Condition –moreover, I would claim that the Novelty Condition actually is a result of this requirement, since it bans access to given information and thus to anaphoric readings. […] When definite expressions rely on contextual information for their interpretation, thus being assigned a deictic or an anaphoric reading, they are typically excluded from proper existentials (for instance, pronouns); in case they satisfy the uniqueness condition by means of linguistic information that is accessible in the sentence, in principle they should not be excluded from existentials (it is the case of “weak” definites and “cataphoric” definites, where definiteness is solved inside the DP). (Leonetti 2008: 24 [my emphasis, XV])

On the second hand, I will argue that regardless of its inherent unsuitability as pivots, pronouns may fit this position when obtaining a contrastive interpretation. Yet, since pronouns cannot get this interpretation for free in this position, because they cannot alternate between a weak/clitic and a strong version, they could only be rescued as pivots when the contrastive reading is obtained by ‘external means’. I will show that (at least) two mechanisms are common for fulfilling this task: the focus particle only and the reinforcer mateix ‘oneself’ (and maybe coordination). In both cases, the pronoun gets a noteworthiness reading, which makes it a suitable pivot. However, in order to present my proposal, a brief explanation is needed of the contrastive nature of pronouns in null subject languages.

4.1 A neo-Gricean View of Contrastive Pronouns Montalbetti (1984) originally pointed out that in Spanish a difference existed between null and strong pronouns regarding variable binding: only the former allowed a bound variable reading (on pronouns in Romance null subject languages, see Luján 1985, Alonso-Ovalle, Fernández-Solera, Frazier & Clifton 2002, Fernández-Soriano 1989, Mayol 2010, Picallo 1994, Rigau 1988b). Hence, the following contrast arises:

Catalan Existentials

(46)

197

Nadiei sabe que proi/éli vendrá. nobody knows that pro/he come.FUT.3SG ‘Nobodyi knows that hei will come’.

Yet, as Montalbetti himself noted, when the strong pronoun is the only option, the bound reading becomes possible: (47)

Ningún miembro del consejoi sabe que no membre of-the council knows that la reunión puede comenzar sin éli. the.F meeting can begin without him ‘[No member of the council]i knows that the meeting can begin without himi’. Rigau (1988b)

Luján (1985) also noticed that a similar fact happened when the pronoun got a contrastive reading, by means of stress or the addition of focal particles as sólo ‘only’: (48)

Ningún estudiantei piensa que (sólo) no student believes that (only) ÉLi/j pasó el examen. he passed.3SG the exam ‘No studenti believes that only hei/j passed the exam’.

If we lexically encoded contrast in Romance overt pronouns, this behaviour could only follow from the existence of two series of personal pronouns: the contrastive ones and the non-contrastive ones. Yet this situation is exactly the same kind of duplicity that Grice (1989) discussed regarding exclusive or. Grice analyses sentences like (49), where a supplementary meaning is attached to the pure semantic value of the disjunctive connective: (49)

The prize is either in the garden or in the attic (but not in both places).

While the basic logical meaning of the disjunction renders the sentence true whenever one of the two coordinates is true (hence, even in the case where both coordinates are true), speakers systematically strengthen this meaning to an exclusive disjunction, as the continuation between parentheses makes explicit. Grice sees this duplicity of meaning as real, but not rooted on lexical items: only the ‘weaker’ inclusive meaning (A or

198

Chapter Five

B or both) is attached to the disjunction. The ‘stronger’ exclusive meaning (A or B, but not both) is generated as a conversational implicature, and it is, hence, pragmatic.10 The pragmatic strategy is simple, but powerful: the speaker utters a disjunction, which in principle allows the inclusive “both A and B”, but if she intended this particular meaning, she had the clearer alternative of uttering a conjunction (A and B) with this particular meaning. Hence, since the utterer’s choice was a less informative option, the listener is prompted to interpret that the speaker didn’t intend the stronger meaning, generating the implicature that “not both”. We can extend Grice’s strategy to pronouns, following the seminal work by Levinson (1987), and Huang (1991, 1994), and recent experimental research by Kaiser (2011) (see Mayol & Clark 2010 and Mayol 2010 for similar ideas couched in the framework of Game Theory). As discussed at length by Huang (1994: 6.5) regarding similar cases in Chinese, we can explain the contrastive reading in neo-Gricean terms as follows. The null/clitic/weak pronoun is the default choice as long as Levinson’s (1987, 2000) Informativeness principle is considered:11 (50)

The I(nformativeness)-principle 'Say as little as necessary', i.e. produce the minimal linguistic information sufficient to achieve your communicational ends,

In the case at hand, this ensures that for purely referential reasons, a null or weak pronoun will be preferred over a strong one. Moreover, the use of a strong pronoun would enter in conflict with another principle, namely the Manner principle, a development of Grice’s Maxim of Manner: (51)

The M(anner)-principle Do not use a prolix, obscure or marked expression without reason.

Hence, if an unmarked form (a null or weak pronoun) is available, the use of a marked one (a strong pronoun) will lead the hearer to infer that the speaker has some special communicative intention, for instance a contrastive disjoint interpretation, as in the previous examples. 10

This is a consequence of Grice’s famous Modified Occam’s Razor Principle Grice (1989: 47): Senses are not to be multiplied beyond necessity. He applies the same strategy to the different meanings of and or numerals. 11 Levinson’s I-principle is loosely related to Grice’s Maxim of Quantity (Grice 1989), and certainly merits the neo-Gricean label.

Catalan Existentials

199

As one can easily appreciate, the neo-Gricean approach correctly predicts that this contrastive implicature associated to the strong pronoun will be absent if strong pronouns are the only option: it is, thus, associated to the choice of the strong pronoun instead of the null/weak alternative. We have just seen that this is exactly what happens in cases like (47)-(48), and—I will argue in the next subsection—in existential constructions as well. To sum up, one can conclude that the contrastive nature of pronouns in Catalan (and other null-subject Romance languages) is not necessarily a lexical feature, but can rather be analysed as a conversational implicature following from the choice between alternatives, in a similar vein to scalar implicatures.

4.2 Back to Existentials and Pronouns How can we relate with existential sentences the neo-Gricean view on contrast in the pronominal system just sketched? Catalan grammarians have noted the surprising incompatibility between existentials and definite pronominal clitics:12 (52)

En in teva your.F *No, not

aquella festa that.F party germana? sister no la not her

també also

hi LOC

hi LOC

havia. had

havia had

la the.F

(Ramos 1998)

Indeed, Catalan tends to treat the internal argument of existentials (and impersonals in general) as the subject of an unaccusative verb, rather than 12

Languages like Spanish or Galician do admit pronominalisation of the pivot by a definite clitic: (i) Estudiantes, los hay bastante buenos. fairly good.PL students them have.3SG(LOC) ‘Students, there are fairly good ones.’ Yet, as discussed by Longa et al. (1998), the pivot is always indefinite and the use of the definite clitic is a ‘recycling’ mechanism to supply the lack of a partitive clitic en/ne ‘of it’. Compare (i) with its Catalan equivalent, where the partitive clitic is in force: (ii) D’estudiants, n’hi ha de força bons. fairly good.PL of-students of.it-LOC have.3SG of ‘Students, there are fairly good ones.’

Chapter Five

200

as an object, as the preference for agreement in most dialects show (see Rigau 1994, 1997 for dialectal differences in Catalan, and RodríguezMondoñedo 2006 for similar facts in Spanish): (53)

N’hi havien of.it-LOC had.3PL ‘There were two (of them).’

dos. two

So then, the only option available is a strong pronoun. According to the neo-Gricean approach defended in subsection 4.1, this raises the prediction that they will not be contrastive in pivot position. However, is this a problem? It is, if we consider Leonetti’s (2008) view that pivots cannot be anaphoric. We are then in the following position: on the one hand, Catalan pronominal pivots cannot be contrastive, for they cannot alternate with a null or clitic version; on the other hand, the pragmatic and informational requirements of pivots reject anaphoric elements. Hence, if nothing else happens, the prediction is that pronouns will be banned altogether from pivot position, which is correct to a great extent. However, my claim in this paper is that we do have some mechanisms to break this vicious circle and provide pronouns with a stronger referential interpretation, which will allow them to become better pivots. These mechanisms have been reviewed in section 3: focus particles like only, reinforcers like oneself and (possibly) coordination. Precisely, the role of només ‘only’ is particularly prominent in allowing pronominal pivots, as many examples indicate (I repeated ex. (38) for the sake of reference): (54)

a.

b.

en aquella central només hi havia jo in that.F store only LOC had I ‘in that store it was only me’ (Les aventures i desventures d'un carregador de mòbil http://apliense.xtec.cat/prestatgeria/a8930043.../bookcon tents.php?) I només hi havia ella and only LOC had she que el pogués aturar. that him could stop ‘And only her was there to stop him.’ (Sebastià Juan Arbó, Hores en blanc. p. 64)

Catalan Existentials

c.

201

d'empleats només hi havia ella i of-employees only LOC had she and una auxiliar. a.F assistant ‘as for employees, it was her and an assistant’ (Rodalies de Catalunya: molta feina per fer. El Punt Avui, 10/11/2010)

Indeed, Beaver et al. (2006: 32) already quantified the effect of only for English existentials: “What we find is that the presence of only makes the NP substantially more existential, by a factor of approximately 10,000.” As is the case with other focus particles, només ‘only’ consists of an asserted part (e.g. in (54)a “nobody other than me was in that store”) and an implicated/presupposed part (e.g. in (54)a “I was in that store”) (see Rooth 1985: 120). Yet, as remarked by Beaver & Clark (2008: 252), this falls short in describing the contribution of only: “The presence of an expectation that something stronger than the prejacent [=the presupposed/ implicated part] is true is an essential part of the meaning of only.” They make their point with the following contrast: (55)

a. b.

I really expected a suite but only got a single room with 2 beds. #I really expected a single room with 2 beds but only got a suite.

In (55a), the prejacent “I got a single room with 2 beds” is evidently weaker than the expectation: “I’ll get a suite”, so only is fine. However, in (55b), it is the other way around: the prejacent “I got a suite” is stronger than the expectation “I’ll get a single room with 2 beds”, which makes the presence of only infelicitous. When we move to existentials and pronouns in (54), one can easily understand that a contrast is made salient between the referent of the pronoun and an expected bigger set of absent individuals. So then, even though the pronoun by itself could not obtain this contrastive reading, as predicted by our neo-Gricean approach to the distribution of pronouns, the presence of only contributes the required contrast for the pronominal pivot to be felicitous. Moreover it is important to note that the contrast must involve some kind of unexpectedness, for particles like almenys ‘at least’ or també ‘also’, which lack this meaning, do not license pronominal pivots. In a search through the Internet, I have found not a single example with

Chapter Five

202

almenys ‘at least’, and only one with també ‘also’, which has been judged as strange by all Catalan speakers consulted: (56)

On hi where LOC ella, Ginebra, she Genevie

ha Lancelot també hi ha has.3SG Lancelot also LOC has.3SG i l'altre, Artús. and the-other Arthur (Vicent Martines Els amants i els altres)

The case of fins i tot ‘even’ merits a comment, for no examples are found with this particle. Two possible factors might explain such a gap. On the one hand, as pointed out by Beaver & Clark (2008: 70-72), even allows an entailment from a weak expectation to a stronger one (it is, thus, upward monotone): (57)

a. b.

I really expected a single room with 2 beds but even got a suite. #I really expected a suite but even got a single room with 2 beds.

Since this entailment is the default one, the contrast regarding the initial expectation is less salient from a pragmatic point of view that in the case of only, which reverses the monotonicity: it allows an entailment from a stronger to a weaker expectation. On the other hand, the reinforcer mateix ‘oneself’ (lit. ‘same’) seems to fulfil this function, which might block the use of fins i tot ‘even’. Indeed, many examples of pronominal pivots are found with the reinforcer mateix ‘oneself’ (I repeat ex. (39) for the sake of reference): (58)

a.

al president li han to-the president to.him have.3PLa començat créixer enemics began to grow enemies –entre els quals hi ha among the.PL which.PL LOC has ell mateix–. he same (Josep M. Espinàs, Una història poc exemplar. El Periódico. 9/07/2008)

Catalan Existentials

b.

c.

203

uns a.PL havia, had

quants, entre els quals hi some.PL among the.PL which.PL LOC jo mateix, […] I same (No volem tombes. Revista de Cambrils. 07/1997) els disset consellers, entre the.PL seventeen delegates among els quals hi ha ell mateix the.PL which.PL LOC has he same (Associació de promotors de Barcelona. Caixa Girona condiciona la fusió a un consell territorial http://apcebcn.cat/?option=com_content&task=view&id =18629&Itemid=9)

The pragmatic role of this element is remarking the noteworthiness of the inclusion of the referent of the pronoun, in a way similar to even, but without the unlikeness component. Note, in this regard, that the sentences yield a presupposition that other people are included in the list, as the incompatibility with només ‘only’ shows: (59)

a.

b.

c.

#al president to-the president començat a began to –entre els among the.PL hi ha LOC has #uns quants, entre a.PL some.PL among només hi havia only LOC had #els disset the.PL seventeen els quals només the.PL which.PL only mateix same

li han to.him have.3PL créixer enemics grow enemies quals només which.PL only ell mateix–. he same els quals the.PL which.PL jo mateix, […] I same consellers, entre delegates among hi ha ell LOC has he

One can now see that the strong contrastive reading just described for només ‘only’ and mateix ‘oneself’ is associated with a noteworthiness

Chapter Five

204

interpretation, which helps us explain the particular behaviour of such pronominal pivots: they are possible, because they are not interpreted anaphorically, but rather as introducing noteworthy referents. This interpretation is quite similar to the one discussed by Ionin (2006) regarding English indefinite this: (60)

a. b.

“...A few years ago, there was this hippie, long-haired, slovenly. He confronted me...” (Prince 1981: 233) There is this man who lives upstairs from me who is driving me mad because he jumps rope at 2 a.m. every night. (Maclaran 1982: 85)

Ionin argues that in these sentences the demonstrative this does not carry its prototypical deictic meaning, but rather a specific one marking noteworthiness. Hence, even though it is clear that we are dealing here with hearer-old referents, the crucial point is that both cases involve a referent with a noteworthy feature known by the speaker, even though the exact identification of the referent is not necessary. We can consider that the highlighting provided to pronominal pivots by only and the reinforce mateix ‘oneself’ conveys noteworthiness, which makes them referentially strong enough to fulfil the requirements of the existential, just as in the case of the demonstrative this in (60). Finally, consider the case of coordinated DPs with the Internet examples repeated from (40): (61)

a.

pels funcionaris en vaga, entre by-the.PL civil.servants on strike among els quals hi havia jo the.F.PL which.PL LOC had I i alguns dels que sou aquí. and some.PL of-the.PLthat are.2PL here ‘by the civil servants on strike, among which I and some of those you are here were.’ (Pasqual Maragall. Solemne signatura de l’acord per a un govern catalanista i d’esquerres a la Generalitat de Catalunya. http://www.unitatdaran.org/2003/12/solemne-signaturade-lacord-per-a-un-govern-catalanista-i-desquerresa-la-generalitat-de-catalunya/)

Catalan Existentials

b.

c.

205

unes quantes dones, entre les a.F.PL some.F.PL women among the.F.PL quals hi havia ella i la Victòria. which.PL LOC had she and the.F Victory ‘few women, among which she and Victoria were.’ (M. Carme Rodríguez Virgili Benvinguda, Marta, p. 180) un sopar on hi havia ell a dinner on LOC had he i la seua dona, […] and the.F his.F wife ‘a dinner which attended he and his wife’ (Ramon Lapiedra, La raó cívica.p. 49)

It is unclear whether these cases should receive the very same treatment in terms of noteworthiness, or rather they should be discarded as enumerative/list existentials, predicted not to show the DE (see Leonetti 2008). While there seems to be a parallelism with the examples with reinforcers in that a group of individuals is highlighted against a previous set, I am not certain that the noteworthiness mechanism should be extended to this case. More research is needed. To sum up, I have argued that pronouns cannot get a contrastive reading in pivot position, but only a anaphoric one because they cannot alternate between a weak/clitic and a strong version. Hence, they are in principle banned as pivots, which cannot be anaphoric. As a consequence, pronouns could only be rescued as pivots when the contrastive reading is obtained by ‘external means’. I have shown that (at least) two mechanisms are particularly suited for this task: the focus particle only and the reinforcer mateix ‘oneself’ (and maybe coordination). In both cases, the pronoun gets a noteworthiness reading, which makes it a suitable pivot. If this explanation is on the right track, a similar effect is predicted to be found in languages respecting the DE, whenever the conditions are provided. I test this prediction in the following section.

4.3 Beyond Catalan The prediction that contrast will improve the felicity of pronominal pivots in languages respecting the DE is fulfilled, even though only in part, for reasons not yet understood. As far as English is concerned, we find similar noteworthiness patterns to the ones we found in Catalan, as the following examples from the Internet attest:

Chapter Five

206

(61)

a.

b.

c. d.

This is the true glory of Aton, Moses, and there are no other gods, because there was only he at first, and all that is or was or will be, he made. (Howard Fast,Moses: The Epic Story of His Rebellion in the Court of Egypt) “There was him and her,” said the small servant, “a sittin’ by the fire, and talking softly together.” (Dickens,The old curiosity shop) There was him and Arthur and the nephew of a Chicagoland gangster… (Duane Schwartz,Fetchenko) There was him and there was his son. And finally there was Mrs Mallory’s adversary, wasn’t there? (F. G. Cottam,The Magdalena Curse)

Even though an accurate corpus analysis should await future research, one gets the impression that only and coordination easily license pronominal pivots, whereas even is less common, in accordance with its weaker contrastive character (see subsection 4.2). In any event, all these cases fit into the picture I have developed: whatever the means, a contrast is built between the pronominal pivot and other members of a salient set, resulting in a noteworthiness reading. This is essentially what we find in the very scarce cases of pronominal pivots in Spanish found by a Google search: (62)

a.

b.

c.

Agapita, que sólo había ella con Agapita that only had.3SG she with ese nombre,… this name (La casa (II): La familia tradicional sanabresa. ) entonces sólo habíamos nosotros. then only had.1PL we (La verdad del amor, ch.12 ) Solo habían ellos dos. only had.3PL they two (Hoja de Lagomar.July 2006. )

Catalan Existentials

207

Spanish is much more restricted than Catalan, Italian or even English in allowing pronominal pivots, which are typically found with solo ‘only’, but less commonly with coordination or never with reinforcers like mismo ‘same’ (cf. Catalan mateix ‘oneself’) or incluso ‘even’ (cf. English even). However, it must be noted that examples exist with definites and proper names in precisely those contexts forbidden for pronouns, as the following examples from the Internet show: (63)

a.

b.

c.

Había solo los vinos más caros(?,) had only the.PL wines more expensive.PL en la tienda. in the.F store ‘There were only the most expensive wines in the store.’ (Leonetti 2008: ex. 34b) parecía que entre ellos había seemed.3SG that between they had.3SG la misma distancia the.F same.F distance ‘it seemed that the same distance separated them’ (Kathryn Ross, Viaje de placer, p. 64) En la celebración, es cierto, in the.F celebration is true había el mismo cansado entusiasmo had.3SG the same tired enthusiasm de todo viejo matrimonio of every old couple ‘The celebration, it is true, had the same tired enthusiasm of any old couple.’ (Sergio Delgado, El corazón de la manzana)

Even though a comparative corpus analysis is needed before drawing any firm conclusion, I have offered strong empirical evidence supporting the initial hypothesis, namely that the behavior of pronouns in existential sentences cannot be explained by standard uniform formulations of the DE, but rather as a confluence of different pragmatic factors, namely the impossibility of pronominal pivots to get a contrastive reading by themselves and the antianaphoric requirements of existential pivots.

208

Chapter Five

5. Conclusions In this chapter I have argued that the Definiteness Effect (DE) found in existential sentences cannot offer an explanation for the behavior of pronominal pivots. On the one hand, I have shown that none of the current analyses of existentials and the DE can explain the general ban against pronominal pivots in languages like Catalan or Italian, which can easily violate the DE, while offering an explanation for the possibility of licensing them in certain contexts, even in languages respecting the DE. My proposal builds on the special nature of existential pivots, which cannot be anaphoric, and the fact that pronouns cannot be contrastive in such a position, due to the lack of alternation with a null/weak counterpart. Hence, by default, pronominal pivots would be anaphoric, and generally forbidden unless their contrastive meaning is supplemented by other means, namely focus particles like only, reinforcers like oneself or coordination. The constrative reading is analyzed as an instance of noteworthiness, which provides the pivot with enough referential independence to fulfil the requirements of the existential construction. In such very particular cases, pronouns are correctly expected to appear in existentials, even in languages respecting the DE, like English or Spanish.

References Abbott, Barbara. 1992. Definiteness, existentials, and the 'list' interpretation. Paper presented at Proceedings of SALT II, Columbus, OH. —. 1993. A pragmatic account of the definiteness effect in existential sentences. Journal of Pragmatics 19: 39-55. —. 1997. Discussion note: Definiteness and existentials. Language 73: 103-8. —. 1999. Support for a unique theory of definites. In T. Matthews & D. Strolovitch (eds.). Proceedings from Semantics and Linguistic Theory IX, 1-15. Ithaca, New York: Cornell University. Aissen, Judith. 1999. Markedness and subject choice in Optimality Theory. Natural Language and Linguistic Theory 17: 673-711. —. 2003. Differential object marking: Iconicity vs. economy. Natural Language & Linguistic Theory 21: 435-483. Alonso-Ovalle, Luis, Susana Fernández-Solera, Lynn Frazier & Charles Clifton. 2002. Null vs. overt pronouns and the topic-focus articulation in Spanish. Italian Journal of Linguistics 14: 151-70.

Catalan Existentials

209

Barwise, K. Jon & Robin Cooper. 1981. Generalized quantifiers in natural language. Linguistics and Philosophy 4: 159-219. Beaver, David & Brady Clark. 2008. Sense and sensitivity. How focus determines meaning. Oxford: Wiley-Blackwell. Beaver, David, Itamar Francez & Daniel Levinson. 2006. Bad subject: (Non-) canonicality and NP distribution in existentials. In E. Georgala & J. Howell (eds.). Proceedings of Semantic and Linguistic Theory XV, 19-43. Cornell University Press. Bentley, Delia. 2004. Definiteness effects. Evidence from sardinian. Transactions of the Philological Society 102: 57-101. —. 2011. Sui costrutti esistenziali sardi. Effetti di definitezza, deissi, evidenzialita. Zeitschrift für romanische Philologie 127: 111-40. —. 2013. Subject canonicality and definiteness effects in Romance theresentences. Language 89: 675-712. Brucart, Josep M. & Gemma Rigau. 2002. La quantificació. In Joan Solà, Maria Rosa Lloret, Joan Mascaró & Manuel Pérez-Saldanya (eds.). Gramàtica del català contemporani, vol. 2, 1517-1589. Barcelona: Empúries. Brucart, Josep M. 2002. Els determinants. In Joan Solà, Maria Rosa Lloret, Joan Mascaró & Manuel Pérez Saldanya (eds.). Gramàtica del català contemporani, vol. 2, 1435-1516. Barcelona: Empúries. Croft, William. 1990. Typology and universals, Cambridge: Cambridge University Press. Cruschina, Silvio. 2012. Focus in existential sentences. In Internet celebration for Luigi Rizzi’s 60th birthday, 19-43. Siena. Dixon, Robert M. W. 1979. A note on Dyirbal ergativity. Chicago Linguistic Society. 90-91. Dobrovie-Sorin, Carmen & Claire Beyssade. 2012. Redefining indefinites. Dordrecht: Springer. Fernández-Soriano, Olga. 1989. Strong pronouns in null-subject languages and the avoid pronoun principle. MIT Working Papers in Linguistics. Francez, Itamar. 2007. Existential propositions. PhD dissertation, Stanford University. Freeze, Raymond. 1992. Existential and other Locatives. Language 68: 553-595. Grice, H. Paul. 1989. Studies in the way of words. Cambridge, Mass.: Harvard University Press. Hoekstra, Teun & René Mulder. 1990. Unergatives as copular verbs: Locational and existential predication. The Linguistic Review 7: 1-79. Huang, Yan. 1991. A neo-Gricean pragmatic theory of anaphora. Journal of Linguistics 27, 301-35.

210

Chapter Five

Huang, Yan. 1994. The syntax and pragmatics of anaphora (Cambridge studies in linguistics). Cambridge: Cambridge University Press. Ionin, Tania. 2006. This is definitely specific: Specificity and definiteness in article systems. Natural Language Semantics 14: 175-234. Kaiser, Elsi. 2011. Salience and contrast effects in reference resolution: The interpretation of Dutch pronouns and demonstratives. Language and Cognitive Processes 26: 1587-1624. Keenan, Edward L. & Jonathan Stavi. 1986. A semantic characterization of natural language determiners. Linguistics and Philosophy 9, 253326. Keenan, Edward L. 1987. A semantic definition of “indefinite NP”. In Eric Reuland & Alice G. B. ter Meulen (eds.). The representation of (in)definiteness, 286-317. Cambridge, Mass.: MIT Press. —. 2003. The definiteness effect: Semantics or pragmatics? Natural Language Semantics 11: 187-216. Kim, Yookyung. 1997. A situation semantic account of existential sentences. PhD dissertation, Stanford University. Kripke, Saul. 1980. Naming and necessity. Cambridge, Mass.: Harvard University Press. Lambrecht, Knud. 2000. When subjects behave like objects: An analysis of the merging of S and O in sentence focus constructions across languages. Studies in Language 24: 611-682. Leonetti, Manuel. 2008. Definiteness effects and the role of the coda in existential constructions. In Alex Klinge & Henrik Hoeg-Müler (eds.). On nominal determination, 131-62. Amsterdam/Philadelphia: John Benjamins. —. 2014. On contrastive readings in the interpretation of NPs/DPs. In Sofiana Chiriacescu (ed.). Proceedings of the VI NEREUS Workshop ‘Theoretical implications at the Syntax / Semantics interface in Romance’. Arbeitspapier 127. Fachbereich Sprachwissenschaft, 99116. Konstanz: Universität Konstanz. Levinson, Stephen C. 1987. Pragmatics and the grammar of anaphora. Journal of Linguistics 23: 379-434. —. 2000. Presumptive meanings. Cambridge, Mass.: MIT Press. Longa, Víctor M., Guillermo Lorenzo & Gemma Rigau. 1998. Subject clitics and clitic recycling: Locative sentences in some Iberian Romance languages. Journal of Linguistics 34: 125-164. Luján, Marta. 1985. Binding properties of overt pronouns in null pronominal languages. Chicago Linguistic Society 21: 123-43. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.

Catalan Existentials

211

Maclaran, Rose. 1982. The semantics and pragmatics of the English demonstratives. PhD dissertation, Cornell University. Mayol, Laia & Robin Clark. 2010. Pronouns in catalan: Games of partial information and the use of linguistic resources. Journal of Pragmatics 42: 781-99. Mayol, Laia. 2010. Contrastive pronouns in null-subject Romance languages. Lingua 120: 2497-2514. McNally, Louise. 1992. An interpretation for the English existential construction. PhD dissertation, University of California, Santa Cruz, CA. —. 2011. Existential sentences. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.). Semantics: An international handbook of natural language and meaning. vol. 2, 1829-1848. Berlin: de Gruyter. Milsark, G. 1974. Existential sentences in English. PhD dissertation: MIT. —. 1977. Toward an explanation of certain peculiarities of the existential construction in english. Linguistic Analysis 3: 1-29. Montalbetti, Mario. 1984. After binding. On the identification of pronouns. PhD dissertation, MIT. Moro, Andrea. 1997. The raising of predicates: Predicative noun phrases and the theory of clause structure. New York: Cambridge University Press. Picallo, Carme. 1994. Catalan possessive pronouns – the Avoid Pronoun Principle revisited. Natural Language & Linguistic Theory 12: 259299. Prince, Ellen F. 1981. Toward a taxonomy of given-new information. In Peter Cole (ed.). Radical pragmatics, 223-255. New York, New York: Academic Press. —. 1992. The ZPG letter: Subjects, definiteness, and information-status. In Sandra Thompson & William Mann (eds.). Discourse description: Diverse analyses of a fundraising text, 295-325. Philadelphia/ Amsterdam: John Benjamins. Ramos, Joan-Rafael. 1998. Presentational sentences in Catalan dialects. Catalan Working Papers in Linguistics 6: 41-58. Rando, Emily & Donna J. Napoli. 1978. Definites in there-sentences. Language 54: 300-313. Rigau, Gemma. 1988a. Els predicats no verbals i l’efecte d'especificitat. Estudi General 8: 51-64. —. 1988b. Strong pronouns. Linguistic Inquiry 19: 503-11.

212

Chapter Five

—. 1994. Catalan presentational sentences and the properties of Agr nodes. In Guglielmo Cinque (ed.). Paths towards Universal Grammar, 343359. Washington: Georgetown University Press. —. 1997. Locative sentences and related constructions in Catalan: esser / haver alternation. In Amaya Mendikoetxea & Myriam UribeEtxebarría (eds.). Theoretical issues at the morphology-syntax interface, 395-421. Universidad del País Vasco. Roberts, Craige. 2011. Only: A case study in projective meaning. The Baltic International Yearbook of Cognition, Logic and Communication 6: 1-59. Rodríguez-Mondoñedo, Miguel. 2006. Spanish existentials and other accusative constructions. In Cedric Boeckx (ed.). Minimalist Essays, 326-394. Amsterdam: John Benjamins. Rooth, Mats. 1985. Association with focus. PhD dissertation, University of Massachusetts, Amherst. Safir, Ken. 1985. Syntactic chains. Cambridge: Cambridge University Press. —. 1987. What explains the definiteness effect? In Eric Reuland & Alice G. B. ter Meulen (eds.). The representation of (in)definiteness, 71-97. Cambridge, Mass.: MIT Press. Soames, Scott. 2002. Beyond rigidity: The unfinished semantic agenda of naming and necessity. Oxford: Oxford University Press. Villalba, Xavier. 2013. Eventive existentials in Catalan and the topic-focus articulation. Rivista di Linguistica 25.1: 147-173. Ward, Gregory & Betty Birner. 1995. Definiteness and the English existential. Language 71: 722-42. —. 1997. Discussion note: Response to Abbott. Language 73: 109-112. Williams, Edwin. 1984. There-insertion. Linguistic Inquiry 15: 131-153. Zucchi, Sandro. 1995. The ingredients of definiteness and the definiteness effect. Natural Language Semantics 3: 33-78.

CHAPTER SIX EXISTENTIALS, POSSESSIVES AND DEFINITENESS IN SAMOYEDIC LANGUAGES1 BEÁTA WAGNER-NAGY

1. Introduction The aim of this paper is to discuss the definiteness effect in some languages without articles. The focus of the investigation is Nganasan (also known as Tavgy), but other Samoyedic languages such as Nenets (Yurak), Enets (Jenissei Samoyed) and Selkup (Ostyak Samoyed) will also be touched upon. For an overview of the Samoyedic languages, see Section 2 below. The phenomenon of the definiteness effect (DE) has been addressed in numerous works since Milsark (1977), but most of the studies investigate English or the Romance languages (cf. Leonetti 2008). The Uralic languages have not yet been studied with respect to this phenomenon. What the notion of the definiteness effect describes is that definite NPs cannot occur in certain syntactic positions, and the pivot of an existential sentence cannot be definite. Accordingly, sentence (1a) is not grammatical since the pivot element does not satisfy the definiteness restiction, while

1

I would like to thank all the people who made this research possible: Maria Brykina, Valentin Gusev, Jean-Luc Lambert, and Sándor Szeverényi, who shared with me their Nganasan field recordings and permitted me to work with them. Of course, my deepest gratitude goes to my Nganasan consultants in UsĢ-Avam and to my colleagues, who helped me organize the expeditions and also my fieldwork. My fieldwork was co-funded by OTKA (Hungarian Research Grant) and FWF (Austrian Research Grant). The present research was supported by the DFG (German Research Grant, grant no. HWA 3153-2/1). Finally, I would like to thank all those who commented on previous versions of this paper.

Chapter Six

214

sentence (1b) is grammatical since it does. (For more on the structure of existential sentences, see Section 3 below). (1)

a. *There is the yellow cat in the room. b. There is a yellow cat in the room.

Several authors have addressed the question of what can modify the pivot element. It has been a widely accepted view since Milsark’s (1977) study that not all DPs can occur in existential sentences. This means that some modifying elements can be categorized as indefinite, while others make the pivot element of the sentence unequivocally definite. The following Hungarian examples are a case in point: (2)

a. A

szobá-ban van room-LOC be.3SG {néhány / sok / Ø} kutya. Ø {some / many / dog ‘There are some/many dogs in the room.’ b. A szobá-ban van DET room-LOC be.3SG {a / ez a} kutya {DET / this DET} dog ‘The / this dog is in the room.’ DET

(p.k.)

(p.k.)

As sentence (2b) demonstrates, in Hungarian, not only falls the definite article but also the determiner ez ‘this’ under the definiteness restriction. In his investigation of English, Milsark (1977: 46) concludes that in English existential sentences universally quantified NP’s cannot occupy the position immediately following the copula, i.e. the position of the pivot element. On the basis of this, determiners can be categorized in two different groups: those that can occur in existential sentences belong to the weak category, whereas those that cannot, belong to the strong category (for Milsark’s classification, see Section 4 below). However, many authors (among others Milsark 1977, Holmback 1984, McNally 2011 etc.) observed that certain morphologically definite noun phrases can be semantically indefinite. An example provided by McNally (2011) illustrates this (3). (3)

There was the mother of a student waiting outside.

Existentials, Possessives and Definiteness in Samoyedic Languages

215

How the definiteness restriction is realized in Samoyedic languages will be demonstrated in detail in Section 3. At this point, I want to point out that these languages neither have definite nor indefinite articles – which makes the expression of definiteness much more complex and complicated than in languages that do have articles. One way, for instance, is through the non-possessive use of possessive suffixes, as is demonstrated in sentence (4) from Nganasan. Here the word mϷu ‘earth’ is marked for the 3SG possessive inflection, which, however, does not express real possessivity. In the Samoyedic languages, the possessive suffix occurs, for example, in sentences describing natural phenomena. (4)

mΩu-ðu sҨürü ƾilҨΩnu þii-mΩΩ earth-3SG snow.GEN under hide-PTCP.PASS.3SGVX ‘The earth is hidden under snow.’ [KNT, 1994] 2

Various authors, among others Lyons (1969), Clark (1978), and Freeze (1992), have pointed out that locative, possessive and existential sentences are closely interrelated with each other, such that possessive and existential sentences can be derived from locative sentences. Later on in this paper I will show how possessive, locative and existential sentences are related in Samoyedic languages, especially in Nganasan, and I will demonstrate the different possibilities of coding (in)definiteness according to the scale Milsark has proposed. Word order together with certain copula triggers the interpretation of existential sentences (all thetic) and, thus, of the DE. The paper is organized as follows. In Section 2 I provide an overview of Samoyedic languages and their typological classification and I briefly describe the database used for the present study. Section 3 discusses the properties of existential sentences, while Section 3.1 concentrates on the existential sentences of Samoyedic languages, with a special focus on copula use. Section 3.2 deals with locative sentences and the ways locative and existential sentences can be differentiated. Section 3.3 describes the characteristics of possessive sentences, demonstrating that possessive and existential sentences are closely interrelated in Samoyedic languages.

2

Abbreviation in glosses are according to Leipzig Glossing Rules with the additions of following items: AUG augmentative, CNG connegative, CO coaffix, CONV converb, DIM diminutive, EL elative, EMPH emphatic element, EP epenthetic vowel, EX existential, EXCL exclamative, INTER interrogative, NAR narrative PAR partitive, PX possessive suffix, REP reportative, vx verbal inflectional suffix

216

Chapter Six

Section 4 shows how Samoyedic languages can express definiteness, whereas Section 5 contains concluding remarks.

2. The Samoyedic Languages and their Typological Overview All Samoyedic languages are severely or critically endangered according to the criteria employed by UNESCO (Moseley 2012) or Krauss (1992). Nganasan is a seriously endangered language with about 50 speakers, which is about 5% of the Nganasan population. According to official statistics, the number of speakers is about twice as much; fieldwork experience, however, shows that the official figures are overly optimistic. Selkup is a language on the verge of extinction. Presently, the northern dialect has most speakers, but their number is estimated at only a few hundred. According to Russian census results from 2010 about 28% of the ethnically Selkup population speaks the language, however in reality the number seems to be much lower. The southern and central dialects are spoken by a total of 10-15 people. Among the languages treated here, Enets is the most endangered language. Only a handful of speakers are left today. According to the latest census of the Russian Federation (2010), there are 21,926 speakers of Nenets altogether, which is about 50% of the ethnically Nenets population and which, compared with other small Samoyedic peoples, is still in a relatively good situation. However, the Nenets language must nevertheless be classified as endangered. Table 1 shows census data from 2002 and 20103 for the Samoyedic peoples in Russia: size of the population and number of speakers. The comparison of the data shows a rapid decline in the number of speakers. 2010 Population Speakers 44,640 21,926

2002 Population Speakers 41,302 31,311

Nenets (Tundra and Forest) Enets (Tundra and 227 43 237 Forest) Nganasan 862 125 834 Selkup 3,649 1,023 4,249 Table 1. Census data on the Samoyedic peoples and languages.

3

119 505 1,641

The census data are taken from and (25 June 2015).

Existentials, Possessives and Definiteness in Samoyedic Languages

217

2.1 Genetic Affiliation Samoyedic languages are part of the Uralic language family, which includes the following language groups and languages: a) Samoyedic languages: Nenets (Forest Nenets and Tundra Nenets), †Yurats, Enets (Forest Enets and Tundra Enets), Nganasan, Selkup (with at least three main varieties), †Mator, †Kamas b) Ugric: Hungarian and the two Ob-Ugric languages spoken along the river Ob: Khanty (with diverse dialectal groupings), Mansi (with great dialectal diversity) c) Finnic (or Baltic Finnic): Finnish, Estonian, Livonian, Ingrian, Votic, Karelian, Veps d) Saamic languages spoken in Norway, Sweden, Finland and Russia: South Saami, Lule Saami, North Saami, Inari Saami, Skolt Saami, Kildin Saami and Ter Saami e) Mordvin, spoken in the Volga area, with two main dialect groups and standard languages: Erzya and Moksha f) Mari with two main dialect groups and standard languages: Hill Mari and Medow Mari g) Permic: Komi (with two literary standards: Komi-Zyryen and KomiPermyak) and Udmurt The following figure shows the family tree of the Uralic languages, it does not, however, contain the names of all varieties and the various different names for each language.

Figure 1. The Uralic language family.

218

Chapter Six

The Uralic languages are spoken in the area extending from Central Europe (Hungarian) to Northern Siberia (Nganasan). Only three languages, Hungarian, Estonian and Finnish have the status of a national language in their country. The “biggest” language is Hungarian. The Hungarian national census of 2011 registered approximately 10 million inhabitants in Hungary, 84.6 % of them are speakers of Hungarian as a native language. In addition, approximately 2–3 million native speakers of Hungarian live outside of Hungary. The smallest language of the Uralic family is Enets with 43 speakers. Since this paper focuses on Samoyedic languages, in the discussion below I will only concentrate on data collected from these languages. The territories where peoples speaking the Samoyedic languages live are part of the Russian Federation, thus the speakers of these languages constitute minority populations. Nenets has the most speakers among the Samoyedic languages and can be divided into two large dialect groups, namely, Tundra Nenets and Forest Nenets. These groups are often considered to be separate languages. The majority of the Nenets speakers (almost 95% of them) speak Tundra Nenets dialects. The Nenets Speakers live in three major administrative districts: Nenets District, Yamal Nenets District and Taymyr Municipal District, which is part of Krasnoyarskiy Kray. The closest relative of Nenets is Enets. The Enets language can be divided into two main dialects, namely, Forest Enets and Tundra Enets, with rather significant differences between the two. Contrarily to Enets and Nenets, Nganasan has almost no dialectal variability: the two main dialects, Avam and Vadey, are very close to each other. The fourth Samoyedic language is Selkup, which is characterized by a large number of dialects. Although the differences between the individual dialects are not always great, they are considerable between what is called northern, central and southern dialect groups. These differences are mostly of a phonological nature, but morphological, syntactical and, naturally, lexical alternations are not rare either.

2.1. Typological Overview This section provides a brief overview of the typological characteristics of Samoyedic languages. These languages are all prevailing agglutinative languages, but the Northern Samoyedic languages (Enets, Nenets and Nganasan) display a highly flectional morphology with regard to case and numeral suffixes: while both can be morphologically separated in Selkup (a Southern Samoyedic language), in Enets, Nenets and Nganasan they are always portmanteau morphemes. Compare the following examples from

Existentials, Possessives and Definiteness in Samoyedic Languages

219

the four Samoyecic languages and from Finnish. Finnish and Selkup do not show flectional characteristics: Nenets: ƾϷno-xϷna ‘in the boat’ [boat-LOC] vs. ƾϷno-xϷҌna ‘in the boats’ [boat-PL.LOC] Enets: mä-kun ‘in the tent’ [tent-LOC] vs. mä-kin ‘in the tents’ [tentPL.LOC] Nganasan: ma-tϷnu ‘in the tent’ [tent-LOC] vs. mað-Ϸ-tinü ‘in the tents’ [tent-EP-PL.LOC] Selkup: mээt-qѠn ‘in the tent’ [tent-LOC] vs. mээt-Ѡ-t-qѠn ‘in the tents’ [tent-EP-PL-LOC] Finnish: talo-ssa [house-LOC] vs. talo-i-ssa ‘in the houses’ [house-PL-LOC] As the above data demonstrate, Selkup has preserved the Proto-Uralic *t plural marker. The plural marker of Nenets is also a reflex of this protolanguage marker but it has become a part of currently used inflection and thus the number marker and the present-day case marker cannot be identified separately in it. The same is true of Nganasan and Enets, where the other Proto-Uralic plural marker, *i, is used, whose reflex can also be found in Finnish non-nominative forms. As far as the synchronic state of the language is concerned, this plural marker cannot be identified separately from the case marker either. (For more historical linguistic background on Samoyedic, see Mikola 2004 and sources given there.) All Samoyedic languages have extensive case systems: the Northern Samoyedic languages have at least seven or eight grammatical cases, while Selkup has 13. Gender, indefinite or definite articles, however, are not features of the Samoyedic languages. But they all show a marker for dual in addition to singular and plural. Nganasan, Nenets and Enets have three conjugation types, namely subjective, objective and reflexive-medial conjugations. The objective conjugation suffixes do not only express the person and number of the subject but also the number of the object. The rules of usage of objective conjugation have not yet been completely clarified for any of these languages. Studies hitherto allow the conclusion that the definiteness and/or indefiniteness of the object is not a decisive factor; the usage of the conjugation types depends on the information structure of the sentence. (For more details, see Körtvély 2005, Siegl 2013, and all references listed there.) All Samoyedic languages in question are head final languages. Nganasan has the freest word order of the four. In the other languages, the preference for an OV word order is much stronger, but pragmatic

220

Chapter Six

organization permits word order changes in these languages as well. The Samoyedic languages are topic-prominent languages: the topic is placed at the beginning of the sentence, whereas the focus is in preverbal position. The Samoyedic languages form subordinate clauses with non-finite verb forms: participles or converbs. These languages are pro-drop languages, thus, the subject or the pronominal object of a sentence is only expressed overtly if it receives extra emphasis or if it is in a topic position. In all other cases it is only the verb inflection that refers to the subject. Additionally, the verb morphology refers not only to the subject but can also refer to the object – and when it does, the object is not expressed overtly in the sentence either. Compare the following Nenets sentences: (5)

a. ladϷ-r hit-2SG>SG.OBJ ‘You hit him/her.’ b. sҨita ladϷ-n (s)he.ACC hit-2SG ‘You hit HIM/HER.’

(Nikolaeva 2014: 386)

(Nikolaeva 2014: 387)

The sentence in (6), shows a typical Nganasan sentence from a narrative text. (6)

ƾonΩΩ ćebtaða-suΩ-mu࣯ lҨüΩ࣯sҨitΩ and tell-PST-1PL Russian.ACC bΩlѠ, ĔaaćΩtΩ-mΩnѠ song.ACC Nganasan.language-PROL kΩiƾi-sҨüΩ-mu࣯ sing-PST-1PL ‘We sang also Russian songs, this song we sang in Nganasan.’ [ChND_080719_Life_nar.0174]

Samoyedic languages express definiteness not by the use of articles but through other means, e.g. with possessive suffixes or word order. In the present paper, I examine two points regarding this: first, how existential, locative and possessive sentences are formed, and second, what kind of determiners can occur before the noun phrase without expressing definiteness. I will concentrate on Nganasan in discussing the latter point.

4

Data codes such as “[XYZ_ (daymonth)year_title of text]” refer to texts. More about the sources see in Section 2.2.

Existentials, Possessives and Definiteness in Samoyedic Languages

221

(For more on the characteristics of the typology of Samoyedic languages, see among others Bakró-Nagy 2012, 2006, Hajdú 1985, and Tauli 1963.)

2.2. Language Data of Endangered Languages The present discussion is based on two types of data. On the one hand, it relies on existing grammatical and typological descriptions (such as, for example, Siegl 2013, Nikolaeva 2014, Tereshchenko 1973, Tereshchenko 1979, and Wagner-Nagy 2011 etc.); on the other hand, on language data derived from texts from fieldwork text collections as well as on elicited data. When discussing Nenets, Enets and Selkup, primarily grammatical descriptions and text collections are used (e.g. Siegl 2013, Nikolaeva 2014, Tereshchenko 1973, Almazova 1961, Kuznecova et al. 1980 and 1993 etc.) due to the fact that no searchable corpus data exist for these languages, unlike Nganasan, for which there is a corpus that is more or less electronically searchable. This corpus includes about 60,000 sentences in total, mostly folklore narratives, but it contains everyday texts as well. A smaller part of the corpus is composed of published texts; however, the majority of the texts are unpublished. Almost all examples cited here are taken from the unpublished part of this corpus. For illustrating grammatical phenomena, I prefer to provide sample sentences from the collected texts over elicited sentences, since syntactic and pragmatic phenomena can rarely be demonstrated with elicited sentences. Regarding these data, I assume that a considerable Russian influence poses some limitations in interpreting them. Examples coded as “[XYZ, year]” are elicited data, while data coded as “[XYZ_ (daymonth) year_title of text]” refer to texts. The first three letters are the initials of the informants; the dates refer to the year of collection. The Nganasan examples are from this database, which is compiled in the context of the Project Corpus building and corpus-based grammatical studies on Nganasan5. At present, the corpus contains 103 glossed and annotated texts that are aligned with the corresponding audio files.

3. Existential, Locative and Possessive Sentences Usually those sentences are regarded existential in which the speaker states something about the existence or nonexistence of something or someone. According to the definition of McNally (2011: 1830), these constructions are specialized or non-canonical. Such sentences have been 5

The project is supported by German Research Found (DFG WA 3153-2/1).

222

Chapter Six

investigated in numerous works (see, for instance, Freeze 2001, Leonetti 2008, and Leonetti in this volume, McNally 2011 etc.). Here I will only reiterate their most important structural characteristics. Authors studying existentials agree that at least three major elements can be identified in existential sentences. Every existential sentence contains an NP (a pivot nominal or theme), the existence or nonexistence of which is stated – usually it serves as the subject of the sentence. According to the definition by Milsark (1977) or Freeze (2001: 941), this entity must be indefinite. If this entity is definite, the sentence is not existential but locational, as in sentence (7b). Sentence (7a) illustrates an existential sentence; because this sentence is specialized, while sentence (7c) is not considered as existential. This sentence entails nothing other than the existence of one yellow cat and it has a canonical structure used in English. (7)

a. There is a yellow cat. b. The yellow cat is there. c. One yellow cat exists.

In most of the world’s languages, existential sentences have a verbal element which is either homophonous with a be-verb as in English or with the verb ‘to have’ as in Catalan or a special verb of existence. In some languages, some other verb occurs whose meaning is bleached in the existential context, as, for instance, in German (geben ‘give’). As will be made clear below in section 3.1, all Samoyedic languages have such a special verb of existence. Another possible element of existential sentences is a coda or locative phrase describing the location in which the given entity is or exists. This latter element is not an obligatory part of an existential sentence. Sentences expressing generic statements (such as four-legged animals exist) do not contain coda (locative elements) either. In numerous languages, existential sentences contain an expletive subject (or according to the definition of Lyons (1969) a deictic particle) like Arabic fii ‘in that, there(in)’, French il or English there. Freeze (1992) calls this sentence type proform existential. Proform existential sentences as such do not occur in Samoyedic languages, but, as we will see, it is historically possible to trace an expletive element which used to occur in existential sentences and which, by now, has grammaticalized as an existential verb. In recent literature on typological locative predication, the phenomenon has often been treated together with existential sentences (see, e.g. Freeze 1992, Payne 1997 and Dryer 2007 etc.). The reason for

Existentials, Possessives and Definiteness in Samoyedic Languages

223

this is that both sentence types contain the same elements. The pivot element of locatives is usually definite, but there are also some sentences where it is indefinite. Which elements express definiteness, however, varies from language to language. The simplest solution is the use of the definite article – which occurs in only one Uralic language, Hungarian. As has been pointed out above, Samoyedic languages do not have articles and, therefore, express the definiteness of the NP by other means (see Section 3 below). The languages of the world code locative predicates in various ways besides marking definiteness, for instance, by changing word order. In locative sentences the (marked or unmarked) definite theme is moved to the subject position – as, for instance, in Russian. Sentence (8a) has an existential meaning, whereas (8b) is a locative predicate – however, the only difference is in the word order. (8)

a. na stol-e ryzhaja koshka on table-LOC yellow cat ‘There is a yellow cat on the table.’ b. ryzhaja koshka na stol-e yellow cat on table-LOC ‘The yellow cat is on the table.’

(p.k.)

(p.k.)

The situation is somewhat more complex in Finnish, where, just like in Russian, there is a change in word order, but it is also accompanied by a change in case marking. The Finnish examples in (9a) and (9b) are existential sentences, while example (9c) illustrates a locative sentence. As can be seen, if the pivot element of the existential is in the nominative, it takes the postverbal position. If the pivot element is coded for the partitive, it can only occupy this preverbal position. In the same position, a coding for the nominative results in a definite interpretation, and the sentence is a locative one. Thus, the word order of the locative sentence is theme– predicate–locative. (9)

a. talo-ssa on lapsi house-LOC be.3SG child ‘There is a child in the house.’ b. lapse-ja on talo-ssa. child-PL.PAR be.3SG house-LOC ‘There were (some) children in the house.’ c. lapsi on talo-ssa child be.3SG house-LOC ‘The child is in the house.’

(p.k.)

(p.k.)

(p.k.)

224

Chapter Six

(For more about existential sentences in Finnish, see Kajander 2013, Huuhmo 2003 and the sources listed there; about definiteness, see e.g. Chestermann 2005.) Before I discuss existential sentences in Samoyedic, a brief reference to a third type of sentence has to be made, namely, to what has been called intransitive possessive sentences. Most Uralic languages do not have a habeo verb, and possession is expressed by a sentence that, when regarding its structure, looks like an existential sentence. The sentential predicate is a verb, which is normally the predicate of existential sentences, i.e. a verb with the meaning ‘to be’, ‘to exist’or ‘to be there’. The possessor does not necessarily have a subject function in the sentence. It is characteristic for some of the Uralic languages that the possessor is marked with a locational (e.g. lative, dative, locative, adessive etc.) or genitive case. The marking of the possessed can vary from language to language; in some languages they are not marked at all, while in others they carry e.g. a possessive suffix. The construction will be illustrated by a Hungarian (10) and a Kamas (11) example. (10)

(11)

Anná-nak van Anna-DAT be.3SG ‘Anna has a book.’

könyv-e book-3SGPX

büüzҨe-n nagur old.man-GEN three i-bi be-PST.3SG ‘The old man had three daughters.’

ko࣯bdo-t daughter-3SGPX

(p.k.)

(Joki 1944: 197)

3.1. Existential Sentences in Samoyedic Languages The present section addresses the question how these three different types of sentences – existential, possessive, and locative – are expressed in Samoyedic languages, with special attention to the strategy of coding definiteness. As we will see, case marking does not play a role here – instead, the choice of verb will be the decisive factor in what interpretation a given sentence obtains. First existential sentences are discussed, followed by locative and possessive ones. Three of the four Samoyedic languages have special existential verbs. The Nenets verb is taĔasҨ, while the Enets one is tonäš. Both verbs are cognates of the existential verb used in Nganasan (see below). A typical characteristic of Samoyedic languages is that the same verb can be used in

Existentials, Possessives and Definiteness in Samoyedic Languages

225

possessive sentences as well (see Section 3.3 below). A Nenets existential sentence is illustrated in (12), an Enets sentence in (13). (12)

Ģiki pædara-xana xo-࣯ that forest-LOC birch-PL ‘There are birches in that forest.’

(13)

tunin soiÿa nääþuk there good girl.youngster bai nääþuk clan.name girl.youngster ‘There is a good girl, a bai girl.’

taĔa-࣯. exist-3PL (Almazova 1961: 39) tonä, exist.3SG

(Siegl 2013: 341)

As sentences (12) and (13) demonstrate, the word order in the existential sentences is locative–theme–copula, which is in agreement with the tendency that Freeze (1992) found in SOV languages. The pivots of the sentences do not have any determiners, so it is safe to say that they are indefinite. Of the Samoyedic languages, Nganasan offers the widest range of possibilities in which existential sentences can be formed, namely, with or without copulas. Sentences without a copula are used very rarely, though. First, I discuss sentences with a copula. One of the most interesting characteristics of Nganasan is that two copulas (tΩisҨa ‘to exist’ and tΩnijsҨa ‘to exist’) can be used in existential sentences. Both copulas were formed by the lexicalization of a pronominal stem (tΩ- ‘this’) + be-verb (isҨa ‘to be’) combination. The pronominal stem tϷ- does not have a locational meaning. In the complex verb tΩnijsҨa (tΩĔijsҨa), the copula is preceded by the adverbial locative form (tϷĔi or tϷni ‘here’) of the tΩ- pronominal stem. Thus, this form can be analysed as representing an expletive proform which by now has evolved to form a fixed part of the lexicalized existential verbs tΩisҨa and tϷnijsҨa. The existential verb tΩisҨa ‘to exist’ appears most often in existential and possessive clauses. This verb can take tense markers, so it can be used in past and future tenses as well, as sentence (14) shows. (14)

ΩmnѠ školΩ tΩi-sҨüΩ here school EX-PST.3SG ‘There used to be a school here.’ [PED04_MyLife_nar.11]

The existential verb tΩnijsҨa ‘to exist’ is used less frequently than tΩisҨa in existential constructions; it is almost exclusively used by the oldest

226

Chapter Six

speakers, but also in their speech it occurs rarely. The use of this copula is illustrated in (15). (15)

ƾunѠrѠ-࣯Ѡa ma-࣯a tΩnɿj-huִ aƾhu short.side.of.tent-AUG tent-AUG EX-REP.3SG ‘They say that there are tents with short flaps.’ [KTD_SeuMelangana_flks.108]

As sentence (14) and (15) show, the word order used in Nganasan is the same as in Enets and Nenets, with the pivot element placed immediately before the verb which can be preceded by a coda (a locative) if there is one. In the text corpora there are some, although, admittedly, very few, sentences in which the word order is the same as in existential sentences but in which the existential verb (isҨa) is used, rather than one of the two copulas, as in (16). (16)

tΩndΩ ma-tΩnu ƾuΩ-lҨaa tѠbѠjkiִ a i-þu. that.GEN tent-LOC one-LIM youth be-CO.3SG ‘In the tent there’s only a boy.’ [KNT_940903_KehyLuu.011]

Copulaless existential sentences are far less frequent than those with a copula. As shown in example (17), these sentences can be formed without a verb form of any kind. (17)

tΩndΩ kuΩćΩ Ĕini this.GEN hill.GEN on ‘On this hill there is a tent.’

ƾu࣯Ωj ma࣯ one tent [JD_00_Musuna_flkd.002]

In this case, the constituent order is locative–theme as well. This, of course, does not allow for any speculations as to where the copula would stand if there were one in the sentence. In any case, it must be noted that this way of forming existential sentences is far less frequent than the two other types discussed above, and it only appears in folkloric texts. In contrast to the Northern Samoyedic languages, in Selkup existential sentences always use the copula, the be-verb ѓѓ-qo. Thus special existential verbs are unknown in Selkup. If there is no explicit morphological marking for the definiteness of the theme, locational and existential sentences only differ from each other concerning word order. In existential sentences, the order of the theme and the locational expression is LOC THEME.

Existentials, Possessives and Definiteness in Samoyedic Languages

(18)

227

qoltoo-qѠnѠt too-qѠn-Ѡt koþi qϷϷlѠ ѓѓ-ƾa river-LOC.1PL lake-LOC-PL many fish be-CO.3SG ‘There are lots of fish in the rivers and lakes.’ (Tereshchenko 1973: 257)

All Samoyedic languages have negative existential verbs. The form of this verb in Tundra Nenets is jaƾkosҨ, in Enets it is ćaguš, while in Selkup it is þääƾkѠqo. There are two existential negation elements in Nganasan: a verb with a complete paradigm (ćaƾgujsҨa), which has cognates in all other Samoyedic languages, and a particle-like negative existential predicate (ćaƾku). For negation in the present tense in Nganasan the existential predicate ćaƾku is usually used. It can only agree with the subject in number. Inflected forms of the verb ćaƾgujsҨa can also be used in the present tense but appear infrequently. In the past or future tense, instead of ćaƾku and the be-verb, a form of the verb ćaƾgujsҨa, inflected for tense or mood, is used. As is illustrated by sentence (19), the word order is exactly the same as in affirmative sentences. Negation does not change the inflection of the pivot (theme), that is, the case marking remains the same. (19)

maa-gΩlҨþΩ koru࣯ ćaƾku, maa ćaƾku. what-EMPH house NEG.EX what NEG.EX.3SGVX ‘There are no houses, nothing at all.’ [TKF_061023_SmallBird_flks.502]

We can conclude, then, that the word order of existential sentences is fixed in all of the Samoyedic languages discussed above: the locative element precedes the theme (pivot element), and the sentence final element is the verb. Northern Samoyedic languages have a special verb which is used in existential sentences and also in possessive sentences (for more on this, see section 3.3. below).

3.2 Locative Sentences in Samoyedic Languages In the following I examine how a typical locative sentence is constructed in two Samoyedic languages, namely in Selkup and Nganasan. As has been pointed out above, there is no special existential verb in Selkup, so both locative and existential sentences use the be-verb. In Selkup, there is a difference in word order: while in existential sentences (20a) the locational element occurs on the left hand side, in locative sentences the theme does, as the example in (20b) illustrates.

Chapter Six

228

(20)

a. mээt-qѠn tээntѠ po-t ѓѓ-ƾээ-tѠt house-LOC broad wood-PL be-CO-3PL ‘There are planks in the house.’ (Kuznecova et al. 1980:170) b. tээntѠ po-t mээt-qѠn ѓѓ-ƾээ-tѠt broad wood-PL house-LOC be-CO-3PL ‘The planks are in the house.’ (Kuznecova et al. 1980:170)

The available data indicates that the sentence in (20b) cannot receive an indefinite interpretation “*Planks are in the house/*There are planks in the house”. This shows that only the change in word order signals that the pivot is interpreted as [+definite]. This supposition is supported by the fact that in sentences where the pivot is unequivocally definite (for instance, because it is marked for possessive and really expresses possession) a word order is used, like the one in (21). (21)

mээl-lѠ kuntaq-qѠt tent-2SG far-LOC ‘Your tent is far.’

ѓѓ-ƾa be-CO.3SG (Tereshchenko 1973: 163)

Among the Uralic languages, this phenomenon is not unique to Selkup. A more distant Uralic language, Finnish, also lacks articles, and the interpretation of the existential and locative sentences is largely determined by the word order (see Section 3 above). The situation is more complicated in Nganasan, where the existential verb introduced above (tϷisҨa) and the be-verb (isҨa) can both occur in a locative sentence, as is illustrated in the sentence pair below in (22) and (23). (22)

taaĔiϷ kϷntϷ-rϷ ƾonϷϷ tϷi-þu. that sledge-2SG another.one EX-CO.3SG ‘There, a little way away, your sledge is still standing.’ [KTD_SeuMelangana_flks.1149]

(23)

tΩtɿ bΩjka࣯a-tu ϷmϷniϷ tajƾiϷ kϷi-tΩnѠ that husband-3SG this.one opposite.GEN side-LOC i-þu. be-CO.3SG ‘That husband of hers can be found on the side opposite her [=the new guest].’ [KTD_SeuMelanga_flks.922]

Existentials, Possessives and Definiteness in Samoyedic Languages

229

Given that in both sentences the pivot is marked for a possessive personal suffix, it can only be definite, and thus the sentence in (22) and (23) cannot be existential. As sentence (23) above shows, the word order is pivot (theme)–coda (locative) –copula [that husband + on the opposite side + is], the exact opposite of an existential sentence. The next two example sentences also illustrate Nganasan locative sentences, and in both of them the coda can be identified unequivocally. (24)

a. tΩni࣯iִ a mej-lҨi-ći […] tΩti KΩhѠ Luu so do-INCH-INF that Partridge Parka hüϷ-tΩ tΩnѠ i-þu. year-LAT there be-CO.3SG ‘During the funeral ceremony […] that Kehy Luu is there the whole time.’ [KNT-940903_KehyLuu.168] b. sѠti ĔuΩ-mti mѠndѠ-tѠ-gΩj, they.DU child-ACC.SG.3DU carry-CO-3DU labsΩ-tu kunsѠ-nѠ i-þu scradle-GEN.3SG inside-LOC be-CO.3SG ‘The two of them take the child in the cradle with them.’ [JSM_090809_Life_nar.300]

The other two Northern Samoyedic languages use a different strategy: in locative constructions, two kinds of copulas can be used. In both languages the BE verb (Nenets ƾæsҨ, Enets eš) or the verb with the meaning ‘be at somebody’s place, be somewhere’ (Nenets mæsҨ6, Enets ƾaš) is used. The word order of the sentence also changes: the coda is placed in a sentence initial position, as (26) and (27) below illustrate. This means that two factors indicate that the pivot element is strong: the use of the verb as well as the word order. The following sentence pair shows the usage of the verbs Nenets ƾæsҨ, Enets eš and ƾaš. As the Enets sentences show, eš can only be used together with a tense marker or a morpheme connected with the marking of tense or mood. The verb ƾaš, in contrast, can also stand alone and must therefore be regarded as the primary BE verb. The constituent order in the case of both copulas is, as can be expected, pivot - coda - copula. Sentence (25) gives a Nenets example, while sentence (26) shows material from Enets.

