Experimental Insights into the Syntax of Romanian Ditransitives 9781501513657, 9781501518072

This book investigates the syntax of Romanian ditransitives building on new experimental data with a view to enable a mo

213 115 1MB

English Pages 303 [306] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Experimental Insights into the Syntax of Romanian Ditransitives
 9781501513657, 9781501518072

Table of contents :
Acknowledgements
Contents
Abbreviations
1 Introduction
2. Romanian direct objects. A comparative analysis with Spanish
3. Linking syntactic structure to interpretation
4. Some considerations on the role of the clitic as a movement trigger
5. On the feature specification of indirect objects
6. Romanian ditransitive configurations
7. General conclusions
Appendices
Bibliography
Index

Citation preview

Alina Tigău Experimental Insights into the Syntax of Romanian Ditransitives

Studies in Generative Grammar

Editors Norbert Corver Harry van der Hulst Founding editors Jan Koster Henk van Riemsdijk

Volume 141

Alina Tigău

Experimental Insights into the Syntax of Romanian Ditransitives

This publication was financially supported by the Alexander von Humboldt-Stiftung

ISBN 978-1-5015-1807-2 e-ISBN (PDF) 978-1-5015-1365-7 e-ISBN (EPUB) 978-1-5015-1347-3 ISSN 0167-4331 Library of Congress Control Number: 2019950221 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2020 Walter de Gruyter, Inc., Berlin/Boston Typesetting: Integra Software Services Pvt. Ltd. Printing and binding: CPI books GmbH, Leck www.degruyter.com

Acknowledgements This book has grown out of my research stay as a Humboldt Fellow at the University of Cologne, Institut für Deutsche Sprache und Literatur I during 2016 – 2018 and my first debt of gratitude goes to the Alexander von Humboldt Stiftung for the generous financial aid and to Klaus von Heusinger, my host and supervisor, for the warm hospitality, his unflinching support and for the endless hours of discussions on various parts of the manuscript and the experiments presented in this book. I am also grateful to my colleagues at UniKöln, Elif Krause, Elyesa Seidel, and Andreas Brocher for their kind help with the statistics and the design of the experiments presented here and to Frederike Weeber, Semra Kizilkaya, Lukasz Jędrzejowski, and Nagihan for their support with the intricate administrative aspects related to university life and life in general while in Cologne. I would not have been able to bring this work to its current shape, had I not had the chance to discuss it with Alexandra Cornilescu, my former PhD supervisor and dear professor at the University of Bucharest to whom I am deeply grateful for the invaluable guidance throughout the past 15 years. Also, I would like to thank the audiences at the Datives in Discourse Workshop, Linguistischer Arbeitskreis and Romanisches Seminar in Cologne, those at the workshop on DOM in Spanish - diachronic change and synchronic variation at the University of Zürich and the Research Seminar at the University of Geneva for their helpful comments on previous drafts of this monograph. The comments and suggestions of the anonymous reviewers from Mouton de Gruyter were likewise most valuable and proved crucial in turning the submitted manuscript into publishable material and the same goes for the help I received from Kirstin Börgen in preparing the final manuscript and Mervin Ebenezer regarding the copy-editing work. I wish to express my deepest gratitude to them all. Lastly, I would like to thank my family back home for their love and constant encouragement and my newly found family in Cologne, Iana, Maggie and Ingo, who stood by me in times of dispair and joy, always ensuring that I get a balanced diet of Chinese, German and Romanian food, work and entertainment.

https://doi.org/10.1515/9781501513657-202

Contents Acknowledgements Abbreviations

V

XI

1 1.1

Introduction 1 A brief overview of the issues touched upon in this book

2

Romanian direct objects. A comparative analysis with Spanish 13 Introduction 13 Specificity 15 Cierto vs. cualquiera 16 Mood 18 Conclusions 25 Scope 25 Extensional QPs 25 Intensional operators 28 Indefinites and strong quantifiers 33 Some contexts which disallow DOMed DOs 34 Interim conclusions 37 Syntactically triggered DOM 38 Interim conclusions 47 Reinforcing the conclusions: two experiments on Romanian DOs 49 Scope dependencies with marked and unmarked DOs 49 An experiment on specificity with marked and unmarked DOs 57

2.1 2.2 2.2.1 2.2.2 2.2.3 2.3 2.3.1 2.3.2 2.3.3 2.4 2.5 2.6 2.7 2.8 2.8.1 2.8.2

3 3.1 3.2 3.2.1 3.2.2 3.2.3 3.3 3.3.1 3.3.2

1

Linking syntactic structure to interpretation 85 Analysing Spanish Data: DOMed DOs move, unmarked DOs do not 86 Movement and DP internal structure 92 KPs vs. DPs 92 Linking syntactic position and semantic composition 95 Conclusions 104 On the differences between Spanish and Romanian DOs 104 Specificity, wide scope and binding 104 Weak indefinites and bare nominals 107

VIII

3.3.3 3.4 3.4.1 3.4.2 3.5 4

Contents

Conclusions 114 On the featural make-up of KPs 115 Sensitivity to animacy and definiteness 115 Integrating animacy and definiteness into syntax Conclusion 129

118

4.1 4.1.1 4.1.2 4.1.3 4.1.4 4.2

Some considerations on the role of the clitic as a movement trigger 131 The pronominal clitic as a movement trigger 131 Background. Evidence for movement 132 An experiment on Subject -DO binding dependencies 138 A theory of cliticisation 147 Deriving clitic doubling structures 151 Conclusion 155

5 5.1 5.1.1 5.1.2 5.2

On the feature specification of indirect objects 157 The featural make-up of the dative IO 157 Background 157 A [Person] feature for Goal DPs 161 Conclusions 164

6 6.1 6.1.1 6.1.2 6.2 6.2.1 6.2.2 6.3

Romanian ditransitive configurations 165 Background 166 Preliminaries 166 Motivation for the experiment and predictions 179 Experimental data on Romanian ditransitives 181 Quantificational dependencies with possessives 182 Quantificational dependencies with anaphors 211 A syntactic analysis of Romanian ditransitive configurations 218 Against two different configurations in Romanian ditransitives 218 Further arguments: semantic uniformity 219 A low position for datives 222 On [Person] and more 224 One basic configuration 227 A derivational analysis for Romanian ditransitives 234 Experimental evidence in favor of the proposed syntactic analysis 242

6.3.1 6.3.2 6.3.3 6.3.4 6.3.5 6.3.6 6.3.7

Contents

Secondary objects 261 Conclusions 268

6.3.8 6.4 7

General conclusions

Appendices Appendix 1

279

Appendix 2

281

Appendix 3

282

Appendix 4

283

Bibliography Index

293

285

271

IX

Abbreviations 1 2 3 ACC ApplP ART CD CL DAT DEF DIST DO DOC DOM DP FEM GEN IND INDEF IO M NP PL POC POSS REFL SG SUBJ VP

first person second person third person accusative Applicative Phrase article clitic doubling clitic dative definite distributive direct object Double Object Construction Differential Object Marking determiner phrase feminine genitive indicative indefinite indirect object masculine noun phrase plural Prepositional Object Construction possessive reflexive singular subjunctive verb phrase

https://doi.org/10.1515/9781501513657-204

1 Introduction 1.1 A brief overview of the issues touched upon in this book Romanian ditransitive configurations have been traditionally analysed on a par with their Romance counterparts, which, in turn, have been argued to pattern with their English correspondents, allowing for a Prepositional Object Construction (1a) and a Double Object Construction (1b). (1)

a. Mary gave a bookTheme to JohnGoal. (POC) b. Mary gave JohnGoal a bookTheme (DOC)

One important difference holding between the English POC and DOC has to do with the c-command relations arising between the two internal arguments. As argued by Barss and Lasnik (1986) and Aoun and Li (1989) a.o., the direct object DP (DO) is hierarchically superior to the indirect object (IO) in the prepositional configuration, while the opposite holds within the double object structure. Various binding dependencies with anaphors, possessives, reciprocals etc. are used to support this claim. Semantic differences arising between the two constructions have been also brought to the fore in an attempt to solve the mystery of the relatedness of the POC with the DOC or the lack thereof. The discussions centering on Romance ditransitives have been initially characterized by an effort to analyse these constructions on a par with their English counterparts. Demonte (1995), Cuervo (2003a) propose, for instance, that Spanish evinces both a POC and a DOC, with the same underlying configuration as their correspondents in English. The difference between the two constructions is claimed to arise from the presence of a dative clitic doubling the IO: POC obtains with non-doubled IOs (2b), while the clitic doubling of the IO gives rise to the equivalent of the English DOC (2a): (2)

a. Pablo le mandó un diccionario a Gabi. Pablo CL.1. DAT.SG sent a.ART dictionary to Gabi ‘Pablo sent Gabi a dictionary.’ b. Pablo mandó un diccionario a Gabi. Pablo sent a.ART dictionary to Gabi ‘Pablo sent a dictionary to Gabi.’ (Cuervo 2003a: 28 p. 30)

https://doi.org/10.1515/9781501513657-001

2

1 Introduction

Romanian has also been analysed along these lines, with a POC correspondent where the DO c-commands an undoubled IO (3a) and a DOC one, where the clitic doubled IO c-commands DO (3b). Just as for Spanish, the dative clitic has been argued to have a structural import in the sense that its presence ensures the reversal of the c-command relations between the two internal arguments: (3)

a. POC: Theme c-commands Goal [VoiceP DPAgentVoice[vPv [PPDPTheme P DPGoal]]] b. DOC: Goal c-commands Theme (clitic doubling) [VoicePDPAgentVoice[vPv [ApplPDPGoal [clAppl] [VP V DPTheme]]]] (Diaconescu and Rivero 2007: 219–220)

More recent work on ditransitives has uncovered various problematic aspects concerning this close parallelism between ditransitives in Romance and English: Ormazabal and Romero (2010) point out, for instance, that it is hard to find semantic differences between the cliticless and the clitic-doubled pattern(s) and that the two internal arguments show symmetric c-command. Pineda (2012, 2013) also argues against the view that the syntactic and semantic differences uncovered between the DOC and the POC in English are also apparent with Spanish ditransitives, where a dative clitic doubling the IO would distinguish between the two configurations. Building on bidirectional c-command facts involving the IO and the DO, she shows that the dative clitic has no structural import. Italian (Giorgi and Longobardi 1991) and French (Harley 2002) have also been argued to allow bi-directional c-command with the two internal arguments of ditransitives. This book comes as a contribution to the domain of Romance ditransitives by putting forth relevant data from Romanian with a view to enable a more accurate understanding of these constructions regarding their underlying configuration(s) and the featural make-up of the two internal arguments. One advantage of the analysis proposed here is that it rests on a strong experimental component. Throughout the book the results of twelve experiments grouped into six bigger blocks will be presented and discussed: the first three experimental blocks uncover various important aspects related to the syntax and semantics of the DO, while the remaining three focus on ditransitives proper, verifying the interactions between the two internal arguments with respect to binding and scope. More than 900 native speakers of Romanian took part in the experiments, mostly students of the University of Bucharest to whom we will forever remain grateful. Consider a list of the experiments in (4):

1.1 A brief overview of the issues touched upon in this book

(4)

3

1. Specificity with marked and unmarked direct objects 2. Scope dependencies with direct objects 3. Binding dependencies holding between the subject DP and the direct object 4. Binding with anaphors (ditransitives) 5. Binding with possessives (ditransitives) 6. Scope dependencies between direct objects and indirect objects

The analysis we end up proposing for Romanian ditransitive configurations is a derivational one, following the initial insights in Larson (1988) and making use of the more recent contributions in Ormazabal and Romero (2010), Pineda (2012, 2013), Georgala (2011), Georgala and Whitman (2008). One single underlying configuration is proposed, with both the IO and the DO merging low within the VP: in line with Larson 2010, we take IO to be part of the verb’s θ-grid, being introduced by the lexical verb and composing inside VP in a syntax quite similar to Larson (1988); the DO merges as the complement of V. We also posit an ApplP taking the lower VP as its complement. This is the high ApplP in Pylkkänen (2002, 2008) but refined in line with Georgala and Whitman (2008), Harada and Larson (2009), and Georgala (2011) in that it may simply function as an expletive (raising) applicative. In the analysis we propose, the Appl head does not introduce any argument into the argument structure of the verb but only participates in case checking and is further endowed with a [uPerson] feature. ApplP may be understood as the αP of López (2012): (5)

ApplP/αP

ei Appl/α

ei Applo/α

VP

ei IO

V’

ei V

DO

The featural make-up of the two internal arguments will also prove crucial in disentangling the various experimentally uncovered aspects having to do with the interaction between the two internal arguments. IOs and differentially object marked DOs will be shown to pattern alike in this respect, carrying a [Person] feature i.e, the syntactic counterpart capturing their sensitivity to the animacy and the definiteness scales (Richards 2008). The exact specification of this feature is

4

1 Introduction

further refined function of whether the respective nominal has been clitic doubled or not. Unmarked DOs have no [Person] feature to contribute. Besides the wide experimental range of data relevant for the syntax of Romanian ditransitives and of the direct object, this book makes several other contributions to the field of Romance linguistics and linguistics in general. Firstly, it establishes a connection between the internal make-up of the DP (direct object or indirect object), its syntax and its interpretation. As known, Romanian is one of those languages which allows for its direct objects to be both differentially object marked (DOM) as well as clitic doubled by means of a pronominal clitic (CD). Clitic doubling of Romanian direct objects (DO) is only possible with differentially object marked DOs; unmarked DOs may not be clitic doubled. Numerous studies have proposed a connection between marking1 and a certain interpretation (e.g., an epistemically specific interpretation for indefinite direct objects), without clearly teasing apart the exact contribution of each of the two mechanisms. One of the merits of this book regards the syntactic and semantic contribution of DOM and CD and the connection between the two mechanisms in this respect: the two types of marked DOs i.e., DOMed DOs and DOMed+CDed DOs are posited to have a different internal featural load, which, in turn, determines their derivation in syntax. Thus, differentially object marked direct objects (DOs) are argued to have KP status, with the K head triggered by an [uPerson] feature in N, which further projects in D. The [Person] feature represents the syntactic reflex of the animacy and definiteness scales, as proposed by Richards (2008). Nominals denoting [+Human] referents which surface as differentially marked DOs carry an unvalued [iPerson] feature, which is copied in D and valued by a merging K (with the feature specification: valued [uPerson]). These DOs are thus KPs, with an [iPerson] feature in N. The KP status of marked DOs is further linked to their syntactic derivation. The proposal follows the general lines of Diesing (1992), who also proposes a correlation between syntactic position and a peculiar interpretation, but it argues for a more indirect relation in this respect along the basic tenets of López (2012), where occupied positions are linked to different modes of semantic interpretation: nominal expressions which stay in situ are interpreted as predicates

1 Farkas (1987), Dobrovie-Sorin (1990) p. 388, Dobrovie-Sorin (1994) p. 234, Cornilescu (2000) p. 103 argue that DOM induces specificity. On the other hand, Steriade (1980), Dobrovie-Sorin (1990) p. 377, Dobrovie-Sorin (1994) p. 224 Gierling (1997) p. 72 ff a.o. point to the correlation between specificity and clitic doubling. Tigău (2011), Chiriacescu & von Heusinger (2011a,b) p.1 a.o. consider that specificity is a joint effect of pe and CD. None of these studies distinguishes between the exact contribution of DOM and that of CD.

1.1 A brief overview of the issues touched upon in this book

5

(via Restrict along the lines of Chung and Ladusaw 2004, 2006) and incorporate into V for case checking, while nominals which scramble are interpreted as arguments as a consequence of having left the VP and merged into a position where they may be interpreted by means of choice functions. The presence of an extra KP layer in the case of DOMed DOs blocks incorporation of these nominals into V and triggers scrambling to Spec αP. Just like their DOMed counterparts, DOMed+CDed direct objects also have a KP status. Nevertheless, they differ with respect to the featural specification of K in that this head is argued to have undergone further bleaching in the sense that its [uPerson] is no longer valued as in the case of single DOMed DOs, but unvalued (along the lines of Cornilescu 2017). Given that the DP complement itself contributes an unvalued [iPerson], the overall feature specification is [iPerson:__]. This difference with respect to feature specification between single DOMed DOs and DOMed+CDed DOs, is shown to have consequences in syntax: DOMed + CDed DOs need to move to a position where they may value their [iPerson:__], while single DOMed DOs do not have to do that since their [iPerson] is valued. These DOs only move to SpecαP given that K prevents incorporation, along the lines of López (2012) extended upon above. The fact that DOMed+CDed DOs need to undergo further movement is supported by the experimental results showing that while DOMed+CDed DOs may bind into the subject DP, single DOMed counterparts are problematic in this respect. Another contribution of this book is to have proposed a parallelism between the syntax of the Romanian direct object and that of indirect object DPs by drawing on similarities with Spanish counterparts where both internal arguments may undergo marking by means of a, a marker originating in the allative preposition a (Lat. ad) (Fábregas 2013). Unlike Spanish, Romanian marks prominent DOs by means of the differential object marker pe (on) derived from the locative preposition on, while indirect objects (IOs) may be either inflectionally marked (they have a dative case ending) or carry the marker la (at) derived from the allative preposition la (at). In spite of the different marking, the competition between marked DOs and IOs is as active in Romanian as it is in Spanish and it is thus incumbent on us to first understand their syntax before we embark on the examination of the ditransitive construction. Both inflectional as well as la-datives are interpretable as Goals, Beneficiaries, Maleficiaries, Affected Experiencers and if one is to generalize over this list of θroles, it seems that they standardly denote human entities. The essential property of the DPs which may realize the dative θ-roles is sensitivity to the animacy hierarchy, a characteristic which they seem to share with DOMed DOs: only nouns prominent on the hierarchy (e.g. humans, animals, corporate bodies) are possible

6

1 Introduction

realizations of the dative θ-roles. Thus, similarly to the DOM marker pe, the dative marker la may be taken to simply signal that its complement is [+Person]. Given this similarity, we posit that dative DPs have a similar internal featural make-up as DOMed direct objects, exhibiting KP status and evincing the same feature specification i.e., valued [uPerson]. Furthermore, clitic doubled IOs pattern on a par with clitic doubled DOs with respect to feature specification: [iPerson:__]. The difference with respect to feature specification of undoubled vs. doubled datives triggers the same derivation in syntax as with DOMed and DOMed+CDed DOs respectively: while undoubled IOs need not move for [Person] checking reasons, doubled IOs do. Cornilescu (2017) shows that this is indeed the case, given that doubled IOs may bind into the subject, while their undoubled counterparts cannot do so. Thus, in an example such as (6a), the dative is argued to bind the possessive in the Subject DP, as shown by co-indexation, while its undoubled counterpart in (6b) is claimed not to be able to do so: (6)

le ajută a. Religia lori/j Religion.DEF.ART their.POSS.3.PL CL.1.DAT.PL helps.3.SG multorai. many.DAT ‘Many people are helped by their religion.’ ajută multorai. b. Religia lor*i/j Religion.DEF.ART their.POSS.3.PL helps.3.SG many.DAT ‘Many people are helped by their religion.’

Finally, the analysis proposed for ditransitive constructions draws on the internal featural make-up of the two internal objects and represents a further point of novelty in this book, which may be relevant not only for Romanian but also for Romance languages in general (see Georgala 2011, 2012, Georgala et al. 2008, Pineda 2012, 2013 for Spanish, Harley 2002 for French, Giorgi and Longobardi 1991 for Italian for a discussion of similar phenomena related to symmetric c-command between the two internal arguments within ditransitives as those identified for Romanian experimentally). The similar internal make-up of the two internal arguments is shown to be instrumental with respect to their interaction and several problematic situations identified experimentally (e.g., co-occurrence of DOMed DOs and clitic doubled IOs) are shown to arise as a consequence of some intervention effects caused by the competition for the same probe between the two internal arguments that bear similar feature specification. By parametrizing the account proposed for Romanian ditransitives, one might explain similar phenomena within ditransitives in Romance.

1.1 A brief overview of the issues touched upon in this book

7

The book is organized as follows: Chapter 2 dwells on some important aspects related to the syntax and semantics of Romanian direct objects from a comparative stance with their Spanish counterparts as described by López (2012). We distinguish between unmarked and marked DOs, with the latter further splitting into differentially marked DOs (DOMed DOs) and clitic doubled and differentially object marked DOs (CDed+DOMed DOs).2 In line with Spanish, marked DOs are shown to have access to a specific reading and to be able to outscope other scope bearing expressions. They are further disallowed in contexts requiring property denoting nominals (e.g., the correspondents of the individual level have in the two languages), and allowed (even required as the only possibility) in contexts necessitating nominals with real argumenthood (small clauses, object control and clause union in causative configurations). In line with López (2012), we analyse marked DOs as KPs, with the DOM marker guaranteeing an type reading and thereby securing argumenthood and individuation. Romanian unmarked DOs differentiate themselves from their Spanish correspondents in that they seem to pattern with marked DOs in allowing for specific and wide scope readings as well as for the occurrence with the three contexts selecting real arguments (small clauses, object control and clause union in causative configurations). Spanish unmarked DOs, on the other hand, never evince a specific interpretation and may not outscope other scope bearing expressions. Furthermore, they are not allowed with small clauses, object control and clause union in causative configurations. In this, they seem to pattern with bare nominals, which exhibit a similar behaviour in both languages. Thus, the relevant parameter differentiating between Romanian and Spanish in this respect seems to relate to the division these languages draw with respect to nominals: while Spanish draws a line between KPs on the one hand and DPs and NPs/NumPs on the other, in Romanian the relevant division is between KPs and DPs on the one hand and NPs/NumPs on the other. These observations find support in two experiments verifying the ability of the three DO types to acquire a specific interpretation and to outscope various scope bearing expressions. The first experiment tests the behaviour of marked and unmarked indefinite DOs when co-occurring with other scope bearing expressions i.e., phrases headed by the extensional quantifier fiecare (every), two

2 Romanian is a DOM language, marking its prominent DOs by means of the functional marker pe, deriving from the lexical preposition p(r)e (on). DOMed DOs may be further clitic doubled and the mechanism as such is dependent on DOM as only DOMed DPs function as candidates for doubling.

8

1 Introduction

intensional operators, the conditional and the modal verb trebuie (must), and negation. As it appears, while marking seems to render the respective DO a more suitable candidate for outscoping other scope bearing expressions, the possibility of acquiring wide scope also remains an option for unmarked DOs. Along the same lines, the second experiment shows that both marked and unmarked DOs pattern somewhat alike with respect to specificity: all DO types lose their otherwise possible specific interpretation when used within contexts forcing a non-specific reading. Even though marking seems to render the respective DO a less suitable candidate for these contexts, neither DOMed DOs nor CDed+DOMed DOs are actually excluded as unacceptable. This shows that both types of marked DOs may read non-specifically and that Romanian does not possess a category of objects which are necessarily interpreted as specific, a conclusion also reached for Spanish by López (2012). Chapter 3 aims at putting forth an analysis of Romanian DOs by parametrizing López’s account in such a way as to accommodate for the experimentally uncovered contrasts between Romanian and Spanish. We retain from López’s analysis the correlation between syntax and semantic interpretation: a) nominal expressions which get interpreted in situ are interpreted as predicates; all nominals which scramble are interpreted as arguments. We also retain the correlation between syntactic position and case-valuation strategies. In-situ objects incorporate, scrambling objects check case by Agree. López (2012) links DO behaviour to its status: unmarked DOs have DP status and as such incorporate into V; marked DOs have an extra KP layer which hinders incorporation into V, thereby forcing the KP to move. The two positions are further linked to two different modes of semantic composition: while incorporating nominals are interpreted via Restrict (Chung and Ladusaw 2004, 2006) in their merge position, nominals undergoing scrambling get interpreted by means of choice functions, which explains their ability to read specifically and to also exhibit a wide scope interpretation when co-occurring with other scope bearing expressions. Given the differences in behaviour holding between unmarked DOs in Spanish and Romanian, we only retain this account for bare nominals and unmarked DOs in their weak readings (property denoting). On the other hand, unmarked DOs which exhibit real argumenthood are analyzed on a par with marked DOs. Romanian marked DOs are posited to scramble and to compose with their predicate in a similar way as their Spanish counterparts. Chapter 3 also contains a discussion regarding the featural make-up of DOMed and unmarked DOs: we posit that marked DOMed DOs are KPs, with K being triggered by the existence of an unvalued [Person] feature in N, which further projects in D. As already announced above, the [Person] feature counts

1.1 A brief overview of the issues touched upon in this book

9

as the syntactic reflex of the animacy and definiteness scales that DOMed DOs have been long argued to be sensitive to (Richards 2008). The DOM marker represents the spell-out of an uninterpretable [Person] feature and its distribution is determined by the local context, following from the interplay of the syntactic features of [person] and [definiteness]. [+Human] denoting nominals may surface as marked DOs (KPs) or unmarked objects. When surfacing as unmarked, these nominals follow in the footsteps of [-Human] nominals, exhibiting no specification as to a syntactic [Person] feature. [+Human] is simply a lexical property of the NP. When [+Human] denoting nominals are, on the other hand, syntactically marked as denoting people, they incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by a merging K (K itself carrying a valued [uPerson] feature). Marked DOs are thus KPs having an additional [iPerson] feature in N. The presence of this unvalued feature in N and then in D triggers the insertion of K. Unmarked DOs, on the other hand, evince a standard DP structure. On a par with DOMed DOs, CDed+DOMed DOs also possess a KP layer. K within these nominals has, however, undergone further bleaching in that the syntactic [uPerson] feature it contributes is unvalued. Consequently, the [iPerson:__] feature of the nominal DP complement may no longer be valued by way of agreement with the KP. An immediate result for these DOs is that they need to find a way to check their [Person] feature and will thus have to scramble even further than DOMed DOs, up to an outer specifier of vP where they would be able to enter agreement with the Person head carrying [iPerson: val]. Chapter 4 focuses on the syntax of CDed+DOMed DOs and couples the proposed account with some experimental data confirming movement of these DPs to a high position in the T area. The analysis we propose closely follows Preminger (2017), where CD is viewed as an instance of long head movement of D° out of the doubled DP. Roughly, CDed+DOMed DOs are argued to start out as KPs, just like their DOMed counterparts, but with a bleached K only contributing an unvalued [uPerson: __]. As a result, the [iPerson:__] feature may no longer be valued by way of agreement with the KP. Agreement taking place between K and the D head results in [iPerson:__] for the entire KP. Given that this feature needs further checking against an appropriate head, the KP will have to scramble all the way to a Person field above the vP (Săvescu 2009). Before reaching SpecvP, however, KP passes through SpecαP checking its [uCase: __] against v, in line with DOMed KPs. It then undergoes further movement to SpecvP and checks its [iPerson:__] against the Pers head. Having reached this site, however, the DO acts as an intervener between the T head

10

1 Introduction

and the Subject DP in the sense that T cannot access the DP Subject for case assignment. Consequently, DO undergoes cliticisation (given that doubled nominals do not act as interveners for other φ agreement and movement operations cf. Preminger 2017). Cliticisation thus counts as a means to remove the DO blocking agreement between T and the Subject DP. The DO obligatorily cliticizes on T, enabling the DP subject to get its case checked. Chapter 5 focuses on the other internal argument from within ditransitive configurations i.e., the IO and aims at uncovering its featural make-up. Similarly to DOMed DOs, Goal DPs are shown to be sensitive to the animacy scale (Aissen 2003) irrespective of whether they carry inflectional or prepositional dative case. This pattern of behaviour justifies a hypothesis according to which IO DPs would also carry a [Person] feature. Also, on a par with DOMed DOs, IOs are posited to have KP status (when surfacing as prepositional datives, the la marker is the exponent of K, while inflectional datives have a silent K). [+Human(like)] denoting Goal nominals thus incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by merging K. K carries a valued [uPerson] feature. The entire KP bears the feature specification [iPerson: val]. With clitic doubled IOs, we posit further bleaching of K (on a par with CDed+DOMed DOs), which only bears an unvalued [Person] feature [uPerson:___]. The entire KP will thus have the feature specification [iPerson: ___] and will have to undergo scrambling up to the Person field on a par with CDed+DOMed DOs. Chapter 6 is devoted to Romanian ditransitives. We start by describing four experiments testing binding dependencies with anaphors and possessives within these constructions. Several important findings are dwelt upon in this respect: Firstly, the two internal arguments are shown to exhibit symmetrical binding abilities in the sense that each of them can bind a possessor or an anaphor contained in the other. Furthermore, the existence of a pronominal clitic doubling the dative argument is shown to have no import on the binding potential of its double. This is a strong argument against the accounts in the literature assuming structural differences between ditransitive configurations containing clitic doubled IOs and their undoubled counterparts (Diaconescu and Rivero 2007). The presence of the dative clitic, on the other hand, appears to have a general impact in the sense of significantly lowering the acceptability scores of the respective items in comparison to their undoubled counterparts. One other important result concerns the surface word order, which seems to be crucial in determining the acceptability of binding relations: cases where the surface word order matches the binding direction seem to fare much better with respect to acceptability rating than those where inverse binding obtains.