6

This verb also goes back to PS (*me֑ - ‘be’ Janhunen 1977: 91). It has only been retained in Nenets and Kamas (mo-).

230

(25)

Chapter Six

texnikuma-wa࣯ Dućinka-xana technical.school-1PL Dudinka-LOC ‘Our technical school is in Dudinka.’

ƾa. be.3SG (Nenyang 2005: 62)

(26)

a. bu ekon ƾa (s)he here be.3SG ‘(S)he is here.’ (Sorokina – Bolina 2001: 89) b. bu ekon ƾa-š (s)he here be-3SG.PST ‘(S)he was here.’ (Sorokina – Bolina 2001: 89) c. äki enþu mäð-iðu࣯ to this person.PL tent-PL.3PL lake.GEN kɟ-xon e-bi-࣯. near-LOC be-PERF-3PL ‘These people, their tents were in the near of the lake.’ (Sorokina-Bolina 2005: 1005/13)

In Nenets, the verb mæsҨ ‘be at somebody’s place, be somewhere’ is used for expressing the location of animate beings. The following examples from Tundra Nenets illustrate locational expressions of animate beings. (27)

a. Säxäko boĐĔica-xana me Seheko hospital-LOC be.3SG ‘Seheko is in the hospital.’ (Nenyang 2005: 57) b. toxolkoda klass-xana me student classroom-LOC be.3SG ‘The student is in the classroom.’ (Almazova 1961: 77)

As is clear from the texts investigated by me and in accordance with Nikolaeva’s (2014: 263) observations as well as confirmed in personal communication by Nikolett Mus, this verb cannot be used in an existential sentence about an animate being (like ‘There is a boy/dog in the house’)? We can conclude, then, that word order and the choice of verb are most important in Enets and Nenets. To summarize, we can state that the choice of verb plays an important role in differentiating existential and locative sentences in Samoyedic languages (except for Selkup), in addition to word order, which is relevant in all of these languages. Table 2 summarizes the similarities in the structure of locative and existential sentences in the four languages under discussion.

Existentials, Possessives and Definiteness in Samoyedic Languages

Language

Construction Existential

Nenets

Locative

Enets

Existential Locative

Existential Nganasan

Locative

Copula taĔasҨ mæsҨ ƾæsҨ tonäš ƾaš eš tϷisҨa tϷnijsҨa isҨa Ø tϷisҨa isҨa

Constituent order Coda pivot cop pivot

coda

cop

Coda

pivot

cop

Pivot Coda ? Coda Coda

coda pivot pivot pivot pivot

cop cop cop cop

pivot

coda

cop

231

Comments [+animate] [-animate]

ѓѓqo Coda pivot cop Existential Selkup Pivot coda cop Locative Table 2. Constituent and Word Order Patterns in Locative and Existential Sentences. The table clearly demonstrates that the differences between existential and locative sentences are to be found in their use of the copula as well as their constituent order. Selkup is the only language which has only one copula. In Nenets the use of the copula already indicates whether the sentence is locative or existential. Nganasan is between these two extremes. In the next section I analyse possessive sentences.

3.3. Possessives in Samoyedic Languages In the Uralic languages in general, and in the Samoyedic languages in particular, predicative possessive sentences are expressed with an existential construction rather than a have-verb. The majority of the Uralic languages belong to the oblique possessive group. (On the classification of possessive constructions, see Stassen 2001, 2009, and 2013). In this type of construction, the grammatical subject of the sentence is the possessed NP, while the possessor is marked with a case suffix. In most cases it is a locational case, less often the genitive or nominative (cf. Wagner-Nagy 2011). The sentential predicate is a verb with the meaning ‘to be’ or ‘to exist’. This means that in the languages under scrutiny here there is a strong connection between real existential sentences and possessive sentences. However, in the possessive sentence, the possessed NP must

232

Chapter Six

bear a possessive suffix which refers to the possessor. The possessor NP may be overt or covert in the sentence, as illustrated by the pair of sentences in (28). Sentence (28a) is an Enets example with overt possessor, while sentence (28b) is a Nganasan one with covert possessor. (28)

a: Forest Enets; b: Nganasan a. moć ne-j tonä I woman-1SG exist.3SG ‘I have a wife.’ (Mikola 1967: 61) b. þiΩðѠr-iִ a࣯ku-þi tϷi-þu. rack-DIM-PL.3DU exist-CO.3SG ‘They two have racks.’ [TKF_97_Djajku_flkd.005]

I now move on to a detailed discussion of the structure of the individual languages. In Nenets, possession can be expressed in different ways, i.e. two different predicates are used in this construction, namely the special existential verb taĔasҨ ‘to exist’ and very rarely the ƾäsҨ ‘to be’ that is normally used in locative sentences. (29)

a. mɚĔ Ĕekɚ-mi, sҨića Ĕe Ĕa-mi tɚĔa. I brother-1SG two woman sister-1SG exist-3SG ‫ދ‬I have one brother and two sisters.‫ތ‬ (Vanuyto 2012: 13) b. pida sҨića Ĕe-da ƾä-wi (s)he two woman-3SGPX be-NAR.3SG ‘He had two wives’ (Lehtisalo 1956: 26)

Also in Enets, two types of possessive construction can be distinguished. The predicate of one type is the verb toneš ‘to exist’, while there is no verb in the other construction. The latter type is most commonly used if the speaker talks about his or her children and stresses their number. In both types, the possessed NP must be marked with the possessive suffix. The sentences below illustrate the case when the speaker uses an overt possessor marked with nominative case. The sentential predicate has to agree in number with the possessed NP. (30)

a. þiki te kasa-ða tonee-bi-sҨ this reindeer man-3SGPX exist-NAR-PST.3SG – tobik – mouse ‘This reindeer had a friend – a mouse’ (Sorokina – Bolina 2005: 170/2)

Existentials, Possessives and Definiteness in Samoyedic Languages

b. teaða moć nehu࣯ now I three ‘I now have three children.’

233

Ĕe-j child-1SGPX (Mikola 1967: 61)

There are several ways to express possession in Nganasan, firstly through existential sentences, secondly with a habeo-verb construction. Only the existential verb (tΩisҨa) can be used in possessive sentences. The possessed NP functions as the grammatical subject of the ‘to exist’-predicate, while the possessor NP is construed in nominative form. The possessed has to agree in number and person with the possessor by means of a possessive personal ending. The following sentence pair illustrates the possessive sentences in Nganasan. (31)

a. mΩnΩ Ĕintuu ƾonΩnΩ buΩða-ĔΩ, mΩnΩ I not myself word-PL.1SG I tΩbtΩ baarbΩ-mΩ tΩi-þu. also chief-1SGPX exist.3SG ‘These aren't my own words, I also have a master.’ [KNT_960809_WildAnimals_flkd.247] b. mΩnΩ nѠ hon-dѠ-m I woman.ACC have-CO-1SG ‘I have a wife.’ [ChND, 2008]

It was mentioned above in Section 3.1, that there is no special existential verb in Selkup. Consequently, Selkup uses the be-verb in possessive sentences. It is typical for Selkup possessive sentences that if the possessor is understood through context, it is generally not overtly expressed in the sentence. The possessed carries the personal possessive suffix referring to the possessor. (32)

ira ilѠ-mpa. šittѠ nälҨa-tѠ old.man live-PST.NAR.3SG two daughter-3SG ѓ-ppѠntѠ be-LATENT.PST.3SG ‘There lived an old man. He had two daughters.’ (Kuznecova et al. 1993: 8/1–2)

As a conclusion, it is safe to state that there is a correlation between existential and possessive sentences in Samoyedic languages. If the language has a special existential verb, it is also used in possessive sentences in which the possessed element is marked by a person marker

234

Chapter Six

co-referenced with the possessor. For more on the structure of possessive sentences, see Wagner-Nagy (2011: 213-264). Before discussing how definiteness is expressed, adnominal possessive structures have to be mentioned briefly. The adnominal possessor construction can contain a pronominal or a lexical possessor. The simplest construction is the construction with an unbound pronominal possessor, however, this is optional and most often absent. But the possessive suffix on the possessed element is obligatory, cross-referencing the possessor in number and person. There are no special possessive pronouns in the Samoyedic languages. Personal pronouns can function as possessive pronouns in the sentence. If the possessor is a lexical element, it must be inflected for the genitive. The possessed can take the possessor suffix, but the affix is not obligatory. The examples in (33), (34) and (35) illustrate these points. Example (35) is a Selkup sentence with a possessive NP. (33)

Enets a. moć koću-i I/my sledge-1SGPX ‘my sledge’ (Siegl 2013: 233) b. muć kasa-Ĕ I companion-GEN.1SG ‘my friend’s net’

poga net (Siegl 2013: 234)

(34)

Nganasan a. (mΩnΩ) ćesѠ-mΩ I/my father-1SGPX ‘my father’ b. maðu-࣯ ban-Ω-࣯ tent-GEN.PL dog-EP-PL ‘the dogs of tents’

(35)

man adҊҨuka-m ugot tabeþ-le I/my grandmother-1SG earlier squirrel.hunt-CONV kwaja-kku-mba go-DUR-PST.NAR.3SG ‘Earlier my grandmother hunted squirrels.’ (Bykonja et al. 1996: 95)

The possessive construction can be modified with a demonstrative, which precedes the possessor if the possession is present, as in the following Enets example.

Existentials, Possessives and Definiteness in Samoyedic Languages

(36)

235

þiki LҨonka poga-ÿa še-sai this Leonid.GEN net-3SGPX hole-COM ‘This one net of Leonid’s has a hole.’ [lit. ‘is with a hole’] (Siegl 2013: 235)

4. Definiteness in Samoyedic Languages So far the existential, locative and possessive sentences of the Samoyedic languages have been discussed. I have pointed out that there is a strong connection between the three sentence types in these languages. The inflection of the pivot element and the word order play a crucial role in deciding what interpretation a sentence is assigned. The previous discussion of data has shown that even though there are no articles in order to code definiteness in the languages discussed here, there are other means. I also examine whether an existential construction can have a possessed DP with a strong pivot element. In this part of the investigation I focus on two languages, Nganasan and Enets. In the following I will propose how the different means should be classified according to Milsark’s distinction. As mentioned in Section 1, Milsark (1977) discussed different types of determiners, demonstrating that some of these make the pivot definite, while others do not. Now I want to examine what kind of determiners an NP can get as well as categorize DPs according to Milsark’s (1977: 46) classification. As a starting point, I will use Milsark’s taxonomy developed for English. (37)

Weak and strong DP’s in English WEAK

STRONG

a something, someone number determiners Ø plural and mass determiner in nonuniversal reading

the demonstratives pronouns possessive DET’s universals: all, every, each any when not polarity item of some Ø DET in universal reading

To illustrate that this classification also holds for the Samoyedic languages, I will give some examples from Nganasan. As I have already stated, definiteness cannot be expressed with an article in Samoyedic languages, but can be signalled by word order change (compare existential

236

Chapter Six

and locative sentences). In Nganasan (and in other Samoyedic languages) a noun phrase is typically head-final: it is composed of an optional determiner and a noun head, thus the basic structure of the noun phrase is as follows: (Det)-(Num/Quant)-(Adj)-Noun. Various demonstratives, adjectives and indefinites etc. occur among the determiners. If an NP is modified by only one adjective, the bare NP can be interpreted either as definite or indefinite, as in (38). The valid interpretation will depend on the syntactic context, specifically on whether the NP already occurred in previous sentences. (38)

hirϷgϷϷ ma࣯ tall tent ‘a/the tall tent’ (constructed example)

In order to examine the definiteness effect in the languages in question, the DPs of the language have to be categorized according to Milsark’s (1977) classification of strong vs. weak DPs. Furthermore, it is necessary to demonstrate that the pivot of the existential sentence and the subject of the individual-level predicate can function according to this classification. Our prior knowledge regarding Nganasan can be summarized as follows: the pivot of existential sentences can be modified with numerals (number determiners or number markers) and with quantifiers. (39)

Weak and strong DP’s in Nganasan strong DPs proper names demonstrative ‘this’, tΩΩ ‘this over there’, ΩmΩniΩ ‘this one’ takΩΩ ‘that over there’, tagΩniΩ ‘that one’ ΩmtѠ ‘this’ tΩti ‘this/that, it’ possessive marker universals such as bϷnsϷ ‘all; whole’; bϷnćika ‘every, all’

weak DPs number determiners NP with number marker ΩmΩ quantifiers: tϷrϷći; ϷmlϷći ‘such’ kanΩraa~kanΩraa࣯kü ‘some’ tanua ‘few, little’

The classification claims that the pivots of existential clauses in Nganasan cannot be modified with a possessive marker. This is indeed so, since, as has been shown in section 3.1 above, the pivot of the existential sentence is never marked for the possessive. Its occurrence can transform a sentence

Existentials, Possessives and Definiteness in Samoyedic Languages

237

into a locative sentence (as in sentences (22) and (23) in Section 3.1 above) or into a possessive sentence (as in 28b above). According to Milsark’s categorization, number marking (in case of Nganasan, the plural and the dual) and number determiners do not necessarily trigger definiteness. The sentence in (40a) shows that a numeral can modify a pivot without making it definite. The same is true for number marking, illustrated in (40b). (40)

a. tѠminiִ a sҨiti tϷjbϷ tϷi-þu. now two bird EX-CO.3SG ‘There are two birds.’ [MDN_97_TwoBirds_flkd.005] b. ou, ƾϷmsu-࣯ tϷi-þu-u-࣯, tΩ ϷmϷniϷ EXL meat-PL EX-CO-EXL-3PL so this.ACC ƾϷmsu maa Ĕi-ƾѠ-ƾ sҨerѠ-࣯ meat.ACC what NEG-INTER-2SG carry-CNG ‘There is some meat there, why don’t you bring some in?’ [KVB_970930_ThreeGirls_flkd.88]

However, the subject of the individual-level predicate is strong if the DP is headed by bϷnsϷ ‘all; whole’, bϷnćika ‘every, all’ or if the subject is in the plural with a generic meaning. (41)

a. Ϸi࣯,

kobtua-rba࣯a-࣯ bϷĔćϷ-࣯ hotuϷ-j girl-AUG-PL all-PL cheek-PL.ACC tunsѠ-tѠ-࣯ tϷ put-CO-3PL PTCL ‘All the girls presented their cheeks (to be kissed).’ [KTD_SeuMelanga_flks.973] EXCL

b. mou ća sϷþϷ-࣯ki-࣯Ϸ: earth.GEN all scrutinize-RES-CO.3SG bϷĔćika-࣯ ĔoðϷ-࣯ heðiִ a-tѠ-࣯ all-PL grass-PL be.stable-CO-3PL ‘S/He looked at the ground: all the grass was in its place. [i.e. where it was supposed to be]’ [KTD_SeuMelangana_flks.057] Quantifiers and universals behave in the same way in Enets. In Enets, similar to Nganasan, the quantifier þukþi ‘all’ can occur either before or after the head (cf. Siegl 2013: 225-226). However, we do not have enough information on this point for either language to be able to explain this

238

Chapter Six

phenomenon or to pinpoint exactly what triggers the change of positioning. I do not have sufficient data about how the quantifier ‘such’ behaves exactly, so its categorization is only hypothetical. But it is safe to claim that the use of the possessive suffix and that of the demonstrative trigger definiteness, whereas numerals, quantifiers and number markers cannot trigger a definite interpretation.

5. Conclusion In this paper I have examined the relationship of possessive, locative and existential sentences in Nganasan. I have sought to answer the question of what linguistic means can be used to modify the NP of an existential sentence without triggering the definiteness effect. What plays a crucial role in the differentiation of existential and locative sentences is the choice of the copula verb and the word order, whereas the presence or absence of a possessive modifier differentiates between an existential and a possessive sentence. A pivot with a possessive modifier can only be interpreted as strong, thus it can appear only in locative or possessive sentences. I am aware that this paper has left some questions unanswered. This is partly due to the fact that the relevant data is limited in many ways, and the lack of a comparative corpus also makes a comparison of these forms difficult. Some issues will require further investigation and clarification, e.g. the question of what exactly determines the choice of copula in existential sentences in Nganasan. Furthermore, it needs to be clarified what elements can be used without a modifier in Nenets, Selkup and Enets without triggering a definite interpretation. What we can certainly be sure of, however, is that the issue of definiteness in Samoyedic languages has not been fully clarified yet – and will require investigation on the basis of more data and, in some cases, crucial input from native speaker consultants.

References Almazova, A. V. 1961. SamouþiteĐ neneckogo jazyka [Nenets self-taught]. Leningrad: Uþpedgiz. Bakró-Nagy, Marianne. 2006. Az uráli nyelvek tipológiai jellemzése [The typological characterization of Uralic languages]. In Kiefer, Ferenc & Siptár, Péter (eds.). Magyar nyelv [Hungarian language], 267-287. Budapest: Akadémiai Kiadó.

Existentials, Possessives and Definiteness in Samoyedic Languages

239

—. 2012. The Uralic languages. In Revue Belge de Philologie et d’Historie 90: 1001-1028. Bykonja, V.V, A.A. Kim, Sh. C. Kuper, N. P. Maksimova & I.A. Iljashenko (eds). 1996. Skazki narymskih seĐkupov [Folktales of Narym Selkups]. Tomsk: TGPU. Chestermann, Andrew. 2005. On definiteness. A study with special reference to English and Finnish. Cambridge: CUP. Clark, Eve V. 1978. Locationals: Existential, Locative and Possessive Constructions. In Joseph H. Greenberg (ed.). Universals of human language: Syntax, 85–126. Stanford: Stanford University Press. Dryer, Matthew S. 2007. Clause types. In Timothy Shopen (ed.). Language Typology and Syntactic Description. Volume I: Clause Structure. Second edition, 224–275. Cambridge: CUP. Freeze, Ray. 1992. Existentials and other locatives. In Language 68: 553– 595. —. 2001. Existential construction. In Martin Haspelmath, Ekkehard König, W. Oestrerreicher & W. Raible (eds.). Language Typology and Language Universals, 941–953. Berlin/New York: de Gruyter. Hajdú, Péter. 1985. The main characteristic features of the Uralic languages. In Acta Linguistica Hungarica 33: 101–112. Holmback, Heather. 1984. An interpretive solution to the definiteness effect problem. In Linguistic Analysis 3: 195–215. Huuhmo, Tuomas. 2003. Incremental existence: The world according to the Finnish existential sentence. In Linguistics 41: 461–493. Janhunen, Juha. 1977. Samojedischer Wortschatz. Gemeinsamojedische Etymologien. Castrenianumin toimitteita, vol. 17. Helsinki, Suomalaisugrilainen Seura. Joki, Aulis J. (ed.). 1944. Kai Donners Kamassisches Wörterbuch nebst Sprachproben und Hauptzüge der Grammatik. Bearbeitet und herausgegeben von A. J. Joki. Helsinki: SUS. Kajander, Mikko. 2013. Suomen eksistentiaalilause toisen kielen oppimisen polulla. Jyväskylä: University of Jyväskylä. Krauss, Michael. 1992. The World's Languages in Crisis. In Language 68.1: 4–10. Körtvély, Erika. 2005. Verb conjugation in Tundra Nenets. Studia UraloAltaica 46. Szeged: SzTE. Kuznecova, A., E. Helimskij & E.V. Gruškina. 1980. Oþerki po seĐkuskomu jazyku [A Grammatical Sketch of Selkup). Moskva: IzdateĐstvo Moskovskogo universiteta. Kuznecova, Ariadna, Olga Kazakeviþ, L. Ju Joffe & Evgenij Helimskij. 1993. Oþerki po seĐkuskomu jazyku. Tazovskij dialekt. [A

240

Chapter Six

Grammatical Sketch of Selkup: The Taz dialect]. Moskva: IzdateĐstvo Moskovskogo universiteta. Lehtisalo, Toivo. 1956. Juraksamojedisches Wörterbuch. Lexica Societatis Fenno-Ugricae XIII. Suomalais-Ugrilainen Seura, Helsinki. Leonetti, Manuel. 2008. Definiteness Effects and the Role of the Coda in Existential Constructions. In Alex Klinge & Henrik Høeg Müller (eds.). Essays on Determination, 131–162. Amsterdam: John Benjamins. Lyons, John. 1969. A note on possessive, existential and locative sentences. In Foundations of Language 3: 390–396. McNally, Louise. 2011. Existential sentences. In C. Maienborn, K. von Heusinger & P. Portner (eds). Semantics: An International Handbook of Natural Language Meaning, Vol. 2, 1829–1848. Berlin: de Gruyter. Mikola, Tibor. 1967. Enzische Sprachmaterialien. In Acta Linguistica Hungarica 17: 59–74. —. 2004. Studien zur Geschichte der samojedischen Sprachen. Aus dem Nachlass herausgegeben von Beáta Wagner-Nagy. Studia uralo-altaica 45. Szeged: SzTE Finnisch-Ugrisches Institut. Milsark, Gary. 1977. Towards an explanation of certain peculiarities of the existential construction in English. In Linguistic Analysis 3: 1–29. Moseley, Christopher. 2012. UNESCO Atlas of the World’s Languages in Danger. Cambridge: World Oral Literature Project. Nenyang, M. A. 2005. Russko-neneckij razgovornik [Russian-Nenets Phrase-book]. Sankt-Peterburg: Drofa. Nikolaeva, Irina. 2014. A Grammar of Tundra Nenets. Berlin: de Gruyter. Payne, Thomas E. 1997. Describing morphosyntax. Cambridge: Cambridge University Press. Siegl, Florian. 2013. Materials on Fprest Enets an Indigenous Language of Nothren Siberia. Mémoires de la Société Finno-Ougrienne 267. Helsinki: SFO. Sorokina, I. P. & D. S. Bolina. 2001. Slovar‘ enecko-russkij i russkijeneckij. [Enets-Russian and Russan-Enets Dictionary]. SanktPeterburg: Prosveshenie. —. 2005. Eneckie teksty [Enets texts]. Sankt-Petersburg: Nauka. Stassen, Leon. 2001. Predicative Possession. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.). Language Typology and Language Universals. Volume. 2, 954–960. Berlin: de Gruyter. —. 2009. Predicative Possession. Oxford/New York: OUP. —. 2013. Predicative Possession. In Matthew S. Dryer & Martin Haspelmath (eds.). The World Atlas of Language Structures Online.

Existentials, Possessives and Definiteness in Samoyedic Languages

241

Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/117, Accessed on 2015-0212.). Tauli, Valter. 1963. Structural tendencies in Uralic languages. Bloomington. Tereshchenko, N. M. 1973. Sintaksis samodijskih jazykov [The syntax of Samoyedic languages]. Leningrad: Nauka. —. 1979. Nganasanskij jazyk. [Nganasan language]. Leningrad: Nauka. Vanuyto, Galina. 2012. Russko-nenetskij razgovornik [Russian–Nenets phrasebook]. Moskva: Ventana-Graf. Wagner-Nagy, Beáta. 2011. On the Typology of Negation in Ob-Ugric and Samoyedic Languages. Mémoires de la Société Finno-Ougrienne 262. Helsinki: SFO.

III DIACHRONIC PERSPECTIVE

CHAPTER SEVEN MODAL VERBS IN GERMAN AND DEFINITENESS EFFECTS ON THE SUBJECT ARGUMENT—FOCUSING ON MODERN STANDARD GERMAN SOLLEN AND MIDDLE HIGH GERMAN SULN “SHALL”. WERNER ABRAHAM AND MAIKO NISHIWAKI

Abstract1 We test Glas’ (1984) generalization that certain features of past marking on the predicate elicit evidential readings focusing the reportative interpretations of the German modal verb sollen. Furthermore, we discuss to what extent contextual factors such as aspectuality of the infinitival complement and grammatical person on the modal predicate trigger definite effects on the embedded infinitivals. In other words, the search is for where the polyfunctionality of German(ic) modal verbs is found to be disambiguated on the (in-)definiteness reading of the embedded construction in question – a phenomenon and grammatical interplay not much investigated so far. The goal of this paper is to use insights from aspectually conditioned definiteness implications on the subject, where normal definite effects cannot be observed for the very reason that there was no regular change between definite, indefinite, and zero articles as well as quantifiers that modern definite effects go by. The languages investigated are modern Standard German and Middle High German.

1

We thank Sonja Zeman, Elisabeth Leiss, Eckart Rupp (all Munich), and three anonymous reviewers for valuable critical input.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 245

1. Introduction: Modality Disambiguated by Grammatical Subject Features? Definiteness effects/DE appear in English existential sentences introduced by there. The postverbal subject is restricted to certain types of reference such that strong quantifiers (every, all, most) and the definite article the lead to unacceptability, while determiners such as a(n), numerals, and bare nominals are possible (Zucchi 1995, Leonetti 2004). This is different in German (Abraham 2015). Definiteness and indefiniteness are categories with clear complementary distribution in modern German (Spec/CP and left vs. right middle fields). Counter to expectation, definite nominal reference triggers evidential readings of MVs, while the interplay between indefinite nominals and root modality turns out to be less stringent. We focus on material from Middle High German - a language period in German when the article differential had not emerged as clearly as it does in modern German. The ensuing article deviates from the common notion of DE in that we take up a solidly attested link between modality and aspect: In constructions with aspectual contexts, the polyfunctional modal verbs in German and other Germanic languages will receive a disambiguated reading. See (1) (cf. Abraham 1989, 1991, 2002, 2005, 2012a,b, van Hout 1998, Hollebrandse et al. 2005, Maché 2013) for must, whose readings as deontic or evidential are contingent upon the aspect status of the embedded predicative. Notice the DE2 in (1c) in that uniqueness on our doctor allows for both modal readings of the main predicate, must. This is in sheer contrast to (1a) and (1b), where, to all appearances, no definiteness effects are triggered. [EMV=evidential reading on modal verb, DMV=deontic/root reading on the MV must]. (1) a. His son must be a doctor predicative imperfective b. His son must become a doctor predicative perfective

2

Æ *DMV/EMV Æ DMV/*EMV

Generally, DEs concern context variants elicited by (in-)definite clausal functions. It means that the definiteness of a given DP-form triggers a particular behavior of a certain element in the syntactic context, thus signaling either its individuality or its specificity or making available certain positions in the clause and disallowing others. This presupposes that one always seeks to go beyond the mere form of definiteness in terms of referential functionality (individualspecific).

246

Chapter Seven

c. His son must be our doctor predicative uniqueness

Æ ?DMV/EMV3

There is, secondly, a clear mereological link between event aspectuality and reference in terms of definiteness and unique identifiability: Aspectual perfectivity/telicity/foregrounding share with definiteness/unique identifiability the mereological outer view, while imperfectivity/atelicity/ backgrounding and indefiniteness/non-uniqueness share the inner view. The most ‘prototypical’ reference for the mereological view is Bach (1981). The linguistic literature on pertinent topics has encapsulated the two views signaled above in various forms and side topics as a firm and trustable background (van Hout 1998, Leiss 2000, Kotin 2008, 2010, 2012). However, we also come across insights that contradict the latter solidly attested views. The goal of this paper is to take up such views by pointing at various contextually dependent properties of modal verbs, among which also definiteness and indefiniteness in terms of determiners, primarily article forms in Middle High German to modern Standard German. The motivation behind this is that, in the history of German, the forms of both the definite and the indefinite article followed a slow and at times relapsing course from the demonstrative function of the definite article form to today’s twofold function. In the first step, it expressed anaphoricity or familiarity at the hands of the hearer. A second step developed the expression of individuality or singularity as selected from a larger set of referents (Heim 1988, Leiss 2000). By contrast, today’s demonstrative functions as referring to either a deictic-gestural goal or, contextually, the rhema in the immediately previous sentence (Abraham 2005). Not only has it sufficiently been clarified that has aspect fundamental links with (in-)definiteness (Abraham 1991, Leiss 2000), but, since aspectuality has equally well established links to modality (Abraham 1989, 2012a,b), that reference at the hands of article forms relate to modality as well. Grammatical person has the force of disambiguating between root and evidential readings of modal verbs. As Nishiwaki (2013) points out, grammatical reference on the subject nominal influences modality readings to the extent that first and second persons trigger, at a frequency 3

The root-modal/deontic reading, DMV, irrespective of the imperfective predicate be, is probably due to a repair mechanism such that the stative predicate becomes subject to an emergent-event reading. This assumption presupposes that the DMVreading is the least logical implication. For a systematic understanding of such repair mechanisms, see Maienborn (2003).

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 247

rate of cases that is beyond chance, root modality on polyfunctional verbs, while impersonal subject reference favour evidential interpretations (as pointed out already by Heine 1995: 26). Since impersonal subjects coincide with the property of [-animate], it is probably correct to deduce that the feature opposition [±animate] takes care of the modality distinction [±root modality] as Diewald (1999: 255f.). Notice that the indefinite subject pronominals man “one”, jeder “everybody”, niemand “nobody” and wer “someone” elicit root deontics (Heine 1995: 36) as well as Mortelmans (2003: 169, 173). Notice also that these facts are in outright contradiction to Traugott’s (1989, 2003) claim, inherent to her conceptual mechanism of epistemizing (Inter)Subjectification, that 1st and 2nd person predicatives should epistemize rather than giving preference to root modality. Fact is that 1st and 2nd person predicatives do the latter. We shall return to this later. Similarly, the feature opposition “definiteness vs. indefiniteness” as subject characteristics imprints the modal reading (Leiss 2011a, Nishiwaki 2013). Definite subjects co-construe with the evidential modal reading (EMV), whereas indefinite subjects do with root readings (DMV). See (2)(3) featuring the modals sollen “shall” and müssen “must”. Notice that, for the evidential meaning of reportative sollen in modern German about the event e, we argue on the following basis: x tells y with the expression e that p (see also Fabricius-Hansen & Sabø 2004, Schenner 2008: 553). Quite arguably, evidentials are usually taken to not only indicate the source of evidence, but also the speaker’s commitment, the strength of which may be lower or stronger than the truth of e.4 (2) a. Ich gehe mal ein BUCH kaufen. Das Buch/Es soll von japanischer Kultur handeln. > DMV ‘I am going to buy a book. The book/It has to deal with Japanese culture.’ b. Ich gehe das Buch mal KAUFEN. Das Buch/Es soll von japanischer Kultur handeln. > EMV ‘I am going to buy the book. The book/It is said to deal with Japanese culture.’

4

More precisely, following Ehrich (2001: 168), the lexical entry of reportative sollenreport may be generalized as: >[soll@]w = Ȝp.[for every world w’ Rw in which the claims of xc in w are true, it holds that w’  p], where xc is understood as the contextually supplied source of the relevant claims. This amounts to saying that ‘sollen(p)’ is equivalent to ‘xc said that p’.

248

Chapter Seven

Both subjects in (2a, b) are formally definite (das Buch), but they refer to antecedents of different referential types. Since in (2a), the indefinite a book is non-referential, the speaker’s utterance cannot capture reference to a prior individual. This anticipates that any reading for soll other than deontic, and specifically evidential, is out. Deontic soll-, by denoting commitment on the subject referent levied by someone outside of speaker and subject referent, presupposes reference to an individual. In contrast, (2b) contains prior reference to an individual book thereby rendering the evidential reading possible. What singles out, and safeguards, the latter reading is that the definite specific das Buch presupposes some, but not the complete, pre-knowledge on the speaker’s side of what the book is about. See (2’a, b) for the divergent references of the Buch “book”, first1 vs. second2 mention, in (2a) and (2b). In (2a and b), both subjects are formally definite, but they refer to antecedent expressions of different semantic types. (2’) a. ein Buch1 - das-DEF Buch2 soll # property-identified by ein, nonsingular entity (Kripke descriptor: reference by property) > DMV b. das Buch1 - das+DEF Buch2 soll # identified singular entity > EMV Deontic muss- is not much different from soll-: strong commitment on subject referent declared by someone outside of speaker and subject referent.5 (3) a. Ich gehe mal ein BUCH kaufen. Das Buch/Es muss von japanischer Kultur handeln. > DMV b. Ich gehe das Buch mal KAUFEN. Das Buch/Es muss von japanischer Kultur handeln. > EMV This appears to confirm that sollen and müssen, with their almost identical root meanings, yield exactly the same generalization on their contingency on the referentiality of the subject nominal: definiteness elicits evidentiality, whereas indefiniteness triggers root/deontic readings on the co-constructional finite modal verbs. In other words, definiteness effects are achieved in a direct way through modality disambiguation.

5

According to Schenner (2008), sollen ‘should’ in its reportative use is truthconditional and does not lexically encode a reduced degree of speaker commitment. Without going into details, we assume that epistemic müssen ‘must’ has a lexical entry quite similar to that of sollen except that no reference is involved to claims on p made by others, but that the contextually supplied source of the relevant claims lies in the speaker’s view of the world himself.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 249

However, searches in corpus investigations on modern German at the hands of Mortelmans (2003: 169, 173) confirm this only in part since the author’s illustrations hold only for animate subject referents. 6 Definite subjects on müssen appear to favour deontic readings obviously for the reason that animate subjects (but not only 1st and 2nd person subjects) take action predicates, which, in turn, go back to dynamic, and often wilful, actions (Diewald 1999: 255ff.). All we can do for the time being is to take it as empirical facts that the MVs sollen and müssen behave differently with respect to our referential subject contexts. It seems that what we have as explanatory scenarios both in the literature and in our illustrations so far in fact lead to incompatible, and even contradictory, claims. (2)௅(3) above are meant to testify to the generalization that definite subjects co-construe with evidential modal readings (EMV), while indefinite subjects do with root readings (DMV). Notice that our hypothesis that indefinite subjects correlate with modal root readings is not contradicted by Mortelmans’ findings. Cf. fn. 6. Let us briefly point out that the position of default accent in (2a, b) is implemental in capturing topic vs. focus material. The position of the direct object/DO, Buch, in the antecedent clause in (2a) is in the right middle field (with reference to the clausal structure of German OV), closer to the verb, than that of Buch in (2b), where it is in the left middle field, farther away from V in clause-last position. The different default accents are telling by themselves. In contrast to (2a), the VP in (2b) does not host any DP. This entails that the finite verb must carry clausal default accent. Thus, by position alone, the DOs are marked for topic as opposed to focus function. The subject DP in the ensuing clauses picks up this information. Simple pronominal subjects can satisfy both functions in the consequent clause, as by es in our case. One important item concerning the (in-)definiteness criterion on modality disambiguation is hidden in the following illustrations, (4)௅(5). While the indefinite article in (4) refers to an unspecific object (nonsingulative, property reference), Notice that Karttunen’s (1976) wellknown diagnostics to distinguish existential indefinites as in (4) from 6

According to Mortelmans (2003: 169, 173), indefinite subjects favor the root reading. What is basic to this result is that such root readings are restricted in the sense that they apply to animate subjects only such that they combine with action verbs. From this, one may conclude that what is behind the collocation is not so much indefiniteness, but animacy and agentive eventuality. Note that, if subject control is behind the subject-modality relation, agentivity on the event-denoting predicate plays a determining role in other contexts as well (Abraham 2013, Ch. 10).

250

Chapter Seven

referential-specific ones as in (5)7 have no bearing on the implications that we are drawing. (4) Ich will/muss einen Stift haben. Hast du vielleicht einen? ‘I want to/must have a pencil. Do you perhaps have one?’ The very same indefinite article refers to a specific object, yet identifiable by and unknown to the hearer as in (5): (5) Ich will/muss einen Stift haben. Der lag eben noch auf dem Tisch. ‘I want to/must have a pencil. The one was still on the table this very moment.’ Specific vs. unspecific, independently from (in-)definiteness (cf. von Heusinger 2011), is thus another criterion to be specific about.

2. Corpus Search for Modern StG sollen The goal of our search is all conjugated forms of the modal verb sollen. To achieve this, Nishiwaki (2013) used the weekly Die Zeit (Online edition) in the period of January 1 to December 31, 2011, restricting her search efforts on the rubric Literature. This corpus consists mainly of reviews of published books; additionally, it contains interviews with authors consequently giving room also to near-oral language, which we thought essential. The present chapter relies heavily on Nishiwaki’s (2013) findings.

2.1. Evaluation of the Corpus Search The corpus that was investigated contained in total 671 tokens of sollen, whereof 597 (89.0%) with root (deontic) readings; sixty three (9.4%) make up the evidential quotative (sollen in the use of “supposed to”); only eleven quotes (1.6%) are evidential sollte “should”. Since the latter use is rare, we excluded it from further discussion. As to the distribution of aspectuality of the infinitival complement, the corpus quotes classifiable as infinitival complements may be characterized as follows: The main distributive properties consist of the opposition [+te7

Brought to our attention by one reviewer. Our caveat in this context also applies to further distinctions of indefinites and specifics as come to mind thinking of von Heusinger (2011) and, more recently, Aguilar-Guevara et al. (2014).

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 251

lic] vs. [–telic]. The general position emanating from previous literature (Abraham 1991, 2002, 2005, Kotin 2008, 2010, 2012, Leiss 2011b, 2012) is not disconfirmed: predicative telicity prompts root (deontic) modal interpretation, whereas, inversely, atelicity elicits the evidential reading. See Table 1 for the findings from the corpus search. To determine the p(robability)-value of our frequencies we apply the Fisher’s-exact test/FET8.

[+telic] [–telic] Ȉ

root (deontic)

evidential

Ȉ

343 (65.6%) 180 (34.4%) 523

27 (42.9%) 36 (57.1%) 63

370 216 586

Table 1. Distribution of modal third person interpretations in the context of [±telicity] on the embedded infinitival complement. According to the FET, the two-tailed P value equals 0.0008. The association between rows (groups) and columns (outcomes) may be considered statistically extremely significant.

2.2. The Distribution of Contextual Definiteness and Indefiniteness Note that, although (6a) below attests to the claim that subject definiteness elicits deontic readings on the modal predicate. In addition, (6b), while definite, allows for an evidential reading in addition to a root one. Obviously, there is a caveat in place considering the fact that the definite denotations, very much in line with the clearly generic plural forms are functionally generics. However, we shall not go into any detail on this issue here, as any further distinction of generics (generic definite vs. bare plural nominals) has no bearing on our conclusions. (6) a. Elma, die Tochter eines armen Bauern, soll keinen Adeligen heiraten … D/*EMV ‘Elma, daughter of a poor farmer, is not supposed to get married to a nobleman.‘

8

The Fisher’s test is called an "exact" test, computing the "exact" two-sided (twotailed) P value. This calculator uses the method of summing small P values.

Chapter Seven

252

b. Elma, (die) Tochter eines armen Bauern, soll einen Adeligen heiraten … D/EMV ‘Elma, daughter of a poor farmer, is supposed to get married to a nobleman.‘ c. Elma, die Tochter eines armen Bauern, soll keinen Adeligen heiraten / kennen ‘Elma, daughter of a poor farmer, is not supposed to get married to / to know a nobleman.‘ (7)

Es soll nichts zu essen geben … ?D/EMV ‘The assumption is that there is nothing to eat.’

The account we proposed for (2) and (3) holds here in a similar way. Since (7) allows for an evidential reading, we may motivate *EMV in (6a) by the identifiability of the subject referent in co-construction with narrow VP-negation. However, the deeper reason for that empirical account is probably the agentivity on the predicate heiraten “marry”. Compare nonagentive predicates such as kennen “know”, which both clearly allows for a respective evidential reading in (6c). Thus, what counts as a definiteness marker is the definite article, demonstrative pronominals, personal pronouns, and possessive pronouns. Now let us look at what the corpus results reveal about the relation between (in-)definiteness and the two types of modal interpretations. deontic

evidential

Ȉ

[+definite] [–definite]

359 (68.6%) 164 (31.4%)

51 (81.0%) 12 (19.0%)

410 176

Ȉ

523

63

586

Table 2. The distribution of deontic and evidential readings triggered by [±definiteness] on the third person subject of the modal verb. FET: twotailed P value equals 0.0576 – association between row (groups) and columns (outcomes) is considered statistically significant.9 9 Given the fact that our present counts on MHG reflect a more or less even distribution of definite and non-definite subjects of deontic modal verbs, the fact that in StMG, there is a much more significant difference in the rate of definite subjects than of indefinite ones seems to indicate a major diachronic change. We do not think that the number of our corpus finds allows for definite conclusions.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 253

Table 2 shows that there is a clear affinity between the definite subject and an evidential reading of the predicative modal (81%). This is quite unexpected since, in terms of mereology, definiteness reflects an outer aspect (perfectivity), which, in turn, is known to be compatible with root modality (Abraham 1989, 1991, 2002). By contrast, indefiniteness reflects inner aspectuality (imperfectivity), which usually goes with evidentiality. The corpus findings contradict this logic on the evidential results. They do not as regards deontic modal readings. In the latter case, definite subjects line up more frequently with deontic readings, a result that would be in line with our mereological expectations. However, the figures to be compared are less convincing (69% vs. 31%).

2.3. Identifiability vs. Definiteness Let us take a step below definiteness and search in accordance with the criterion of (non-)identifiability on the subject referent. According to Lyons (1999: 6f.), a referent is identifiable not only if he is familiar to the hearer, but also if the hearer can find him because of his properties (close to the Kripke descriptor, i.e. ‘attributive reference’). (8) a. Günter Amendt […] starb am vergangenen Sonntag mit 71 Jahren infolge eines schweren Autounfalls. Der Unfallverursacher soll unter Drogeneinfluss gestanden haben. (14.03.) ‚G.A. […] died last Sunday at 71 in a heavy car accident. The person causing this is said to have been drugged.’ b. Und falls die Idee dahinter sein sollte, dass die Ukraine ins Glück finden könnte, … (22.04.) ‘And just in the case that it should be the idea that the Ukraine should be happy,…’ In (8a, b), identifiability is coded by the definite article in a combination of contextual anaphoricity and everyday knowledge.10 The illustration in (8b), for example, relates to the cataphoric use of the definite article. The subordinate clause introduced with dass “that” functions as the antecedent to the definite subject.11 10

Lyons (1999: 4) calls this use of the article “bridging cross-reference” or associative. 11 As with plenty of historical documents, one cannot exclude the possibility of reported speech and logophoricity is based on speaker’s beliefs in assessing (8b). Notice, however, that this would hardly limit the value of our comparisons as we pursue the exegetic tradition in the field of historical philology.

Chapter Seven

254

More generally, though, definite nominal phrases need not identify the referent. Among such definiteness markers is the neutral personal pronoun with impersonal verbs and the definite article in generic usage. Recall illustration (2a, b) above, where the criterion of definiteness as an identifiable entity played a major role in the identification of modal deonticity as opposed to evidentiality (cf. Section 1). Based on the criterion [±identifiable], the relation between subject reference and modality readings is quite different. See Table 3. deontic

evidential

Ȉ

[+identifiable] [–identifiable]

249 (47.6%) 274 (52.4%)

48 (76.2%) 15 (23.8%)

297 289

Ȉ

523

63

586

Table 3. Distribution of identifiability on subject referents on deontic and evidential readings. FET: two-tailed P value is less than 0.0001. Association between rows (groups) and columns (outcomes) is considered statistically extremely significant. The comparison of Table 3 with Table 2 yields that, for evidential readings, identifiable subject referents continue to be dominant. For the deontic readings, however, the distribution has decidedly changed: Roughly 30% of the definite nominal phrases share the feature [–identifiable].12 This yields the general picture of a close affinity of subject identifiability and modality. By contrast, the relation between deonticity and subject identifiability turns out to be insignificant. No results of significance were drawn with respect to subject (non-)identifiability and deonticity; see Table 3 above.

2.4. Individual-Level and Stage-Level Predicates There is a strong affinity between root-deontic modal verbs and telic infinitival complements. Yet, such infinitival complements co-construe also with evidential MVs when referring to an event completed not until in the future. See (9). 12

Justification: 359 individual counts collocate root modality and subject definiteness (Table 2). 249 counts have an individually identifiable subject (Table 3). The difference, i.e. 110 individual counts, are therefore unidentifiable subject nouns, i.e. 30% of the definite subjects.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 255

(9) Der Hermann dürfte (bis) morgen ein Buch auslesen (from Maché 2008: 410f.) ‘Herman should finish reading the book until tomorrow.’ Beyond that, deontic MVs may occur also with infinitival complements, i.e. with atelic complements, given that the verb’s stativity amounts to a temporary state (stage level property; cf. Maché 2008: 394ff.). See (10), which may have a deontic as well as an evidential reading. (10) Der Gerdsch muss gescheit sein ‘Gerdsch must be clever.’ The infinitival complement gescheit sein “be clever” denotes an individual-level (essential) property. 13 The MV muss “must” in (11) is evidential by preference. However, in certain contexts this individual-level predicate may receive a deontic reading. See (12). (11) Auf der Uni muss der Gerdsch dann gescheit sein. ‘At college, Gerdsch must be clever.’ One may paraphrase (12) doubly (Maché 2008: 395). (12) a. ‘When at college, Gerdsch must behave cleverly.’ b. ‘When Gerdsch will be studying he will necessarily become cleverer.’ Both deontic interpretations do not refer to the invariant characteristic of the subject person, but to his individual-level property. From such examples as above Maché draws the conclusion that, given definite singulative subject nominals, the deontic MVs do not tolerate individual-level predicates with an invariant characteristic. Under the subject-nominal conditions just discussed, the root-deontic and evidential interpretations of MVs are contingent upon Carlson’s feature opposition [+essential] vs. [–essential], with which the invariance vs. variance of the property of the subject referent on the individual-level predicate is described. 13 In Carlson’s (1980) terminology, individual-level (event) property relates to an invariant characteristic of a lexical predicate (‘essential predicate’). The opposed concept is the stage-level (event) property referring to a temporary state or property of the subject referent. Many grammatical phenomena are sensitive to this difference (cf. Kratzer 1995: 125).

256

Chapter Seven

In sum, it deserves maintaining that the two feature oppositions on the infinitival complements, [±telic] and [±essential] behave as reading-distinguishing context factors. As Maché (2008: 395) has shown, any indefinite marked subject nominal should be subject to the scrutinizing question whether the feature opposition [±essential] plays a role on state predicates with respect to the modal trigger on the complements. In the corpus analysis in the present discussion, the two features [±telic] and [±essential] will be co-considered on the tests for subject characteristics and their impact on modal readings.14

3. Summary and Conclusion 3.1. Definiteness Effects and Corpus Generalizations The result of our corpus investigation is challenging. What is it that lies at the bottom of subject definiteness and its power to contextually elicit evidentiality on co-constructional modal verbs? Let us pursue this question on (2a), repeated here as (13a), which may be reformulated as (13a') without any loss or addition of meaning. Note that the second sentence, (13a'), works as a modifying restrictive attribute of the indefinite nominal phrase. The set of books from which the speaker has to select one is subject to the constraint expressed by the relative clause. However, note that in contrast to (2a), repeated here as (13a), the paraphrase in (2b)/(13b) results in ungrammaticality as in (13b'): (13) a. Ich gehe mal ein Buch kaufen. ‘I’m going to buy a book.’ Das Buch soll von japanischer Kultur handeln. = (2a) ‘The book has to be about Japanese culture.’ a'. Ich gehe mal ein Buch kaufen, das von japanischer Kultur handeln soll. ‘I will buy a book which has to be about Japanese culture.’ b. Ich gehe mal das Buch kaufen. ‘I will buy the book.’ Das Buch soll von japanischer Kultur handeln. = (2b) b'. *Ich gehe mal das Buch kaufen, das von japanischer Kultur handeln soll. ‘I will buy the book which has to be on Japanese culture.’ 14 According to Nishiwaki (2013), for individual subjects typically deontic sollen is compatible with the feature [+essential] on individual-level predicates.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 257

Formal article definiteness as expressed in (13a), das Buch in the apodosis or relative clause, describes attributively, whereas in (13b) it is directly referential. The two readings of relative clauses, restrictive in (13a') vs. unrestrictive as in (13b'), can be told apart by means of insertion of modal particles, in our case ja. The relative clause in (13b') is possible only on the non-restrictive reading as in (13c). (13) c. Ich gehe mal das Buch kaufen, das ja von japanischer Kultur handeln soll.15 ‘I will buy the book, which as we all know is on Japanese culture.’ The paraphrase in (13c) suggests that in (13b) the existence of a book is presupposed and that something is asserted about this book. The modalized sentence in (13b) represents the structure “presupposition – assertion”.16 By contrast, the second sentence in (13a) has the function of an assertion. As the paraphrase in (13a') indicates it modifies the object in the first clause whose referent is new information for the hearer.

3.2. Definiteness Effects and Explanatory Generalizations There is good reason to assume that the information structural/IS result (‘new information for the hearer’) developed in Section 3.1 is typical of evidential contextuality and co-constructionality. Evidential MVs express the subjective appreciation of the proposition at the hands of the speaker. The speaker’s presupposition need not capture the entire asserted proposition since the hearer may know much of the assertion in the first place. It may capture just part of the assertion. In IS-terms the subject referent is an unmarked topic (cf. Lambrecht 1994: 131f.). This explains 15

By contrast, the relative clause in (13a') may only be restrictive. The nonrestrictive relative clause Ich gehe mal ein Buch kaufen, das ja von japanischer Kultur handeln soll is unacceptable, since non-sensical as the insertion of the modal particle/MP, ja, renders the relative clause presuppositional (Abraham 2012a). This is so because restrictive relative clauses are incompatible with MPs, whereas non-restrictive relatives allow for the insertion of MPs: viz. *JENER Hans, der ja/doch dort drüben wohnt, ist blind. “That Hans living PART over there is blind” vs. Hans, der ja/doch dort drüben wohnt, ist blind. “Hans living over there is blind.” This exceptionless complementary distribution is because of the foregrounded restrictive relative clauses and focused attributes, whereas nonrestrictive ones background and defocus attributes to the head noun. (cf. Abraham 2012a, Abraham 2012b: 80ff.) 16 As regards the terminology, we follow Lambrecht (1994: 52).

258

Chapter Seven

the dominance of subject definiteness in evidential contexts. In (14) we lay down once more what our discussion has yielded so far. (14) a. [CP DP[+Def] [Comp [left Middle field [right Middle field [VP (V) mussEMVfinite]]]]] b. [CP DP[-Def] [Comp [left Middle field [right Middle field [VP (V) mussDMVfinite]]]]] Based on the readings of the German MVs sollen and müssen, we expect identical results based on subject reference. Deontic MUSS-/SOLL- both mean strong commitment on the subject referent declared by someone outside of speaker and subject referent (weakened by past subjunctive, müsste, obviously for the reason that, since the commitment existed in the past and was not satisfied, its strength is weakened by past fact – as has been noted before). One may note cross-clausal restrictions on article words and pronominals as in (16a, b) – compare the respective subject determiners in (15a) as opposed to (15b). (15) a. Ich gehe mal ein BUCH kaufen. Es/Ein solches/*Dieses Buch muss von japanischer Kultur handeln. > DMV ‘I’ll be going to buy a book. It/Such a/This book must deal with Japanese culture.’ b. Ich gehe das Buch mal KAUFEN. Es/*Ein solches/Dieses Buch muss von japanischer Kultur handeln. > EMV ‘I’ll be going to buy the book. It/Such a/This book must deal with Japanese culture.’ Recall what we presupposed in terms of contextually elicited definiteness effects at the hand of the clausal subjects. (16) MÜSSEN/SOLLEN readings on the basis of subject reference as in (14a, b): a. singulative >> EMV as in (14a) – with individuality, singulativity Æ derived, speaker-perspectivized EMV b. property >> DMV as in (14b) – with unindividuality, nonsingulativity Æ original DMV Now, let us expand our distributional basis with (17)-(20). Notice that we test on the type of eventivity of the predicate (sehen “see; catch sight of”

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 259

with EXP-subject as opposed to einholen “run up to” with AGENT-subject), which yields the diverse modal readings after “>>”. Notice the diverging CLAUSAL ACCENTS signaling different information structural contexts. (17)

MÜSSEN/SOLLEN on THETA ROLE:

i. ii.

(18)

Er muss den Segler SEHEN17 / EINHOLEN >> EMV / DMV he must the sailing boat see / run up to Er MUSS den Segler sehen/einholen >> EMV-DMV / DMV-*EMV

MÜSSEN/SOLLEN THETA ROLE:

iii.

iv. v.

Object definiteness with EXPERIENCER/AGENT

on Object indefiniteness with EXPERIENCER/AGENT

Er muss einen SEGLER sehen/einholen he must a sailing boat see / run up to >> EMV-DMV / DMV-EMV Er MUSS einen Segler sehen/einholen >> EMV-DMV / DMV-*EMV Er muss EINEN Segler SEHEN/EINHOLEN >> DMV-EMV / DMV-*EMV

We draw the following interim conclusion as in (19a-d). (19) a. with Verum focus as in (18iv)18: only DMV is yielded b. with AGENT-subject on einholen “run up to”: DMV due to [+perfective]+AG c. with non-AGENT subject on sehen “see”: EMV due to [-perfective]+EXP d. all (i)-(v) irrespective of definiteness feature on DO

(20)

MÜSSEN/SOLLEN on

vi.

17

V-perfectivity: Er muss den Segler EINHOLEN >> DMV Perfective+Definiteness he must the sailing boat run up to >> DMV/*EMV

One reviewer’s critical remark prompts our explanation that sehen is telic on the reading of “erblicken; catch sight of” and atelic on its ongoing reading “keep seeing”. 18 Verum focus consists in prosodic emphasis on a grammatical (non-referential) item (as long as it can carry accent as an unbound morpheme). For a modern discussion of Verum focus see Lohnstein 2012, 2014, Lohnstein & Stommel 2009.

Chapter Seven

260

Er muss einen SEGLER einholen>>DMV Perfective+Indefiniteness >> DMV/*EMV The conclusion drawn based on (20) is (21). vii.

(21) with Perfectivity (= AGENT subject) Æ DMV19 irrespective of Definiteness (22)

MÜSSEN/SOLLEN on

V-imperfectivity: Er muss den Segler SEHEN >> EMV Imperfective+Definiteness >> EMV ix. Er muss einen SEGLER sehen >> EMV Imperfective+Indefiniteness >> EMV x. Er muss einen Segler SEHEN >> DMV/*EMV Imperfective+Indefiniteness >> DMV The conclusion drawn based on (22) is (23). (23) with imperfectivity (= subject non-AGENTIVITY) Æ EMV except with Verum focus on MV viii.

All of this together yields that theta role opposition [±AGENT] plays the deciding role, whereas [±DEFINITENESS] on the subject reference as well as [±ASPECT] on the predicate play a secondary, theta role-derived role. Notice that the predicate/event type is also contingent on the theta role of the main predicate. In other words, nothing bars us from concluding that predicative theta roles mediate what looks like definiteness effects in the first place. In addition, Verum focus always decides in favour of root meaning, i.e. in our case of DMV. Now, let us draw inferences on what we found to hold so far. Assume that VP with respect to the internal argument/iA plays a role, which is different from external argument/eA orientation. Then, in line with our findings so far on (i)-(x), this should be the case because of the secondary criteria, definiteness and aspect, while on the criterion of subject theta role 19

Clearly, telicity/perfectivity and agentivity are not always equivalent in terms of their eliciting force as the following illustration shows: Er muss das/ein Buch lesen “He must read the/a book.”>> DMV, where “lesen/read” is lexically atelic/imperfective. Possibly, the correlation between agentivity and deonticity can be explained based on agentivity triggering narrow scope of the modal verb. We shall not pursue this further at this point.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 261

the results should be in line with one another. Is that so? In (2a), with an unindividuated book (i.e., with ein Buch as a classeme= subject in the consequent clause), the yield is root modality, while with an individuated subject referent as in (2b) (i.e. das Buch as a single, definite, and identified subject referent) the result is evidential modality on the MV-predicate sollen/müssen “shall/must”. Let us have a look at the predicate handeln “act”: In (2a), with the class property reading on the indefinite subject, it is definitely more AGENTIVE than handeln in (2b) with the individuated subject referent in that there is a clear commitment laid on the book to be of a certain content. Let us say that the subject referent is a ‘derived committer’ in this case. By contrast, in (2b), where handeln predicates over a specific, individuated subject referent, the consequent clause has no option other than epistemize the MV given the individual identification of the referent in the antecedent clause. It/The/*This book is no longer a (derived) committer – one that appears to be possible only as long as this option can be counted as one out of several in a class. What counts as a definiteness effect on the MV (in terms of root vs. evidential reading) is thus whether the subject reference (in the consequent clause, in our specific case) is reference-open, or token-open, enough to allow for the root modal reading. This means that the subject reference is one of type or genericity or class membership, not of token or individuated status. What is at the bottom, however, is a modality hierarchy in the sense of ‘let root modality count (= be merged) as long as there are no constraining factors to be heeded – when the latter is the case “merge evidentiality”’. Clearly, in this case root readings are due to options of property – i.e. of class membership – rather than to time and referent individual, i.e. subjective, constraints. Note that something like the latter generalization lies at the bottom of the fact that in German MVs can be used as full verbs without infinitival embedding, but, though not necessarily, nominal objects instead: Er kann das – er muss jetzt – er darf das – ich will alles – er soll nur etc. Dutch is even more inventive on this phenomenon (Abraham 1989). Modal root readings are the default readings. If they were not, evidential uses would be possible instead of (20)-(23) – something which is not the case, however.

3.3. Agentivity on the Subject In what follows we add material testing our main hypothesis that agentivity on the subject triggers a deontic, or root, reading, whereas non-agentivity correlates with an evidentials reading. The material used stems from the Modern Standard German corpus that Nishiwaki (2013) was based on

262

Chapter Seven

(i.e. the weekly Die Zeit (online version, dated Jan. 1 to Dec. 31. 2011, column Literatur).The results are shown in the following table. evidential™ quotative [+agentivity] 238 (39.9%) 14 (22.2%) 252 [–agentivity] 359 (60.1%) 49 (77.8%) 408 ™ 597 63 660 Table 4. Distribution of the feature Agentivity on the subject nominal. FET: two-tailed P value equals 0.0062. Association between rows (groups) and columns (outcomes) is considered statistically very significant. Deontic

Table 4 confirms for MStG that for evidential readings the feature of nonagentivity is preferred. For deontic readings, we found the proportion to be less clear, with a slight numerical propensity for non-agentivity of root readings. This somewhat lopsided relation prompts the following remarks as background for our findings. As features for agentivity, we instantiated [+human] and [+intentional/voluntary] applied to both institutions and bodies of organizations as illustrated in (24)-(25). (24) Da sollen sich die Polizei und die Justiz drum kümmern. (8-11, 2011) ‘This the police and bodies of justice should take care of.’ (25) Eichborn sollte ursprünglich zum 1. Juli nach Berlin ziehen. (6-16) ‘E. was supposed to move to Berlin on July 1st.’

4. The MHG Modal Verb suln “shall” in Iwein – its Contextual Modal Triggers 4.1. Research Material So far we drew the conclusion that context factors such as aspectuality support specific readings on grammatical person and (in-)definiteness of the subject. In the previous Section 3.2, we highlighted the fact that the features [±telic/perfective] and [±definite] appear to play just a secondary role in the respective eliciting contexts, while [±agentive] appears to play a major role. In the present section on material from Middle High German/MHG (1050-1350 B.C.), a historical period preceding Modern StG (1550௅), we pursue the goal of unfolding those context factors, which correlate with

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 263

root and evidential modal readings on the MHG modal verb suln “shall”. The equivalence between MHG suln and MStE shall exists only by etymological comparison. Today’s use of MStE shall has lost much of its original modal-root use. On pursuing the use in MHG we intend to consider specifically the following criteria: aspectuality of the infinitival complement, the grammatical person of the subject, and the status of (in-) definiteness of the subject referent. It is the latter features that have been shown to own affinity with root and evidential modality (cf. Abraham 1989, 1991, Diewald 1999, Heine 1995, Nishiwaki 2013). In pursuit of the goals mentioned above we have to co-compute the well-known fact that sollen + infinitive is used to denote future tense in all older historical periods of German (both Old High German (750-1050) and Middle High German (1050-1350) and even later; cf. Ebert et al. 1993: §S163; DWB: XVI, 1488; Paul 2007: §S13; Sonderegger 2003: 335). It is for this reason that we shall lay focus on those corpus finds in which the temporal future reading is dissolved in favor of other, and specifically modal, readings. We listed and evaluated each instance of an inflected form of Middle High German/MHG suln “shall” in the verse novel Iwein, which consisting of some ca. 8200 verses was written around 1200. The text contains 206 suln-tokens, of which 19 tokens had no infinitival form. Of the 187 suln-tokens, 131 (70.0%) are documented in the present indicative. The subjunctive occurs only twice. We restrict our investigation to the present indicative since the change between deontic and evidential readings occur only in this mood and tense. Recall that we focus on the notion and use of evidential sollen only.20

20

This restriction by itself legitimates to some degree that we did not take up the detailed discussion of the different reference types of definiteness, genericity, and specificity (see BondaĜuk et al. 2014 and Aguilar-Guevara et al. 2014 as most recent overviews). In addition, it would not make sense to speculate on reference distinctions, either in terms of article forms or fine-grained semantic classes of the various nominals referred to in the MHG texts. A similar position has been adopted with respect to clause-structural distinctions of MHG and modern German. Clearly, V3 and many post-Vlast positions have been documented for MHG (Tomaselli 1995 and others). Likewise, discourse-related serializations played a far greater role in the written tradition of MHG than in normative modern German (Lenerz 1984, Axel 2007, Jäger 2008, to name but a few). Accordingly, it is not only legitimate, but also necessary, to compare the two language periods in greater detail. While, for example, we are quite aware of the exemplary investigation into the referential status of the numeral ein “one” in MHG (Donhauser & Petrova 2012), this and similar topics play no role in our own investigation.

Chapter Seven

264

4.2. Feature Affinities with Root/Deontic and Evidential-quotative Readings From the analysis of our MHG text Iwein we draw the conclusion that approximately 80% of the quotes in the Present Indicative have a deontic reading, whereas the frequency of the evidential quotative is less than 3%. See Table 5. evidentialquotative number of quotes 101 24 6 (percentage) (77.1%) (18.3%) (4.6%) Table 5. Distribution of readings on suln in the Present Indicative. Reading

Deontic

Future

As regards the notorious difficulty to distinguish temporal future from modal readings, the following is to be remarked. This particular difficulty may reside in certain instances of god taken to be the modal source. See (26) where the future reading receives preference. (26) dâ von sol sich mîn senediu nôt, therefore shall I my desired pain, ob got wil, unz an mînen tôt if God will, until my death niemer volenden: not and never lay off. (1811ff.: English translation by WA ) One may read (26) deontically since, although god’s will is hypothetical, the alternative is a future reading based on the temporal unz an mînem tôt “until my death”, which foregrounds the event before the meaning of volenden “complete; lay off”. Each of the illustrations suggesting future readings (eight in total) contain lexically telic or prefixed verbs. We conclude that suln contains little, if any, “purely temporal semantic potential” taking as a point of derivation the future implication that emanates from telic, and perfective, predicates and which does not have to be encoded by an extra auxiliary. See (27) as a concrete testimony for this assumption. (27) mir sol des strîtes vür komen Sir Gawein will mîn her Gâwein: be first in our battle. des enist zwîvel dehein, There is no doubt als schiere so er des strîtes gert, that, as soon as he wishes to fight, ern werdes vür mich gewert. he will have this right before me. (914-918: English translation by WA based on Cramer 2001)

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 265

According to Paul (2007: §S13) the modal verb in line 1, sol “shall”, marks future tense. However, it might just as well stand for the speaker’s assumption about who can start the fight. Yet, generally, clear future uses are rare to find (Behaghel 1924: §689). According to our count, about 80 % of the suln-tokens have a deontic interpretation in Iwein. The modal source can be third person and the speaker as in (28) and (29). (28) er sprach նich sol um mitten tac He said: ࡐI must around noon time morgen komen an eine stat tomorrow come to a place dar mich ein vrouwe komen bat where a lady asked me to come, diu mir vil gedienet hât, who offered me a great service. (4742-4745: English translation by WA based on Cramer 2001) (29) sus wart dâ urloup genomen zeinem ganzen jâre. ouch swuor sî des, zewâre, und beliber iht vürbaz, ez wære iemer ir haz. […] Sî sprach ‘iu ist daz wol erkant daz unser êre und unser lant vil gar ûf der wâge lît, ir enkumt uns wider enzît, daz ez uns wol geschaden mac. hiute ist der ahte tac nâch den sunewenden: dâ sol daz jârzil enden. sô kumt benamen ode ê, ode ichn warte iuwer niht mê.

This he took farewell for a whole year to ride off. However, she swore that, would he stay away longer, she would take a grudge at him forever. She said: ‘You know precisely that our honor and our country are at stake, and that, if you do not return on time, great misfortune may come about. Today is the eighth day after solstice, when the deadline is over. Do come then or earlier, or else I would not expect you any longer. (2924ff.: our translation relating to Cramer 2001)

The two quotes, (28) and (29), show beyond doubt that it was ‘she’ herself that set the time thus making clear that she is the modal source.21 In (28), the third person is the modal source, whereas in (29) the third person refers to the speaker. As for indirectly coded definiteness effects we shall take up contexts of aspectuality on the embedded infinitival complements. 21

To be sure, ‘she’ stands for two non-identical persons: in (28) it refers to a court woman, while in (29) it is the queen. However, this does not touch upon our conclusions.

266

Chapter Seven

4.3. The Distribution of Aspectuality of the Infinitival Complement Work by Abraham (1989: 145f., 1991: 109, 115), Diewald (1999: 257ff.), Heine (1995: 25f.), Kotin (2010, 2012) and Maché (2013) has demonstrated beyond doubt that the root reading on modal verbs is triggered by the telicity of the entire infinitival complement. This was confirmed for present StG by a corpus-supported case study (Table 1 in Section 2.1). Identical context dependence has shown to hold for atelicity and evidential-quotative readings. As before, the reason for including aspect in this study on definiteness effects is that mereologically referential definiteness shares with perfective aspect the outer event view, while, image-reflecting, referential indefiniteness is on a par with the inner viewpoint (Smith 1997). Iwein-MHG appears to share with StG the affinity of modal deonticity with aspectual telicity on the infinitival complements. Two thirds of the Iwein-quotes in deontic usage correlate with telic infinitive complements. Inversely, there is no direct and statistically relevant co-constructional link between the evidential-quotative reading atelicity. See Table 6. deontic evidential-quotative22 [+ telic] 69 (68.3%) 5 (83.3%) [– telic] 32 (31.7%) 1 (16.7%) ™ 101 (100%) 6 (100%) Table 6. Distribution of aspect on the infinitival complement.23 FET: twotailed P value equals 0.6638. Association between rows (groups) and columns (outcomes) is considered not statistically significant.

22 It is often cumbersome to decide on the epistemic status of the modals. Even the authors of the present paper may want to decide differently on second or third readings. Given this difficulty, we hang on to our first decisions, which led to the present frequency percentages. 23 We feel that the following is a necessary comment prompted by Zeman’s critical remark about our identification of the infinitival complements as epistemic quotative readings since in atelic contexts: The predicates in questions are komen “come”, geschehen (twice) “occur”, zebrechen “break up”, stechen “prick, sting”, and diezen “sound hard”. We take them to be telic except for the latter to be atelic. However, other than Zeman, we take stechen “prick, sting” to be a telic verb with its occurrence in (30) below meaning “hurting by stabbing/with one stab” rather than an atelic “be pricking, unpleasant on touching“ such as through thistles.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 267

4.4. The Distribution of the Deictic and the Grammatical Person There is a fundamental difference between the first and second person markings, on the one hand, and the third person marking. The latter is the only truly grammatical person in the sense that the identification of the referent is context based by way of anaphoricity (or, more rarely, cataphoricity). The precise reference of the two first persons (I – you – we – you) in what is usually seen as belonging to one common paradigm with the third person in the older and modern grammars, is not grammatically retrievable. Their reference is identifiable only by deictic means (Diewald 1991).24 In our corpus of MHG Iwein, the third grammatical person is overwhelmingly more frequent in both the deontic and the evidentialquotative readings. Third persons – and this is why we have added this section in the present discussion – bring about definiteness effects to the extent that they share anaphoricity with the definite article and other determiners. This finding from the MHG corpus search thus turns out to be in line with our search on modern StG in the previous section. deontic evidential-quotative 1st sg/pl 34 (33.7%) 0 2nd sg/pl 20 (19.8%) 1 (16.7%) 3rd sg/pl 47 (46.5%) 5 (83.3%) ™ 101 (100%) 6 (100%) Table 7. Distribution of the deictic and grammatical person. FET1-2: P value 0.3818 – not statistically significant // FET2-3: P value 0.6660 – not statistically significant // FET1-3: P value 0.1519 – not statistically significant. Comparison with the finds from modern German newspapers with the present MHG text (Nishiwaki 2013, Table 2) yields that in “Iwein” first and second person marking occurs far more often. This is due to the frequent dialogues in the MHG epic Iwein.

24

No doubt, as one reviewer points out, there are also generic contexts elicited by the second person singular in German as in Da kannst du einfach nichts machen! “There is nothing you can do.” Here we stick to the general and default identifiability in terms of factual deixis.

268

Chapter Seven

4.5. Definiteness Marked on the Subject Recall our illustrations in (2) and (13), which led to our conclusion as to the link between referential codes on the subject nominal and the root-deontic/evidential-quotative readings of sollen (parallel to Leiss (2011a), Heine (1995: 21), and Johannesson (1976: 51), the latter on English must). At this point, we take up historical material from the MHG Iwein corpus. The question is whether the referential subject features used for mStG, [+definite] and [+identifiable], respectively, are elicited by MHG suln, which is the equivalent of StG sollen. First and second (non-grammatical) persons will not be considered, since their reference is outside of anaphoric retrievability. In modern mStG, the collocative distribution of (in-)definite subject and DMV is different from that in MHG. The number of definite subject nominals in mStG is twice as high as that of indefinite ones (see Table 2). As in mStG, the discriminating subject features, [±definiteness], have an almost convergent distribution. See also Table 8. Thus, no changes on modal deonticity are relevant reflecting a result identical with mStG. The distribution is identical under the criterial feature of [±identifiability]. See Table 9. deontic evidential-quotative [+ definite] 25 (53.2%) 5 (100%) [– definite] 22 (46.8%) 0 ™ 47 (100%) 5 (100%) Table 8. Distribution of definiteness on the subject nominal in the 3rd person. FET: P value 0.0650 – not statistically significant. Deontic evidential-quotative [+ identifiable] 22 (46.8%) 3 (60%) [– identifiable] 25 (53.2%) 2 (40%) ™ 47 (100%) 5 (100%) Table 9. Distribution of identifiability on the subject nominal in the third person. FET: P value 0.6624 – not statistically significant. As Table 9 shows, the 3rd person evidential quotative occurs only five times in Iwein. The respective subject nominals are represented by the weak demonstrative pronoun der, functionally relative “who” or the personal pronoun er “he”. Both are definite because of their presuppositional status. Are these definite expressions in line with their referents? See (30).

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 269

(30) swer iuch mit lêre bestât,

Whoever wishes to impress you with admonition, deist ein verlorniu arbeit. his efforts will be worth nothing. irn sult iuwer gewonheit You will not let go your habits durch nieman zebrechen. for anyone else. der humbel der sol stechen: The humble bee stings, ouch ist reht daz der mist wherever there is dung, stinke swâ der ist: it stinks there, der hornûz der sol diezen. and the hornet is humming. (202-209: translation WA based on Cramer 2001)

The antecedent nominals of the two demonstrative pronouns in (30) – der humbel “the humble bee” and der hornûz “the hornet” – are generic functions of the definite codes. Therefore, the two relative pronouns do not refer to any contextually identifiable referent. However, the remaining three personal and relative pronouns do: Their reference is identifiable on the codes of definiteness. It is true that the two instances of sult, sol “should, shall”, 2nd and 3rd persons, in (30) above are not evidential quotatives in the narrow sense since the information delivered by the propositions does not stem from unknown sources. 25 However, what may just as well trigger quotative readings is when you take the quotes to be in the 2nd person, which would indeed make an evidential-quotative reading plausible. Consider the verse translation “You will not let go your habits for anyone else”, which is compared with the behavior of the humble bee’s or the hornet’s habit of stinging and supports the quotative interpretation. While not a quotative in the stricter sense, the information about the proposition appears to be due to common sense and common knowledge. One may assume that it represents an uncompleted state of grammaticalization toward a full evidential quotative.26 25

There may not be an epistemic-quotative reading in place here (comment by Sonja Zeman). See fn. 9 with respect to this uncertainty. 26 According to empirical evidence documented by Fritz (1997: 46), Müller (2001: 244ff.), and Zeman (2013), quotative readings of sollen can be found all along since MHG. (i.e. as early as from 1050 onwards). This is in solid contradiction to Traugott (1989), Diewald (1999), Gamon (1993), Nordlinger & Traugott (1997) as well as others, who claim that the epistemic readings derive from their early root stages through grammaticalization and subjectification. See the following MHG lines, for example (gleaned from Diewald 1999: 420). sölch ellen was ûf in gezalt: sehs ritter solter hân gevalt,

270

Chapter Seven

As before in mStG, the evidential-quotative reading of suln is elicited by the definiteness code of the clausal subjects and, in a majority of instances, with its feature of identifiability.

5. Summary and Prospective View Our discussion of specific DEs in mStG and MHG has prompted one main result. The context factors that yield a clear affinity between modal and reference codes in mStG are relevant also in the older period of MHG – several hundreds of years earlier. Telicity on the infinitival complement elicits a modal deontic reading. In addition, theta role of eventive [±Agentivity] plays another deciding role, whereas [±Definiteness] on the subject reference as well as [±Aspect] on the predicate play a secondary, theta-role derived role. This is the hardest result in our discussion. Recall that aspectual inferences are based on the mereological of outer and bounded referentiality, just like definiteness. The converse relation, that between atelicity and the evidential-quotative reading (both sharing the inner/unbounded event characteristic) on the MV sollen was not confirmed, neither in MHG nor in mStG, although somewhat better recognizable in the latter period – see Table 1. The grammatical subject feature of the third person pronouns is dominant in mStG as well as in MHG triggering definiteness effects in terms of disambiguating the fundamental polyfunctionality of MVs. To be true, the MHG MV sollen confirms affinity between modal evidential-quotative uses and the definiteness and identifiability of the subject referent. However, this result rests on too small an empirical basis to warrant a solid conclusion on definiteness effects in modal contexts. MHG suln has no role to play in the relation between the subject properties of (in-)definiteness and (non)identifiability. Although mStG sollen is linked to a majority of definite codes, many of them are at a far distance from something individual and presupposed rendering no more than half of them nonidentifiable (Tables 2 and 3 in Section 2.2 and 2.3). Table 10 reflects our results in compact form. The indication Indirect DEs means that the DEs are mediated by other grammatical contingencies such as co-constructional aspect in ways explainable on a mereological basis. The index Secondary DEs relates to the fact that for mStG we saw reason to relate modal disambiguation to subject theta properties rather

die gein im kœmen ûf ein velt. (Parzival IV, 197, 17ff.) ‘Such fighting zeal was ascribed to him: six knights he was said to have killed who stood up against him in a field.’ (translation WA)

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 271

than subject reference properties. However, for the corpus-based facts in MHG we fell back on factors of subject nominal reference. Correlation between telicity and deontics – Indirect definiteness effects Correlation between atelicity and evidentiality – Indirect definiteness effects Dominance of the 3rd person on both readings – Secondary definiteness effects Correlation between subject theta and evidentiality – Secondary definiteness effects Correlation between identifiability and evidentiality – Secondary definiteness effects Correlation between (non)identifiability and deontics – Secondary definiteness effects

MHG

mStG

yes

Yes

no

Yes

yes

Yes

yes

No

(yes)

Yes

no

No

Table 10. Reading affinities and definiteness effects on the basis of contextual factors in MHG vs. modern StG. In conclusion, we add that what is still open is an investigation into the historical periods of Old High German (750-1050) and Early New High German (1350-1650). For one, these are the historical periods completing the history of German from scratch. In addition, second, this would confirm whether the longitudinal result, which our discussion suggests, would receive its final and lasting value.

6. Caveat – and Open Issue There is a dilemma. Our conclusion about what the corpus counts on mStG and MHG yield with respect to the relation between subject definiteness and modal evidentiality on the predicate, on the one hand, and earlier solid accounts of the relation between imperfective aspect and modal evidentiality in the verbal complex, on the other hand, are not compatible with one another. In addition, the relation between verbal aspect and co-constructional verbal modality is symmetrical between and . By contrast, the force of indefiniteness on the subject referent has no triggering equivalent on the modal side ( has no support by the corpus counts). It seems that all one may say is that, syntactically speaking, the two phenomena have nothing in common. The obvious divergence in mereological terms ( as opposed to ) may be rooted in the syntax-categorial differences. The subject-predicate relation anchored in nominal (in-)definiteness and subject-predicate agreement, assumedly, anchored in T/INFL, thus represents an N:V relation. In contrast, the finite-MV: infinitival-V relation is one under direct c-command in the lowest layer of VP, which in AuxOV-German is far away from the syntactic subject node. Another vexing issue is why the N:V relation prompts definiteness effects on evidential modality, but not on root modality. And the third issue is whether accounts on the empirical basis of the modal sollen can be transferred to each of the other modal verbs in German (and possibly Germanic) in just the same way. We want to leave it at this diagnostic statement. It appears to be a puzzle for the time being. In fact, we believe we have no other choice.

References Primary References Hartmann von Aue. Iwein. 4., überarbeitete Auflage. Text der siebenten Ausgabe vou G. F. Benecke, K. Lachmann and L. Wolff. Übersetzung und Nachwort von Thomas Cramer. Berlin: de Gruyter 2001. Hartmann von Aue. Iwein. Aus dem Mittelhochdeutschen übertragen, mit Anmerkungen und einem Nachwort versehen von Max Wehrli. Zürich: Manesse 1988. Wolfram von Eschenbach. Parzival. Studienausgabe. 2. Auflage. Mittelhochdeutscher Text nach der sechsten Ausgabe von Karl Lachmann. Übersetzung von Peter Knecht. Mit Einführungen zum Text der Lachmannschen Ausgabe und in Probleme der ‚Parzival‘Interpretation von Bernd Schirok. Berlin / New York: de Gruyter 2003.

Secondary References Abraham, Werner. 1989. Syntaktische Korrelate zum Lesartenwechsel zwischen epistemischen und deontisch-volitiven Modalverben. Groninger Arbeiten zur germanistischen Linguistik (GAGL) 30: 145-166. —.1991. Modalverben in der Germania. In Eijiró Iwasaki (ed.). Begegnungen mit dem „Fremden“. Grenzen – Tradition – Vergleiche. Akten des VIII. Internationalen Germanistenkongresses in Tokyo 1990. Vol. 4, 109-118. München: iudicium.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 273

—. 2002. Modal verbs: evidentials in German and English. In Sj. Barbiers, F. Beukema & W.v.d. Wurff (eds.). Modality and its interaction with the verbal system [Linguistik Aktuell/Linguistics Today 47], 19-50. Amsterdam: John Benjamins. —. 2005. Event arguments and modal verbs. In C. Maienborn & A. Wöllstein (eds.). Event arguments: Foundations and applications [Linguistische Arbeiten 501], 243-276. Tübingen: M. Niemeyer. —. 2012a. Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. derivational structure? In Werner Abraham & Elisabeth Leiss (eds.). Modality and theory of mind elements across languages [Trends in Linguistics. Studies and Monographs 243], 67-108. Berlin: de Gruyter. —. 2012b. Covert patterns of modality in typological comparison. In Werner Abraham & Elisabeth Leiss (eds.). Covert patterns of modality, 386-439. Newcastle upon Tyne: Cambridge Scholars. —. 2013. Deutsche Syntax im Sprachenvergleich. Grundlegung einer typologischen Syntax des Deutschen [Studien zur deutschen Grammatik 41]. Tübingen: Stauffenburg. —. 2015 (submitted). Thetic sentences at the level of modality - an exercise in modal event quantification. Universities of Munich and Vienna July 5, 2014. Abraham, Werner & Elisabeth Leiss (eds.). 2009. Modalität. Epistemizität und Evidentialität bei Modalverb, Adverb, Modalpartikel und Modus. [Studien zur deutschen Grammatik 77]. Tübingen: Stauffenburg. —. (eds.). 2012. Modality and theory of mind elements across languages [Trends in Linguistics. Studies and Monographs 243]. Berlin: de Gruyter. Aguilar-Guevara, Ana, Bert Le Bruyn & Joost Zwarts (eds.). 2014. Weak referentiality [Linguistik Aktuell/Linguistics Today 219]. Amsterdam: John Benjamins. Axel, Katrin. 2007. Studies on Old High German syntax. Left sentence periphery, verb placement and verb-second [Linguistik Aktuell/Linguistics Today 112]. Amsterdam: John Benjamins. Bach, Emmon. 1981. On time, tense, and aspect: An essay in English metaphysics. In Peter Cole (ed.). Radical pragmatics, 63-8l. New York: Academic Press. Behaghel, Otto. 1924. Deutsche Syntax. Eine geschichtliche Darstellung. Bd. II. Heidelberg: Winter.

274

Chapter Seven

BondaĜuk, Anna, Gréte Dalmi & Alexander Grosu (eds.). 2014. Advances in the Syntax of DPs. Structure, agreement, and case [Linguistik Aktuell/Linguistics Today 217]. Amsterdam: John Benjamins. Carlson, Gregory N. 1980. Reference to kinds in English. Revised version of authorૅs doctoral thesis, University of Massachusetts, Amherst, 1977. New York: Garland Publishing. Diewald, Gabriele. 1991. Deixis und Textsorten im Deutschen [Reihe germanistische Linguistik 118]. Tübingen: Niemeyer. —. 1999. Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität [Reihe Germanistische Linguistik 208]. Tübingen: Niemeyer. Donhauser, Karin & Svetlana Petrova. 2012. Sprachliche Strategien zur Aktivierung und Deaktivierung von Diskursreferenten in deutschsprachigen Texten des Mittelalters. In I. Kasten (ed.). UnVerfügbarkeit. Paragrana. Internationale Zeitschrift für Historische Anthropologie 21.2: 159-176. DWB = Deutsches Wörterbuch von Jacob und Wilhelm Grimm. 18541971. 33 Bände. Nachdruck der Erstausgabe. München: Deutscher Taschenbuch Verlag. 1984. Ebert, Robert Peter, Oskar Reichmann, Hans-Joachim Solms & KlausPeter Wegera. 1993. Frühneuhochdeutsche Grammatik. Tübingen: Niemeyer. Ehrich, Veronika. 2001. Was ‘nicht müssen’ und ‘nicht können’ (nicht) bedeuten können: Zum Skopus der Negation bei den Modalverben des Deutschen. In R. Müller & M. Reis (eds.). Modalität und Modalverben im Deutschen [Linguistische Berichte Sonderheft 9], 149–176. Hamburg: Buske. Fabricius-Hansen, Cathrine & Kjell J. Sabø. 2004. In a meditative mood: the semantics of the German reportive subjunctive. Natural Language Semantics 12: 213–257. Fritz, Gerd. 1997. Historische Semantik der Modalverben. Problemskizze – exemplarische Analysen – Forschungsüberblick. In G. Fritz & Th. Gloning (eds.). Untersuchungen zur semantischen Entwicklungsgeschichte der Modalverben im Deutschen [Reihe Germanistische Linguistik 187], 1-157. Tübingen: Niemeyer. Gamon, David. 1993. On the development of epistemicity in the German modal verbs mögen and müssen. Folia Linguistica Historica 14: 125176. Glas, Reinhold. 1984. sollen im heutigen Deutsch: Bedeutung u. Gebrauch in der Schriftsprache [Studien zur deutschen Grammatik 27]. Tübingen: Narr.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 275

Heim, Irene. 1988. The semantics of definite and indefinite noun phrases. New York: Garland. Heine, Bernd. 1995. Agent-oriented vs. evidential modality. Some observations on German modals. In J. Bybee & S. Fleischman (eds.). Modality in grammar and discourse [Typological Studies in Language 32], 17-53. Amsterdam: John Benjamins. Hollebrandse, Bart, Angeliek van Hout & Co Vet (eds.). 2005. Crosslinguistic views on tense, aspect, and modality [Cahiers Chronos 13]. Amsterdam: Rodopi. Hout, Angeliek van. 1998. Event semantics of verb frame alternations: A case study of Dutch and its acquisition. New York: Garland. Jäger, Agnes. 2008. History of German negation [Linguistik Aktuell/ Linguistics Today 118]. Amsterdam: John Benjamins. Johannesson, Nils-Lennart. 1976. The English modal auxiliaries. A stratificational account. Stockholm: Almqvist & Wiksell international. Karttunen, Lauri. 1976. Discourse referents. In J. McCawley (ed.). Syntax and Semantics 7: Notes from the Linguistic Underground, 363–385. New York: Academic Press. Kotin, Michail L. 2008. Zu den Affinitäten zwischen Modalität und Aspekt: eine germanisch-slavische Fallstudie. Die Welt der Slaven LIII/1: 116-140. —. 2010. Zur referenziellen Identität von Tempus- und Modusformen. In A. Kątny & A. Socka (eds.). Modalität / Temporalität in kontrastiver und typologischer Sicht, 29-38. Frankfurt a. M.: P. Lang. —. 2012. Preterito-prezentnyje glagoly i razvitije kategorii modal´nosti v germaĔskich jazykach [Preterit presents and the development of the category of modality in the Germania]. In D. B. Nikuliþeva (ed.). Dinamiþeskije procesy v germaĔskich jazykach. Materialy ýetvjortych lingvistiþeslich þtenij pamjati V. N. Jarcevoj, 19-29. Moskva – Kaluga: IP Šilin I. V. (Izdatel’stvo „Eidos”). Kratzer, Angelika. 1995. Stage-level and individual-level predicates. In G. N. Carlson & F. J. Pelletier (eds.). The generic book, 125-175. Chicago: The University of Chicago Press. Lambrecht, Knud. 1994. Information structure and sentence form. Topic, focus, and the mental representations of discourse referents [Cambridge Studies in Linguistics 71]. Cambridge: CUP. Leiss, Elisabeth. 2000. Artikel und Aspekt. Die grammatischen Muster von Definitheit [Studia linguistica Germanica 55]. Berlin: W. de Gruyter. —. 2011a. Abbau des Artikels im Gegenwartsdeutschen. Lecture at German-Japanese Summer School for Linguistics. August 10, 2011. Ms. LMU Munich.

276

Chapter Seven

—. 2011b. Lexikalische versus grammatikalische Epistemizität und Evidentialität: Plädoyer für eine klare Trennung von Lexikon und Grammatik. In G. Diewald & E. Smirnova (eds.). Modalität und Evidentialität. Modality and evidentiality [Fokus. LinguistischPhilologische Studien 37], 149-169. Trier: Wissenschaftlicher Verlag. —. 2012. Epistemicity, evidentiality, and theory of mind (ToM). In W. Abraham & E. Leiss (eds.). Modality and theory of mind elements across languages [Trends in Linguistics. Studies and Monographs 243], 39-65. Berlin: de Gruyter. Lenerz, Jürgen. 1984. Syntaktischer Wandel und Grammatiktheorie: eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Leonetti, Manuel. 2004. Definiteness effects and the role of the coda in existential constructions. Ms. University of Alcalá. Internet version. Lohnstein, Horst. 2012. Verumfokus – Satzmodus – Wahrheit. In H. Lohnstein & H. Blühdorn (eds.). Wahrheit – Fokus – Negation [Linguistische Berichte Sonderheft 18], 2-37. Hamburg: Buske. —. 2014. Verum focus. In C. Fery & S. Ishihara (eds.). Handbook of Information Structure, 1-29. Oxford: Oxford University Press. Lohnstein, Horst & H. Stommel. 2009. Verum focus and phases. In K. Grohmann & P. Panageotidis (eds.). Linguistic Analysis 35 [Special Issue. Phase Edge Investigations], 109–140. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Maché, Jakob. 2008. The autopsy of a modal – insights from the historical development of German. In W. Abraham & E. Leiss (eds.). Modalityaspect interfaces. Implications and typological solutions, 385-415. Amsterdam: John Benjamins. —. 2013. On black magic – how evidential modifiers emerge. Unpublished Ph.D. Dissertation FU-Berlin. Maienborn, Claudia. 2003. Die logische Form von Kopulasätzen [studia grammatica 56]. Berlin: Akademieverlag. Mortelmans, Tanja. 2003. The ‘subjective’ effects of negation and past subjunctive on deontic modals. The case of German dürfen and sollen. In F. Lenz (ed.). Deictic conceptualisation of space, time and person [Pragmatics & Beyond New Series 112], 153-182. Amsterdam: John Benjamins. Müller, Reimar. 2001. Modalverben, Infinitheit und Negation im ProsaLancelot. In R. Müller & M. Reis (eds.). Modalität und Modalverben im Deutschen [Linguistische Berichte Sonderheft 9], 239-261. Hamburg: Buske.

German sollen/suln “shall” – Definiteness Effects in Verbal Complements 277

Nishiwaki, Maiko. 2013. Modalverben und die (In)Definitheit des Subjekts – unter besonderer Berücksichtigung von sollen. In W. Abraham & E. Leiss (eds.). Funktionen von Modalität [Linguistik – Impulse & Tendenzen 55], 231-248. Berlin: de Gruyter. Nordlinger, Rachel & Elizabeth Traugott. 1997. Scope and the development of evidential modality: evidence from ought to. English Language and Linguistics 1.2: 295-317. Paul, Hermann. 2007. Mittelhochdeutsche Grammatik. 25th edition. Neu bearbeitet von Th. Klein, H.-J. Solms & K.-P. Wegera. Mit einer Syntax von Ingeborg Schöbler, neubearbeitet und erweitert von H.-P. Prell. Tübingen: Niemeyer. Schenner, Mathias. 2008. Double face evidentials in German: Reportative ‘sollen’ and ‘wollen’. In A. Gronn (ed.). Embedded Contexts. Proceedings of SuB12, 552–566. Oslo: ILOS. Smith, Carlotta M. 1997. The parameter of aspect. First edition 1991. Dordrecht: Kluwer Academic Press. Sonderegger, Stefan. 2003. Althochdeutsche Sprache und Literatur. Eine Einführung in das älteste Deutsch. Darstellung und Grammatik. 3rd, revised and considerably expanded edition. Berlin: de Gruyter. Tomaselli, Alessandra. 1995. Cases of verb theird in Old High German. In Adrian Battaye & Ian Roberts (eds.). Clause structure and language change, 345-369. New York: Oxford University Press. Traugott, Elizabeth Closs. 1989. On the rise of evidential readings in English: an example of subjectification in semantic change. In Language 57: 33-65. —. 2003. From subjectification to intersubjectification. In R. Hickey (ed.). Motives for language change, 124-142. Cambridge: CUP. von Heusinger, Klaus. 2011. Specificity. In K. von Heusinger, C. Maienborn & P. Portner (eds.). Semantics: An International Handbook of Natural Language Meaning Vol 2, 1024–1057. Berlin: de Gruyter. Zeman, Sonja. 2013. Zur Diachronie der Modalverben: sollen zwischen Temporalität, Modalität und Evidentialität. In W. Abraham & E. Leiss (eds.). Funktionen von Modalität [Linguistik – Impulse & Tendenzen 55], 335-366. Berlin: de Gruyter. Zucchi, Alessandro. 1995. The ingredients of definiteness and the definiteness effect. In Natural Language Semantics 3: 33-78.

CHAPTER EIGHT THE EVOLUTION OF DEFINITENESS EFFECTS WITH FRENCH IL Y A FROM 1300 TO TODAY CHARLOTTE COY

1. Introduction1 French il y a is a classical instance of the so called existentials (for a definition see McNally 2011) and shows a definiteness effect in that it is much more frequently employed with an indefinite noun phrase than with a so-called “strong noun phrase”, i.e. a noun phrase determined by a definite determiner, a demonstrative determiner, a possessive determiner, a universal quantifier, a strong personal pronoun or a proper noun (for a general discussion of the definiteness restriction in French see among others Cannings 1978, Etchegoyhen & Tsoulas 1998, Martin 2002, Dobrovie-Sorin & Beyssade 2012: 103-126). 2 However, strong noun phrases are not totally excluded after il y a (1). (1)

Il y avait la Seine en bas. ‘There was the Seine below.’

(E. de Goncourt, 1861)

In a synchronic corpus study on newspaper articles, Meulleman (2012) counted 14% of strong noun phrases after il y a and observes that French il y a is much more permissive than Spanish hay with regard to the definiteness restriction (cf. also Leonetti 2008). She takes this to be one sign for a more advanced grammaticalization of il y a, as compared to

 1

Many thanks to Amelie Meixner, Bilke Ullrich, Christian Rödelsperger, Petr Makarov and Teresa Fay for helping me check and classify the data and for computational support. This article is dedicated to the memory of Prof. Dr. Peter Koch (1951-2014). 2 For ease of discussion, I will also use the term “strong determiners” to refer to the whole group of determiners and other strong noun phrases listed here.

Evolution of Definiteness Effects with French il y a from 1300 to Today 279

Spanish hay and Italian c’è (Meulleman 2012). With regard to earlier stages of French, however, we know almost nothing about the definiteness effect. Il y a goes back to the Vulgar Latin existential construction habet and has been present at all stages of French (Maillard 1985: 73-74). Marchello-Nizia (1997: 139) states that in the Middle French period, (il) (y) a is generally followed by an indefinite noun phrase, but she does not elaborate further on the issue. This diachronic corpus study thus aims to look in more detail into the dynamics of the definiteness effect.

2. Four Categories of Exceptions from the Definiteness Restriction On the basis of the literature on the definiteness effect in French and other languages, I adopted four categories of exceptions to the definiteness effect. These categories most succesfully accounted for the classification of the corpus data. The clearest case of exception is the superlative (2). Although formed with a definite determiner it occurs easily after il y a or other impersonal constructions and has been acknowledged as one exception to the definiteness restriction in French, Spanish, and English (Suñer 1982: 79-82, Abbott 1993, Ward & Birner 1995: 736-737, Etchegoyhen & Tsoulas 1998, McNally 2011). (2)3

S' il y a le moindre doute If EXPL there has the least doubt ‘If there is the least doubt’ (G. Flaubert, Correspondance, 1868)

The second class of exceptions are cataphoric noun phrases (3), which are also described as exceptions to the definiteness effect in the literature on French, Spanish, and English (Suñer 1982: 82-84, McNally 2011, Ward & Birner 1995: 735-737, Etchegoyhen & Tsoulas 1998). (3)

il y avait les lignes suivantes, [...]: "les […] EXPL there had the lines following ‘There were the following lines [...]: “the great wars [...]’ (A. Dumas Père, Le comte de Monte Christo, 1846)

 3

All examples are taken from Frantext. I use the following abbreviations: EXPL= expletive subject pronoun, REFL= reflexive pronoun, SUBJ= subjunctive.

280

Chapter Eight

These first two classes are fairly clear-cut and the literature agrees on the less definite or less referential character of the noun phrases. The third class of exceptions can roughly be described as containing list reading and reminder cases (4). Like superlatives and cataphoric noun phrases, list reading and reminder cases are well established exceptions to the definiteness restriction for French, Spanish, and English (Suñer 1982: 86-90, Abbott 1993, Ward & Birner 1995, McNally 2011, Dobrovie-Sorin & Beyssade 2012: 110-111). (4)

[Speaking about which woman to marry]: Il y a la Louise, la Sylvaine la Claudie EXPL there has the Louise the Sylvaine the Claudie ‘There is Louise, Sylvaine, Claudie’ (G. Sand, La Mare au diable, 1846)

They allow for referential, anaphoric and thus highly definite noun phrases, even for proper nouns. In addition to the characterization as items on a list, the noun phrases have also been said to receive an availability or a reminder reading, and to require a “salient open proposition” (Ward & Birner 1995: 734) or a contextualization by a “predicative slot” (Abbott 1993: 44). For the remaining part of the article, I adopt Abbott’s (1993) approach of list reading cases as contextualized existentials. This explains the availability reading (the purpose to which the noun phrase is available is described by the context/ predicative slot) and the reminder cases as well (the existence of the noun phrase is always reminded at some purpose). Additionally, I include a further class of counter-examples, specific to Romance, within the category of contextualized existentials. There exists, mainly in French (Dobrovie-Sorin & Beyssade 2012: 106-110) and to a minor extent also in Italian and Catalan (Leonetti 2008: 141, Cruschina 2012: 97-104), a special construction that has been called “eventive existential” (Dobrovie-Sorin & Beyssade 2012: 106). It has a bipartite structure il y a NP qui VP ‘there is NP who/which VP’ and serves to make a statement about an event in which the referent of the noun phrase participates, rather than expressing mere existence of the noun phrase (5).

Evolution of Definiteness Effects with French il y a from 1300 to Today 281

(5)

il y a le garde champêtre qui s’ est EXPL there has the policeman who REFL is cassé le bras... broken the arm ‘The policeman broke his arm’ (E. Zola, Une page d’amour, 1878)

The construction is similar to the classic contextualized existentials in that it obligatorily requires a contextualization of the noun phrase. Indeed, the relative clause cannot be omitted (see Dobrovie-Sorin & Beyssade 2012: 107). However, the contextualization of the noun phrase takes place at two different points. For the classic list reading, availability, or reminder cases, the contextualization happens before the existential sentence is uttered: “the predicative slot must be fixed before they [= proper nouns and anaphoric definites] are introduced”, states Abbott (1993: 44). In the eventive existentials it is the other way around: there is no specific preceding context (the construction is preferably employed as an answer to the question Qu’est-ce qu’il y a ? ‘What happened? / What’s the matter?’) but there is an obligatory predication following the existential construction, a necessary subsequent contextualization of the noun phrase by the help of the relative clause. The fourth class of exceptions are instances of location or presence (6). Just like the contextualized existentials, they also accept real definite, referential noun phrases. (6)

Il y avait la Seine en bas EXPL there had the Seine below ‘There was the Seine below’ (E. de Goncourt, Soeur Philomène, 1861)

The close link between existence and location has been pointed out by many authors (among others Lyons 1967, Clark 1978, Freeze 1992, Partee & Borschev 2007, Koch 2012) and locational explanations have been used to account for a certain class of exceptions to the definiteness restriction in Italian (most recently Leonetti 2008, Cruschina 2012).4 The French il y a

 4

A major difference between French and Italian is the use of another existential verb. Italian employs a form of the copula essere ‘to be’ and a locative proform ci ‘there’. This structure facilitates a direct interpretation in terms of locative predication for the class of the so-called “inverse locatives” (Cruschina 2012: 86). French il y a on the other hand contains the locative proform y ‘there’ but does not use a form of the copula être ‘to be’, which is used in canonical locative predications (e.g. Le livre est sur la table ‘The book is on the table’). Instead, the

282

Chapter Eight

construction has generally not been described as allowing a locational use, with the exception of Koch 2012. In the present corpus study, I found clear empirical evidence for a locative-presentative use of il y a, and this as early as the 16th century. In these cases, the sentence is not about the mere existence of the entity, but rather about its presence at a certain place, in other words about its location. Additionally, there are some further types of exceptions for the quantifier tout ‘every’ and the demonstrative determiner ce ‘this’. The former is known for its quantification over kinds reading (see McNally 2011: 1842-1843, Dobrovie-Sorin & Beyssade 2012: 112-117, and the literature cited therein) and is also quite frequently used in combination with an indefinite noun phrase in the corpus (see below, section 4.4). The latter can be used in the style of an “indefinite this” (v. Heusinger 2011: 1053-1054, cf. also Ward & Birner 1995: 738-739). The use of restrictive ne … que, that also allows for definite noun phrases after il y a, will not be adressed in this paper (but cf. Zimmermann, this volume).

3. Presentation of the Corpus Study 3.1. Corpus Make Up and Methodology I used the Frantext corpus 5 which covers the whole history of literary French. It contains more than 270 million words distributed in over 4000 literary texts going back to the 12th century. My study is focussed on Middle and Modern French because these periods are best documented in the corpus and easier to query with Frantext’s search algorithm. Concerning il y a in Old French, I can however report, that I had a look at all occurrences of the 3rd person singular of avoir in the 11 medieval texts included in Frantext at the time of my study6, and that I did not find any

 3rd person singular of otherwise possessive avoir ‘to have’ is employed. This renders a direct locative interpretation in form of an “inverse locative” difficult. However, Koch 2012 argues convincingly that there is a locative-presentational use connected to il y a, so that I will not pursue this issue further. 5 Base textuelle FRANTEXT, ATILF - CNRS & Université de Lorraine. . Frantext is continuously revised and enlarged. The data for my study is taken from the version that was online in autumn and winter 2013. All further revisions and enlargements of Frantext (e.g. the inclusion of 29 additional medieval texts in October 2014) are not considered in this study. 6 The expression il y a already functioned as an existential in Old French. However, neither the expletive subject pronoun il nor the originally locative

Evolution of Definiteness Effects with French il y a from 1300 to Today 283

occurrences of an existential (il) (y) a followed by a strong noun phrase. The definiteness restriction thus seeems to hold for Old French, a finding that is in line with other studies on the definiteness effect in Old French (Zimmermann, this volume). In order to obtain the relevant data on definiteness effects in Middle and Modern French, I searched for (il) y a7 in all tenses and moods with either a definite determiner (le 8 ), a demonstrative determiner (ce), a possessive determiner (mon), a strong personal pronoun9 (moi), a universal quantifier (tout ‘all/every’, chaque ‘each’, chacun ‘each one’) or a proper noun (identified by upper case matching) immediately following it. I searched for both affirmative as well as negated sentences, but not for restrictive ne…que.

3.2. Overall Results Only the search for il y a + chaque/chacun/chacune did not return any results. This quantifier really seems to be excluded after il y a (cf. also Dobrovie-Sorin & Beyssade 2012: 114, n. 13). In contrast, all the other strong determiners that I was searching for, and also proper nouns, are well represented in the corpus, and their first occurrences appear quite early (table 3.2-1). All the data on tout, proper nouns, possessive article, strong personal pronouns, chaque, chacun, and the occurrences of the definite and demonstrative article before 1880 were checked for false positives and classified, according to the categories described in section 2, by hand. The data of il y a + definite determiner later than 1880 was down sampled by performing a random choice of 100 occurrences per 20-year slice. Likewise, the data for il y a + demonstrative determiner later than 1880 was down sampled to 40 occurrences per 20-year slice. The complete data on demonstrative determiners was additionally checked for false positives.

 pronoun y were obligatory, so that the only reliable feature that I could search for was the 3rd person singular of avoir. 7 I did not include the expletive il in the search expression because it can be missing both in the early texts from the Middle Ages and the Renaissance and in the late texts of the 19th and 20th century, where the omission of il can be a marker of an informal, colloquial language style. 8 I only cite the singular masculine form of the determiners and the 1st person singular form of the pronouns, but I searched of course for the whole paradigm. 9 French has two series of pronouns, clitic and non-clitic pronouns. Only the nonclitic pronouns, usually referred to as “strong personal pronouns”, are of interest here.

284

Chapter Eight

There is a marked frequency increase of occurrences of il y a + strong noun phrase in the 19th and 20th centuries (figure 3.2-1). This increase holds for all determiners taken together, and it holds equally for each determiner considered in its own right. At the same time, the frequency of il y a, regardless whether followed by a strong noun phrase or not, is fluctuating around a stable value, so that the frequency increase of il y a + strong noun phrase really means a substantial growth of this very construction. All in all, the frequency of il y a immediately followed by a strong determiner remains however very low, even at its peak value in the 20th century where it amounts to around 6% of all instances of il y a in the corpus.

Figure 3.2-1. Frequency of il y a (y-axis on the left) and of il y a + strong noun phrase (y-axis on the right) per number of words per period in the corpus.

determiner/ type of noun # year of first phrase occurrence definite article (le) 6931 1351 demonstrative article (ce) 1301 1435 proper noun 1297 1443 tout 1104 1342 possessive article (mon) 579 1421 strong personal pronoun (moi) 170 1759 chaque, chacun 0 ---Table 3.2-1. Number of occurrences and year of first occurrence of il y a + strong noun phrase in the corpus.

Evolution of Definiteness Effects with French il y a from 1300 to Today 285

As you can see in table 3.2-1, there is considerable variation in the number of occurrences and the year of first occurrence. The definite article le is by far the most common strong determiner after il y a, and is recorded very early in this construction. Tout, the demonstrative ce and proper nouns are the second most common strong determiners, with tout slightly less common but appearing a bit earlier than ce and the proper nouns. The possessive determiner mon also occurs very early in this construction, but at a lower frequency.10 Strikingly, the strong personal pronouns moi, toi, etc., appear at an even lower frequency, and also much later. I relate this fact to their inherent deictic value and very strong definite feature. Especially the 1st and 2nd person pronouns are known to rank highest among semantic definiteness scales, and to display in consequence the strongest definiteness effects (see Beaver, Francez & Levinson 2005, Bentley 2013). In addition to their different dates of first occurrences, the determiners vary also with respect to the density of their first examples (figure 3.2-2). While le is almost constantly found from the first half of the 14th century onwards, the other determiners need more time to become established in the il y a construction. Although the first examples for all of them except strong personal pronouns show up no later than the first half of the 15th century, the early occurrences are rather isolated and spread loosely over the decades.

 10

These different frequencies seem not to be only an effect of the lower overall frequency of mon as compared to the overall frequency of le. If the proportions of each determiner after il y a were the same as in the whole corpus, relative to the frequency of the definite determiner, the possessive determiner should appear approximately 4 times more often after il y a than it actually does, and the frequency of the strong personal pronouns after il y a should even be around 15 times higher. On the other hand, tout should be less frequent than it actually is. Only the demonstrative determiner has the expected frequency. This amounts to saying that il y a accepts the definite determiner and the demonstrative determiner equally well. Tout is accepted better than the definite and the demonstrative determiner, probably because some of its occurrences are in reality indefinite noun phrases (see below in the section on tout). The possessive determiner and the strong personal pronouns are markedly underrepresented.

286

Chapter Eight

Figure 3.2-2. Early occurrences per decade (decades marked by “x” contain at least one occurrence of il y a + the determiner in question).

Around 1600, we observe a first evolutionary step, when tout and ce join le in occurring regularly in every decade. Mon, although appearing quite early for the first time, takes around 200 years longer to establish itself after il y a, and also the proper nouns only close the gap around 1800. Mon and proper nouns also display a similar distribution of occurrences: while for tout approximately 13% and for ce approximately 7% of occurrences date before 1800, mon and proper nouns only count 2% of their occurrences before 1800. Furthermore, they are also close in their referential behaviour. Whereas tout, definite and demonstrative determiners are readily used non-referentially or with cataphoric reference, both proper nouns and possessively determined noun phrases have a strong tendency towards referential uses. Their recurrent use after 1800 indicates thus a second step of grammaticalization, followed by a third step in the later 19th century (~1870), when the strong personal pronouns join in.

4. Evolution of Definiteness Effects: Early Occurrences and Development per Determiner As shown in the preceding section, the frequency of strong noun phrases following il y a increases remarkably, especially during the 19th century. The following sections will show that this is not only the effect of a quantitative increase in tokens, but also a real qualitative evolution of types. Splitting up the analysis into different kinds of strong noun phrases accounts for subtle differences in their development. In addition to more

Evolution of Definiteness Effects with French il y a from 1300 to Today 287

general trends, each strong noun phrase has also its own characteristic way of evolving in the il y a context.

4.1. The Definite Determiner (le) The definite determiner le is by far the most common and, in the earlier stages, the most regularly occuring strong determiner after il y a. Most of the early data is made up of superlatives (7a), but there are also list readings (8) or cataphoric noun phrases (9). All of the early examples have a reduced or special kind of referentiality in common, e.g. superlative (7a, b), or type rather than token-denoting noun phrases (8) (for type vs. token denoting noun phrases and their relation to the definiteness effect see Beaver, Francez & Levinson 2005: 25, Dobrovie-Sorin & Beyssade 2012: 114). Quantifiers determined by a definite determiner are a special case. I counted them together with the superlatives because they very often convey a superlative meaning and are morphologically related to the superlative, e.g. le moins de ‘the least’, le plus de ‘the most’, la plupart de ‘the biggest part of/ most’ (7b). (7)

a.

b.

(8)

dedenz ces murs il y a le plus bel iardin EXPL there has the most beautiful garden ‘Within these walls there is the most beautiful garden’ (J. de Mandeville, Voyages, approx. 1360) c'est de son mestier dont il y a le moins de livres. EXPL there has the least of books ‘It is on his subject that there are the least books’ (G. Patin, Lettres, 1649) Les yssues du vel coustent […] deux blans: […]

y a there has

la teste et la fraze et la pance et les .IIII. piez the head and the mesentery and the paunch and the 4 feet ‘You pay two blancs for calf innards [...]: […] the head, the mesentery, the paunch and the four hooves’ (no author, Le menagier de Paris, 1394)

288

(9)

Chapter Eight

Parmi les sentences […], il y a la suivante: EXPL there has the following l'humilité est le chemin de l'honneur. ‘Among the maxims […], there is the following: humility is the way of honor’ (J. de Héroard, Histoire particulière de Louis XIII, 1605-1610)

Over the course of time, the frequency of superlatives and (mostly superlative) quantifiers, reaching over 40% before 1800, decreases remarkably and remains at roughly 5% from the later 19th century onward (figure 4.1-1). It looks as if there was an abrupt drop of frequency in the second half of the 19th century, but zooming in reveals a constant development of decreasing frequency (figure 4.1-2). Both contextualized existentials (list reading, reminder, il y a qui) and locative uses, i.e. types of exceptions that allow for definite, referential noun phrases, gradually gain importance and account for over 85% of occurrences in the later 20th and beginning 21st century. Cataphoric constructions vary between 4% and 12%.

Figure 4.1-1. Types of occurrences of il y a + definite determiner.

Evolution of Definiteness Effects with French il y a from 1300 to Today 289

Figure 4.1-2. Superlatives and quantifiers after il y a in the 19th century.

4.2. The Demonstrative Determiner (ce) The first examples of il y a + ce in my corpus are all produced by Juvenal des Ursins. The first two occurrences are exactly parallel and display a special structure with the past participle escripte from écrire ‘to write’ that modifies the existential il y a (10, note the absence of expletive il). The next example, still from Juvenal des Ursins, features a classic il y a construction, without further modification (11). The use of the demonstrative determiner ce in a cataphoric way, as displayed in this example, will remain the typical use for this combination in the next centuries. There is quite a gap between these early hits and the next occurrences of il y a immediately followed by a demonstrative determiner. All of them are either cataphoric, like example (11), or of the special construction with past participle of écrire, just like in the very first cases. (10)

y avoit ceste lettre F .III. fois escripte there had this letter F 3 times written ‘There was this letter F written three times’ (J. Juvenal des Ursins, AUDITE CELI, 1435)

290

(11)

Chapter Eight

il y a ceste clause Quos per se vel […] EXPL there has this clause ‘There is this clause Quos per se vel […] (J. Juvenal des Ursins, A, A, A, nescio loqui, 1445)

In the subsequent centuries, the cataphoric use of ce remains overwhelmingly dominant just until the second half of the 19th century (figure 4.2-1). It is only then that the types of constructions become more varied. “Indefinite this” (12), locations and contextualized existentials gain ground. (12)

il y avait ce multimillionaire américain EXPL there had this multimillionaire American ‘There was this American multimillionaire’ (M. Leiris, La Règle du jeu, 1948)

However, cataphoric constructions stay the most frequent category of il y a + demonstrative determiner, with still up to nearly 40% of occurrences in the 20th and beginning 21st century.

Figure 4.2-1. Cataphoric noun phrases among occurrences of il y a + ce.

4.3. Proper Nouns The first occurrences of il y a followed immediately by a proper noun are nearly all of a special constructional type, the so-called “eventive



Evolution of Definiteness Effects with French il y a from 1300 to Today 291

existential”, which is characterized by a relative clause predicating some action or state to the noun phrase (13, 14, 15). (13)

n' not

y aura Dieu, saint ne sainte, qui n’ there will.have God saint nor saint who not apparçoive ton orgueil. realize.SUBJ your arrogance ‘There won’t be neither God, nor any saint, who will not be aware of your arrogance’ (Charles d’Orléans, Rondeaux, 1443)

(14)

Il n' y a Georget ne Marquet qui EXPL not there has Georget nor Marquet who d' elles se sache eschapper from them REFL know.SUBJ escape ‘Neither Georget nor Marquet know how to escape them’ (N. de la Chesnaye, La Condamnation de Banquet, 1508)

The referentiality of the noun phrases in the earliest occurrences is considerably reduced through an irrealis-context, signaled by negation of il y a and subjunctive mood in the relative clause (13, 14). Moreover, the names Georget and Marquet in example (14) neither refer to real world persons nor to figures of the play, but are used as common place first names to refer to the ordinary man (cf. the corresponding comment in the edition by Koopmans & Verhuyck (de la Chesnaye 1991 [1508]: 283)). The first real referential proper noun after il y a dates from 1548, nearly a hundred years later than the first example, and the construction is again an eventive existential (15). (15)

Il y eut Siclere qui se leva EXPL there had Siclere who got up ‘Siclere got up’ (N. du Fail, Les Baliverneries d'Eutrapel, 1548)

Whereas the il y a qui construction stays the most important construction for proper nouns before 1800, it is then overtaken by list reading and reminder cases, and, from the second half of the 19th century, also by location (figure 4.3-1). The by far most frequent construction for il y a + proper nouns becomes, from the 19th century onwards, the list reading. Especially private letters, diaries and memoirs abound with occurrences of the type “yesterday dinner with x, y, and z” (16).

292

(16)

Chapter Eight

jeudi 27 mars. […] Nous dînons chez Gisette. Il y a Gautier, sa femme […], Peyrat, […] EXPL there has Gautier, his wife […] Peyrat, […] ‘Thursday, march 27. […] We have dinner at Gisette’s place. There is Gautier, his wife, Peyrat, […]’ (E. de Goncourt & J. de Goncourt, Journal : mémoires de la vie littéraire, 1863)

Figure 4.3-1. Types of occurrences of il y a + proper noun.

4.4. tout The first occurrences of il y a + tout fall into two groups. One is the classic quantification over kinds reading (17), which is the only use of universal quantifiers that is acknowledged for in the literature (McNally 2011: 1842-1843, Dobrovie-Sorin & Beyssade 2012: 112-115, and the literature cited therein). There are both implicit quantification over kinds (17a), and explicit quantification over kinds with toute sorte de/toute espèce de ‘each/every kind/species of’ (17b).

Evolution of Definiteness Effects with French il y a from 1300 to Today 293

(17)

a.

y avoit toutes autres toilles tainctes comme there had all other cloth dyed like bougrans futaines de toutes sortes, sarges et sayettes ‘There were all other (kinds of) coloured cloths, like bougrans11, all sorts of futaines, sarges and sayettes’ (A. de la Vigne, Le voyage de Naples, 1495)

b.

il y avoit toutes sortes de vacations EXPL there had all kinds of occupations ‘There were all kinds of occupation’ (F. Garasse, La Doctrine curieuse des beaux-esprits de ce temps, 1623)

The second construction, especially frequent in the beginning, is an intensification of some indefinite determiner (18). All early occurrences of il y a + tout that are not quantification over kinds, are of the type tout plein de ‘a lot of’. In the beginning, these occurrences outnumber by far the quantification over kinds: 14 out of the first 20 occurrences (these first 20 occurrences stretching out over the huge time span from 1342-1631) are il y a + tout plein de constructions. (18)

il y avoit tout plein d' herbes EXPL there had all full of grass ‘There was a lot of grass’ (P. Crignon, La Navigation de Jean et Raoul Parmentier, approx. 1529)

The use of tout as an intensifier of some indefinite determiner remains throughout the centuries. It seems limited to a stereo-typical tout plein de in the first centuries just until 1800, but becomes more diverse later on, e.g. tout un tas de ‘a lot of’, toute une série de ‘a whole series of’, and occurs also quite frequently with the idiomatic expressions il y a lieu de and il y a intérêt à (becoming il y a tout lieu de, il y a tout intérêt à ‘there is every (good) reason to’).12 All centuries taken together, the occurrences

 11 Unfortunately, I could not find the English translations for bougran, futaine, sarge and sayette. They all refer to different kinds of cloth or fabric (see Dictionnaire de Moyen Français, ). 12 Cf. for a parallel case in English Ward & Birner (1995: 739) who classify every reason in their example There is every reason to study them as “false definite” and comment: “every reason doesn't mean literally each one of a set of reasons, but rather something more like 'good reasons' or 'many reasons’”.

294

Chapter Eight

of tout + indefinite noun phrase or idiomatic expression amount to 34% of the data, i.e. they explain already a third of the occurrences of il y a + tout. The use of tout as an intensifier does not remain restricted to indefinite noun phrases. From the second half of the 17th century onwards, I find it also with definite determiners, possessives and demonstratives in the corpus. It then conveys some notion of globality or completeness. The types of exceptions correspond exactly to the types found without tout, i.e. cataphoric uses, contextualized existentials (= list reading, reminder cases, il y a qui constructions (19)) and locational uses. (19)

Y avait toute la famille qui nous suivait there had all the family who us followed ‘The whole family followed us’ (B. Blier, Les Valseuses, 1972)

Tout can also be used as a noun, meaning ‘everything’. It then refers to some indefinite globality or completeness. Very often, it is modified by an infinitive, e.g. il y a tout à craindre ‘we should fear the worst’, or resumes a list of characteristic parts or properties of an object or event (21). (21)

Des restaurants, des bars, des pharmaciens, des coiffeurs, [...], il y a tout ici, sauf des trains EXPL there has all here except trains ‘Restaurants, bars, chemist’s, hairdresser, [...], there is everything here but trains.’ (P. Morand, New-York, 1930)

Diachronically, I observe a clear development towards the usual exceptions, i.e. the contextualized existentials, cataphoric noun phrases and locations (figure 4.4-1.). The intensification of indefinite noun phrases remains stable at around a third of occurrences. The quantification over kinds is especially prevalent at earlier periods and loses its dominant position later on.

Evolution of Definiteness Effects with French il y a from 1300 to Today 295

Figure 4.4-1. Types of occurrences of il y a + tout.

4.5. The Possessive Determiner (mon) The first 10 occurrences of il y a + possessive determiner stretch out over a huge period (1421, 1480, 1511, 1543, 1560, 1578, 1629, 1711, 1794, 1822) and it is only from the 1830s that the construction is recorded with more stability. The very early occurrences are ambiguous between existentials and personal possessive constructions (21). There are also cases of reminder uses (22) and locationals (23). (21)

de tous les jeunes escuyers

n' y avoit son pareil not there had his equal ‘He had not/ there was not his equal among all the young squires’ (B. Aneau, Alector ou Le Coq : histoire fabuleuse, 1560)

(22)

Il y a notre Madame la Baillive, par exemple. EXPL there has our Madame la Baillive for example ‘There is our Madame la Baillive, for example’ (F. Carton Dancourt, Les Eaux de Bourbon, 1711)

296

(23)

Chapter Eight

“Cherchez là dedans […] il y a ma clef. EXPL there has my key ‘Search therein […] there is my key’ (Chamfort, Caractères et anecdotes, 1794)

Contrary to what we observe with proper nouns, tout and the definite determiner le, the locative use of il y a + possessive determiner does not increase with time (figure 4.5-1.). Instead, there is a slight decrease, although the locations stay at a high level representing nearly 30% of occurrences in the late 20th and beginning 21st centuries. The il y a qui construction, constitutive for the proper nouns, plays no large role in the early stages of il y a + possessive determiner. Its frequency constantly increases, however, and reaches 14% in the second half of the 20th century and beginning of the 21st century. Parallel to the observations for proper nouns, the list reading and reminder cases (24) are by far the dominant construction for il y a + possessive determiner, with a jump of frequency at the beginning of the 20th century. (24)

Il y avait mon père, […] ma mère, […] EXPL there had my father my mother ‘There was my father, my mother, […]’ (E. Ollivier, Mille eaux, 1999)

Figure 4.5-1. Types of occurrences of il y a + possessive.

 

Evolution of Definiteness Effects with French il y a from 1300 to Today 297

4.6. Strong Personal Pronouns (moi) The first example of il y a + strong personal pronoun dates from 1759. It is followed by a huge gap, with the next occurrences dating from 1844, 1849, 1857,... The vast majority of the early examples are contextualized existentials with a reminder function and chiefly occur in dialogues (25). (25)

il n'y a plus de manichéens dans le monde. [dit Candide] – il y a moi, dit Martin EXPL there has me said Martin ‘“There are no more Manicheans in the world” [said Candide] “There is me”, Martin replied’ (Voltaire, Candide, ou l'Optimisme, 1759)

There is a considerable decrease of reminder uses in the 20th century (figure 4.6-1.). Although they remain frequent and make up nearly a quarter of examples in the late 20th and beginning 21st centuries, they are overtaken by the list reading and lie equal to the locative uses (26). (26)

Dans le jardin, il y avait toi et le type américain EXPL there had you and the guy American ‘In the garden, there were you and the American guy’ (Le Clézio, Le Procès-verbal, 1963)

The switch in dominating category goes hand in hand with an extension of the construction to narrative passages. While il y a + strong personal pronoun overwhelmingly occurs in dialogue texts in the beginning (over 80% of the occurrences before 1900 appear in dialogue texts), this tendency is reversed in the middle of the 20th century (1900-1950: 60%; 1950-1980: 43%; after 1980: 18% in dialogue texts).

298

Chapter Eight

Figure 4.6-1. Types of occurrences of il y a + strong personal pronoun.

5. Conclusions The definiteness effect in French appears with different strength, depending on the century. Il y a in combination with strong determiners or strong noun phrases (definite determiner, demonstrative, possessive, strong personal pronoun, proper noun, quantifier tout) is already recorded very early in Middle and Modern French. The temporal development of first occurrences follows a trend of increasing definiteness: the first strong determiners after il y a are the least definite ones, namely the definite determiner and tout as an intensifier of indefinite noun phrases. More definite determiners or noun phrases, such as proper nouns and possessives, although first recorded quite early, take considerably longer to establish themselves in the existential construction, and their early occurrences are either ambiguous or feature noun phrases with reduced referentiality. Strong personal pronouns, the most definite elements, follow even later, and their frequency stays extremely low. The overall frequency of strong determiners after il y a increases remarkably in the 19th and 20th century and reaches around 6% in contemporary Modern French. Parallel to the increase in frequency, we observe an evolution of types of exceptions to the definiteness restriction. Whereas in the beginning the non-referential types like superlatives and cataphoric descriptions dominate the occurrences, referential types like list reading, reminder and locations constantly extend their importance.

Evolution of Definiteness Effects with French il y a from 1300 to Today 299

Moreover, each determiner starts typically with one type of exception, e.g. superlatives for le, cataphoric use for the demonstrative ce, intensification of plein de and quantification over kinds for tout, il y a qui for proper nouns, and reminder use for strong personal pronouns. Once this first exception to the definiteness restriction is established, other types can follow.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of Pragmatics 19: 39-55. Base textuelle FRANTEXT, ATILF - CNRS & Université de Lorraine. . Beaver, David, Francez, Itamar & Levinson, Dmitry. 2005. Bad subject: (Non-)canonicality and noun phrase distribution in existentials. In Efthymia Georgala & Jonathan Howell (eds.). Proceedings of SALT XV, 19-43. Ithaca, NY: Cornell University. Bentley, Delia. 2013. Subject canonicality and definiteness effects in Romance there-sentences. In Language 89.4: 675-712. Cannings, Peter. 1978. Definiteness and relevance: the semantic unity of il y a. In Margarita Suñer (ed.). Contemporary Studies in Romance Linguistics, 62-89. Washington D.C.: Georgetown University Press. Clark, Eve. 1978. Locationals: existential, locative, and possessive constructions. In Joseph Greenberg (ed.). Universals of Human Language IV: Syntax, 85-126. Stanford, CA: Stanford University Press. Cruschina, Silvio. 2012. Focus in existential sentences. In Valentina Bianchi & Cristiano Chesi (eds.). Enjoy Linguistics! Papers offered to Luigi Rizzi on the occasion of his 60th birthday, 77-107. Siena: CISCL Press. de la Chesnaye, Nicolas. 1991 [1508]. La Condamnation de Banquet. Edited by Jelle Koopmans & Paul Verhuyck. Genève: Droz. Dobrovie-Sorin, Carmen & Beyssade, Claire. 2012. Redefining Indefinites. Dordrecht: Springer. Etchegoyhen, Thierry & Tsoulas, George. 1998. Thetic and categorical, attributive and referential. Towards an explanation of definiteness effects. In Armin Schwegler, Bernard Tranel & Myriam UribeEtxebarria (eds.). Romance Linguistics, 81-95. Amsterdam, Philadelphia: John Benjamins. Freeze, Ray .1992. Existentials and other locatives. In Language 68: 553595.