1.1 A brief overview of the issues touched upon in this book

11

Finally, some problematic situations were identified: the co-occurrence of DOMed DOs with CDed IOs was granted very low acceptability scores by the respondents in both directions of binding. This effect was not noticed with the counterparts of these patterns in the experiments conducted with unmarked DOs and CDed+DOMed DOs, which prompts us to hypothesize that the lower acceptability has to do with the internal structure of the DPs involved. In view of the experimental findings, we argue in favour of a derivational account for ditransitives. The symmetric binding potential of the two internal arguments obtains as a consequence of their relative hierarchical order within the VP (see above) combined with subsequent movement for reasons of case assignment and [Person] checking. We further posit some priority criteria with respect to feature valuation between the two objects: DO has general priority over IO, but this may change function of the feature specification of the two objects. The system we propose allows us to account for all the patterns assessed as acceptable in the three experiments as well as to explain the problematic cases where a DOMed DO interacts with a clitic doubled IO. In the latter case, the analysis draws on the internal featural make-up of the two internal arguments and shows that the problem amounts to a locality issue: in this particular case, DOMed DOs carry [iPerson] and only need to check their case. The CDed IO, on the other hand, needs to check both case and its [iPerson:___] feature. Since IO has more features to verify, it will gain priority over DO (closer proximity to Appl is also important). The IO enters an Agreement relation with the Appl head (specified as [uPerson: val]) and checks both its case and [iPers: __] feature. The [uPerson: val] feature of Appl is EPP and the IO moves to SpecAppP. As such, it acts as an intervener for the DO, which may no longer move to a Spec of Appl to get its case feature valued by v. Chapter 7 contains the general conclusions to this book.

2 Romanian direct objects. A comparative analysis with Spanish 2.1 Introduction This chapter represents a first step towards our investigation of Romanian ditransitive configurations and focuses on some relevant aspects concerning the behaviour of one of their internal arguments i.e., the direct object (DO). Romanian ditransitives evince a number of problems in view of the properties of their two internal arguments: a) the possibility of clitic doubling (CD) both the direct object (DO) as well as the indirect object (IO); b) the possibility of differentially marking (DOM) the DO; c) the availability of both inflectional and prepositional markers for the dative DP. This chapter will only dwell on properties (a) and (b), relevant for the DO syntax, and will additionally touch upon various other important aspects in this respect e.g., the behaviour of the various types of DOs regarding specificity, scope dependencies, binding a.o. Properties (a) and (c), relevant for the syntax of the IO, will be touched upon in a subsequent chapter. From a Romance perspective, Romanian ranges closest to Spanish, which also allows for its DOs to be clitic doubled and differentially marked. Unlike Spanish, however, where there is a clear connection between DOM and dative marking, Romanian employs different marking mechanisms for DOM and Dative DPs. In (7a), the DO mulți studenți (many students) bears pe (on), the DOM marker, which has evolved from the locative preposition p(r)e (on) and has been also clitic doubled by means of the pronominal clitic I (them.Acc), while in (7b) the IO studenți (students) shows up in its inflected form studenților and its prepositional form la studenți (with la having derived from the directional preposition at, to). In the Spanish examples under (8) both the DO and the IO carry the same a marker originating in the allative preposition a (Lat. ad) (Fábregas 2013): (7)

a. I-am ajutat pe mulți studenți CL.3.ACC.PL-have.1.SG helped DOM many students în cariera mea. in career mine.POSS ‘I have helped many students in my career.’ b. Am dat note studenților/ la studenți. Have.1.SG given marks student.DAT.PL to student.ACC.PL ‘I gave marks to the students.’

https://doi.org/10.1515/9781501513657-002

14

(8)

2 Romanian direct objects. A comparative analysis with Spanish

a. Ho encontrado a la Have.1.SG met DOM DEF.ART ‘I met the girl.’ b. Le enviaron el libro CL.3.DAT.SG sent DEF.ART book ‘They sent the book to the girl.’

niña. girl a la niña. DOM DEF.ART girl

In spite of the different marking, the competition between marked DOs and IOs is as active in Romanian as it is in Spanish. It is therefore necessary to first understand the syntax of the marked and unmarked DOs in Romanian before we embark upon the examination of the ditransitive construction. This is the aim of this chapter and the subsequent one. The current chapter is rather descriptive in its aim and consists of two major parts: we first examine the behaviour of Romanian DOs from a comparative stance with their Spanish counterparts, analysing the potential of both marked3 and unmarked DOs to allow for specific readings, to outscope other scope bearing expressions, and to show up in certain contexts selecting either property denoting nominals or DPs possessing real argumenthood. It will be shown that while Spanish draws a clearcut distinction between marked DOs on the one hand and unmarked DOs and bare nominals on the other, Romanian seems to group marked and unmarked objects together and to distinguish between these DPs on the one hand and bare nominals on the other: marked and unmarked DOs roughly have the same distribution in Romanian, in sharp contrast with Spanish. The discussion regarding the behaviour of marked and unmarked DOs will be backed up by some experimental findings in the second part of this chapter. The two experiments presented will further expose the contrasts in behaviour between Romanian and Spanish DOs and will pave the way towards an analysis building on López (2012) and parametrizing this account in such as way as to accommodate for the Romanian data. The analysis proper constitutes the topic of the subsequent chapter.

3 By marked DOs we mean both those DOs which only carry DOM as well as those DOs that have also been clitic doubled besides being differentially marked. The former will be labelled as DOMed DOs, while the latter will be referred to as CDed+DOMed DOs.

2.2 Specificity

15

2.2 Specificity One of the most widely discussed topics in relation to the behaviour of Romanian marked and unmarked DOs has been their ability to acquire a specific reading. Both DOMed and CDed+DOMed indefinite DOs have been argued to have a nonambiguous specific interpretation and Romanian linguists differ in their opinions as to which of the two mechanisms i.e., DOM or CD bears responsibility for this interpretation. Thus, Farkas (1987), Dobrovie-Sorin (1990) p. 388, Dobrovie-Sorin (1994) p. 234, and Cornilescu (2000) p. 103 claim that the DOM marker pe induces specificity, while Steriade (1980), Dobrovie-Sorin (1990) p. 377, Dobrovie-Sorin (1994) p. 224 and Gierling (1997) p. 72 ff a.o. establish a correlation between specificity and clitic doubling. Tigău (2011), Chiriacescu & von Heusinger (2011a,b) p.1 a. o. argue, on the other hand, that specificity is a joint effect of pe and CD. The case of Spanish is quite similar in this respect, with DOMed DOs having been argued to evince a specific interpretation. As pointed out in López (2012) a.o., the addition of the DOM marker a or the subtraction thereof has an interpretive import: while objects thus marked may acquire a specific reading, unmarked objects may not do so (9): (9)

María busca a/ø un traductor de alemán. Maria seeks DOM/ø INDEF.ART translator of German ‘Maria is looking for a German translator.’

With respect to the type of specificity that marked indefinite DOs may evince, López argues in favour of epistemic and partitive specificity (Farkas 1994/ 1995). Unmarked indefinites, on the other hand, are claimed to read unambiguously non-specifically. The examples under (10) are presented as a case in point: while variant (a) allows for the use of both a marked and an unmarked indefinite DO, due to the fact that the student in question may be known to the speaker or not, variant (b) only allows for a marked indefinite DO, given that the referent denoted by un hijo mío (a son of mine) is known to the speaker: (10) a. Ayer vi a/ø un estudiante en la Yesterday saw.1.SG DOM/ø INDEF.ART student in DEF.ART biblioteca. library Yesterday I saw a student in the library.’

16

2 Romanian direct objects. A comparative analysis with Spanish

b. Ayer vi a/*ø un hijo mío en Yesterday saw.1.SG DOM/ø INDEF.ART son mine.POSS in la biblioteca. DEF.ART library ‘Yesterday I saw a son of mine in the library.’ (López 2012: 33, p. 17)

2.2.1 Cierto vs. cualquiera As expected, a modifier such as a certain, which has been shown to foreground an epistemic interpretation by forcing the referent denoted by the indefinite DP to become salient to the speaker’s mind, triggers the use of a. Expectedly, an unmarked indefinite is ungrammatical in this case: (11) Juan buscó a/*ø un cierto futbolista Juan sought.3.SG DOM/*ø INDEF.ART certain soccer player ‘Juan looked for a certain soccer player.’ (López 2012: 34, p. 17) On the other hand, a marked DOs lose their specific reading when modified by a free choice indefinite such as cualquiera (any). As expected, unmarked indefinite DOs are also fine in this context: (12) Juan buscó a/ø un futbolista cualquiera. Juan sought.3.SG DOM/ø INDEF.ART soccer player any ‘Juan looked for a soccer player, no matter who.’ (López 2012: 35, p. 17) López (2012) also draws a connection between a marked indefinite DOs and partitive specificity in Spanish. Thus, in example (13) below it is argued that only the marked DO is suitable on account of the partitive reading it has, while its unmarked counterpart is out, given that it may not acquire a partitive interpretation: Context: Some gentlemen came into the room (13) Me presentaron a/*ø uno de ellos. 1.ACC.SG introduced.3.PL DOM/*ø one of them ‘I was introduced to one of them.’ (López 2012: 36, p. 17) As already pointed out, Romanian has been ranged with those languages allowing DOM and the same interpretive effect has been signalled for the pe marker

2.2 Specificity

17

involved in this mechanism. Farkas (1987), and Dobrovie-Sorin (1990, 1994) a.o. point out that pe disambiguates the interpretation of the indefinite DO o secretară (a secretary) towards a specific one. In (14b) below the indefinite is argued to only have a specific reading.4 (14) a. Caut o secretară. Look.1.SG INDEF.ART. secretary ‘I am looking for a secretary.’ (Dobrovie-Sorin 1994: 48a p. 224) b. Caut pe o secretară. Look.1.SG DOM INDEF.ART. secretary ‘I am looking for a secretary.’ (Dobrovie-Sorin 1994: 61a p. 234) Along the same lines, Cornilescu (2000) argues that DOMed indefinite DOs are epistemically specific in the sense of Farkas (1994/1995): (15) . . . unde să vizitez pe niște vechi și buni frieteni. where SUBJ. visit.1.SG DOM some old and good friend.PL ‘where I should visit PE some good old friends.’ (Cornilescu 2000: 30 p. 102) Note, however, that unmarked indefinite DOs may also give rise to epistemically specific readings. Thus, unlike their Spanish counterparts, these DPs may be allowed in a context such as (16) below, even though one may well interpret the referent of the unmarked DO as known to the speaker: (16) Ieri am văzut pe/ø un copil de-al Yesterday have.1.SG seen DOM/ø INDEF.ART. child of meu în bibliotecă. mine.POSS. in library ‘Yesterday I saw a child of mine in the library.’ Also, Romanian unmarked indefinites behave differently in that they allow cooccurrence with the equivalent of a certain and enable a specific interpretation. In (17) below, both the DOMed and the unmarked DO are possible:

4 Romanian DOM exhibits further complications in that it is more often than not accompanied by Clitic Doubling and many linguists have argued that the two mechanisms actually pertain to the same phenomenon (Dobrovie-Sorin 1990, Gierling 1997, Tigău 2010, Chiriacescu & von Heusinger 2011a, b)

18

2 Romanian direct objects. A comparative analysis with Spanish

(17) Ion caută pe/ø un anumit fotbalist. John seek.3.SG DOM/ø INDEF.ART. certain soccer player ‘John is looking for a certain soccer player.’

2.2.2 Mood Mood has been long used as a tool to tease a specific interpretation apart from a non-specific one (Rivero 1979). In example (18) below, the DO is modified by a relative clause whose predicate is employed in the subjunctive mood. This triggers a non-specific interpretation on the indefinite una gestora (a manager). As pointed out by López (2012), the indefinite DP in this case may be either unmarked or differentially object marked by means of a. On the other hand, when the mood employed in the relative clause is the indicative as in (19), the sentence is ungrammatical if the DO does not carry the a marker. Furthermore, the indefinite DP may only be read specifically in this context i.e., ‘There is a specific manager who spoke German and whom Maria was looking for’. (18) María buscó a/ø una gestora que hablara Maria sought DOM/ø INDEF.ART. manager that spoke.SUBJ. alemán. German ‘Maria was looking for a manager that spoke.Subj German.’ (López 2012: 4, p. 1) (19) María buscó a/*ø una gestora que hablaba Maria sought.3.SG DOM/*ø INDEF.ART. manager that spoke.IND alemán. German ‘Maria was looking for a manager that spoke.Ind German.’ (López 2012: 5, p. 2) The subjunctive test may be further complicated through the combination with a certain or qualquiera, with the result that the former may be felicitously employed with the indicative mood and an a marked indefinite (20a), while the latter forces the use of the subjunctive mood and only allows the use of unmarked indefinite DOs (20b).

2.2 Specificity

19

(20) a. Maria buscó a/*ø una cierta gestora que Maria sought.3.SG DOM/*ø INDEF.ART. certain manager that habla/ *hable alemán. speak.3.SG.IND speak.3.SG.SUBJ. German. ‘Maria looked for a certain manager that speaks.IND/*SUBJ German.’ b. Maria buscó *a/ø una gestora cualquiera que Maria sought.3.SG *DOM/ø INDEF.ART. manager any that *habla/ *hable alemán. speak.3.SG.IND speak.3.SG.SUBJ. German ‘Maria looked for a manager (no matter what) that *speaks.IND/SUBJ German.’ (López 2012: 39, 40, p. 19) Farkas (1982, p.109–130) and Dobrovie-Sorin (1994) discuss similar constraints with respect to Romanian differentially marked indefinites. Thus, while the unmarked indefinite in (21a) may be interpreted either as specific or non-specific, it may only read non-specifically in (21b) and specifically in (21c). As it seems, in line with the Spanish data, the mood in the relative clause modifying the indefinite DO in Romanian has an important impact on the (specific vs. nonspecific) interpretation of this DP. Nevertheless, Romanian DOs behave differently from their Spanish counterparts when it comes to the use of unmarked indefinites with relatives containing the indicative mood (21c), which are quite felicitous in these contexts and have a specific reading. (21) a. Maria caută o croitoreasă. Maria search.3.SG INDEF.ART. seamstress. ‘Maria is looking for a seamstress.’ b. Maria caută o croitoreasă care să îi Maria search.3.SG INDEF.ART. seamstress that SUBJ. CL.3.SG.DAT. facă rochia de mireasă. make.3.SG dress.DEF.ART. of bride ‘Maria is looking for a seamstress who should make her bridal gown.’ c. Maria caută o croitoreasă care Maria search.3.SG INDEF.ART. seamstress that i-a făcut rochia de mireasă. CL.3.SG.DAT-have.3.SG made dress.DEF.ART. of bride ‘Maria is looking for a seamstress who made her bridal gown.’ With respect to the use of DOM in these contexts, Dobrovie-Sorin (1994) has argued that marked indefinite DOs acquire a specific interpretation. As a consequence, these DPs would not be expected in a context such as (21b), which forces a non-

20

2 Romanian direct objects. A comparative analysis with Spanish

specific interpretation (see also Tigău 2016 for a more extensive discussion). Tigău and von Heusinger (ms.) test these predictions for Romanian DOM in a number of contexts forcing non-specific readings and reach the conclusion that pe marked indefinite DOs may in fact occur in such contexts, even though the examples in question do not get as high acceptability scores as their counterparts containing unmarked DOs. What is even more interesting about such examples is that Clitic Doubled and DOMed indefinites seem to fare even better in the tested contexts forcing a non-specific interpretation. This is contrary to most studies on Romanian CD, which have viewed the pronominal clitic as a specificity trigger (Dobrovie-Sorin 1994, Tigău 2010, 2016, von Heusinger and Chiriacescu 2011a, 2012). Consider the following example tested by Tigău and von Heusinger (ms):5 (22) Context: Examenul de lingvistică a fost foarte greu și majoritatea studenților au picat. ‘The linguistics exam was very difficult and most of the students failed.’ a. Aceștia caută acum câțiva seminariști care să These search.3.PL now some seminar tutors that SUBJ. le explice din nou materia. CL.3.DAT.PL. explain again subject matter ‘They are now looking for some seminar tutors who would explain the subject matter to them once more.’ b. Aceștia caută acum pe câțiva seminariști care să These search.3.PL now DOM some seminar tutors that SUBJ. le explice din nou materia. CL.3.DAT.PL. explain again subject matter.DEF.ART. ‘They are now looking for some seminar tutors who would explain the subject matter to them once more.’ caută acum pe câțiva seminariști c. Aceștia Îi These CL.3.ACC.PL search.3.PL now DOM some seminar tutors care să le explice din nou materia. that SUBJ. CL.3.DAT.PL. explain again subject matter.DEF.ART. ‘They are now looking for some seminar tutors who would explain the subject matter to them once more.’

5 Tigău and von Heusinger (ms) test the behaviour of indefinite DOs in four contexts forcing a non-specific interpretation: cel mult (at most)/cel puțin (at least), the distributive câte, the free choice indefinite oarecare (any), and subjunctive vs. indicative mood in a relative clause modifying the DO. The resulting items are assessed by 60 native speakers of Romanian. For an extensive presentation of the experiment proper see section 8.2.

2.2 Specificity

21

The unmarked indefinite DO is expectedly felicitous in this kind of context, reaching a score of 6,19 GM on a 7 rung scale (where 1 represents the lowest acceptability score, while 7, the highest). The differentially marked DO in (22b) does not fare as well, with a 3,43 GM score, but seems to be quite near the acceptability threshold (we considered GM of 3,5 as a sign that the sentence was assessed as marginally acceptable).6 Other similar tested examples fare much better. Consider (23) below, where the variant containing a DOMed DO modified by a relative in the subjunctive mood reaches a GM of 4,29 on the same acceptability scale: (23) Context: Maria nu dispune de foarte mulți bani și trebuie să elibereze apartamentul până la sfârșitul lunii. ‘Maria does not have so much money and needs to move out of her apartment by the end of the month.’ a. Ea vrea să cheme câțiva prieteni care să She want.3.SG SUBJ. call some friend.PL. that SUBJ. o ajute cu mobila și CL.3.ACC.FEM.SG help with furniture.DEF.ART. and obiectele mai grele. object.DEF.ART.PL. more heavy.PL. ‘She wants to call some friends who might help her with the furniture and the heavier objects.’ b. Ea vrea să cheme pe câțiva prieteni care să She want.3.SG SUBJ. call DOM some friend.PL. that SUBJ. o ajute cu mobila și obiectele CL.3.ACC.SG help with furniture.DEF.ART. and object.DEF.ART.PL. mai grele. more heavy.PL. ‘She wants to call some friends who might help her with the furniture and the heavier objects.’ c. Ea vrea să-i cheme pe câțiva prieteni She want.3.SG SUBJ.-CL.3.ACC.PL call DOM some Friend.PL. care să o ajute cu mobila și that SUBJ. CL.3.ACC.SG help with furniture.DEF.ART. and

6 Sentence (b) is in fact one of the items with the lowest acceptability score. Most items containing single DOMed DOs fare better, with scores higher than the acceptability threshold set for 3,5 (e.g., 3,61/3,57/3,90/3,66/3,85/4,29 etc.). Only two other items of this type get scores lower than 3,5 i.e., 3,29 and 2,95.

22

2 Romanian direct objects. A comparative analysis with Spanish

obiectele mai grele. object.DEF.ART.PL. more heavy.PL. ‘She wants to call some friends who might help her with the furniture and the heavier objects.’ Note also that the contexts containing clitic doubled and DOMed indefinites fare even better than their counterparts where only single DOMed indefinites are employed: variants (22c) and (23c) get acceptability scores of 5,43 and 4,81 GM respectively.7 This shows, in line with López (2012) that DOM (or the pronominal clitic) should not be actually looked upon as pieces of morphology inducing specificity. Rather, marking arises out of syntactic reasons and the availability of a specific interpretation comes as a consequence of a certain mode of semantic composition of the object with its predicate, which is in turn dependent on a peculiar syntactic position (either within VP or in SpecαP). Along the same lines, DOMed indefinite DOs in Romanian may combine with free choice indefinite oarecare (any) being bereft of their specific interpretation. The same is at stake for unmarked indefinites: (24) Alege pe/ø un coleg oarecare și Pick DOM/ø INDEF.ART colleague any and roagă-l să te ajute. ask.2.SG-CL.3.ACC.SG SUBJ. 2.ACC.SG help.3.SG ‘Pick up any colleague and ask him to help you.’ When combining the subjunctive test with expressions such as oarecare (any) or un anumit (a certain), the result is again somewhat different from the Spanish data with respect to the behaviour of unmarked indefinites: while pe marked DOs behave similarly to their a marked counterparts in allowing cooccurence with the indicative and a certain, contrary to their Spanish counterparts, Romanian unmarked indefinites allow co-occurrence with the indicative and a certain (26b). Both marked and unmarked indefinites are felicitous in the context of qualquiera and the subjunctive (25a). The combination between the subjunctive with a certain seems to be problematic both for marked and unmarked indefinites, as expected (26a). Finally, oarecare and the indicative seem problematic together both for marked and unmarked DOs (25b).

7 In fact, most contexts containing clitic doubled and DOMed indefinite DOs get scores higher than the acceptability threshold and even higher than their counterparts containing single DOMed DOs; Some items get values as high as 5,42/4,66/4,80.

2.2 Specificity

23

(25) a. Caut pe/ø un student oarecare care să Search.1.SG DOM/ø INDEF.ART Student any who SUBJ. vorbească bine englezește la clasă. speak well English at Class ‘I am looking for a (any) student who might speak English well.’ b. ?Caut pe/ø un student oarecare care a vorbit Search.1.SG DOM/ø INDEF.ART student any who has spoken bine englezește la clasă. well English at class ‘I am looking for a (any) student who spoke English well.’ (26) a. ? Caut pe/ø un anumit student care să Search.1.SG DOM/ø INDEF.ART certain student who SUBJ. vorbească bine englezește la clasă. speak well English at class ‘I am looking for a certain student who might speak English well.’ b. Caut pe/ø un anumit student care a vorbit Search.1.SG DOM/ø INDEF.ART certain student who has spoken bine englezește la clasă. well English at class ‘I am looking for a (any) student who spoke English well.’ Finally, Romanian unmarked indefinites may be used in contexts where they are interpreted as partitively specific: Context: Some gentlemen entered the room. (27) Mi-au prezentat pe/ø unul dintre ei. 1.DAT.SG-have.3.PL presented DOM/ø one of them ‘They introduced one of them to me.’ The conclusion we draw from these examples is that DOMed indefinite DOs may be used in contexts forcing non-specific interpretations, similarly to their Spanish counterparts. Consider also the example below, where a clitic doubled and DOMed indefinite may be used with a generic interpretation, suggesting that the contribution of DOM or CD does not necessarily amount to specificity. (28) Dacă angajatorul îl concediază pe un If employer.DEF.ART. CL.3.ACC.SG fire.3.SG DOM INDEF.ART. angajat încălcând prevederile Codului Muncii employee breaking provisions.DEF.ART. Code.GEN. Labour.GEN

24

2 Romanian direct objects. A comparative analysis with Spanish

este amendat cu 5000 de RON. is fined with 5000 of RON ‘If an employer fires an employee and by so doing breaks the provisions of the Labour Code, he faces a fine of 5000 RON.’8 The internet provides, in fact, quite a rich range of data supporting the hypothesis that neither DOM nor CD should be viewed as specificity triggers and that their contribution is rather a syntactic one. Consider some samples in (29) below: (29) a. Cea mai gravă înjurătură este să-l numești The most serious Curse is SUBJ-CL.3.ACC.SG name pe un bucureștean . . . moldovean DOM INDEF.ART. Bucharester . . . Moldavian ‘The worst curse for a Bucharester is to call him a Moldavian.’ b. Ce Îl enervează pe un bărbat la o what 3.ACC.SG irritates DOM INDEF.ART. man at INDEF.ART. femeie? woman ‘What is it that a man finds annoying in a woman?’ c. Noi publicul apreciem talentul, dar nu suntem We public.DEF.ART. appreciate talent.DEF.ART. but not are.1.PL cei mai nimeriți a desemna pe un concurent sau the more suitable to choose DOM INDEF.ART. contestant or altul ca fiind câștigător. another as being winner. ‘We, the public, appreciate talent, but we are not the most appropriate (judges) when it comes to choosing one contestant over another as the winner.’9 d. Faptul de a trimite în judecată sau de a Fact.DEF.ART. of to send in trial or of to condamna pe o persoană, știind că este condemn DOM INDEF.ART. person knowing that is

8 (http://cristianpaun.finantare.ro/2010/09/23/munca-e-bratara-de-aur/), Access date: 10.06.2018 9 (http://vocearomaniei.protv.ro/video/vocea-romaniei-semifinala-2016-ramona-nerra-one-moretry.html), Access date: 10.06.2018

2.3 Scope

25

nevinovată, constituie infracțiunea de a . . . innocent constitutes crime.DEF.ART. of to . . . ‘Prosecuting or condemning someone known as innocent constitutes the crime of . . . ’10

2.2.3 Conclusions Building on the data presented above, we may draw the following two conclusions with respect to Romanian: unlike Spanish indefinites, Romanian unmarked indefinites may acquire a specific interpretation. As such, they may be used in the context of anumit (a certain) or may be modified by relative clauses in the indicative mood, where they get a specific interpretation. (Remember that according to López, Spanish unmarked indefinites cannot be specific.) Similarly to their Spanish counterparts, DOMed indefinite DOs allow both a specific as well as a non-specific interpretation.

2.3 Scope When interacting with other scope bearing expressions, marked DOs have been claimed to favour a wide scope interpretation both in Romanian (Dobrovie-Sorin 1994, Tigău 2010) as well as in Spanish (López 2012). This section aims to uncover the behaviour of both marked and unmarked DOs in the context of extensional QPs, intensional operators and negation. It will be shown that, just like with respect to their potential to evince specific readings, Spanish DOs differentiate themselves from their Romanian counterparts in that only marked DOs may acquire a wide scope reading in the former language, while both marked and unmarked DOs may outscope other scope bearing expressions in the latter language.