300

Chapter Eight

Koch, Peter. 2012. Location, existence, and possession: A constructionaltypological exploration. In Linguistics 50.3: 533–603. Leonetti, Manuel. 2008. Definiteness effects and the role of the coda in existential constructions. In Alex Klinge & Henrik Høeg-Muller (eds.). Essays on nominal determination: from morphology to discourse management, 131-162. Amsterdam, Philadelphia: John Benjamins. Lyons, John. 1967. A note on possessive, existential and locative sentences. In Foundations of Language 3: 390-396. Maillard, Michel. 1985. L’impersonnel français de “il” à “ça”. In Jaques Chocheyras (ed.). Autour de l’impersonnel, 63-118. Grenoble: Ellug. Marchello-Nizia, Christiane. 1997. La langue française aux XIVe et XVe siècles. Paris: Nathan. Martin, Robert. 2002. Sur le statut sémantique de il y a. In LINX 47: 7987. McNally, Louise. 2011. Existential sentences. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.). Semantics: An International Handbook of Natural Language Meaning, Vol. 2, 18291848. Berlin: Mouton de Gruyter. Meulleman, Machteld. 2012. Degrees of grammaticalization in three Romance languages: A comparative analysis of existential constructions. In Folia Linguistica 46.2: 417-451. Partee, Barbara & Borschev, Vladimir. 2007. Existential sentences, BE, and the genitive of negation in Russian. In Ileana Comorovski & Klaus von Heusinger (eds.). Existence: Semantics and Syntax, 147-190. Dordrecht: Springer. Suñer, Margarita. 1982. Syntax and Semantics of Spanish Presentational Sentence-Types. Washington D.C.: Georgetown University Press. von Heusinger, Klaus. 2011. Specificity. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.). Semantics: An International Handbook of Natural Language Meaning, Vol. 2, 1024-1057. Berlin: Mouton de Gruyter. Ward, Gregory & Birner, Betty. 1995. Definiteness and the English existential. In Language 71.4: 722-742.

CHAPTER NINE EXISTENTIALS VS. UNACCUSATIVES: THE DEFINITENESS RESTRICTION IN ROMANCE SUSANN FISCHER

1. Introduction1 Since Milsark (1974, 1977), who documented this phenomenon for English, it has been a well-known fact that certain positions or contexts within sentences require a noun phrase occurring there to have a particular value >– Def@ or to be interpreted as having such a value. Restrictions of this kind are termed definiteness restrictions (DR) and have also been found and discussed with respect to Romance languages (Belletti 1988, Leonetti 2008, Bentley, Ciconte & Cruschina 2013, among many others). Consider the following examples from English and French2 illustrating that definite (or strong) determiners are excluded in unaccusative (1b, 1d) and existential (2b, 2d) constructions, whereas indefinite (or weak) determiners are allowed, thus exhibiting a clear-cut DR.

1

Preliminary versions of this paper have been presented at the “VI NEREUS workshop: Theoretical implications of the syntax/semantics interface in Romance”, held at the University of Cologne in 2012, at the “Definiteness workshop” held at the University of Hamburg in 2014, and at the Grammar Seminar, University of Lund in 2011. Thanks to the audiences for their comments and remarks. I would particularly like to thank the two anonymous reviewers, and additionally Sarah Jobus, Mario Navarro, Elisa Sorrenti, Jorge Vega Vilanova, Xavier Villalba, and my co-editors for interesting discussions and great comments. All remaining errors are entirely my own. 2 In order to show which examples stem from which language, I will use the following abbreviations: ModE = Modern English, OCat = Old Catalan, ModCat = Modern Catalan, OF = Old French, ModF = Modern French, ModIta = Modern Italian, OSp = Old Spanish, ModSp = Modern European Spanish, LAS = Latin American Spanish.

Chapter Nine

302

(1)

(2)

a. b. c. d.

There arrives a train. *There arrives the train. Il arrive un train. * Il arrive le train.

a. b. c. d.

There is a dog in the garden. *There is the dog in the garden. Il y a un chien dans le jardin. *Il y a le chien dans le jardin.

ModE ModF

ModE ModF

The DR has been discussed within different components of grammar: syntax, semantics, and pragmatics, as well as a combination thereof. However, so far no explanation has been able to entirely account for the phenomenon. One of the problems seems to be that existential constructions have been paired up with unaccusative constructions (Belletti 1988, Perlmutter 1978, Lyons 1999, among others) or unaccusatives with existentials (Carnie & Harley 2005, Leonetti 2008) as if they both belonged to the same type of sentence/construction. This paper is an attempt to bring together the information about the DR concerning the two constructions that is scattered throughout different papers. New data (from some old and some new Romance languages) will be provided showing that there are more differences than similarities between these constructions. On the basis of these data and the results of the comparison between the two constructions, I will argue that the DR in unaccusatives is a syntactic restriction and has to be explained in syntactic terms, whereas the DR in existentials is a semantic/pragmatic restriction that has to be explained by referring to the semantics of the constructions and the pragmatics of the context. I will show that the DR in unaccusatives is dependent on whether a language is a null-subject language or not, whereas the DR in existentials depends on the interrelation of word-order and information structure – in some varieties animacy seems to play a role as well. Section two will give a short overview of the variability of the DR in unaccusatives and existentials in four Romance languages. Section three will summarise the differences between the two constructions concerning structure and semantics. Section four will discuss different analyses that have been proposed over the last years in order to account for the DR in the two constructions and different languages. In section five new data will be presented showing that the explanations put forward so far are not compatible with the diachronic development of existential constructions in Romance, nor with the variability concerning the elements triggering the

Unaccusatives vs. Existentials

303

DR in two varieties of Spanish. It will be shown that additional features, namely animacy and word-order, need to be included in order to better understand the DR in existentials, whereas this is not necessary for unaccusatives, where the null-subject parameter is relevant. Section six will summarise and discuss the findings, and attempt an explanation that is based on the insight that existentials and unaccusatives have to be analysed differently.

2. The DR in Romance The DR shows a variable behaviour in the different Romance languages. Consider the following examples from four Romance languages (French, Spanish, Catalan, and Italian) of unaccusative (3, 5, 7, and 9) and existential (4, 6, 8, and 10) constructions. The French examples exhibit the DR in unaccusative (3a, b) as well as in existential (4a,b) constructions. French, being a non-null-subject language, needs the expletive il whenever the pivot is located postverbally. Furthermore, the verb is conjugated with être (to be), as illustrated in example (3a). Existential constructions consist of the expletive il, a locative clitic y, and the verb avoir (to have), as illustrated in example (4a). (3)

a.

b. (4)

a.

b.

Il est arrivé une fille. there is arrived a girl ‘There has arrived a girl.’ *Il est arrivé la fille.

ModF

Il y a une fille dans le jardin. there LOC has a girl in the garden ‘There is a girl in the garden.’ *Il y a la fille dans le jardin.

ModF

Spanish shows the DR only in existentials (6b). Since Spanish is a nullsubject language, no expletive is needed when the subject DP is in a postverbal position. The unaccusative verb in (5a) and (5b) is conjugated with haber (to have). The existential construction shows the lexicalised hay (there is/there are). From a diachronic perspective, the origin of the verb form hay is an incorporation process of the locative y (there) – derived from the Latin adverb ibi – into the third person singular form of the verb haber (to have).

Chapter Nine

304

(5)

a.

b. (6)

a.

b.

Ha llegado una chica. has arrived a girl ‘There arrived a girl.’ Ha llegado la chica.

ModSp

Hay una chica en el jardín. has a girl in the garden ‘There is a girl in the garden.’ *Hay la chica en el jardín.

ModSp

Catalan does not present a definiteness restriction in unaccusatives (7), and there are quite a few exceptions in existential constructions (8).3 Like Spanish it is a null-subject language and therefore it does not need an expletive when DP subjects are located postverbally. Like French, Catalan has a locative clitic hi (there). (7)

a.

Va arribar una noia. a girl ‘There arrived a girl.’ Va arribar la noia.

ModCat

Hi

ModCat

PAST arrive

b. (8)

a.

ha una noia al jardí. has a girl in.the garden ‘There is a girl in the garden.’ Hi ha la noia al jardí.

LOC

b.

Italian looks similar to Catalan in that the DR does not apply in unaccusatives (9), and exceptions to the DR can be found with regard to existentials (10). Like Spanish and Catalan, it is a null-subject language, thus postverbal subjects do not need to be completed by an expletive. Like French and Catalan, the existential construction makes use of a clitic locative ci. (9)

a.

b.

3

È arrivata una ragazza. is arrived a girl ‘There arrived a girl.’ È arrivata la ragazza.

For a somewhat different view, see Leonetti (2008, and this volume).

ModIta

Unaccusatives vs. Existentials

(10)

a.

C’è

una ragazza in giardino. a girl in garden ‘There is a girl in the garden.’ C’è la ragazza in giardino.4

305

ModIta

LOC’is

b.

The availability of definite postverbal subjects in Spanish, Catalan, and Italian without the definiteness restriction has often been termed “free inversion” and has been interpreted as a characteristic of null-subject languages, thus explaining why French as a non-null-subject language shows the DR, in contrast to, for example, Italian or Catalan. However, something else must be at stake concerning existentials, since even though Catalan, Italian, and Spanish are null-subject languages, only Spanish shows a clear DR in existentials, while Italian and Catalan are less restrictive.

3. Structural and Semantic Properties The data in chapter two demonstrated a difference between existentials and unaccusatives concerning the DR. In the following, the structural and semantic properties of the two constructions will be presented and discussed.

3.1 Syntax and Semantics of Unaccusatives Unaccusatives belong to the class of intransitive verbs. It is generally assumed that they include a single argument within their subcategorisation frame, namely the internal argument of the VP as illustrated in (11) (cf. Belletti 1988, Belletti & Bianchi this volume, Perlmutter 1978, and others).

4

Leonetti (2008, this volume) as well as Cruschina (2015, this volume) argue that this kind of example can be interpreted as a “pseudo-existential locative construction”, and that the DR holds in Italian with respect to genuine existentials. See, however, Sorrenti (2015), who shows that the DR does not always apply in Italian existentials. It seems to depend on where the people using Standard Italian live, i.e. what dialect of Italian they speak in addition to Standard Italian.

Chapter Nine

306

(11)

Unaccusatives VP | V’ 2 V°

DP

Thus, the postverbal subject or rather the syntactic pivot of these constructions is an object in disguise, and the theta role which is assigned to this argument is never [AGENT] but rather [THEME]. This is the reason why they have also been discussed together with psych verbs5 (Belletti & Rizzi 1988). A number of structural criteria have been proposed in order to detect unaccusativity.6 The most well-known criterion is auxiliary selection in languages that use two different temporal auxiliaries (have and be) for analytic past/perfect verb forms: e.g. German: ich bin gefallen (I have fallen); French: je suis tombé; Italian: sono caduto. It has also been suggested that be encodes aspectuality (Sorace 2000, Grewendorf 1989). More specifically, it has been proposed that unaccusative verbs tend to express a telic change of state or location (12), while verbs choosing have tend to express an agentive activity which does not involve telicity (13). (12)

a. b.

La fille est arrivée. La ragazza è arrivata. the girl is arrived ‘The girl has arrived.’

(13)

a.

La fille

5

a

parlé.

ModF ModIta

ModF

Psych verbs are verbs that show an oblique subject, mi piace, me gusta, m’agrada (me like), that is never [AGENT] but rather [EXPERIENCER]. It has been controversially discussed whether these subjects are true syntactic subjects or whether they are objects that have been moved to the subject position (cf. Belletti & Rizzi 1988, Gutiérrez-Bravo 2006, Fischer 2004). 6 It should be noted that the diagnostics are not always perfect in detecting unaccusativity; see, e.g., the discussions in Alexiadou, Anagnostopoulou & Everaert 2004 and Mackenzie 2006.

Unaccusatives vs. Existentials

b.

La ragazza ha parlato. the girl has spoken ‘The girl has spoken.’

307

ModIta

In Italian7 some verbs can be constructed with either essere or habere denoting an aspectual difference (14). (14)

a.

b.

Le mele sono marcite. the apples are rotted ‘The apples are rotten.’ Le mele hanno marcito al sole. the apple have rotted in-the sun ‘The apples have been rotting in the sun.’

ModIta

ModIta

Sono marcite expresses a completed process, i.e. that the apples are actually rotten, whereas hanno marcito denotes the duration of the process. Another criterion is passivisation. Under the assumption that the crucial characteristic of the passive is the absorption of the subject theta role, it is clear that no such absorption is possible in the case of unaccusatives (15). (15)

a. b. c. d. e.

*The man was arrived *La fille a été arrivée *La ragazza è stata arrivata *La chica ha sido llegado *La noia ha estat arribata

ModE ModF ModIta ModSp ModCat

The literature on –er nominalisation has established the so-called external argument generalisation, i.e. –er nominals typically denote the external argument of the underlying predicate, irrespective of the specific thematic role the argument has (Rappaport Hovav & Levin 1992, Schäfer 2008).

7

The same applies to Southern German, e.g. in Franconian and Swabian the difference between using be vs. have with the following (and many other) verbs clearly denotes an aspectual difference: (i) Ich bin gestanden. vs. Ich habe gestanden. I am stood I have stood ‘I stood.’ ‘I have been standing.’ (ii) Ich bin in den Raum getanzt. vs. Ich habe drei Stunden getanzt. I am in the room danced I have three hours danced ‘I danced into the room.’ ‘I have been dancing for three hours.’

Chapter Nine

308

Since the one argument in unaccusative verbs is the internal argument and not the external one, these verbs do not allow –er nominalisation (16). (16)

a. b. c. d. e.

*Arriver *Arrivateur *Arrivatore *Llegador *Arrivador

ModE ModF ModIta ModSp ModCat

En cliticisation in French (17) and ne cliticisation in Italian (18) is another criterion that has been proposed in order to detect unaccusativity. It has been claimed that the partitive clitics en/ne can only cliticise out of the object position (Burzio 1986, Müller 2000). This being so, the behaviour of these clitics is used to test whether the quantified argument is an internal (17a, 18a) or an external argument (17b, 18b). (17)

a.

b.

(18)

a.

Trois en arriveront demain. three PART arrive.FUT tomorrow ‘Three of them will arrive tomorrow.’ *Trois en para‫ۺ‬tront demain. three PART appear.FUT tomorrow ‘Three of them will appear tomorrow.’ Ne

sono arrivato molti. are arrived many ‘Many (people) have arrived.’ *Ne hanno telefonato molti. PART have called many ‘Many (people) have called.’

ModF

ModF

ModIta

PART

b.

ModIta

In Catalan, however, both the internal (19a) and the external (19b) argument can be substituted by the partitive clitic en. Thus it is obvious that the partitive clitic in Catalan cannot detect the internal argument. Since all other tests also hold for the Catalan unaccusatives, this might be due to the categorical status of the partitive clitic in this language (cf. Bonet 1991). (19)

a.

En

van

arribar molts. arrived many ‘Many (people) arrived.’

PART PAST

ModCat

Unaccusatives vs. Existentials

b.

En

van

trucar molts. called many ‘Many (people) called.’

309

ModCat

PART PAST

Summarising, it can be said that unaccusatives denote a telic change of state or location and that the criteria applied show that the syntactic pivot of unaccusative constructions seems to be the internal argument and not the external argument. Furthermore, this argument never carries the thematic role [AGENT].

3.2 Syntax and Semantics of Existentials Existential sentences are used to express a proposition about the existence or presence of someone or something. They primarily introduce a new referent into the discourse; that is why they are sometimes also called thetic sentences, i.e. purely thematic, or topicless. Alternatively, if they are assumed to have a theme-rheme structure, the theme is hypothesised to be a location rather than the referent of the pivot (Leonetti 2008, McNally 2011). Concerning the underlying structure of existentials, different views are held in the literature. In English, be is a copula and as such it appears that the single argument is a complement (cf. Lyons 1999: 238). As a matter of fact, the intransitive verbs which occur in existential sentences have sometimes been analysed as “unaccusatives” (Perlmutter 1978, Burzio 1986), i.e. they subcategorise for an NP-complement but do not assign Case (see also Maling 1988: 170). Under this view, the structure is like the structure in (11) for unaccusatives. In general, however, copular verbs such as to be, être, essere, and haber are said to be defective; they are not capable of assigning a theta role (for a different view, see Löbel 2009). Moro (1991, 1997), building on Heggie (1988), argues for a small clause analysis where the small clause consists of an NP/PP and an expletive, and where the expletive always needs to move to SpecIP (20).

Chapter Nine

310

(20)

Existentials IP 3 PP/NP I’ | 3 therei I° VP 3 V SC | 3 be NP PP/NP | ti

Languages differ in the way they express existential meanings. In many languages an expletive subject is required like English there or French il; however, there are many languages without the need for an expletive subject, e.g. null-subject languages like Spanish, Catalan, or Italian. Existential constructions do not always contain a verb,8 and if there is one, it is often homophonous with a verb meaning ‘to be’ or ‘to have’. Italian uses c’è, 3SG of the verb esserci ‫ދ‬to be.LOC‫ ;ތ‬French il y a and Catalan hi ha use the 3SG of the verb avoir/haver ‫ދ‬to have‫ ތ‬together with a locative clitic; Spanish makes use of the element hay, which shows an incorporated locative together with the 3SG verbal form of haber ‫ދ‬to have‫ތ‬. Although Italian, Catalan, and French use the locative clitic in other contexts as well, it is generally assumed that its use in existential constructions is lexicalised. Thus, what all existential sentences seem to include is such a locative, either as an expletive (English there), or as a clitic (Italian ci, Catalan hi, French y), or incorporated into the verb (Spanish hay). The presence of such an element is the reason why existential sentences are often grouped and discussed together with locative constructions (e.g. Freeze 1992). Thus, with the exception of English, unaccusative constructions in the languages investigated here – in contrast to existential constructions – do not seem to include locative elements,9 ignoring for the moment how semantically bleached these elements are in the existential constructions.

8

For example, Russian doesn’t need a copula in present tense sentences. For a different view, see Alexiadou & Schäfer (2011), who assume that a locative is always included in unaccusative sentences, either covertly or overtly.

9

Unaccusatives vs. Existentials

311

Another difference between existentials and unaccusatives is that the former do not express a telic change of state or location but rather as mentioned before the existence and presence of someone or something. It seems correct to claim that only atelic events are possible in existential constructions. Milsark (1977: 12) argues that those adjectives which are permitted in the coda of existential constructions denote states (21a and 22a, b, c, d,), i.e. they are state-descriptive predicates in Milsark’s (1977) terminology, and stage-level predicates in more modern terminology. According to McNally (2011: 1845), this fact seems to be uncontroversial. In her analysis she proposes that the coda serves to restrict the spatiotemporal parameters within which the referent of the pivot is instantiated. Under this view, individual-level predicates are ruled out because they lack the ability to provide the necessary spatio-temporal restriction.10 (21)

a. b.

There is a man drunk. *There is a man intelligent.

ModE

(22)

a. a’. b. b’. c. c’. d. d’.

Il y a un homme ivre. *Il y a un homme intelligent. Hay un hombre borracho. *Hay un hombre inteligente. C’è un uomo ubriaco. *C’è un uomo intelligente.11 Hi ha un home borratxo. *Hi ha un home intel.ligent.

ModF ModSp ModIta ModCat

Carnie & Harley (2005: 46) argue – on the basis of standard tests for presuppositionality – that presupposition of the process component of the bipartite event structure in telics is forced by the assertion of the endpoint component and that presupposed elements are not pragmatically consistent with existentials. In their view telic predicates are therefore ruled out in existential constructions (Carnie & Harley 2005: 47). Summarising the facts so far, the structure and semantics of 10

In Bianchi & Chesi (2014) a novel analysis is proposed that derives the difference between preverbal vs. postverbal subject positions, on the one hand, and individual- vs. state-level (and in/definiteness) on the other by a top-down-oriented computation where reconstruction no longer needs to be stipulated. 11 This sentence as interpreted with an existential meaning is ungrammatical with the individual-level adjective. However, if ci is interpreted as a deictic element as in “Here is an intelligent man”, then the sentence is grammatical (Elisa Sorrento p.c.)

Chapter Nine

312

existentials seem to vary in the different languages and also compared to unaccusative constructions: there are telic predicates in unaccusatives vs. atelic predicates in existentials; locative elements are used in existential constructions, whereas unaccusatives (with the exception of English) do not include any locative element. Regarding the similarities in both constructions, the pivot never denotes the thematic role [AGENT]. As for unaccusatives, the available tests have shown that the argument is never the external but rather the internal argument, whereas in existential sentences the results are not as clear-cut. The tests for unaccusativity cannot be applied, since copular verbs do not subcategorise for internal and/or external arguments (cf. Lohndal 2006, Löbel 2009). The pivot of existentials seems to be a small clause or an NP adjoined to the VP (see Moro 1997, Leonetti 2008) and not the internal argument of the copula verb. Thus, in order to explain the DR, we need to take into account the syntax, the argument structure, and the semantics of existential and unaccusative constructions. All in all, it seems we need to focus more on their differences than on their similarities.

4. Previous Analyses Many analyses have been advanced in order to explain the DR in unaccusatives and existentials. In the following, a short overview will be given.

4.1 Unaccusatives For a long time, the definiteness effect in Modern English and Modern French unaccusatives has been explained by Diesing’s (1992) Mapping Hypothesis, i.e. syntactic positions are directly mapped to semantic interpretation and the observed definiteness restrictions in (23c) and (24c) are due to the low VP-internal position of the subjects. These low subjects are mapped to nuclear scope and therefore get a weak existential reading. (23)

a. b. c. d.

There suddenly >VP arrives a train@. A train suddenly arrives. *There suddenly >VP arrives the train@. The train suddenly arrives.

ModE

Unaccusatives vs. Existentials

(24)

a.

b.

c. d.

Il arrive souvent >VP une there arrives often a ‘There often arrives a girl.’ Une fille arrive. a girl arrives ‘A girl arrives.’ *Il arrive souvent [VP la there arrives often the La fille arrive. the girl arrives ‘The girl arrives.’

fille@. girl

313

ModF

fille]. girl

It has also been shown for Modern Catalan (25) and Modern Spanish (cf. Vallduví 2002, Ordóñez 1998, Sheehan 2004 for many Spanish examples) that in VS orders the verb and postverbal subject may be interrupted by adverbials. (25)

a.

b.

Ahir va tornar a l’India un funcionari. ModCat yesterday PAST return to India a official ‘Yesterday, a functionary returned to India.’ I el cap de setmana arribarà tranquilament la Nuria... & the end of week arrive.FUT calmly the Nuria ‘And on the weekend, Nuria will calmly arrive.’

Under Diesing’s Mapping Hypothesis these subjects should get a weak existential reading when they are inside VP. However, in Catalan and Spanish definite subjects in the postverbal position also follow the adverbials and are presumably in a VP internal position; nevertheless they don’t seem to get a weak existential reading. The SV/VS alternation without any DR that we see in (26) and (27) has often been called free inversion and is argued to be a characteristic of null-subject languages. (26)

a.

La chica llega. the girl arrives ‘The girl arrives.’

ModSp

Chapter Nine

314

(27)

b.

Llega la chica. arrives the girl ‘There arrives the girl.’12

a.

La noia arriba. the girl arrives ‘The girl arrives.’ Arriba la noia. arrives the girl ‘There arrives the girl.’

b.

ModCat

Under the general concept of minimalism, where movement needs to be triggered and is a “last resort” option in order to check uninterpretable features, free inversion is no longer an acceptable explanation. In order to explain the lack of the DR in null-subject languages, Barbosa (1995) and Alexiadou & Anagnostopoulou (1998) propose that the DR only applies in those languages where an XP needs to check the [EPP] feature in I°/T°, i.e. in non-null-subject languages. By contrast, in null-subject languages the EPP is checked by the verb, i.e. by a head (X°), and that is why no DR applies. As a consequence of the verb checking the EPP in null-subject languages, all preverbal subjects need to be analysed as topics. Sheehan (2004) (see also Silva-Villar 1998), building on the idea of a split in how to check the [EPP] and the relation to the DR, specifies the argument further. She argues that the possibility of checking the [EPP] in T° in nonnull-subject languages is dependent on the expletive: if an expletive is introduced, the definiteness restriction applies; if no expletive is introduced, no DR applies. Looking at the old strata of English13 (28) and also French14 (29), we notice that these explanations do not seem to hold.

12 In the following, many examples are translated into English with a definite DP in unaccusative and/or existential constructions. The English examples containing definite DPs are all ungrammatical; nevertheless, I include this free translation into English in order to demonstrate the differences between the examples. 13 All examples of Old and Middle English are taken – if not indicated otherwise – from the York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE). Detailed information on the YCOE is available at http://www-users.york.ac.uk/ ~lang22/YCOE/. 14 Examples of Old French – if not indicated otherwise – are taken from the Nouveau Corpus d’Amsterdam: Corpus informatique de textes littéraires d’àncien français (1150–1350). Detailed information on the Old French corpus is available at http://www.uni-stuttgart.de/lingrom/stein/corpus.

Unaccusatives vs. Existentials

(28)

a.

b.

c.

d.

(29)

a.

b.

315

and cuman yfele men and bereafian hine OE and came evil men and bereave him ‘and there came evil men and bereaved him’ (coapollo, ApT:17.22.36) & come a culur se briht as þeah ha bearnde of heouene & come a dove so bright as though it burnt of heaven ‘And there came a dove from heaven so bright as though it burnt.’ (Margarete 89.564, cited in Biberauer & Roberts 2003) þonne he cume when he comes ‘when he comes’ (cocura,CP:49.379.18.2564) þere ben 5 provinces LOC are 5 provinces ‘Five provinces are there ....’ (Mandeville 29.28) lors vint un abbe au chevalier OF then came an abbot to the knight ‘then came an abbot to the knight’ (joinv, 180) il i vienge uns chevaliers qui les maintiegne. it there come a knight who them protects ‘There would come a knight who protects them.” (Perceval 2315)

In (28a,b) no expletive is used in the unaccusative construction as if Old English were a null-subject language; nevertheless the DR seems to apply here. In (28c) and many other sentences subject pronominals are attested where in Spanish or other null-subject languages no subject pronominal would be used. The status of Old English as being a null-subject language is not clear (cf. O. Fischer et al. 2000, Fischer 2010, among many others). However, in the very first texts expletive drop is the only available means in unaccusative constructions. þere (there) used as an expletive and not as a locative adverbial appears only from the 15th century onwards (cf. Ingham 2001), thus in (28d) there needs to be interpreted as a deictic element, a locative. Concerning the French examples, we see an expletive and the DR applying in (29b), and in (29a) an indefinite postverbal subject without an expletive.15 15

As is well known in studying diachronic data, we are faced with a lack of negative evidence that is impossible to overcome. However, it again needs mentioning that not finding clear counter-examples is not unequivocal evidence that there does not exist such an example; it could just mean that so far we have

Chapter Nine

316

If Old French and Old English are analysed as null-subject languages, the verb checks the [EPP] and thus no DR should apply (cf. Alexiadou & Anagnostopoulou 1998). However, if they are analysed as non-null-subject languages, the expletive can obviously be dropped (28b) and we should then find sentences where the expletive is dropped and the DR does not apply. In Old English (28) and Old French (29), however, we see the DR applying independently of whether the expletive is dropped or not (and so far I haven’t found any counter-examples). Traditional approaches analyse Old French (Boucher 2003) and Old English (Kroch 2002) as null-subject languages. However, it has been argued by Roberts (1995) and Rinke & Meisel (2009) that Old French did not exhibit the stereotypical syntactic behaviour of a null-subject language. Fischer (2010) and Zimmermann (2014) argue that Old French is not a null-subject language but a non-null-subject language. The same has been proposed for Old English (O. Fischer et al. 2000, Fischer 2010): “Old English did not allow referential pro drop and only limited expletive pro drop” (O. Fischer et al. 2000: 39). In Fischer (2010), I have proposed that Old French and Old English are indeed non-null-subject languages and this is why the DR always applies independently of whether an expletive is introduced or not. However since Old English and also Old French show object shift (30 and 31) and VSO orders (Biberauer & Roberts 2003, Zaring 1998), the inverted subject can be outside VP and is therefore close enough to the verb to check the EPP feature, thus no expletive needs to be introduced. (30)

(31)

a.

(...) þei shuld no meyhier haue OE they should no mayor have ‘they were not allowed to have a major’ (Chronicles 62,23)

b.

(...) þæt he mehte his feorh generian that he want his live save ‘(...) that he wanted to save his life’ (Or 2.5.48.18) (cited in Biberauer & Roberts 2003)

a.

(...) por coi avés vos ce fait why have you this made ‘(...) why have you made this’

not found it. However, for more extensive data on the DR in Old and Middle French, see Zimmermann (2014), who reaches the same conclusion.

Unaccusatives vs. Existentials

b.

317

(...) les gens qui ont accoustumé a ce faire, the people who have accustomed to this make ‘(...) the people who are accustomed to doing this.’ (StL189) (cited in Zaring 1998: 321)

In those sentences where an expletive is introduced, the postverbal subject is in a different position, presumably inside the VP; in any case it is not close enough to the verb to check the [EPP]. It is in these sentences that an expletive is needed to check the [EPP] in I°/T°. Independently of whether it is the expletive or the postverbal subject that checks the [EPP], the DR applies, in contrast to in null-subject languages, where the verb itself checks the EPP and no DR applies.16 Under this view (Fischer 2010), the DR in unaccusatives is subject to syntactic considerations.

4.2 Existentials Many theories have been proposed in order to account for the DR in existential constructions. We find syntactic (Freeze 1992, Safir 1985, Moro 1997), pragmatic (Abbott 1993, 2005, Lumsden 1988), and semantic (Milsark 1974, Keenan 1976) approaches as well as combinations thereof (Leonetti 2008, McNally 2011). The syntactic approaches with respect to English (Freeze 1992, Moro 1997) are similar to those proposed for unaccusatives. Thus there constructions are derived from their copular sentences by specific transformational rules. (32)

16

a. b. c. d. e.

There is a burglar in the basement. Il y a un voleur dans le sous-sol. Hay un ladrón en el sótano. C’è un ladro nel scantinato. Hi ha un lladre al soterrani. has a burglar in the basement ‘There is a burglar in the basement.’

ModE ModF ModSp ModIta ModCat

Of course, we do find examples with a definite DP in unaccusative sentences; however, these examples are the ones that we find in Modern French and Modern English as well (cf. Gaatone 1970, Vet 1981, and see the discussion on the exceptions and a complete list of references in Zimmermann in this volume). Examples are of the type: There is the man who is always late; Il y a le tour d’Eiffel, Notre-Dame et beaucoup de choses intéressantes.

Chapter Nine

318

(33)

a. b. c. d. e.

A burglar is in the basement . Un voleur est dans le sous-sol. Un ladrón está en el sótano. Un ladro è nel scantinato. Un lladre es al soterrani. a burglar is in the basement ‘A burglar is in the basement.’

ModE ModF ModSp ModIta ModCat

In languages other than English this analysis is difficult to maintain (already under a GB analysis) since different verbs are used for existential and other copular sentences. Compare (32a), an existential sentence which uses hay, to (33c), which uses estar and which cannot be interpreted as an existential. This shows that these constructions do not share the same underlying structure. However, in minimalist terms the assumption runs into difficulties for English as well, since the numeration sets of sentences with the expletive (there) and without the expletive is different and therefore the two sentences cannot compete for spell-out. In pragmatic approaches (e.g. Abbott 1993: 44) existential constructions are considered to represent thetic sentences introducing a hearer-new referent into the discourse. The sentences in (32) can open up a conversation, while those in (33) are far more restricted, since they cannot serve to introduce a new referent or a new situation into the discourse in an “out-of-the-blue” context: What happened? (34)

a. b. c. d.

Well there was a burglar in the basement. Bon, il y avait un voleur dans le sous-sol. Pues, había un ladrón en el sótano. Beh, c’èra un ladro nel scantinato.

ModE ModF ModSp ModIta

(35)

a. b. c. d.

*Well, a burglar was in the basement. * Bon, un voleur était dans le sous-sol. *Pues, un ladrón estaba en el sótano. *Beh, un ladro era nel scantinato.

ModE ModF ModSp ModIta

This shows that (32, 34) are existential constructions whereas (33, 35) cannot be analysed as such. Milsark (1974, 1977), who presented the first extensive study on the semantics of existentials, provides a typology for determiners. He classifies the different determiners and noun phrases as either strong or

Unaccusatives vs. Existentials

319

weak. According to him, the weak determiners (a, many, some) involve cardinality, whereas the strong ones (the, every, all) are quantificational. Milsark (1977, see also Lyons 1999: 240) analyses there be as an existential quantifier which has to bind a variable. (36)

a. b. c.

There is a man in the garden. *There is the man in the garden. The man is in the garden.

ModE

Sentence (36a) makes a claim of existence about the entity described by the complement phrase, but (36c) doesn’t. The DR applies in (36b) since strong determiners or noun phrases are themselves quantificational, i.e. a definite noun phrase complement does not provide a variable that is free for binding. The weak determiner in (36a), however, is not necessarily quantificational and therefore provides a variable that the existential quantifier can bind. Semantic-pragmatic approaches like for example Leonetti (2008) and McNally (2011) also include the syntactic component in their considerations. Leonetti (2008) thoroughly investigates the different possible structures that have been discussed under the heading “existentials” and provides data showing that there are at least three types of there-constructions (37). (37)

a. b. c.

Proper existential : There is a dog in the garden. Eventive existential: There is a dog barking. Enumerative existential (list): There is Peter, Mary, and Paul.

Proper existentials – the only true existentials in Leonetti’s (2008) view (see also Milsark 1974: 20) – always allow a coda that provides a specification of the location, or another relevant relation to the discourse. He argues that indefinites in proper existentials are compatible with a locative coda either within the same VP or dislocated, whereas definites need the coda to be dislocated17 in all cases. The fact that definite DPs are acceptable with a narrow focus reading when the coda is dislocated was already noticed by Belletti (1988). See the Italian (38) examples from Leonetti (2008) and the Catalan (39) examples created accordingly.

17 The dislocation of the coda in these examples is marked by the comma, which indicates a prosodic break.

Chapter Nine

320

(38) a. a’.

b. b’. (39) a. a’.

b. b’.

ModIta C’è una statua di Michelangelo, in Piazza della Signoria. C’è una statua di Michelangelo in Piazza della Signoria. there’is a statue of Michelangelo in P. d. S. ‘There is a statue of Michelangelo in the P. d. S.’ C’è la statua di Michelangelo, in Piazza della Signoria. ??C’è la statua di Michelangelo in Piazza della Signoria. ModCat Hi ha una estàtua de Joaquín Suárez, a la Plaça del Pi. Hi ha una estàtua de Joaquín Suárez a la Plaça del Pi. LOC has a statue of J. S. at the P. d. P. ‘There is a statue of Joaquín Suárez in the Plaça del Pi.‫ތ‬ Hi havia la estàtua de Joaquín Suárez, a la Plaça del Pi. ?Hi havia la estàtua de Joaquín Suárez a la Plaça del Pi.

The a/a’-sentences in (38) and (39), with or without a prosodic break, are true existentials, while the b/b’-sentences in (38) and (39) are not (cf. Leonetti 2008). The definite articles in these sentences do not seem to license a coda (38b’) and (39b’). However, even in the Italian examples some speakers do not completely rule out sentence (38b’) without a prosodic break. Furthermore, it needs to be noted that there is a difference in the acceptance of these sentences between Italian and Catalan speakers. Catalan speakers are less strict in allowing a definite DP with a nondislocated coda.18 This seems like a convincing argument, although I will show (section 5) that the situation is even more complex than the distribution proposed here. Another piece needs to be added to the jigsaw of existentials and this is word-order in general. Italian and Catalan allow VOS/VXS with a narrow focus on the postverbal subject (see Leonetti 2008, Vallduví 1993, 2002), which is exactly the word-order and the position in which definite DPs are allowed. VSX orders are excluded in Modern Catalan (Vallduví 2002) and also in Italian when the subject is definite (Belletti 1988, Leonetti 2008). It seems reasonable to argue that this position is excluded due to the fact that Catalan as well as Italian need to explicitly mark informational partitions with respect to marked word-orders, e.g. via dislocation or cleft sentences, 18

Three of the five Catalan native speakers from Barcelona with whom I checked these sentences rated these kinds of examples (39b’, a definite DP without a prosodic break) with a ‫ތ?ދ‬, two rated them as being o.k.

Unaccusatives vs. Existentials

321

and do not accept complex strings (Vallduví 1993, 2002). This can be seen in existentials as well as in other sentence types, as has also been discussed with respect to subject inversion in French by Kampers-Mahne et al. (2004: 572ff). It is a well-known fact that word-order plays an important role in information structure, even in languages that are no longer considered to be “topic prominent languages” (Lehmann 1976, see also Calboli 1978). What we see in Catalan is exactly this: definite DPs are allowed in existentials since they do not immediately follow the verb, the canonical word-order in Catalan being VXS (Vallduví 1993, 2002), and it seems that something similar is at stake in Italian. This postverbal subject position is presumably outside VP and thus escapes the weak existential interpretation. Thus, it seems plausible to assume that this position is a grammaticalised topic position (cf. Lehmann 1976, Li & Thompson 1975) where the subject/the pivot DP can be located in Modern Catalan.

5. Some More Data In the following, some new data from historical Catalan and historical European Spanish as well as Modern Latin American Spanish will be discussed, arguing that word-order and its connection to information structure is in fact one of the main ingredients of the DR.

5.1 Some New Data from Modern Languages Building on Abbott’s (2005) intuition that the weak-strong distinction should not be taken as a binary opposition, but should rather be seen as a scale, Zielke (2012) carried out a study on the L2 acquisition of the DR in Spanish existentials and could indeed establish a tripartite distinction. The focus of her study was the acquisition of the DR in L2 European Spanish and L2 Latin American Spanish by L1 German and L1 Turkish speakers. Using an acceptability judgment task, her main goal was to show that L1 Germans outperform the L1 Turkish group since the DR plays out similarly in German and Spanish but differently in Turkish. As expected by Zielke (2012), the L1 German group outperformed the L1 Turkish group on this phenomenon due to positive L1 transfer. However, in my view another very interesting result is the tripartite distinction concerning the expressions triggering the DR that was manifested in the control group

Chapter Nine

322

of Latin American Spanish19 speakers as compared to the European Spanish speakers. (40)

weak expressions > less strong expressions > strong expressions -indefinite article -definite article -weak quantifier referent: inanimate -zero article, bare noun -strong quantifiers

-definite article referent: animate uniquely identifiable -strong quantifiers -proper names -possessives

The native speakers of Latin American Spanish (LAS) only showed the DR in strong expressions, whereas the native speakers of Modern European Spanish (ModSp) showed the DR both with less strong and with strong expressions (Table 1). weak expressions

ModSp LAS

no DR no DR

strong expressions less strong expressions DR

strong expressions

DR

Table 1. The weak/strong distinction in ES and LAS (Zielke 2012: 30). What the two languages seem to share is that strong expressions that are specified for [+ animate] can never occur in existential sentences. Zielke (2012) therefore argues for a tight connection between definiteness and animacy. See the following example of Latin American Spanish, where the animate definite DP is forbidden (41a), but the inanimate definite DP is accepted by the speakers (41b); both sentences are inacceptable in European Spanish. (41)

19

(a)

*Hay todas las personas de la empresa. LAS has all the people of the press ‘There are all the people from the press.’ (b) Hay todas las películas. has all the films ‘There are all the films.’ (Zielke 2012: 88)

Zielke (2012) uses the term Latin American Spanish since the speakers of her control group were from different countries, namely from Peru, Ecuador, Columbia, and Mexico, and because there was no difference among these speakers concerning their judgements of the DR.

Unaccusatives vs. Existentials

323

5.2 Some New Data from Old Languages Looking at Old Spanish20 and Old Catalan, we see that no change is detectable concerning the two constructions. In Old Spanish the DR was already active (42–44) in existentials only. (42)

Entre esas compañas de casa, OSp among these neighbours of house avié un omne bueno, que perdiera el viso; had a man good that lost the radiance ‘Among these neighbours of the house, there was a good man that lost his glamour;’ (Berceo: “Martirio de San Lorenzo”. 19/3)

(43)

En in /

(44)

(...) ca en l‫ތ‬otro ay 50 enxiemplos et en éste ay ciento. since in the’other are 50 examples and in this are 100 ‘Since in the other there are fifty examples and in this one there are one hundred.’ (Don Juan Manuel: “El Conde Lucanor”. Tercera Parte, p 288)

Coloña la rica, cabeza de regnado, OSp colony the rich head of reign avié un monasterio de Sant Piedro. was a monastery of Saint Peter ‘In the rich colony, head of the reign, was a monastery of Saint Peter.’ (Berceo: “Milagros de Nuestra Señora”. § 160)

Sentence (42) shows an indefinite animate DP, (43) an indefinite inanimate DP, and in (44) we see an inanimate plural DP. In unaccusatives no change seems to have happened from Old Spanish to Modern Spanish either. Already in the Old Spanish texts unaccusative sentences without the DR are attested (45) and (46). (45)

20

(...) que llegue la sennal de la otra estrella; OSp that arrive the signal of the other star ‘... that there arrived the signal of the other star;’ (Alfonso X, Folio 97R)

Regarding the different OSp texts: Milagros de Nuestra Señora and Martirio de San Lorenzo, Gonzalo Berceo, ~1252; El Conde Lucanor, Don Juan Manuel, ~1335; General Estoria, Las siete partidas, Setenario, Alfonso X, ~1270; Calila e Dimma, Anonymous, between 1195 and 1207; and Mio Cid, ~1207. Thanks to Mario Navarro, who checked the texts for existentials and unaccusatives and who created the excel table for further investigation.

324

(46)

Chapter Nine

Destroyda es Moab. & las sus cibdades. subieron los enemigos destroyed is M. & the his townsfolk ascended the enemies sobrellos. over.them ‘Moab and his townsfolk are destroyed. There ascended the enemies to them.’ (Alfonso X, Folio 97V)

In both unaccusative constructions a definite DP is used. In (45) it is an inanimate DP whereas in (46) it is a definite animate plural DP. The Catalan data is not as clear-cut; here we get a deviant pattern. On the one hand, concerning the existential construction we detect a change compared to Modern Catalan; on the other hand, no change is attested concerning the unaccusative construction. In Modern Catalan, as discussed in section three, existentials are accepted by native speakers of Catalan that allow definite DPs not only with a dislocated coda but also without a dislocated coda. In Old Catalan this doesn’t seem to be the case. No definite DP has been attested in any of the three Old Catalan texts we examined21 nor in the corpus CICA,22 which we examined with respect to existentials and unaccusatives. Thus, all existentials showed a strict DR with inanimate (50) and animate DPs (47–49) in contrast to Modern Catalan. (47)

Fèlix--dix l’ermità--, en una terra havia un rey OCat Felix says the’hermit in a land was a king qui molt amava justícia who much loved justice ‘Felix, says the hermit, there was a king in a land who loved justice very much.’ (Llull, Felix 30, 5-6)

(48)

E entorn de tots los dessusdits havia molts falcons, and around of all the said had many hawks ‘And around of all the ? there were many hawks,’ (Metge, Somni 18, 6-8)

21 The Catalan examples presented here are either from CICA or from the following three medieval texts: Llibre de Meravelles, Ramon Llull, ~1288; Lo Somni, Bernard Metge, ~1399; and Curial E Güelfa, Anonymous, between 1435 and 1462. Thanks to Jorge Vega Vilanova, who went through the texts and extracted all the existential and unaccusative examples. 22 Thanks to Xavier Villalba, who checked the texts included in CICA for existential sentences and who made his excel tables of existentials available to me.

Unaccusatives vs. Existentials

325

(49)

car aquí havia dos cavallers qui volien defendre per since here were two knights who wanted defend for batalla la honor de la duquessa battle the honour of the duchess ‘since here there were two knights who wanted to defend by battle the honour of the duchess.’ (Curial 67, 4-5)

(50)

encare.t prec que, si y ha alcunes coses, still.2SG.ACC plea that if LOC has some things ‘I ask you now that if there are some things,’ (CICA_Diàlegs St. Grego, 61)

Another development that took place from Old to Modern Catalan was a change in word-order that was more dramatic than, for example, in Spanish. In Old Catalan all different word-orders were possible, including VSO/VSX (Fischer 2010, 2013). In Modern Catalan word-order got fixed on VOS/VXS (Vallduví 2002), also allowing SVO/SVX but never VSO/VSX. Thus, I suggest that only when word-order got fixed on VOS/VXS was the DR no longer obligatory in Catalan existentials, assuming that the reason for this is the fact that postverbal subjects in Catalan (existentials) are allowed in a position outside VP where they can receive a topic interpretation and are thus not subject to the DR. No change has been attested over the centuries concerning the unaccusative constructions. Already in Old Catalan, definite DPs (52) and proper names (51) are possible. Thus, no DR applies in unaccusatives.

(51)

OCat Tantost e sens triga vengueren Jacob e Curial immediately and without delay arrived Jacob and Curial ‘There arrived Jacob and Curial immediately and without delay.’ (Curial 72/2)

(52) Mas aprés dos dies (...), arribà lo prior, e tots eren more after two days arrived the prior, and all were tornats sinó Ypòlit. returned but Y. ‘After two days (...), there arrived the prior, and all others had returned except for Y.’ (CICA_Martorell, Tirant, 729, 7). For a better overview of where the DR has been attested in the different languages over the centuries, two tables which include the different factors

Chapter Nine

326

mentioned above (null-subject language (NSL), non-null-subject language (NNSL), weak DPs, strong-inanimate DPs, strong-animate DPs) are given below. See table (2) for the DR in unaccusatives and table (3) for the DR in existentials.23 Weak Expressions indefinite articles, weak quantifiers, zero articles & bare nouns NNSL NNSL NNSL NNSL NSL NSL NSL NSL

OEng ModEng OFr ModFr ModSp LAS OCat ModCat

no DR no DR no DR no DR no DR no DR no DR no DR

Strong Expressions definite articles definite article + inanimate, + animate, strong strong quantifiers quantifiers + + animate, inanimate proper names DR DR DR DR DR DR DR DR no DR no DR no DR no DR no DR no DR no DR no DR

Table 2. Definiteness restrictions in unaccusatives. In unaccusatives there is a clear-cut difference between non-null-subject languages and null-subject languages. The DR applies in all non-nullsubject languages whereas it does not apply in null-subject languages, irrespective of whether the strong DPs are animate or inanimate.

23

I have included all those languages in the tables for which either diachronic or synchronic varieties have been investigated. Italian is not included since I have only looked at normative Modern Italian.

Unaccusatives vs. Existentials Weak Expressions indefinite articles, weak quantifiers, zero articles & bare nouns NNSL NNSL NNSL NNSL NSL NSL NSL NSL

OEng ModEng OFr ModFr ModSp LAS OCat ModCat

no DR no DR no DR no DR no DR no DR no DR no DR

327

Strong Expressions definite articles + definite article + inanimate, animate, strong strong quantifiers quantifiers + + inanimate animate, proper names DR DR DR DR DR DR DR DR DR DR no DR DR DR DR no DR no DR

Table 3. Definiteness restrictions in existentials. Concerning the existentials, the picture is not as clear-cut. It seems more factors need to be included in explaining the occurrence of the DR. First, we need to differentiate between strong animate DPs and strong inanimate DPs, and, as has been shown with respect to the Catalan data, it seems we also need to include word-order since it is directly connected to information structure in our interpretation of the data.

6. Attempting a Summary In summarising the findings on unaccusative and existential constructions concerning the different languages presented here, we get a complex picture with regard to existentials and a less complex picture with respect to unaccusatives. It seems correct to claim that whether the DR applies in unaccusatives or not is dependent on whether a language is a null-subject language or not. Independently of whether an expletive is introduced or not, independently of whether an animate DP or an inanimate DP is used, and also independently of word-order, the DR applies in non-null-subject languages and it does not apply in null-subject languages. It has been argued that the reason for this is the fact that in null-subject languages it is the verb that checks the EPP, allowing postverbal subjects in different positions, also outside the VP, and thus escaping a weak existential interpretation (Alexiadou & Anagnostopoulou 1998, Fischer 2010). Hence, I propose that the DR in unaccusatives is a syntactic phenomenon that needs to be explained by syntactic features.

328

Chapter Nine

Concerning the existentials the picture is much more complex. The DR applies irrespective of whether a language is a non-null-subject language or not. It has been shown that the DR in existentials is far more dependent on pragmatic and semantic features than in unaccusatives. First, the data presented here has shown that the weak-strong distinction is not sufficient to explain the occurrences of the DR in Latin American Spanish as compared to European Spanish. In order to capture this difference, the semantic feature [+/í animate] needs to be included (cf. Zielke 2012). Second, word-order and its connection to information structure plays an important role in explaining the DR. Data was presented showing that in Old Catalan the DR applied obligatorily in all existential sentences. Only after the word-order has been fixed on VOS/VXS are existential sentences attested in which the DR does not apply, assuming that this postverbal position is outside the VP, allowing it to be a topic position. It seems to be correct that the reason for this lies in the ways information structure is mapped onto syntax in Modern Catalan compared to the system in Old Catalan. Thus, contrary to Lyon’s (1999: 46) claim that the DR “is more likely to be a semantic or pragmatic constraint than a syntactic one”, I suggest that the DR in unaccusatives has to be explained in syntactic terms, whereas the DR in existentials is a syntactic, semantic, and pragmatic restriction that has to be explained by referring to the semantics of the constructions, the pragmatics of the context, and also the word-order, i.e. the information structure, of the language investigated.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of pragmatics 19, 39-55. —. 2005. Definiteness and Indefiniteness. In L. Horn & G. Ward (eds.): Handbook of Pragmatics, 122-149. Oxford: Blackwell Alexiadou, Artemis & Elena Anagnostopoulou. 1998. Parametrizing AGR: Word-Order, V-Movement, EPP-checking. In Natural Language and Linguistic Theory 16: 491-539. Alexiadou, Artemis, Elena Anagnostopoulou & Martin Everaert (eds.). 2004. The Unaccusativity Puzzle: Explorations of the Syntax Semantic Interface. Oxford: Oxford University Press. Alexiadou, Artemis & Florian Schäfer. 2011. An Unaccusativity Diagnostic at the Syntax-Semantics Interface: there-Insertion, Indefinites and Restitutive again. In Ingo Reich, Eva Horch & Dennis Pauly (eds.). Sinn und Bedeutung 15. Proceedings of the 2010 Annual

Unaccusatives vs. Existentials

329

Conference of the Gesellschaft für Semantik, 101-115. Saarbrücken: Universaar, Saarland University Press. Barbosa, Pilar. 1995. Null Subjects. Cambrigde, Mass.: MIT Press. Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry, 19: 1-34. Belletti, Adriana & Valentina Bianchi.This volume. Definiteness Effects and Unaccusative Subbjects: An Overview and some new Thoughts. Belletti, Adriana & Luigi Rizzi. 1988. Psych-Verbs and Theta-Theory. In Natural Language and Linguistic Theory 6: 291-352. Bentley, Delia, Francesco Maria Ciconte & Silvio Cruschina. 2013. Existential Constructions in Crosslinguistic Perspective. Special issue of Italian Journal of Linguistics 25.1: 1-13. Biberauer, Teresa & Ian Roberts. 2003. Parameter interaction and word order change in the history of English. Paper presented at the conference on Comparative Diachronic Syntax. University of Leiden, 29.-30. August 2003. Bianchi, Valentina & Cristina Chesi. 2014. Subject Islands, Reconstruction, and the Flow of the Computation. In Linguistic Inquiry 45.4: 525-569. Bonet, Eulalia. 1991. Morphology after syntax: Pronominal clitics in Romance. PhD-Dissertation. Cambridge, Mass.: MIT. Boucher, Paul. 2003. Determiner Phrases in Old and Modern French. In M. Coene & Y. D’Hulst (eds.). From NP to DP, Vol. 1. The syntax and semantics of noun phrases, 47-69. Amsterdam: Benjamins. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. Calboli, Gualtiero. 1978. Die Entwicklung der klassischen Sprachen und die Beziehung zwischen Satzbau, Wortstellung und Artikel. In Indogermanische Forschungen 83: 197-261. Carnie, Andrew & Heidi Harley. 2005. Existential Impersonals. In Studia Linguistica 59.1: 46-65. Cruschina, Silvio. 2015. Patterns of variation in existential constructions. In Isogloss. A journal on variation of Romance and Iberian languages 1.1: 33-65 Diesing, Molly. 1992. Indefinites. Cambrigde, Mass.: MIT Fischer, Olga, Ans van Kemenade, Willem Koopmann & Wulf van der Wurff. 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Fischer, Susann. 2004. The diachronic relationship between quirky subjects and stylistic fronting. In P. Bhaskararao & K.V. Subbarao (eds.). Non-nominative Subjects, Vol.1, 193-212. Amsterdam: Benjamins.

330

Chapter Nine

—. 2010. Word-order as a trigger for grammaticalisation. Amsterdam: Benjamins. —. 2013. Subjects, subject positions and word-order: Old Romance vs. Old Germanic. In Jörg Fleischer & Horst Simon (eds.). Comparing Diachronies, 119-141. Berlin/Boston: Walter de Gruyter. Freeze, Ray. 1992. Existentials and other locatives. In Language 68.3: 553–595. Gaatone, David. 1970. La transformation impersonnelle en français. In Le Français Moderne 38: 389-411. Grewendorf, Günter. 1989. Ergativity in German. Berlin: de Gruyter. Gutiérrez-Bravo, Rodrigo. 2006. A reinterpretation of quirky subjects and related phenomena in Spanish. In Jean Pierre Montreuil & Chiyo Nishida (eds). New Perspectives in Romance Linguistics, 127-142. Amsterdam: John Benjamins. Heggie, Lorie. 1988. The Syntax of Copular Structures. PhD-Dissertation, University of Southern California. Ingham, Richard. 2001. The structure and function of expletive there in premodern English. Reading Working Papers in Linguistics 5. 231249. Keenan, Edward. 1976. Towards a universal definition of ‘subject’. In Charles N. Li (ed.). Subject and topic, 303-334. New York: Academic Press. Kampers-Mahne, Brigitte, Jean-Marie Marandin, Frank Drijkoningen Jenny Doetjes & Aafke Hulk. 2004. Subject NP Inversion. In Francis Corblin & Henriëtte de Swart (eds). Handbook of French Semantics, 553–579. Stanford, CA: CSLI. Kroch, Anthony. 2002. Syntactic change. In M. Baltin & C. Collins, (eds.). The handbook of contemporary syntactic theory, 699-729. Oxford: Blackwell. Lehmann, Winfred P. 1976. From Topic to Subject in Indo-European. In Charles N. Li (ed.). Subject and Topic, 447-456. New York: Academic Press. Leonetti, Manuel. 2008. Definiteness Effects and the Role of the Coda in Existential Constructions. In A. Klinge & H. Hoeg-Müller (eds.). On Nominal Determination, 131-162. Amsterdam: John Benjamins. —. This volume. Definiteneness Effects: The Interplay of Information Structure and Pragmatics. Li, Charles N. & Sandra A. Thompson. 1975. Subject and Topic a new typology of languages. In Charles N. Li (ed.). Subject and Topic, 457489. New York: Academic Press.

Unaccusatives vs. Existentials

331

Löbel, Elisabeth. 2009. Copular verbs and argument structure: participant vs. non-participant roles. In Theoretical Linguistics 26.3: 229-258. Lohndal, Terje. 2006. The phrase structure of the copular. In Working Papers in Scandinavian Syntax 78: 37-75. Lumsden, Michael. 1988. Existential Sentences: Their Structure and Meaning. London: Routledge. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Mackenzie, Ian. 2006. Unaccusative Verbs in Romance Languages. New York: Palgrave. Maling, Joan. 1988. Variations on a theme: Existential Sentences in Swedish and Icelandic. In Denise Fekete & Zofia Laubitz (eds.). McGill Working Papers in Linguistics. Special Issue on Comparative Germanic Syntax, 168-191. Montreal: McGill University. McNally, Louise. 2011. Existential Sentences. In C. Maienborn, K.von Heusinger & P. Portner (eds.). Semantics: an international handbook of natural language meaning, 1829-1848. Berlin: De Gruyter. Milsark, G.L. 1974. Existential Sentences in English. Ph-Dissertation. Cambridge, Mass.: MIT —. 1977. Toward an Explanation of Certain Peculiarities of the Existential Construction in English. In Linguistic Analysis 3: 1-29. Moro, Andrea. 1991. The raising of predicates: Copula, expletives and existence. In Lisa L.S., Cheng & Hamida Demirdash (eds.). More papers on wh-movement, 119–181. Cambridge, Mass.: MITWPL. Moro, Andrea. 1997. The raising of predicates. Predicative noun phrases and the theory of clause structure. Cambridge: CUP. Müller, Natascha. 2000. Ergative und unergative Verben aus romanistischer Sicht. In BLicK: 43-52. Ordóñez, Francisco. 1998. Postverbal Asymmetries in Spanish. In Natural Language and Linguistic Theory 16: 313-346. Perlmutter, David M. 1978. Impersonal Passives & the unaccusative hypothesis, In Berkeley Lingusitic Society 4: 157-189. Rappaport Hovav, Malka & Beth Levin. 1992. -er Nominals: Implications for a Theory of Argument Structure. In T. Stowell & E. Wehrli, (eds.). Syntax and Semantics 26: Syntax and the Lexicon, 127-153. New York: Academic Press. Rinke, Esther & Jürgen M. Meisel. 2009. Subject-Inversion in Old French: Syntax and Information Structure, In Georg A. Kaiser & Eva-Maria Remberger (eds.). Proceedings of the Workshop "Null-subjects, expletives and locatives in Romance", 93-130. Konstanz: Universität Konstanz (= Arbeitspapier, 123). Safir, Kenneth. 1985. Syntactic Chains. Cambridge: CUP.

332

Chapter Nine

Schäfer, Florian. 2008. Event Denoting –er Nominalization in German. In Working Papers of the SFB 732 Incremental Specification in Context 01 (2008): 173-187. Sheehan, Michelle. 2004. An interface analysis of verb subject inversion in Romance. In DWPapers in Linguistics 10. Silva-Villar, Luis. 1998. Subject positions and the roles of CP. In Schwegler, B. Tranel & M. Uribe-Etxebarria (eds.). Romance Linguistics. Theoretical Perspectives, 247-270. Amsterdam: Benjamins. Sorace, Antonella. 2000. Gradients in Auxiliary Selection with intransitive Verbs. In Language 76.1: 859-890. Sorrenti, Elisa. 2015. Opacity effects and (in)definiteness: A contrastive analysis in languages with and without articles. PhD-Dissertation. Universität Hamburg. Vallduví, Enric. 1993. Catalan as VOS: Evidence from information packaging, In William J. Ashby, Marianne Mithun, Giorgio Perissinotto & Eduardo Raposo (eds.). Linguistic Perspectives on the Romance languages, 335-350. Amsterdam: Benjamins. —. 2002. L’oració com a unitat informativa. In J. Solà, M. R. Lloret, J. Mascara & M. P. Saldanya (eds.). Gramàtica del català contemporaní. Sintaxi, Vol. 2, 1221-1276. Barcelona: Empúries. Vet, Co. 1981. Les constructions impersonnelles en français: Une approche dans le cadre de la Grammaire Fonctionnelle de S. C. Dik. In Travaux de Linguistique 8: 49-64. Zaring, Laurie. 1998. Object Shift in Old French. In A. Schwegler, B. Tranel & M. Uribe-Etxebarria (eds.). Romance Linguistics: Theoretical Perspectives, 319-332. Amsterdam: Benjamins. Zielke, Marina. 2012. The Aquisition of the Definiteness Effects in L2 Spanish. M.A. Thesis. Universität Hamburg. Zimmermann, Michael. 2014. Expletive and Referential Subject Pronouns in Medieval French. Berlin: De Gruyter. —. This volume. Definiteness effects in French unaccusatives.

CHAPTER TEN VERBAL ASPECT AND DEFINITENESS EFFECTS IN CATALAN ABSOLUTE SMALL CLAUSES JORGE VEGA VILANOVA

1. Introduction The term Absolute Small Clause (ASC) designates a variety of adjunct constructions that can convey modal, temporal, causal, conditional, or similar information to the main clause. Independently of their interpretation, ASCs show some constant distinctive morphosyntactic properties, which have already been analysed extensively for many languages, such as Italian (Belletti 1992), French (Kayne 1989: 35) and Spanish (Hernanz 1991, López 1994, Gunnarson 1994, Marín 2002, Bruno 2011). The main focus of these studies, however, has been the syntactic characterization distinguishing ASCs from other clausal structures and, more specifically, features like word order, verb movement, or case assignment. The aim of this paper is to provide a new analysis of the syntactic behaviour of ASCs in Catalan by covering some features – definiteness and aspect– that are yet to be properly treated. Although there are some cross-linguistic variations – even in closely related languages – ASCs show some unique properties which seem to have a universal validity. An ASC is generally made of a predicate and a DP. The predicate is usually, but not necessarily, a past participle or deverbal adjective (e.g. acabat, ‘finished’, or ple, ‘full’ in Catalan). Other grammatical categories may form an ASC if they are able to fulfil similar predicative functions as the former categories. The nominal referent of predication may be implicit in some cases. Optionally, other elements such as prepositions, particles, or adverbial locations may introduce the clause. They usually stress the aspectual or temporal value of the ASC with regard to the main clause. In this paper I concentrate on participle clauses without any introducing element, assuming that the main lines of my analysis should apply, in principle and with little adaptation, to other types of

334

Chapter Ten

ASCs. However, there is very limited specific research on the restrictions on the DP itself. Examples (1), (2) and (3), differing only in the determiner of the DP, show a sharp contrast in grammaticality. Whereas sentence (1) is perfectly acceptable, (2) and (3), with a bare NP (‘deures’) or an indefinite DP (‘molts deures’) are categorically ruled out. This suggests that, in addition to the usually observed characteristics of ASCs, there is some kind of constraint on the referential value of the DP. (1)

[ASC Acabats els deures], podeu sortir a finish.PSTPRT.MPL the homework.MPL can.2PL go-out to jugar. play ‘Once you finish your homework, you can go and play.’

(2)

* [ASC Acabats deures], podeu sortir a jugar.

(3)

* [ASC Acabats molts deures], podeu sortir a jugar. ‘Once you finish much homework, you can go and play.’1

If one tries to characterize this construction, the first issue is to identify the kind of phenomenon we are dealing with: is this a semantic, pragmatic, or syntactic phenomenon? Assuming that there is a tight connection between verbal aspect and nominal reference (Krifka 1989, Leiss 2000, Fischer 2005), I will argue that the restrictions on this DP are fundamentally determined by structural factors. I also claim that although they are similar in nature, the definiteness restriction one finds in existential sentences should be treated separately from the definiteness restriction applying on ASCs. The influence of semantics and pragmatics in this phenomenon is not immediately ruled out, but a structural account has the advantage that it can explain the relevant data while relating them to apparently unconnected phenomena. I show that a diachronic change induces modifications on the aspectual features of past participles from Old Catalan to Modern Catalan. For independent reasons (case assignment, Diesing’s (1992) Mapping Hypothesis), the object DP stays in a special syntactic relation with the participle during the derivation. I argue that the perfective feature of the participle imposes the definiteness restriction

1

In these and following examples the acceptability of the sentences can be improved by introducing the ASC with ‘un cop’, once, or similar expressions. They are often used in the English translations. The alternative without ‘un cop’ is, although less natural, still accepted as grammatical.

Catalan Absolute Small Clauses

335

illustrated in (1) to (3): since definiteness is tightly related to perfectivity, only definite DPs can co-occur with the past participle in ASCs. The paper is organized as follows: first, I describe the syntactic features that have been postulated for ASCs in Romance languages so far. I especially emphasize the organization of the functional projections within the ASC. In section 3, I provide some theoretical background about definiteness and aspect with special attention to Catalan data. Next, I discuss some relevant Old Catalan data. In section 5, I propose an analysis for language change, which also considers the interplay between definiteness and aspect. In the last section I sum up some of the results of the analysis, and outline some implications for language change and the development of verbal aspect and definiteness in Romance languages.

2. Syntactic Properties of Absolute Small Clauses Absolute Small Clauses are usually characterized as predicative constructions exhibiting a reduced clausal structure. They are composed of an adjective or past participle and a DP to which the predication applies. This relation is frequently visible in the obligatory morphological agreement of nominal features (number and gender) in the predicate and DP. Beyond this, there is little consensus about what the syntactic structure of ASCs should look like. The relative constituent order is subject to parametrical variation and may be more or less strongly fixed. Although ASCs are endued with a reduced syntactic structure in comparison with full clauses, it is not really clear which functional categories (C, T, Neg…), if any, are missing. The relevance of this discussion becomes apparent when we attempt to explain the constraints on the ASCs. Diachronically, this construction could be taken as the result of a progressive erosion of a much permissive (and richer) structure. Some changes to the functional material of the clause would activate specific restrictions that can be found nowadays in ASCs across Romance. Unfortunately, we have only limited information about ASCs in Catalan, and a detailed analysis of the evolution of this construction from Old Romance to Modern Romance languages has largely been neglected. Thus, in the remainder of this section, I summarize the main properties of ASCs in Italian and Spanish, two rather extensively studied Romance languages, before checking their adequacy in connection with Catalan ASCs.

336

Chapter Ten

2.1 ‘Quirkiness’ of Absolute Small Clauses It is commonly assumed that ASCs can only support a limited amount of functional structure. This fact gives rise to some of the peculiarities or ‘quirks’ of ASCs. How precisely this reduced structure is understood, however, may vary from one approach to the other. Some of the most prominent ‘quirks’ of ASCs in Romance languages are summarized by Alcázar and Saltarelli (2007). In their account, Italian ASCs have quite a simple syntactic structure lacking the C, T, and Ȟ* projections altogether: [VP V O]. This clearly contradicts Belletti (1992), who postulated a more complex (but not complete) functional layer. From a different point of view, López (1994) delimits the functional features of ASCs by comparing them to a similar clausal construction with rather different syntactic behaviour, namely gerundival clauses. The adjunct clause in (5), for instance, can overtly encode tense (through the auxiliary verb) and negation (through the adverbial no). This functional material is lacking in ASCs, as shown by the ungrammaticality of (4). The construction in (5) is known as gerundival construction (‘Absolute Full Clause’ in López's terminology). (4)

(*No) Desaparecidas las joyas, not disappear.PSTPRT.FPL the jewels.FPL buscaron al ladrón searched-for.3PL to-the thief ‘The jewels missing, they looked for the thief.’

(5)

No habiendo encontrado pruebas, llamaron not have.GER find.PSTPRT.MSG proof.FPL called.3PL a la policía. to the police ‘Since they couldn't find any proof, they called the police.’

However, the distinction between ASCs and gerundival clauses might be less clear-cut than suggested above. In fact, at least some native speakers accept sentences of type (4), if certain conditions are met. Thus, a revision of the assumed properties of ASCs is needed before probing into the analysis of the definiteness restrictions prevailing in ASCs.

Catalan Absolute Small Clauses

337

2.1.1 ASCs and Negation To begin with, it has often been taken for granted that ASCs cannot be negated. However, there is some data challenging this assumption. The sentences in (6) are acceptable for many native speakers, although they are rather marked. (6a) and (6b) clearly contrast in acceptability with (4), the only difference between them being the lexical verb (desaparecer vs. empezar and encontrar). Coordination, especially when linked to contrastive focus, seems to improve even sentences with verbs like desaparecer, as shown in (6c) and (6d). (6)

a. (Aún) No empezada la carrera, los hinchas (yet) not begin.PSTPRT.FSG the race.FSG the fans ya aplaudían. already clapped ‘Before the race began, the fans were already clapping.’ b. No encontradas las joyas, tuvieron que seguir not find.PSTPRT.FPL the jewels.FPL had.3PL to keep buscando. searching ‘Since they didn’t find the jewels, they kept looking for them.’ c. Robadas, no desaparecidas, las steal.PSTPRT.FPL not disappear.PSTPRT.FPL the joyas, buscaron al ladrón. jewels.FPL searched-for.3PL to-the thief ‘As the jewels had been stolen, not disappeared, they looked for the thief.’ d. Perdidos y aún no encontrados los lose.PSTPRT.MPL and yet not find.PSTPRT.MPL the hilos de conversación, ... threads.MPL of conversation ‘Lost and not yet found the threads of the conversation, ...’

It is, therefore, less than evident that small clauses are devoid of a functional structure. In contrast to what has often been assumed, negation is possible under certain circumstances. More precisely, the examples leading to the exclusion of NegP from ASCs seem to be rather biased. There is a difference in inner aspect between verbs like desaparecer and verbs like encontrar or empezar. The first group gets an atelic interpretation in absolute clauses and disallows negation, while the second one has a telic reading and is generally compatible with the negation. The

338

Chapter Ten

same effect can be found in finite adjunct clauses like (7) and (8). The presence or lack of negation does not seem to be directly related to either a reduced syntactic structure of the ASC or to lexical idiosyncrasies. Instead, it seems to be correlated with the aspectual configuration of the eventuality and more concretely with telicity. Previous studies only took into account verbs favouring atelic readings and thereby led to the conclusion that ASCs cannot be negated, contrary to the facts. (7)

a. Una vez / Cuando hubieron desaparecido once / when had.3PL disappear.PSTPRT.DEF joyas, ... jewels.FPL ‘As soon as the jewels disappeared, …’ b. * Una vez / Cuando no hayan desaparecido las joyas, ... * ‘As soon as the jewels did not disappear, ...’

(8)

a. Una vez / Cuando hayan encontrado once / when had.3PL find.PSTPRT.DEF las joyas,... the jewels.FPL ‘As soon as they find the jewels, …’ b. ? Una vez / Cuando no hayan encontrado las joyas,... * ‘As soon as they do not find the jewels, …’

las the

2.1.2 ASCs, Tense, Voice and Aspect Assuming that negation is available in the structure of small clauses, it is appealing to think that other functional projections may be present as well. There is enough evidence to conclude that the TP could be partially, but not completely, projected in ASCs. First, examples like (9b) and (10b), for Italian and Spanish respectively, suggest that tense is not encoded in the structure of ASCs. Auxiliary verbs yield ungrammatical results. Unlike De Miguel (1992), one could argue that tense is not conveyed by ASCs. In most cases, however, the ASC is interpreted as anterior to the main clause. In the same fashion as other adverbial values are inferred from the context, this anteriority value does not need to be syntactically represented; it is rather a pragmatic inference departing from other related features. Could the aspectual value (in general, perfectivity) be also contextually inferred? Or is there any positive evidence for some functional projection hosting aspectual features? I argue that there are at least two further functional

Catalan Absolute Small Clauses

339

features encoded in the participle: voice and external aspect. The fact that perfectivity can be pragmatically translated into anteriority supports the claim that tense in ASCs is inferential. I come back to this point in section 2.1.4. (9) a. Desaparecidas las joyas, (...) b. * Habido/as desaparecido/as have.PSTPRT.DEF/FPL disappear.PSTPRT.DEF/FPL las joyas, (...) the jewels.FPL (10)

a.

b.

Arrivata Maria, Gianni tirò arrive.PSTPRT.FSG M., G. escaped un sospiro di sollievo. a sigh of relief ‘Maria [having] arrived, Gianni was relieved.’ * Stato/a arrivato/a be.PSTPRT.DEF/FSG come.PSTPRT.DEF/FSG Maria, (...) Maria

Related to this, it has been claimed that the predicate of ASCs must be a Stage Level one (SL), with Individual Levels (IL) predicates being excluded (Kratzer 1995, basing on Carlson 1977). This explains the ungrammaticality of the Spanish examples in (11) and (12), all they taken from López (1994: 46 and ff.). As shown in (11a), it is not possible to have an ASC with an IL predicate like saber ‘to know’. Such sentences are not better if the indefinite DP is substituted by a definite one (las fórmulas instead of muchas matemáticas) (11b). Ungrammaticality does not arise any more with an SL predicate like aprender. The same explanation applies to (12): hablar is an IL predicate (it means the capacity of an individual to speak a language, but not the event of speaking that language in a certain moment) and therefore ruled out, while decir, an SL predicate with a definite argumental DP, is fully grammatical. (11)

a. * Sabidas muchas matemáticas, (...) known.PSTPRT.FPL many maths.FPL ? ‘Knowing Maths very well, (...)’ b. * Sabidas las fórmulas, (...) known.PSTPRT.FPL the formulas.FPL ? ‘Knowing the formulas, (...)’

340

Chapter Ten

c. Aprendidas las fórmulas, podéis learn.PSTPRT.FPL the formulas.FPL, can.2PL salir a jugar go-out to play ‘Having learnt the formulas, you can go to the break.’ (12)

a. * Hablado español, (...) speak.PSTPRT.MSG Spanish.MSG ? ‘Speaking Spanish, (...)’ b. Dicho todo lo que tenían que decir, (...) say.PSTPRT.MSG all that had to say ‘Once they said all they had to say, (...)’

As suggested by Arche (2012), the distinction between SL and IL predicates could be the result of a syntactic operation. The restriction on this kind of ‘aspectual’ differentiation would hint at a sort of functional projection placed in a low structural position. Additionally, there are good arguments to believe that ASCs in Spanish are passive clauses. On the one hand, only verbs with internal arguments (transitive and unaccusative verbs) are possible in absolute constructions (López 1994: 46, Belletti 2006). Following Burzio’s Generalization (Burzio 1986: 178), the object DP must occupy a derived subject position where it can get assigned nominative case. We must remember that ASCs have been used as a classical test to discern between unaccusative and unergative intransitive verbs in Modern Romance languages: only unaccusative verbs can appear in this construction. The only way to introduce the external argument of a transitive verb (although this is subject to some cross-linguistic variation in acceptability among Romance languages) is with an agentive by-phrase as in (13)2. On the other hand, case marking of personal pronouns provides another argument for the analysis of ASCs as passive clauses in Spanish. As (14) shows, only nominative forms, but not accusative (or oblique) ones, are accepted in this construction3. 2

In the Minimalist framework of Alcázar & Saltarelli (2007), the impossibility to insert an external argument in an A-position of the VP leads to the conclusion that neither T nor Ȟ* are available in ASCs. For these authors, thus, the passive character of ASCs is taken as evidence that this construction has an even smaller range of functional projections than usually postulated. 3 It is so in Spanish and Catalan, but it must not necessarily be so in every other Romance language or in each of their varieties. As Belletti (2006: 82) shows, Italian ASCs also admit accusative clitics (i). She, thus, concludes that the small clause in Italian do not have a passive structure. However, the same restrictions on

Catalan Absolute Small Clauses

341

(13)

Encontradas las joyas por la policía, la señora find.PSTPRT.FPL the jewels.FPL by the police, the woman se tranquilizó. calmed down ‘As soon as the police found the jewels, the woman calmed down.’

(14)

Desmayada *mí / yo, (...) faint.PSTPRT.FSG *me/I ‘When I fainted, (...)’

Furthermore, as suggested by Bartra-Kaufmann (2007) for main clauses, passive sentences in Modern Romance languages do not allow bare noun subjects. If ASCs are passive clauses, it is expected that such restrictions also apply. The ungrammaticality of the examples of ASCs with bare nouns (2), (12a), and (15), therefore, conforms to this expectation. In summary, it can be argued that Spanish ASCs should be analysed as passive constructions, as the preceding examples illustrate. This, however, still leaves the restrictions on definiteness unexplained. (15)

* Aprendidas fórmulas, podéis salir learn.PSTPRT.FSG formulas.FSG, can.2PL go-out jugar play ? ‘Having learnt formulas, you can go to the break.’

a to

2.1.3 ASCs, Case Assignment and the CP-domain Another important issue on the study of Romance ASCs has been case assignment. Considering the assumption that these constructions lack (almost) every functional projection over the VP, what can then be an appropriate case assigner for the argumental DP? Different accounts have proposed different mechanisms to explain case assignment. According to Alcázar & Saltarelli (2007), a possible solution for this problem would acceptable predicates and verb classes hold for Italian ASCs (only unaccusative and transitive verbs, and SL predicates are grammatical). These restrictions deserve additional clarification and they challenge a cross-linguistic explanation for the phenomenon. (i) Conosciutami / Salutatala / Incontratici known-me.ACC / greeted-her.ACC / met-us.ACC ‘[Having] known me / greeted her / met us, (...)’

Chapter Ten

342

appeal to the Inheritance Model of Chomsky (2005), who claims that C can be responsible for nominative case assignment in absence of a full set of phi-features in T. If this be so, ASCs would skip the T-domain and directly project the CP-layer. Also, Belletti (1992) postulates an obligatory C projection with some classes of verbs (i.e., with unaccusatives). She also derives the fixed verb-initial word order in Italian and Spanish from this structure. However, the sentences in (16) show that there is no possible landing site for dislocated elements (be it focus or topic) in ASCs. (16a) has a fronted focus in the ASC and results in ungrammaticality. (16c), in opposition to (16b), has a clitic left-dislocated constituent. Even without the usually compulsory clitic doubling, the sentence is out. (16)

a. *[SC[Por la policía]FOC encontradas las joyas], la by the police find.PSTPRT.FPL the jewels.FPL the señora se tranquilizó. woman calmed down b. Una vez entregado el premio al Once give.PSTPRT.MSG the price.MSG to-the ganador, procedieron con la ceremonia. winner, proceeded.3PL with the ceremony ‘After giving the price to the winner, they proceeded with the ceremony.’ c. * [Al ganador, una vez entregado(le) el premio], to-the winner once give.PSTPRT.MSG the price.MSG procedieron con la ceremonia. proceeded.3PL with the ceremony

Therefore, there is no evidence for the presence of a CP projection, but there must be at least some functional structure in the TP-domain. Case assignment would, therefore, be assigned within the TP. Although this claim will be questioned afterwards, it seems clear that the TP in ASCs could be complex enough to host syntactic features such as aspect, voice, and even case assignment.

2.1.4 ASCs and Aspect So far, I have argued that ASC cannot only have the reduced structure that Alcázar and Saltarelli (2007) proposed for them, namely [V O], but they must be composed of some additional functional projections: [TP[VP[VP V (DP)]. Now, the object DP and the past participle (or adjective) must agree in gender and number. Assuming that ASCs are passive clauses, this is not

Catalan Absolute Small Clauses

343

unexpected. As in (17), the underlying object shows obligatory participle agreement, and also triggers number and person agreement with the auxiliary verb from the derived subject position. On its way to Spec,TP, the object DP would be forced to pass through the syntactic position responsible for object agreement (e.g. AgrOP). The same phenomenon is observed in contexts where the object, for one reason or another, has to move over the verb, be it through cliticization, wh-movement, or other syntactic operations (see Taraldsen 1987, Lefebvre 1988, Kayne 1989, Parodi 1995, Belletti 2006 among many others). A similar movement could be postulated for the object DP in ASCs. (17) [TPLas joyasi [T fueron [AGROP t’i robadasj [VP tj ti.]]] the jewels.FPL were steal.PSTPRT.FPL ‘The jewels were stolen.’ However, this movement is problematic with regard to two respects: first, it does not account for the obligatory verb-initial order of ASCs; and second, unaccusative verbs, which do not require participle agreement in finite clauses in Modern Catalan, trigger obligatory agreement in ASCs. As for the second question, I cannot provide a definitive answer here. I presume that the object movement in ASCs triggers participle agreement on the way over the verb (possibly for the necessity of case assignment), while the movement of the underlying object in finite clauses would involve different mechanisms, leading to default morphology of the participle and no overt object-verb agreement. As for the first question, the surface word order points to a recursive verb movement, implying additional functional material in the TP, as already proposed in the preceding sections.4 In some stage of the derivation, the object DP must be placed in a position that triggers object agreement, labelled AgrO for the sake of convenience. Moreover, such a projection has often been claimed to be related to the aspectual configuration of the clause: aspect can be understood as an interplay between verbal and nominal quantification, i.e., the sentential aspect is expressed either in the verbal, or in the nominal morphology of the object (through case marking or definiteness), or in a 4

Under the assumption that agreement with DPs in situ is possible in Romance languages (e.g. Alexiadou & Anagnostopoulou 1998, Chomsky 2000 and subsequent work), the discussion about constituent order in ASCs loses ground. Nevertheless, it does not change anything in the claim that there must be some more material in the TP of ASCs than usually admitted, irrespective of whether the functional features hosted there attract the XP they agree with or not.

Chapter Ten

344

combination of both factors (Krifka 1989, Leiss 2000, and Fischer 2005). ASCs encode the distinction between IL and SL predicates as well as inner aspect. Intuitively, ASCs may also carry external aspect information. Travis (1991) and Bruno (2011), for instance, claim that there is a low syntactic position, within the ȞP, that confers perfectivity on the clause. Thus, perfect aspect would be obtained independent of the contextual contribution, without resource to pragmatic inferences, against Alcázar & Saltarelli (2007: 10 fn. 6) and De Miguel (1992: 63 and ff.). Moreover, the insertion of some adverbial locations reinforcing the aspectual interpretation of the small clause (una vez ‘once’, todavía, aún ‘still’) seems to support the claim that aspect is syntactically realized in ASCs. (18)

a. Una vez encontrado Once find.PSTPRT.MSG b. Aún desaparecidas Still disappear.PSTPRT.FPL

el ladrón... the thief.MSG las joyas... the jewels.FPL

In order to explain the main properties of ASCs (word order, case assignment, and participle agreement), a more detailed syntactic structure over the VP seems to be needed. In the following section I discuss some Modern Catalan data, which, to my knowledge, have not yet been analysed, before I turn to the definiteness restrictions on the DP.

2.2 Quirkiness of Catalan ASCs As already mentioned in the preceding sections, there are some variations in the realization of ASCs among Romance languages. For this reason, I show which properties can be established for Catalan before turning to the discussion of definiteness effects in section 3 and the diachronic evolution of Catalan ASCs in section 4.

2.2.1 ASCs and Negation Catalan ASCs share most properties with ASCs in other Romance languages.5 There are, however, some relevant differences suggesting a more rigid clausal structure than in Italian or Spanish. 5 If not otherwise indicated, all Catalan examples and grammaticality judgments correspond to speakers of the dialect spoken in Tortosa, a transition dialect between Occidental and Oriental Catalan. Although the main lines of the analysis should hold for other dialects of Catalan, variation should not be excluded without further research.

Catalan Absolute Small Clauses

345

Parallel sentences to Spanish (6), (7), and (8) are also possible in Catalan, but they sound more marked. Negation with deverbal adjectives is generally acceptable (19a). Negative sentences like (19b), with the adverbial reinforcement encara, are also grammatical. ASCs with a simple participle, however, are felt as only marginally acceptable (19c). Coordination does not substantially improve their acceptability (19d). So, there is enough evidence that NegP is present in Catalan ASCs. Telicity and predicate class also seem to play a role in Catalan ASCs: (19c), with a telic verb, is slightly better than (19e), with an atelic one; individual-level predicates are excluded (20). (19)

a. No satisfeta amb la resposta,... not satisfied.FSG with the answer,… ‘Unsatisfied with the answer, ...’ b. Encara no començada la cursa,... still not begin.PSTPRT.FSG the race.FSG, ... ‘Before the race began,…’ c. ? No trobades les joies... not find.PSTPRT.FPL the jewels.FPL d. ? Robades, no desaparegudes, les steal.PSTPRT.FPL not disappear.PSTPRT.FPL the joies,... jewels.FPL ‘As the jewels had been stolen, not disappeared, …’ e. * No desaparegudes les joies, ... not disappear.PSTPRT.FPL the jewels.FPL

(20)

a. * Parlat català (per nosaltres), l'hostaler speak.PSTPRT.MSG Catalan (by us) the host va ser més acollidor was more friendly ‘Since we spoke Catalan, the host was friendlier.’

346

Chapter Ten

b. * Sabuda la lliçó, podreu sortir a know.PSTPRT.FSG the lesson can.FUT.2PL exit to l'esbarjo.6 the break ‘Only if you know the lesson, you will be allowed to take the break.’

2.2.2 ASCs, Tense and Voice Tense is not available in Catalan ASCs, as shown by the fact that auxiliary verbs are categorically excluded (21a). The interpretation of ‘anteriority’, as in Spanish, would be inferred from the perfective aspect of this construction (see below). Whenever tense needs to be made explicit, the ASC is replaced by a gerundival clause (21b). (21)

a. * Hagut vingut l'últim have.PSTPRT.DEFAULT come.PSTPRT.MSG the last alumne, van començar la classe. pupil.MSG, began.3PL the class ‘When the last one pupil had come, they started the lessons.’ b. Havent vingut l'últim alumne,... have.GER come.PSTPRT.DEFAULT the last pupil

As in the Spanish data, only transitive and unaccusative verbs can be found in Catalan ASCs. They can also be considered as passive clauses. Evidence for this comes from the pronominal substitution (only nominative pronouns like ella are acceptable, but its accusative counterpart la is out (22a)) and from the possibility to introduce the agent argument with an agentive by-phrase (22b) (However, it usually remains implicit. Actually, in most cases, the agentive adjunct leads to odd or even slightly unacceptable constructions). (22)

6

a. Triada ella / *-la entre totes choose.PSTPRT.FSG she / *her among all candidates,... candidates ‘Since she was chosen among all candidates,...’

les the

As an anonymous reviewer correctly remarks, this sentence could be (marginally) acceptable for some speakers. In this case, however, the verb would not be interpreted as an IL predicate, but rather as a SL one, closer to the meaning of aprendre ‘to learn’.

Catalan Absolute Small Clauses

347

b. (Un cop) comprada la llet (per la (once) buy.PSTPRT.FSG the milk.FSG (by the mare), farem el pastís. mother), make.Fut.1PL the cake ‘As soon as (the mother) will buy the milk, we'll make the cake.’ Furthermore, the fact that bare nouns are not permitted in ASCs is an expected consequence of them being derived subjects in a passive clause. Bare nouns present a quite restricted distribution in Romance languages: they cannot appear as subjects of unergative verbs under normal conditions, and they also cannot be the derived subjects of passive sentences, as Bartra-Kaufmann (2007) shows. However, as an anonymous reviewer correctly points out, there are several counterexamples to this rule. The clauses in (23) to (25) are perfectly acceptable: (23)

a. Tancades portes i finestres... close.PSTPRT.FPL door.FPL and window.FPL ‘After closing (all) doors and windows...’ b. Donades circumstàncies especials... give.PSTPRT.FPL circumstance.FPL especial.FPL ‘Given especial circumstances...’

(24)

a. Un cop vista una, ja les once see.PSTPRT.FSG one.FSG already them.FPL has vistes totes. have.2SG see.PSTPRT.FPL all.FPL ‘When you see one of them, you have already seen all of them.’

(25)

Fet esment... make.PSTPRT.MSG mention.MSG ‘Having mentioned...’

(26)

Un cop portada faldilla, ja no pots once wear.PSTPRT.FSG skirt.FSG yet not can.2SG posar-te pantalons. dress trousers ‘When you are used to wear skirt, you cannot wear trousers anymore.’

348

Chapter Ten

In (23) and (24), the bare noun is only possible if it can be interpreted as specific. Hence, (23a) does not mean an arbitrary quantity of windows and doors getting closed, but rather all of the available windows and doors in a specific place. A similar reasoning applies to (23b): it is some concrete circumstances that are referred to but not overtly expressed. (24a) can be similarly interpreted as referentially bounded (i.e. specific): the indefinite pronoun una refers to a subpart of the object in the main clause. It, therefore, gets a specific reading (cf. Enç 1991 and related work on the specificity of indefinite DPs). (25) should be analysed in a different way. The object here can be considered as a component of a verb-noun collocation. Fer esment, in fact, is equivalent to esmentar. The case of (26) is more complex. Here, the DP is taken non-referentially, similar to the Italian sentence in (27). Un lavoro in the Italian example has also a generic interpretation. Alternatively, portar faldilla and portar pantalons can be understood as fixed expressions that do not admit any nominal modification of the object DP. In any case, these examples do not seem to undermine the general claim that ASCs are passive clauses and consequently they show a strong tendency to reject bare nouns in subject position. (27)

Finito un lavoro, è piacevole prendersi una vacanza. ‘finished a task it is nice to take a vacation’ (from Alcázar & Saltarelli 2005: 5)

2.2.3 ASCs and the CP-domain There is no evidence in Catalan for a CP-domain in ASCs. Similar to (16), the sentences in (28) illustrate that neither focus fronting nor left dislocation is possible in this construction. With this respect, Catalan shows the same restrictions as Spanish ASCs. (28)

a. * [FOCPer la policia] trobades les joies by the police find.PSTPRT.FPL the jewels.FPL (i no pas pel majordom), ... (and not by-the butler) b. * [TOP Al guanyador, un cop donat(-li) to-the winner once give.PSTPRT.MSG(-him) el premi, va començar la cerimònia. the price.MSG began the ceremony ‘After giving the price to the winner, they proceeded with the ceremony.’

Catalan Absolute Small Clauses

349

2.2.4 ASCs and Aspect Finally, word order constraints and the obligatory participle agreement in Catalan ASCs are equivalent to Spanish patterns. Hence, with the exception of some aspectual adverbial expressions like un cop, tan bon punt, encara, etc., these clauses are strictly verb-initial, and the past participle (or deverbal adjective) must agree in number and gender with the object DP. As illustrated by the examples given so far, Catalan ASCs do not significantly differ from other Romance languages. The structure proposed for Spanish ASCs seems to be justified for their Catalan counterpart. Not surprisingly, Catalan shows the same definiteness restriction attested in Spanish and other Romance languages (cf. Marín 2002) and hence the ungrammaticality of (29): (29)

* (Un cop) cuites unes poques patates, (once) cook.PSTPRT.FPL some few potato.FPL podem començar a preparar la carn. can.1PL begin to prepare the meat ‘When a few potatoes are cooked, we can begin to prepare the meat.’

At a first glance, we are dealing here with a definiteness restriction similar to the one found in existential constructions and perhaps also in sentences with unaccusative verbs. The first question is, then, in which domain the definiteness effect originates. Since the interpretation of the ASCs as temporal, concessive, conditional, causative, or other adjunct type has a tight relation with pragmatic factors, it could be argued for a pragmatic account to explain the ungrammaticality of (29). Bare nouns and indefinite DPs would violate some pragmatic requirements of the clause. It is generally assumed that determiners are roughly linked to information structure: definite determiners usually introduce old or already known referents into the discourse, while indefinite determiners convey new information. This means that ASCs are only able to contribute old information to the utterance. Whenever new information is going to be provided within a clausal adjunct, its realization is always a gerundival clause (like (21b)). Why should these two closely related adjunct constructions differ in such a marked way? There seems to be no a priori reason excluding new referents from one structure but not from the other. To consider the definiteness restriction in ASCs as a consequence of pragmatic requirements is, thus, unmotivated. On the contrary, the

350

Chapter Ten

limitations on the object DP in ASCs (only definite DPs being admitted) lead to a pragmatic effect with the consequence that only old/known information, retrievable from the preceding discourse or from the extralinguistic context, can be encoded within the ASC. In section 2.1.4, I claimed that ASCs must have an AspP over the VP. In the following, I pursue the idea that the ungrammaticality of (29) is closely related to precisely this syntactic projection. A last remark is due at this point. The usage of ASCs is felt as belonging to a high register of the language, and it generally has an archaic flavour. Many of the adduced examples are more or less marked and not very productive in informal speech. Instead, the gerundival clause is used with increasing frequency. The progressive replacement of ASCs by gerundival clauses (or even full-fledged subordinate clauses) suggests the possibility of an ongoing language change. In the next sections I also explore the relation between aspect and reference on the one hand, and language change on the other. I claim that an alteration in the aspectual system in Old Romance could have motivated the properties observed in ASCs nowadays. To start with, I expose some basic notions about definiteness and definiteness restrictions (which has received an intensive attention, especially for English existential sentences) before turning to the analysis of Old Catalan data.

3. Some Remarks on Definiteness and Aspect Since the 1970s there are numerous works on the possibilities and restrictions on the DP in existential or presentational sentences. The fact that (30) is perfectly grammatical, but not (31), has given rise to the socalled Definiteness Effect (DE): the restriction for certain DPs (more specifically, those with a certain type of determiners) to appear in some syntactic positions. (30)

There is a wolf at the door.

(31)

* There is the wolf at the door. (examples from Milsark 1977: 4)

The DE posed two main problems to the linguistic theory. On the one hand, it was not (and it is still not) clear how to define such a vague concept as ‘definiteness’. Even if the distribution of lexical determiners could be clearly distinguished between definite and indefinite ones, there are still some controversial cases (indefinite use of demonstratives,

Catalan Absolute Small Clauses

351

problematic categorization of universal quantifiers, etc.). This picture is also blurred due to the overlapping of different constructions with similar surface structures (list readings, deictic uses of there... See Lyons (1999) for further details). On the other hand, the very nature of the DE is not uniformly defined. Recent accounts oscillate from a more syntactic perspective to a semantic or pragmatic one (cf. Lyons 1999, Leonetti 2008, Villalba 2013). Since determiners are usually used to distinguish new from old entities in the discourse, the DE could be intuitively ascribed to pragmatic requirements or information structure constraints. However, this cannot be the whole explanation, due to the fact that the DE is subject to parametrization, even in closely related languages. Italian, for instance, has presumably no definiteness effects (or only a very weak restriction), while Catalan has a rather flexible adaptation of DE. At last, French and Spanish have a more restrictive ‘version’ of it. Why should pragmatic factors have such different repercussions in different languages? The alternative view seems more convincing: both semantic and syntactic factors would play an important role in the configuration of existential sentences. The discussion on definiteness and DE has been extended to other apparently unrelated structures such as superlatives, internal-head relatives, and others (Lyons 1999: 246 and ff.). The connection between existential constructions and unaccusativity, for instance, has been amply debated (e.g. Belletti 1988, Fischer 2013). Examples like those in (32) are reminiscent of the type of constraint found in existential constructions. Hence, an account unifying the DE in existential and unaccusative sentences is desirable. (32)

a. ?* There comes the wolf. b. There comes a wolf.

What one can conclude from these observations is that (a) definiteness cannot be defined only from a semantic and pragmatic perspective, but it rather needs a complex view combining syntactic and semantic-pragmatic factors; and (b) definiteness has far-reaching implications in a variety of grammatical structures so that in some way or another it interacts with the syntactic configuration of the clause leading to definiteness restrictions in some specific positions of the sentence. Definiteness is, therefore, sensitive to the morphosyntactic context, and it is not the mere consequence of some discourse restrictions. If this is on the right track, it can be assumed that definiteness may interact with other grammatical features. This is indeed what seems to

352

Chapter Ten

occur with verbal aspect. It has often been claimed that both features (definiteness and aspect) are only different expressions of one and the same phenomenon, on the nominal or on the verbal domain, respectively (Krifka 1989, Leiss 2000 and 2007, Fischer 2005). For example, Leiss (2007) illustrates the reciprocal influence between aspectual marking on the verb form or verb stem and the case marking on the object DP in languages like Russian and Finnish in connection with the development of structured article systems. She proposes a general explanation for the language change process, which can be extended to other language families. She shows that the evolution of the definite article in Germanic languages (Old Icelandic, Gothic, and Old High German) has something to do with the weakening and loss of aspectual distinctions. So, a new way of marking the disappearing aspectual features of the verb is applied to the object DP itself (33): Gothic verbs lose explicit perfective marking (as the prefix ga-) whenever the definite object is accompanied by a definite determiner, albeit maintaining the perfective meaning. In the same way, Finnish can distinguish the aspectual characterization of events by marking the direct object with partitive (for imperfect aspect) or genitive case (for perfect aspect) (34). (33)

jah ullai waurÞun CONJ. full.NOM.PL.M.ST became.3SG.PAST allai modis in Þizai all:NOM.PL.M.ST mood.GEN.SG.M PREP DEM.DAT.SG.F swnagogein hausjandans Þata synagogue.DAT.SG.F hear.PRES.PART DEM.ACC.SG.N ‘They were all filled with wrath in the synagogue, as they heard these things.’ (from Leiss 2007: 90-91)

(34)

a. Kirjotin kirjeen write.PAST.1.SG. letter.GEN ‘I finished writing the letter.’ (‘I wrote the entire letter.’) b. Kirjotin kirjettä write.PAST.1.SG letter.PART ‘I was writing a/the letter. (from Leiss 2007: 80)

Leiss further claims that ‘the focus of attention in definiteness research should be continually on aspect and related factors’ (2007: 98). Taking this into consideration, the question is how this correlation can be extrapolated to Romance languages. Is there explicit evidence for the connection between aspect and definiteness? On the one side, Romance languages

Catalan Absolute Small Clauses

353

generally maintain an aspectual distinction in non-present tenses. Catalan has different past forms depending on aspect: feia (‘(s)he made’, imperfect) vs. féu (‘(s)he made’, perfect). Additionally, there are many compound tense forms with the auxiliary haver and a past participle (ha fet, havia fet, va haver fet, hagué fet…). All of them are perfect past tenses. Analytic past tense forms in Romance have developed from the Late Latin structure HABERE (full verb) + participle small clause (Macpherson 1967, Smith 1995 and others), and in some Romance varieties they have partially or completely substituted the synthetic ones. The past participle within the Modern Romance verbal paradigm has a new functional load: it carries out passive morphology, and it is now an overt marker for perfectivity rather than for past tense. Interestingly, when these changes happened, the Romance article system emerged and got established. Therefore, I claim that the co-occurrence of these two changes (grammaticalization of the participle and emergence of articles) should not be considered coincidental, accepting the assumption that definiteness and verbal aspect are manifestations of the same phenomenon. Furthermore, it is not uncommon that changes in a specific domain in the language affect or cause reorganization in other related domains. So, for example, changes in the syntactic status of clitic pronouns come along with changes in word order, object agreement, etc. Changes in word order and overt case morphology also seem to be interrelated (see Fischer 2010). When a function becomes opaque as the result of language change, this language might resort to other mechanisms to compensate the increasing ambiguity. Hence, the loss of case endings tends to make difficult the identification of grammatical functions within the clause and therefore the relative position of the arguments could be used as a hint at coding grammatical functions. Stark (2002) shows that the propagation of a systematic (and, thus, grammaticalized) usage of definite and indefinite articles in Old Italian begins between the 13th century and the 15th century. At that time, compound tense verb forms became definitively fixated in Romance languages. Thus, the correlation proposed by Leiss for Germanic languages seems to hold for Romance languages, too. All these facts suggest that it is really the interplay of grammatical features and not only pragmatic constraints triggering the emergence of Romance articles. However, the full picture is even more intricate than just suggested. A categorical distinction between languages like Russian, with a productive aspectual morphology but without articles, and languages like English, lacking overt aspect morphology but with a highly developed article system, is evasive. A summary typological examination is enough to prove the inadequacy of this claim. Romance languages have still several devices

354

Chapter Ten

to express the verbal aspect. Thus, Catalan has a systematic distinction between imperfect and perfect past tenses. Strictly speaking, there was no necessity in Old Catalan to develop an article system. Otherwise, if one considers the absence of the overt aspect in presence tenses as a sufficient motivation for the rise of Romance articles, why did not they already appear in Latin? As for present tenses, definiteness turns out to be decisive to delimit the verbal aspect. It could be the case that the aspectual interpretation of a sentence with a present tense verb is constructed compositionally, that is, the aspectual features of the sentence are derived from the direct object and the context. Provided that the following utterances are used in a neutral context, (35) is usually interpreted as imperfective (habitual or progressive), having an indefinite direct object, while (36) tends to be interpreted as perfective since the direct object is a definite DP. As soon as additional contextual cues are provided, the interpretation of (36) may vary considerably. These examples indicate that there is a link between the direct object and aspect, which can be overcome under certain circumstances. As for past tenses, where the aspectual distinction is still effective in Romance languages, I assume that changes in participles may have had more profound consequences than usually assumed. I illustrate this point with the aid of Old Catalan data in the following sections. (35)

El teu pare talla llenya. the your father cut.PRES.3SG wood ‘Your father cuts / is cutting wood.’

(36)

El teu pare talla la llenya. the your father cut.PRES.3SG the wood ‘Your father cuts (is cutting) the wood.’

To sum up, I have argued that definiteness interacts with perfectivity in a more or less obvious way in Romance languages. I, therefore, conceive the definiteness effect from a syntactic perspective. Morphological verbal aspect, lexical aspect, and definiteness may, indeed, be involved in the global aspectual conceptualization of the eventuality of the utterance. The consideration of some diachronic data will confirm the association of these seemingly unrelated features and shed some light on the motivation of the DE found in ASCs.

Catalan Absolute Small Clauses

355

4. Old Catalan Data To my knowledge, there are no works dealing with the structure of ASCs in Modern Catalan. Studies about their diachronic evolution are lacking too. However, there has been an important language change with relation to absolute clauses. Latin does not show any restriction in the equivalent construction, namely absolute ablative (37). These sentences allow a nonspecific reading of the DP object. The combinatory possibilities in ASCs decrease over time: DPs with an indefinite reading disappear from ASCs quite early, bare nouns and quantified DPs remain only in a few expressions in Modern Catalan, as shown above. (37)

a. compluribus expugnatis oppidis, (...) ‘when several towns had been captured’ (Caesar, De Bello Gallico 3-14, 56 B.C) b. obsidibusque datis, (...) ‘after hostages had been given’ (Caesar, De Bello Gallico, 3-1, 56 B.C.)

What are then the syntactic differences between Latin and Romance languages? What has triggered this language change? Why do all Romance languages change towards a similar direction, so that the definite DP is becoming (almost) obligatory in ASCs? I try to provide answers to these and similar questions in the remainder of this section. Assuming that both aspect and definiteness are only two sides of the same coin, changes on the one side are expected to cause alterations on the other side as well. In order to account for the development of the DE in absolute clauses, we have to take a look at the evolution of the past participle (reanalysed in Old Catalan as passive or perfect aspect marker) and the emergence of the current determiner system.

4.1 The Corpus In order to test the relation between definiteness and the aspectual configuration of a clause, I gathered a sample of sentences from different Old Catalan prose texts from the 13th century to the 16th century7. The aim 7

The main criterion for the selection of the corpus texts was their relative closeness to spontaneous speech. For a period where literacy was practically confined to clergy, it is difficult to find documents that are not loaded with argumentative figures proceeding from a philosophical and religious discourse.

356

Chapter Ten

of my corpus search was to find out the whole range of structures that ASC constructions could adopt in Old Catalan, and identify similarities and differences with the grammaticality judgments presented above for Modern Catalan. Even if some possible options only appear marginally (data may over-represent or under-represent real use), this would already justify the supposition of an ongoing language change. If we are able to isolate the features that first disappeared from those which persisted longer (whether they are commonly used or only as a residue of the initial structure), perhaps we would also be able to determine what was the primary trigger for the language change. Unfortunately, the analysis of definiteness and aspect encounters an additional difficulty: both phenomena are placed at the interface between morphosyntax and semantics-pragmatics. This is, the relevant interpretive cues are placed in the immediate context of the utterance and not in the utterance itself. Furthermore, interface phenomena are usually more unstable and prone to variation. For these reasons, it is expected that phenomena related to aspect and definiteness show during a certain period of time an important amount of variability before they get grammaticalized. Although variability has almost disappeared from ASCs in all Romance languages nowadays, the achieved structure differs from one language to the other in some minor details (e.g., word order, case marking, and the possibility to overtly express the agent). If we could understand more precisely which feature or feature combination gave rise to the definiteness restriction in ASCs, the micro-parametric differences among Romance languages might be attributed to other independent factors. Therefore, I assume that at least two features are relevant and must be taken into account when analysing Old Catalan data: aspectual features of the past participle and DP structure. As for the first one, I claim that past participles lose their temporal content (past tense) and shift towards the status of aspectual and voice markers. I rely on the reanalysis account already outlined in Macpherson (1967), who describes the grammaticalization The selected texts, however, have a dialogical composition, which means that they often try to reflect the spoken language. Within the available medieval repertoire, therefore, these texts can provide some valuable data. Another problem related with the corpus comes from the fact that written documents provide almost exclusively positive evidence. Furthermore, the quantitative results can only be taken with caution: although some structure may have been more or less productive in normal speech, written data do not necessarily supply information about the usage rate of such construction. Likewise, the absence of evidence for a specific construction does not automatically imply its ungrammaticality.

Catalan Absolute Small Clauses

357

process of past participles to compound verb tenses. As for the nominal syntax, it is interesting to notice the syntactic conditions under which each type of DP (specific or non-specific, definite or indefinite) is used. I will address this point with the aid of Diesing’s (1992) Mapping Hypothesis.

4.2 The Data All sentences containing an adverbial small clause headed by a past participle were excerpted from the selected texts.8 For the sake of consistency, adnominal uses of participles —i.e., participles in adjectival function— were excluded from the corpus, even if they behaved like verbal clauses (taking DP complements instead of a PP). All the found sentences were encoded for several morphosyntactic features such as phifeatures of the object, word order, case, etc. Some of the characteristics mentioned in section 2.2 for Modern Catalan are found in Old Catalan as well. Unergative verbs, for example, are not attested in the corpus, but only unaccusative and transitive verbs. Again, this does not mean that ASCs with unergative verbs were ungrammatical, but at least they must have been very rare or marginal already in the middle of the 13th century. However, we find some unexpected data. The ASC in (38), with the verb saber, ‘to know’, would be ruled out in Modern Catalan, since saber is an individual-level predicate and only stage-level predicates can appear in this construction in Modern Catalan. (38)

8

Sabuda per mi la sua dolorosa mort know.PSTPRT.FSG by me the her painful death.FSG ‘As I found out her painful death...’ (Somni 86)

A fragment with similar extension (the first 100 pages approximately) was analysed for each text. The texts, that cover a period of almost 400 years, are the following ones: (a) Llibre de Meravelles, philosophical-religious prose text by Ramon Llull, on the second half of the 13th century. (b) Lo somni, philosophical prose tale by Bernat Metge, on the first half of the 14th century. (c) Curial e Güelfa, anonymous novel of cavalry, of the middle of the 15th century. (d) La fi del comte d'Urgell, anonymous chronical prose text, on the second half of the 15th century. (e) Col·loquis de la insigne ciutat de Tortosa, prose dialogues by Cristòfor Despuig, on the second half of the 16th century. The citation of the examples in the text refers to the title of the source (Meravelles, Somni, Curial, Urgell and Tortosa, respectively) followed by the page number of the used edition (see references at the end of section 7).

358

Chapter Ten

One of the most evident differences is, however, the variable word order in Old Catalan in comparison with Modern Romance languages. So, while the order V - DP (39) is found with a very high frequency from the 13th century onwards, the opposite pattern (40) can still be found at a lower rate until the 16th century. After this moment, we find a clear tendency to place the past participle before the object DP, similar to the modern pattern. (39)

a. dexades moltes memorables coses, (...) leave.PSTPRT.FPL many memorable thing.FPL ‘having left away many memorable things, (...)’ (Urgell 29) b. feta donació del regne make.PSTPRT.FSG donation.FSG of-the kingdom ‘having transferred the kingdom...’ (Tortosa 93)

(40)

i aquella oberta i lo que en ella and that.FSG (=the box) open.PSTPRT.FSG and the that in her venia, vist come.PST.3SG see.PSTPRT ‘and having opened it and seen what was inside...’ (Tortosa 76)

Additionally, these examples show that the DP of the Old Catalan ASC did not have any restriction on the determiner. The DP can be introduced by a definite determiner (38) (or it can be intrinsically marked for definiteness, as the demonstrative pronoun in [40]) or by an indefinite determiner (39a), and it can even be a bare noun (39b). Furthermore, there is a strong preference to let the DP in situ, that is, to the right of the past participle, if it is indefinite. Definite DPs, especially in early texts, show a more flexible placement in the clause; they appear at the right of the verb but can also move over the past participle to the clause-initial position. Negation particles can be inserted in Old Catalan ASCs (41). This means that at least some functional categories are projected over the VP in the T-domain, as I have already argued for Modern Catalan. As for voice, some of the examples show unambiguously that ASCs in Old Catalan are already passive clauses: (38) can only be interpreted as passive because the agent argument appears in an agentive by-phrase adjunct, per mi 9. 9

This kind of agentive adjuncts appears very rarely in the corpus. This should not be interpreted as a marked feature of ASCs but rather a consequence of the

Catalan Absolute Small Clauses

359

However, there are also some intriguing cases. In (42) the argumental DP is represented by an accusative clitic instead of the expected nominative personal pronoun ella. This sentence, however, does not provide solid evidence for Old Catalan ASCs as active constructions for three reasons: (a) it is an isolated case; (b) by the time Curial was written, the Italian influence on Catalan was arguably quite intensive due to the political situation (expansion of the Crown of Aragon over Southern Italy, for instance); and (c) the complex structure of (42), with a causative verb dominating an infinitive clause, could be responsible for particular case requirements on the DP (the object clitic la does not necessarily rise to the higher verb feta since it can get accusative case assigned in the embedded clause, directly from the embedded verb encastar). Such accusative objects are exceptional, even in Old Catalan, and thus it can be assumed that ASCs are passive clauses from the first available documents. (41)

(42)

No constreta la voluntat not contrain.PSTPRT.FSG the free will.FSG ‘without constrains on their free will...’

(Meravelles 47)

a. feta -la encastar en un leó make.PSTPRT.FSG -her.ACC.FSG mount in a lion d’or of gold ‘and then she ordered to mount it in a golden lion’ (Curial 76)

From the discussion so far it has become clear that the clausal structure of ASCs in Old Catalan was at least as rich as the Modern Catalan structure. Evidence for (or against) some projection over TP in the CP-domain, however, could not be attested in the consulted texts so far (i.e. overt material at the left edge: fronted focus, dislocated topics, connectors...). Furthermore, even if one speculates that the verb movement to the initial position could be an instance of V-to-T-to-C movement, there are no obvious arguments to confirm this assumption. Even more, verb movement seems to be optional in Old Catalan. Does this mean that the rise of the verb to the CP can take place overtly or covertly in Old Catalan (but only overtly in Modern Catalan)? Thus, it would be preferable not to postulate higher projections than the TP by the virtue of economy reasons, if not strictly necessary. Verb movement, in fact, can be explained by

argument demotion of the passive voice. Passive is used when the agent does not matter in the utterance and thus can easily be deleted.

360

Chapter Ten

means of formal features in the TP, as I elaborate later on. So, provisionally, the following structure could be adopted for Old Catalan ASCs: [TP[VP[VP V (DP)]. Finally, the relative frequency of bare nouns in Old Catalan ASCs is noteworthy. This observation is in consonance with Bartra-Kaufmann (2007). Since ASCs are a subtype of passive clause, they are subject to the same restrictions on the nominal domain, the only difference being the absence of the auxiliary verb ésser, ‘to be’, in main clauses. The examples in (43) show that bare nouns in ASCs can be found in texts of all periods in the corpus. In all cases, the bare noun follows the past participle. Remember that indefinite DPs are not accepted in Modern Catalan, with the exception of the cases discussed in (23) to (26) and some fossilized expressions like (44). (43)

a. Haüda entre ells grans dicepció have.PSTPRT.FSG among them big disappointment.FSG ‘Since they were extremely disappointing about that...’ (Somni 105) b. Aguda deliberatió e consell have.PSTPRT.FSG deliberation.FSG and advice.MSG ‘having deliberated and discussed...’ (Urgell 45) c. Feta processió e gràcies make.PSTPRT.FSG procession.FSG and thanks.FPL ‘Having hold a procession and an act of thanksgiving...’ (Curial 95) d. feta donació del regne make.PSTPRT.FSG donation.FSG of-the kingdom ‘having transferred the kingdom…’ (Tortosa 93)

(44)

a. Passats alguns anys,... go-by.PSTPRT.MPL several years.MPL ‘after a couple of years...’ (Urgell 84, also possible in Modern Catalan)

Although indefinite DPs or bare nouns are found very frequently in Old Catalan texts, it is still more common in all consulted documents to find ASCs with definite DPs. Such data provide a rough idea of what kind of structures were acceptable in this period, independent of their degree of markedness. Moreover, some properties of Catalan ASCs undergo alterations during these centuries. The definite DPs gain ground until they

Catalan Absolute Small Clauses

361

become obligatory in this position. Indefinite DPs are consequently dismissed. However, while the decline of bare nouns in ASCs runs parallel to the ungrammaticality of bare noun subjects in passive sentences due to structural differences on the DP in Modern Catalan with respect to Old Catalan (Bartra-Kaufmann 2007), there is still no direct explanation for the unacceptability of the indefinite DPs. A closer look at related phenomena can shed light on the phenomenon with which we are dealing. Old Catalan participles are losing some of their properties at that time. Past participles are getting grammaticalized in all Romance languages, as is evident from the weakening or loss of participle agreement in main clauses (Portuguese and Spanish completely lost it; French, Catalan, and Italian show participle agreement with increasing restrictions). Even more, I assume that past participles as well as present participles have shifted their main temporal meaning and become predominantly marked for aspect. (45) illustrates how present participles can introduce an absolute clause triggering participle agreement in number (gender is morphologically invisible in the present participle). This pattern was quite common in the 13th and 14th centuries. (45)

molts hòmens (...), que, ignorants son aveniment, many people (...) REL ignore.PRPRT.PL their future són anats a foch perdurabla. be.3PL go.PSTPRT.MPL to fire eternal ‘many people (...) who, ignoring their destiny, went to the eternal fire.’ (Meravelles 117)

Gerunds, another non-finite verb form found in Old Catalan texts, are often used in verbal periphrases with aspectual meanings. Gerunds and present participles share the same surface form, but the former, in contrast with the latter, are uninflected. From the 14th century onwards the rate of agreed present participles clearly decreases (the data from the 13th century cannot be taken into consideration since the number of tokens is too small). This is shown in table 1. In the 16th century there are no more examples of inflected present participle in the corpus. Apparently, present participles got confused with gerunds in this period, which led to the loss of agreement.

362

Chapter Ten

Total of +Agreement -Agreement Pres. Prt. Meravelles (13th century) 7 2 5 Somni (14th century) 79 51 (64.56%) 28 (35.44%) Curial (15th century) 52 9 (17.31%) 43 (82.69%) Tortosa (16th century) 29 0 (0%) 29 (100%) Table 1. Agreement of present participle absolute clauses in Old Catalan. The newly developed gerundival clause is different from the ASC because (a) participial agreement is out; (b) aspect-tense distinctions can be introduced by the auxiliary verb haver; and (c) it does not show any definiteness effects. In contrast to this, participle agreement in ASCs is maintained and is still compulsory, and such clauses cannot convey temporal and aspectual distinctions. I suggest that the old dichotomy between present and past participles, based on a temporal differentiation, is given up in Old Catalan. Instead, the participles adopt an exclusively aspectual value. Indirectly, this contributes to the convergence of present participles and gerunds. The main difference between gerundival clauses and ASCs is the fixed aspectual value of the past participle. It is appealing to think that aspect is also involved in other characteristic properties of the ASCs, namely participle agreement and definiteness of the DP.

4.3 Language Change in ASCs As I suggested in the preceding section, the definiteness restriction on ASCs must have emerged along with a reorganization of the functional features of participles. Furthermore, I have shown that an aspect phrase over the VP complex must be present in ASCs. Adopting a concept of agreement as the morphological reflex of a Spec-Head relation (e.g. Kayne 1989), overt participle agreement would require DP and participle to be in a Spec-Head configuration at some point of the derivation. This means that there are two possible scenarios: either (i) agreement occurs low in the structure —maybe in a VoiceP within the VP— with the consequent movement of the participle to the aspect phrase, AspP, in the T-domain to check its aspect features, or (ii) AspP is the relevant projection for participle agreement, which would then imply that there must be a CP as a landing site for the subsequent verb movement to the first position of the

Catalan Absolute Small Clauses

363

clause.10 This second step would be independently motivated and could take place overtly or covertly. Travis (1991: 8) claims that ‘there is a link between the derived object position and the +/- completed aspect of the verb’. In this sense, the second option seems to be intuitively more adequate. I, therefore, extend the alleged structure for Catalan ASCs to include a CP over the TP. The only difference, then, with main clauses is that ASCs have a defective T° lacking tense features. Further evidence for the connection of participle agreement with AspP and definiteness is found in the following examples: (46)

a. Con no li havien mostrat en sa juventut then not him had.3PL show.PSTPRT.DEF in his youth tanta de sciència per la qual (...) so-much of science.FSG by which (...) ‘As nobody offered him so much knowledge in his youth, so that (...)’ (Meravelles 139) b. Follia han fet folly.FSG have.3PL do.PSTPRT.DEF ‘They commited a folly.’ (Curial 51) c. Que ell ha vist d‘aquestos hàbits that he have.3SG see.PSTPRT.DEF of this vestment. MPL ‘that he has already seen such vestments.’ (Tortosa 133)

When past participles in Old Catalan developed from being a more or less autonomous element of a small clause to belonging to the verbal paradigm, they gained the possibility of remaining uninflected in certain cases. In other words, participle agreement with the object DP is not obligatory any more. Sentences like (46a), with the default form of the past participle (the same as for masculine singular), were quite rare in the 13th century, but they became more and more frequent at the end of the Middle Ages (46c). Interestingly, as shown in table 2 and table 3, definiteness correlates with the rate of past participle agreement in Old Catalan. So, once participle agreement became optional around the 13th century, the majority of definite DPs still triggered agreement (69–89% of all definite objects in the examined documents triggers participle agreement, hence only 10– 30% of the participles show default agreement). Indefinite DPs, however, oscillate between agreement and default marking, but globally there is a

10 In both scenarios, case assignment could be considered as a consequence of this agreement relation (see Chomsky 2000 and related work).

364

Chapter Ten

slight tendency to let participles uninflected.11 This is even more obvious at the 16th century, where only a few cases of indefinite objects trigger agreement. Furthermore, most of the definite DPs in Col·loquis follow this generalization rule: with post-verbal full DPs or pre-verbal clitics, participles are inflected for number and person; when the object is shifted through wh-movement (questions or relative sentences), participle agreement is disfavoured. This situation strongly resembles Modern Italian (Belletti 2006): agreement is obligatory with 3rd person clitics but excluded with wh-moved objects. Now, these data can be interpreted as follows: only when the object can occupy a specifier position in relation to the participle in the course of the derivation, it results in agreement. Hence, this difference is decisive between main clauses and ASCs: the former, unlike the latter, do not require object movement. Total of [+Def] +Agr -Agr DPs (%) (%) Meravelles (13th century) 142 76.76 23.24 Somni (14th century) 75 77.33 22.67 Curial (15th century) 92 89.13 10.87 Col·loquis (16th century) 56 69.64 30.36 Table 2. Past participle agreement in main clauses with definite DPs in Old Catalan texts. Total of [-Def] +Agr -Agr DPs (%) (%) Meravelles (13th century) 8 62.5 37.5 Somni (14th century) 46 43.48 56.52 Curial (15th century) 25 80 20 Col·loquis (16th century) 23 13.04 86.96 Table 3. Past participle agreement in main clauses with indefinite DPs in Old Catalan texts. To recapitulate, in this section I have shown a series of seemingly unrelated changes emerging in Old Catalan: grammaticalization of the past participle in compound tense verbs with the gradual loss of participle agreement, collapse of present participle and gerund, emergence (or re-organization) of 11

Curial shows a rather archaic language in many features for its period. For this reason, the results must be taken with caution. However, despite the high rates of agreement, the same differences between definite and indefinite DPs could be attested: indefinite DPs show more variability than definite ones.

Catalan Absolute Small Clauses

365

the functional projections of the DP and emergence of definiteness effects in ASCs. All these phenomena involve two common features: definiteness and aspect. Past participles get detached from their temporal meaning (evidence for this is the erosion of the T-domain in ASCs) and are limited to aspectual features that are checked in AspP, distinct from other aspectual levels of the event (IL vs. SL predicates, or inner aspect). At the same time, the new syntactic configuration of the participle will have consequences on object movement. Given that the past participle is marked for the perfect aspect, only definite objects would naturally move to Spec,AspP, thereby triggering participle agreement. Indefinite objects would be more reluctant to be placed in this syntactic position, since they have a feature incompatible with the perfect aspect. In this way, it is expected that optional agreement first appears with indefinite DPs. Object movement becomes obsolete in these cases. Another manifestation of this principle is found in Old Catalan ASCs: definite DPs can be freely placed before or after the verb (object movement is possible and usual), while indefinite DPs are found predominantly after the verb (in situ). This supports the idea that object shift depends on definiteness (e.g. Diesing (1992) and Leiss (2000)). Since the landing site of the moved object in ASCs is at least the specifier of AspP, the postulated interaction between definiteness and aspect gets further confirmation. However, the main question about ASCs in Modern Catalan persists: why are indefinite DPs systematically excluded from ASCs? What can be the motivation for such a strong restriction? There is no apparent reason rendering post-verbal indefinite DPs, perhaps with default agreement, unacceptable.

5. Two Types of Definiteness Effects: Existential Sentences against ASCs In this section I focus on the analysis of Catalan ASCs departing from another point of view to explain the ban on indefinite DPs. In fact, existential and presentational sentences share some relevant properties with ASCs. In both cases only SL predicates are allowed, and both constructions show definiteness restrictions (although in completely different directions). Furthermore, it has been claimed that the DP argument or pivot by existential sentences would constitute a sort of small clause together with its coda (e.g., a locative expression) (47). Such an analysis closely resembles the analysis of ASCs most commonly assumed (48): the participle or deverbal adjective would have the same function as the coda since it contains a predication about the DP (see Leonetti (2008: 140) for further discussion, but also Villalba (2013) for well-founded arguments against this view).

366

(47)

Chapter Ten

a. Hi ha [SC [DP un gos] [CODA al jardí] ]. LOC has a dog in-the garden ‘There is a dog in the garden.’ b. Hi ha [SC[DPla Maria] [CODA esperant-se fa estona]]. LOC has the Mary.FSG waiting-REFL makes while ‘Mary has been waiting for a while.’ (Villalba 2013: 151)

(48)

[SC [V Acabats] [DP els deures] ], podeu finish.PSTPRT.MPL the homework.MPL can.2PL sortir a jugar. go-out to play ‘Once you finish your homework, you can go and play.’

However, there are compelling reasons to take apart one structure from the other. The DE can be suspended by eventive existentials, i.e., existentials with a sentential coda, as in (47b). By ASCs, one cannot find any context where the definiteness restrictions can be annulated. Not only the DE leads to different outcomes in (47) and (48), but there are also differences in word order and the syntactic status of the alleged small clause. While the predicate is clause-initial in ASCs (the verb acabats in (48)), it follows the DP in existential constructions (the locative PP al jardí in (47a) or the sentential coda esperantse fa estona in (47b)). More striking is the difference in syntactic status, as illustrated in (49). Existential constructions arguably occupy an argumental position and therefore they allow for wh-extraction (49a). ASCs, on the contrary, are clausal adjuncts and represent islands for extraction. (49)

a. Quèi dius que hi ha ti al jardí? what say.2SG that LOC has in-the garden ‘What do you say there is in the garden?’ b. * Quèi dius que acabats ti, podeu sortir what say.2SG that finish.PSTPRT.MPL can.2PL go-out a jugar? to play *‘What do you say, when finished, you can go and play?’

Moreover, the pragmatic conditions applying to both existential constructions and ASCs are quite different. Existential sentences are presentational constructions, which means that they must introduce a new referent in the immediate discourse in such a manner that the presence of

Catalan Absolute Small Clauses

367

indefinite determiners is occasionally justifiable from a pragmatic point of view. The secondary uses of there-is constructions with definite DPs (list reading, locative existentials, etc.) are language-specific and usually get different analyses. Hence, all these facts point to a fundamental difference between these structures, and a different account for the DE in ASCs seems to be needed. To this purpose, I adopt here some ideas taken from Diesing (1992) about object movement and definiteness. Her account of definiteness attempts to combine syntactic structure with semantic properties of the clause. According to her, material remaining within the VP is mapped onto the nuclear scope of quantification, i.e., it is interpreted within an existential closure. Material outside the VP is mapped into a restrictive clause, where it gets quantificational force. This splitting rule is formulated under the ‘Mapping Hypothesis’ (MH) and applies at least by the level of Logical Form (Diesing 1992: 10). This has important consequences for the interpretation and classification of indefinite determiners. Virtually, a DP can only receive existential interpretation if it is placed within the VP at LF. DPs that are not to be interpreted within the existential closure (e.g., quantificational indefinite determiners) must be moved out of the VP by LF. The position of DPs in Old Catalan ASCs seems to reflect Diesing’s assumptions: indefinite or bare nouns are rare in pre-verbal position (they remain within the existential closure without any movement), while definite DPs may appear in both positions (they must move, either overtly or covertly). Along these lines, this account links object shift to specificity and presuppositionality. Moved DPs receive a presuppositional reading, i.e., they are connected to the preceding discourse, linguistic context or the extra-linguistic situation, unlike DPs that remain in situ at LF. Applying this general mechanism to past participles, and combining this with language economy considerations, it is possible to model how the different agreement patterns emerge. Non-specific indefinite direct objects in main clauses are not forced to leave the existential closure and therefore they preferably remain in situ. Case checking also does not require movement for the sake of economy of the derivation, and it takes place through long-distance agreement12. The relevant Spec-Head relation for participle agreement is not met overtly, and it declines until it disappears. This explains the variability found in Old Catalan indefinite objects, as a manifestation of an ongoing process of language change. Definite DPs, however, must leave the VP in some moment of the derivation (before or 12

Or the DP gets assigned inherent partitive case, as Belletti (1988) suggests.