2.3.1 Extensional QPs Besides favouring a specific interpretation, Spanish DOMed indefinite DOs seem prone towards a wide scope reading when interacting with extensional

10 (https://www.luju.ro/institutii/min-justitiei/si-dorel-cui-ramane-partenera-de-completa-vestitului-dorel-matei-de-la-curtea-de-apel-bucuresti-judecatoarea-anamaria-tranca-sepregateste-de-un-post-caldut-intr-o-institutie-ue-anamaria-tranca-s-a-inscris-in-procedura-pentruselectarea-adjunctului-membr), Access date: 10.06.2018

26

2 Romanian direct objects. A comparative analysis with Spanish

quantifiers and various sentence operators. Thus, in (30a) below, the indefinite a una mujer (a woman) may outscope the universal quantifier todo hombre (every man), allowing for an interpretation according to which ‘there was a woman such that every man loved‘. The narrow scope reading of the marked indefinite also remains an option. (30b) evinces the same interpretive import of the a marker, with the DOMed DO allowing for both a narrow scope reading and a wide scope one. (30) a. Todo hombre amó a Every man loved DOM ‘Every man loved a woman.’ 9>∀ ∀>9 b. La mayoría de los The most of DEF.ART. mujer. woman Most men loved a woman.’ 9 > Most Most > 9

una mujer INDEF.ART. woman

hombres amó a una men loved DOM INDEF.ART.

(López 2012: 25a,b, p. 13)

The unmarked counterpart of the indefinite DO in (30) above may not, on the other hand, outcope the universal quantifier. The only available reading in (31) is the dependent one, according to which for every man there exists (at least) a woman such that the man in question loves. (31) a. Todo hombre amó una mujer. Every man loved INDEF.ART. woman ‘Every man loved a woman.’ * 9>∀ ∀>9 b. La mayoría de los hombres amó una DEF.ART. most of DEF.ART. men loved INDEF.ART. mujer. woman Most men loved a woman.’ * 9 > Most Most > 9 (López 2012: 26a,b, p. 10)

2.3 Scope

27

Romanian data pattern with the Spanish ones in that DOMed indefinite DOs may outscope the quantifiers in the subject position in both (32a) and (32b). The broad scope reading is actually preferred over the narrow scope one: (32) a. Fiecare bărbat a iubit pe o femeie Every man has loved DOM INDEF.ART. woman era tânăr. was young ‘Every man loved a woman when he was young.’ 9>∀ ∀>9 b. Majoritatea bărbaților au iubit pe o Most men Have loved DOM INDEF.ART. în tinerețe. in youth ‘Most men loved a woman in their youth.’ 9 > Most Most > 9

pe când when

femeie woman

Unlike Spanish, however, unmarked indefinite DOs seem to allow both a narrow scope interpretation and a wide scope one with respect to the quantifier in subject position: (33) a. Fiecare bărbat a iubit o femeie Every man has loved INDEF.ART. woman era tânăr. was young ‘Every man loved a woman when he was young.’ 9>∀ ∀>9 b. Majoritatea bărbaților au iubit o Most men Have loved INDEF.ART. ‘Most men loved a woman in their youth.’ 9 > Most Most > 9

pe când when

femeie în tinerețe. woman in youth

Marked indefinite DOs may also outscope negation. In fact, (34a) shows that the only available interpretation is the broad scope one. The unmarked indefinite on the other hand only allows for a narrow scope interpretation (34b):

28

2 Romanian direct objects. A comparative analysis with Spanish

(34) a. Juan no amó a una mujer. Juan not loved DOM INDEF.ART. woman ‘There was a woman such that Juan did not love’ *Juan did not love any woman. 9>: *: > 9 (López 2012: 25c, p. 10) b. Juan no amó una mujer. Juan not loved INDEF.ART. woman *‘There was a woman such that Juan did not love’ ‘Juan did not love any woman‘. *9 > : :>9 (López 2012: 26c, p. 10) Again, the Romanian data pattern with the Spanish ones only in what the marked DO is concerned i.e., such DPs may outscope the negation operator (35a), the wide scope interpretation being actually favoured. Unmarked DOs may also do so, however, contrary to their Spanish counterparts (35b): (35) a. Ion nu suportă pe o femeie John not stand.3.SG DOM INDEF.ART. woman în casa lui. in house his There is a woman such that Juan does not stand in his house.’ ‘Juan does not stand any woman in his house. 9 >: : >9 b. Ion nu suportă o femeie în casa lui. John not stand.3.SG INDEF.ART. woman in house his ‘There is a woman such that Juan does not stand in his house.’ ‘Juan does not stand any woman in his house. 9>: :>9 2.3.2 Intensional operators Another observation made by López (2012) is that Spanish unmarked indefinite DOs may not outscope the conditional in (36) below and that the only interpretation in this case is one according to which the indefinite has narrow scope

2.3 Scope

29

(36b). A marked indefinites, on the other hand, may acquire wide scope with respect to the conditional (36a): (36) a. Si Lud invita a un filósofo, Bert se If Lud invite.3.SG DOM INDEF.ART. philosopher Bert REFL. ofenderá. take offense ‘If Lud invites a philosopher, Bert will be offended.’ 9> ! ! >9 b. Si Lud invita ø filósofo, Bert se If Lud invite.3.SG ø INDEF.ART. philosopher Bert REFL. ofenderá. take offense ‘If Lud invites a philosopher, Bert will be offended.’ *9 > ! ! >9 (López 2012: 6, p. 2) The same goes for the marked indefinite in (37), which may outscope both the conditional and the universal quantifier. Note also, that an intermediate scope interpretation, according to which the friend may be dependent on the universal quantifier, is also available. Unmarked indefinites, on the other hand, only allow for the narrowest scope reading in this context (37b): (37) a. Todo el mundo está convencido de que si All DEF.ART. world is convinced of that if invito a un amigo mío a la invite.1.SG DOM INDEF.ART. friend mine.POSS. to DEF.ART. fiesta, será un desastre. party will be INDEF.ART. disaster ‘Everybody is convinced that if I invite a friend of mine to the party, it will be a disaster. 1. ∀ > ! > 9 2. ∀ > 9 > ! 3. 9 > ∀ > ! (López 2012: 25f, p. 14) b. Todo el mundo está convencido de que si All DEF.ART. world is convinced of that if

30

2 Romanian direct objects. A comparative analysis with Spanish

invito un amigo mío a la fiesta, invite.1.SG INDEF.ART. friend mine.POSS. to DEF.ART. party será un desastre. will be INDEF.ART. disaster ‘Everybody is convinced that if I invite a friend of mine to the party, it will be a disaster. 1. ∀ > ! > 9 2. *∀ > 9 > ! 3. *9 > ∀ > ! (López 2012: 26f, p. 15) Romanian data are similar to the Spanish ones up to a certain point. Consider: (38) a. Dacă Ion admiră o actriță, Maria îl If John admire.3.SG INDEF.ART. actress Mary CL.3.ACC.SG va certa. will scold ‘If John admires an actress, Mary will scold him.’ 9 > ! ! >9 b. Dacă Ion (o) admiră pe o actriță, If John (CL.3.ACC.SG) admire.3.SG DOM INDEF.ART. actress Maria îl va certa. Mary CL.3.ACC.SG will scold ‘If John admires an actress, Mary will scold him.’ 9 > ! ! >9 In line with the Spanish data, DOMed indefinite DOs may acquire wide scope with respect to the conditional and it might well be the case that the only reading available to the marked DO in (38) is the specific one. Nevertheless, unmarked Romanian indefinites seem to also be able to outscope the conditional, as opposed to their Spanish counterparts. It seems to us that variant (38a) allows a reading according to which there is a certain actress such that if John happens to admire that actress, Mary will get upset. When it comes to interacting with more than one scope taking expression, Romanian marked indefinite DOs behave on a par with their Spanish counterparts. In (39a) pe un student de-al meu (a student of mine) may outscope the universal quantifier as well as the conditional, enabling a reading according to which ‘there is a certain student of mine such that, if I happen to invite him,

2.3 Scope

31

the party will turn out to be a disaster’. An intermediate reading according to which the student in question varies in relation to the colleagues is also available. Finally, the narrow scope interpretation remains an option. (39) a. Toți colegii mei sunt convinși că dacă invit All colleagues mine.POSS are convinced that if invite.1.SG pe un student de-al meu la petrecere, va DOM INDEF.ART. student of mine.POSS to party will ieși un dezastru. turn out INDEF.ART. disaster ‘All my colleagues are convinced that if I invite a student of mine to the party, it will turn out a disaster.’ 1. ∀ > ! > 9 2. ∀ > 9 > ! 3. 9 > ∀ > ! b. Toți colegii mei sunt convinși că dacă invit All colleagues mine.POSS are convinced that if invite.1.SG un student de-al meu la petrecere, va ieși INDEF.ART. student of mine.POSS to party will turn out un dezastru. INDEF.ART. disaster ‘All my colleagues are convinced that if I invite a student of mine to the party, it will turn out a disaster.’ 1. ∀ > ! > 9 2. ∀ > 9 > ! 3. 9 > ∀ > ! Unlike their Spanish counterparts, unmarked indefinite DOs in Romanian allow for all of the above mentioned readings, being thus able to outscope both the universal quantifier and/or the conditional. López (2012) makes some claims with respect to the scope taking abilities of indefinite objects, which do not seem to be in accordance with the Romanian data as presented above. Thus, in (40a), the DOMed indefinite is claimed to be able to outscope the conditional, while its unmarked counterpart in (40b) is claimed not to have such an outscoping potential. We believe that this is not actually the case and the experiment presented in Section 8 will show that both marked and unmarked indefinites may ouscope conditionals.

32

2 Romanian direct objects. A comparative analysis with Spanish

(40) a. Dacă Bert invită pe un filosof, Lud se If Bert invites DOM INDEF.ART. philosopher Lud REFL. va supăra. will get upset ‘If Bert invites a philosopher, Lud will be annoyed.’ ! >9 9> ! b. Dacă Bert invită un filosof, Lud se If Bert invites INDEF.ART. philosopher Lud REFL. va supăra. will get upset ‘If Bert invites a philosopher, Lud will be annoyed.’ ! >9 (López 2012: 114, 115, p. 142) *9 > ! A similar observation may be drawn with respect to the interaction between an indefinite DO and a modal verb: the unmarked indefinite un coordonator (a coordinator) in (41a) may outscope the modal of necessity and enable a reading according to which ‘there is a certain professor in the commission such that Michael must select‘. The use of DOM and CD in the same context seems to tilt the balance towards a wide scope reading of the indefinite. Nevertheless, a narrow scope reading in (41b) and (41c) remains an option: (41) a. Mihai trebuie să aleagă un coordonator dintre Michael must SUBJ. choose INDEF.ART. supervisor from profesorii din comisie. professors.DEF.ART. from comission ‘Michael must choose a supervisor from the professors in the commission.’ b. Mihai trebuie să aleagă pe un coordonator Michael must SUBJ. choose DOM INDEF.ART. supervisor dintre profesorii din comisie. from professors.DEF.ART. from comission ‘Michael must choose a supervisor from the professors in the commission.’

2.3 Scope

33

c. Mihai trebuie să îl aleagă pe un Michael must SUBJ. CL.3.SG choose DOM INDEF.ART. coordonator dintre profesorii din comisie. supervisor from professors.DEF.ART. from comission ‘Michael must choose a supervisor from the professors in the commission.’ Thus, unmarked indefinites in Romanian pattern differently than Spanish unmarked indefinite DOs in that they may be modified by a relative clause containing an indicative mood, in which case they allow for a specific reading. Furthermore, they seem to also be able to outscope quantifiers, negation, and intensional operators such as the conditional or modals. 2.3.3 Indefinites and strong quantifiers López (2012) also signals a difference between indefinites and strong quantifiers pointing out that the latter seem to be more rigid with respect to scope readings. Thus, unlike indefinites, strong quantifiers in object position may not outscope weak quantifiers occupying the subject position (42): (42) a. Un hombre ama a toda mujer. INDEF.ART. man love.3.SG DOM every woman ‘A man loves every woman.’ 9>∀ (López 2012: 7, p. 2) *∀ > 9 b. Ayer visitó un hombre a la mayoria de Yesterday visited INDEF.ART. man DOM DEF.ART most of las mujeres prisioneras. DEF.ART. women prisoners ‘Yesteray, a man visited most of the female prisoners.’ 9 > Most *Most > 9 (López 2012: 28, p. 16) Again, Romanian strong quantifiers evince somewhat different properties (43): although the DO fiecare femeie (every woman) strongly prefers a narrow scope reading, a wide scope reading remains a possibility.

34

2 Romanian direct objects. A comparative analysis with Spanish

(43) Un bărbat iubește pe fiecare femeie. INDEF.ART. man loves DOM every woman ‘A man loves every woman.’ 9>∀ ∀>9 The same goes for the DO headed by most in (44), which may outscope the existential quantifier even if the preferred reading is the narrow scope one: (44) Ieri un bărbat a vizitat pe majoritatea femeilor Yesterday INDEF.ART. man has visited DOM most women din închisoare. from prison. ‘A man visited most women in prison yesterday.’ 9 > Most Most > 9 Building on examples such as (43) and (44) above, López comes to the conclusion that the accusative a marker only allows a wide scope reading when prefixing an indefinite DO. As seen, Romanian data do not seem to support such a claim.

2.4 Some contexts which disallow DOMed DOs Building on Bleam (2005), López shows that a marked DOs are infelicitous as complements of the existential haber (have) and the possessor or relator tener (have). (45) En el patio hay *a/ø un niño. In DEF.ART. yard HABER *a/ø INDEF.ART. child ‘There is a boy in the yard.’ (López 2012: 44, p. 20) While haber always selects unmarked indefinites that are property denoting, tener allows for both a marked and unmarked DOs. A difference in interpretation is, however, at stake. In line with Bleam (1999, 2005), López distinguishes between an individual level tener and a stage-level one. In (46a), tener functions as an individual level predicate and is the equivalent of own in that the

2.4 Some contexts which disallow DOMed DOs

35

possession relationship is not associated with or restricted by a particular spatio-temporal location. In this particular context the use of DOM is disallowed. (46b), on the other hand, prompts the stage level reading and the use of a is permitted: (46) a. María tiene *a/ø tres hijos. Mary has *a/ø three children ‘Maria has three children.’ b. María tiene a/ø un hijo en el ejército. Maria has a/ø a child in the army ‘Maria has a son in the army.’ (López 2012: 45, 48, p. 20, 21) (47) I have a car. ≫ a. I own a car b. I have a car (with me today)

(individual-level) (stage level)

Romanian exhibits the same pattern as Spanish tener in that the verb a avea (have) allows for both a stage level reading and an individual level one, giving rise to the same definiteness effect. Notice also that the individual-level reading of have may only allow for an indefinite complement; the definite bicicleta (the bike) is not permitted under this reading, but allowed with the stage-level interpretation as shown in (48b): (48) a. Am bicicletă/ *bicicleta. Individual-level Have.1.SG bike bike.DEF.ART. ‘I own a bike.’ b. Am bicicleta (cu mine). Stage-level Have.1.SG bike.DEF.ART. (with me) ‘I have the bike (with me).’ The Romanian data pattern with the Spanish ones: while the complement of the individual level have always stays unmarked (49a), the complement of the stage-level have allows for the use of DOM (49b): (49) a. Maria are (*pe) o soră. Mary has (*DOM) INDEF.ART. sister ‘Mary has a sister.’ b. Am pe o soră (de-a lui Mihai la mine Have.1.SG DOM CL.3.SG sister of Michael at me

36

2 Romanian direct objects. A comparative analysis with Spanish

luna asta). month this ‘I have one of Michael’s sisters living with me this month.’ The individual level reading of a avea may only select a property-denoting complement, not allowing a pe marked DP as its complement, since such DPs are never property denoting. The complement of the stage-level a avea may be pe marked in this case. This is so because the stage-level a avea denotes a property of events (type and its complement is syntactically represented as a small clause containing a subject DP and a spatio-temporal predicate. As such, the subject DP may be of e or type, which allows the use of pe. Thus, pe marked DPs may never be used with individual-level a avea as it only selects property-denoting complements, but they may be used as complements of the stage level a avea as e or type DPs are allowed as subject of the small clause selected by this verb. In this respect, we may view the discussion above as an argument that pe marked nominals are not property denoting but real arguments. Another argument provided by Bleam in favour of the fact that the a marker is a denotation filter comes from the realm of the so called non-incorporating verbs such as love/hate, which do not allow a property denoting NP complement. Following Carlson (1977), Bleam notices that the bare plural objects of these verbs are always generic and never acquire an existential interpretation (Kim hates apples/ Kim loves linguists.). This might be explained if one assumes that the verbs in question do not allow a property denoting NP complement. This is borne out by Romanian data where bare plurals are not allowed as complements of love/hate:11 (50) *Ion urăște profesori. John hate.3.SG. teachers ‘John hates teachers.’ These non-incorporating verbs may take definite and pe marked complements: (51) Ion urăște profesorii/ pe profesori. John hate.3.SG. teachers.DEF.ART. DOM teachers ‘John hates teachers.’

11 This is also pointed out in Cornilescu (2000).

2.5 Interim conclusions

37

Finally, interactions with various operators may also give us a clue as to the semantic type of pe marked DPs. As already discussed in the previous sections, the mood of the relative clause modifying an NP signals whether the nominal is interpreted as opaque or transparent in intensional contexts: if the nominal is interpreted as transparent, the relative clause modifying it will be in the indicative mood. If the nominal is opaque, then the modifying relative clause will be in the subjunctive mood. Pe indefinites may be interpreted either as opaque or as transparent and this is why they are compatible with both a subjunctive and an indicative relative clause. The fact the pe DPs may be modified by an indicative relative clause shows that these DPs are strong i.e., have argumental denotations of type e or . Furthermore, the interaction with intensional verbs such as a vrea (want), a dori (wish) may also strengthen this argument: bare singulars, which are property denoting nominals, are opaque in this context. When pe marked nominals are employed in the same context, they may be interpreted as transparent: (52) Ion dorește menajeră vs. Ion dorește pe o John wishes house keeper. John wishes DOM INDEF.ART. ‘John wants a house keeper.’ ‘John wants a house keeper from this menajeră din acest hotel. house keeper from this hotel hotel.’ Apparently, the formal type-theoretic properties of nominals are relevant when it comes to pe marking.12 The presence of pe places constraints on the denotations of the DPs it marks. The DPs marked in this way always have an objectlevel reading and, as for their specific denotations, these DPs always select only argument denotations i.e. e (i.e. object) or (i.e. generalized quantifier). On the other hand, these DPs never have a property reading i.e. .

2.5 Interim conclusions The four sections above have uncovered a number of similarities and differences regarding the behaviour of marked and unmarked DOs in Romanian and Spanish. Firstly, in both Spanish and Romanian, DOMed DOs may evince specific

12 The Romanian data are confirmed by Cornilescu (2000) who, following Bleam (1999), argues that pe marked DPs are never property denoting.

38

2 Romanian direct objects. A comparative analysis with Spanish

as well as non-specific readings, exhibiting a propensity for the former interpretation. Unmarked objects, on the other hand, contrast in that while Spanish DOs never acquire specific readings, their Romanian counterparts may give rise to such an interpretation. Furthermore, it seems that neither language has a class of objects that are always interpreted as specific, which points to the fact that specificity as such is not directly encoded in these languages.13 With respect to DO scope taking potential, the two languages pattern alike when it comes to marked DOs, which seem to be able to ouscope other scope bearing expressions. On the other hand, the same contrast is at stake with respect to unmarked indefinite DOs: Spanish DOs may not outscope Subject QPs, conditionals, or the negative operator, while Romanian unmarked DOs may do so. Finally, the language specific counterparts of have in both Spanish and Romanian never allow marked DOs in their individual level reading, being able to only combine with property denoting nominals, while the stage level reading of the respective verbs allow marked DOs. This points to the fact that marked DOs only have argument denotations i.e. e (object) or ≪ e,t>t> (generalized quantifier). On the other hand, these DPs never have a property reading i.e. .

2.6 Syntactically triggered DOM Drawing away from the semantic aspects involving scope and specificity, López (2012) considers some syntactic phenomena where the use of DOM seems to be compulsory in the absence of any possible semantic trigger. Small clause complements represent one relevant case in this respect: in (53a), the argument of

13 As observed by López (2012), one such language might be Persian, where indefinites marked by means of the DOM -râ are necessarily specific. In (1), where the indefinite monši (assistant) has a non-specific reading, the use of -râ is disallowed: (1)

(*-râ) mi-gard-e. Jān dombāl-e monšii John search-EZ assistant (*DOM) CONT-look-3SG ‘John is looking for an assistant.’ Ui bāyad ālmāni be-dun-e. she/he must German SUBJ-know-3SG ‘She/he must know German.’

(López 2012: 14, p. 109)

2.6 Syntactically triggered DOM

39

the small clause complement is obligatorily a marked. Unmarked indefinites or bare plurals (which reject DOM) are discarded as ungrammatical: (53) a. Considero a/*ø un estudiante inteligente. Consider.1.SG. a/*ø INDEF.ART. student intelligent ‘I consider a student intelligent.’ (López 2012: 19, p. 10) b. *El profesor consideró a/ø estudiantes inteligentes DEF.ART. professor considered DOM/ø students intelligent ‘The professor considered students intelligent.’ (López 2012: 57, p. 23) Note that the DOM requirement in this particular case does not arise from any specificity considerations. (54) contains non-specific affectees, which are nonetheless necessarily marked. The use of the subjunctive mood in the relative ensures that the specific reading is out: (54) Juan no considera honrado a/*ø un hombre John not consider.3.SG honest DOM/*ø INDEF.ART. man que acepte sobornos. that accepts bribes ‘Juan does not consider honest a man that accepts bribes.’ (López 2012: 68, p. 25) The small clause data in Romanian pattern with the Spanish ones up to a certain extent. While it is true that preference is given to the DOMed marked nominals, the unmarked ones are also accepted (55b). Bare plurals, on the other hand, are rejected as ungrammatical (55c): (55) a. (Îl) consider pe un student inteligent. CL.3.ACC.SG consider.1.SG DOM INDEF.ART. student inteligent ‘I consider a student intelligent.’ b. Consider un student inteligent. consider.1.SG INDEF.ART. student inteligent ‘I consider a student intelligent.’ c. *Consider studenți inteligenți consider.1.SG students intelligent.PL ‘I consider students intelligent.’

40

2 Romanian direct objects. A comparative analysis with Spanish

The use of unmarked indefinites in this context seems quite spread. Here are some relevant examples that came out as a result of a simple Google search: (56) a. Eu consider unii posesori de R1 și R6 cazuri speciale. I consider.1.SG some owners of R1 and R6 cases special.PL ‘I consider some owners of R1 and R6 to be special cases.’14 b. Și eu consider multe fete de aici prietene. And I consider.1.SG many girls of here friends ‘I also consider many girls here as my friends.’15 c. Mă refer la faptul că eu consider câțiva oameni Me refer.1.SG. at fact that I consider.1.SG some people capabili să urce pe tronul României. capable.PL SUBJ. mount on throne.DEF.ART. Romania.GEN ‘I refer to the fact that I consider some people capable to lead Romania.’16 Note further that both marked and unmarked indefinites are allowed as small clause arguments when modified by a relative containing subjunctive, with a preference for the unmarked variant: (57) Ion nu ar considera John not would.1.SG consider care să accepte mită de la who SUBJ. accept bribe from ‘John would not consider to be bribes from the electors.’

cinstit (pe) un politician honest (DOM) INDEF.ART. politician alegători. electors honest a politician who would accept

The Romanian examples discussed above disconfirm the analysis proposed in López (2012) for this language. Note that he builds this analysis on inaccurate data: in (58a, b) it is argued that only pe marked DPs are allowed:

14 (http://www.motociclism.ro/forum/index.php?/topic/747673-motocicleta-moarta-debatranete/page-3), Access date: 10.06.2018 15 (http://www.culinar.ro/forum/continut/pentru-cei-mici/26350/ingrediente-mai-mult-saumai-putin-nocive-ptr-copii/page-2), Access date: 10.06.2018 16 (https://forum.softpedia.com/topic/873827-ce-vei-vota-la-referendum/page__st__612), Access date: 10.06.2018

2.6 Syntactically triggered DOM

41

(58) a. Ion (îl) consideră pe/*ø student tâmpit. John CL.3.ACC.SG. consider.3.SG. DOM/*ø student stupid ‘John considers the student stupid.’ b. Obama (l)-a numit pe/*ø senator Obama CL.3.ACC.SG.-has.3.SG. appointed DOM/*ø senator ministrul sănătății. minister health.GEN ‘Obama designated the senator Secretary of Health.’ (López 2012: 119, p. 143) An unmarked indefinite is, in fact, quite felicitous in this context, as also shown above: (59) Ion consideră un student tâmpit. John consider.3.SG. INDEF.ART. student stupid ‘John considers the student stupid.’ (López 2012: 119, p. 143) The object of an object control predicate also requires the use of obligatory a marking: (60) Juan forzó a/*ø un niño a hacer los John forced DOM/*ø INDEF.ART. child to do DEF.ART. deberes. homework ‘John forced a child to do his homework’ (López 2012: 67, p. 25) Again, the use of DOM does not obey specificity requirements as such marked objects may be modified by relatives containing a subjunctive. (61) María forzaria a/*ø una empleada que Maria force.COND DOM/*ø INDEF.ART. employee.FEM that tuviera depresion a venir al trabajo. had.SUBJ depression to come to work ‘Mary would force an employee who were depressed to come to work.’ (López 2012: 71, p. 25) López (2012) claims that objects in Romanian object control configurations behave similarly to their Spanish counterparts and provides the following example:

42

2 Romanian direct objects. A comparative analysis with Spanish

(62) Ion l-a forțat pe/*ø un băiat John CL.3.ACC.SG.-has.3.SG forced DOM/*ø INDEF.ART. boy să-i facă temele. SUBJ.-CL.3.DAT.SG. do homework.DEF.ART.PL ‘John forced a boy to do his homework.’ (López 2012: 121, p. 144) We find the use of an unmarked indefinite perfectly acceptable in this context: (63) Ion a forțat un băiat să-i John has.3.SG forced INDEF.ART. boy SUBJ.-CL.3.DAT.SG. facă temele. do homework.DEF.ART.PL ‘John forced a boy to do his homework.’ Thus, just like with small clauses, Romanian objects in object control contexts may bear DOM. Nevertheless, marking is not obligatory as unmarked objects are also allowed (64a). Bare plurals, on the other hand, are rejected (64b). (64) a. Șeful a foțat (pe) mulți angajați să Boss.DEF.ART. has.3.SG forced (DOM) many employees SUBJ. rămână la serviciu peste program. remain at work over schedule ‘The boss forced many employees to do overtime.’ b. ?Șeful a foțat angajați să rămână la Boss.DEF.ART. has.3.SG forced employees SUBJ. remain at serviciu peste program.17 work over schedule ‘The boss forced many employees to do overtime.’ Further examples from various websites confirm the differences holding between the behaviour of unmarked indefinites in this context in Spanish and Romanian:

17 This example might be acceptable under a reading according to which one explains the more recent actions of the boss in questions along the lines of (1): (1)

So what‘s new about your boss? He forced employees to do overtime.

2.6 Syntactically triggered DOM

43

(65) Angajatorul a forțat un angajat să Employer.DEF.ART. has.3.SG. forced INDEF.ART. employee SUBJ. își dea demisia. REFL. give resignation.DEF.ART. ‘The employer forced an employee to resign.’18 (66) . . . a forțat mulți părinți să transfere altora Has.3.SG. forced many parents SUBJ. transfer others îndatoririle parentale. duties.DEF.ART. parental.PL. ‘ . . . has forced many parents to transfer their parental duties to others.’ (Munteanu 2016: 113) A third syntactic context where DOM is required in Spanish is that provided by accusative affected arguments in clause union. In the causative construction presented under (67) below, the causee in the accusative is necessarily DOMed. An unmarked indefinite or a bare plural is out. (67) a. María hizo llegar tarde a/*ø Mary made arrive late DOM/*ø ‘Maria made a boy arrive late.’ b. * María hizo llegar tarde a/ø Mary made arrive late DOM/ø ‘Maria made boys arrive late.’

un niño. INDEF.ART. child niños. boys (López 2012: 58, p. 24)

Permissive constructions seem to obey the same restrictions in requiring a marking of the affectee and rejecting unmarked indefinites and bare plurals: (68) María dejó trabajar los domingos a/ø un Maria let work DEF.ART. Sundays DOM/ø INDEF.ART. empleado. employee ‘Mary allowed an employee to work on Sundays.’ (López 2012: 62, p. 24)

18 (https://www.avocatnet.ro/forum/discutie_593913/Poti-sa-demisionezi-si-fara-preaviz. html*1), Access date: 10.06.2018

44

2 Romanian direct objects. A comparative analysis with Spanish

Again, a marking is not triggered by specificity considerations as may be seen from (69, 70), where the relative clause modifying the affectee contains the subjunctive mood. (69) María hace quedarse en clase a/*ø un niño que Mary does stay.INF in class DOM/*ø INDEF.ART boy who no haya terminado les deberes. not has.SUBJ finishes DEF.ART. duties ‘Maria makes a boy that has not finished the assignment stay in class.’ (70) María no dejaría salir a/*ø nungún niño que Mary not let.COND leave.INF. DOM/*ø no boy who no haya terminado los deberes. not has.SUBJ. finished DEF.ART. duties ‘Maria would not let any boy who has not finished the assignments go out.’ (López 2012: 69, 70, p. 25) López (2012) makes similar claims about Romanian: in (71), DOM is argued to be obligatory: (71) Ion l-a lăsat pe/*ø un copil John CL.3.ACC.SG.-has.3.SG let DOM/*ø INDEF.ART. child să joace Nintendo. SUBJ. play Nintendo. ‘John let a child play Nintendo.’ (López 2012: 120, p. 144) This, is nevertheless, not true, as unmarked indefinites are acceptable in these configurations: (72) Ion a lăsat un copil să joace Nintendo. John has.3.SG let INDEF.ART. child SUBJ. play Nintendo. ‘John let a child play Nintendo.’ Note, however, that Romanian causative constructions actually amount raising to object sentences and, as such, they are not relevant for the current discussion. In the causative-permissive structure, for instance, the use of DOMed and unmarked affectees is allowed, as well as that of bare plurals, contrary to what is argued in López (2012):

2.6 Syntactically triggered DOM

45

(73) a. A lăsat (pe) un străin să-i Has.3.SG. let (DOM) INDEF.ART. stranger SUBJ.-CL.3.DAT.SG. intre în casă, fără să se gândească la consecințe. enter in house without SUBJ. REFL. think of consequences ‘He let a stranger enter his house without thinking about the consequences.’ b. A lăsat străini să-i intre în casă, Has.3.SG. Let strangers SUBJ.-CL.3.DAT.SG. enter in house fără să se gândească la consecințe. without SUBJ. REFL. think of consequences ‘He let a stranger enter his house without thinking about the consequences.’ López (2012) also draws attention to the behaviour of perception verbs in clause union structures, which follow the same pattern in only allowing marked indefinites and disallowing their unmarked counterparts and bare plurals: (74) María vio cae a/*ø un niño. Mary saw fall DOM/*ø INDEF.ART. boy ‘Mary saw a boy fall.’

(López 2012: 64, p. 24)

Romanian perception verbs allow both marked and unmarked indefinites. Bare plurals are also permitted. (75) a. Am văzut (pe) un vecin lucrând Have.1.SG seen (DOM) INDEF.ART. neighbour working și duminica. and Sunday ‘I have seen a neighbour working even on Sundays.’ b. Am văzut vecini lucrând și duminica. Have.1.SG seen neighbour working and Sunday ‘I have seen a neighbour working even on Sundays.’ Another obligatory context where DOM is required in Spanish is that supplied by object wh-phrases irrespective of whether this object moves to SpecC or remains in-situ as in multiple wh-sentences.

46

2 Romanian direct objects. A comparative analysis with Spanish

(76) a. ¿ A/*ø quién has llamado? DOM/*ø who have.2.SG called ‘Who did you call?‘ b. ¿ Quién dijiste que llamó a/*ø quién? Who said.2.SG that called DOM/*ø who ‘Who did you say called who?’ (López 2012: 72, 73, p. 26) The same holds for Romanian wh- object phrase care (who/which), where DOM is obligatory (Dobrovie-Sorin 1994): (77) Dacă ar fi să optezi între rochia roșie și If would be SUBJ. opt between dress.DEF.ART. red and cea albastră, pe care ai alege-o? DEF.ART. blue DOM which would.2.SG choose-CL.3.ACC.SG ‘If you were to select between the red and the blue dress, which would you choose?’ This kind of phenomena, where specificity does not represent a requirement for the use of DOM, prompt López to propose that the mechanism as such might not be completely ruled in by the definiteness and the animacy scales (Aissen 2003) but that the phonological realization of marking comes as a consequence of the suitable environmental conditions in which the DO may find itself and that scrambling represents a prerequisite for these conditions. DOM is thus argued to be the morphological expression of a syntactic configuration (López 2012, p. 28). With respect to the structures presented above, Spanish seems to draw a distinction between marked indefinite DOs on the one hand and unmarked indefinites and bare plurals on the other: while the former seem to be required in these contexts, the latter are rejected. Romanian patterns with Spanish up to a certain extent in allowing DOMed indefinite DOs in these configurations and disallowing bare plurals. Nevertheless, Romanian also allows unmarked indefinites to occur in small clauses, clause union and object control configurations, as opposed to Spanish. Note also that there is a further distinction between unmarked indefinites and bare plurals with respect to scope in Romanian: while, as seen above, unmarked indefinite DOs may outscope various scope taking expressions, contrary to their Spanish counterparts, bare plurals may not do so. In (78) below, the bare plural colleagues may not outscope the conditional: there exists no particular group of colleagues such that if Mary happens to invite them, Michael will get annoyed.

2.7 Interim conclusions

47

(78) Dacă Maria invită colegi de serviciu la petrecere, If Mary invite.3.SG. colleagues of work at party Mihai se va supăra. Michael REFL. will get upset ‘If Mary invites colleagues from work to the party, Michael will get annoyed.’ The same goes for example (79), where the bare plural cannot outscope the universal quantifier: a reading according to which there is a group of messengers, such that the ruler of the country sent to all the kings in the neighbouring countries, is not possible. The only possible interpretation is one in which the messengers distribute function of the kings. (79) Domnul țării a trimis soli tuturor Ruler.DEF.ART. country.GEN has.3.SG sent messengers all.DAT împăraților vecini pentru a cere ajutor împotriva kings.DAT neighbor.PL for to ask help against cotropitorilor. invaders.GEN ‘The ruler sent messengers to all the kings in the neighbourhood to ask for help against the invaders.’ Thus, the partitioning that Romanian seems to operate between marked DOs and unmarked DOs on the one hand and bare plurals on the other hand stays coherent throughout all the cases discussed so far. In Spanish, the split differentiates between marked DOs on the one hand and unmarked counterparts and bare plurals on the other.

2.7 Interim conclusions The first part of the current chapter focused on a comparison between Romanian and Spanish DOs using López (2012) as a starting point for this discussion. Both Spanish and Romanian DOMed DOs seem to pattern alike in allowing for a specific interpretation and a wide scope reading; the non-specific and the narrow scope interpretations remain an option with these DPs however. The two languages differ with respect to the behaviour of unmarked DOs: while these DPs always get interpreted non-specifically and may only exhibit a narrow scope reading when co-occurring with other scope bearing expressions, their Romanian counterparts pattern with the DOMed DOs, allowing specific as well as wide scope interpretations.

48

2 Romanian direct objects. A comparative analysis with Spanish

A further point of similarity between Romanian and Spanish marked DOs has to do with the occurrence of these DPs within the context of have: while this verb only enables co-occurrence with unmarked DOs in its individual level reading, marked DOs are allowed with its stage level variant. These restrictions point, in turn, that marked DOs have real argumenthood and never denote properties. López (2012) further discusses three contexts which require the use of DOM in Spanish: small clauses, object control and clause union in causative configurations. These are, however, constructions where no specificity requirements are at stake and which, as a consequence, prompt López (2012) to propose that DOM is not completely governed by definiteness/animacy requirements, but that a more syntactic approach might be in order. As far as the Romanian data are concerned, we noted an important difference with respect to Spanish in that not only DOMed indefinites are allowed in these contexts, but that unmarked indefinites seem to also be acceptable. Bare plurals, similarly to their counterparts in Spanish, are rejected from these configurations. Thus, while Spanish draws a line between DOMed objects on the one hand and unmarked and bare nominals on the other, Romanian seems to group marked and unmarked indefinite objects together when it comes to the aforementioned contexts and to distinguish between these DPs on the one hand and bare plurals on the other. The next part of this chapter shows a more detailed picture of the behaviour of Romanian marked and unmarked DOs by building on some new experimental findings uncovering the scope abilities of DOs in the context of various scope taking expressions (Universal QPs, Conditional, Modal Verbs, negation). This will enable us to better capture the differences in behaviour holding between Spanish and Romanian DOs. The uncovered facts will, in turn, tell us more about the syntax of unmarked DOs: if, as proposed by López (2012), the availability of a wide scope interpretation comes as a consequence of a certain mode of semantic composition of the object with its predicate, which is in turn dependent on a peculiar syntactic position (either within VP or in SpecαP), then the availability of a wide scope interpretation for Romanian unmarked DOs would point to their capacity of leaving their merge position within VP. Thus, on a par with DOMed DOs, unmarked DOs would be allowed to reach SpecαP, unlike their Spanish counterparts. A proposal as to how one should fine-grain the account in López (2012) so as to accommodate for the behaviour of Romanian unmarked DOs will be advanced in a subsequent chapter.

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

49

2.8 Reinforcing the conclusions: two experiments on Romanian DOs 2.8.1 Scope dependencies with marked and unmarked DOs When comparing the behaviour of Spanish indefinite DOs regarding their interaction with other scope bearing expressions to that of their Romanian counterparts, we noticed a clearcut difference regarding unmarked DOs: while these DPs may not outscope extensional QPs, conditionals, modal verbs or negation in Spanish and hence may only acquire a narrow scope interpretation, for Romanian unmarked DOs wide scope readings seem to be available. In this, Romanian unmarked DOs pattern with DOMed DOs. Such a difference of behaviour, if attested, would tell us more about the syntax of unmarked DOs. The experiment described in this section serves this very particular purpose of verifying whether one may differentiate between Romanian marked and unmarked DOs with respect to their behaviour when used in the context of other scope bearing expressions. 2.8.1.1 Design In this experiment we checked the behaviour of indefinite DOs within four contexts: that of the universal QP fiecare (everybody), the conditional, the modal verb trebuie (must), and negation. For each context, 12 relevant sentences were created. Each sentence had three variants of its own due to the fact that the indefinite DO employed could be of three types: unmarked DO, DOMed DO and CDed+DOMed DO. We hence ended up having a total of 144 items in need of testing, as shown in Table (1) below:

Table 1: Tested items and contexts. every QP

Conditional

Modal

Negation

unmarked DO









DOMed DO









CDed+DOMed DO









The items were further distributed into three different questionnaires using the Latin square method for an even distribution. To each questionnaire 30 fillers were added, grouped into 10 expectedly unacceptable items, 10 completely

50

2 Romanian direct objects. A comparative analysis with Spanish

acceptable items and 10 average items with respect to acceptability (the fillers were separately checked for acceptability in a smaller, informal experiment). Each questionnaire ended up having a number of 78 items. The questionnaires were formatted as Google online forms in such a way that the potential respondent could only access one item at a time, without having the possibility of going back or forth. The experiment was carried out in two stages: an initial norming experiment probing for the interpretation (i.e., wide scope vs. narrow scope) assigned to the indefinite DO in each of the tested items, and a second experiment where the respondents were required to assess the items on a 7-rung scale (with 1 being the lowest acceptability score). Each questionnaire was assessed by at least 20 native speakers of Romanian such that more than 60 people took part in the first stage of the experiment, followed by other (at least) 60 respondents who participated in the second stage of the experiment. The results thus obtained were then verified and outliers were removed i.e., questionnaires in which more than 7 fillers had been wrongly assessed. Remaining questionnaires entered statistical analysis. The examples under (80) – (83) below represent sample items tested during the experiment proper: Universal QP (80) a. În urma sondajului s-a dovedit că fiecare Following survey.DAT REFL.-has.3.SG proved that every cetățean român respectă puțini miniști din citizen Romanian respects few ministers from actualul guvern. current.DEF.ART. government ‘The survey showed that every Romanian citizen respects few ministers from the current government.’ – unmarked DO b. În urma sondajului s-a dovedit că fiecare Following survey.DAT REFL.-has.3.SG proved that every cetățean român respectă pe puțini miniști din citizen Romanian respects DOM few ministers from actualul guvern. government current.DEF.ART. ‘The survey showed that every Romanian citizen respects few ministers from the current government.’ – DOMed DO

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

51

c. În urma sondajului s-a dovedit că fiecare Following survey.DAT REFL.-has.3.SG proved that every Îi respectă pe puțini miniști cetățean român citizen Romanian CL.3.ACC.SG. respects DOM few ministers din Actualul guvern. from current.DEF.ART. government ‘The survey showed that every Romanian citizen respects few ministers from the current government.’ – unmarked DO Conditional (81) a. Dacă Mihai invită un coleg la masă, If Michael invite INDEF.ART. colleague to dinner soția lui pleacă de acasă. wife.DEF.ART. his.POSS leave.3.SG from home ‘If Michael invites a colleague to dinner, his wife leaves home.’ – unmarked DO b. Dacă Mihai invită pe un coleg la masă, If Michael invite DOM INDEF.ART. colleague to dinner soția lui pleacă de acasă. wife.DEF.ART. his.POSS leave.3.SG from home ‘If Michael invites a colleague to dinner, his wife leaves home.’ – DOMed DO c. Dacă Mihai îl invită pe un coleg If Michael CL.3.ACC.SG. invite DOM INDEF.ART. colleague la masă, soția lui pleacă de acasă. To dinner wife.DEF.ART. his.POSS leave.3.SG from home ‘If Michael invites a colleague to dinner, his wife leaves home.’ – CDed+DOMed DO

Modal (82) a. La conferința anuală, Departamentul de Engleză trebuie At conference annual Department.DEF.ART. of English must să invite un lingvist de seamă. SUBJ. invite INDEF.ART. linguist of note ‘The English department must invite a famous linguist to their annual conference.’ – unmarked DO

52

2 Romanian direct objects. A comparative analysis with Spanish

b. La conferința anuală, Departamentul de Engleză trebuie At conference annual Department.DEF.ART. of English must să invite pe un lingvist de seamă. SUBJ. invite DOM INDEF.ART. linguist of note ‘The English department must invite a famous linguist to their annual conference.’ – DOMed DO c. La conferința anuală, Departamentul de Engleză trebuie At conference annual Department.DEF.ART. of English must să-l invite pe un lingvist de seamă. SUBJ.-CL.3.ACC.SG invite DOM INDEF.ART. linguist of note ‘The English department must invite a famous linguist to their annual conference.’ – CDed+DOMed DO Negation (83) a. Directorul nu a concediat câțiva angajați. Manager.DEF.ART. not has fired some employees ‘The manager did not fire some employees.’– unmarked DO b. Directorul nu a concediat pe câțiva angajați. Manager.DEF.ART. not has fired DOM some employees ‘The manager did not fire some employees.’– DOMed DO c. Directorul nu i-a concediat pe câțiva Manager.DEF.ART. not CL.3.ACC.SG.-has fired DOM some angajați. employees ‘The manager did not fire some employees.’ – CDed+DOMed DO As already mentioned, the experiment was carried out in two stages: we started with a norming experiment probing for the interpretation i.e., wide scope vs. narrow scope assigned to the indefinite DO in each of the tested items. In (84), each tested item is followed by a set of questions eliciting the available interpretations for the respective indefinite DO; thus, for each type of DO, we wanted to find out whether a narrow scope or wide scope interpretation (or both) is possible. (84) a. Fiecare student care frecventează Biblioteca Every student who frequent.3.SG Library.DEF.ART.

53

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

Academiei a întâlnit câțiva academicieni de renume. Academy.GEN has met some academicians of renown ‘Every student who frequents the Library of the Academy has met some renowned academicians.’ From the fragment above I understand: i.

The academicians who were met differed from student to student

yes

no

ii. All students frequenting the library of the Academy have met the same academicians

yes

no

iii. Both (i) and (ii)

yes

no

b. Dacă l-aș vedea pe un profesor If CL.3.ACC.SG.-would.1.SG. see DOM INDEF.ART. professor că se comportă urât cu studenții, that REFL. behave.3.SG badly with students.DEF.ART. m-aș plânge la decan. REFL.-would.1.SG complain to dean ‘If I were to see a professor behaving badly towards students, I would complain to the dean.’ From the fragment above I understand: i.

If I were to see a professor behaving, badly towards students, no matter who that might be, I would complain to the dean.

yes

no

ii. If a certain professor were to behave badly towards students, I would complain to the dean.

yes

no

iii. Both (i) and (ii)

yes

no

In a separate experiment, we tested the acceptability of the items used in the first experiment. In this particular case, the respondents were required to assess the items on a scale ranging from 1 to 7 (one being the lowest acceptability score).

2.8.1.2 Results and discussion When considering the acceptability19 scores assigned to the items in the four tested contexts, we notice that they range towards the higher end of the seven-rung

19 See Appendix 4 for the statistics.

54

2 Romanian direct objects. A comparative analysis with Spanish

acceptability scale, with values over the GM 4,5. Sentences containing DOMed DOs without CD seem to fare somewhat worse than their counterparts which, as also noticed with other experiments,20 seems to be a general trend, most probably due to the fact that the grammaticalisation of CD+DOM as one marking mechanism is quite advanced (see Avram 2014, Dobrovie-Sorin 1994 for a discussion in this respect): Table 2: Acceptability values. CD+DOM DO

DOM

unmarked DO

Universal QP

,

,

,

Conditional

,

,

,

Modal

,

, 

,

Negation

,

,

,

Acceptability Scope DO

Universal QP

Conditional

Modal

unmarked DO

DOMed DO

Cded+DOMed DO

unmarked DO

DOMed DO

Cded+DOMed DO

unmarked DO

DOMed DO

Cded+DOMed DO

unmarked DO

DOMed DO

Cded+DOMed DO

7 6 5 4 3 2 1

Negation

Graph 1: Acceptability scores.

The norming experiment, which accompanied the one probing for acceptability, uncovered a different behaviour of the three types of DO function of the context of occurrence at stake. Thus, when co-occurring with the extensional universal QP fiecare, unmarked and DOMed DOs seem to strongly favour a narrow scope reading, while CDed+DOMed DOs are preferably interpreted as having wide scope (still allowing, however, the possibility of a narrow scope interpretation). This difference between CDed+DOMed DOs on the one hand and DOMed and unmarked DOs on the other hand is no longer apparent when these DPs occur in

20 The observation will be extended upon in the next sections.

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

55

the context of intensional operators. With conditionals, for instance, all the three DO types exhibit a strong preference for the narrow scope interpretation, with over 65% of the respondents considering that this is the only way of interpretation available. The wide scope reading remains, however, available and the percentages in this respect are quite similar for all DO types with somewhat lower scores for unmarked DOs: CDed+DOMed: 17%+18%, DOMed DO: 12%+19%, unmarked DO: 11,6%+ 11,77%.21 The tendency towards a narrow scope interpretation is maintained for all DO types in the context of the modal verb trebuie (must). The wide scope interpretation seems, however, to be more readily available in this case, with 31%+17% of respondents choosing this reading for CDed+DOMed DOs, 27%+21% selecting it for DOMed DOs and 16,37%+15% opting for it in the case of unmarked DOs (see Table 3 below). Table 3: GMs contexts and DO types.

Universal QP

Wide scope

Narrow scope

Both

CDed+DOMed DO

%

%

%

DOMed DO

%

%

%

,%

,%

,%

CDed+DOMed DO

%

%

%

DOMed DO

%

%

%

,%

,%

,%

%

%

%

unmarked DO Conditional

unmarked DO Modal

CDed+DOMed DO

%

%

%

,%

,%

,%

CDed+DOMed DO

%

%

%

DOMed DO

%

%

%

,%

,%

,%

DOMed DO unmarked DO Negation

unmarked DO

Interestingly, when interacting with the negative operator, all DO types appear to strongly favour a wide scope interpretation, with over 70% of the respondents

21 The first figure represents the percentage of respondents who considered the wide scope reading as the only acceptable reading, while the second percentage represents the proportion of respondents for whom both the narrow scope reading and the wide scope one were available.

56

2 Romanian direct objects. A comparative analysis with Spanish

selecting this reading as the only available option. The narrow scope reading remains possible, with unmarked DOs seemingly more inclined towards it: CDed+DOMed: 5%+8%, DOMed DO: 8%+9%, unmarked DO: 13, 98%+ 14, 23% (see Appendix 4 for the statistics).

DO scope

87%

83%

76.59%

Universal QP

Conditional Series1

27% 21%

16.37% 15.00%

Modal Series2

5% 8%

8% 9%

13.98% 14.23% unmarked DO

CDed+DOMed DO

17%

unmarked DO

DOMed DO

CDed+DOMed DO

unmarked DO

19% 12% 11.64% 11.77%

52%

DOMed DO

31% 23% 20.18% 17% 19% 16.66% 18%

71.78%

CDed+DOMed DO

52%

unmarked DO

66%

68.63%

69%

DOMed DO

57%

63.15%

DOMed DO

CDed+DOMed DO

100% 90% 80% 70% 60% 50% 40% 44% 42% 30% 20% 10% 14% 0%

Negation

Series3

Graph 2: Norming – Series 1: wide scope; Series 2: narrow scope; Series 3: both.

According to the experimental results, the behaviour of the three types of DO differs function of the scope bearing expression with which these DPs co-occur. Even so, what is important to notice is the fact that there are no contexts in which any of the DO types may not have a wide scope interpretation. All DO types seem able to outscope universal QPs, the conditional, the modal trebuie and the negation operator even if to a different extent. 2.8.1.3 Conclusions This section dwelled on some experimental findings concerning the scope behaviour of unmarked and marked DOs in an effort to strengthen the intuitions extended upon in the previous sections and according to which Romanian DOs differ from their Spanish counterparts, contrary to what is being claimed in López (2012). The results show that all types of DOs allow both for a narrow scope and a wide scope interpretation and that preference for one or the other reading is actually given by the type of scope bearing expression with which they interact:

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

57

when co-occurring with the universal QP fiecare, there seems to be a distinction between the behaviour of CDed+DOMed DOs on the one hand, which seem to favour a wide scope interpretation, and the behaviour of unmarked and DOMed DOs on the other hand, which seem more prone towards a narrow scope reading. The opposite readings remain, however, an option: CDed+DOMed DOs may be interpreted as having narrow scope (42% + 14%). In their turn, unmarked and DOMed DOs may allow for a wide scope interpretation (16,66% + 20,18% and 23% + 19% respectively). With intensional operators (conditional and modal must) on the other hand, this difference is no longer apparent, all the three types of DOs clearly evincing a strong preference for the narrow scope reading. The wide scope interpretation remains, nevertheless, an option, which is quite significant for those contexts where the DO co-occurs with the modal trebuie (must): CDed+DOMed: 17%+18% (Conditional) and 31%+17% (Modal); DOMed DO: 12%+19% (Conditional) and 27%+21% (Modal); unmarked DO: 11,6%+ 11,77% (Conditional) and 15%+15% (Modal). Finally, in the context of the Negation operator, all types of DO seem to strongly favour a wide scope interpretation. The narrow scope interpretation, on the other hand, appears to remain a possibility, but this is quite small with CDed+DOMed DOs and DOMed DOs and higher with unmarked DOs. When considering the general behaviour of these indefinite DOs in the four tested contexts, what we notice is that marking seems to render the respective DO a better candidate for outscoping another scope bearing expression. Nevertheless, the narrow scope interpretation remains an option. Conversely, lack of marking with the DO seems to tilt the balance towards a narrow scope interpretation, which does not, however, preclude the possibility for a wide scope interpretation. In this, Romanian unmarked DOs behave differently from their Spanish counterparts as depicted in López (2012). Marked DOs in Romanian pattern with their Spanish counterparts in allowing both a strong and a weak reading.

2.8.2 An experiment on specificity with marked and unmarked DOs Both DOMed and CDed+DOMed indefinites represent possible DO marking strategies in Romanian and have been argued to evince a non-ambiguous specific interpretation. Thus, Farkas (1987), Dobrovie-Sorin (1990, 1994) and Cornilescu (2000) argue that pe induces specificity. On the other hand, Steriade (1980) and Gierling (1997) point to the correlation between specificity and clitic doubling. Tigău (2011), Chiriacescu & von Heusinger (2011a, b) consider that specificity is a joint effect of pe and CD and Tigău (2016) stresses on the functional unity of CD+DOM as a

58

2 Romanian direct objects. A comparative analysis with Spanish

clause-level construction having a dedicated semantics and pragmatics related to integrals (Hornstein et al. 1995), Uriagereka 2001, 2002, 2005). The purpose of this section is to present an experiment on Romanian DOs, whose results point to the fact that neither the DOM marker, nor the pronominal clitic within CD structures is a specificity trigger. More specifically, both DOMed DOs and their CDed + DOMed counterparts seem to be able to occur in contexts forcing non-specific readings where they lose their specific interpretation. We test the behaviour of unmarked, DOMed and DOMed+CDed indefinite DOs when inserted in four contexts that have been long acknowledged to force a nonspecific interpretation. The contexts were built around the following expressions: cel mult (at most)/cel puțin (at least), the distributive câte, the free choice indefinite oarecare (any), relative clauses modifying the indefinite DO and containing a predicate in the subjunctive mood.22 Consider first the context where an indefinite DO is preceded by cel mult (at most)/cel puțin (at least). In (85a), the indefinite doi prieteni (two friends) allows both a reading according to which John visits the same two friends every evening, as well as one where the set of two friends varies from one evening to the next. When the indefinite is preceded by cel mult (at most), as in (85b), the only available reading for the indefinite is the non-specific and narrow scope one. As it seems, the cel mult/cel puțin phrases disambiguate between the two possible specific and the non-specific interpretations in favour of the latter. (85) a. Ion vizitează doi prieteni în fiecare seară. John visits two friends in every evening ‘John visits two friends every evening.’ b. Ion vizitează cel mult doi prieteni în fiecare seară. John visits at most two friends in every evening ‘John visits at most two friends every evening.’ If both DOM and CD trigger specific interpretations on DO DPs, then a natural expectation would be for such DOs to be discarded from the context of cel mult/cel puțin. (86) below shows sample variants with cel mult/cel puțin that were tested in the experiment. Variant (a) contains an unmarked indefinite preceded by cel mult (at most), while variants (b) and (c) contain DOMed and DOMed+CDed indefinite DOs respectively within the same context:

22 See Tigău (2016) for an initial discussion on the behaviour of marked and unmarked DOs within these contexts.

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

59

(86) Context: Legile privind adopția s-au modificat de curând. ‘The laws concerning adoption have undergone modification recently.’ a. Astfel, indiferent de statutul social, familiile Thus irrespective of status.DEF.ART: social families.DEF.ART. pot adopta cel mult doi copii aflați în grija statului. may adopt at most two children found in care state.GEN Thus, families may adopt at most two children found in the care of the state, irrespective of their social status.’ b. Astfel, indiferent de statutul social, familiile Thus irrespective of status.DEF.ART: social families.DEF.ART. pot adopta pe cel mult doi copii aflați în grija statului. may adopt DOM at most two children found in care state.GEN Thus, families may adopt at most two children found in the care of the state, irrespective of their social status.’ c. Astfel, indiferent de statutul social, familiile Thus irrespective of status.DEF.ART: social families.DEF.ART. îi pot adopta pe cel mult doi copii aflați în CL.3.ACC.PL. may adopt DOM at most two children found in grija statului. care state.GEN Thus, families may adopt at most two children found in the care of the state, irrespective of their social status.’ Mood represents another ingredient with an impact on the specific/non-specific reading of indefinite DOs. As known, the grammatical mood of a predicate from within a relative clause modifying a nominal was proposed as a specificity test as early as Rivero (1979). Farkas (1982) extensively discusses subjunctive relatives and argues that the same phenomenon as in Spanish is at stake in Romanian. The same is observed in Dobrovie-Sorin (1994). The main idea behind examples such as the ones under (87) is that, while the unmodified indefinite DO in (a) may exhibit both a specific and a non-specific reading, the same DP may only read non-specifically in (b), given that it has been modified by a relative clause containing a subjunctive predicate. In (c), where the mood in the relative clause is the indicative, the DO may only acquire a specific interpretation: (87) a. Ion caută un instalator. John seeks INDEF.ART. plumber ‘John is looking for a plumber.’