368

Chapter Ten

after spell-out) to escape the existential closure. If the ASC has only a reduced structure lacking any functional material (e.g., it is only a VP), the MH encounters a problem: it would predict that either no DE would arise, since the tree splitting is not available, or only indefinite DPs would appear in a sort of existential closure, contrary to fact. Is the MH, then, only valid for main clauses, being small clauses exempt of its restrictions? Or is there an analogous tree splitting taking place within the VP complex so that the VP, for example, would represent the existential closure and the ȞP would contain the restrictive clause? Both solutions have an ad hoc flavour. An extended syntactic structure for ASCs, as proposed in this paper, has the advantage that the Mapping Hypothesis gets a unified application among different clausal structures. Further differences between ASCs and existential constructions have now become evident. The DP in existentials is supposed to be obligatorily mapped into the existential closure and is integrated in a low position of the main clause headed by the existential verb. Even if a small clause analysis were correct, its structure would significantly differ from that of ASCs: it is the existential verb that carries out tense, aspect, and mood features but not the coda. ASCs, having a larger syntactic structure, can show a tree splitting following MH. Different to gerundival clauses, ASCs are marked for the perfect aspect. Moreover, object-verb agreement is necessary in ASCs in the form of participle agreement, probably due to case requirements. Along with the change in aspectual features of the participle, the constraints on the sort of DPs that can be placed in the specifier position of the participle may vary. In Old Catalan the DP is more flexible, while in Modern Catalan it must be compatible with the perfect aspect of the participle. As a consequence of this, indefinite DPs are gradually excluded from this construction. Therefore, it is the perfect aspect that forces a definite determiner to appear in the argument of ASCs in Catalan. In both cases, namely ASCs and existential constructions, some definiteness restrictions have been attested. However, the motivation for these restrictions has a very different source in each case.

6. Conclusions In the present work I have analysed some properties of Catalan absolute small clauses which are often overlooked. Although there is an important amount of literature dealing with the ‘quirky’ properties of ASCs in diverse Romance languages, until now there is no specific study on the particular properties of their object DP. ASCs show, indeed, a sort of definiteness restriction (parallel to the definiteness effect claimed for

Catalan Absolute Small Clauses

369

existential sentences) that deserves a more detailed analysis. I have, therefore, presented some relevant data of Modern Catalan and compared the results with the characteristics already found in other Romance languages. In the course of this discussion it has become evident that a diachronic perspective can enrich our general picture about ASCs. I have shown that, contrary to what is usually assumed, Catalan (and also Spanish) ASCs are structurally not very much reduced. There is, indeed, enough evidence for voice, negation and aspect phrases (supporting the suggestions of Travis (1991), Marín (2002) and Bruno (2011)). I have then put AspP in relation with word order, case, and definiteness restrictions in Catalan ASCs, based on proposals put forward by Krifka (1989), Leiss (2000) and (2007), Fischer (2005) and Stark (2002). The examination of Old Catalan data provides evidence for simultaneous changes in the two domains of verbal aspect (grammaticalization of past participles, collapse of present participle and gerund) and definiteness (rise of the modern determiner system). This fact corroborates the relation between aspect and definiteness. Furthermore, the decline of participle agreement (as experimented in all Romance languages with different outcomes) seems to be also influenced by the definiteness feature of the object DP. Changes in the aspectual features of participles, in conjunction with the requirements of the Mapping Hypothesis as proposed by Diesing (1992), have led to the emergence of a definiteness restriction in ASCs. I have also shown that, in spite of some similarities, the DE in existential sentences and in ASCs are intrinsically different in nature: each of these constructions has a different structural composition and plays a different role in the utterance (adjunct vs. argument). Even more, I have demonstrated that formal features alone are sufficient to account for the observed DE in ASCs so that no semantic-pragmatic requirements need to be postulated. Of course, many of these claims are still subject to revision, waiting for a larger corpus of Old Catalan data. Also, cross-linguistic comparisons can help verify the general ideas of this chapter. In any case, the results presented here, though provisory, suggest that disparate phenomena can have intricate connections. Therefore, further research is still very promising and can enrich our knowledge of how functional features are syntactically manifested, and which mechanisms link verbal and nominal morphology within the syntax.

370

Chapter Ten

References Alcázar, Asier & Mario Saltarelli. 2007. The quirky case of participial clauses. In Sergio Baauw, Frank Drijkoningen & Manuela Pinto (eds.). Romance Languages and Linguistic Theory, 1-18. Utrecht: Benjamins. Alexiadou, Artemis & Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order, V-movement, and EPP-checking. In Natural Language and Linguistic Theory 16: 491-540. Arche, María J. 2012. On the aspectuality of the individual-level / stagelevel dichotomy. In Borealis: An International Journal of Hispanic Linguistics 1.2: 109-132. Bartra-Kaufmann, Anna. 2007. Determinerless noun phrases in Old Romance passives. In Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds.). Nominal Determination. Typology, context constraints and historical emergence, 257-286. Amsterdam: Benjamins. Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry 19: 1-34. —. 1992. Agreement and Case in Past Participle Clauses. In Tim Stowell & Eric Wehlri (eds.). Syntax and Semantics. Vol. 26, 21-44. New York: Academic Press. —. 2006. (Past) Participle Agreement. In Martin Everaert & Henk van Riemsdijk (eds.). The Blackwell Companion to Syntax, 493-521. Malden, MA: Blackwell. Bruno, James V. 2011. Absolute Constructions: Telicity, Abstract Case, and Micro-variation. In Luis A. Ortiz-López (ed.). Selected Proceedings of the 13th Hispanic Linguistics Symposium. Somerville, MA: Cascadilla. Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel. Carlson, Gregory N. 1977. Reference to Kinds in English. PhDDissertation. Amherst: University of Massachusetts. Caesar, Gaius Iulius. 1986 [56 B.C.]. Commentarii de bello Gallico. Munich: Artemis. Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Roger Martin, David Michaels & Juan Uriagereka (eds.). Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, 89-156. Cambridge, MA: MIT Press. —. 2005. On Phases. In Robert Freidin, Carlos P. Otero & María Luisa Zubizarreta (eds.). Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud, 133-166. Cambridge, MA: MIT Press.

Catalan Absolute Small Clauses

371

De Miguel, Elena. 1992. El aspecto en la sintaxis del español: Perfectividad e impersonalidad. Madrid: Ediciones de la UAM. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Enç, Mürvet. 1991. The semantics of specificity. In Linguistic Inquiry 22: 1-25. Fischer, Susann. 2005. The interplay between Aspect and Reference. In Klaus von Heusinger, Georg Kaiser & Elisabeth Stark (eds.). Proceedings of the Workshop ‘Specificity and the Evolution/ Emergence of Nominal Determinations Systems in Romance’, 1-18. Konstanz: Arbeitspapier 119. Fachbereich Sprachwissenschaft, Universität Konstanz. —. 2010. Word-order change as a source of grammaticalisation. Amsterdam: Benjamins. —. 2013. Unaccusatives vs. Existentials: Explorations of the SyntaxSemantics Interface. In Sofiana Chiriacescu (ed.). Theoretical implications at the syntax-semantics interface in Romance: Proceedings of the VI Nereus International Workshop, 1-21. Konstanz: Arbeitspapier 127. Fachbereich Sprachwissenschaft, Universität Konstanz. Gunnarson, Kjell-Åke. 1994. Small clauses and absolute constructions in Spanish. In Probus 7: 125-171. Hernanz, María Luisa. 1991. Spanish absolute constructions and aspect. In Catalan Working Papers in Linguistics 1: 75-128. Kayne, Richard. 1989. Facets of Romance Past Participle Agreement. In Paola Benincà (ed.). Dialect Variation and the Theory of Grammar, 85-103. Dordrecht: Foris. Kratzer, Angelika. 1995. Stage-level and Individual-level Predicates. In Gregory N. Carlson & Francis J. Pelletier (eds.). The Generic Book, 125-175. Chicago: Chicago University Press. Krifka, Manfred. 1989. Nominal Reference, Temporal Constitution and Quantification in Event Semantics. In Renate Bartsch, Johan van Benthem & Peter van Emde Boas (eds.). Semantics and Contextual Expressions, 75-115. Dordrecht: Foris. Lefebvre, Claire. 1988. Past Participle Agreement in French: Agreement = Case. In David Birdsong & Jean-Pierre Montreuil (eds.). Advances in Romance Linguistics, 233-253. Dordrecht: Foris. Leiss, Elisabeth. 2000. Artikel und Aspekt. Die grammatischen Muster von Definitheit. Berlin/New York: De Gruyter. —. 2007. Covert patterns of definiteness/indefiniteness and aspectuality in Old Icelandic, Gothic and Old High German. In Elisabeth Stark, Elisabeth Leiss & Werner Abraham (eds.). Nominal Determination:

372

Chapter Ten

Typology, context constraints, and historical emergence, 73-102. Amsterdam: Benjamins. Leonetti, Manuel. 2008. Definiteness effect and the role of the coda in existential constructions. In Henrik Høeg Müller & Alex Klinge (eds.). Essays on Nominal Determination: From morphology to discourse management, 131-162. Amsterdam: Benjamins. López, Luis. 1994. The Internal Structure of Absolute Small Clauses. In Catalan Working Papers in Linguistics 4.1: 45-92. Lyons, Christopher. 1999. Indefiniteness. Cambrigde: Cambridge University Press. Macpherson, Ian. 1967. Past Participle Agreement in Old Spanish: Transitive Verbs. In Bulletin of Hispanic Studies 44.4: 241-254. Marín, Rafael. 2002. De nuevo sobre construcciones absolutas. In Círculo de Lingüística Aplicada a la Comunicación. [available at http://www.ucm.es/info/circulo/no10/marin.htm.] Milsark, Gary L. 1977. Toward an Explanation of Certain Peculiarities of the Existential Construction in English. In Linguistic Analysis 3.1: 129. Parodi, Claudia. 1995. Participle Agreement and Object Shift in Old Spanish: A Minimalist Theory Approach. In Héctor Campos & Paula Kempchinsky (eds.). Evolution and Revolution in Linguistic theory, 276-301. Washington, DC: Georgetown University Press. Smith, John Charles. 1995. Agreement between past participle and direct object in Catalan: the hypothesis of Castilian influence revisited. In Jacek Fisiak (ed.). Linguistic Change under Contact Conditions, 271289. Berlin: Mouton de Gruyter. Stark, Elisabeth. 2002. Indefiniteness and Specificity in Old Italian Texts. In Journal of Semantics 19.3: 315-332. Taraldsen, Tarald. 1987. Clitic/Participle Agreement and Auxiliary Alternation in Romance. In Carol Neidle & Rafael Gómez Cedeño (eds.). Studies in Romance Languages, 263-281. Washington, DC: Georgetown University Press. Travis, Lisa deMena. 1991. Derived Objects, Inner Aspect, and the Structure of VP. Paper presented at the 22nd Annual Meeting of the North East Linguistics Society (NELS 22), University of Delaware. Villalba, Xavier. 2013. Eventive existentials in Catalan and the topic-focus articulation. In Italian Journal of Linguistics/Rivista di Linguistica 25.1: 147-173.

Catalan Absolute Small Clauses

373

Consulted Literature for the Old Catalan Corpus a. Anonymous. 2007 [ca. 1450?]. Curial e Güelfa. Toulouse: Anacharsis. b. Anonymous. 1931 [1462-72]. La fi del comte d'Urgell. Barcelona: Barcino. c. Despuig, Cristòfor. 1996 [1557]. Los col·loquis de la insigne ciutat de Tortosa, Barcelona: Curial Edicions Catalanes. d. Llull, Ramon. 1931 [1287-89]. Llibre de Meravelles. Vol. I. Barcelona: Barcino. e. Metge, Bernat. 1925 [1399] Lo somni. Barcelona: Barcino.

CHAPTER ELEVEN DEFINITENESS EFFECTS IN THE HISTORY OF FRENCH: THE CASE OF PRESENTATIONALS1 MICHAEL ZIMMERMANN

1. Introduction Impersonal constructions (ICs) in Modern Standard French, the presentday formal variety dating back to the 17th century, stand out due to the properties (i) that the finite verb is invariably in the third person singular (cf. Ex. 1a and Ex. 2a) and (ii) that the overt expletive, i.e. semantically vacuous, subject pronoun il ‘it’ is generally mandatory (cf. Ex. 1b and Ex. 2b). (1)

a. b.

1

Il arrive(*nt) it.NOM.3SG arrive.3SG/3PL *(Il) arrive it.NOM.3SG arrive.3SG ‘There arrive three girls.’

trois three trois three

filles.2 girl.FEM.3PL filles. girl.FEM.3PL

For helpful comments on a draft version of this paper, I would like to thank two anonymous reviewers, the editors of this volume, Georg A. Kaiser as well as the audience of the workshop ‘Definiteness Effects’ at the 34th Annual Conference of the German Linguistic Society (DGfS). 2 The following abbreviations are used in the glossings of the examples in this paper: NOM, OBL for nominative and oblique case; 1, 2, 3 for first, second, and third person; SG, PL for singular and plural; MAS, FEM for masculine and feminine; INF for infinitive; PAST for past tense; REFL for reflexive.

Definiteness Effects in the History of French

(2)

a.

b.

375

Il a/ *ont été volé à la it.NOM.3SG have.3SG have.3PL been stolen at the.FEM.3SG bibliothèque trois livres. library.FEM.3SG three book.MAS.3PL *(Il) a été volé à la it .NOM.3SG have.3SG been stolen at the.FEM.3SG bibliothèque trois livres. library.FEM.3SG three book.MAS.3PL ‘Three books were stolen at the library.’

ICs with either an inherently intransitive verb such as in Ex. 1 or with a transitive verb used intransitively (Martinet 1975, Gaatone 1976, 1991, 1994) as in Ex. 2 are commonly referred to as presentationals. Regarding these constructions in Modern Standard French, it is generally argued that the ‘Definiteness Effect’ (DE) holds, i.e. the requirement that the postverbal DP be not definite (Belletti 1988, Alexiadou & Anagnostopoulou 1998, Sheehan 2006). This is illustrated in Ex. 3 and Ex. 4 by the ungrammaticality of presentational ICs, in which the postverbal DP features a definite/demonstrative/possessive determiner (Moignet 1971, Seelbach 1971, 1978, KĊsik 1975, 1991, Rivière 1979, 1981) and, respectively, constitutes a proper name or a non-clitic subject personal/ demonstrative/possessive pronoun (Zribi-Hertz 1982, Belletti 1988, Le Goffic 1993). (3)

*Il arrive les / ces / ses it.NOM.3SG arrive.3SG the.3PL this.3PL his/her.3PL filles. girl.FEM.3PL

(4)

*Il it.NOM.3SG elles / she.FEM.3PL

arrive Jeanne (et Marie) / arrive.3SG Jeanne.FEM.3SG and Marie.FEM.3SG celles-là / les siennes. this.FEM.3PL the.3PL his/her.FEM.3PL

Still, it is by and far acknowledged in the literature that the postverbal DP in presentational ICs in Modern Standard French can in exceptional cases be definite (Gaatone 1970, Vet 1981, Lazard 1994, Eguzkitza & Kaiser 1999), viz. when one or several of the following highly restricted set of conditions are met:

Chapter Eleven

376

-

(5)

-

(6)

use of restrictive ne ... que ‘only’ (Gross 1968, KĊsik 1975, 1986, Stéfanini 1975, Picabia 1986): Il n’ arrive que les filles. it.NOM.3SG not arrive.3SG that the.3PL girl.FEM.3PL ‘Only the girls arrive.’ use of the verbs rester ‘to remain’ and manquer ‘to lack’ (Martin 1970, Seelbach 1971, 1978, KĊsik 1975, 1984, 1991, Stéfanini 1975): a.

b.

-

(7)

-

(8)

Il reste les filles. it.NOM.3SG remain.3SG the.3PL girl.FEM.3PL ‘There remain the girls.’ Il manque les filles. it.NOM.3SG be.missing.3SG the.3PL girl.FEM.3PL ‘The girls are missing.’

‘cataphoricity’ of the postverbal DP, i.e. coreferentiality with (part of) the directly ensuing co(n)text (Lerch 1931, Eskénazi 1968, KĊsik 1984, 1986, 1991, 1998, Picabia 1986): Il a été volé à la bibliothèque it.NOM.3SG have.3SG been stolen at the.FEM.3SG library.FEM.3SG les livres suivants : Sapir, the.3PL book.MAS.3PL following.MAS.PL Sapir.MAS.3SG Benveniste et Guillaume. (Picabia 1986: 235) Benveniste.MAS.3SG and Guillaume.MAS.3SG ‘The following books were stolen at the library: Sapir, Benveniste, and Guillaume.’ list item status of the postverbal DP in the context of an enumeration (Seelbach 1978, Rivière 1981, Picabia 1986, Le Goffic 1993): Il a dormi ici Jean, Paul it.NOM.3SG have.3SG slept here Jean.MAS.3SG Paul.MAS.3SG et François. (Picabia 1986: 235) and François.MAS.3SG ‘Jean, Paul, and François slept here.’

Definiteness Effects in the History of French

-

‘post-determination’ (Vet 1981: 51) of the postverbal DP in the very sentence in which it occurs by ‘restrictive complements’ (KĊsik 1991: 190) such as “a restrictive relative, a superlative, or some PPs” (Etchegoyhen & Tsoulas 1998: 81) (cf. also Gaatone 1970, Seelbach 1971, 1978, KĊsik 1975, 1984, Rivière 1981):

(9)

a.

b.

c.

-

Il arrive les filles dont it.NOM.3SG arrive.3SG the.3PL girl.FEM.3PL of.who.OBL je t’ ai parlé. I.NOM.1SG you.OBL.2SG have.1SG spoken ‘The girls I told you about arrive.’ Il arrive les filles les plus it.NOM.3SG arrive.3SG the.3PL girl.FEM.3PL the.3PL most belles du monde. beautiful.FEM.PL of.the.MAS.3SG earth.MAS.3SG ‘The most beautiful girls in the world arrive.’ Il arrive les filles d’ à it.NOM.3SG arrive.3SG the.3PL girl.FEM.3PL of to côté. side.FEM.3SG ‘The girls from next door arrive.’

‘non-specificity’ of the postverbal DP (Gaatone 1970, KĊsik 1984, 1991):

(10)

-

377

Il lui venait la it.NOM.3SG he/she.OBL.3SG be.coming.PAST.3SG the.FEM.3SG sueur au front. sweat.FEM.3SG to.the.MAS.3SG forehead.MAS.3SG (Les Hommes de bonne volonté, quoted from Gaatone 1970: 399) ‘It brought the sweat to his/her brow.’

participation of the postverbal DP in a fixed idiomatic expression which is passivized:

(11)

Il a été cassé la croûte it.NOM.3SG have.3SG been broken the.FEM.3SG crust .FEM.3SG récemment sur cette table. recently on this.FEM.3SG table.FEM.3SG (Zribi-Hertz 1982: 146) ‘One had recently something to eat at this table.’

Chapter Eleven

378

Since these exceptional cases constitute an altogether small and welldefined set, the DE is usually argued to generally hold in presentational ICs in Modern Standard French. Given the general analysis of this variety as a non-null subject language,3 such a view is essentially in line with the common claim of a correlation between the DE and the non-null subject property: The one difference between the null expletive and overt expletives in nonnull-subject languages is that the null expletive is not obligatorily associated with ‘definiteness effect’; that is, the DP associate of the expletive (the DP that, although not in the canonical syntactic position, functions as the subject argument) must have a weak determiner, roughly speaking an indefinite. (Roberts 2007: 15)

Being thus a non-null subject language in which the DE (generally) holds, Modern Standard French evidently substantiates this common claim. Unlike its modern formal variety, the medieval, i.e. pre-17th century, stages of French are usually analyzed as null subject languages (Adams 1987, Roberts 1993, Vance 1997), since subject pronouns are not consistently overt, as illustrated in Ex. 12 for Old French (9th-13th century) and, respectively, Middle French (14th-16th century). (12)

3

Old French a. Ne pot intrer en la ciutat ; not can.PAST.3SG enter.INF in the.FEM.3SG city.FEM.3SG (saint-léger, p.166) ‘He could not enter the city;’

This general analysis has explicitly been challenged by Sportiche (1999) and Roberts (2010) who crucially claim that postverbal clitic subject pronouns constitute inflectional affixes principally instantiating uninterpretable ij-features of the T and, respectively, C level. Space considerations prevent me from going into a discussion of this claim; I shall therefore merely point out that the empirical evidence put forward for such a claim, viz. distributional restrictions of postverbal clitic subject pronouns as compared to their preverbal counterparts, is far from uncontroversial.

Definiteness Effects in the History of French

379

Middle French b. Et pour celer son meurdre, and for conceal.INF his.MAS.3SG murder.MAS.3SG feit bruler le cors make.3SG burn.INF the.MAS.3SG body.MAS.3SG du pauure trépassé, (…) (heptaméron, p.22) of.the.MAS.3SG poor.SG deceased.MAS.3SG ‘And to conceal his murder, he has the body of the poor deceased burnt, (…)’ Yet, null subjects are not freely allowed in these stages of French. In effect, it is generally argued that subject pronouns can only be null when specific structural conditions are met (Foulet 1928, Franzén 1939, Skårup 1975). Unless this is the case, subject pronouns are necessarily overt, whence the comparatively frequent occurrence of overt subject pronouns in Old and Middle French. In view of this intriguing trait, the question arises as to whether or not the DE holds in presentational ICs in the medieval stages of French. In this regard, Section 2 will discuss previous investigations, showing that each of these rests on an altogether scarce and ambivalent empirical basis, a state of affairs which has led to opposing views and, thus, does not allow for a definitive answer to said question. Based on the results from analyzing a large diachronic data corpus, Section 3 will investigate the issue of the DE on a broader empirical basis and will essentially argue that, like in the modern standard stage, the DE generally holds in presentational ICs in the medieval stages of French, a deduction which is in fact expected under a principled alternative analysis of these stages as non-null subject languages.

2. Previous Investigations into the DE in Medieval French In contrast to Modern Standard French, the issue of whether or not the DE holds in presentational ICs in Old and Middle French has so far been tackled by only a few researchers—who have in fact put forward opposing views: while according to de Bakker (1995, 1997), the DE does not hold in Old French, Fischer (2009) maintains that it does hold. Given these opposing views, the empirical evidence on which said investigations are based as well as its analysis will presently be discussed. De Bakker (1997: 156) argues that “[t]he Definiteness Effect that applies to the Modern French il-construction does not seem to exist in Old French […] [since] [b]oth definite and indefinite NPs appear as logical

380

Chapter Eleven

subject in the il-construction, although the indefinite ones seem to be in the majority.” This “variation in definiteness” (de Bakker 1997: 197) is claimed to persist until the very end of the Middle French period and inferred from the set of constructions in Ex. 13. (13)

Old French a. Il i vont ci IL.NOM.3 there go.3PL this.NOM.MAS.3PL viel prestre et old.NOM.MAS.3PL priest.NOM.MAS.3PL and cil viel this.NOM.MAS.3PL old.NOM.MAS.3PL clop et cil lame.person.NOM.MAS.3PL and this.NOM.MAS.3PL manke (…) one.armed.person.NOM.MAS.3PL (Aucassin et Nicolette, quoted from de Bakker 1997: 157) ‘Those that go there are the old priests and the old lame as well as the one-armed persons (…)’ b. Il i fu li IL.NOM.3 there be.PAST.3SG the.NOM.MAS.3SG qens de Pontiu et ses count.NOM.MAS.3SG of Pontieu and his.NOM.MAS.3SG oncles li quens uncle.NOM.MAS.3SG the.NOM.MAS.3SG count.NOM.MAS.3SG de Saint Pol, (…) of Saint.MAS.3SG Pol.MAS.3SG (La fille du compte de Pontieu, quoted from de Bakker 1997: 157) ‘The count of Pontieu and his uncle, the count of Saint Pol, were there (…)’ c. Il sortirent les vieux. IL.NOM.3 go.out.PAST.3PL the.3PL old.people.MAS.3PL (de Bakker 1997: 172,fn.51) ‘The old people went out.’

As reflected by the small number of examples, the empirical evidence in favor of the view that the DE does not hold in Medieval French is relatively scarce. Moreover, it appears that these examples are not ideally suited to decide on the issue of the DE in at least Old French.

Definiteness Effects in the History of French

381

As far as Ex. 13a-b are concerned, de Bakker (1997: 158) herself points out that the relevant constructions “cannot be interpreted as unambiguous proof of definite NPs in the il-construction”. In effect, these constructions might just as well be considered personals involving right dislocation of the (masculine) postverbal DPs, and this all the more so in light of the following morphological facts: (i) in Old French, the form of the overt masculine third person singular as well as plural subject pronouns is, just like that of the default overt expletive subject pronoun, il; (ii) in Ex. 13a-b which date from a period when the two-case system morphologically distinguishing the masculine nominative from the masculine oblique is still by and far observed, the postverbal DPs are nominative; (iii) at least in Ex. 13a, plural verbal agreement obtains (cf. also Dufresne & Dupuis 2007). Note with regard to the latter fact that the singular verbal agreement in Ex. 13b does not undermine such an alternative view, since “this is the regular subject-verb agreement pattern in Old French for coordinated subjects” (Arteaga 1994: 145). Essentially, even if the pertinent constructions in Ex. 13a-b above prove to be impersonal, they are not ideally suited to decide on the issue of the DE in at least Old French. For both of these meet one of the highly restricted set of conditions under which definite postverbal DPs are by way of exception allowed in presentational ICs in Modern Standard French, a language otherwise strictly showing the DE (cf. Section 1): the post-determination of the postverbal DP by a restrictive complement.4 As follows from Ex. 13b and the more comprehensive Ex. 14 incorporating Ex. 13a, the definite postverbal DPs are post-determined separately by a PP and, respectively, jointly by a restrictive relative.

4

As pointed out by an anonymous reviewer, the definite postverbal DPs in Ex. 13b also meet the condition of list item status, as the pertinent construction might alternatively be considered a presentational IC featuring an enumeration.

Chapter Eleven

382

(14)

Il

i vont ci viel there go.3PL this.NOM.MAS.3PL old.NOM.MAS.3PL prestre et cil viel priest.NOM.MAS.3PL and this.NOM.MAS.3PL old.NOM.MAS.3PL clop et cil lame.person.NOM.MAS.3PL and this.NOM.MAS.3PL manke qui tote jor et one.armed.person.NOM.MAS.3PL who all.SG day.OBL.MAS.3SG and tote nuit cropent devant ces all.SG night.FEM.3SG crouch.3PL before this.OBL.MAS.3PL autex et en ces viés creutes, (…) altar.OBL.MAS.3PL and in this.3PL old.FEM.PL crypt.FEM.3PL (Aucassin et Nicolette, quoted from de Bakker 1997: 153) ‘Those that go there are the old priests and the old lame as well as the one-armed persons who all crouch night and day before the altars and in the old crypts, (…)’ IL.NOM.3

Regarding Ex. 13c above which seems to represent an unequivocal piece of evidence in favor of the view that the DE does not hold in at least Old French, de Bakker (1997: 173,fn.51) draws attention to the fact that she has “not been able to verify this example” and therefore quotes it from Gamillscheg (1957). As it stands, Gamillscheg (1957: 521) does not quote said example, but rather employs it as a label—“the type il sortirent les vieux”—to refer to a specific set of constructions discussed in Horning (1880). An examination of all of Horning’s examples reveals that neither Ex. 13c, nor any other pertinent example with a definite postverbal DP from the oldest stage of French is found in this work.5 It thus appears that

5

The discussion in Horning (1880) does, however, feature one relevant example from Middle French, given in (i). (i)

Il luy vint en mémoire les it.NOM.3SG he.OBL.MAS.3SG come.PAST.3SG in mind.FEM.3SG the.3PL paroles que le héraut d’ word.FEM.3PL which.OBL the.MAS.3SG herald.MAS.3SG of Angleterre luy avoit dites ; England.FEM.3SG he.OBL.MAS.3SG be.having.PAST.3SG said (comines, p.87) ‘The words which the herald from England had told him came to his mind;’

Definiteness Effects in the History of French

383

Ex. 13c does not represent a genuine Old French presentational IC with a definite postverbal DP. In contrast to de Bakker, Fischer (2009: 50) argues that “Old French […] display[s] D[efiniteness]R[estrictions] in unaccusative constructions”, i.e. (at least) in a subset of the constructions discussed in this paper. Under this view, Old French would not differ from Modern Standard French. Given Fischer’s (2009: 53) alternative analysis of Old French as a non-null subject language, this is in fact what one would expect in view of the common claim of a correlation between the DE and the non-null subject property (cf. Section 1). The set of constructions from which Fischer deduces that the DE holds in Old and, by extension, Middle French is given in Ex. 15 along with her glossings and translations. (15)

Old French a. quant il durent ariver une turmente (…) when it must arrive a storm (Lai d’Eliduc, quoted from Fischer 2009: 53) ‘when a storm will arrive (…)’ b. en toteneis est arrivez plusurs reis (…) in now is arrive some kings (Lai d’Eliduc, quoted from Fischer 2009: 53) ‘and now some kings have arrived (…)’ c. en walcres arriva rois antiaumes (…) in w. arrive kings antiaumes (Chronique rimée de Philippe Mouskes, quoted from Fischer 2009: 53) ‘in Walcres arrive the antiaume Kings (…)’

As exposed by the small number of examples, the empirical evidence in favor of the view that the DE holds in Medieval French is quite scarce. Besides, it seems that these examples are not ideally suited to decide on the issue of the DE in Medieval French, since presentational ICs with an indefinite postverbal DP such as those in Ex. 15 a-b are typical not only

Like the relevant constructions in Ex. 13a-b, however, that in (i) is not ideally suited to decide on the issue of the DE in this stage of French, since the definite postverbal DP is post-determined by a restrictive relative and hereby meets one of the highly restricted set of conditions under which definite postverbal DPs are possible in presentational ICs in Modern Standard French.

384

Chapter Eleven

of languages in which the DE holds, but also of those in which it does not hold. Moreover, as reflected by Fischer’s translation of Ex. 15c, the relevant construction seems to represent evidence against her view that the DE holds in at least Old French. Yet, as shown in Ex. 16 by the directly preceding as well as ensuing context of the construction in Ex. 15c, this construction proves to be personal, rather than impersonal and, thus, does not represent evidence against Fischer’s view. (16)

Tantost la tormente ciésa, en as.soon.as the.FEM.3SG storm.FEM.3SG cease.PAST.3SG in Walcres ariva. Rois Walcre arrive.PAST.3SG King.NOM.MAS.3SG Antiaumes quant il le Antiaume.NOM.MAS.3SG when he.NOM.MAS.3SG it.OBL.3SG sot, D’ Engletière, (…), Li know.PAST.3SG from England.FEM.3SG he.OBL.MAS.3SG envoia (…) (chronique rimée, p.49) send.PAST.3SG ‘As soon as the storm ceased, he arrived at the isle of Walcre. When he heard of it, King Antiaume sent him (…) from England (…)’

To conclude, the few previous investigations into the issue of the DE in presentational ICs in Old French have put forward opposing views, and this evidently for reasons of both the scarcity of their empirical basis and its ambivalent nature. In the following, said issue will be investigated on a broader empirical basis by means of the analysis of a large diachronic data corpus from which it is essentially inferred that, in line with Fischer, the DE generally holds in presentational ICs in the medieval stages of French.

3. A New Investigation into the DE in Medieval French In order to more promisingly tackle the issue of whether or not the DE holds in presentational ICs in Old and Middle French, substantial empirical evidence is needed. Given, however, as noted by de Bakker (1997: 192), “the rarity of definite NPs, even in Old French” in presentational ICs (cf. also Dufresne & Dupuis 2007), such evidence is not readily at hand. Evidently, the only means to gather a broad(er) empirical basis is the analysis of a large diachronic data corpus. In this regard, the corpus underlying Zimmermann (2014) has been chosen which covers the

Definiteness Effects in the History of French

385

medieval stages of French on the basis of (extracts from)6 eleven texts selected in accordance with an elaborate list of criteria. As follows from the analysis of this data corpus with regard to the overall frequency of ICs in finite declaratives, there are in total 735 instances of such constructions. Of these, only 55 instances or 7.5% constitute presentational ICs.7 Tab. 1 gives the numbers of the established presentational ICs with a(n) (in)definite postverbal DP, and this separately for constructions with overt and, respectively, null expletive subject pronouns.8 definite DP indefinite DP text (dating) Ø il / ce Ȉ Ø il / ce Ȉ roland (1125-50) 3 0 0 0 3 0 livre reis (1170) 9 0 1 0 9 1 saint graal (1220) 1 1 0 1 2 1 conquête (1290) 1 0 0 0 1 0 saint louis (1330-40) 1 2 1 2 3 3 livre des fais (1409) 3 0 4 2 3 6 galien (1450) 0 0 0 1 0 1 abregé (1450-75) 3 0 5 1 3 6 minute (1498-1515) 1 0 0 1 1 1 heptaméron (1553) 0 0 2 3 0 5 registre-journal (1580-1606) 0 0 1 5 0 6 Table 1. Absolute numbers of presentational ICs with an (in)definite postverbal DP. 6

An extract of 1,500 clauses was chosen from each text to allow for the capture of both a large number and a great(er) variety of primarily impersonal constructions. 7 The bulk of relevant constructions are genuine or intrinsic impersonals, above all existentials with y avoir ‘to be’, while constructions such as those under investigation in this paper, generally considered impersonal variants of the corresponding personal constructions, are comparatively infrequent. As to why this is so—an issue raised by an anonymous reviewer—, one may speculate that in the medieval stages of French, just like in its modern standard stage, personal constructions are preferred to impersonals, much like actives to passives. 8 Note that in addition to the occurrence of overt il, that of overt ce is encountered in ICs in the medieval stages of French. Unlike the former, the latter is highly restricted, in that it is only found with two specific kinds of IC, viz. (i) that featuring the verb venir ‘to come’ followed by either an infinitive or a DP and (ii) that featuring the verb être ‘to be’ directly preceded by the adverbial pronoun en ‘of it’ and introduced by the concessive expression quoi que ‘whatever’ (cf. Zimmermann (2014) for discussion). Of these two specific kinds of IC, (a subset of) the former must presently also be taken into consideration.

386

Chapter Eleven

As ensues from these results, throughout Old and Middle French, instances of presentational ICs with an indefinite postverbal DP such as in Ex. 17 are encountered along those with a definite postverbal DP, illustrated in Ex. 18.9 (17)

9

Old French a. ‘ Sire, rien ne me falt.’ (livre reis, p.114) Sire nothing not I.OBL.1SG lack.3SG ‘‘Sire, I do not need anything.’’ b. « Seignor, il m’ avint Sir, it.NOM.3SG I.OBL.1SG happen.PAST.3SG anuit en mon dormant in.the.night.FEM.3SG in my.MAS.3SG sleeping.MAS.3SG une avision si merveilleuse que (…) a.FEM.3SG vision.FEM.3SG so strange.FEM.SG that (saint graal, p.304) “Sir, while asleep last night, I had a very strange vision, namely that (…)’’

Note with regard to Ex. 17a as well as Ex. 18a-b which feature presentational ICs with experiencer verbs in Old French that it has recently been claimed that in such constructions, the subject position is occupied by an oblique pronoun, taken to constitute a quirky subject of the Modern Icelandic type (Mathieu 2006, Fischer 2010). Given, however, that the occurrence of the (nominative) expletive subject pronoun il ‘it’ has always been a feature of the French language (Büchtemann 1912, Zimmermann 2009, 2014) in whose medieval stages subject pronouns can be null when specific structural conditions are met (Adams 1987, Roberts 1993, Vance 1997, Zimmermann 2014) (cf. also footnote 16), one might alternatively argue that the occurrence of elements in the subject position of ICs in Old (as well as Middle) French is restricted to expletives. In line with traditional analyses, the occurrence of (non-clitic) oblique pronouns in first position in ICs with experiencer verbs might under this alternative view be accounted for in terms of the Tobler-Mussafia law (Mussafia 1886, Tobler 1912) which incorporates the insight that in general, (clitic) oblique pronouns are banned from occurring preverbally when the finite verb is in first position.

Definiteness Effects in the History of French

387

Middle French c. Ou flum devant le in.the.MAS.3SG river.MAS.3SG before the.MAS.3SG roy avoit une galie de king.MAS.3SG be.having.3SG a.FEM.3SG galley.FEM.3SG of Genevois la ou il ne Genevan.MAS.3PL there where it.NOM.3SG not parois que un seul be.appearing.3SG that a.MAS.3SG sole.MAS.SG home desur. (saint louis, p.184) man.MAS.3SG on.top ‘On the river, before the king, there was a galley of the Genevans on which there was but one man on the bridge.’ d. (…), et a un chacun autre (…) and to a.MAS.3SG everyone.MAS.3SG other.SG qui demanderoit la jouste, who.NOM would.ask.for.3SG the.FEM.3SG duel.FEM.3SG seroit delivré .V. coups de would.be .3SG given five blow.MAS.3PL of rochet. (livre des faits, p.67) peak.of.the.lance.MAS.3SG ‘(…), and every single one (…) who would ask for a duel, would receive five blows with the peak of the lance.’ e. (…) qu’ il arriva ung that it.NOM.3SG arrive.PAST.3SG a.MAS.3SG messaige moult fort lassé. message.MAS.3SG very strong.MAS.SG unhappy.MAS.SG (galien, p.124) ‘(…) that an extremely bad massage arrived.’ f. (…) des choses dont ci dessus of.the.3PL thing.FEM.3PL of.which.OBL there below est faicte brieve mencion10 (…) (abregé, p.169) be.3SG made short.FEM.SG mention.FEM.3SG ‘(…) of the things which are briefly mentioned right below (…)’

10

Note that just like the definite article (cf. footnote 11), the indefinite article is not necessarily present in the medieval stages of French, again a phenomenon commonly encountered in these stages.

Chapter Eleven

388

g.

h.

i.

(18)

(…) celle Katherine luy this.FEM.3SG Katherine.FEM.3SG she.OBL.FEM.3SG dist qu’ il venoit be.saying.PAST.3SG that it.NOM.3SG be.coming.PAST.3SG une femme, (…) (minute, p.143) a.FEM.3SG woman.FEM.3SG ‘(…) this Katherine told her that there would come a woman, (…)’ Mais vous vous contenterez que but you.NOM.2PL REFL.2PL will.be.satisfied.2PL that jamais n’ en fut veu un never not of.it be.PAST.3SG seen a.MAS.3SG pareil. (heptaméron, p.19) same.MAS.SG ‘But you will be satisfied that a similar one has never been seen.’ Car il n’ y mourust point for it.NOM.3SG not there be.dying.PAST.3SG not cinquante hommes (…) (registre-journal, p.204) fifty man.MAS.3PL ‘For there died less than fifty men (…)’

Old French a. Einz que om alast before that one.NOM.3SG be.going.PAST.3SG un sul arpent de a.OBL.MAS.3SG sole.OBL.MAS.SG arpent.OBL.MAS.3SG of camp, Falt li flat.land.OBL.MAS.3SG lack.3SG he.OBL.MAS.3SG le coer, (…) (roland, p.420f.) the.OBL.MAS.3SG heart.OBL.MAS.3SG ‘Before one had crossed one sole arpent of the flat land, his heart grows faint, (…)’

Definiteness Effects in the History of French

b.

c.

d.

389

É guarde les and obey.2SG the.OBL.MAS.3PL cumandemenz nostre commandment.OBL.MAS.3PL our.OBL.MAS.3SG Seignur é çó que tu Lord.OBL.MAS.3SG and this.3SG which.OBL you.NOM.2SG saveras que li plarrad, (…) will.know.2SG which.OBL he.OBL.MAS.3SG will.please.3SG (livre reis, p.113) ‘And obey the commandments of our Lord and that of which you will know that it will please him, (…)’ (…), se il avenist chose if it.NOM.3SG be.happening.PAST.3SG thing.FEM.3SG que les bestes sauvages oississent that the.3PL beast.FEM.3PL wild.PL leave.PAST.3PL des forez, (…)11 (saint graal, p.303) of.the.3PL woods.FEM.3PL ‘(…), if it was the case that the wild beasts left the woods, (…)’ Onques es navies ne remest que never on.the.3PL nave.FEM.3PL not remain.PAST.3SG that les mariniers ; (conquête, p.106) the.OBL.MAS.3PL sailor.OBL.MAS.3PL ‘Only the sailors remained on board the naves;’

Middle French e. Quant ce vint le when it.3SG come.PAST.3SG the.MAS.3SG soir, (…) (saint louis, p.202) evening.MAS.3SG ‘When it got dark, (…)’ 11

Note that in Old and Middle French, the definite article is not necessarily present, a phenomenon commonly encountered in these stages of the language, whence the assignment of presentational ICs lacking an article such as in Ex. 18c to the set of constructions with a definite postverbal DP. Still, this assignment is possibly too liberal: according to Foulet (1928: 49), the definite article is generally absent in the medieval stages of French when “a noun is taken in a vague and indefinite sense”, while its presence ensues “to mark the definiteness”. If this is indeed the case, then presentational ICs without an article had better be considered constructions with an indefinite, rather than a definite postverbal DP.

Chapter Eleven

390

f.

g.

h.

Si fait ycy a notter le then make.3SG here to mention.INF the.MAS.3SG grant courage et bonne great.MAS.SG courage.MAS.3SG and good.FEM.SG voulenté que les vaillans will.FEM.3SG which.OBL the.3PL brave.MAS.PL François ont tous jours Frenchman.MAS.3PL have.3PL all.MAS.3PL day.MAS.3PL eu (…) (livre des fais, p.91) had ‘Then it must be mentioned the great courage and the good will which the brave Frenchman have always had (…)’ Et en icelle est declairé en brief and in this.FEM.3SG be.3SG registered in short la genealogie des vraiz the.FEM.3SG genealogy.FEM.3SG of.the.3PL true.MAS.PL roys de France. (abregé, p.114) king.MAS.3PL of France.FEM.3SG ‘And in this is briefly given the genealogy of the true kings of France.’ (…), sur lequel ilz on which.MAS.3SG they.NOM.MAS.3PL firent serment solempnel de faire make.PAST.3PL oath.MAS.3SG solemn.MAS.SG of make.INF tout ce que leur all.MAS.3SG this.3SG which.OBL they.OBL.3PL avoit esté commandé. (minute, p.87) be.having.PAST.3SG been ordered ‘(…), on which they took the solemn oath to do exactly as they had been ordered to do.’

From this initial finding one might be induced to conclude that the DE does not hold in presentational ICs in the medieval stages of French (pro de Bakker 1995, 1997 and contra Fischer 2009). Yet, as has become evident in the context of the discussion of the empirical evidence de Bakker draws on (cf. Section 2), some scepticism seems to be in order when it comes to definite postverbal DPs in presentational ICs in French. For in principle, such DPs do not necessarily indicate that the DE does not hold, given that at least in the modern standard variety of the language, postverbal DPs are by way of exception allowed to be definite when one

Definiteness Effects in the History of French

391

or several of the highly restricted set of conditions outlined in Section 1 is met. As it turns out, all of the presentational ICs in Ex. 18 above meet one of these conditions. In particular, in Ex. 18a and Ex. 18e, the definite postverbal DP is non-specific; in Ex. 18d, restrictive ne ... que ‘only’ is made use of; in Ex. 18c and Ex. 18f as well as in Ex. 18g, the definite postverbal DP is post-determined by a restrictive relative and, respectively, a PP; and in Ex. 18b and Ex. 18h, the definite DPs whose preverbal positioning results from their raising from an originally postverbal position is likewise post-determined by a restrictive relative.12 Essentially, in addition to those illustrated in Ex. 18 above, all of the other established presentational ICs with a definite postverbal DP meet one of the highly restricted set of conditions allowing by way of exception for the definiteness of the postverbal DP in Modern Standard French. In effect, aside from the relevant construction in Ex. 19 below which meets the condition of the ‘cataphoricity’ of the postverbal DP, all of the other established pertinent constructions which space considerations prevent from being likewise given meet the same conditions as the majority of those in Ex. 18 above, viz. ‘non-specificity’ and, respectively, ‘post-determination’ —by either a restrictive relative or a PP—of the postverbal DP. (19)

(…), a icellui Loÿs son to this.MAS.3SG Louis.MAS.3SG his.MAS.3SG pere fut monstré tele father.MAS.3SG be.PAST.3SG shown such.FEM.3SG vision, c’ est assavoir que vision.FEM.3SG this.3SG be.3SG be.known.INF that ledit Phelippe tenoit en said.MAS.3SG Philipp.MAS.3SG be.holding.PAST.3SG in sa main (…) (abregé, p.136) his.FEM.3SG hand.FEM.3SG ‘(…), this Louis, his father, had a specific vision, namely that said Philipp was holding in his hand (…)’

This shows that the decision on the issue of the DE in the medieval stages of French should not be made exclusively on the basis of instances of presentational ICs with a definite postverbal DP which by way of 12

The definiteness of the postverbal DP in Ex. 18a might alternatively be explained by the verb falloir. For given that in Old French, it means ‘to lack’, this verb can be readily equated with the Modern Standard French manquer ‘to lack’ which by way of exception allows the postverbal DP to be definite.

392

Chapter Eleven

exception are allowed for in (the standard variety of) a later stage of the language for which it is usually acknowledged that the DE otherwise strictly holds. Given that, apart from those just discussed, none of the other established instances of presentational ICs have a definite postverbal DP, the state of affairs regarding the DE in such constructions in Old and Middle French thus proves to be identical to that in Modern Standard French.13 In light of this crucial finding, it seems to be a natural step to give the question as to whether or not the DE holds in presentational ICs in the former stages of French the same answer as that usually given with regard to the present-day formal variety, viz. that the DE generally holds (contra de Bakker 1995, 1997 and pro Fischer 2009).14 As noted in Section 1, Old and Middle French are, unlike Modern Standard French, usually analyzed as null subject languages. Still, as illustrated in Ex. 17b-c, Ex. 17e, Ex. 17g, Ex. 17i, Ex. 18c as well as Ex. 18e above, the medieval stages of French witness a trait otherwise 13

As pointed out by two anonymous reviewers, one must not readily conclude from the absence in the medieval stages of French of presentational ICs with a definite postverbal DP other than those presently established, i.e. by way of exception allowed for in (the standard variety of) a later stage of the language, that such ICs did not exist and, thus, were ungrammatical. Yet, given that not one single such IC shows up in a large data corpus and that to my knowledge, no such IC has been reported in the literature, I shall, until proven false by pertinent empirical evidence, tentatively assume that these constructions are non-extant and, thus, ungrammatical in Old and Middle French. 14 An anonymous reviewer points out that the state of affairs regarding the DE as reflected by the results in Tab. 1 might alternatively be considered evidence in favor of a developing, rather than a materialized DE, and this in light of the following two observations: (i) in Old French, contrary to Middle French, presentational ICs with a definite postverbal DP outweigh those with an indefinite postverbal DP; (ii) unlike the former, the latter generally have an overt expletive subject pronoun. Even though these observations are, descriptively speaking, accurate, it must essentially be borne in mind that without exception, all of the established instances of presentational ICs with a definite postverbal DP are allowed in (the standard variety of) a later stage of the language which is otherwise generally argued to stand out due to a strict DE. Crucially, this applies to all such instances regardless of the respective medieval stage of French taken into consideration as well as of the (non-)occurrence of an overt expletive subject pronoun. Accordingly, the observed distributional differences regarding (in)definite postverbal DPs in Old and Middle French had better not be considered to follow from principled reasons, but rather from mere coincidence. As for the issue of the occurrence of overt expletive subject pronouns in the medieval stages of French which is, as in the modern standard variety, not categorically excluded with a definite postverbal DP, cf. the discussion in footnote 16.

Definiteness Effects in the History of French

393

characteristic of non-null subject languages (cf. Rizzi 1986, Alexiadou & Anagnostopoulou 1998, Haider 2001): the occurrence of overt pronouns which are in every respect identical to those in the subject position of ICs in non-null subject languages, i.e. overt subject expletives (Zimmermann 2009, 2014). Even though these pronouns are, unlike in the modern standard variety, not consistently overt in Old and Middle French, the frequency of their occurrence is far from marginal. In fact, the occurrence of overt expletive subject pronouns in these stages of French is at times (highly) frequent, especially when embedded clauses are taken into consideration, as shown by the results in Fig. 1 from analyzing the established diachronic data corpus (Zimmermann 2014). 100% 90% 21/23

80%

38/44

17/20

70%

42/49

23/27

22/29

60% 23/37

50% 40%

21/45

30%

9/26

20% 10%

2/10

2/14

0%

Figure 1. Frequency of overt expletive subject pronouns in embedded declaratives.

Regarding ICs, then, Old and Middle French appear to stand out due to an idiosyncrasy which might be readily considered evidence against an otherwise empirically well motivated correlation pertaining to the null subject property, viz. the consistent non-occurrence of overt expletive subject pronouns.15 Note, however, that this unexpected state of affairs 15

For extensive discussion on the occurrence of overt subject pronouns in ICs in a range of (consistent as well as partial) null subject languages, cf. Silva-Villar (1998), Holmberg & Nikanne (2002), Carrilho (2005), Hinzelin (2006) as well as Hinzelin & Kaiser (2007) who ultimately analyze these elements as either discourse markers or topic/CP expletives, rather than TP expletives. Cf. Zimmermann (2015) for extensive discussion on the issue of the analysis of ce, cela, and ça in Modern Colloquial French, the informal spoken variety, as expletives.

Chapter Eleven

394

essentially rests on the analysis of these stages of French as null subject languages. Interestingly, Old and Middle French likewise stand out due to the (highly) frequent occurrence of overt referential subject pronouns, particularly in embedded clauses, as shown in Fig. 2 (Zimmermann 2014)—again a puzzling phenomenon under the analysis of these stages as null subject languages. 100% 90%

422/432 257/278

80%

547/576 213/242 165/187

345/354 433/471

262/324

70%

154/212

60% 50%

418/434

72/133

40% 30% 20% 10% 0%

Figure 2. Frequency of overt referential subject pronouns in embedded declaratives.

In light of these both conceptually and typologically unexpected findings, it appears that an alternative analysis of the medieval stages of French as non-null subject languages is more adequate.16 16 An anonymous reviewer wonders whether with regard to 12th century Old French, it would be more appropriate to argue for a period of gradual grammaticalisation of the non-null subject property, rather than for its status as a non-null subject language, given both the theoretical need of a transitional period for the loss of the null subject property from Latin, a null subject language, to French and the empirical observation that in Fig. 1 and Fig. 2, the earliest, i.e. 12th century, texts differ considerably from the later ones in the frequency of the occurrence of overt subject pronouns. Even though such a view would in principle be conceivable, there are a number of reasons against it. First, both 12th century texts witness the occurrence of overt expletive subject pronouns, i.e. purely grammatical elements whose occurrence is exclusively conditioned by morphosyntactic factors, viz. the need for a finite sentence to have an overt subject (= the

Definiteness Effects in the History of French

395

Essentially, under such an analysis which has recently been put forward by Fischer (2009, 2010) and Zimmermann (2009, 2014), not only does the occurrence of overt expletive subject pronouns as well as of overt referential subject pronouns in unmarked contexts in Old and Middle French follow naturally, but also the general presence of the DE. Given the common claim of a correlation between the DE and the non-null subject property, the crucial finding that in the history of French, the postverbal DP in presentational ICs is indefinite, unless one or several of a highly restricted set of conditions is met, is in fact expected under the view that French has always been a non-null subject language.

References Adams, Marianne. 1987. Old French, Null Subjects, and Verb Second Phenomena. Ph.D. dissertation, University of California, Los Angeles. Alexiadou, Artemis & Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order, V-movement and EPP-checking. In Natural Language and Linguistic Theory 16: 491-539. Arteaga, Deborah. 1994. Impersonal constructions in old French. In Michael L. Mazzola (ed.). Issues and Theory in Romance Linguistics: Selected Papers from the Linguistic Symposium on Romance Languages XXIII, April 1-4, 1993, 141-156. Washington D.C.: Georgetown University Press. De Bakker, Cecile. 1995. Synchronic and diachronic variation in the French il-construction. In Marcel den Dikken & Kees Hengeveld

EPP of Chomsky (1982)). Crucially, such overt pronouns have been widely acknowledged to be characteristic of non-null subject languages only. Second, at least the earliest 12th century text constitutes a verse text. As is well known, such texts ever so often feature a number of peculiarities directly affecting the morphosyntax (Zimmermann 2014). Third, both 12th century texts feature a comparatively small number of relevant constructions. Fourth and finally, as noted in Section 1, the medieval stages of French do not freely allow null subjects, but only under specific structural conditions which are roughly tantamount to verb second configurations (Adams 1987, Roberts 1993, Vance 1997, Zimmermann 2014). Given the general claim that these configurations are particularly typical of 12th century Old French, the observed discrepancy in the occurrence of overt subject pronouns is in fact expected. Essentially, up to the 16th century, subject pronouns, in particular expletives, are null under the very same conditions as in the 12th century (Büchtemann 1912, Zimmermann 2014). From these insights it thus appears that the transitional period for the loss of the null subject property precedes the documented history of French.

396

Chapter Eleven

(eds.). Linguistics in the Netherlands 1995, 1-12. Amsterdam: Benjamins. —. 1997. Germanic and Romance Inversion in French. A Diachronic Study. The Hague: Holland Academic Graphics. Belletti, Adriana. 1988. The case of unaccusatives. In Linguistic Inquiry 19: 1-34. Büchtemann, Albert. 1912. Neutrales il im Altfranzösischen. Halle an der Saale: s.n. Carrilho, Ernestina. 2005. Expletive ele in European Portuguese Dialects. Doctoral dissertation, Universidade de Lisboa. Chomsky, Noam. 1982. Lectures on Government and Binding. The Pisa Lectures. Dordrecht: Foris, second edition. Dufresne, Monique & Fernande Dupuis. 2007. Expletives, number and language change. In José Camacho & al. (eds.). Romance Linguistics 2006. Selected Papers from the 36th Linguistics Symposium on Romance Languages, New Brunswick, March 31-April 2, 2006, 111125. Amsterdam: Benjamins. Eguzkitza, Andolin & Georg A. Kaiser. 1999. Postverbal subjects in Romance and German: Some notes on the Unaccusative Hypothesis. In Lingua 109: 195-219. Eskénazi, André. 1968. Note sur les constructions impersonnelles du français contemporain. In Revue Romane 3: 97-115. Etchegoyhen, Thierry & George Tsoulas. 1998. Thetic and categorical, attributive and referential. Towards an explanation of definiteness Effects. In: Schwegler, Armin, Bernard Tranel & Myriam UribeEtxebarria, (eds.). Romance Linguistics. Theoretical Perspectives. Selected Papers from the 27th Linguistic Symposium on Romance Languages, Irvine, 20-22 February 1997, 81-95. Amsterdam: Benjamins. Fischer, Susann. 2009. Expletives, definiteness and word-order in Romance. Accounting for the Differences between Spanish/Catalan and French. In Espinal, Maria, Manuel Leonetti & Louise McNally (eds.). Proceedings of the IV Nereus International Workshop Definiteness and DP Structure in Romance Languages, 45-61. Konstanz: Fachbereich Sprachwissenschaft, Universität Konstanz. —. 2010. Word-Order Change as a Source of Grammaticalisation. Amsterdam: Benjamins. Franzén, Torsten. 1939. Étude sur la syntaxe des pronoms personnels sujets en ancien français. Uppsala: Almqvist & Wiksel. Foulet, Lucien. 1928. Petite syntaxe de l’ancien français. Paris: Champion, third revised edition.

Definiteness Effects in the History of French

397

Gaatone, David. 1970. La transformation impersonnelle en français. In Le Français Moderne 38: 389-411. —. 1976. Il doit y avoir - *il faut y avoir. A propos de la ‘montée du sujet’. In: Revue Romane 11: 245-266. —. 1991. Il y a impersonnel et impersonnel: sur les contraintes et les finalités de la construction impersonnelle. In Michel Maillard (ed.). L’impersonnel. Mécanismes linguistiques et fonctionnements littéraires. Actes du colloque tenu à l’Université Stendhal de Grenoble du 17 au 19 mai 1990, 133-138. Grenoble: Ceditel. —. 1994. Passif, impersonnel et passif impersonnel en français: quelques réflexions. In L’information grammaticale 62: 42-44. Gamillscheg, Ernst. 1957. Historische französische Syntax. Tübingen: Niemeyer. Gross, Maurice. 1968. Grammaire transformationnelle du français. Syntaxe du verbe. Paris: Larousse, second revised and augmented edition. Haider, Hubert. 2001. Parametrisierung in der generativen Grammatik. In Martin Haspelmath & al. (eds.). Language Typology and Language Universals. An International Handbook, Volume 1, 283-293. Berlin: De Gruyter. Hinzelin, Marc-Olivier. 2006. Die neutralen Pronomina ‘ell’, ‘això’, ‘allò’, ‘açò’ und ‘ço’ in unpersönlichen Konstruktionen in diachroner und dialektaler Variation. In Claus Pusch (ed.). La gramática pronominal del catalá: variació – evolució – funció / The Grammar of Catalan Pronouns: Variation – Evolution – Function, 47-84. Aachen: Shaker. Hinzelin, Marc-Olivier & Georg A. Kaiser. 2007. El pronombre ello en el léxico del español dominicano. In Wiltrud Mihatsch & Monika Sokol (eds.). Language Contact and Language Change in the Caribbean and Beyond / Lenguas en contacto y cambio lingüístico en el Caribe y más allá, 171-188. Frankfurt am Main: Lang. Holmberg, Anders & Urpo Nikanne. 2002. Expletives, subjects, and topics in Finnish. In Peter Svenonius (ed.). Subjects, Expletives, and the EPP, 71-105. Oxford: Oxford University Press. Horning, Adolf. 1880. Le Pronom neutre il en langue d’oïl. Son origine, son extension. In Romanische Studien 4: 229-272. KĊsik, Marek. 1975. La prédétermination du substantif en séquence dans les phrases du type Il est venu un homme. In Kwartalnik neofilologiczny 22: 83-90. —. 1984. La phrase impersonnelle – problèmes de description. In Neophilologica 4: 28-40.

398

Chapter Eleven

—. 1986. Description de la phrase impersonnelle. Entre la syntaxe, la sémantique et la rhétorique. In Linguistique générale et linguistique romane. Histoire de la grammaire. Actes du XVIIe Congrès International de Linguistique et Philologie Romanes, Aien-Provence, 29 août-3 septembre 1983, 249-263. Aix-en-Provence: Université de Provence. —. 1991. Impersonnel et progression du texte. In Michel Maillard (ed.). L’impersonnel. Mécanismes linguistiques et fonctionnements littéraires. Actes du Colloque tenu à l’Université Stendhal, de Grenoble du 17 au 19 mai 1990, 183-192. Grenoble: Ceditel. —. 1998. Cataphore, impersonnel et présupposition existentielle. In Lubelskie materialy neofilologiczne 22: 217-224. Lazard, Gilbert. 1994. L’actant H: Sujet ou objet? In Bulletin de la Société de Linguistique de Paris 89: 1-28. Le Goffic, Pierre. 1993. Grammaire de la phrase française. Paris: Hachette. Lerch, Eugen. 1931. Hauptprobleme der französischen Sprache. Besonderes. Braunschweig: Westermann. Martin, Robert. 1970. La transformation impersonnelle. In Revue de Linguistique Romane 34: 377-394. Martinet, Hanne. 1975. Les variantes impersonnelles d’énoncés en français. In La Linguistique 11: 75-86. Mathieu, Eric. 2006. Quirky subjects in Old French. In Studia Linguistica 60: 282-312. Moignet, Gérard. 1971. Verbe unipersonnel et voix verbale. In Travaux de linguistique et de littérature 9: 267-282. Mussafia, Adolfo. 1886. Una particolarità sintattica della lingua italiana dei primi secoli. In Graziadio I. Ascoli & al. (eds.). Miscellanea di filologia e linguistica. In memoria di Napoleone Caix e Ugo Angelo Canello, 255-261. Florence: Successori Le Monnier. Picabia, Lélia. 1986. Remarques sur l’interprétation indéfinie. In Jean David & Georges Kleiber (eds.). Déterminants: syntaxe et sémantique. Actes du Colloque International de linguistique organisé par la Faculté des Lettres et Sciences Humaines de Metz, Centre d’Analyse Syntaxique, 6-8 décembre 1984, 227-245. Metz: Centre d’Analyse Syntaxique, Faculté des Lettres et Sciences Humaines. Rivière, Nicole. 1979. Problèmes de l’intégration de l’impersonnel dans une théorie linguistique. In Le Français Moderne 47: 289-311. —. 1981. La construction impersonnelle en français contemporain. SaintSulpice-de-Favières: Favard.

Definiteness Effects in the History of French

399

Rizzi, Luigi. 1986. Null objects in Italian and the Theory of pro. In Linguistic Inquiry 17: 501-557. Roberts, Ian. 1993. Verbs and Diachronic syntax. A Comparative History of English and French. Dordrecht: Kluwer. —. 2007. Introduction. The Null Subject Parameter. In Ian Roberts (ed.). Comparative Grammar. Critical Concepts in Linguistics Volume II. The Null Subject Parameter, 1-44. London: Routledge. —. 2010. Varieties of French and the Null Subject Parameter. In Theresa Biberauer & al. (eds.). Parametric Variation: Null Subjects in Minimalist Theory, 303-327. Cambridge: Cambridge University Press. Seelbach, Dieter. 1971. Die Extrapositionsregel im Französischen. In Dieter Wunderlich (ed.). Probleme und Fortschritte der Transformationsgrammatik, 219-245. München: Hueber. —. 1978. Transformationsregeln im Französischen aus der Sicht der historischen und romanischen Syntax. Heidelberg: Winter. Sheehan, Michele Louise. 2006. The EPP and null subjects in Romance. Ph.D. dissertation, Newcastle University. Silva-Villar, Luis. 1998. Subject positions and the roles of CP. In Schwegler, Armin, Bernard Tranel & Myriam Uribe-Etxebarria (eds.). Romance Linguistics. Theoretical Perspectives. Selected Papers from the 27th Linguistic Symposium on Romance Languages, Irvine, 20-22 February 1997, 247-270. Amsterdam: Benjamins. Skårup, Povl. 1975. Les premières zones de la proposition en ancien français. Essai de syntaxe de position. Copenhagen: Akademisk Forlag. Sportiche, Dominique. 1999. Subject clitics in French and Romance. Complex inversion and clitic doubling. In Kyle Johnson & Ian Roberts (eds.). Beyond Principles and Parameters. Essays in Memory of Osvaldo Jaeggli, 189-221. Dordrecht: Kluwer. Stéfanini, Jean. 1975. Grammaire et sémantique. In De Vriendt, Sera, Jean Dierickx & Marc Wilmet (eds.). Grammaire générative et psychomécanique du langage, 99-109. Paris: Didier. Tobler, Adolf. 1912. Rezension von Jules Le Coultre, De l’ordre des mots dans Chrestien de Troyes, Dresden 1875. In Adolf Tobler (ed.). Vermischte Beiträge. Der vermischten Beiträge zur französischen Grammatik fünfte Reihe, 395-415. Leipzig: Hirzel. Vance, Barbara. 1997. Syntactic Change in Medieval French. Verb-Second and Null Subjects. Dordrecht: Kluwer. Vet, Co. 1981. Les constructions impersonnelles en français: Une approche dans le cadre de la Grammaire Fonctionnelle de S. C. Dik. In Travaux de Linguistique 8: 49-64.