60

2 Romanian direct objects. A comparative analysis with Spanish

b. Ion caută un instalator care să îi John seeks INDEF.ART. plumber who SUBJ. CL.3.DAT.SG. repare chiuveta. fix sink.DEF.ART. ‘John is looking for a plumber who should repair his sink.’ c. Ion caută un instalator care i-a John seeks INDEF.ART. plumber who CL.3.DAT.SG.-has reparat chiuveta anul trecut. fixed sink.DEF.ART. year.DEF.ART. last ‘John is looking for a plumber who repaired his sink last year.’ Again, just as in the case of DOs preceded by cel mult/cel puțin (at most/at least), DOMed and CDed+DOMed indefinites are expected to be infelicitous in the context of a relative containing a predicate in the subjunctive and to fare quite well when the predicate in the modifying relative surfaces in the indicative mood. (88) below contains actual examples used in the experiment, with (a) showing the behaviour of unmarked indefinite DO, and (b) and (c) containing DOMed and CDed+DOMed counterparts, respectivelly. Note that only the case of relatives with their predicate in the subjunctive was considered in the experiment: (88) Context: Examenul de lingvistică a fost foarte greu și majoritatea studenților au picat. ‘The exam in linguistics was very difficult and most of the students failed.’ a. Aceștia caută acum câțiva seminariști care să These search now some teaching assistants who SUBJ. le explice din nou materia. CL.3.DAT.PL. explain again subject matter.DEF.ART. ‘They are now looking for some teaching assistants who might explain the subject matter again to them.’ b. Aceștia caută acum pe câțiva seminariști care să These search now DOM some teaching assistants who SUBJ. le explice din nou materia. CL.3.DAT.PL. explain again subject matter.DEF.ART. ‘They are now looking for some teaching assistants who might explain the subject matter again to them.’ caută acum pe câțiva seminariști c. Aceștia Îi These CL.3.ACC.SG. search now DOM some teaching assistants

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

61

care să le explice din nou materia. who SUBJ. CL.3.DAT.PL. explain again subject matter.DEF.ART. ‘They are now looking for some teaching assistants who might explain the subject matter again to them.’ The third context considered for the experiment was one in which the indefinite DO showed up in the vicinity of the free choice indefinite oarecare (any), with the expectation that unmarked DOs would lose the availability of a specific interpretation but remain acceptable, while marked DOs i.e., both DOMed and CDed+DOMed variants would be rejected. The examples under (89) below show that an unmarked indefinite DO may only read non-specifically when modified by oarecare: in (89a), John may be looking for a specific plumber whom he has in mind and who might help him with the bathroom. In (89b), on the other hand, where the DO carries the modification of oarecare, the specific reading is out. (89) a. Ion caută un instalator pentru niște lucrări John seeks INDEF.ART. plumber for some works de rutină la baie. of routine at bathroom ‘John is looking for a plumber (whoever he might be) for some routine job in the bathroom.’ b. Ion caută un instalator oarecare pentru niște lucrări John seeks INDEF.ART. plumber any for some works de rutină la baie. of routine at bathroom ‘John is looking for a plumber (whoever he might be) for some routine job in the bathroom.’ Note that oarecare may be combined with the effects of a modifying relative clause containing a predicate in the subjunctive/indicative: (90a) shows that the combination of an indefinite DO modified by oarecare with a relative clause in the subjunctive amounts to a restriction on the available readings for the DO, with the result that only the non-specific one remains available; (90b), on the other hand, seems to show that there is a clash as to the semantic import of oarecare, which forces a non-specific interpretation of the indefinite, and the use of the indicative mood within the modifying relative clause, which induces a specific reading. The result is somewhat problematic, marginally allowing for a partitive reading, however: the student in question may be looked upon as pertaining to a group of fellow students who happened to speak English well at a certain point in the classroom.

62

2 Romanian direct objects. A comparative analysis with Spanish

(90) a. Caut un student oarecare care să vorbească Search.1.SG INDEF.ART. student any who SUBJ. speak bine englezește la clasă. well English at class ‘I am looking for a (any) student who might speak English well.’ b. ?Caut un student oarecare care a vorbit Search.1.SG INDEF.ART. student any who has spoken bine englezește la clasă. well English at class ‘I am looking for a (any) student who might speak English well.’ Consider (91) with an actual sample item from the experiment: (91) Context: Românii încă nu au învățat să meargă la psiholog pentru a-și rezolva problemele emoționale. ‘Romanians haven‘t yet learnt to go to the psychologist in order to solve their emotional problems.’ a. Atunci când trec prin situații grele, ei solicită Then when go.3.PL through situation difficult they ask un prieten oarecare, în fața căruia își descarcă INDEF.ART. friend any in front whom REFL. unburden sufletul. soul.DEF.ART. ‘When they go through difficult situations, they resort to a friend of theirs (no matter who) and pour our their hearts to them.’ b. Atunci când trec prin situații grele, ei solicită Then when go.3.PL through situation difficult they ask pe un prieten oarecare, în fața căruia își DOM INDEF.ART. friend any in front whom REFL. descarcă sufletul. unburden soul.DEF.ART. ‘When they go through difficult situations, they resort to a friend of theirs (no matter who) and pour our their hearts to them.’ ei c. Atunci când trec prin situații grele, Then when go.3.PL through situation difficult they îl solicită pe un prieten oarecare, în fața CL.3.ACC.SG. ask DOM INDEF.ART. friend any in front

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

63

căruia își descarcă sufletul. whom REFL. unburden soul.DEF.ART. ‘When they go through difficult situations, they resort to a friend of theirs (no matter who) and pour our their hearts to them.’ The fourth context we verified experimentally with respect to the behaviour of indefinite DOs is the one supplied by the distributive determiner câte. This determiner is shown by Farkas (2001) to restrict the possible interpretations of an indefinite to a non-specific and a narrow scope one. Thus, while the indefinite DO a colleague in (92) may exhibit a specific reading according to which all the members of the department voted for the same colleague, as suggested by the continuation in (a’), a non-specific interpretation is also possible, as shown by the continuation in (a‘‘), which suggests that the members of the department selected different candidates from the list. When the distributive câte is employed in this context, the specific reading (and wide scope reading) of the indefinite is out. As such, only continuation (a’’) would be suitable in this case. (92) Context: Șeful de departament se alege prin vot secret. ‘The head of the department is elected by resorting to secret voting.’ a. Ieri membrii departamentului au ales toți Yesterday members.DEF.ART. department.GEN have selected all un coleg de pe lista de candidați. INDEF.ART. colleague on list.DEF.ART. of candidates ‘Yesterday, the members of the department have all selected a colleague from the list of candidates.’ a’. Acesta a fost domnul Popescu. This has been Mr. Popescu ‘This was Mr. Popescu.’ a’’.Dar nici unul dintre cei selectați nu a reușit să But none of those selected.PL not has managed SUBJ. obțină majoritatea din primul tur. obtain majority.DEF.ART. from first.DEF.ART. round But none of those selected managed to obtain the majority of votes in the first round.’ b. Ieri membrii departamentului au ales toți Yesterday members.DEF.ART. department.GEN have selected all câte un coleg de pe lista de candidați. DIST. INDEF.ART. colleague on list.DEF.ART. of candidates ‘Yesterday, the members of the department have all selected a colleague from the list of candidates.’

64

2 Romanian direct objects. A comparative analysis with Spanish

b’. *Acesta a fost domnul Popescu. This has been Mr. Popescu ‘This was Mr. Popescu.’ b’’.Dar nici unul dintre cei selectați nu a reușit să But none of those selected.PL not has managed SUBJ. obțină majoritatea din primul tur. obtain majority.DEF.ART. from first.DEF.ART. round But none of those selected managed to obtain the majority of votes in the first round.’ Example (93) contains a sample item used in the experiment, where the (a) variant aims at verifying the behaviour of the unmarked indefinite DO doi candidați (two candidates), (b) contains its DOMed counterpart and (c), the CDed+DOMed variant. (93) Context: La examenele de limbă Cambridge, probele scrise se dau individual, iar cele orale pe grupuri. ‘At Cambridge language exams, the written examinations are individual-based, while the oral ones are group based’ a. Aici comisia de examen evaluează câte doi Here commission.DEF.ART. of exam assesses DIST. two candidați o data, punând foarte mare accent pe candidates once placing very big stress on interacțiunea dintre ei. interaction.DEF.ART. between them Here, the exam commission assesses two candidates at a time granting great importance to the interaction between them.’ b. Aici Comisia de examen evaluează pe câte Here commission.DEF.ART. of exam assesses DOM DIST. doi candidați o data, punând foarte mare accent pe two candidates once placing very big stress on interacțiunea dintre ei. interaction.DEF.ART. between them Here, the exam commission assesses two candidates at a time granting great importance to the interaction between them.’ c. Aici comisia de examen îi evaluează Here commission.DEF.ART. of exam CL.3.ACC.PL assesses pe câte doi candidați o data, punând foarte mare DOM DIST. two candidates once placing very big

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

65

accent pe interacțiunea dintre ei. stress on interaction.DEF.ART. between them Here, the exam commission assesses two candidates at a time granting great importance to the interaction between them.’ As discussed above, the expectations for the experiment were that, while unmarked indefinites should fare well in these contexts, losing their specific interpretation, DOMed and CDed+DOMed indefinite DOs, on the other hand, should be discarded as unacceptable, given that both DOM and CD function as mechanisms which ensure a specific reading on the marked DPs. 2.8.2.1 Experiment design The experiment tested the behaviour of unmarked, DOMed and CDed+DOMed indefinite DOs within the four contexts described in the previous section. Nine lexicalisations were created for each context, which were further varied function of the type of indefinite employed, resulting a total of 108 items as shown in Table (4) below: Table 4: Lexicalisations with unmarked/DOMed/CDed+DOMed indefinite DOs. cel mult /cel puțin

câte

oarecare

mood

bare DOM CD+DOM bare DOM CD+DOM bare DOM CD+DOM bare DOM CD+DOM 























The items thus obtained were further distributed into three different questionnaires, employing the Latin square method for an even randomization. Fillers were added to the resulting questionnaires (grouped into completely unacceptable items, completely acceptable items and items of average acceptability rating). The fillers were previously tested informally with a small group of respondents who did not take part in the main experiment. The questionnaires were formatted as Google online forms in such a way that the potential respondent could only access one item at a time, without having the possibility of going back or forth. Each questionnaire was assessed by at least 20 native speakers of Romanian, such that more than 60 people took part in the experiment. The results thus obtained were then verified and outliers were removed i.e., questionnaires in which more than 7 fillers had been wrongly assessed. Remaining questionnaires entered statistical analysis (see Appendix 3).

66

2 Romanian direct objects. A comparative analysis with Spanish

2.8.2.2 Expected results – experimental hypotheses Our main hypotheses in what concerns the current experiment built on various studies on Romanian DOM and CD, which stress on the role of both these mechanisms as specificity triggers.23 As such, the expectation we had with respect to the behaviour of indefinites in contexts inducing a non-specific interpretation differed function of DO marking or the lack thereof. More specifically, unmarked indefinite DOs were expected to fare well in these contexts and to lose their specific interpretation (which is generally available otherwise). Drawing on Tigău (2016) where it is argued that DOMed indefinite DOs seem to behave on a par with their unmarked counterparts in these contexts, i.e., to fare quite well with respect to acceptability rating and to lose their specific interpretation (which seems to be preferred otherwise Cornilescu 2000, Dobrovie-Sorin 1994). Finally, CDed+DOMed indefinites were expected to be totally discarded from such contexts. This hypothesis is formulated in an extensive discussion in Tigău (2016) with respect to the contribution of the pronominal clitic as a specificity trigger and finds support in the behaviour of clitic left dislocated DOs which do not get DOMed (the only situation in which one may draw DOM and CD apart and study the contribution of the clitic in isolation24). Consider the examples under (94) in this respect: in (94a) the CLLDed DO is necessarily DOMed and clitic resumed and is interpreted as specific i.e., there is a certain corrupt politician such that all the existing journals have criticized, irrespective of their political orientation. As expected, the use of the distributive câte with such an example leads to ungrammaticality (94b). Thus, an indefinite whose only available interpretation is a specific one, is infelicitous within the context of the distributive inducing non-specificity. Finally, consider (94c), where the DO has been left dislocated without being clitic resumed: a non-specific reading becomes available for the indefinite in this case and the use of câte is no longer rejected as ungrammatical (94d):

23 Steriade (1980), Farkas (1987), Dobrovie-Sorin (1990, 1994), Cornilescu (2000), Gierling (1997), Tigău (2010, 2016), Chiriacescu & von Heusinger (2011a, b) a.o. 24 As pointed out in various studies, the CD instances represent a subset of the DOMed instances in the sense that a clitic doubled DO is also necessarily DOMed. (Note that the opposite does not hold). DOM and CD have, however, arose at different points in time, with the first already at work in the XVIth c. and the second acquiring increasing importance in the XVIIth c. (Stan 2009, Tigău 2010). The interdependence between the pronominal clitic and differential object marking by means of pe seems to have grown in time to such an extent that some Romanian linguists analyse the two phenomena as the expression of one mechanism at the clause level (Cornilescu and Dobrovie-Sorin 2008, Tigău 2015 a.o.).

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

67

(94) a. Pe un politician corupt l-au DOM INDEF.ART. politician corrupt CL.3.ACC.SG-have.3.PL criticat toate ziarele, indiferent de orientare criticized all newspapers.DEF.ART. irrespective of orientation politică. political ‘All the journals have criticized a corrupt politician, irrespective of his political orientation.’ b. *Pe câte un politician corupt l-au DOM DIST. INDEF.ART. politician corrupt CL.3.ACC.SG-have.3.PL criticat toate ziarele, indiferent de orientare criticized all newspapers.DEF.ART. irrespective of orientation politică. political ‘All the journals have criticized a corrupt politician, irrespective of his political orientation.’ c. Un politician corupt vor critica oricând toate INDEF.ART. politician corrupt will.3.PL. criticise anytime all ziarele, indiferent de orientare politică. newspapers.DEF.ART. irrespective of orientation political ‘All the journals will criticize a corrupt politician, irrespective of his political orientation.’ d. Câte un politician corupt vor critica oricând DIST. INDEF.ART. politician corrupt will.3.PL. criticise anytime toate ziarele, indiferent de orientare politică. all newspapers.DEF.ART. irrespective of orientation political ‘All the journals will criticize a corrupt politician, irrespective of his political orientation.’ As it seems, CLLD behaves on a par with CD in that indefinites thus marked appear to only be able to evince the specific reading, which explains the impossibility of co-occurrence with câte. Note, however, that, just like in the case of CD, one may not exactly point out which of the two mechanisms (DOM or the clitic) is responsible for the specific interpretation of the dislocated indefinite which further clashes with the non-specific reading forced on the DO by câte. Consider now the case of CLLDed indefinite DOs denoting an inanimate referent: what is crucial about such examples is that the left dislocated DO may only be clitic resumed. DOM is out, given that these DPs denote inanimate referents. We start first by noticing that the left dislocated indefinite două cărți (two books) in (95a) may evince both a specific and a non-specific reading; in (95b),

68

2 Romanian direct objects. A comparative analysis with Spanish

on the other hand, only the specific interpretation seems to be available (i.e., one in which every student read the same two specific books written by Tolstoi). Given that this restriction with respect to the interpretation of the dislocated DO seems to come up as a consequence of clitic resumption (in the absence of DOM), one might presuppose that the clitic alone is at the core of the specific interpretation (while DOM serves a different purpose). The hypothesis is further strengthened when considering (95c) and (95d) below, containing the respective dislocated DOs in the company of the distributive câte: while the undoubled DO allows co-occurrence with câte exhibiting a non-specific reading, the CLLDed variant does not fare very well in this respect. (95) a. Două cărți de Tolstoi a citit fiecare two books of Tolstoi has read every student de la Litere. student from Letters ‘Every student from the Faculty of Letters has Tolstoy.’ b. Două cărți de Tolstoi le-a citit two books of Tolstoi CL.3.ACC.PL.-has read fiecare student de la Litere. every student from Letters ‘Every student from the Faculty of Letters has Tolstoy.’ c. Câte două cărți de Tolstoi a citit fiecare DIST. two books of Tolstoi has read every student de la Litere. student from Letters ‘Every student from the Faculty of Letters has Tolstoy.’ d. ?Câte două Cărți de Tolstoi le-a DIST. two books of Tolstoi CL.3.ACC.PL.-has fiecare student de la Litere. every student from Letters ‘Every student from the Faculty of Letters has Tolstoy.’

read two books by

read two books by

read two books by citit read

read two books by

As it seems, CLLDed indefinite DOs have a specific reading even in the absence of DOM. Unmarked left dislocated DOs behave on a par with their unmoved counterparts. This state of affairs lends strong support to the hypothesis that

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

69

the pronominal clitic (and not DOM) is the actual trigger for specific readings25 and it was accepted as one of the hypotheses put forth with respect to the current experiment. Summarizing the discussion so far, the following three hypotheses were formulated with respect to the behaviour of indefinite DOs in the four contexts forcing non-specific readings: 1. Unmarked indefinite DOs would be found to range quite high on the acceptability scale and would only be interpreted as non-specific. 2. DOMed indefinite DOs would function on a par with their unmarked counterparts in the sense that their occurrence within the four contexts would be rated as acceptable (on a non specific interpretation). 3. DOMed and CDed indefinite DOs would be rejected from these contexts, given that the pronominal clitic doubling them induces a specific interpretation and as such, there will be a clash between such DPs and their context of occurrence which forces non-specificity.

2.8.2.3 Results and discussion 2.8.2.3.1 Preliminary remarks All the four contexts forcing non-specificity exhibit the same general pattern: while unmarked indefinites seem to be the most felicitous, DOMed indefinites and CDed+DOMed ones fare significantly worse. Indeed, the acceptability of unmarked indefinites seems to go unquestioned across the four contexts, all items getting assigned acceptability values ranging between 6 and 7 on the 7-rung acceptability scale by over 73,33% of the respondents. Note also that none of the 60 respondents assessed any of the sentences containing unmarked indefinite DOs as unacceptable (with scores lower than 2 on the scale) or as rather acceptable (score up to 3). 96.66% of the respondents gave scores above 4, considering these items as ranging from quite acceptable to completely acceptable, and only 3.33% of the participants gave scores between 3 and 4. Marked DOs, whether simply DOMed or CDed+DOMed, fared significantly worse than their unmarked counterparts, a result which would be expected under a hypothesis that DOM and CD are specificity triggers. The graph below presents us with the mean values sorted function of the acceptability scores assigned to the

25 Building on the case of CLLDed and (non-DOMed) indefinites, Tigău (2016) analyses the pronominal clitic as the sole specficity trigger. The clitic is argued to function as a function restricting the output of the embedding functions interpreting the respective marked indefinite in DRT terms.

70

2 Romanian direct objects. A comparative analysis with Spanish

Sorted function of unmarked DOs 8.00 7.00 6.00 5.00 4.00 3.00 2.00 1.00 Q3.15 Q3.6 Q3.4 Q3.9 Q2.16 Q1.12 Q3.3 Q3.2 Q3.17 Q1.7 Q2.7 Q1.10 Q1.11 Q2.21 Q2.22 Q2.20 Q1.8 Q1.13 Q3.16 Q2.14 Q1.18 Q1.6 Q2.2 Q3.1 Q2.17 Q3.18 Q3.11 Q1.5 Q2.1 Q1.17

0.00

Series1

Series2

Series3

Graph 3: Results sorted function of unmarked DOs.

GMs in contexts

câte

oarecare

4,10

3,61

3,79

CD + DOM + indefinite

DOM+indefinite

3,62

bare indefinite

3,68

CD + DOM + indefinite

bare indefinite

CD + DOM + indefinite

DOM+indefinite

cel mult.cel puțin

3,53

DOM+indefinite

4,31 3,57

DOM+indefinite

6,34

5,95

bare indefinite

6,04

CD + DOM + indefinite

6,24

bare indefinite

7 6 5 4 3 2 1

mood

Graph 4: GMs contexts and DO types. Table 5: GMs contexts and DO types.

cel mult/cel puțin

câte

bare indefinite DOM+indefinite CD + DOM + indefinite bare indefinite DOM+indefinite CD + DOM + indefinite

, , , , , ,

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

71

Table 5 (continued ) oarecare

bare indefinite DOM+indefinite CD + DOM + indefinite bare indefinite DOM+indefinite CD + DOM + indefinite

mood

, , , , , ,

items containing unmarked indefinites. As it may be noticed, unmarked indefinites fare significantly better than their marked counterparts. This trend is a general one, with only one questionnaire assigning a higher acceptability score to the CDed+DOMed variant than to its unmarked counterpart (Q3.6).26 The statistical tests uncovered no significant difference with respect to the behaviour of the three types of DO function of the context: the same DO type exhibits the same behaviour irrespective of the context wherein it has been inserted. The graph below thus shows the differences of DO behaviour within the merged contexts:

Contexts merged 7 6

6.146825397

5 4 3.974867725 3.591269841

3 2 1 bare indefinite

DOM+indefinite

CD + DOM + indefinite

Graph 5: Three DO types, merged contexts.

What is, however, unexpected is that the variants containing DOMed or CDed+DOMed DOs were not completely discarded as ungrammatical, but actually assigned scores towards the mid area of the seven-rung acceptability scale.

26 Series 1 represents the acceptability values assigned to items containing CDed+DOMed; series 2 shows the scores assigned to the DOMed variants, series 3 presents the results obtained for the unmarked indefinites.

72

2 Romanian direct objects. A comparative analysis with Spanish

Thus, as the table below shows, the results concerning marked DOs surpass the acceptability threshold of 3,5 GM on a 7-rung acceptability scale for all the four contexts, with values ranging from 3,53 GM (for câte contexts) to 4,31 GM (for cel mult/cel puțin). Indeed, only 6,66 % (with respect to CDed+DOM DOs) and 18,33 % (with respect to the single DOMed variants) of the respondents respectively consider the marked variants to be totally out, assigning scores of 1 or 2 on the acceptability scale. More than half of the respondents supplied scores ranging above the acceptability threshold of 3,5: 61,66 % for the CDed+DOMed variants and 55% for the DOMed one. Note, however, that the instances containing marked DOs did not usually get higher scores from the other end of the acceptability scale: only 8,33 assigned scores of 6 and 7 to CDed+DOMed DO and an even smaller minority did so for the DOMed variants i.e., 6,66%. Consider Table (6) below extending on the results: Table 6: Differentiated GM scores. Type

GM

GM> 

CL+pe /–,% /–,% /–,% /–% /–.% pe / -,% /–% /–% /–% /–.% ; / / /–,% /–.% /–,%

Consider also the graph below showing a comparison with respect to acceptability scores between items containing indefinite DOs and fillers. We grouped fillers into three categories function of how these items fared in a preliminary informal experiment: very bad/very good/average. Note that the fillers labelled as very bad, got very low acceptability scores, with a GM of 1,77, while fillers considered to be very good received scores higher than 6 on the seven-rung acceptability scale i.e., 6,19 GM. As the graph below shows, marked indefinite DOs, whether DOMed or CDed+DOMed fared much better than the bad fillers with respect to acceptability ratings. One further unexpected result had to do with the differences in acceptability uncovered for the DOMed variants as opposed to their CDed+DOMed counterparts. These unexpected results disconfirmed both our hypothesis with respect to the behaviour of single DOMed variants (i.e., that DOMed DPs behave on a par with their unmarked counterparts getting good scores and allowing only the non-specific reading), as well as our hypothesis concerning the impossibility of CDed+DOMed DOs in the four contexts forcing a non-specific interpretation. Thus, contrary to our expectations, DOMed DOs seemed to fare worse than the CDed+DOMed variants (see Table 7):

73

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

Items and Fillers

cel mult.cel puținFillersFillers

câte

FillersFillers

oarecare

FillersFillers

mood

very good

very bad

CD + DOM + indefinite

DOM+indefinite

bare indefinite

very good

very bad

CD + DOM + indefinite

DOM+indefinite

bare indefinite

very good

very bad

CD + DOM + indefinite

DOM+indefinite

bare indefinite

very good

very bad

CD + DOM + indefinite

DOM+indefinite

bare indefinite

7 6 5 4 3 2 1

FillersFillers

Graph 6: Acceptability of items vs. fillers.

Table 7: Contexts and marked DOs.

DOMed DO CDed+DOMed DO

Cel mult/cel puțin

Câte

Oarecare

Mood

. GM , GM

, ,

, ,

, ,

Indeed, while 40% of the respondents assessed contexts with DOMed indefinites as rather unacceptable assigning scores below 3 on a 7-rung acceptability scale, only 21% of the respondents assessed CDed+DOMed indefinites in a similar way. The trend stays the same if we consider the number of respondents who assessed contexts with DOMed indefinites and CDed+DOMed indefinites as completely unacceptable, with more people considering that this is the case for DOMed items (18,33%) and only 6,66% of respondents evaluating CDed+DOMed indefinites in a similar way. The trend of considering the DOMed variant as worse than its CDed+DOMed counterpart is general: There are 15 questionnaires out of 60 which assess contexts with DOMed indefinites as more acceptable than their CDed+DOMed counterparts. Nevertheless, the differences in score values are quite small and therefore insignificant. Only in the case of one respondent is the value assigned to the CD+DOM variant significantly lower than for the DOMed variant (Q2.21: CD+DOM: 2,75 vs. DOM: 4,25). The graph below shows the results sorted function of the acceptability rating of CDed+DOM instances. As may be seen, the DOM variants are generally assigned lower acceptability values, with only few exceptions:

74

2 Romanian direct objects. A comparative analysis with Spanish

Sorted according to CL+pe : CL+pe; pe, unmarked 8.00 7.00 6.00 5.00 4.00 3.00 2.00 0.00

Q3.15 Q1.7 Q2.8 Q1.9 Q2.19 Q1.19 Q1.1 Q3.21 Q1.16 Q2.5 Q1.4 Q3.4 Q1.8 Q2.3 Q3.14 Q1.20 Q2.7 Q3.3 Q3.9 Q2.12 Q3.22 Q3.7 Q2.10 Q1.2 Q3.5 Q2.2 Q2.20 Q3.2 Q1.17 Q3.20

1.00

Series1

Series2

Series3

Graph 7: Results sorted function of CDed+DOMed DOs.

Conversely, there are 42 questionnaires which assign higher acceptability scores to the contexts containing CDed+DOMed indefinites. Again, the differences between score values are quite small, and only 5 questionnaires draw a significant difference between the CDed+DOM variant and the DOMed one: Table 8: Significant differences between scores for CDed+DOMed DOs vs. DOMed DOs. Questionnaire Q. Q. Q. Q. Q.

CD+DOM

DOM

. . . . .

. .  . .

The following graph presents the results sorted function of the DOM instances and it shows quite clearly how the CDed+DOMed generally get higher acceptability values. Thus, as a conclusion on the acceptability differences between the DOMed and the CDed+DOMed variants, we could say that the latter are slightly preferred over the former. There is, however, a general understanding to mark both DOMed and CDed+DOMed along the mid value area of the scale. Leaving aside the general lower score received by DOMed variants with respect to their CDed+DOMed variants, an even more unexpected result in view of our initial hypothesis is the fact that CDed+DOMed instances were at all accepted in the four contexts forcing non-specific readings. This is indeed surprising under the hypothesis that the pronominal clitic is the actual specificity trigger in these

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

75

Sorted function of DOM 8.00 7.00 6.00 5.00 4.00 3.00 2.00 1.00 Q1.5 Q2.4 Q2.14 Q3.15 Q3.4 Q2.7 Q1.11 Q3.21 Q1.3 Q2.17 Q1.1 Q2.19 Q2.10 Q3.3 Q2.15 Q2.22 Q2.12 Q3.7 Q3.9 Q2.1 Q1.8 Q3.13 Q1.18 Q1.15 Q3.1 Q3.22 Q1.6 Q1.17 Q2.20 Q3.11

0.00

Series1

Series2

Series3

Graph 8: Results sorted function of DOMed DOs.

configurations (and not DOM). As shown in Table 5, the scores assigned to CDed+DOMed structures range between 3,68 and 4,31 prompting one with the idea that these configurations are generally looked upon by the respondents as rather/fairly acceptable. Another aspect which might be of interest is the relative uniform distribution of values we noticed with respect to the CDed+DOMed structure: respondents distribute evenly, assigning values ranging from 1,5 to 6,67 as may be seen in the following graph:

CL+pe 7.00 6.00 5.00 4.00 3.00 2.00 1.00

Q2 .4 Q2 .1 9 Q1 .3 Q3 .2 1 Q2 .1 7 Q1 .4 Q3 .1 7 Q2 .3 Q2 .2 2 Q2 .7 Q3 .6 Q2 .1 2 Q3 .1 3 Q2 .1 0 Q2 .1 Q2 .2 Q3 .1 8 Q1 .1 7

Q1

.7

0.00

Graph 9: Acceptability scores for CDed+DOMed DOs.

76

2 Romanian direct objects. A comparative analysis with Spanish

2.8.2.3.2 Accounting for the results In view of the results presented above, one has to discuss the following aspects in relation to the hypotheses initially put forth, which only seemed confirmed to a certain extent: 1. The fact that unmarked DOs expectedly obtain higher scores on the acceptability scale. 2. The fact that DOM DOs get the lowest acceptability scores. These results should be discussed in relation to our initial hypothesis that DOMed DOs behave on a par with their unmarked counterparts in allowing occurrence within the four contexts and being forced to only evince their non-specific reading in this case. 3. The unexpectedly higher scores obtained by the CD+DOMed variants in the context of our initial hypothesis that these DOs would be completely discarded from the four contexts. 4. Finally, we need to explain why the CDed+DOMed variants fared better than their DOMed counterparts: 1. As already pointed out (1), the acceptability scores obtained for unmarked DOs were expectedly high: unmarked indefinites in Romanian may allow both a specific, as well as a non-specific reading and may acquire both a wide scope and a narrow scope interpretation when co-occuring with other scope taking expressions. As such, the non-specific reading is generally available with unmarked indefinites and will be the only one to get actualized when the respective DO gets inserted within a context forcing a non-specific reading. This is precisely what happened with the unmarked DOs tested in the current experiment. 2. In order to account for (2) above, several things need to be considered. Let us first dwell upon the initial research hypothesis according to which DOMed DOs should fare quite well within the four contexts tested and behave similarly to their unmarked counterparts in this respect. Tigău (2016) discusses at length specificity with DOMed indefinite DOs and reaches the conclusion that DOM is not a specificity trigger. More specifically, there are situations where a marked DO does not read specifically. Consider to this end example (96) below, where DOM is obligatorily required with bare quantifiers such as nimeni (nobody), cineva (somebody), oricine (anybody) which are never specific:

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

77

(96) Am auzit zgomot de pași, dar nu am văzut Have.1.SG. heard noise of steps but not have.1.SG. seen pe/*ø nimeni. DOM/*ø nobody ‘I heard footsteps but I did not see anyone.’ Consider also (97) below, where the DOMed indefinite may only be interpreted in the scope of the negative operator. In this respect, the DOMed DO (97a) behaves on a par with its unmarked variant; the CDed+DOMed variant (97b) is however discarded from such a context and this was indeed expected, given that such DPs were hypothesized to always read specifically and, as such, to outscope the negative operator, a reading which, however, the current context does not allow,27 hence the impossibility of CD+DOM: (97) a. N-am văzut (pe) un singur profesor în Not-have.1.SG seen (DOM) INDEF.ART. single professor in toată viața mea, care să se poarte all life.DEF.ART. mine.POSS who SUBJ. REFL. behave așa urât cu studenții cum o face el. so badly with students.DEF.ART. how it does he ‘I haven‘t seen a single professor in my whole life who should behave so badly with the students as he does.’ b. ?Nu l-am văzut (pe) un singur Not CL.3.ACC.SG.-have.1.SG seen (DOM) INDEF.ART. single profesor în toată viața mea, care să se professor in all life.DEF.ART. Mine.POSS who SUBJ. REFL. poarte așa urât cu studenții cum o face el. behave so badly with students.DEF.ART. how it does he ‘I haven‘t seen a single professor in my whole life who should behave so badly with the students as he does.’ The same may be seen in (98), where the DOMed DO favours a reading whereby it is interpreted within the scope of the intensional operator and nonspecifically. A reading according to which there is a certain professor who, if I happened to see him doing some terrible deed, I would report on remains a

27 Indeed, the combination of the negative operator and singur (sole) forces a non-specific interpretation.

78

2 Romanian direct objects. A comparative analysis with Spanish

possibility. The specific reading is, however, strongly favoured (possibly the only one available) in the (98b) variant, where the DO is CDed+DOMed. (98) a. Dacă aș vedea (pe) un singur profesor că If would.1.SG. see (DOM) INDEF.ART. single professor that face asta, m-aș plânge la decan. does this REFL.1.SG.-would.1.SG. complain at dean ‘If I were to see a single professor doing this, I would complain to the dean.’ b. Dacă l-aș vedea (pe) un If CL.3.ACC.SG.-would.1.SG. see (DOM) INDEF.ART. singur profesor că face asta, m-aș plânge single professor that does this REFL.1.SG.-would.1.SG. complain la decan. at dean ‘If I were to see a single professor doing this, I would complain to the dean.’ The two examples presented above thus show that the non-specific interpretation remains an option for the DOMed indefinite DO and that it may actually be actualized in an appropriate context. Hence the expectation that DOMed DOs behave in a similar way to their unmarked counterparts within the four tested contexts forcing a non-specific reading. Given the behaviour of DOMed indefinite DOs in the contexts presented above, we expected a similar behaviour of these DPs within the four tested contexts (which had been designed in such a way as to actualize the non-specific interpretation on the DO). In this respect, the DOMed variants, given that they allowed a non-specific interpretation, were expected to behave on a par with unmarked indefinite DOs and allow a non-specific reading, the only one possible in these circumstances. Thus, the significantly lower results with respect to acceptability, which the items containing DOMed DOs get, are quite unexpected and need an explanation. To this end, note first, that the fact that the DOMed counterpart was almost always assessed as less felicitous is in line with a similar observation in Avram (2014), which also discusses the low acceptability scores some respondents assign to sentences where single DOMed DOs are employed. The results rest on an experiment involving two acceptability sentence questionnaires from 23 native speakers of Romanian (age 20–57) and Avram (2014) argues that the participants fall into two categories: those that always clitic double the pe marked DP and those who allow for ‘single’ pe marking besides having the option of CD+DOM.

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

79

There seems to exist, thus, a consistent group of respondents for whom single DOM marking is simply unacceptable. The speakers in question seem to always pair CD with DOM and to find this combination as the only one available for DO marking. To these speakers, the CD+DOM seems to function as a complex clausal construction with functional unity, in line with its Spanish counterpart described in David (2014, 2015).28 Based on these results, Avram (2014) proposes two instances of pe: the ‘single’ pe, as a semantic gender marker (as described by Cornilescu 2000) and the pe in CD structures, which has been bleached of its role as a gender marker and which is taken to function as a mere accusative case marker.29 Tigău (2015) makes a similar proposal, analysing CDed+DOMed structures as integrals along the lines of (Hornstein et al. 1995), Uriagereka 2001, 2002, 2005). A similar tendency to mark DOMed DOs as less acceptable than their CDed+DOMed counterparts is noticed in another experiment carried out by Tigău and von Heusinger (ms. ditransitives). Theirs is a more complex endeavour testing binding within ditransitives and being comprised of a series of three distinct experiments differentiated one from the other through the type of DO involved.30 A general observation put forth with respect to those instances where single DOMed DOs are employed is that these DPs tend to fare worse than the items where CDed+DOMed counterparts are used. Thus, while items containing DOMed DOs never get scores higher than 4,43, sentences containing CDed+DOMed DOs reach acceptability values of 5,52. Moreover, the DOMed variants obtain very low scores on some binding dependencies (e.g., 2,64 or 2,36) and generally get scores lower than the acceptability threshold of 3.5 on a 7-rung acceptability scale. Instances containing CDed+DOMed DOs mostly get assessed as higher than or very close to the acceptability threshold (G.M.s 4,51/ 3,52/ 3,47/ 3,18/ 3,67/ 3,42/

28 It is argued that the complex construction arises because CD and DOM have converging interpretive effects, both enhance prominence and accessibility (in the sense of Ariel (1988, 1991) of the DP referent. DOM and CD are both ‘encapsulated in a core accessibility scenario, namely, the one in which the speaker and the addressee have maximal accessibility to the discourse referents, which are physically observable in the event [David, 2015: 125], from which the construction has been shown to gradually extend. 29 Diachronic studies seem to support the view of a ‘pendulating‘ pe, with a XVIth c. pe sensitive to animacy and a XVIIth c. one which has been bleached functioning as a generalized accusative case assigner and marking both animate and inanimate direct object DPs (Stan 2009). 30 DO differ with respect to DOM and CD marking: Experiment 1 checks binding between unmarked DOs and IOs; experiment 2 focuses on binding relations with DOMed DOs and IOs, and experiment 3 checks binding dependencies with CDed+DOMed DOs and IOs.

80

2 Romanian direct objects. A comparative analysis with Spanish

5,52/ 3,51); these instances never receive scores lower than 3, as it is the case with DOMed variants.31 Given the experimental results, obtained independently of our own undertaking, which show a general dispreference for DOMed DOs in the absence of CD, we might consider that a possible explanation of the low acceptability results in this particular case would actually have to do less with their potential for a specific reading or for the loss of that reading in the four tested contexts, and more with the general frown on the phenomenon of DOM marking (without CD) as such. Note that as early as 1994, Dobrovie-Sorin (1994) already signalled such a tendency at stake. 3. The hypothesis advanced with respect to CDed+DOMed indefinites was that these DPs would be disallowed from contexts inducing a non-specific interpretation. This expectation rested on data concerning inanimate CLLDed DOs, which do not need DOM and which seemed to disallow non-specific readings and be rejected from contexts forcing such a reading on them. The expectation, however, proved to be wrong as instances containing CDed+DOMed indefinite DOs generally received scores ranging around the mid area of the acceptability scale. This shows that these examples are considered not altogether unacceptable, but rather acceptable. The explanation we would like to propose follows the line of argument advanced in (2) above and takes CD+DOM as a complex mechanism where the combination of DOM and CD exhibit functional unity. Furthermore, the pronominal clitic seems to function less like a specificity trigger and more like a syntactic one in the sense that it triggers movement out of the VP into a position where the DO has access to a mode of semantic composition (i.e., choice functions), which may give rise to a specific interpretation along the lines of López (2012).32 Having a specific interpretation is, however, not mandatory (as it is also the case with DOMed indefinite DOs.). Thus, just like López (2012) proposes for Spanish marked DOs, the specific interpretation is not linked with a piece of morphology i.e., the pronominal clitic showing up on T, but comes as a result of a peculiar mode of semantic composition enabled in syntax by way of movement to an appropriate position. In the case of Romanian CDed+DOMed DOs movement is mediated by the clitic itself.

31 For an extensive presentation of the Experiments on ditransitives see Chapter 6. 32 This is position P3 in Lòpez (2012), a position within the TP area, wherefrom the scrambled DO may bind into the subject DP, for instance.

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

81

Finally, consider some examples taken from the internet, where CDed+DOMed indefinite DOs seem to be quite comfortable on a non-specific interpretation: (99) a. Am fost învățate că sexul, Have.1.PL. been taught.FEM.PL. that sex.DEF.ART. aspectul fizic, ceea ce faci îl atrag aspect.DEF.ART. physical what do.2.SG. CL.3.ACC.SG. attract pe un barbat. DOM INDEF.ART. man ‘We have been taught that sex, physical appearance, what one does attract a man.’33 b. Pe o femeie o pierzi cu fiecare DOM INDEF.ART. woman CL.3.ACC.FEM.SG. lose with every greșeală pe care o faci. mistake DOM which CL.3.ACC.FEM.SG. do.2.SG ‘You lose a woman with every mistake that you make.’34 4. As already discussed, the DOMed variants fared worse that their CDed+DOMed counterparts in the four contexts triggering a non-specific interpretation, whereas our expectations were actually the opposite i.e., while DOMed variants were expected to range quite high with respect to acceptability scores, CDed+DOMed variants had been hypothesized to get rejected as unacceptable. The explanation advanced for this state of affairs has to do with the general dispreference for single DOM instances. On the contrary, the use of CD+DOM as a mechanism with functional unity renders instances containing CDed+DOMed DOs as quite acceptable, even in the absence of a specific interpretation with these DPs. As proposed, the specific interpretation is an indirect consequence of CD marking, in the sense that the pronominal clitic triggers the movement of the CDed DP to a position where the mode of composition with the predicate is such that a specific interpretation is available. Thus, rather than directly pairing doubling by means of a clitic with specificity, the connection between syntax and interpretation is argued to be an indirect one in line with López (2012). The examples taken from the internet, wherein CD+DOMed indefinite DOs surface in contexts where they get nonspecific interpretations strengthens this hypothesis.

33 (http://andradadan.com/psihologia-secreta-prin-care-se-indragosteste-un-barbat/), Access date: 10.06.2018 34 (http://devorbacutine.eu/pe-o-femeie-o-pierzi-cu-fiecare-greseala-pe-care-o-faci-cu-fiecare-pas -incorect-si-cu-fiecare-dorinta-imposibila-si-o-mai-pierzi-din-ipocrizie/), Access date: 10.06.2018

82

2 Romanian direct objects. A comparative analysis with Spanish

2.8.2.4 Conclusion The experiment presented in this section has uncovered a few interesting facts with respect to the behaviour of Romanian marked and unmarked indefinite DOs within contexts forcing a non-specific interpretation. The following conclusions were reached: All respondents exhibit the same pattern of behaviour with respect to unmarked indefinites, by assessing the relevant contexts as high on the acceptability scale, as expected: unmarked indefinites thus employed receive a nonspecific interpretation. DOMed indefinites and CDed+DOMed ones fare significantly worse in these contexts. However, neither DOMed DOs nor CDed+DOMed DOs are totally out when compared to the bad fillers. This shows that both types of marked DOs may read non-specifically and that Romanian does not possess a category of objects which are necessarily interpreted as specific (as it seems to be the case with Persian DOM -râ, where definites thus marked are necessarily interpreted as specific and where one may argue that specificity is encoded in the grammar).35 Thus, neither the DOM marked nor the pronominal clitic in CDed constructions guarantees specificity. The behaviour of our respondents with respect to marked indefinites does not tell us much about the division of labour with respect to specificity between DOM and CD. Contrary to our hypothesis, according to which the DOMed indefinites should have ranked high on the acceptability scale on a par with unmarked indefinites, DOMed DPs are generally assessed as more infelicitous than their CDed+DOMed counterparts (which we expected to be totally out). The experimental data support a syntactic approach to DOMed and CDed DOs along the model of Spanish marked DOs (López 2012): the DOM marker guarantees an type reading on the nominal it marks, which secures argumenthood and individuation. As a consequence, marked DPs may not remain in-situ, as this is a position where only property denoting nominals may occupy, given that they compose with the predicate via incorporation. Marked DPs need to actually scramble to a position where they may compose with the predicate via Function Application, as expected for nominals that function as real arguments. The target position secures a peculiar mode of semantic composition as well as a mode of interpretation of the scrambled DO (via a choice function). Specificity thus arises as a possible reading deriving from this mode of semantic composition and not as a consequence of DOM marking.

35 Indeed, -râ is impossible with sentences containing an indefinite modified by a relative in the subjunctive as in (1), unlike Romanian or Spanish equivalent DOMed DOs:

2.8 Reinforcing the conclusions: two experiments on Romanian DOs

83

CDed+DOMed DOs scramble even further than their DOMed counterparts, presumably passing through the position that DOMed DOs reach. They also get composed with the predicate in a similar way as their DOMed counterparts, which allows us to expect the same readings with both marked variants. The next chapter will present an account of Romanian DOs by patametrizing the analysis in López (2012) just sketched.

3 Linking syntactic structure to interpretation In the previous chapter we discussed a number of similarities and differences with respect to the behaviour of marked and unmarked DOs in Romanian and Spanish. On the one hand, we observed that marked DOs in the two languages pattern together with respect to specificity, scope dependencies, the ban on functioning as complements of the individual-level have, and the occurrence in small clauses, object control and clause union in causative configurations. On the other hand, it was shown that the two languages differ with respect to unmarked DOs: while Spanish unmarked DOs always read non-specifically, never outscope other scope bearing expressions, occur as complements of the individual-level have, never surface in small clauses, object control and clause union in causative configurations, the Romanian counterparts pattern with marked DOs in allowing for specific and wide scope readings, and being able to occur in small clauses, object control and clause union in causative configurations. One final observation had to do with the behaviour of bare nominals, which seem to pattern alike in the two languages, following the model of unmarked DOs in Spanish. In this chapter we turn to an enlightening analysis of Spanish DOs proposed by López (2012), and put forth our own proposal for Romanian DOs parametrizing López’s account in such a way as to capture the contrasts between Romanian and Spanish with respect to both the syntax and the interpretation of the DO in the two languages. We retain from López’s analysis the correlation between syntax and semantic interpretation: a) nominal expressions which are interpreted in situ are interpreted as predicates; all nominals which scramble are interpreted as arguments. We also retain the correlation between syntactic position and case-valuation strategies. In situ objects incorporate, scrambling objects check case by Agree. Briefly, we will argue that the parameter differentiating between Romanian and Spanish lies in the types of nominals that must scramble: Spanish distinguishes between KPs and DPs, only allowing the former (i.e., marked DOs) to scramble, while the latter incorporate. In Romanian, on the other hand, both KPs and DPs may scramble. The relevant division in this case is between NP/NumPs which incorporate and DPs/KPs which scramble. Remember that scrambling is associated with argumenthood. As shown by Longobardi (1994 and seq.), the locus of argumenthood is the Determiner, the presence of the Determiner is sufficient for reference. Nominals that have supplementary marking of referential status (i.e., K, CD) naturally also count as arguments and must scramble. Both DPs and KPs combine with the verb by https://doi.org/10.1515/9781501513657-003

86

3 Linking syntactic structure to interpretation

functional application. The fact that scrambling secures a referential interpretation also goes in the direction of Diesing (1992)’s account. The outline of this chapter is the following: sections 1 and 2 focus on López (2012)’s account of Spanish DOs; section 3 extends on the problems raised for this account by Romanian DOs and parametrizes the analysis put forth in López (2012) in such a way as to accommodate for the observed differences. Section 4 advances a proposal with respect to the featural make-up of Romanian DOs.

3.1 Analysing Spanish Data: DOMed DOs move, unmarked DOs do not Drawing on Spanish data López (2012) makes a distinction between DOMed and unmarked indefinite DOs, arguing that while the former may acquire a specific interpretation, the latter do not engender such a reading. He further suggests that this interpretive difference boils down to a syntactic one in that DOMed DOs have to undergo short scrambling from their base generated position to a position where they may be case licensed, as opposed to their unmarked counterparts, which stay in situ and incorporate with the verb. López (2012) does not, however, argue in favour of either a direct mapping between syntax and semantics, or one such relationship between morphology and semantics. If the connection between a piece of morphology and a certain interpretation were indeed a direct one, then the use of DOM would always lead to a specific interpretation, whereas Spanish (and Romanian data) show that DOMed indefinite DOs are prone only to such an interpretation. As to the connection between syntax and semantics, López (2012) stresses on the fact that it is indirect in the sense that the syntactic position that the DO occupies affects the mode of semantic composition, and this ultimately affects the interpretation of the sentence. The different modes of semantic composition would thus explain the lack of specific readings with unmarked indefinites as well as the availability of a specific interpretation with marked indefinites, given that the former get interpreted by means of Restrict36 (cf. Chung and Ladusaw 2004, 2006), while the

36 Chung and Ladusaw (2004, 2006) propose an operation, Restrict, which enables the combination of the DO with its predicate by way of conjunction. As a consequence, the predicate remains unsaturated. López (2012) argues that unmarked indefinite DOs get interpreted by means of this mechanism, which explains why they necessarily have a narrow scope reading.

3.1 Analysing Spanish Data: DOMed DOs move, unmarked DOs do not

87

latter end up in a position where they get interpreted by means of Choice Functions, in line with Reinhart (1997).37,38 According to López (2012), indefinite DOs which remain in their merge position may only be composed with their selecting predicate by means of Restrict. On the other hand, DOs which undergo (short) scrambling may only be interpreted by choice functions. The landing site of the (short-) scrambled DO is a position governed by small v, wherefrom the DP may not c-command the subject or other scope- taking operators within the clause. The fact that the DO in question may acquire a wide scope or a specific interpretation is a consequence of it having been interpreted by means of a choice function. In line with the basic tenets of Minimalism (Chomsky 2000), López (2012) assumes that transitive predicates have an abstract light verb in their structure which introduces the External Argument (EA). Also, Accusative case is assumed to originate in v. Building on Travis (1992, 2010), Koizumi (1995), Baker and Collins (2006), López further assumes a functional category lying between vP and VP, which he labels α (27). This category is argued to be a conglomerate of properties in that it may on the one hand function as an applicative projection introducing

37 A choice function variable shifts a DP from a property denotation to an entity one e. As a consequence, the DP may then be composed by Functional Application. Reinhart (1997) proposes that the choice function variable may be bound by an existential quantifier, which may be merged at different points within the derivational tree. This enables a suitable explanation for the different scopes that indefinites may give rise to e.g., the intermediate scope of the DO a friend in (1) below: (1)

Every woman is convinced that if John invites a friend of his to the party, it will be a disaster.

∀x9fCHðf Þ½womanðxÞ ! convincedðx, ½inviteðJohn, f ðfriend of mineÞ, partyÞ ! disasterðpartyÞÞ ‘For every woman x there exists a (choice function that picks out a) friend of John’s such that if John invites the friend picked out by the choice function to the party, x is convinced it will be a disaster’ (López 2012: 15b, p. 7) 38 In line with Diesing (1992), López (2012) argues that there is a correlation between the syntactic position occupied by the indefinite DO and its ability to acquire a specific reading. Unlike Diesing, however, López proposes an indirect mapping between configuration and semantic interpretation in that this mapping is mediated by the kind of operations which may apply in order to construct the compositional meaning of the verb and its complement. López thus pairs the syntactic positions occupied by the DO with different modes of semantic composition. See also Tigău (2010) for the relevance of the Mapping Hypothesis in Diesing (1992) with respect to the Romanian indefinite DOs.

88

3 Linking syntactic structure to interpretation

indirect objects into the argument structure of the verb while assigning case to the direct object at the same time (Marantz 1993, Pylkkänen 2002, 2008); this αAppl further differentiates itself from a simpler α which only serves for case reasons. Moreover, α may also function as a head related to inner aspect, ensuring telicity or boundedness. (100) vP ri v’ EA ri αP v ri α’ ri α VP ri V DO

SpecαP39 is argued to be the landing position of DOMed DOs, which undergo short scrambling for reasons of case.40 One argument supporting the hypothesis that DOMed DOs move from their base generated position comes from binding dependencies with the IO and the subject: the marked DO may bind the IO but it may not bind into the Subject.

39 This position is different from the SpecvP position, which has also been argued to function as a landing site for marked object movement, as in Torrego (1998), Rodríguez-Mondoñedo (2007). López (2012) argues that SpecvP may not be the landing site of marked DOs given that these DPs may never bind into the Subject DP, a prediction which should be borne out if the DOMed DO were to move into a position wherefrom it can c-command the EA (1): (1)

vP ri v’ DODOM ri v’ EA ri v

40 Movement is assumed to be triggered by an unvalued feature within the item that moves and targets the next specifier up. Movement ends when the unvalued feature finds itself governed by a suitable probe. The links of a chain created by means of movement get joined together by means of feature sharing and the resulting dependency is formally identical to dependencies created by means of Agree. (López 2012: 35).

3.1 Analysing Spanish Data: DOMed DOs move, unmarked DOs do not

89

Unmarked DOs stay in situ and will not be able to bind either the IO or the subject DP. (101) below shows this at stake: (101)

vP ri v’ EA ri v

αP ri DODOM α’ ri IO α’ ri α VP ri V DO

(López 2012: 17, p. 40) Spanish data seem to support these claims: in (102) the presence of DOM gives rise to a reading according to which there is a pairing between prisoners and sons, which would not be available otherwise. The availability of such a quantifier-variable interpretation shows that the DOMed DO finds itself in a position wherefrom it may c-command the possessive within the IO.41 When the DO is left unmarked, the quantifier-variable reading is, expectedly, no longer available due to the fact that (as shown in (101)) this DO will never occupy a position wherefrom c-command into the IO might be possible.42 (102) [Context: What did the enemies do? The enemies delivered X to Y and Z to W, but. . ..] ningúni Los enemigos no entregaron a sui hijo a/ø DEF.ART. enemies not delivered to his son DOM/ø no prisionero. prisoner ‘The enemies did not delivered any prisoner to his son.’ (López 2012: 18, p. 41)

41 With respect to this example, López (2012) discusses the remarks in Leonetti (2004), where it is argued that the availability of the quantifier – variable reading comes as a consequence of the specific reading of the marked DO and not necessarily from the different c-command relations mediated by a. López argues that the signalled binding dependency could not have been the outcome of specificity given that this reading is very weak for the DP ningúni prisionero. 42 See, however, Pineda (2013) for a discussion on the bidirectional c-command potential of the two internal arguments within Spanish ditransitives.