Chapter Eleven

400

Zimmermann, Michael. 2009. On the evolution of expletive subject pronouns in Old French. In Georg A. Kaiser & Eva-Maria Remberger (eds.). Proceedings of the Workshop Null-subjects, Expletives, and Locatives in Romance, 63-92. Konstanz: Fachbereich Sprachwissenschaft, Universität Konstanz. —. 2014. Expletive and Referential Subject Pronouns in Medieval French. Berlin: De Gruyter. —. 2015. On the (non-)referentiality of ce, cela, and ça in colloquial French. In Zeitschrift für französische Sprache und Literatur 125: 116. Zribi-Hertz, Anne. 1982. La morphologie verbale passive en français: essai d’explication. In Jacqueline Guéron (ed.). Grammaire transformationnelle. Théorie et méthodologies, 127-153. Saint-Denis: Presses Universitaires de Vincennes.

Medieval French Editions Consulted abregé chronique rimée comines

conquête

galien heptaméron livre des fais livre reis

Noël de Fribois. Abregé des croniques de France. Kathleen Daly & Gillette Labory (eds.). Paris: Champion. 2006. Philippe Mouskes. Chronique rimée. Tome II. Frédéric de Reiffenberg (ed.). Bruxelles: Hayez. 1838. Nouvelle collection des mémoires pour servir à l’histoire de France, depuis le XIIIe siècle jusqu’à la fin du XVIII. Tome IV. Joseph-François Michaud & Jean-Joseph-François Poujoulat (eds.). Paris: Éditeur du commentaire analytique du Code civil. 1837. Josfroi de Vileharduyn. La conqueste de Costentinoble. D’après le manuscrit n°2137 de la B.N.. Nancy: Centre de Recherches et d’Applications Linguistiques, Université de Nancy II. 1978. Galien le Restoré en prose. Hans-Erich Keller & Nikki L. Kaltenbach (eds.). Paris: Champion. 1998. Marguerite de Navarre. Nouvelles. Yves Le Hir (ed.). Paris: Presses Universitaires de France. 1967. Le livre des fais du bon messire Jehan le Maingre dit Bouciquaut, Mareschal de France et Gouverneur de Jennes. Denis Lalande (ed.). Geneva: Droz. 1985. Li Quatre Livre des Reis. Die Bücher Samuelis und der Könige in einer französischen Bearbeitung des 12. Jahrhunderts. Nach der ältesten Handschrift unter Benutzung der neu aufgefundenen Handschriften. Ernst

Definiteness Effects in the History of French

minute

registre-journal roland saint graal saint-léger

saint louis

401

R. Curtius (ed.). Dresden: Gesellschaft für Romanische Literatur. 1911. La minute française des interrogatoires de Jeanne la Pucelle. D’après le Réquisitoire de Jean d’Estivet et les manuscrits de d’Urfé et d’Orléans. Paul Doncoeur (ed.). Melun: Argences. 1952. Pierre de l’Estoile. Registre-Journal du Règne de Henri III. Tome I. Madeleine Lazard & Gilbert Schrenck (eds.). Geneva: Droz. 1992. La Chanson de Roland. Cesare Segre (ed.). Milano: Ricciardi. 1971. L’Estoire del saint Graal. Tome II. Jean-Paul Ponceau (ed.). Paris: Champion. 1997. Saint-Léger. Étude de la langue du manuscrit de Clermont-Ferrand suivie d’une édition critique du texte avec commentaire et glossaire. Joseph Linskill (ed.). Geneva: Slatkine. 1974. Joinville. Vie de saint Louis. Jacques Monfrin (ed.). Paris: Dunod. 1995.

IV BILINGUAL/SECOND LANGUAGE ACQUISITION PERSPECTIVE

CHAPTER TWELVE DEFINITENESS EFFECTS IN GERMAN-TURKISH BILINGUALS ACQUIRING ENGLISH AS THIRD LANGUAGE TANJA KUPISCH

1. Introduction1 The present paper investigates transfer in the context of third language (L3) acquisition based on the definiteness effect. Third language acquisition is generally understood as the acquisition of any language acquired after the first foreign language. In other words, an L3 can also be a person's fourth or fifth language.2 In a slightly different manner, the present paper deals with a language which is chronologically the third one –which is why we use the label “L3”í, but which represents the first foreign language. More specifically, we will look at the English of Turkish heritage speakers in Germany who were exposed to Turkish from birth and who acquired German either (quasi) simultaneously (ages 0-3), or as an early second language (ages 4-7). English represents their third language in order of acquisition and the first language that was learnt in a typical foreign language setting, i.e., through explicit instruction at school. 1

The idea of investigating transfer of the DE originated at McGill University in 2007-2008 with my colleagues Alyona Belikova (University of of Calgary), Paul Hagström (Boston University), Öner Özçelik (Indiana University Bloomington) and Lydia White (McGill). I wish to thank them for inspiration, continuous exchange of ideas and feedback. The test material used here (and in the other two contributions in this book section) was developed together with them. I further wish to thank Ilse Stangen for her dedication in this project on Turkish heritage speakers, Deniz Akpinar and Cigdem Güney for supporting the data collection, and Anika Lloyd-Smith and Henrik Gyllstad for their work on proficiency-related aspects. Many thanks to Esther Rinke for valuable feedback. 2 In the literature, the abbreviations L3 and Ln acquisition are often used interchangeably.

The DE in German-Turkish Speakers with L3 English

405

In this respect, the population studied here differs from the typical learners in previous L3 acquisition research, but the question which transfer mechanisms apply is equally relevant. In the present study transfer is investigated with respect to the Definiteness Effect (henceforth DE). The DE differs systematically between Turkish and German, which means that potential transfer into L3 English should lead to different outcomes, depending on the transfer source (and provided there is transfer). As outlined in various other papers in this volume, the DE is a syntactic restriction on the occurrence of definite NPs. This generalization is valid for positive and negative existentials, as illustrated in (1) and (2) for English. (1)

a. b.

There is a cat in my garden. *There is the cat in my garden.

(2)

a. b.

There isn’t a cat in my garden. *There isn’t the cat in my garden.

German, like English, has definite and indefinite articles and displays similar effects as English, as illustrated in (3) and (4) (though see Zielke, this volume, for differences between German and English). (3)

a. b.

Es ist eine Katze in meinem Garten. *Es ist die Katze in meinem Garten.

(4)

a. b.

Es ist keine Katze in meinem Garten. *Es ist nicht die Katze in meinem Garten.

Turkish has no definite articles and no fully grammaticalized indefinite articles, but it nevertheless displays definiteness effects in existential sentences.3 Positive existential sentences are formed with the verb var ‘exist’ and they show the DE (see 5), similar to English and German. However, the DE is absent in negative existentials, formed with the verb yok ‘not exist’ (see 6), unlike in English and German.

3

The numeral bir ‘one’ can be used in some of the contexts where German and English require indefinite articles, but this is not relevant to the subsequent discussion.

Chapter Twelve

406

(5)

a.

b.

(6)

a.

b.

Bahçe-de birkaç çocuk var. garden-LOC some child exist ‘There are some children in the garden.’ *Her çocuk bahçe-de var. every child garden-LOC exist ‘There is every child in the garden.’ Bahçe-de çok a÷aç yok. garden-LOC many tree not-exist ‘There aren’t many trees in the garden.’ Ali bahçe-de yok. Ali garden-LOC not-exist ‘There isn’t Ali in the garden.’

In summary, German and English both show the DE in positive and negative existentials. Turkish shows the same effects in positive existentials, but not in negative existentials. Thus, if learners transfer from Turkish into English, they might have problems recognizing definiteness violations in negative existentials, while having no problems with positive existentials. The goals of the present contribution are to investigate the transfer source when heritage speakers acquire foreign languages and to discuss the implications of our results for current theoretical debates in L3 acquisition research. This paper is structured as follows: Section 2 provides further information on the DE in English as compared to German and Turkish, summarizes previous findings in acquisition research and formulates predictions. Section 3 outlines current issues in L3 acquisition research. Participants, experimental conditions and results are presented in section 4. The paper concludes with a discussion and conclusion in sections 5 and 6.

2. The DE in English Compared to German and Turkish As mentioned in the previous section, English shows the DE in positive and negative existentials. This means that existential sentences with indefinite articles and other weak determiners (e.g. some, many) are allowed, while strong determiners like definite articles, proper names, possessive pronouns and strong quantifiers (e.g. all, every) are disallowed. This is illustrated in (7) for positive (affirmative) existentials and in (8) for negative existentials.

The DE in German-Turkish Speakers with L3 English

(7)

a. b. c. d.

There's a very cute cat in the pet store. There are many cats in the pet store. *There's the phone here for you now. *There's your dog at my house.

(weak) (weak) (strong) (strong)

(8)

a. b. c. d.

There isn't a suitable bowl here. There aren’t many cats in the pet store. *There isn't the phone here for you. *There isn't your dog at my house.

(weak) (weak) (strong) (strong)

407

The German sentences corresponding to (7) and (8) result in parallel grammaticality judgments; see examples (3-4). Turkish has no articles, but with other weak and strong determiners, the sentences corresponding to (7) and (8a,b) would result in the same grammaticality judgments as in German and English. However, the Turkish equivalents to (8c,d), which contain strong determiners in negative existentials, would be grammatical (see 6a,b). The relevant properties are summarized in Table 1 (see also Zielke, this volume, for further examples). Positive existentials Negative existentials

English Yes

German yes

Turkish Yes

Yes

yes

no

Table 1. Existence of the DE in English, German and Turkish Previous work on L1 Turkish learners of L2 English has shown that the DE can be acquired at more advanced levels of L2 proficiency (White, Belikova, Hagström, Kupisch & Özçelik, 2012). Furthermore, bilingual heritage speakers of Turkish in Germany acquire the DE like monolinguals in their two languages, i.e., they accept definite NPs in negative contexts in Turkish, while rejecting them in German (Kupisch, Belikova, Özçelik, Stangen & White, forthcoming). Thus, the DE does not appear to transfer, neither at advanced levels of L2 proficiency nor between the two languages of early bilinguals. The question arises whether transfer is also absent when early bilingual heritage speakers acquire a foreign language. In other words, do they transfer (or not) from one or both of their early acquired languages. If they transfer, the following scenarios are possible: (i) Transfer from L2 German, which results in ceiling performance in positive and negative existentials in L3 English.

408

Chapter Twelve

(ii) Transfer from L1 Turkish, which results in different success rates in positive vs. negative existentials with better performance in positive existentials (where Turkish and English pattern alike).

3. L3 Acquisition and Source of Transfer Studying L3 acquisition as an independent field (as compared to “classical” L2 acquisition studies) has become increasingly important, because a growing number of people are multilingual with more than two languages. L3 studies have so far focussed on investigating the source of transfer, because since the earliest studies it has become clear that in such multilingual situations transfer does not necessarily occur from the L1. Furthermore, L3 acquisition studies have mostly involved late bilinguals, i.e., L3 learners who grew up monolingually and learnt a second language when the L1 acquisition process had already been completed (though see Gallardo del Puerto, 2007 and Polinsky, 2015 for exceptions). The present study will take a fresh look at multiple language acquisition from the perspective of heritage speakers (henceforth HSs) acquiring a foreign language. These early bilinguals differ from late bilinguals in starting to acquire a second language (here referred to as L2, German) when the acquisition of their first language (here referred to as L1, Turkish) is still ongoing. Furthermore, they are exceptional in terms of the dominance relation between their L1 and their L2. While late bilinguals tend to be more proficient in their L1 than in their L2, many HSs become dominant in their L2, the language of their national environment, over time. Previous studies in L3 acquisition have been discussed in the light of three (partially competing) models: the Cumulative Enhancement Model (CEM), the Typological Primacy Model (TPM), and the L2 Status Factor Model (L2SFM). Except for the CEM, these models were meant to account for transfer at the initial stages of acquisition. Although we will be dealing with advanced learners in this paper, our findings might be relevant to those models, because there is evidence that transfer also occurs at very advanced levels of L2 proficiency, and supposedly also L3 proficiency (e.g., Lardiere, 2006; White, 2003). The CEM maintains that any prior language acquisition can facilitate subsequent language acquisition, irrespective of the order in which these languages were acquired. Transfer occurs on a property-by-property basis. It only occurs when it is facilitative; otherwise it is blocked (Flynn, Foley, & Vinnitskaya, 2004). For the property investigated here, the prediction would be that HSs will have no problems with the DE due to facilitative transfer from German.

The DE in German-Turkish Speakers with L3 English

409

The TPM claims that multilingual transfer is determined by structural similarities between languages (Rothman, 2011, 2015). In other words, transfer, positive or negative, will occur from a language that is typologically close. As mentioned above, the TPM has been proposed as a model for the initial stages in L3 acquisition, where transfer means that one previous system is transferred completely, i.e., including all language components (phonology, morphology, syntax, etc.). The problem of testing this model with learners at more advanced levels of proficiency is they may have simply acquired a property (rather than transferring successfully). Moreover, early bilinguals might not need to transfer either the L1 or L2 completely. Rather, they might be able to transfer on a property-by-property basis, because they have had more bilingual experience and thus can control transfer more systematically (Rothman, 2015: 187). Alternative to the TPM, the L2 Status Factor Model (L2SFM) predicts —again for learners at the initial stages of acquisition— that L3 transfer will occur from the L2 because L2 and L3 acquisition involve cognitively similar processes, specifically declarative memory when it comes to syntax, morphology and phonology. L1 acquisition, by contrast, relies on different cognitive mechanisms from L2 and L3 acquisition, involving procedural memory. It is debatable to what extend the L2SFM applies to HSs because in the L2SFM the difference between L1 and L2 acquisition partially rests on the type of learning (Bardel & Falk, 2007, 2012), implicit and explicit respectively, and partially on one’s assumptions about critical periods in language acquisition. Although HSs acquire their L3 at different ages, their L2 is often acquired in an implicit fashion, which makes the L1-L2 distinction more elusive. Moreover, the predictions of the L2SFM do not necessarily extend to advanced learners, who may have turned explicit into implicit knowledge (Paradis, 2009). In this paper, we bring another factor into play, namely the role of language dominance. In case we can establish transfer from either German or Turkish into English, the question is whether it can be related to the speakers’ relative proficiency in their two early acquired languages. In the research to date, the role of language dominance with respect to L3 transfer has remained rather vague. The reason might be that the learners in most previous studies acquired their L2 at age 10 or even later, which in most cases implies that their L1 proficiency exceeded L2 proficiency, and that their L1 development had been completed before the onset of L2 acquisition. However, an increasingly common learning scenario nowadays is represented by HSs who acquire an early L2, which becomes their dominant language over time. Lloyd-Smith, Gyllstad & Kupisch

410

Chapter Twelve

(forthcoming) have tested this scenario with German-Turkish learners of English, focussing on the question whether speakers who are phonologically more proficient in Turkish than in German are more likely to have a Turkish accent when speaking their L3 English than speakers who are more proficient in German. It was indeed the case that speakers who had a native-like pronunciation in Turkish were more likely to be identified as L1 Turkish speakers (and not as L1 German speakers) when speaking English. In the present study, we will look at a subset of these speakers, focussing on the DE. Our question is whether heritage speakers of Turkish transfer from Turkish, their L1, or from German, their L2. Since German is structurally similar to English with respect to the DE in existential sentences, transfer from German should lead to ceiling performance. By contrast, transfer from Turkish should lead to comparatively more problems with negative existentials because Turkish allows definite NPs here, while English doesn’t. Both the general issue of the transfer source and the more specific question whether a relation exists between the dominant language and the transfer source are relevant for L3 acquisition research and bilingualism.

4. Study on the DE in L3 English 4.1 Participants The participants were 12 adult German-Turkish bilingual speakers aged between 20 and 42 years (M = 25), recruited in Hamburg, Northern Germany. The data were part of a larger dataset, previously studied in Lloyd-Smith et al. (forthcoming) with regard to global accent. For the purpose of the present study, the participants with the clearest dominance profile were selected (see below). Most of our participants were born in Northern Germany and all spoke standard varieties of German and Turkish. They all acquired Turkish from birth, and Turkish was the predominant language in their homes when they were growing up. Their parents were native speakers of Turkish born in Turkey, who spoke Turkish with one another as well as with their children. Among the siblings, German and Turkish were used. All participants held German high school diplomas. At the time of testing, most were using more German than Turkish and considered themselves to be either equally proficient in both languages or more proficient in German despite being fluent in both languages. Participants differ with respect to their AoO in German. Some can be considered simultaneous bilinguals (2L1s) with an AoO between 0-3 years (n = 7), others were early L2 learners (eL2s) with

The DE in German-Turkish Speakers with L3 English

411

an AoO between 4-7 (n = 5) (see Table 2). Since this age distinction turned out to be irrelevant in previous studies, we will not consider it further here. Instead, we will refer to all speakers generally as “early bilinguals”. All 12 speakers learnt English at school at the ages between 8 and 11. None of them had additional training in English, i.e., beyond school, and most had been in an English-speaking country for between 1 and 6 weeks. Their self-estimated proficiency was either intermediate or advanced. For the purpose of the present investigation, we distinguished three groups of speakers, who differed in terms of their relative language dominance in German and Turkish. Our dominance classification is based on two measures. One measure is a score representing the quantity and quality of the speakers’ contact with Turkish in the past and at the time of testing (in the following referred to as “TUS” for “Turkish Use Score”; see Table 2). This score was based on a questionnaire that all informants had filled out prior to testing. A high score on the TUS score results from frequent use of Turkish with many different conversation partners and in different situations (e.g., at home and work), frequent and long visits to Turkey and use of the language in different modes (speaking, writing, reading, listening), and relatively frequent use of Turkish. We did not calculate an equivalent score for German since all participants had intensive contact with German and did not differ noticeably in this respect. We further used the results of the participants’ perceived phonological proficiency in German and Turkish from a previous study (see two rightmost columns in Table 2). A high number on the phonological proficiency scale means that the speakers has been deemed native-like by a high number of different native speakers of the respective language (max. score = 15). All numbers are taken from Lloyd-Smith et al. (forthcoming). The following three groups resulted from this classification: (i)

Turkish-dominant bilinguals who tended to be perceived as native speakers of Turkish (and foreign speakers of German) and who used Turkish with a relatively high frequency.4 (ii) German-dominant bilinguals who tended to be perceived as native speakers of German (and foreign speakers of Turkish) and who used Turkish with a relatively low frequency.

4

There is a small chance that we underestimated the proficiency of these speakers in German, since the proficiency score in German is merely based on their perceived global accent.

Chapter Twelve

412

(iii) Balanced bilinguals perceived to be native speakers of Turkish who used both German and Turkish with a relatively high frequency.

(i)

(ii)

(iii)

Participant

AoO in German

TUS score

BUR GUL ULK EME DUY ARZ TUL SIB HAK AYS GOK PIN

2L1 2L1 eL2 2L1 2L1 eL1 eL2 2L1 eL2 eL2 2L1 2L1

14 15.5 16 17 7 7.5 9.5 10 14 14 15,5 16

Phon. proficiency Turkish 11 15 13 13 1 4 2 3 13 11 12 12

Phon. proficiency German 2 1 2 2 14 11 13 13 15 12 13 12

Table 2. Overview of participants and relative proficiency in German and Turkish. We wanted to test whether with respect to the DE in L3 English these bilinguals would transfer from the language in which they were most proficient. The predictions would thus be: I

Turkish-dominant bilinguals transfer from English. Therefore, have problems with the DE in negative existentials, overaccepting strong DPs. II German-dominant bilinguals transfer from German. Therefore, they would have no problems with the DE due to positive transfer from German.

The dominance-based account does not make any predictions for balanced learners, but we included them to see if any systematic patterns emerge from that group.

4.2 Design and Methods The task was a bi-modal acceptability judgment task (AJT). The test items were identical to those in White et al. (2012) but fewer in number (n = 70). The test items consisted of a context and a test sentence and were presented orally as well as in writing on a computer screen, and they were

The DE in German-Turkish Speakers with L3 English

413

automatically randomized anew for each participant. The participants were instructed to read and listen to the stimuli, and to repeat the sentence if it sounded fine to them in the given context, or to provide a spoken correction if they thought it did not sound good. Response time was limited and corresponded to the length of the stimulus sentence (read by a native speaker) multiplied by two. The reason for using an oral task was to obtain spontaneous responses and to prevent the participants from trying to access explicit knowledge of what they might have thought the test was about. Each test sentence was preceded by a short context. The context was important since some of the items were grammatical but infelicitous in the given context. An example is presented in (9), which illustrates a negative existential involving a possessive marked DP (i.e., a definite expression). Such existentials are grammatical in Turkish but not in German. If a participant appropriately rejects this English sentence, cross-linguistic influence from Turkish is unlikely because transfer from Turkish should have resulted in acceptance. (9)

Elizabeth's pet dog has run away. She calls John and asks if he has seen it. He looks around and replies: There isn't your dog here.

14 sentence types were tested, with 5 test items per type, including grammatical and ungrammatical existentials (n = 40) and sentences controlling for other aspects of (in)definiteness (n = 30). There were 8 ungrammatical types (for a total of 40 ungrammatical items) and 6 grammatical types (for a total of 30 grammatical items). The examples in (10) illustrate grammatical and ungrammatical affirmative existentials with articles and other determiners. The grammatical cases involved DPs with indefinite articles. The ungrammatical cases involved strong DPs with definite articles (10b), proper names (10c) and possessives (10d). These items test whether the L3ers are sensitive to the DE in situations where strong and weak determiners behave similarly in German and Turkish. Test Sentences with Affirmative Existentials (10)

a. b.

Grammatical with indefinite article (n = 5) There's a very cute cat in the pet store! Ungrammatical with definite article (n = 5) There's the phone here for you now.

Chapter Twelve

414

c. d.

Ungrammatical with proper name (n = 5) There's Mr. Jones here already. Ungrammatical with possessive (n = 5) There's your dog at my house.

The items in (11) illustrate the equivalent sentence types for negative existentials. Note that the both the German and the Turkish equivalent of (11a), which contains a weak DP, is grammatical. By contrast, the items in (11b-d), which include strong DPs, test sensitivity to the DE when Turkish and German behave differently, as is the case with DPs containing proper names, possessives or quantifier. In other words, the equivalents of (11bd) are ungrammatical in German but grammatical in Turkish. Test Sentences with Negative Existentials (11)

a. b. c. d.

Grammatical with indefinite article (n = 5) There isn't a suitable bowl here. Ungrammatical with definite article (n=5) There isn't the package here yet. Ungrammatical with possessive pronoun (n=5) There isn't your dog here. Ungrammatical with proper name (n=5) There isn't Mr. Jones here yet.

The test conditions in English and the grammaticality of their equivalents in the two early acquired languages (i.e. the potential transfer sources) are summarized in Table 3.

The DE in German-Turkish Speakers with L3 English

Test conditions DP Grammatical (no. of items) type in English Positive Indef./ yes existential weak (n=15) Positive Def./ no existential strong (n=5) Negative Indef./ yes existential weak (n=5) Negative Def./ no existential strong (n=15) Table 3. Summary of test conditions.

415

Grammatical in German yes

Grammatical in Turkish no

no

yes

yes

no

no

no

In addition to these 40 test items, there were 30 control items and fillers. The examples in (12) show grammatical list (12a) and deictic (12b) sentences. List readings were included in order to determine whether participants simply follow a strategy of accepting all instances of there’s+indefinite, while rejecting all instances of there’s+definite. In other words, if participants accept a sentence like Look, there's the mouse again, we can rule out that rejection of there’s+definite is a result of chunk-based learning. Further, these (grammatical) items helped counterbalancing the amount of ungrammatical definites in the test conditions. (12)

a. b.

Look, there's the mouse again. Well, for a start, there's the moon.

Finally, the sentences in (13) illustrate the sentence types that did not involve existentials or there. Since the definite article in the existential test items is always ungrammatical, sentences like (13a) are included to exclude the possibility that participants simply adopt a strategy of always rejecting all items with definite articles. Conversely, since all test items involving the indefinite article in existential contexts are grammatical, sentences like (13b) through (13d) can be used to ensure that subjects were not biased towards accepting all indefinite DPs. Moreover, these items served the purpose of counterbalancing the number of grammatical and ungrammatical indefinites.

Chapter Twelve

416

Control Items (13)

a. b. c. d.

Grammatical definite subject The next job candidate is waiting outside. Grammatical indefinite subject An apple would be nice. Ungrammatical indefinite subject A bird looks hungry. Ungrammatical indefinite object No, I've seen a movie already.

4.3. Results This section summarizes the results, starting with affirmative existentials. Figures 1 and 2 show the number of appropriate responses, Figure 1 for grammatical test sentences (containing weak DPs) and Figure 2 for ungrammatical test sentences (containing strong DPs). Given the low number of participants in each of the three groups, individual results rather than group means are shown. For weak affirmatives, a response counted as appropriate if the stimulus sentence was repeated without any correction concerning the weak DP in question. For strong affirmatives, a response counted as appropriate if the stimulus sentence was corrected and if the correction concerned the strong DP, i.e., if this DP was replaced by a weak DP, as in (14a), or if its order had been changed, as in (14b). Corrections which did not involve the weak or strong DP were counted as if no correction had been made. Failure to respond within the given time was also counted as if no relevant correction had been made. (14)

Examples of relevant corrections Stimulus: There isn’t your dog here. a. There isn’t a dog here. b. Your dog isn’t here

As shown in Figure 1, all participants accepted sentences with weak DPs in the majority of cases, i.e. repeated them as they were. Exceptions include corrections of “there is” to “here is” or “it is” (e.g., Here’s a package for you at my house), which were counted as relevant since they involved a change from an existential to a non-existential sentence. There are no noticeable differences between the three groups. The correction patterns for sentences with strong DPs, illustrated in Figure 2, show individual variation in the participants’ ability to correct

The DE in German-Turkish Speakers with L3 English

417

ungrammatical sentences. The comparison of the two figures further shows a bias towards accepting sentences, including sentences that are not acceptable in English. Again, there seem to be no group-specific patterns. In all three groups, there are participants who perform relatively well and others who perform relatively badly.

Figure 1. Acceptance of weak DPs in affirmative and negative existentials (absolute number of appropriate acceptance, max score=5).

Figure 2: Relevant corrections of strong DPs in affirmative and negative existentials (absolute number of appropriate corrections, max score=15).

418

Chapter Twelve

Figure 3 compares the results for positive and negative existentials. This time, performance is shown as % of correct responses from the total of responses. The comparison shows that the number of correct responses is similar for affirmative and negative existentials, i.e. speakers who perform well with positive existentials tend to perform well with negative existentials, too.

Figure 3. Correct responses (%) in affirmative and negative existentials.

5. Discussion The major goal of this study was to determine whether German-Turkish bilinguals show systematic transfer into L3 English from their early acquired languages. Negative existentials played a crucial role in answering this question, for Turkish and English pattern differently here, while German and English are similar. Transfer from German would predict systematic corrections. By contrast, Turkish allows strong DPs with negative existentials, so that in these contexts transfer from L1 Turkish could lead to (over)acceptance of strong DPs in English, while strong DPs in positive sentences should be rejected. We further asked whether speakers with a comparatively high proficiency in Turkish would be more likely to transfer from Turkish, based on the assumption that language proficiency determines the transfer source. Thus, especially the Turkish-dominant bilinguals, possibly also balanced bilinguals, were predicted to have particular problems with negative sentences. Alternatively, transfer could have been determined by typological

The DE in German-Turkish Speakers with L3 English

419

proximity. In that case, all participants should have performed very well, given structural similarities between German and English with regard to the DE.

5.1 Absence of Transfer The results illustrated above (see especially Figure 3) did not confirm any of these expectations. The bilinguals performed better with acceptable sentences than with non-acceptable sentences, as has been observed in previous acceptability judgment tasks with L2 speakers and HSs. However, most bilinguals obtained similar results for affirmative and negative sentences. In other words, some bilinguals performed well with both positive and negative existentials (GUL, EME, DUY, ARZ, HAK, AYS, PIN), while others had problems with both sentences types. There doesn’t seem to be an advantage with affirmative as compared to negative existentials. This suggests that (i) Turkish is not a source of negative transfer, and (ii) transfer does not explain the learners’ general behavior. Moreover, speakers whose German is strong do not have an advantage over Turkish-dominant speakers, suggesting that relatively proficiency in the early acquired language does not determine the speakers’ performance in L3 English, contrary to the predictions in I and II above.

5.2 Alternative Explanations One might speculate whether the expected effects were absent because the speakers were too advanced to transfer from their previously acquired languages. However, advanced proficiency would have implied good performance in general, which was not the case for all speakers. Conversely, the learners’ proficiency may have been too low to deal with the test sentences. If the latter was correct, there should be a relation between their performance in the DE experiment and their overall proficiency. In order to explore this possibility, the results of the DE experiment were compared to other proficiency-related data from participants that was available to us. These included: (i) age of exposure to English, (ii) self-estimated proficiency, (iii) affective relation with English, (iii) residence in an English-speaking country, (iv) perceived foreign accent.

420

Chapter Twelve

Age at first exposure to English. The participants did not differ substantially in this respect (range 8-12 years). Nevertheless, these years may be crucial such that earlier exposure to English promoted better performance in this language. However, the participants with the earliest age of exposure (8-9 years) were EME, BUR and SIB, and in the DE experiment they did not behave similarly. ARZ, GÖK, HAK, ULK, TUL, AYS, DAY were exposed to English relatively late (11-12 years). They also perform very differently from each other, but not necessarily worse than the earlier acquirers. Thus, age of onset cannot explain the patterns found in this study. Self-rated proficiency. The 12 participants reported intermediate to advanced overall language skills in English in reading, writing, listening, and speaking. Those reporting advanced skills íGÖK, PIN, ULK, AYS, DUYí ended up performing differently from each other in the DE test. The same holds for the other 7 participants who considered themselves to be advanced. The latter were different from each other, and not all of them outperformed the bilinguals at intermediate proficiency levels. Affective factors. Most participants claimed that they liked to speak English, except for EME, SIB and TUL. EME specified that she did not like her pronunciation, while SIB and TUL were generally uncomfortable using English. The latter two were also the ones who had most problems with the task. Therefore, there might be a relation between overall performance and affective factors. Residence in an English-speaking country. All but two speakers had been to an English-speaking country. These two participants (ARZ, SIB) performed different from each other in the DE experiment. The only speaker who had spent a year in an English-speaking country performed very well, but so did others who had spent less than two weeks in a comparable environment (EME, AYS, DUY, GÜL). So, length of time in an English-speaking environment only cannot explain the results either. Perceived foreign accent. Finally, the results from the DE experiment were compared to those of a previous study on the participants’ foreign accent when speaking English (Lloyd-Smith et al., forthcoming). In that study, native speakers of English rated spontaneous speech samples of our participants in terms of perceived accent strength (1 = native-like, 7 = strong foreign accent). On average, the participants’ accents were rated 3.5 (range 2.5-4.7). TUL and SIB both had a comparatively strong accent and did not perform very well on the DE experiment (see Figure 3). On the other hand, ARZ und EME also had relatively strong accents but performed well in the DE experiment. ULK and GOK also perform well in the DE experiment but ULK has comparatively strong accent (4.65), while

The DE in German-Turkish Speakers with L3 English

421

GOK has a rather mild accent (2.75). Thus, there seems to be no relation between perceived accent strength and performance in the DE experiment. In short, comparison of the DE data to other information about the bilinguals does not provide conclusive evidence that one particular factor determines the bilinguals’ performance. The data also appears to speak against systematic transfer at lower levels of proficiency. However, a potential limitation of this study is the absence of an independent proficiency test. The data on self-perceived proficiency and foreign accent may be related to general proficiency, but they cannot be seen as representing all aspects of language. For instance, speakers who are not particularly good at pronouncing foreign languages may have excellent grammar skills. This is why our findings cannot be taken to mean that proficiency is generally irrelevant.

6. Conclusions The study was concerned with the question whether L3 speakers of English with German and Turkish as their early acquired languages, show systematic transfer from any of these two languages. Another question was whether Turkish-dominant bilinguals with relatively lower proficiency in German would be more likely to transfer in the domain of negative existentials where Turkish differs from English (and German). Contrary to expectations, the results show general problems with strong DPs in both positive and negative existentials. Since negative transfer would obtain in negative existentials only, transfer alone cannot account for the results. Moreover, no systematic relation emerged between the observed patterns and language dominance: the Turkish-dominant bilinguals did not systematically differ from the other speakers, as they were not particularly prone to negative transfer from Turkish. Good performance in the DE experiment might be related to overall proficiency in English, but we could not provide firm evidence if favour of this claim. The results are somewhat inconclusive when it comes to the question whether L3 transfer models apply to heritage speakers like those we tested. Systematic transfer from German, as predicted by both the L2SFM and the TPM, should have resulted in ceiling performance, contrary to our results. This does not imply that the findings of the present study invalidate these models since both the L2SFM and the TPM make predictions for speakers at the initial state of L2 acquisition. Moreover, the L2SFM envisages L2ers who were formally instructed in their L2, which is not generally true of our participants. On the other hand, since transfer is known to be persistent at advanced levels of proficiency with regard to other grammatical

422

Chapter Twelve

properties, e.g., articles, it remains an open question why the DE does not transfer (see also Snape & Sekigami and Zielke, this volume).

References Bardel, Camilla & Ylva Falk. 2007. The role of the second language in third language acquisition: the case of Germanic syntax. In Second Language Research 23.4: 459-484. —. 2012. The L2 status factor and the declarative/procedural distinction. In Jennifer Cabrelli Amaro, Suzanne Flynn, & Jason Rothman (eds.). Third Language Acquisition in Adulthood, 61-78. Amsterdam: John Benjamins. Flynn, Suzanne, Claire Foley & Ianna Vinnitskaya. 2004. The cumulativeenhancement model for language acquisition: Comparing adults' and children's patterns of development in first, second and third language acquisition of relative clauses. In International Journal of Multilingualism 1.1: 3-16. Gallardo del Puerto, Francisco. 2007. Is L3 phonological competence affected by the learner's level of bilingualism? In International Journal of Multilingualism. Kupisch, Tanja, Alyona Belikova, Öner Özçelik, Ilse Stangen & Lydia White. Forthcoming. On complete acquisition in heritage speakers: The definiteness effect in German-Turkish bilinguals. In Linguistic Approaches to Bilingualism. Lardiere, Donna. 2006. The Fossilized Steady State in Second Language Acquisition. Mahwah NJ: Lawrence Erlbaum. Lloyd-Smith, Anika, Henrik Gyllstad & Tanja Kupisch. Forthcoming. On L3 transfer when the L2 becomes the dominant language: Global accent in Turkish heritage learners acquiring English as an L3. In Linguistic Approaches to Bilingualism. Paradis, Michel. 2009. Declarative and Procedural Determinants of Second Languages. Amsterdam: John Benjamins. Polinsky, Maria. 2015. When L1 becomes an L3: Do heritage speakers make better L3 learners? In Bilingualism: Language and Cognition 18.2: 163-178. Rothman, Jason. 2011. L3 syntactic transfer selectivity and typological determinacy: The typological primacy model. In Second Language Research 27.1: 107-127. —. 2015. Linguistic and cognitive motivations of the Typological Primacy Model (TPM) of third language (L3) transfer: Timing of acquisition

The DE in German-Turkish Speakers with L3 English

423

and proficiency considered. In Bilingualism Language and Cognition 18.2: 179-190. White, Lydia. 2003. Fossilization in steady state L2 grammars, persistent problems with inflectional morphology. In Bilingualism: Language and Cognition 6: 129-141. White, Lydia; Alyona Belikova, Paul Hagström, Tanja Kupisch & Öner Özçelik. 2012. Restrictions on definiteness in second language acquisition: affirmative and negative existentials in the L2 English of Turkish and Russian speakers. In Linguistic Approaches to Bilingualism 2.1: 54-89.

CHAPTER THIRTEEN JAPANESE SPEAKERS’ L2 ACQUISITION OF THE ENGLISH DEFINITENESS EFFECT NEAL SNAPE AND SETSU SEKIGAMI

1. Introduction This paper investigates the acquisition of the Definiteness Restriction that exists in English existential there constructions. Specifically, the aim of our study is to look at judgments by Japanese second language (L2) learners on affirmative and negative existential sentences to see whether they accept violations of existential there constructions such as, e.g., *There is the man in the room. The example is considered a violation because of the use of the strong definite determiner the. As an existential sentence, it is ungrammatical as only weak determiners such as the indefinite determiner a are grammatical. Native speakers of English are not likely to accept existential sentences with strong determiners when presented in a context that clearly gives an existential interpretation, but when it comes to judgments by Japanese L2 learners it is unclear if they would be able to make a distinction as Japanese is an article-less language. At first glance, any first language (L1) transfer effects from Japanese may hinder the L2 learner in acquiring articles because Japanese has no morphological markers like articles, apart from demonstratives like sono (the/that) (Wakabayashi, 1998). Indeed, a growing number of studies have investigated whether L2 learners can acquire articles (the & a) in English (Huebner, 1985; Thomas, 1989; Robertson, 2000; White, 2003 among others). Some of these studies have focused on the acquisition of article semantics (Ionin, Ko & Wexler, 2004) whilst others have looked at the acquisition of prosodic structures to represent articles as target-like or nontarget like elements (Goad & White, 2004). Studies that have exclusively investigated article acquisition by Japanese L2 learners found that learners (i) often failed to produce articles in spoken production where they are required (see example (1)) (e.g., Shirahata, 1995), (ii) experienced

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

425

difficulties with selecting the appropriate article in a discourse situation (see (2)) (Butler, 2002; Snape, 2013; Snape, García-Mayo & Gürel, 2013), (iii) the count-mass distinction is a source of errors in article choice (see (3)) (Snape, 2008; Inagaki & Barner, 2009, Inagaki, 2014). (1)

*I’m looking for book.1

(2)

#I’m going to meet the friend tomorrow.

(3)

*I need an information.

Possible explanations put forward to try and account for the difficulties learners have with articles in spoken production range from the Missing Surface Inflectional Hypothesis (Prévost & White, 2000) to the Prosodic Transfer Hypothesis (Goad & White, 2004). Essentially, the two accounts attempt to show why L2 learners fail to map functional morphology like articles to the phonological output. The Feature Reassembly Hypothesis (Lardiere, 2009) is a two-step process of mapping and (re)assembling formal feature matrices. For the first step, any similarities between the L1 and L2 should be relatively straightforward for learners as they should be able to map the features from the L1 to L2. For example, Japanesespeaking learners of English have shown that they can successfully map the Japanese past tense feature of –ta onto the English regular past tense marker –ed since they share the same meaning, i.e., to indicate an event that took place in the past (Al-Hamad et al., 2002). The second step involves feature reassembly where features can be added or deleted. This is achieved gradually by making adjustments based on input-based evidence for meaning and usage. (Slabakova et al., 2014). In the case of articles, Japanese has demonstratives, one of which resembles a definite article in certain contexts, but crucially the difficulty for learners without a formal article system is that reassembly of features available in the L1 which must be mapped onto specific morphological items in the L2, i.e., the and a, may take time because the English article system is semantically and pragmatically complex. Despite the problems Japanese learners have with articles, they, like other L2 learners, seem to fair much better in the use of the There is a … existential construction (Thomas, 1989; White, 2003; Lardiere, 2004). It turns out that the existential construction is less problematic for L2 1 * represents an ungrammatical sentence and # indicates that the sentence is pragmatically odd in a discourse situation.

426

Chapter Thirteen

learners in regards to using the appropriate article, the indefinite article in There is a … existential construction. Recent work by White (2008a, b), King, Steinhauer and White (2006) and White et al. (2009, 2012) have explored whether L2 learners from different L1 backgrounds (Mandarin Chinese, Russian and Turkish) can make distinctions between grammatical existential sentences and ungrammatical sentences. Their collective findings show that even if the L1 lacks articles, as is the case for Mandarin Chinese and Russian, L2 learners were able to successfully make judgments about affirmative and negative existential sentences (see examples in section 2 below). In the case of existentials in Japanese there are two particular verbs that can be combined with the nominal marker ga to provide an existential interpretation, ga + ar-u and ga + ir-u, and discussion about these two constructions is given in section 3. Notwithstanding the difficulties Japanese L2 learners have with articles, we show that our Japanese participants are able to recognize the Definiteness Restriction (Milsark, 1974, 1977) or the Definiteness Effect (DE) in grammatical existential constructions. The participants were instructed to make judgments on a range of sentence types and they were asked to rate the sentences on a scale as being grammatical (natural within a context) or ungrammatical (unnatural within a context) on an acceptability judgment task (AJT). We compare the learners’ performance with native English speakers’ performance in this small-scale study and demonstrate that in contrast to previous findings on the L2 acquisition of articles, Japanese learners at high proficiency levels can make a distinction between indefinite and definite DPs, the former being grammatical, e.g., There is a fly in my soup, versus the latter being ungrammatical, e.g., *There is the fly in my soup.

2. The Definiteness Effect in English There are a number of approaches to explain the definiteness restriction (Rando & Napoli, 1978; Abbott, 1993; Safir, 1987; Zucchi, 1995, amongst others), but for the purpose of our study we examine how the DE functions in languages like English and Japanese. In English, definiteness is expressed by articles as well as by other determiners (demonstratives, possessives, numerals, quantifiers). In relation to existential sentences, English allows indefinite DPs, such as the one in (4), but not definite DPs in (5). The restriction on affirmative definite DPs extends to negative existentials in (7), not for negative indefinite DPs, as in (6). (4)

There is a student in the classroom.

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

(5)

*There is the student in the classroom.

(6)

There isn’t a student in the classroom.

(7)

*There isn’t the student in the classroom.

427

Milsark (1977) calls definite DPs, which cannot occur in there existential constructions, strong expressions and indefinite DPs are weak expressions. The types of weak and strong expressions are featured in Table 1. Weak (indefinite DPs)

Strong (definite DPs)

indefinite (a)

definite (the)

Quantifiers (some, many, few, no)

quantifiers (all, most, every, each)

cardinal numbers (one, two, three etc.)

demonstratives, possessives (this, that, my, his, their)

bare plurals, mass nouns personal pronouns, proper names (chairs, furniture) (I, he, she, John, Mary) Table 1. Weak and strong expressions in English (adapted from White et al. 2012: p. 56). There are other occurrences of there BE sentences without an existential interpretation (Lyons, 1999; Kearnes, 2011). For example, Lyons (1999) suggests that the deictic locative adverb there (meaning ‘in/to that place’) accompanying the verb to be allows a definite DP or a proper name as its complement, unlike existential there, as in examples (8) and (9). (8)

There’s the man you’re looking for, behind those bushes.

(9)

As she turned the corner she froze; there was Joe, staring straight at her. __ (Lyons, 1999)

There is also a phonological reduction of existential there (from [ðԥz] for there is and [ðԥrԥ] for there are) to the unstressed form thr, unlike the deictic form there. Furthermore, existential there is typically referred to as a pleonastic, a semantically empty filler in subject position which does not

Chapter Thirteen

428

add any further meaning to a sentence, as (10) illustrates since (10a) and (10b) do not differ in meaning. (10)

a. b.

There is a man in the room. A man is in the room.

According to Lyons (1999), there are three other types of there. List reading there is where definite DPs and proper names can be used, as in (11).  (11)

a. b.

What have we got to eat? Well, there’s the chicken, the bacon, and that cheese you bought.

Presentational there can introduce a new entity into the discourse situation, as shown in (12). (12)

There are only three rooms available.

The final there construction is task there which can take a weak or strong DP. Task there differs from the other types of there constructions as shown in (13). Task refers to a number of things one has to carry out. (13)

There are some of / most of the trees to prune and spray.

In sum, the examples for (8-13) clearly show how the deictic, list, presentation and task readings differ from examples (4-7) above with the existential readings. Here, we are only concerned with whether Japanese L2 learners can make a distinction between grammatical sentences (as in (4) and (6)) and ungrammatical sentences (as in (5) and (7)).

3. The Definiteness Effect in Japanese Japanese allows a Subject Object Verb (SOV) existential reading shown in examples (14-22) below. Though the ga ir-u (nominative marker + animate exists) / ga ar-u (nominative marker + inanimate exists) construction resembles the English There is a … construction, the meaning of ga ir-u / ga ar-u can also have the same interpretation as to have and to be since Japanese has no overt expletive. In Japanese, the existential meaning can be expressed by the verb ir-u for animacy or by ar-u for inanimacy. Examples (14) and (15) are two types of affirmative

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

429

existential sentences with an SOV reading. Only weak expressions can occur in Japanese possessive sentences, as the example in (14) is grammatical but (15) is ungrammatical because strong DPs are not possible. (14)

SOV reading Watashi ni wa {ooku / ikuraka} no zaisan ga ar-u. I DAT a lot of / little fortune NOM exist-NPST ‘I have {a lot of / little} fortune’.

(15)

SOV reading *Watashi ni wa {subete / hotondo} no zaisan ga ar-u. I DAT all / most fortune NOM exist-NPST ‘I have {all / most} fortune’.

The locative and possessive constructions appear to be the same, i.e., ga ir-u, but in English the existential is There is a … and the possessive uses the verb to have.2 In other words, there are two distinct structures in English and only one in Japanese. The difference between these two readings in Japanese is that the possessive construction has a restriction as it only allows indefinite DPs. Negative existentials in Japanese, like Turkish and Russian, do not show a definite/indefinite contrast, as weak and strong DPs are acceptable in negated existential constructions as the following examples (16-18) indicate. (16)

SOV reading Gakusei ga i-nai. student NOM exist-NPST-NEG not there. ‘There isn’t a student / There aren’t any students’.

The negative example in (16) can be read as There isn’t a student or A student isn’t here. (17)

SOV reading John / ano gakusei ga i-nai. John / that student NOM exist-NPST-NEG not there ‘There isn’t John / that student’.

2 As one reviewer notes, existentials are formed with have in other languages, e.g. French and some dialects of German.

430

(18)

Chapter Thirteen

SOV reading Ano gakusei-tachi ga i-nai. those students NOM exist-NPST-NEG ‘There aren’t those students’.

The sentences in (17) and (18) sound infelicitous with an ‘existential’ meaning. Probably, this is because the proper noun or the determiner that or those trigger ‘an existential presupposition’. Strong DPs already imply the existence of John, that student, those students. Thus, the sentences in (17) and (18) sound better if John, that student and those students have been previously established and are already presupposed. The example in (19) demonstrates that weak expressions such as ooku no (many), takusan no (much), suu-nin no (a few people of), syousuu no (a small number of), and numerals like san-nin ijyou no (at least three people)are possible as they are indefinite NPs. (19)

SOV reading Waratta {ooku no / takusan no / suu-nin no / syousuu no / laughed many much a few a small number of san-nin ijyou no} gakusei ga ir-u. three-CL at least student NOM exist-NPST ‘There are / many / much / a few / a small number of / at least three / students who laughed’. 3

There is a restriction for noun phrases with modifiers expressing quantification such as subete no (all of), daibubun no (most of), and noun phrases with demonstrative adjectives sono, ano and kono. All of the following examples from (20-22) show that strong expressions are not possible unless they occur in a locative existential sentence (LSV) like (23) (Kageyama & Kishimoto, 2011). (20)

3

SOV reading *Subete no waratta gakusei ga ir-u. all laughed student NOM exist-NPST ‘There are all students who laughed’.

If the quantifiers come before waratta (laughed), the example sentence in (19) sounds better. This might be because the quantifier should quantify over O[laughed(x) šstudent(x)]. Probably, gakusei (student) needs to combine with waratta (laughed) before it combines with a quantifier.

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

431

(21)

SOV reading *Daibubun no waratta gakusei ga ir-u. most of laughed student NOM exist-NPST ‘There are most of the students who laughed’.

(22)

SOV reading *Sono / *ano / *kono waratta gakusei ga ir-u. the / that / this laughed student NOM exist-NPST ‘There is the / that / this student who laughed’.

(23)

LSV reading Kyoushitsu ni sono / ano / kono / subete no / daibubun no classroom LOC the / that / this all / most of waratta gakusei ga ir-u. laughed student NOM exist-NPST ‘There are {the / that / this / all / most of } student(s) who laughed in the classroom’.

To summarize, Japanese has existential ga a-ru and ga ir-u constructions. For affirmative existentials, strong DPs are grammatical in locative existential sentences (see 23)), but in SOV and possessive constructions (see (14-15)), strong DPs are ungrammatical. Conversely, negative existentials can be indefinite (weak DPs) (see 16)) or definite (strong DPs) (see 17-18)) as there is no restriction. The main differences between English and Japanese are summarized in Table 2 below. Language

Affirmative Existentials

Negative Existentials

English Japanese

DE DE DE in SOV sentences No DE No DE in locative sentences Table 2. Affirmative and Negative Existentials in English and Japanese. In this paper, we examine whether L2 learners whose first language is Japanese know how restrictions on definiteness work in English. The fact that Japanese L2 learners typically struggle with the English article system does not mean they do not know how definiteness functions in regards to the use of the definite article and the indefinite article. White et al. (2012) point out that the DE in English is rather subtle and learners may not know that there is a restriction. In Japanese, affirmative and negative existentials can be strong or weak DPs unless the context provides further information

432

Chapter Thirteen

where it can be assumed that the ga ir-u / ga ar-u construction has a definite or indefinite interpretation. Therefore, differences between the L1 and the L2 may result in cross-linguistic influence where learners simply accept affirmative and negative grammatical and ungrammatical types of existential sentences. On the other hand, if learners are sensitive to the DE it may indicate that there is some universal principle guiding the acquisition process. In other words, some things come for free in the acquisition process and it seems that the acquisition of cross-linguistic semantics, in this case the DE in L2 English, is somewhat easier than L2 morphology (Slabakova, 2008) or phonology (Goad, White & Steele, 2003). 

4. Previous L2 Research on the Definiteness Effect L2 studies from the 80ǯs and 90ǯs that investigated the acquisition of articles by L2 learners tended to largely focus on the L2 learnersǯ ability to acquire the definite and indefinite articles using a variety of different tasks (e.g., Parrish, 1987; Thomas, 1989; Chaudron & Parker, 1990). However, a number of recent studies have tried to zero in on specific areas of difficulty learners have with articles. Ionin and colleagues (2004, 2008) have found that L2 learners fluctuate between the semantic features [±definite] and [±specific] in article choice by comparing them with native English controls. Learners failed to correctly identify which article is appropriate for the discourse context when presented with short written dialogues in the form of a forced choice elicitation task. In a collection of studies featured in García-Mayo & Hawkins (2009), there is a range of phenomena investigated including article choice, the prosodic representation of articles, generic reference, article production, and the online processing of articles. It is only up until recently that more attention has been given to existential constructions and the definiteness effect. Studies by White (2003) and Lardiere (2004), for instance, examined within their datasets spoken production that includes the use of there existentials. White (2003) examined a Turkish native speaker of English by collecting spontaneous production data through a series of interviews. Lardiere (2004) studied a L1 Chinese L2 English learner who had reached a steady state in English. Both studies reported no violation of the DE. White (2008a) looked at the DE in Turkish speaking and Mandarin Chinese speaking English L2 learners by way of using an elicited production task (see also White, 2008b). The participants were instructed to tell a story by looking at a sequence of pictures. Their English proficiency was intermediate. The study revealed no DE violations. King,

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

433

Steinhauer and White (2006) conducted an ERP study, which investigated brain function in L2 acquisition. King et al. (2006) asked Chinese and French participants to provide grammaticality judgments on sentences testing the English DE. The results show that only high intermediate Chinese L2 learners distinguished ungrammatical sentences from grammatical sentences. The low intermediate Chinese L2 learners failed to make the distinction between grammatical (e.g., I hope that there is a light in the closet) and ungrammatical sentences (e.g., *I hope there is the light in the closet) as both sentence types were accepted. The authors found that generally the more proficient Chinese and French learners showed P600 effects, which are generally associated with Ǯnative-likeǯ grammaticality judgements. White et al. (2012) addressed issues related to previous studies as they pointed out four problems: (i)

Past findings were dependent on spontaneous production data and as a result there were rather infrequent productions of there constructions. There may not have been many contexts where a there construction was required, thus spontaneous spoken production may not be the most appropriate method to look for DE violations. (ii) Grammaticality judgment tasks without contexts are unlikely to illicit appropriate answers since there are other uses of there such as deictic, list, presentation and task readings (see section 2 above for examples). In other words, existential there + indefinite DP is appropriate given the right context. If the context is not clearly an existential one then learners could judge all there type constructions as grammatical since they are exposed to input with all types of there readings. (iii) The role of L1 transfer must be accounted for since languages without articles may have other determiners (see Table 1 above) that feature in either weak or strong DPs. Any judgments on affirmative existentials in L2 English may be the direct result of L1 transfer effects. In order to eliminate L1 transfer, White et al. (2012) included negative existential contexts since learners are only ever likely to receive error-driven acquisition from affirmative existentials. “If L2ers prove not to have problems with negative existentials when the L1 and L2 differ, this suggests that transfer alone cannot be responsible for L2ers’ success” (White et al., 2012; p. 63). (iv) Learners may be sensitive to frequency of there existentials in the input and/or the effects of instruction in the L2 classroom.

434

Chapter Thirteen

Therefore, it is necessary to administer a task where learners have to make judgments not just on there existentials but a whole range of there sentences that include deictic, list, presentation and task readings plus negative existentials (e.g., There isn’t a man in the room). This way, if learners had learnt a rule that there must follow an indefinite DP, i.e., There is a …., other there readings would reveal if this is the case, or whether learners can acquire all types of there sentences. Both Russian and Turkish display a DE with determiners other than articles, but only in positive existentials, not in negative existentials. The study by White et al. (2012) found that advanced Russian and Turkish learners of English could perform just as well as native speakers on judgments of definiteness effects. Thus, the authors argue that the participantsǯ ability to judge English sentences appropriately cannot be due to L1 transfer. Rather, the authors argue that L2 learners have access to universal semantics and are able to make judgments on subtle restrictions of definiteness in the L2.

5. Our Study Our main research question is whether Japanese L2 learners respect the DE in English existentials. We used the same test sentences as White et al. (2012) to investigate whether Japanese learners could detect ungrammatical sentences (DE violations) and accept grammatical existential constructions. The motivation for replicating White et al.’s (2012) study was to find out if Japanese L2 learners are able to recognize the DE restriction given that Japanese allows both weak and strong DPs with the ga ir-u / ga ar-u constructions in affirmative and negative existentials. Crucially, Japanese differs from Russian and Turkish as Japanese allows strong DPs in affirmative sentences for locative existentials as specified in (23) above; only when the sentence is SOV, a strong DP is ungrammatical. Japanese is the same as Russian and Turkish with respect to negative sentences. We hypothesize that the L2 learners will perform i.) well in judging affirmative existentials as Japanese has the DE and similar morphology to English articles, such as sono (the/that), albeit only with SOV sentences, ii.) poorly in the negative DE since Japanese does not have the DE in negative sentences.

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

435

5.1 Participants Six native speakers of English and 10 L1 Japanese L2 English learners participated in the study. We used the recent Test of English for International Communication (TOEIC) scores to place the learners into two proficiency groups. The highest score one can achieve on the TOEIC test is 990. The average TOEIC score amongst our participants is 842.5. The range is 780 to 965. Five participants scored in the range of 840-965 and were therefore classified as advanced. The remaining five participants were classified as upper-intermediate learners with scores in the range of 780-805.

5.2 Task The task was created in the form of a written acceptability judgment task. Each test sentence was preceded by a short context. As in White et al. (2012), participants were instructed to read the context and then to decide whether the sentence that appeared beneath it was natural or unnatural in the context. In other words, they were not asked to provide a grammaticality judgment for the sentence by itself: although all ungrammatical items in the task are either unacceptable or infelicitous in the given context, many of them were in fact acceptable in other contexts. If participants judged a sentence to be unnatural, they were asked to provide a correction. The test items did not appear prior in the contexts so as to avoid any priming effects. The test sentences, created by White et al. (2012), also avoid a repeat bias of the definite article. For example, if the context mentions the doctor as in (26), and the test sentence asks participants to make a judgment on There’s the doctor here already, some learners may judge the test sentence to be acceptable since the doctor, i.e., Dr. Salter, is mentioned in both context and test sentence. However, if they reject such examples we can rule out it is due to presupposition failure. In other words, if they reject such sentences then there must be some other reason for their judgment other than presupposition. (26)

Unnatural affirmative existential item (strong) Anne is feeling sick, so she makes an appointment to see Dr. Salter. She arrives early and the nurse tells her to go right in, saying: There’s the doctor here already.

Two sentences from each of the 18-sentence types were used so the task had 36 items in total for this small-scale study. Students were asked to

436

Chapter Thirteen

participate in our study during a busy semester of classes, thus fewer items were included in our version of the task due to time restrictions. 8 are grammatical types, 10 ungrammatical types, including 6 control sentences. Some grammatical and ungrammatical examples of the types of contexts featured in the task are provided in examples (27-32) below. Examples (27-30) are some of the test items and examples (31-32) features some of the control items in the AJT. (27)

Grammatical positive indefinite article Lisa and Denis were moving into a new apartment. Lisa was worried about how to get their furniture upstairs, because the elevator looked too small. Denis said: Don’t worry, there’s also a big elevator in this building.

(28)

Ungrammatical positive definite article Tom has to make copies of a report but he can't find a photocopier. He asks the secretary what to do. She says: There’s the most reliable copy machine downstairs.

(29)

Grammatical negative indefinite article Nick and Sarah were looking for an apartment. They found one that Nick liked, but Sarah was not sure. She said: There isn’t a very nice view from the living room.

(30)

Ungrammatical negative definite article John was expecting a present from his mother but it had not arrived. He called the post office to find out where it was. The agent checked, and then told him: There isn’t the package here yet.

The test items are a mix of grammatical and ungrammatical sentences. The positive items in (27) and (28) test whether L2 learners are sensitive to the DE in situations where strong and weak determiners behave similarly in the L1 and L2, despite differences in Japanese between SOV and LSV type sentences. The negative items in (29) and (30) test whether L2 learners are able to detect differences between Japanese and English with respect to the DE since Japanese has no DE. Japanese L2 learners are predicted to perform worse on the negative existential sentences because of L1 transfer effects.

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

(31)

(32)

437

Grammatical control indefinite subject The teacher needs a red pen. She asks a student to bring her one. The student comes back and says that she couldn’t find any pens. The teacher says: Never mind. A pencil would be fine instead. Ungrammatical control indefinite subject Bill likes to cook, but he has not been able to do so recently. He tells his friend: A kitchen stove is broken.

The control items appear in non-existential sentences. (31) is a grammatical sentence. In the context there is no presupposition of a particular pencil, thus the indefinite article is appropriate, referring to a non-specific pencil. (32) is ungrammatical because there is the presupposition that if Bill likes to cook there must be a stove in his kitchen, so only the definite article is appropriate, e.g., The kitchen stove is broken. The idea of having the control items was to check whether L2 learners could judge indefinites appropriately in non-existential sentences.

5.3 Procedure Our version of White et al.’s (2012) task did not appear on a computer screen but instead the participants were asked to complete a paper-andpencil task. The participants were encouraged to write any comments (in English or Japanese) on the task itself if they felt the continuing sentence was strange. We hoped that by asking for comments we could see if participants could express explicitly why a sentence was odd within the presented context. For example, a participant read the sentence ‘There are most of my books in my office already’, and she circled ‘unnatural’ but she did not know how to write a correction in English, so she wrote the following in Japanese (translated into English) ‘I thought it sounds strange because most is after there are in this sentence’. There was no time limit, but participants were encouraged to read each context and then move on to the next one without going back to the previous one.

5.4 Results The results in Figure 1 of the native English control speakers show that there is a high level of acceptance (above 95.4%) for the three grammatical categories (control, positive and negative). For the Japanese L2 learners, Figure 1 indicates the following two things: Firstly, the

438

Chapter Thirteen

advanced learners’ and the intermediate learners’ judgments are highly accurate for grammatical sentences. Secondly, judgments of the ungrammatical sentences shown in Figure 2 reveal noteworthy differences between the three groups. The advanced learners were able to reject ungrammatical sentences better than the intermediate learners even though the rate was far from a native-like level.

Figure 1. Percentages of acceptance for grammatical (natural) sentences.

Figure 2. Percentages of rejection for ungrammatical (unnatural) sentences.



Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

439

Figure 2 reveals that the native control speakers correctly rejected ungrammatical existential sentences over 87.5% of the time. The most difficult items for both the L2 learner proficiency groups was the ungrammatical positive category, e.g., There is the … All learners failed to strongly reject the strong definite DP in existential contexts: The advanced group rejected them 33% of the time and the intermediate group 12% of the time, respectively. Rejection of the ungrammatical control and negative types were higher than the positive category with the advanced group successfully rejecting both categories over 50%. The intermediate group only managed to be more successful in their rejection of the ungrammatical control category (see (32) above). Figure 3 shows the two proficiency groups and indicates the easiest and the most difficult categories for all the participants. The grammatical category proved to be easier to judge compared with the ungrammatical category.

Figure 3. Correct answers (%) to items by L2 proficiency.

The three figures of results do not provide the complete picture of the two learner groups. Other factors such as each individual’s AJT score and length of stay in an English speaking country may reveal some between and / or within group differences.

Chapter Thirteen

440

5.4.1 Relationship between Proficiency Level, AJT Scores and Length of Stay Most of our participants had studied in a program that allowed them to spend a year at a university as an exchange student, typically somewhere in the U.S. Given that high exposure to input may play a role in how accurate learners are at judging grammatical and ungrammatical existentials, we looked at their TOEIC test scores and compared them with their total scores on the AJT (number of correct responses) and their time spent in English-speaking countries to see whether there was any correlation between high proficiency in English, high scores on the task and length of stay. The comparisons are shown in Table 3.4 Participant ID

TOEIC

Total score Length of stay (no. of (out of 36 items months) on the AJT) J1 965 27 60 J2 920 30 9 J3 900 19 11 J4 865 16 12 J5 840 17 36 J6 805 13 0 J7 785 23 11 J8 785 15 12 J9 780 17 less than 1 month J10 780 17 12 Table 3. TOEIC, total AJT score and length of stay. As Table 3 shows, no participant managed to complete the AJT without any incorrect judgments. Clearly, the level of proficiency is an important factor in being able to make a distinction between grammatically appropriate existentials and ungrammatical ones. However, participants who spent a lot of time in an English-speaking country did not perform better than those participants who had shorter experiences of living 4

One anonymous reviewer commented that Age of Acquisition may be a relevant factor in how well the participants performed on the AJT. This is unlikely as almost all the participants started acquiring English at the same age – around 12 years old.