90

3 Linking syntactic structure to interpretation

The same bound interpretation obtains with the DOMed DO a nadie in (103) below: nadiei (103) Los enemigo no entregaron a sui hijo a DEF.ART. enemies not delivered.3.PL. DOM his son DOM nobody ‘The enemies delivered no one to his son.’ (López 2012: 19, p. 41) DOMed DOs may also bind a dative reflexive, while their unmarked counterparts may not. In (104a), the unmarked DO un hombre stays in situ and may not bind the anaphor within the dative DP. This binding dependency becomes possible once the DO is DOMed and this may be accounted for if the marked DO finds itself in a position wherefrom it may c-command the IO i.e., Spec, αP. (Note also that this binding dependency obtains even in the absence of a specific interpretation of the marked DO as demonstrated by the possibility of using the free-choice indefinite cualquiera as a modifier. Hence, the bound reading obtains as a consequence of having moved the marked DO into a position wherefrom it may c-command the dative reflexive and not from the specific interpretation of this DO, contra Leonetti 2004): (104) a. *María le entregó a Mary CL.3.DAT.SG. delivered DOM ‘Maria delivered a man to himself.’ b. *María le entregó a Mary CL.3.DAT.SG. delivered DOM un hombre. (cualquiera). INDEF.ART. man (no matter who) ‘Maria delivered a man to himself.’

sí mismo un hombre. himself INDEF.ART. man sí mismo a un himself DOM INDEF.ART.

(López 2012: 20b,c, p. 41)

A DOMed DO may not, however, bind into the Subject DP: this is so due to the fact that the position that the marked DO reaches by undergoing scrambling does not c-command the EA. Example (105) shows this at stake: a quantifier – variable interpretation is not possible in this case, given that the a marked DO ningún niño may not bind the possessive within the IO.43

43 The quantifier-variable interpretation may obtain, but only if the DOMed DO is also clitic right dislocated or p-moved/scrambled: in (1) the clitic-right-dislocated DO may bind the EA, as argued by López (2012) building on Villalba (2000). An explanation for the existence of such readings is that the right dislocated DO has moved to a position above the in situ position of the EA:

3.1 Analysing Spanish Data: DOMed DOs move, unmarked DOs do not

91

(105) [Context: So, what happened yesterday?] ningúni niño. Ayer no atacó su*i propio padre a Yesterday not attacked his own father DOM no boy ‘His own father attacked no boy yesterday.’ (López 2012: 22, p. 43) The binding data thus suggest that while Spanish unmarked DOs stay low within the VP and never leave their in situ position as complements of the lexical verb, marked DOs move to a position wherefrom they may c-command IOs but in which they are c-commanded by the EA. This position is Spec, αP, as argued by López (2012), contra Torrego (1998) or Mondoñedo (2007). López (2012) applies this analysis to the three syntactic contexts where a marking is obligatory in the absence of any specificity requirements i.e., small clauses, clause union, object control. He shows that the DO in all these contexts does not occupy the V complement position, which prevents it to incorporate and forces scrambling. Given that only DOMed DOs may scramble, small clause subjects are realized as marked DOs. (106) exemplifies this at stake with small clauses, where the DO is to be found in the subject position.

(1)

[Context: What happened yesterday to every boy?] Ayer lo atacó sui proprio padre a cada niño. Yesterday CL.3.ACC.SG. attacked his own father DOM each boy ‘His own father attacked every boy yesterday.’ (López 2012: 23, p. 43)

In (2) DO p-movement/scrambling has taken place and the DO has reached a position to the left of the one occupied by the EA. As such, the DO c-commands the EA and a quantifiervariable interpretation becomes available: (2)

[Context: What happened yesterday to every boy?] Ayer atacó a cadai niño sui proprio padre. Yesterday attacked DOM every boy his own father ‘His own father attacked every boy yesterday.’

(López 2012: 27, p. 44)

92

(106)

3 Linking syntactic structure to interpretation

vP ri v’ EA ro v αP ep NomPDOM α’ a un estudiante ri α VP ri V SC ri tNomP AP

3.2 Movement and DP internal structure 3.2.1 KPs vs. DPs As seen above, the data coming from binding dependencies support the hypothesis that while DOMed DOs in Spanish undergo short scrambling to a position between vP and VP, unmarked DOs stay in situ within VP. This section discusses the reasons underlying DOMed DO movement and relates this mechanism to case theory and the internal structure of the DO. López (2012) stresses the connection between the size of the nominal phrase and the case-marking strategy this nominal phrase goes by. Thus, nominals which do not project a full structure but only function as NPs or #Ps (along the lines of Cardinaletti and Starke 1999, Déchaine and Wiltschko 2002 or Ghomeshi 2008, Farkas and de Swart 2003, Martí 2008) check case in situ, by incorporating into the verb. Marked DOs, on the other hand, have a more comprehensive structure, projecting a K functional head (in line with Bittner and Hale 1996, Brugè and Brugger 1996, Brugè 2000, a.o.) as in (107) below: (107)

KP ei K[uC] DP ei D[uC] #P ei NP[uC] #[uC]

(López 2012: 32, p. 49)

3.2 Movement and DP internal structure

93

This internal structure that marked DOs evince blocks incorporation. As a consequence, these phrases need to move in order to acquire case. Movement to SpecαP is thus triggered by the [uC] of the marked DO. This allows v to probe the marked DO and to assign it [Acc] as described in (108) below:

(108)

vP ri v’ EA ri v[Acc] αP ri KP[uC] α’ ro α’ IO ri α VP ri V tKP

(López 2012: 33, p. 49) Spanish unmarked DOs are argued to project a smaller structure, as DPs, NPs or #Ps (109) and to undergo a case-checking mechanism based on incorporation.

(109)

DP ei D[uC] #P ei #[uC] NP[uC]

(López 2012: 30, p. 47) The type of incorporation at stake is, however, one which does not presuppose actual morphological incorporation of the DO into the lexical verb – it merely involves the copying of the highest nominal head onto the verb. The [uC] feature of the DO will then be checked once the lexical verb incorporates into v, which contains a valued [Acc] feature (110).44

44 López (2012) draws an important difference between KPs on the one hand and smaller nominal phrases on the other, proposing that the latter do not count as phases and that, as a consequence, they are sent to the interpretive systems together with the lexical verb. Since DPs/#Ps/NPs do not have phasal status, they cannot move either and as such may only satisfy

94

3 Linking syntactic structure to interpretation

V’ ei V D/#P ei ei V tD/# NP D/#[uC]

(110) a.

z__________________m v’ ep v VP ei ti v[Acc] tV V ei V D/#[uC] ! z--------------m

b.

(López 2012: 31, p. 48) The case of bare plurals may also be explained as an instance of incorporation. One important characteristic of these nominals is that they may never be DOMed: (111) Ayer vimos *a/ø mujeres. Yesterday saw.1.PL. *DOM/ø women ‘Yesterday we saw women.’

(López 2012: 37, p. 52)

One further important property of these nominals is that they do not undergo movement and must always remain postverbal as may be seen in (112), where movement of the bare plural unaccusative subject hombres in (112b) leads to ungrammaticality: (112) a. Llegaron hombres de todas partes. Arrived men of all parts ‘Men arrived from everywhere.’

their case requirements by way of incorporating into V. The option of moving to SpecαP is never available to them. Marked DOs, which are KPs, and thus have phasal structure are self contained and able to move. (As we will see, unlike their Spanish counterparts, Romanian unmarked DOs seem to be able to undergo movement even if they have a smaller, possibly DP, internal structure).

3.2 Movement and DP internal structure

b. *Hombres llegaron de todas partes. Men arrived of all parts ‘Men arrived from everywhere.’

95

(López 2012: 38, p. 52)

Building on previous work by Martí (2008), López explains the impossibility of a marking and movement as a consequence of their defective internal structure i.e., as #P, which enables them to check case by means of incorporation. Case requirements in this particular situation are thus satisfied without recourse to movement, as there is no K head blocking incorporation and eliciting movement. In the analysis put forth by López (2012) Accusative case may be checked in two different ways function of the internal structure of the DO in question: unmarked DOs and bare plurals have a defective structural make-up, DP/#P, and check case by means of incorporation into the lexical verb (by having their [uC] feature checked when V further incorporates into v through feature sharing i.e., v values the [uC] feature of the D/# head incorporated into V and then feature sharing takes place as a consequence of the syntactic dependency between the incorporated D/# head and its copy). Marked DOs, on the other hand check their [uC] feature by moving into a position, SpecαP, wherefrom they may be probed by the Accusative case assigning head v. Movement is triggered by the need to check Accusative case in a situation in which incorporation into V is blocked. Thus, contra other approaches to DO case assignment e.g., Belletti (1988), de Hoop (1996) a.o., López (2012) argues in favour of one single type of Case which is generally assigned by way of incorporation, unless the internal make-up of the nominal in question prevents it, in which case movement takes place as an alternative solution. The problematic situations involving small clauses, clause union and object control, where DOM was shown to be obligatory, also find an explanation, since the position occupied by the marked DO is one in which incorporation is prevented. Bare plurals are expectedly disallowed in these contexts as their defective internal structure would not allow them to check case otherwise than by means of incorporation.

3.2.2 Linking syntactic position and semantic composition Spanish data seem to point to a clearcut division between marked DOs on the one hand and unmarked DOs and bare plurals on the other, with respect to

96

3 Linking syntactic structure to interpretation

the different interpretive effects concerning specificity, scope, or binding that these nominals exhibit. López (2012) draws a connection between the different interpretation potential of these nominals and the syntactic position in which they surface, arguing that while marked DOs undergo short scrambling to a landing site lying between v and V, unmarked DOs and bare plurals stay in situ. Movement of the marked DO arises as a consequence of the internal make-up of the respective DP, which hinders regular Accusative case checking by means of incorporation. López further connects the in situ and ex situ positions with two different ways of semantic composition. In this, he partially builds on Diesing (1992) by adopting the view that scrambling to a position outside the VP is connected to the interpretation of the nominal phrase. Also, indefinite DOs which remain in their merge position inside the VP may only have a weak reading. Specificity and wide scope interpretation thus arise as a consequence of syntactic configuration.45 Nevertheless, López (2012) departs from Diesing (1992) in that he argues against a direct mapping between syntax and semantics. More specifically, the syntactic position a nominal ends into is linked to specific modes of semantic composition, which the respective syntactic position renders available. Short scrambling of the marked DO amounts to a precondition for the application of a choice function, which in turn accounts for the possibility of specific and wide scope readings with that DO. López (2012) builds on Reinhart (1997), who provides an explanation as to how wide scope interpretation may obtain independently of syntax by proposing choice functions as a solution. Example (113) below shows this at work: the indefinite starts out as and undergoes type shifting to under the effect of a choice function f which picks an individual from the set denoted by the (initially) property denoting nominal. The operation as such is called Satisfy (Chung and Ladusaw 2004). The type shifted nominal is then combined with the predicate via Functional Application. Finally, existential closure applies to f in line with Reinhart (1997): (113) Mary called a man. Lexical array: V:λx λe [called’(x)(e)]46

45 As such these readings do not depend on the actual morphological realization of K as a. The DOM marker is not a specificity trigger and only gets realized when the appropriate contextual conditions are met. This would explain why a marked DPs are not always specific or exhibit wide scope. See López (2012) p. 60 ff for an account of the spell out of K as a. 46 Building on Kratzer (1996), López (2012) adopts a compositional semantics for a syntactic theory according to which the external argument is introduced by v. Also, both v and V possess

3.2 Movement and DP internal structure

97

NomP: man’ v: λy λe [Init (y)(e)] NomP: Mary’ Derivation of VP: Satisfy: 1. Type-shift: man’! f(man’) 2. Merge (λx λe [called‘(x)(e)], f(man’)) 3. Function Application: λe [called’(e)( f(man‘))], Derivation of v: Merge (λyλe [Init (y)(e)], λe [called’(e)( f(man’))]) Event Identification: λyλe [Init (y)(e) ^ called’(e)( f(man’))] Merge (Mary’, λyλe [Init (y)(e) ^ called’(e)( f(man’))]) Function Application: λe [Init (Mary’)(e) ^ called’(e)( f(man’))] Existential Closure (e): 9e [Init (Mary’)(e) ^ called’(e)( f(man’))] Existential Closure (f): 9f 9e [Init (Mary’)(e) ^ called’(e)( f(man’))] (López 2012: 6, p. 75) This model of semantic composition may be used to account for the wide and intermediary scope readings that marked indefinites have been shown to allow. López (2012) claims that these interpretations depend on the exact point where Existential Closure applies. In (114) variant (a) is a representation of how the intermediary scope interpretation of the indefinite obtains, while (b) is an account of the wide scope reading of the same nominal: in (a) the existential quantifier closing the choice function variable is inserted such that it may be ccommanded by the universal quantifier; in (b) on the other hand, the existential is found to the left of the universal quantifier:

an event argument in their predicate structure and they conjoin if their event arguments are compatible via Event Identification. In the semantic structure thus obtained, the two event arguments end up being bound by the same lambda operator. Both v and V are analysed as functions: V is a function which takes events (s) as its arguments and spills out truth values: ; v is a function from entities to a function taking events as input and having truth values as output: . V and v function as the input to Event Identification, a rule whose output is a function from entities to functions from events to truth values . Consider (1) below, where the second line represents this at stake in lambda format and where v is rendered as Init i.e., denoting initiation in line with Ramchand 2008: (1)

v V Event Identification

! λ x λ e Init (x)(e) λV’(a’)(e) λ x λ e Init (x)(e) ^ λV’(a’)(e)

(López 2012: 2, p. 73)

98

3 Linking syntactic structure to interpretation

(114) Every woman saw a man. a. ∀y [woman (y)] 9f 9e [[Init (y)(e)] ^ [seen’(f (man’)) (e)]] ‘For every woman y there is a choice function f which picks out a man and there is an event e such that y saw the man picked out by f in e.’ b. 9f ∀y [woman (y)] 9e [[Init (y)(e)] ^ [seen’(f (man’)) (e)]] ’There is a choice function f that picks out a man and for every woman y there is an event e such that y saw the man picked out by f in e.’ (López 2012: 7, p. 76) Besides being able to acquire intermediary/wide scope readings with respect to other scope taking expressions, marked indefinites also evince specific interpretation. López (2012) argues that the type of specificity involved in this case is epistemic specificity (Farkas 1994/1995) or specificity as referential anchoring (von Heusinger 2002), which amounts to anchoring the discourse referent contributed by the respective indefinite DP to that of the speaker (epistemic specificity) or to that of the subject of the sentence (referential anchoring). The solution proposed by López (2012) as to the implementation of specific readings with marked indefinites presupposes the subscription of the choice function interpreting the marked DO as in (115a), with the interpretation in (115b): (115) a.

KP = fspeaker/subject (DP) ru K DP g fspeaker/subject

b. Juan vio a una cierta mujer. John saw DOM INDEF.ART. certain woman ‘Juan saw a certain woman.’ 9f CH (f) ^ Juan saw fspeaker (woman) ‘There is a choice function f, which picks out a woman, known to the speaker, and Juan saw the individual picked out by f.’ (López 2012: 20, 21, p. 83) Note at this point that, even though a choice function would in principle enable any indefinite to acquire wide scope with respect to another QP irrespective of its syntactic position, López (2012) argues that, at least for Spanish, (short) scrambling is actually a prerequisite for a choice function to apply: it is only those indefinite DOs which scramble to SpecαP that may outscope other QPs, even though

3.2 Movement and DP internal structure

99

they do not reach a c-commanding position with respect to these QPs. Apparently, short scrambling represents a precondition to choice function application. Unmarked DOs, which are of type never leave their in situ position and compose with a verb via Restrict (Chung and Ladusaw 2004), which crucially differs from Function Application (the operation by means of which marked DOs combine with the verb), in that the predicate does not get saturated in the process. More specifically, instead of saturating the predicate, the DP is merely conjoined with it, with the result that is acts as a restrictive modifier on this predicate. Thus, if the predicate is a function ranging over a domain of entities, and there is a property denoting DP that combines with it via Restrict, the application of this operation would result in a restriction of the domain to a subdomain enclosing only those entities that have the property denoted by the respective DP. Example (115) shows this at work: (116) Mary called a man. Lexical array: V:λx λe [called’(x)(e)] NomP: man’ v: λy λe [Init (y)(e)] NomP: Mary’ Derivation of VP: Merge (λx λe [called’(x)(e)], man’) Restrict: λx λe [called’(x)(e) ^ man’ (x)] Existential closure (x): λx λe [called’(x)(e) ^ man’ (x)]= λe 9x[called’(x)(e) ^ man’ (x)] Derivation of v: Merge (λyλe [Init (y)(e)], λe 9x[called’(x)(e) ^ man’ (x)]) Event Identification: λyλe [Init (y)(e) ^ 9x[called’(x)(e) ^ man’ (x)]] Merge (Mary’, λyλe [Init (y)(e) ^ 9x[called’(x)(e) ^ man’ (x)]]) Function Application: λe [Init (Mary’)(e) ^ 9x[called’(x)(e) ^ man’ (x)]] Existential Closure (e): 9e [Init (Mary’)(e) ^ 9x[called’(x)(e) ^ man’ (x)]] (López 2012: 5, p. 75) Just like in the case of marked indefinites, which acquire intermediate and/or wide scope with respect to other QPs, the mode of semantic composition may account for the narrow scope readings that Spanish unmarked indefinite DOs always acquire when co-occurring with other scope taking expressions. (117) shows this at work: the main idea is to allow the predicate to be saturated

100

3 Linking syntactic structure to interpretation

before existential closure applies to its event argument. An existential quantifier is consequently merged beforehand, binding the indefinite: (117) Every woman saw a man Derivation of VP: Merge (λx λe [saw’(sx)(e)], man’) Restrict: λx λe [saw’(x)(e) ^ man’ (x)] Existential closure (x): λx λe [saw’(x)(e) ^ man’ (x)] = λe 9x[saw’(x)(e) ^ man’ (x)] As discussed above, bare plurals behave on a par with unmarked indefinites in that they may only acquire a narrow scope interpretation with respect to other scope bearing expressions and never evince specific readings. This behaviour is argued to come as a consequence of the deficient internal structure of these nominals, which are of category #P and which lack the higher functional categories KP and DP. Given that they are nominal expressions, bare plurals bear a [uC] feature, which they check by way of incorporating into the lexical V, similarly to unmarked indefinites. Since they undergo incorporation, the only available mode of semantic composition for these nominals would be Restrict as proposed in (118): (118) Mary picked mushrooms. Lexical array: V:λx λe [pick’(x)(e)] NomP: mushrooms’ v: λy λe [Init (y)(e)] NomP: Mary’ Derivation of VP: Merge (λx λe [picked’(x)(e)], mushrooms’) Restrict: λx λe [picked’(x)(e) ^ mushrooms’ (x)] Existential closure (x): λx λe [picked’(x)(e) ^ mushrooms’ (x)]= λe 9x[picked’(x)(e) ^ mushrooms’ (x)] Derivation of v: Merge (λyλe [Init (y)(e)], λe 9x[picked’(x)(e) ^ mushrooms’ (x)]) Event Identification: λyλe [Init (y)(e) ^ 9x[picked’(x)(e) ^ mushrooms’ (x)]] Merge (Mary’, λyλe [Init (y)(e) ^ 9x[picked’(x)(e) ^ mushrooms’ (x)]])

3.2 Movement and DP internal structure

101

Function Application: λe [Init (Mary’)(e) ^ 9x[picked’(x)(e) ^ mushrooms’ (x)]] Existential Closure (e): 9e [Init (Mary’)(e) ^ 9x[picked’(x)(e) ^ mushrooms’ (x)]] These two ways of semantic composition thus account for the different interpretive potential of marked and unmarked DOs: unmarked indefinites compose by means of Restrict and as such they will necessarily have narrow scope with respect to other scope taking expressions (extensional QPs, intensional operators a.o.). Marked DOs compose by means of Satisfy and may outscope other scope taking expressions. The two modes of semantic composition are linked to two distinct syntactic positions: Satisfy applies at the level of Spec,αP, where marked DOs scramble, while Restrict interprets unmarked DOs which stay in situ. Scrambling to Spec,αP is triggered by the DP type of the indefinite DO. Marked DOs are KPs and have real argumenthood and are of type , while their unmarked counterparts have DP status and are property denoting nominals of type .47 López (2012) associates K with a semantic function which takes a nominal of type as its input and type-shifts it to . As such, K amounts to a choice function variable,48 as may be seen in (119): (119)

KP = f (DP) ru K DP g f

(López 2012: 13, p. 78) By associating K with a choice function, López (2012) manages to relate DOM, choice functions, and scrambling: K shifts the denotation of the DP it combines

47 In this López (2012) partially adopts the distinctions in Bleam (2005), who also analyses unmarked DOs as property-type nominals which incorporate into the verb, while allowing marked DOs to be interpreted by Function Application. Bleam, however, lumps marked indefinites with strong quantifiers and proposes the same ≪e,t>t> type for these nominals, disregarding their different scope taking abilities i.e., strong quantifiers are scope-rigid, as opposed to marked indefinites. 48 Note that K is only associated with a choice function when it combines with nominals of type . When combining with nominals of type or ≪e,t>t>, K will no longer be associated with the type-shifting function. K remains, however, crucial for semantics as it forces scrambling to SpecαP.

102

3 Linking syntactic structure to interpretation

with from to , securing argumenthood. The KP thus formed may no longer combine with the verb via incorporation and needs to move into a position wherefrom it may be assigned case by v. KPs then semantically combine with the predicate via Function Application. Example (120) below shows this at work: the marked indefinite DO has KP status and has to undergo short scrambling to SpecαP in order to get case by having its [uC] feature valued by v, given that K hinders incorporation with the lexical verb. KP leaves a trace behind, of type , which may combine with V. The lexical verb then combines with α, which may either be composed with V by Event Identification (if it functions as an applicative introducing an IO) or simply contribute inner aspect. The moved KP (f(man)‘) introduces a lambda operator (along the assumptions of Heim and Kratzer 1998) and gets to function as the argument of saw‘ by way of lambda conversion. v is then merged and the event arguments of v and V are then conjoined via Event identification; the EA is further merged and composed with the predicate by regular Function Application. Finally, Existential closure applies to e and then to the choice function (b): (120) a. María vio a un hombre. Mary saw DOM INDEF.ART. man ‘Mary saw a man.’ b.



’ ’

’ ’

’ ’









’ ’





’ ’

’ ’ ’ ’

3.2 Movement and DP internal structure

103

As already seen, unmarked DOs combine via Restrict. Variant (b) in (121) below shows how this comes about: (121) a. María vio un hombre. Mary saw INDEF.ART. man ‘Mary saw a man.’ b.

Ǝe [Init (Mary’) (e) Λ Ǝx[saw’ (e)] (x) Λ man (x)] wp vP = λe [Init (Mary’) (e) Λ Ǝx[saw’ (e)] (x) Λ man (x)] Ǝe wp v’ = λyλe [Init (y) (e) Λ Ǝx[saw’ (e)] (x) Λ man (x)] Mary’ wp o v αP λyλe [Ini (y) (e)] wp α’ wu VP = Ǝx[λe [saw’ (e)] (x) Λ man (x)] αo wi VP = λx[λe [saw’ (e)] (x) Λ man (x)] Ǝx wi λx [λe [saw’(e)] (x)] man’(x)

Note that not only marked indefinites are KPs, but all the nominals that may be DOMed. Definite descriptions, proper names and quantified noun phrases also bear DOM in Spanish and are analysed as KPs. KPs denote individuals or generalized quantifiers and as such may not occupy the complement position of the verb as they do not have the appropriate denotation i.e., the property, one. Thus, they need to reach those positions where individuals i.e., property instantiations (Carlson 2003) get introduced i.e., either αP or vP. Both αP and vP denote a token event, when an argument is merged at this level, an event type is produced. Thus, it is only weak indefinites or bare plurals that are allowed to stay in the complement position of V. Weak indefinites and bare plurals are nominals which exhibit a smaller internal structure. They get merged as complements of V, a position which may only host property denoting nominals. Lexical verbs which combine in this way denote eventualities and the combination of a nominal of type with these verbs would only yield another, more specific, event type obtained by way of restricting the one denoted by the verb in isolation. Thus, the internal make-up of nominal phrases has consequences on their syntax, forcing movement or permitting them to stay in their merge positions. As a consequence, case is checked either by way of agreement with v or by means of incorporation of the V. Also, semantic interpretation derives as a consequence of the connection existing between syntactic positions and modes of semantic composition.

104

3 Linking syntactic structure to interpretation

3.2.3 Conclusions López (2012) proposes an account of indefinite DOs, which presupposes an indirect connection between syntactic position and semantic interpretation. More specifically, it is argued that the choice between moving a nominal expression and allowing it to remain in its in situ position has consequences on the way in which the respective nominal semantically composes with the predicate: while nominals staying in situ get composed by means of Restrict, nominals that leave the VP are subject to another mode of semantic composition, undergoing type-shifting under the effect of a choice function and then combining with the predicate via regular Function Application. Leaving or staying within the merge position is influenced by the internal make-up of the nominal: DPs/#Ps are nominals and stay in situ, while nominals with or denotations are KPs and forced to move, given that K hinders incorporation; also, the lexical verb itself may only compose with property denoting nominals. As a consequence, KPs move to SpecαP, where individual denoting nominals may be introduced into the structure. The two different ways of semantic composition has consequences on the interpretation of the nominals involved: KPs may acquire specific and wide scope readings as they get interpreted by means of a choice function, which allows such a reading. DPs/#Ps on the other hand are never subject to an interpretation by means of a choice function and never acquire specific or wide scope readings. Note also that all KPs which have status and move to SpecαP do not necessarily have to spell out as a marked nominals: a marking presupposes reaching SpecαP, but this is not a sufficient condition in this respect. Rather, whether K is realized as a or as ø represents a matter which is resolved at PF and which is sensitive to the immediate context of KP. López (2012) enumerates several relevant aspects in this respect: the properties of the DP which K takes as complement (definiteness etc); the properties of the NP within the DP i.e., animacy; aspectual properties etc.

3.3 On the differences between Spanish and Romanian DOs 3.3.1 Specificity, wide scope and binding As shown above, Romanian DOs differ from their Spanish counterparts in that both DOMed and unmarked indefinites may be interpreted as specific and may outscope other scope bearing expressions, whether extensional or intensional. Furthermore, both marked and unmarked DOs are allowed as arguments in small

3.3 On the differences between Spanish and Romanian DOs

105

clauses, clause union and object control contexts. Bare plurals, on the other hand pattern with their Spanish counterparts, in disallowing such readings. Binding is also somewhat problematic in that both DOMed and unmarked DOs seem to be able to bind into the IO.49 In an experiment concentrating on possible binding dependences between the DO and the IO, which will be extended upon in the subsequent sections, von Heusinger and Tigău (ms.) show that a quantifier-variable reading is available both with pe marked and unmarked DOs in examples such as (122). Note that binding holds irrespective of the word order of the two arguments. (122) Teroriștii au predat comandantului săui Terrorists.DEF.ART. have surrendered commander.DAT his (pe) fiecare soldati luat prisoner. DOM every soldier taken prisoner ‘The terrorists handed in every soldier take hostage to each commander.’ The same is at stake in (123a) below, where the DO may not be read specifically, such that the argument put forth in Leonetti (2004) according to which the availability of the quantifier – variable reading comes as a consequence of the specific reading of the marked DO and not necessarily from the c-command relations per se. It seems to us, that a quantifier-variable reading is available both with the marked as well as with the unmarked DO, with a preference, however for the marked variant. Note that, when trying to use a negative quantifier nimeni (no one) DOM is obligatory, even if no specific reading may be available.50

49 Note, however, that this might not be a relevant argument, given that some studies on Romanian propose a very low merge position for the IO, which is c-commanded by the DO (Cornilescu et al. 2017a, b). One argument in favour of such a position is the possibility of allowing two dative DPs within the same sentence. The lower dative is always a core dative and may be bound by a quantifier within the DO: (1)

Context: How did you help the Browns marry all their daughters? Le-am prezentat fiecare fată unui amic de-al meu. CL.3.DAT.PL. presented every girl INDEF.ART.DAT friend of mine ‘I have introduced every girl of theirs to a friend of mine.’