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

441

abroad. For example, J1 has the highest TOEIC score (965) and has spent the most time in an English-speaking environment (60 months) but does not have the highest total score for correct judgments on the AJT (27) when compared with J2 (a score of 30) who has only spent 9 months abroad. There is a weak correlation between total score on the AJT and length of stay (r= 0.418). Furthermore, J5 has a high TOEIC score (840) and has spent considerable time abroad (36 months) but only scored 17 on the AJT.5 To sum up, the findings of our study show that participants’ level of proficiency in the L2 is an important factor in being able to distinguish between grammatical and ungrammatical items on the AJT.

6. Discussion and Conclusion To summarize, we found that Japanese L2 learners of English at advanced levels of proficiency are able to differentiate between grammatical and ungrammatical affirmative and negative existential sentences. The intermediate learners can correctly identify grammatical existentials but cannot detect the ungrammaticality of strong DPs in existentials. The native English control group performed as expected in that they accepted grammatical sentences in existential contexts and rejected ungrammatical ones. Our results are very similar to those found in the study by White et al. (2012) where the proficiency level in the L2 matters. It is possible to gain some understanding of how Japanese L2 learners judge the DE in English from these findings. For the grammatical sentences, the intermediate group performed as accurately as the advanced and native speaker control groups. This accuracy on grammatical sentences may seem somewhat surprising as the intermediate learners seem to be able to distinguish between sentences in the AJT that are existential and those that are not (control, list, deictic and other grammatical items). In other words, there is not a general preference to accept all indefinite weak DPs for all types of there constructions and reject definite strong DPs otherwise we would find much greater differences between each of the categories illustrated in Figure 3 above. But, one must be careful in how one interprets the findings: the low rejection rate of ungrammatical sentences by the intermediate group might reflect a general over acceptance of any construction regardless of whether there is a definite or indefinite article present. At intermediate level, learners cannot clearly discriminate between grammatical and ungrammatical sentences in contexts where 5

Unfortunately, we do not have details about the amount of exposure to input the learners received whilst living abroad.

442

Chapter Thirteen

either There is a … or There is the … Intermediate learners might be more influenced by their L1 when compared with the advanced learners as the intermediate group accepted almost all the ungrammatical sentences in the task. These types of sentences are grammatical in Japanese. The fact that the advanced learners performed better than the intermediate learners overall may suggest that sensitivity to the English DE has something to do with the learner’s English proficiency level. The presence of the DE in Japanese is likely to aid acquisition of the affirmative existentials in L2 English. However, the absence of articles in Japanese means that learners have to work out that the DE is only expressed via weak expressions like indefinite a, not definite the. Furthermore, negative existentials in Japanese can be expressed by weak of strong expressions, but in English only by weak expressions. As our findings demonstrate, it is possible for advanced learners to accept weak expressions in L2 English and reject strong expressions in both affirmative and negative existential sentences. It is important to point out the limitations of our study. This was a small-scale study with relatively small groups of participants. As a result, we were unable to conduct any statistical analyses on the results. Furthermore, we did not include all the items in our version of the AJT as we created a shortened version of the task due to time constraints. It is highly possible that the complete version of White et al.’s (2012) task could yield different results to ours because Japanese learners would have to provide even more judgments across a wider variety of sentence types. Furthermore, we did not test the different orders SOV and LSV systematically in the task. In order to gain a more complete picture of the L2 learners’ interlanguage grammars of how the DE plays out, we would need to include affirmative SOV and LSV sentences since the AJT largely consisted of LSV sentences. Another important difference to note between White et al. (2012) and our study is that we administered the AJT as a paper-and-pencil task rather than on a computer screen. Our method may have inadvertently led to a more careful analysis of the contexts and test sentences even though we instructed the participants not to linger over each of the contexts. In other words, there were potential opportunities for learners to give more careful consideration to each of the contexts when presented to them as a written task rather than via a computer where they had to read individual contexts on a screen.6 6

One anonymous reviewer suggests that it may be more natural to have the participants listen to the test items rather than read them as they would sound more natural than the L2 learners reading the test items themselves and placing their own intonation while parsing the sentences. For the L2 learners, the test items may

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

443

To conclude, what we find is that level of proficiency in the L2 is an important factor. The advanced learners performed better than the intermediate learners, which suggests that sensitiveness to the English DE has something to do with the learner’s level of English; the higher the proficiency level the more likely one is able to detect ungrammaticality. With regards to acquiring an article system, it seems that the acquisition of the DE is somewhat easier in comparison with the acquisition of semantic features related to articles (i.e., definiteness and specificity). It may be that there is a difference in the acquisition process between these two phenomena which needs to be explored further.

Acknowledgements We wish to thank Lydia White, Tanja Kupisch, Alyona Belikova, Paul Hagstrom and Öner Özçelik for sharing the AJT stimuli with us. Furthermore, we are grateful to Keiko Yoshimura and Hironobu Hosoi for helpful comments and suggestions regarding how definiteness effects are borne out in Japanese ga ar-u / ga ir-u constructions. Any errors that remain are solely our own.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. Journal of Pragmatics, 19.1: 39-55. Al-Hamad, Mohammad, Eidah Al-Malki, Gabriela Casillas, Florencia Franceschina, Roger Hawkins, Jim Hawthorne, Daniela Karadzovska, Kazue Kato, Sarah A. Liszka, Cristóbal Lozano, Shiro Ojima, Natsumi Okuwaki & Emma Thomas. 2002. Interpretation of English tense morphology by advanced L2 speakers. In Susan H. Foster-Cohen, Tanja Ruthenberg & Marie Louise Poschen (eds.). EUROSLA Yearbook 2, 49-69. Amsterdam: John Benjamins. Butler, Yuko. 2002. Second language learners’ theories on the use of English articles: An analysis of the metalinguistic knowledge used by

sound more unnatural because the intonation they place on the test items is nontarget like. We agree that reading rather than listening may play a role in the L2 learners’ judgements, but clearly from White et al.’s (2012), Zielke’s (this volume) and our results all show that the more proficient the L2 learner is, the more likely she/he can make a target-like judgement. Thus, it may not necessarily be the task modality, but rather level of proficiency which is important.

444

Chapter Thirteen

Japanese students in acquiring the English article system. Studies in Second Language Acquisition, 24: 451-481. Chaudron, Craig & Kate Parker. 1990. Discourse markedness and structural markedness: the acquisition of English noun phrases. Studies in Second Language Acquisition, 12: 43-64. García-Mayo, María P. & Roger Hawkins (eds). 2009. Second language acquisition of articles: Empirical findings and theoretical implications. Amsterdam: John Benjamins. Goad, Heather & Lydia White. 2004. Ultimate attainment of L2 inflections: Effects of L1 prosodic structure. In Susan Foster-Cohen, Mitsuhiko Ota, Michael Sharwood Smith & Antonella Sorace (eds.). EUROSLA Yearbook 4, 119-145. Amsterdam: John Benjamins. Goad, Heather, Lydia White & Jeffrey Steele. 2003. Missing inflection in L2 acquisition: Defective syntax or L1-constrained prosodic representations?. Canadian Journal of Linguistics, 48: 243-263. Huebner, Thom. 1985. System and variability in interlanguage syntax. Language Learning, 35: 141-163. Inagaki, Shunji & David Barner. 2009. Countability in absence of count syntax: Evidence from Japanese quantity judgments. In Shunji Inagaki, Makiko Hirakawa, Yahiro Hirakawa, Setsuko Arita, Hiromi Morikawa, Mineharu Nakayama, Hidetoshi Sirai & Jessica Tsubakita (eds.). Papers from the Eighth Annual Conference of the Japanese Society for Language Sciences (Studies in Language Sciences 8), 111125. Tokyo: Kurosio Publishers. Inagaki, Shunji. 2014. Syntax–semantics mappings as a source of difficulty in Japanese speakers’ acquisition of the mass–count distinction in English. Bilingualism: Language and Cognition, 17 (3): 464-477. Ionin, Tania, Heejeong Ko & Kenneth Wexler. 2004. Article semantics in L2 acquisition: The role of specificity. Language Acquisition, 12(1): 369. Ionin, Tania, Maria L. Zubizarreta & Salvador B. Maldonado. 2008. Sources of linguistic knowledge in the second language acquisition of English articles, Lingua, 118: 554-576. Kageyama, Taro & Hideki Kishimoto. 2011. Nichieitaisyo: Meishinoimito-koubun [Comparison of Japanese and English: Meaning and Construction of Nouns] Tokyo: Taishukan-Shoten. Kearnes, Kate. 2011. Semantics. Palgrave Modern Linguistics Series. Palgrave Macmillan. King, Erin, Karsten Steinhauer & Lydia White. 2006. The definiteness effect in L2 acquisition: What can event-related brain potentials tell

Japanese Speakers’ L2 Acquisition of the English Definiteness Effect

445

us? Paper presented at Generative Approaches to Second Language Acquisition 8, Banff, April 2006. Lardiere, Donna. 2004. Knowledge of definiteness despite variable article omission in second language acquisition. In Alejna Brugos, Linnea Micciulla & Christine E. Smith (eds.). Proceedings of the 28th Annual Boston University Conference on Language Development, 328-339. Somerville, MA: Cascadilla Press. —. 2009. Some thoughts on a contrastive analysis of features in second language acquisition. Second Language Research, 25(2): 173-227. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Milsark, Gary L. 1974. Existential Sentences in English, Doctoral dissertation, MIT. —. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3: 1-29. Parrish, Betsy. 1987. A new look at methodologies in the study of article acquisition for learners of ESL. Language Learning, 37: 361-83. Prévost, Philippe & Lydia White. 2000. Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research, 16: 103-133. Rando, Emily & Donna J. Napoli. 1978. Definites in there-sentences. Language, 54: 300-313. Robertson, Daniel. 2000. Variability in the use of the English article system by Chinese learners of English. Second Language Research, 16: 135-172. Safir, Ken. 1987. What explains the definiteness effect? In Eric J. Reuland & Alice G. B. Ter Meulen (eds.). The representation of (in)definiteness, 71-97. Cambridge, MA: M.I.T. Press. Shirahata, Tomohiko. 1995. The Japanese learner’s long-term attainment in English articles. JACET Bulletin, 26: 113-130. Slabakova, Roumyana. 2008. Meaning in the second language. Berlin: Mouton de Gruyter.͒ Slabakova, Roumyana, Tania L. Leal & Judith Liskin-Gasparro. 2014. We have moved on: current concepts and positions in generative SLA. Applied Linguistics. Snape, Neal. 2008. Resetting the Nominal Mapping Parameter in L2 English: Definite article use and the count–mass distinction. Bilingualism: Language and Cognition, 11: 63-79. —. 2013. Japanese and Spanish adult learners of English: L2 acquisition of generic reference. Studies in Language Sciences: Journal of the

446

Chapter Thirteen

Japanese Society for Language Sciences, 12, 70-94, Kaitakusha, Tokyo, Japan. Snape, Neal, María. P. García-Mayo & Ayúe Gürel. 2013. L1 transfer in article selection for generic reference by Spanish, Turkish and Japanese L2 learners. International Journal of English Studies, 13(1): 1-28. Thomas, Margret. 1989. The acquisition of English articles by first- and second-language learners. Applied Psycholinguistics, 10: 335-355. Wakabayashi, Shigenori. 1998. Systematicity in the use of the definite article by Japanese learners of English. Bulletin of Gunma Prefectural Women’s University, 19: 91-107. White, Lydia. 2003. Fossilization in steady state L2 grammars: persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6: 129-141. —. 2008a. Definiteness effects in the L2 English of Mandarin and Turkish speakers. In Harvey Chan, Heather Jacob & Enkeleida Kapia (eds.). Proceedings of BUCLD 32, 550-561. Somerville, MA: Cascadilla Press. —. 2008b. Different? Yes. Fundamentally? No. Definiteness Effects in the L2 English of Mandarin Speakers. In Proceedings of the 9th Generative Approaches to Second Language Acquisition Conference (GASLA 2007), (eds.). Roumyana Slabakova, Jason Rothman, Paula Kempchinsky & Elena Gavruseva, 251-261. Somerville, MA: Cascadilla Proceedings Project. White, Lydia, Alyona Belikova, Paul Hagstrom, Tanja Kupisch & Öner Özçelik. 2009. Restrictions on definiteness in L2 English. In J. Chandlee, M. Franchini, S. Lord, & G.M. Rheiner (eds.). Proceedings of the 33rd Annual Boston University Conference on Language Development, 622-633. Somerville, MA: Cascadilla Press. —. 2012. Restrictions on definiteness in second language acquisition. Affirmative and negative existentials in the L2 English of Turkish and Russian speakers. Linguistic Approaches to Bilingualism 2.1: 56-84. Zucchi, Alessandro. 1995. The ingredients of definiteness and the definiteness effects. Natural Language Semantics, 3: 33-78.

CHAPTER FOURTEEN THE ACQUISITION OF THE DEFINITENESS EFFECT IN THE L2 EUROPEAN SPANISH OF L1 GERMAN AND L1 TURKISH SPEAKERS MARINA ZIELKE

1. Introduction It is commonly assumed that existential constructions (ECs) exhibit a socalled Definiteness Effect according to which only indefinite, or, in Milsark’s (1974, 1977) terms, “weak” expressions such as the indefinite article or weak quantifiers such as some are allowed in proper existentials, whereas definite, or “strong”, expressions such as the definite article and quantifiers such as every, most, among others, are ruled out. This holds true for European Spanish: (1)

a.

b.

Hay un libro muy interesante en la biblioteca. there-is a book very interesting in the library ‘There’s a very interesting book in the library.’ *Hay el libro muy interesante en la biblioteca. there-is the book very interesting in the library ‘There’s the very interesting book in the library.’

While the indefinte article in (a) is fine, the definite one in (b) is not. This can be accounted for in pragmatic terms if we assume that the function of an EC is to introduce a new referent into the discourse (Abbott 1993: 44). Hence, a referent that has been introduced into the discourse before, or is otherwise familiar to or uniquely identifiable by the hearer, cannot present new information and is therefore ruled out. In this paper I investigate the acquisition of the Definitenes Effect (DE) in ECs in European Spanish by monolingual L1 speakers of German and Turkish.

Chapter Fourteen

448

Whereas the DE in European Spanish and German plays out largely the same, Turkish differs from European Spanish and German in that positive ECs show a DE (2), whereas negative ECs do not (3). In negative ECs, strong expressions such as the proper noun Ali are grammatical: (2)

*Dükkanda hala Ali var. store-LOC still Ali exist ‘There is still Ali at the store.’ (White et al. 2012: 58)

(3)

Dükkanda Ali yok. store-LOC Ali not-exist ‘There isn’t Ali at the store.’ (White Belikova, Hagstrom, Kupisch & Özçelik 2012: 58)

German and European Spanish (as well as English) do not license proper names or any other strong expressions in ECs: (4)

a. b.

*Es gibt Ali im Laden.1 *Hay Ali en la tienda. there-is Ali in the store ‘There is Ali at the store.’

According to Diesing’s (1992) Mapping Hypothesis existential readings arise whenever the subject is situated within the “nucelar scope” of a sentence, i.e. within VP, whereas subjects positioned higher than VP receive a “quantificational/referential”, i.e. non-existential, reading. In Turkish, existential closure only applies in positive ECs but is absent in negative ones2, hence there is no DE in Turkish in positive ECs. L1 speakers of Turkish acquiring European Spanish as an L2 thus have to understand that in Spanish both positive and negative ECs show a DE. German L1 speakers studying European Spanish as an L2, on the other hand, can resort to positive transfer from their L1.

1

In German, there are two existential expressions: es gibt and es ist/sind which behave very differently syntactically (no agreement vs. agreement with the subject) and semantically. See 2.2. 2 On the basis of Kelepir’s (2001) analysis, White et al. (2012) assume that yok ‘not-exist’ is a participle and remains low in the structure and that existential closure applies earlier, i.e. lower in the structure in negative than in positive ECs. However, it is unclear why yok should move higher than var as both are participles possibly generated within VP which move to a locative X° (Kelepir 2001).

The Acquisition of the Definiteness Effect in L2 European Spanish

449

Previous studies on the acquisition of the DE in ECs have focused on the L2 English of L1 speakers of Chinese, Russian and Turkish (White 2008, White et al. 2012) and Spanish (Palacio-Martínez & Martínez-Insua 2006). These studies showed that the DE is acquired rather ‘effortlessly’ even by L1 speakers of languages that do not have articles. On the basis of these results, I hypothesize that the DE in European Spanish, too, should be acquired rather ‘effortlessly’. Due to greater similiarities in how the DE plays out in Spanish and German, I further hypothesize that the L1 German group will outperform the L1 Turkish group due to positive L1 transfer. What has to be borne in mind is that for the speakers investigated in this study, European Spanish is not a proper L2 but an L3, as they have learnt at least English and, possibly, another language before European Spanish. Hence, it is possible that there is L2 transfer from their respective L2s. The present article, however, is not an L3 study in that it systematically tries to detect L1 and L2 transfer sources, but is meant to take a first look at the acquisition of the DE in the context of Ln acquisition and, importantly, in a language other than English. As such, it might provide valuable information about the DE and its acquisition in Romance. The paper is organized as follows: First, I will give a brief overview on the link between ECs, Diesing’s (1992) Mapping Hypothesis and the DE. Then I will show how the DE works in European Spanish, German and Turkish ECs. In section 3 the experimental design of the study will be presented. In Section 4 and 5 I will present and discuss the results. A conclusion as well as indications for further research will be given in Section 6.

2. Existential Constructions and the Definiteness Effect ECs serve to introduce a hearer-new referent into the discourse (Abbott 1993: 44). More often than not, a language has a special construction by which this can be achieved, usually consisting of a verb meaning ‘be’, ‘have’, or a special (invariant) form derived from one of these (Lyons 1967: 391). In English, the expression there is/there are carries out this function; in Spanish, this is done by means of the impersonal verb hay. The structures are non-canonical, i.e. they have a special structural makeup and are different from the corresponding copular sentences, hence it is obvious that they are assigned a special pragmatic function (McNally 2011: 1820). Crucially, the contexts in which the sentences in (5) can

Chapter Fourteen

450

occur differ considerably from the pragmatic contexts in which the sentences in (6) are licensed: (5)

a. b.

There's a spider in the basement. Hay una araña en el sótano. there-is a spider in the basement. ‘There's a spider in the basement.’

(6)

a. b.

A spider is in the basement. Una araña está en el sótano. a spider is in the basement ‘A spider is in the basement.’

While (5 a-b) can serve to start a conversation or a new subject matter within a conversation, the canonical sentences in (6 a-b) cannot introduce a new referent or situation into the discourse in an out-of-the-blue context: The ordering of the constituents in ECs is in line with Diesing’s (1992) Mapping Hypothesis according to which material within VP is assigned the “nuclear scope” of the sentence, i.e. a weak/existential reading, whereas material outside VP receives a strong, or referential/quantificational reading. Thus, in an EC the subject remains inside VP, whereas it moves to a higher position within the structure in a canonical sentence: (7)

[There [VP is a cat in the garden]]

(existential/weak reading)

(8)

[The cat [VP is in the garden]]

(referential/strong reading)

In the case of ECs versus ‘canonical’ sentences, the distinction between the two readings is also marked morphologically (a cat vs. the cat), at least in languages with articles. The tendency for definite and indefinte NPs to be in complementary distribution in existential and non-existential sentences supports the idea that there is a fixed position in the sentence structure where existential and non-existential readings are assigned. The need to move the subject to a position lower in the structure (and, in some languages, fill the canonical subject position by an expletive pronoun) in ECs can be neatly accounted for by Diesing’s Mapping Hypothesis: (9)

Existential: a. There is a cat in the garden. b. There is *the cat in the garden.

The Acquisition of the Definiteness Effect in L2 European Spanish

(10)

Non-existential: a. ?A cat is in the garden. b. The cat is in the garden.

451

(weak reading)

Milsark (1974, 1977) observed that there is a restriction as to what kind of expressions are allowed in ECs. Only indefinite, or “weak” expressions can be used, whereas definite, or “strong” expressions are ruled out. The fact that strong expressions are referential and that their existence is already established by means of their definiteness makes them incompatible with a structure introducing a hearer-new referent into the discourse (Abbott 1993: 44). Strong expressions are often considered to have a quantificational force denoting how many individuals of a given set confirm to the property denoted by the EC; i.e. there is a “quantification over individuals in their whole temporal extendedness” (Musan 1996: 167), e.g. most/all/no-one of the students. Weak expressions, on the other hand, are exempt from any quantificational meaning. Within Milsark's classification, not only articles are understood to be either definite or indefinite, but also demonstratives, pronouns, possessives, etc. Following Beaver, Francez & Levinson (2005) and Leonetti (2008), I analyze the DE in terms of markedness rather than as a categorical “ban” (Beaver et al. 2005: 22), assuming that definite NPs are not ruled out from ECs per se, but are significantly more marked than indefinite NPs and require a very specific pragmatic context in order to be felicitous. This is due to the fact that, cross-linguistically, the canonical subject position tends to be occupied by a topic constituent rather than a focus constituent. Topics tend to be definite as they have already been introduced into the discourse whereas focused elements are more likely to be indefinite and to appear at a lower position in the sentence (see Beaver et al. 2005). The DE can therefore be considered a natural consequence of the distribution of topicalized and focalized material. In line with the aforementioned authors, I assume a markedness scale for the different expressions: Unmarked - a - some, many, few, several - cardinal numbers - zero articles (with bare plurals, mass nouns)

Marked - the - all, most, every, each - demonstratives - possessives - proper names - pronouns Table 1. Marked and unmarked expressions in existential constructions (a modified version of White et al. 2008: 5)

Chapter Fourteen

452

English has many structures that formally look like ECs but whose interpretation is essentially different such as “list” readings (Rando & Napoli 1978), e.g. Look! There is Alex. Other sentences that formally look like ECs comprise “kind readings” (Musan 1996: 169), “extremely readings” (Lyons 1999: 247), deicitc readings and weak definites. This shows that (there) be is also used in other, non-existential constructions.

2.1 European Spanish European Spanish ECs are formed by the impersonal verb hay ‘there is’, which is an amalgamation of the 3rd person singular of the Latin verb HABERE ‘have’ and the locative adverb IBI ‘there’. They are subject to the DE: (11)

a.

Hay {un gato/algunos gatos} en el jardín there-is a cat/some cats/many cats in the garden ‘There is/are a cat/some cats/many cats in the garden.’ *Hay {el gato/todos los gatos} en el jardín there-is the cat/all (the)cats in the garden ‘There is/are the cat/every cat/all the cats in the garden.’

b.

The weak/strong distinction is very stable in European Spanish with weak definites being one of the few apparent ‘exceptions’ to the DE, although weak definites, of course, are only formally strong but semantically weak: (12)

No hay ni el más mínimo not there-is even the smallest ‘There isn’t the smallest error in his thesis.’

error error

en su tesina. in his thesis

. The fact that the constituent order of hay sentences is different from the order in non-existential sentences and that, again, the NP of an EC (13a) is positioned lower in the syntactic structure than in non-existential sentences (13b) is in accord with the idea that the order of the material plays a major role in the assignment of existential and non-existential readings: (13)

a.

Hay un hombre en la calle. there-is a man in the street ‘There is a man on the street.’ (existential/weak reading)

The Acquisition of the Definiteness Effect in L2 European Spanish

b.

453

Un/El hombre está en la calle. a /the man is in the street ‘A/The man is on the street.’ (non-existential reading)

The same is true for negative ECs: (14)

a.

b.

No hay un hombre en la calle. not there-is a man in the street ‘There is a man on the street.’ (existential/weak reading) Ningún hombre está en la calle. no man is in the street ‘No man is on the street.’ (non-existential reading)

Both in positive and negative ECs, the canonical subject position is empty and the subject remains in a Spec position lower than that, probably in Spec,VP (Rodríguez-Mondoñedo in press). In sum, both positive and negative ECs in European Spanish are subject to the DE and the default existential expression hay has an essentially existential force that is only marginally applicable to non-existential, ‘DE-related’, sentence types.

2.2 German German has three main structures by which ECs can be formed: es gibt, es sind as well as regular copular sentences: (15)

a.

b.

c.

Es gibt viele Wolkenkratzer in New York. there-are many skyscrapers in New York ‘There are many skyscrapers in New York.’ Es sind heute viele Leute auf der Straße. there-are today many people on the street ‘There are many people on the street today.’ In meinem Haus ist gerade eine schwarze Katze. in my house is right-now a black cat ‘There is a black cat at my house right now.’

Es gibt implies permanence (15a) and cannot be used to describe an accidental or temporary property such as in (15b). This is reflected by the fact that temporal adverbs such as gerade ‘right now’ are incompatible with es-gibt sentences. Instead, a canonical structure is needed whenever temporal adverbs are present (Czinglar 1999):

Chapter Fourteen

454

(16)

Es gibt (*gerade) einen Verrückten in unserem Haus. there-is right-now a lunatic in our house ‘There is a lunatic in our house right now.’ (Czinglar 1999: 58)

German thus displays a lexical restriction on ECs absent both in European Spanish and Turkish (and English, for that matter) but common to Scandinavian languages such as Danish, Norwegian and Swedish. Es gibt can only assert permanent existence (see Sawyer 1973 and Czinglar 1999 for a more detailed account). Besides the lexical restriction illustrated above, German, too, displays a DE in both positive (17a-b) and negative ECs (17c-d): (17)

a.

b.

c.

d.

*Es gibt hier alle grauen Katzen there-is here all grey cats Intended: ‘There are all grey cats.’ *Es ist sein Bruder heute in der Schule. there-is not his brother today in the school Intended: ‘There is his brother at school today.’ ??Es gibt hier nicht alle grauen Katzen. there-is here all grey cats Intended: ‘There isn’t all grey cats.’ *Es ist nicht sein Bruder in der Schule heute. there-is not his brother in the school today Intended: ‘There isn’t his brother at school today.’

Just as in Spanish, weak definites are licensed in ECs due to their underlying indefiniteness: (18)

Es gab nicht den geringsten Fehler in seiner Arbeit. there wasn’t the slightest error in his thesis ‘There wasn’t the slightest error in his thesis.’

German ECs, just like European Spanish ECs, are both subject to the DE and the DE applies both in positive and in negative ECs.

2.3 Turkish Turkish ECs are formed by means of the participles var ‘exist’, and yok ‘not exist’. Just as in German, English and European Spanish, ECs in Turkish may or may not contain a (locative) coda:

The Acquisition of the Definiteness Effect in L2 European Spanish

(19)

a.

b.

455

Sepet-te kedi var-d. basket-LOC cat exist-PAST ‘There was a cat in the basket.’ (Kelepir 2001: 175) Kedi var-d. cat exist-PAST ‘There was a cat.’

If there is a locative, it has to precede the pivot. Any other position renders the sentence ungrammatical3: (20)

a.

b. c.

Bahçe-de bir köpek var . garden-LOC one dog exist ‘There is a dog in the garden.’ *Bir köpek bahçe-de var. *Bahçe-de bir köpek var. (examples taken from Erguvanl-Taylan 1987: 217)

Specificity in Turkish is marked by an accusative marking on the noun. Proper Turkish ECs, however, do not display ACC-marking of the pivot. This stresses the fact that the pivot receives a non-specific interpretation since nouns that do not bear accusative case are always interpreted as nonspecific in this language (Enç 1991: 4). Also, the sentence-initial position is reserved for definite NPs; an indefinite NP can only follow a definite one in a sentence (Kuno 1971). This ties in with Diesing’s Mapping Hypothesis according to which indefinite material has to remain inside the VP. In fact, a non-presupposed NP is excluded from this position (unless it is a contrastive topic or bears generic reference) (Tura 1986): (21)

a.

?Bir kedi sepet-te-y-di-ø a cat basket-LOC-COPULA-PAST-3.sg ‘A cat was in the basket’ (Kelepir 2001: 176)

Turkish differs crucially from European Spanish and German in that positive existentials display a DE (22), whereas negative existentials (23) do not: 3

German, English and Spanish sentences whose first constituent is a locative are usually referred to as locatives. However, the fact that the locative must be in first position in Turkish does not challenge the fact that these sentences are existential as their reading is clearly an existential as opposed to a locative one.

456

Chapter Fourteen

(22)

*Dükkanda hala Ali var. store-LOC still Ali exist ‘There is still Ali at the store.’ (White et al. 2012: 58)

(23)

Dükkanda Ali yok. store-LOC Ali not-exist ‘There isn’t Ali at the store.’ (White et al. 2012: 58)

Thus, existential closure only seems to apply in positive but not in negative ECs, or the domain over which it applies is smaller so that the subject is not existentially bound. White et al. (2012) assume that this has to do with the position of negation and the fact that negated ECs, unlike in German and Spanish, are not formed by means of the regular form of sentential negation in Turkish. According to their analysis yok ‘not exist’ is a participle and remains within VP. Existential closure would thus be triggered earlier in negative ECs than in positive ones. Turkish, too, allows for some weak definites to occur in sentences that formally look like ECs. However, most of them do not formally include any strong expresssions as definiteness is not morphologically marked in articles in Turkish. Hence, they do not represent any apparent ‘exception’ to the DE as is the case with European Spanish and German. Instead, they are formally identical to regular indefinites expressing ‘any’: (24)

Hicbir fikrim yok. Slightest idea mine not-exist Reading 1: ‘I don’t have the slightest idea/There isn’t the slightest idea.’ Reading 2: ‘I don’t have any idea/There isn’t any idea.’ (Zielke 2013: 206)

In summary, Turkish differs from both European Spanish and German in that only positive ECs show a DE, whereas negative ones do not. Table 1 sums up how the three languages behave with respect to the DE: Language Positive Existentials Negative Existentials European Spanish DE DE German DE DE Turkish DE no DE Table 2. Overview of the DE in European Spanish, German and Turkish.

The Acquisition of the Definiteness Effect in L2 European Spanish

457

3. Previous Research and Hypotheses So far, there has only been a handful of studies investigating the acquisition of the DE in ECs within L2 acquisition. All of these studies have focused on the acquisition of English as an L2 (Palacio-Martínez & Martínez-Insua 2006, White 2008, White et al. 2012). Although the acquisition of an L2 article system has been reported to be rather problematic, especially if a speaker’s L1 does not have articles, or does not morphologically mark definiteness (Thomas 1989, Robertson 2000, among others), studies on the acquisition of the DE clearly contradict the view that the acquisition of an article system and/or definiteness is problematic per se. White (2008) and White et al. (2012) have shown that both intermediate as well as advanced L1 speakers of Mandarin, Russian and Turkish acquired the DE of English ECs rather ‘effortlessly’ although they had never had any explicit training on how English ECs work. On the basis of these results, I hypothesize (i) (ii)

that the DE in European Spanish, too, should be acquired rather ‘effortlessly’; and that the L1 German group will outperform the L1 Turkish group due to positive L1 transfer as the DE plays out largely the same in European Spanish and German.

4. The Study This study investigates the acqusition of the DE in European Spanish ECs by 14 L1 speakers of German and 10 L1 speakers of Turkish. The data used for this study have been obtained by means of an acceptability judgement task (AJT) as well as by a Spanish cloze test to obtain an idea of the overall proficiency level of the subjects.

4.1 Participants and Methodology Two groups are included: (i) 14 L1 German speakers and (ii) 10 L1 Turkish speakers acquiring Spanish. Both groups were tested in their respective home countries, i.e. Germany and Turkey. L1 German speakers

458

Chapter Fourteen

All 14 speakers are monolingual, although some of them speak Low German or Bavarian dialects4. The majority of the subjects were students of the University of Hamburg and the Ruhr University of Bochum in Germany. Their age was between 23 and 31 years (mean: 25;1). All participants had spent time in Spain ranging from 7 days to a year and a month. The speakers differed as to how long they had been studying Spanish for, ranging from 2 to 15 years. Strictly speaking, Spanish was the speakers’ L3 as English is a compulsory subject from the 5th grade on in Germany. L1 Turkish Speakers All 10 speakers were monolingual speakers of Turkish, although some of them reported to speak additional varieties of Turkish mostly from the Black Sea region. All were tested at Bo÷aziçi University in Istanbul. Their ages were between 22 and 35 years (mean: 25;8). Whereas some of the participants had spent time in Spain, others did not. The speakers also differed as to how long they had been learning Spanish, ranging from 1 year to 10 years. Overall, these speakers had been learning Spanish for a shorter period of time than the L1 German group. As with the L1 German speakers, all learners of Spanish were studying Spanish as an L3, not as an L2, as it is common to first learn English at school in Turkey.

Age Age of Acquisition Time spent in Spain (years; months; days) Time Spanish has been studied for

L1 German (n=14) 23 – 31 yrs 14 – 21yrs 0;0;7 – 1;1

L1 Turkish (n=10) 22 – 35 yrs 17 – 25 yrs 0;0 – 2;0

2 – 15 yrs

1 – 10 yrs

Table 3. Background information on the L2 groups. One of the factors that can have an influence on L2/L3 acquisition is the speakers’ proficiency level (Falk & Bardel 2010: 196ff.). The speakers’ overall proficiency level was tested by means of a Spanish cloze test. The scores of the cloze test ranged from between 13 and 36 (mean: 24) out of a possible 44 points for the L1 German group and from between 11.5 and 29 (mean: 20.4) for the L1 Turkish group. L2ers who scored less than 22, 4

One L1 German subject reported to have received Korean input when she was a child. Since she claimed to have forgotten how to speak this language and considered herself to be a native speaker of German only, she has been included in the analysis.

The Acquisition of the Definiteness Effect in L2 European Spanish

459

which is 50% of what was possible, were classified as Intermediate. The proficiency level of those who scored between 22 to 44 was categorized as Advanced. The proficiency level of 4 of the L1 German speakers was classified as intermediate, 10 learners of this group were classified as advanced. In the L1 Turkish group there were 4 intermediate and 4 advanced learners of Spanish (2 speakers of the L1 Turkish group were not included in the analysis (see next section) so that the final number of Turkish speakers was 8): L1 German (n=14) L1 Turkish (n=8) - 4 intermediate (mean score: - 4 intermediate (mean score: 16.75) 14.25) - 10 advanced (mean score: 24.25) - 4 advanced (mean score: 26.5) Table 4. Number and mean scores of intermediate and advanced speakers.

4.2 Acceptability Judgment Task The speakers’ performance on the DE in Spanish was tested by means of an AJT consisting of a total of 72 items. The test was bi-modal, i.e. the items were presented both in written and in audio form (read by a male native speaker of Colombian Spanish)5. The L2 learners were given between 15 and 20 seconds for the evaluation of the items and were asked to state whether they thought the item was either correct/natural or incorrect/unnatural within the given context. In all cases a sufficient amount of context had been provided. Then, according to their judgment they had to either repeat the item when they deemed it correct/natural or correct the item when they considered it to sound incorrect/unnatural. Their responses were recorded by means of an M-Audio MicroTrack II audio device. The final test design included the following test items:

5 The task was based on White et al.’s (2012) test design. The items were roughly balanced for grammaticality.

Chapter Fourteen

460

1

No. 5

2

5

3

5

4

5

5

5

6

5

7

5

Item Grammatical Positive ECs: Hay + indefinite articles (un, una, etc.) Grammatical Positive ECs: Hay + weak quantifiers (algunos, muchos, etc.) Grammatical Positive ECs: Hay + weak quantifiers (ningunos, etc.) Ungrammatical Positive ECs: Hay + possessives (mi, tu, su) Ungrammatical Negative ECs: Hay + possessives (mi, tu, su) Ungrammatical Positive ECs: Hay + proper names (Carmen, Daniel, etc.) Ungrammatical Negative ECs: Hay + proper names (Carmen, Daniel, etc.)

Table 5. Test items testing (un)grammatical positive and negative existentials. The individual items comprised a short context followed by a highlighted sentence that was to be judged. An example is given in (25): (25)

Context: Carmen se olvidó de las llaves en el trabajo. Llamó a su secretaria. Ella le dijo: Test sentence: No hay tus llaves aquí. ‘Carmen had left her keys at work. She called the secretary who told her: There aren’t your keys here.‘

In addition, there were 37 control items. Expressions that are formally definite but semantically indefinite such as el menor problema ‘the slightest problem’ etc. belong to the class of weak expressions although formally they look like strong ones. 5 items testing whether the speakers were able to identify semantic indefiniteness in proper ECs, even if it is formally definite, were included. The test also included 2 items controlling for the L2ers’ treatment of extremely readings, i.e. hay sentences that are formally strong but semantically weak, such as English There was the most beautiful flower in the garden:

The Acquisition of the Definiteness Effect in L2 European Spanish

8

No. 5

9

2

461

Item Grammatical Negative ECs: Hay + weak definites Grammatical Positive ECs: Hay + definite article in extremely readings

Table 6. Control items testing grammatical hay-sentences with weak definites. These items are especially important as they show whether definites are rejected categorically or not. Other control items tested the general use of the (in)definite article to see whether possible problems with the weak/strong distinction resulted from difficulties with the Spanish article system as such. Also, the general use of the verb estar was controlled for as this is the verb that can usually be used whenever hay is not possible due to the definiteness of an expression. Other control items included deictic sentences including estar investigating whether the L2ers have acquired the fact that hay is purely existential and can never be used in a deictic sentence, whereas estar denotes presence, not existence. Also, there were 10 more items testing the L2ers’ acceptance of hay in contextualized/list reading sentences:

10

No. 5

11

5

12 13 14

5 5 5

15

5

Control items Grammatical sentences testing the general use of the definite article Ungrammatical sentences testing the general use of the indefinite article Grammatical sentences testing the general use of estar Grammatical deictic sentences with estar Grammatical sentences testing contextualized/list readings with estar Ungrammatical contextualized/list readings with hay

Table 7. Control items testing (un)grammatical contextualized/list readings. Within my analysis, only those speakers who performed about 80% accurately on the following control items were included in the analysis: (i) L2ers that have understood the overall concept of (in)definitenes. Only if the use of the (in)definite article is correct, the judgments on the test items become reliable – otherwise the speakers might just be generally unsure about the use of articles in any construction, existential or not, as well as about the general concept of (in)definiteness;

462

Chapter Fourteen

(ii) L2ers who have acquired the general use of the verb estar. Speakers have to have understood that estar is a copular verb and can occur with strong expressions, i.e. precisely with those expressions that cannot occur with hay. Only if they master this verb, they can actually use it to render an ungrammatical hay sentence grammatical (*Hoy no hay Julia en clase ‘Today there isn’t Julia in class’ would then be correctly changed to Hoy no está Julia en clase ‘Today Julia isn’t in class’). The exclusion of L2ers who did not perform at least 80% accurately on these control items makes sure that the difficulties they may have with the DE cannot be attributed to problems concerning aspects of article use unrelated to the DE. In total, 2 L1 Turkish speakers had to be counted out. Of the L1 German group no L2er had to be discarded from the analysis6. Corrections in which hay was changed for verbs other than estar were deemed relevant, even if the correction as such was ungrammatical: (26)

*Tu hermano es en el equipo. [L1 German speaker] your brother is in the team ‘Your brother is on the team.’

Althought the use of ser ‘be’ is ungrammatical here, this error can be attributed to difficulties with the distinction between the two copular verbs ser and estar. What is important, however, is that this L2er correctly rejected hay in this sentences since tu hermano ‘your brother’ is a strong expression and therefore incompatible with hay.

4.3 Results Figure 1 shows the overall acceptance of weak and strong expressions in proper positive and negative European Spanish ECs:

6

This is probably due to the fact that German and European Spanish are more closely related so that no problems arose for L1 German speakers as far as the use of articles and the copular verb estar was concerned. Turkish, on the other hand, lacks both phenomena. Also the overall level of proficiency of the L1 Turkish group was slightly lower than for the L1 German group.

The Acquisition of the Definiteness Effect in L2 European Spanish

100

463

Controls L1 German L1 Turkish

50

0 Pos + weak Neg + weak Pos + strong Neg + strong

Figure 1. Acceptance of weak and strong expressions in positive and negative ECs.

Figure 1 shows that both L1 groups correctly accepted almost all ECs – both positive and negative – that included weak expressions and thus coincide with the monolingual controls. The L1 German group accepted 92.9% of the positive ECs containing a weak expression and 88.6% of the negative ones. Similarly, the L1 Turkish group accepted 95% of the positive ECs including a weak expression and 97.5% of negative ECs. Although, on the whole, both groups tended to reject strong expressions in both positive and negative ECs, the L1 Turkish group accepted these items more readily: Whereas the L1 German group accepted strong expressions in positive ECs only in 12.6% and in negative ECs in 12.1%, the acceptance rate of the L1 Turkish group is more than twice as high: 28.8% of the positive ECs and 27.5% of the negative ECs containing a strong expression were deemed grammatical. Figure 2 shows whether a difference in the acceptance of strong expressions for the different proficiency levels of Spanish within the two groups could be found:

Chapter Fourteen

464

100 Controls L1 German Intermediate L1 German Advanced L1 Turkish Intermediate L1 Turkish Advanced

50

0 Pos + weak Neg + weak

Pos +

Neg +

strong

strong

Figure 2. Acceptance of weak and strong expressions in positive and negative ECs.

Within both groups the intermediate speakers accepted a higher number of strong expressions in ECs than the advanced speakers of the same L1: The intermediate L1 German speakers accepted strong expressions in about 20% in positive, and 25% in negative ECs, whereas the advanced L1 German speakers only accepted strong expressionsin about 11% in positive, and 7% in negative ECs. For the L1 Turkish speakers the differences in performance are even bigger: The intermediate L1 Turkish speakers accepted slightly more than 50% of strong expressions in positive as well as negative ECs. The advanced L1 Turkish speakers, on the other hand, although accepting some strong expressions (5%) in positive ECs, correctly rejected all strong expressions in negative ECs. The advanced L1 Turkish group thus constitutes the group that performed most target-like among all four. Figure 2 also shows that weak expressions were largely accepted by all groups. However, there are some minor differences between the intermediate and the advanced speakers of each of the two L1 groups: In both groups the advanced speakers rejected ECs containing weak expressions more often than the intermediate speakers, even though these items were perfectly grammatical. However, these rejections are only minor and most items were deemed grammatical by the two advanced groups. Figure 3 shows the distribution of strong expressions – namely possessives and proper names – in those cases in which they were accepted and shows whether there are differences in the acceptance rates for the different types of strong expressions within the four groups (the

The Acquisition of the Definiteness Effect in L2 European Spanish

465

control group is not included as this group accepted close to no proper names or possessives):

100

Proper names Possessives

50

0 L1 German Intermediate

L1 German Advanced

L1 Turkish Intermediate

L1 Turkish Advanced

Figure 3. Acceptance of possessives and proper names in positive ECs.

‘100’ refers to all ungrammatical items that were deemed grammatical by the L2ers. As can be seen in Figure 3, both L1 groups accepted more proper names than possessives in positive ECs. The advanced L1 Turkish group only accepted proper names in ECs. Figure 4 shows what the distribution of incorrectly accepted possessives and proper names looks like for negative ECs (again, the control group has been left out, as there are no instances in which proper names or possessives were accepted in negative ECs):

100

Proper names Possessives

50

0 L1 German Intermediate

L1 German Advanced

L1 Turkish Intermediate

L1 Turkish Advanced

Figure 4. Acceptance of possessives and proper names in negative ECs.

466

Chapter Fourteen

Figure 4 shows that more possessives than proper names have been deemed grammatical in negative ECs. The L1 Turkish group accepted possessives and proper names each about 50% of the time. The L1 German groups shows a preference for possessives. Whereas they accepted 35.3% of negative ECs including proper names, their acceptance of possessives was 64.7%. In order to determine whether the L2ers rejected strong expressions categorically, we have to take a look at their performance on weak definites in ECs and extremely readings. Weak definites and extremely readings are formally definite but semantically indefinite. Their semantic indefiniteness licenses them in proper ECs in European Spanish. In order for the DE to be fully acquired, these items have to be judged as grammatical, hence a correlation between the correct acceptance of hay in these control items on the one hand and accuracy on the weak/strong distincion on the other is necessary. Table 8 shows whether such a correlation is present:

L1 German intermediate L1 German advanced L1 Turkish itermediate L1 Turkish advanced

Accuray on the weak/strong distinction 85.0% 93.4% 68.6% 95.9%

Acceptance of control items (weak definites & extremely readings) 50.0% 56.0% 25.0% 45.0%

Table 8. The L2ers’ accuracy on the weak/strong distinction and their acceptance of weak definites and extremely readings. Table 8 shows the average of each of the groups’ accuracy rate on the weak/strong distinction and the average of their acceptance rates of weak definites and extremely readings, i.e. sentences in which strong expressions are licensed because they are semantically weak in ECs. An L2er that has acquired the DE of ECs properly should accept weak definites such as ni el menor ‘no the slightest’ and extremely readings in proper ECs, too, i.e. ideally, both numbers should be ‘100’. Otherwise it is impossible to tell whether the speaker overgeneralizes the ‘rule’ that weak expressions are allowed in ECs whereas strong ones are not. Table 8 shows that none of the four groups as a whole has accepted all control items containing weak definites but that this did not necessarily impair their performance on the weak/strong distinction. The intermediate L1 German group performed relatively well on the weak/strong distinction but did not score 100% on the control items. The advanced L1 German speakers performed better on the weak/strong distinction, but overall only

The Acquisition of the Definiteness Effect in L2 European Spanish

467

accepted slightly more weak definites in ECs and extremely readings than the intermediate speakers. The results for both L1 Turkish groups are very similar: The accuracy rates for the acceptance of weak definites and extremely readings amount to more or less half of the items in the intermediate group. Interestingly, the advanced L1 Turkish group shows the lowest acceptance rate for weak definites and extremely readings on the one hand, and, as was already mentioned, the most target-like performance on the weak/strong distinction, on the other. In sum, the results show that both the intermediate and the advanced L1 German group performed more target-like than the intermediate L1 Turkish group. However, the advanded L1 Turkish speakers accepted fewer strong expressions in positive and in negative ECs than the intermediate L1 Turkish groups and both L1 German groups. What the intermediate and advanced L1 Turkish groups have in common is that the strong expressions most readily accepted were proper names. However, whereas the intermediate L1 Turkish group accepted proper names and to a lesser extent than possessives both in positive and negative ECs, the advanced L1 Turkish speakers only accepted these expressions in positive ECs but not in negative ECs. Both L1 German groups only marginally accepted strong expressions in positive and negative ECs with the advanced L1 German group rejecting more of them than the intermediate L1 German speakers. None of the groups accepted all grammatical control items containing weak definites and extremely readings in proper ECs. The group who most readily accepted them was the advanced L1 German group. The advanced L1 Turkish group, on the other hand, presents a clear ‘gap’ in this respect with their acceptance rates of weak definites and extremely readings being very low but their accuracy rates on the weak/strong distinction being very high.

5. Discussion I hypothesized that (i) the DE in European Spanish would be acquired rather ‘effortlessly’; and that (ii) the L1 German group would outperform the L1 Turkish group due to positive L1 transfer. These hypotheses are only partially confirmed: (i) The Spanish DE is acquired effortlessly. As was shown, with the exception of the intermediate L1 Turkish group, both groups acquired the DE relatively well. This is also true for the intermediate L1 German speakers, although most of them scored poorly on the cloze test. Weak expressions were correctly accepted both in positive and negative ECs and

468

Chapter Fourteen

did not pose a problem for any of the four groups investigated. However, especially the intermediate L1 Turkish group tended to accept strong expressions both in positive and negative ECs. Their acquisition of the DE can therefore not be considered complete. The group that performed most target-like as far as the rejection of strong expressions is concerned, was the advanced L1 Turkish group. As in White et al.’s (2012) study, the speakers’ proficiency level had an impact on the speakers’ performance on the weak/strong distinction, with the L1 Turkish group showing a stronger ‘gap’ between the intermediate and advanced speakers than the L1 German group. What has to be borne in mind, though, is that the cut-off point for intermediate and advanced in my study was not the same as in White et al.’s (2012). (ii) The L1 German group outperforms the L1 Turkish group. This hypothesis, too, can only be partially confirmed. The intermediate L1 German speakers performed more target-like than the intermediate L1 Turkish speakers. However, within the group of advanced speakers, the L1 Turkish speakers outperformed the L1 German group. These results suggest that L1 German speakers acquire the DE at an earlier learning stage. The results of this study hint at other interesting facts about the DE, some of which I will discuss briefly. Inter-linguistic differences. In White et al.’s (2012) study, the intermediate L1 Turkish speakers were more target-like on the DE in L2 English than the subjects of the present study were in L2 European Spanish. This raises the interesting question of why the DE should be acquired less ‘effortlessly’ in English than in Spanish. Another question that has yet to be addressed is why the advanced L1 Turkish group accepted more strong expressions in negative ECs in L2 English than in L2 Spanish. However, when comparing the results of the two studies the following has to be borne in mind: (i) The categories of intermediate and advanced proficiency do not coincide in the two studies, (ii) the test design was not identical in both studies as different control items had to be included, even though the present study was inspired by White et al.’s (2012) test design, (iii) lastly, the fact that English was an L2 for the speakers investigated by White et al. (2012), whereas Spanish represented an L3 for the speakers investigated in this study might have had an influence on the learners’ performance (see next section). L1 & L2 transfer. European Spanish is, in fact, the speakers’ L3. As such, transfer from any of the speakers’ L2 cannot be ruled out, especially as the L2 has often been considered to have a special status in L3 acquisition. Falk & Bardel (2010), for instance, argue for a so-called ‘L2

The Acquisition of the Definiteness Effect in L2 European Spanish

469

status factor’ according to which the L2 is the priviledged source of transfer in L3 acquisition. Another factor that has been shown to play a role in L3 acquisition is the proficiency level in the background languages (again, see Falk & Bardel 2010) as well as actual or perceived typological proximity between languages (Rothman 2011), among others. However, since the speakers investigated here did not carry out the same task in all their background L2s it is difficult to tell whether deviations in the treatment of the DE might be due to L2 transfer. Further research should, of course, include a systematic investigation of the speakers’ background languages as well as their proficiency level of all their L2s. Acceptance of proper names in positive existentials. My study has shown that although the advanced L1 Turkish speakers did not accept any strong expressions in negative ECs, they did accept strong expressions in positive ECs. Interestingly, White et al.’s (2012) results, too, show that the advanced speakers accepted strong expressions in positive ECs, although Turkish positive ECs show a DE. The data further show that the strong expressions most readily accepted by all L1 Turkish speakers were proper names. The intermediate L1 Turkish group was the group that most readily accepted these strong expressions in both positive and negative ECs. But the advanced L1 Turkish group, too, accepted many of them in positive ECs, although they correctly rejected all strong expressions in negative ECs. I argue that there are three constructions that might be responsible for this acceptance of strong expressions in positive but not in negative sentences by the advanced L1 Turkish group: Firstly, there is a so-called ‘Genitive possessor construction’ in Turkish usually requiring the participle var – the same participle used in Turkish ECs – that does not abide by the DE (Kelepir 2001: 192): (27)

Kz-n alt parma÷- var. girl-GEN six fingers-POSS exist ‘The girl has six fingers’

It is possible that the L2ers’ confounded existential var, which does abide by the weak/strong distinction, with the possessive var of these constructions. Their acceptance of strong expressions in positive ECs might thus be attributed to formal similiarities between these two structures. The results of the present study have further shown that the strong expression most readily accepted by both proficiency groups were proper names. Interestingly, proper names can occur in Turkish ECs whenever they constitute a contrastive topic (28a) or the NP is focused (28b) (both examples are taken from Kelepir 2001: 211):

Chapter Fourteen

470

(28)

a.

b.

Banyo-da HASAN m var. bathroom-LOC Hasan PART. exist ‘Is it Hasan who is in the bathroom?’ Hasan var ya. Hasan exist INTERJ. ‘There is Hasan.’

However, if this is indeed the case it is not clear why in White et al.’s (2012) study, the intermediate L1 Turkish speakers tended to accept possessives instead of proper names in L2 English. It is unclear why possessives should be accepted more readily in English than in Spanish. Frequency Effects. In many languages the expression used in ECs is used in other contexts as well. This is true for Turkish var which is also used in Genitive-Possessor and contrastive-topic constructions. The fact that Turkish has no lexical verb meaning ‘have’ and thus has to resort to var/yok more often (Arabam var ‘I have a car’ (literally: ‘My car exists’)), too, stresses the fact, that these items appear very frequently in the input, probably with a higher frequency than in German, where existential readings can be expressed by regular copular sentences without the need to employ a default existential expression at all (Gestern waren viele Leute auf der Straße ‘Yesterday, there were many people in the street.’ (literally: ‘Yesterday, many people were in the street.’). A high occurrence of existential look-alike structures might thus have an influence on the acquisition process. Acceptance of weak definites and extremely readings. Taking a look at these items sheds some light on whether strong expressions were rejected categorically, even when they are grammatical as is the case with weak definites. Remember – weak definites (often in the form of extremely readings) in ECs are grammatical, as they are semantically indefinite. Within none of the four groups, weak definites and extremely readings, were accepted across the board. By contrast, even the advanced speakers of both L1 groups were rather reluctant to accept these items. The highest acceptance rate for these items was found within the advanced L1 German group. Interestingly, a ‘gap’ between the accuracy on the weak/strong distinction and the acceptance of weak definite and extremely reading control items within the advanced L1 Turkish group was found. Apart from one speaker, none of the advanced L1 Turkish speakers accepted more than 40% of these items, although this group performed extremely well on the weak/strong distinction in all other ECs. It is possible that both the L1 German and L1 Turkish groups were simply not familiar with these weak definites such as ni el menor ‘not the slightest’ or ni el más mínimo ‘not the tiniest’. Also, these expressions often occur in sentence types

The Acquisition of the Definiteness Effect in L2 European Spanish

471

other than existentials such as No tengo (ni) la menor duda ‘I don’t (even) have the slightest idea’ rather than in in proper ECs. Their reluctance towards accepting these items would thus not stem from difficulties as far as the weak/strong distinction is concerned but from the distribution of this more or less ‘fixed’ expression in Spanish. Test design. Although acceptability judgments enable us to obtain intuitive data, they have limitations. In addition to grammaticality/ acceptability judgments, it might make sense for future studies to complement this study by (semi-)spontaneous elicitation data to obtain a more complete picture of the learners’ interlanguage. Although the experimental design used for this study included a high number of control items, it is clear that even more control items should be included in future test designs aimed at testing the acquisition of the DE. Control items for the L1 Turkish group should include Spanish existential look-alike structures lacking the DE that have direct equivalents in Turkish as in this language var can be used in possessive, contrastive topic as well as in existential structures. Lastly, it is also necessary to point out that the results obtained in this study have mostly been linked to the speakers’ overall proficiency level. However, besides their proficiency level, it might be necessary to determine the speakers’ exposure to their background langauges, since both factors in combination rather than proficiency alone have been reported to account for language dominance and ensuing transfer patterns (see Wang 2013). Since these additional items render the test very long, the test sessions might be split into two sessions in order to avoid fatigue.

6. Conclusion I investigated the acquisition of the DE in European Spanish existential sentences by intermediate and advanced L1 speakers of German and Turkish. Whereas both German and European Spanish abide by the DE according to which strong expressions are ruled out from proper existential sentences, in Turkish only positive existential sentences display a DE. I hypothesized that the L1 German speakers should outperform the L1 Turkish speakers on the performance on the weak/strong distinction of European Spanish existential sentences due to positive L1 transfer but that the DE in general should be acquired relatively well as was shown in previous research on the acquisition of the DE in L2 English. Both hypotheses were only partly confirmed. On the whole, the DE was acquired very well by all groups except the intermediate L1 Turkish group, which shows that proficiency clearly improved the accuracy rate on

472

Chapter Fourteen

this phenomenon, at least for the L1 Turkish speakers. Although the intermediate L1 German speakers did outperform the intermediate L1 Turkish speakers on the weak/strong distinction, this does not hold true for the two advanced groups. Here, the advanced L1 Turkish speakers outperformed the advanced L1 German speakers. However, a closer look at the advanced L1 Turkish group reveals that they often applied the DE in negative but not in positive existential sentences. Very tentatively I assumed that this might be due to formal similarities between positive existential sentences and three other, non-existential sentences in Turkish containing the existential particle var that do not show a DE. What is more, all L1 Turkish speakers showed a bias towards accepting proper names in positive (and within the intermediate L1 Turkish group also in negative) existential sentences. I argued that this might be due to formal similarities to other structures in Turkish that do not show a DE but not necessarily to the acquisition of the DE itself. Previous studies showed that the intermediate L1 Turkish speakers particularly were more target-like on the DE in L2 English than the subjects of the present study in European Spanish. However, the two studies are not directly comparable. Most notably, English is the speakers’ L2 in White et al.’s (2012) study, whereas the speakers of my study were L3 speakers of European Spanish, so that differences between the acquisition of this phenomenon in English on the one hand and Spanish on the other might be due to general differences beetween L2 and L3 acquisition. This clearly calls for follow-up studies that do not only investigate this phenomenon in different languages and language families but both within L2 and L3 acquisition. It is clear that the present study can only give a general idea of what the acquisition process of the DE looks like in a language other than English.

References Abbott, Barbara. 1993. A pragmatic account of the definiteness effect in existential sentences. In Journal of Pragmatics 19: 39-55. Beaver, David; Itamar Francez & Dimitry Levington. 2005. Bad subject: (Non-)canonicality and NP Distribution in Existentials. In Efthymia Georgala and Jonathan Howell (eds.). SALT XV, 19-43. Ithaca, NY: Cornell University. Birner, Betty J. & Gregory Ward. 1998. Information Status and Noncanonical Word Order in English. Amsterdam/Philadelphia: John Benjamins.

The Acquisition of the Definiteness Effect in L2 European Spanish

473

Côté, Marie-Hélène. 1999. Quantification Over Individuals and Events and the Syntax-Semantics Interface: The Case of Existential Constructions. In Kimary N. Shahin, Susan Blake & Eun-Sook Kim (eds.). The Proceedings of the Seventeenth West Coast Conference on Formal Linguistics, 147-161. Stanford, CA:CSLI. Czinglar, Christine. 1999/2000. Pure Existentials as Individual-Level Predicates: Evidence from Germanic. In Wiener Linguistische Gazette 64-65: 55-82. Diesing, Molly. 1992. Indefintes. MIT Press, Cambridge, MA. Enç, Mürvet. 1991. The semantics of specificity. In Linguistic Inquiry 22: 1-25. Erguvanl-Taylan, Eser. 1987. The role of semantic features in Turkish word order. Folia Linguistica. In Folia Linguistica 21.2-4: 215-229. Falk, Ylva & Camilla Bardel. 2010. The study of the role of the background languages in third language acqusition. The state of the art. In International Review of Applied Linguistics in Language Teaching (IRAL) 48: 185-219. Kelepir, Meltem. 2001. Topics in Turkish Syntax: Clausal Structure and Scope. Unpublished Ph.D. dissertation, Massachusetts Institute of Technology. Kuno, Susumo. 1971. The Position of Locatives in Existential Sentences. In Linguistic Inquiry 2.3: 333-378. Leonetti, Manuel. 2008. Definiteness effects and the role of the coda in existential constructions. In Hendrik Høeg-Müller & Alex Klinge (eds.). Essays on nominal determination: From morphology to discourse management, 131-162. Amsterdam: Benjamins. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Lyons, John. 1967. A note on possessive, existential and locative sentences. In Foundations of Language 3.4: 390-396. McNally, Louise. 2011. Existential Sentences. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.). Semantics: an international handbook of natural language meaning, 1829-1848. Berlin: Walter de Gruyter Press. Milsark, Gary Lee. 1974. Existential Sentences in English. Ph.D. Dissertation. Massachusetts Institute of Technology (MIT). —. 1977. Toward an explanation of certain peculiarities of the existential construction in English. In Linguistic Analysis 3: 1-29. Musan, Renate. 1996. There Constructions Revisited. In Teresa Galloway & Justin Spence (eds.). Proceedings of SALT VI, 167-184. Ithaca, N.Y.: Cornell University.

474

Chapter Fourteen

Palacio-Martínez, Ignacio & Ana Martínez-Insua. 2006. Connecting linguistic description and language teaching: native and learner use of existential there. In International Journal of Applied Linguistics 16.2: 214-232. Rando, Emily & Donna Jo Napoli. 1978. Definites in there-sentences. In Language 54.2: 300–313. Robertson, Daniel. 2000. Variability in the use of the English article system by Chinese learners of English. In Second Language Research, 16: 135–172. Rodríguez-Mondoñedo, Miguel. In press. A restriction on the Definiteness Effect in Spanish. In Proceedings of the 37th North-Eastern Linguistics Society (NELS 37). Rothman, Jason. 2011. L3 syntactic transfer selectivity and typological determinacy: The typological primacy model. In Second Language Research 27.1: 107-127. Sawyer, Janet. 1973. Existential Sentences: A Linguistic Universal? In American Speech 48: 239-245. Thomas, Margaret. 1989. The acquisition of English articles by first- and second-language learners. In Applied Psycholinguistics 10: 335-355. Tura, Sabahat Sansa. 1986. Definiteness and referentiality: non-verbal sentences. In Dan I. Slobin & Karl Zimmer (eds.). Studies in Turkish Linguistics, 165-194. Amsterdam/Philadelphia: John Benjamins Publishing Company. Wang, Xin. 2013. Language dominance in translation priming: Evidence from balanced and unbalanced Chinese–English bilinguals. In The Quarterly Journal of Experimental Psychology 66: 727-743. White, Lydia. 2008. Different? Yes. Fundamentally? No. Definiteness Effects in the L2 English of Mandarin Speakers. In Roumyana Slabakova et al. (eds.). Proceedings of the 9th Generative Approaches to Second Language Acquisition Conference (GASLA 2007), 251-261. Sommerville, MA: Cascadilla Proceedings Project. White, Lydia, Alyona Belikova, Paul Hagstrom, Tanja Kupisch & Öner Özçelik. 2012. Restrictions on definiteness in second language acquisition. Affirmative and negative existentials in the L2 English of Turkish and Russian Speakers. In Linguistic Approaches to Bilingualism 2.1: 54-89. Zielke, Marina. 2013. The Acquisition of (one Kind) of Weak Definites in L2 Spanish. In Revista da ABRALIN 12.1: 201-232.