50 Cornilescu (2000) discusses these examples in relation to negative quantifiers denoting inanimate referents, which are never DOMed and proposes that pe is a marker of personal i.e., +human, gender: (1)

Nu am văzut pe nimeni/ (*pe) nimic. Not have.1.SG. seen DOM no one (*DOM) nothing ‘I have not seen anybody/ anything.’

106

3 Linking syntactic structure to interpretation

The DO may still bind the IO in this case. Unfortunately, one may no longer distinguish between a DOMed variant and a non-marked one of the DO in this case. (123) a. Teroriștii nu au predat comandantului săui Terrorists.DEF.ART. not have surrendered commander.DAT his (pe) nici un soldat luat prizonieri. DOM no soldier taken prisoner ‘The terrorists handed no soldier taken prisoner to his commander.’ b. Teroriștii nu au predat comandantului săui Terrorists.DEF.ART. not have surrendered commander.DAT his (pe) nimeni luat prizonieri. DOM no one taken prisoner ‘The terrorists handed no one taken prisoner to his commander.’ Cornilescu et al. (2017a) experimentally study binding with anaphors within ditransitive configurations.51 Examples such as (124) are thus investigated and it is argued that the DO may bind the anaphor within the IO. Incidentally, in this particular case, we find the unmarked DO to fare better than its unmarked counterpart with respect to binding: (124) Este vorba despre un roman SF în care, la un moment dat, personajul principal, Bill, ajunge să prezinte chiar ei înseșii (pe) o prietenăi, dar la o altă vârstă. Autorul vrea sa arate astfel că există lumi paralele și că aceiași indivizi pot evolua diferit în contexte diferite. Lit. ‘This is a Science Fiction novel where, at some point, the main character, Bill, ends up introducing to herselfi a friendi, but at a different age. The author wants to show that parallel worlds exist and that the same individuals may evolve differently in different contexts.’ With respect to binding dependences involving the subject DP, neither the DOMed DO nor its unmarked variant seem to be able to bind into the subject: a quantifiervariable reading is not available.52

51 See also the experimental results regarding anaphor binding within ditransitives in Chapter 6. 52 Cornilescu and Dobrovie-Sorin (2008) draw attention to the fact that a clitic doubled and DOMed DO may bind into the subject. Thus the CD and DOMed DO pe nici un copil (any child) is said to bind the possessive within the dative DP:

3.3 On the differences between Spanish and Romanian DOs

107

(125) Ieri, la ședința cu părinții, nu a criticat Yesterday at meeting with parents.DEF.ART. not has criticised nici un copili. tatăl săui pe Father.DEF.ART. his DOM no child ‘At the parents meeting yesterday hisi father did not criticise any childi.’ The data concerning Romanian DOs described above seem to parallel to a certain extent with Spanish DOs. Thus, in both languages marked DOs may evince specific and wide scope readings, may bind a possessive pronoun or an anaphor within the IO and may not bind into the subject DP. This similar behavior suggests that Romanian marked DOs may be analysed on a par with their Spanish counterparts and argued to scramble into SpecαP in line with López (2012). Just as with Spanish DOMed DOs, movement is triggered by case reasons: the marked DO, having a KP internal structure which hinders case checking by incorporation, is forced out from its in situ position to a landing site where it may be probed by the v head. Unlike in Spanish, however, Romanian unmarked DOs behave on a par with their marked counterparts with respect to all the relevant phenomena (specificity, wide scope, binding, small clauses, clause union, object control). This similarity in behaviour prompts us to propose that these DOs also undergo the same checking mechanism by way of scrambling to SpecαP. Reaching this position would allow both marked and unmarked DOs to also acquire specific and wide scope readings as they would be interpreted with the aid of choice functions, and would also account for the binding facts. Thus, the parameter which differentiates between Romanian and Spanish addresses the types of nominals which may scramble. As it seems, Romanian allows both KPs as well as (some) DPs to scramble, while Spanish only allows the scrambling of KPs.

3.3.2 Weak indefinites and bare nominals Not all unmarked indefinites exhibit strong readings. In (126) the unmarked indefinite un șoc (a schock) may only allow a narrow scope reading with respect to the QP subject.

(1)

Ieri, la ședința cu părinții, nu l-a criticat Yesterday at meeting with parents.DEF.ART. not CL.3.ACC.SG.-has criticised tatăl săui pe nici un copili. Father.DEF.ART. his DOM no child ‘At the parents meeting yesterday hisi father did not criticise any childi.’

108

3 Linking syntactic structure to interpretation

(126) Orice tânăr care nu învață de acasă să își Any young man who not learn from home SUBJ. REFL.DAT poarte singur de grijă, va suferi un șoc când va carry alone of care will.3.SG suffer a shock when will.3.SG pleca la facultate într-alt oraș. leave at college in-other city ‘Any young man who does not learn from home how to take care of himself, will suffer a shock when he has to go to the university in a different city.’ Similarly, the unmarked indefinite o rujeolă (a measles) may not outscope the QP subject or the conditional in (127). (127) Dacă nu se iau măsuri, toți copiii din If not REFL. take.3.SG. measures all children.DEF.ART. of școală vor face o rujeolă de toată frumusețea. school will.3.PL make INDEF.ART. measles of all beauty ‘If measures are not take, all the children in school will catch really serious measles.’ In the examples above, the unmarked indefinite DOs may only have a property reading and behave on a par with bare plurals. Indeed, while marked DOs seem to only have status, unmarked DOs may have both and status. Consider also the following contexts, which require the use of nominals with property reading and disallows nominals denoting entities. As it may be seen unmarked indefinites and bare plurals are allowed in these contexts. Marked indefinites are out (for an extensive discussion see Cornilescu 2000, pp. 31 ff). Consider first the case of post-verbal subjects with se (reflexive) passive. As shown by Cornilescu (2000), only nominals denoting properties may be allowed in this context. As a consequence, bare plurals are allowed in this context (128b), while proper names, given their (only) type reading are discarded as ungrammatical (128a). Indefinites are also allowed in this context, which shows that they may be reanalyzed as NP/NumP and have a property reading: (128) a. S-a adus Maria de la școală. REFL.-has brought Mary from school ‘Mary was brought from school.’ b. S-au băgat banane la aprozar. REFL.-has brought bananas at store ‘They have supplied the store with bananas.’

3.3 On the differences between Spanish and Romanian DOs

109

c. S-a adus de curând o girafă la REFL.-has brought of soon INDEF.ART. giraffe at grădina zoologică. garden zoological ‘The zoo was endowed with a giraffe.’ As also noticed above, unmarked indefinites may function as the DO of have in its individual-object reading, a situation in which bare singulars and plurals are also allowed, while marked DOs are discarded. The observation was first drawn by Cornilescu (2000) building on Bleam (1999): (129) a. Maria are copii deștepți. Mary has children smart.PL. ‘Mary has intelligent children.’ b. Maria are un copil deștept. Mary has INDEF.ART. child smart ‘Mary has an intelligent child.’ c. *Maria are pe un copil deștept.53 Mary has DOM INDEF.ART. child smart ‘Mary has an intelligent child.’ Some intensional verbs impose the same restrictions: cere (ask), dori (wish), a pretinde (claim), vrea (want): (130) a. Maria dorește copii deștepți. Mary wishes children smart.PL ‘Mary wants intelligent children.’ b. Maria dorește un copil/ ?pe un copil deștept. Mary wishes INDEF.ART. child DOM INDEF.ART. child smart ‘Mary wants an intelligent child.’

53 This example may become acceptable under a reading along the lines of (1) below. Note that the meaning in this case is, however, significantly different: (1)

Maria are pe un copil deștept care îi va Mary has DOM INDEF.ART. child smart who CL.3.DAT.SG. will.3.SG. rezolva problema în câteva minute. solve problem.DEF.ART. in few minutes ‘Mary knows a smart child who will solve the problem in a few minutes.’

110

3 Linking syntactic structure to interpretation

Tănase-Dogaru (2009) analyzes bare objects i.e., determineless nominals (bare singulars and plurals): (131) a. Maria vrea bărbat bogat. Mary wants man rich ‘Mary wants a rich husband.’ b. Maria n-are copii. Mary not-has children ‘Mary does not have children.’ As it seems, these nominals never evince specific readings and may not outscope extensional QPs or intensional operators. In fact, as pointed out as early as Carlson (1977), these nominals have the narrowest scope. Indeed, the bare plural colegi (colleagues) in (132a) may only exhibit a dependent reading on the QP subject according to which the colleagues who get invited to dinner vary function of the employee. A wide scope reading interpretation according to which ‘there are some colleagues such that any employee would invite to dinner’ is completely out. Along the same lines, the bare plural in (132b) cannot outscope the conditional as no particular group of colleagues is targeted such that if they get invited, John gets upset: (132) a. Orice angajat mai invită colegi la masă din când în când Any employee more invite colleagues at dinner sometimes chiar dacă soția se supără. even if wife.DEF.ART. REFL. get annoyed ‘Any employee invites colleagues to dinner from time to time, even though this might upset the wife.’ b. Dacă Maria invită colegi la masă, Ion se supără. If Mary invites colleaguse at dinner John REFL. get annoyed ‘If Mary invites colleagues to dinner, John gets upset.’ In (132) above, the constituent ‘inviting colleagues‘ designates a particular type of complex activity, where the bare nominal acts as a restrictive modifier on the predicate. Tănase-Dogaru (2009) shows that bare nominals are ‘small size nominals’ and draws the conclusion that bare objects should be analyzed as NPs or at most NumPs which can only combine with the verb by way of incorporation – at least when they merge in complement position. One further argument supporting this hypothesis is that bare singulars may only merge as VP complements. Similar

3.3 On the differences between Spanish and Romanian DOs

111

observations have been drawn with respect to Spanish bare plurals, which also seem to obey the restriction of only surfacing in post-verbal position (Brugè and Brugger 1996). The contexts presented above show that unmarked indefinites may pattern with bare nominals: they require property denoting nominals (bare plurals, bare singulars) and may accommodate unmarked indefinites which may also be property denoting. In this particular case, indefinites are reanalyzed as NP/NumP and remain in situ and incorporate into the verb (Tănase-Dogaru 2009). As a consequence, they get interpreted via Restrict and never acquire specific, or wide scope readings. Building on Cornilescu and Nicolae (2012), Cornilescu and Tigău (ms.), we propose that indefinites in their weak reading (as properties) are projected as NPs: in this particular case, the D head is not projected as such and the weak indefinite article functions as a quantifer or as an adjective, occupying the Specifier position of the nominal head: (133)

nP ru Art n’ [idef] ru un n NP g N [+N] [iφ] [udef] elev

This is in line with the proposals already advanced for definite descriptions, which may also exhibit property readings and which have been analyzed as NPs in this case (Cornilescu and Nicolae 2012). Consider first the examples under (134), where the definite nominal exhibits a property reading. In this case the definite expression seems to be a NP/NumP, where the definite article functions as the specifier (135) along the lines of Emonds (1985) who argues that determiners are in fact former adjectives and as such function as specifiers of nominal heads. Cornilescu and Nicolae (2012 ff) support this hypothesis of the reanalysis of the DP by drawing attention to the fact that the definite article is a suffix in Romanian. Mark that reanalysis is only necessary for property readings of definites. Strong determiners will follow the lines of analysis in (136) since they are referential:

112

3 Linking syntactic structure to interpretation

(134) a lua trenul / a bate drumurile to take train.DEF.ART. to beat roads.DEF.ART. ‘to take the train/ to wander the roads’ (135)

nP ru Art n’ [idef] ru n NP g N [+N] [iφ] [udef] elev+(u)l pupil+the

(136)

DP ru D nP [idef] ru n’ ru n NP g N [+N] [iφ] [udef] elev+(u)l pupil+the

On a par with definite descriptions in their weak readings and bare plurals, weak indefinites will thus be analysed as NPs. This analysis is strengthened by the fact that, with definite descriptions, the D head has no phonological context, given the suffix status of the definite determiner. As discussed, marked indefinites along with proper names or personal pronouns are rejected from these contexts. This is because these nominals only have the object-level, type reading. The contribution of K is to secure argumenthood. As a consequence, nominals thus marked may only be interpreted as arguments (never as properties) and need to scramble to a position where they may combine with the verb by means of functional application. In this respect,

3.3 On the differences between Spanish and Romanian DOs

113

Romanian marked indefinites pattern with marked DOs in Spanish. Bleam (1999, 2005) shows that a-marked nominals denote individuals or generalized quantifiers and argues that the DOM marker acts as a filter on the DP denotation, filtering away the property reading. On the other hand, unmarked indefinites may function as DPs and as such may give rise to the same readings as marked indefinites (specificity, wide scope, binding dependencies) and undergo scrambling to SpecαP for case reasons. The fact that they reach this position, enables them to get interpreted via Choice Functions, on a par with their marked counterparts. Thus, DOM seems to have grammaticalized to such a point that both marked and unmarked indefinites may evince the same readings. A suitable analysis of Romanian data should thus account for the contrasts between Romanian and Spanish unmarked indefinites: the Spanish a-stripped nominals may only be property denoting combining with the verb by means of incorporation and never leaving their merge position. Romanian unmarked indefinite DOs exhibit a two-fold behaviour: as DPs, they may have argumental status and pattern on a par with marked DOs, or they may undergo reanalysis to NP/NumP and behave similarly to other property denoting nominals (e.g., bare plurals). In this latter case, they closely follow in the footsteps of Spanish unmarked DOs. The basic tenets of DO analysis in the two languages may be schematically represented as in table (9) below. As it seems, while DPs in Spanish incorporate (López 2012), Romanian only allows for the incorporation of nominals with an even smaller internal structure i.e., NP/NumP; nominals with DP status in Romanian do not incorporate: Table 9: DP status in Spanish and Romanian. Language

DOMed DOs

non-DOMed indefinite DOs

bare plurals

Spanish

argument status or KPs

properties

DPs/#Ps

properties

DPs/#Ps

Romanian

argument status or KPs

argument status or DPs

properties

NP/NumP

properties

NP/NumP

The parameter that differentiates between Spanish and Romanian thus has to do with the types of nominals that must undergo scrambling: while scrambling is restricted to KPs in Spanish, Romanian allows both KPs and DPs to

114

3 Linking syntactic structure to interpretation

scramble. Also, while Spanish DPs incorporate, Romanian allows incorporation only with nominal expressions whose internal structure is smaller than DP i.e., NP/NumP (137). (137)

KP > DP > NumP > NP

As argued by Longobardi (1994 ff), the locus of argumenthood is the determiner, which is sufficient for reference. Consequently, nominal expressions with supplementary marking of referential status also count as arguments and, as such, must undergo scrambling and compose with the predicate by means of Function Application.

3.3.3 Conclusions One first conclusion drawn with respect to Romanian DOs in this section was that they merge as c-selected complements of V and need to check Accusative case. In line with López (2012) and general minimalist assumptions, we take v to be the source of Accusative case. On a par with Spanish, Romanian resorts to two strategies of case valuation: DOs projecting as NPs/NumPs check case in situ by means of (semantic) incorporation. This amounts to having their [uCase] feature valued when the lexical verb mounts to v. Incorporating objects are interpreted as predicates/ properties (entities of type ) and form complex predicates with V. DOs projecting as DPs or KPs in Romanian may not incorporate and have to move to a higher position between v and V, where they verify their [uCase] feature by way of agreement with v. The fact that these nominals undergo scrambling finds support in the availability of strong readings i.e., wide scope, specificity a.o. Scrambling DOs are interpreted as arguments and combine with the verb by Function Application. Marked and unmarked DOs exhibit the same distribution in Romanian, in sharp contrast with Spanish where marked DOs pattern differently from their unmarked counterparts with respect to scope readings, specificity, binding etc. Note, however, that Romanian exhibits some syntactic differences between marked and unmarked objects, which have to be explained: firstly, marked DOs may be CD-ed, while their unmarked counterparts may not. This difference does not follow from anything we have discussed so far and it will be addressed later. Secondly, marked DOs necessarily scramble, while this remains an option for unmarked objects, which may scramble. This difference follows from the fact that, while DPs may be re-analyzed as NP/NumPs, KPs may not.

3.4 On the featural make-up of KPs

115

Finally, scrambled DOs, whether marked or unmarked are semantically equivalent, in that they evince the same scope properties, their reference may be stable (“epistemically specific”) or unstable (non-specific). Specificity does not trigger DOM, instead, DOM is merely associated with a cluster of pragmatic effects. Note that Romanian is not the only language where DOM has undergone grammaticalization to such a point that both marked and unmarked DOs have the same range of readings. Kannada seems to display the same properties (Lidz 2006).

3.4 On the featural make-up of KPs This section is devoted to the examination of the featural make-up of marked and unmarked DOs in Romanian and argues in favour of the hypothesis that the marker K of DOM, which surfaces as pe54 is the spell out of a [Person] feature. Analysing Romanian DOM along these lines (cf. Cornilescu 2000, Cornilescu and Tigău ms.), will allow us to better predict when this mechanism is obligatory, optional or rejected in Romanian, a fact which seems to generally depend on the interplay between definiteness and [Person] within nominal expressions.

3.4.1 Sensitivity to animacy and definiteness In all the languages that have DOM, it has been noticed that this mechanism is sensitive to the animacy and the definiteness hierarchies (Aissen 2003). Romanian DOM makes no difference in this respect, DOM marking DOs which are animate, usually human or human-like, or at least considered human (e.g., pets, large animals); inanimate DOs are rejected as ungrammatical when DOMed (138). In (139) below, sentences (a) and (b) are grammatical, with DOM marking a human referent denoting DO in (a), and an animate one in (b). (139c) on the 54 Pe has evolved from the locative preposition p(r)e (on) and currently exhibits a similar behaviour to the Spanish DOM marker a, marking more proeminent DOs. Diachronic studies seem to support the view of a ‘pendulating‘ pe, with a XVIth c. pe sensitive to animacy and a XVIIth c. one which had been bleached functioning as a generalized Accusative case assigner and marking both animate and inanimate direct object DPs (Stan 2009). More recent studies (Cornilescu et al. 2017b, Cornilescu 2017) capitalize on this distinction and posit a [iPerson] feature on pe marking undoubled DOs, as opposed to the pe in clitic doubled DOs which as analyzed as a mere case marker, having undergone further grammaticalization. This account seems to also find support in recent experimental findings (Avram 2014, Cornilescu et al. 2017a, Tigău and von Heusinger ms. Ditransitives).

116

3 Linking syntactic structure to interpretation

other hand is out, as the DO denotes an inanimate referent. In this respect, DOM marks a class of atypical DOs, in the sense that these DPs are more Subject-like (through their agentivity); typical DOs are inanimate and interpreted as Themes: (138) Animacy scale Human > Animate > Inanimate (139) a. Maria ajută la teme pe colegii care nu Mary helps at homework.PL. DOM colleagues.DEF.ART. who not înțeleg lecția în clasă. understand lesson.DEF.ART. in class ‘Mary helps the colleagues who have not understood the lessons during class with their homework.’ b. Ion mângăie pe cățel și îi dădu să John patted DOM puppy and CL.3.DAT.SG. gave SUBJ. bea apă. drink water ‘John patted the puppy and gave it some water.’ c. *Maria admira pe grădina plină de flori. Mary admired DOM garden.DEF.ART. full.FEM of flowers ‘Mary admired the garden full of flowers.’ The fact that DOM is sensitive to animacy may also be seen in the contrasts exhibited by bare quantifiers: while QPs denoting [human] referents are obligatorily marked, their counterparts pointing to inanimate referents reject DOM: (140) [+human] cineva (someone) oricine (anyone) nimeni (nobody)

[-human] ceva (something) orice (anything) nimic (nothing)

(141) Cred că am auzit pe cineva/ (*pe) ceva. Think.1.SG. that have.1.SG. heard DOM someone (*DOM) something ‘I think I have heard someone/something.’ Romanian DOM is also sensitive to the definiteness hierarchy: individuation has been shown to be an important ingredient, with DOM being associated with DPs denoting particularized referents or referents which are familiar or salient to the speaker. In (52) below the general expectation is to find DOM with those DOs that range higher on the scale with respect to individuation. Farkas and

3.4 On the featural make-up of KPs

117

von Heusinger (2003) point that individuation should be understood in terms of stability of reference across situations that a certain nominal expression exhibits. Thus, personal pronouns and proper names are obligatorily DOMed since they are rigid designators and hence referentially stable (143a, b). At the other end of the scale, non-specific indefinites are never marked, given their nonstable reference (53c). Definite descriptions are optionally DOMed (53d): (142) Definiteness Scale Personal pronouns > Proper names > definite descriptions > specific indefinites > non- specific indefinites (143) a. Ion m-a ajutat și pe mine/ *mine la John me.ACC-has helped and DOM me.Acc me.Acc at teme. homework. ‘John also helped me with the homework.’ b. Ion ajută pe Maria/ *Maria la teme de câte ori John helps DOM Mary Mary at homework whenever poate. can.3.SG ‘John helps Mary with her homework whenever he can.’ c. Un guvern corupt nu va ajuta *pe INDEF.ART. government corrupt not will.3.SG. help DOM oameni nevoiași, ci va fura cât mai mult. people poor.PL but will.3.SG. steal as much as possible ‘A corrupt government will not help poor people, but will steal as much as possible.’ d. Prin acest program, statul vrea să ajute (pe) Through this program state.DEF.ART. wants SUBJ. help DOM oamenii nevoiași. people.DEF.ART. poor.PL ‘Through this program, the state wants to help the poor.’ Lexical phrases headed by strong quantifiers pattern with definite descriptions allowing DOM (144a, b). Note that the use of DOM in this particular situation is discarded as ungrammatical with QPs denoting inanimate referents (144c, d): (144) a. Maria a ajutat (pe) toți colegii la teme. Mary has helped DOM all colleagues at homework ‘Mary has helped all the colleagues with their homework.’

118

3 Linking syntactic structure to interpretation

b. Ion a ajutat (pe) fiecare coleg la teme. John has helped DOM every colleague at homework ‘John helped every colleague with his homework.’ c. Maria a citit *pe toate cărțile din bibliotecă. Mary has read DOM all books from library ‘Mary has read all the books in the library.’ d. Maria a citit *pe fiecare carte din bibliotecă. Mary has read DOM each book from library ‘Mary has read every book in the library.’ An interesting aspect noticed by Cornilescu (2000) in relation to strong quantifiers is their different bevahiour with respect to DOM when they function as strong pronominal quantifiers. Thus, as opposed to their determiner use presented in (144) above which is sensitive to animacy, strong pronominal quantifiers obligatorily require DOM irrespective of whether they denote an animate (145a) or an inanimate referent (145b): (145) a. Maria a ajutat pe/*ø toți la teme. Mary has helped DOM/*ø all at homework ‘Mary has helped everybody with their homework.’ b. Maria nu a citit multe cărți din bibliotecă, dar Mary not has read many books from library but Ion le-a citit pe/*ø toate. John CL.3.ACC.PL-has read DOM/*ø all ‘Mary has not read many books from the library, but John read them all.’ These data, which have been so far left unexplained, show that, while DOM surfaces in connection to animacy and definiteness, a purely semantic or pragmatic approach is not sufficient. This section will try to provide an answer to this puzzle by building on the internal structure of the KP and the role of K itself (for a more detailed analysis see Cornilescu and Tigău ms.)

3.4.2 Integrating animacy and definiteness into syntax In the previous sections it has been shown that the contribution of K is to shift the interpretation of the nominal it marks from to (and to leave it unchanged if the nominal in question already denoted entities or functioned as a generalized quantifier). Hence, K has the role of securing argumenthood in the sense that it is called for when a person denoting nominal is explicitly marked

3.4 On the featural make-up of KPs

119

for an object-level reading.55As a consequence, it will be obligatory with those nominals that only allow an type reading and excluded with nominals that lack such readings. This explains, for instance, why personal pronouns or person denoting proper names obligatorily get DOM56 as they may only have type denotations and why DOM is rejected with bare plurals and bare singulars, which are only property denoting nominals. Another important aspect discussed so far has to do with the fact that Romanian allows both KPs and DPs to scramble. Thus, since both these types of nominals scramble and are interpreted as arguments, the projection of the KP layer is determined by the local context i.e., mainly by the internal feature structure of the nominal expression. Building on Fábregas (2013), Cornilescu and Tigău (ms.) list the following morpho-syntactic properties as relevant factors triggering the insertion of K at PF: (146)

1. The properties of the determiner that K combines with: definite vs. indefinite determiners. 2. The properties of the NP contained within the DP: whether the N is [+HUMAN] and whether this lexical property is syntactically interpreted as a [Person] feature. 3. Other properties of vP e.g., the semantic category of the subject, aspect a.o.57

55 Evidence that DOMed DOs may only have an object level reading was discussed in the previous section where it was shown that these DPs never surface in contexts requiring the property reading e.g., individual level have etc. Proper names and personal pronouns are also rejected from these contexts. 56 In fact, person denoting proper names and personal pronouns are the only nominals obligatorily getting DOMed. 57 Cornilescu and Tigău (ms.) discuss examples such as the one under (1), where the semantics of the subject DP seems to be relevant for the use of DOM on the DO in the sense that, while with agentive subjects both DOMed and unmarked DOs are allowed, with inanimate subjects there is a clear-cut preference for DOMed DOs (1a,b): (1)

a. Haina face pe/*ø om/ ?omul. Coat makes DOM/*ø man man.DEF.ART. ‘Clothes make the man.’ surprinde pe/*ø bătrân/ ?bătrânul adormit. b. . . . imaginea image.DEF.ART. catches DOM/*ø old man old man.DEF.ART. asleep ‘The image catches the old man asleep.’ Cornilescu and Tigău (ms.): (105), p. 32

120

3 Linking syntactic structure to interpretation

In line with López (2012), we consider the DOM marker pe as a functional category labelled as K (Tigău 2016) and extending the functional domain of the nominal phrase. Since K is external to definiteness and indefiniteness markers, it will select a DP complement acting as a phrasal affix as in (147): (147)

KP eo K[uC] DP pe ei #P/NP D[uC] femeie o

In line with Cornilescu and Tigău (ms.), we take pe to be a functional category whose distribution follows from the fact that it represents the spell-out of a valued or unvalued [iPerson] feature.58 As seen above, definiteness represents another feature which plays a critical part in the distribution of DOM, besides the [Person] feature. Following Richards (2008), we take [Person] to be the syntactic counterpart of both the definiteness and animacy scales: “Our claim is that [Person] in the syntax is just animacy/definiteness at the (semantic) interface. That is, we assume that there is a single, discrete, binary property [± person] whose presence vs. absence correlates with high vs. low interpretations (on the definiteness and animacy scales) in the semantic component” Richards (2008): p. 139. Richards (2008) thus argues that [Person] belongs in fact to both the animacy and the definiteness scales as below: (1)

Person /animacy scale [+Person] (DP) | [-Person] NP 1/2 personal pron. > animate (3-p. pron./nouns > inanimate (3-p. pronoun/ nouns ← likelihood/obligatoriness of animacy

(2)

Person/definiteness scale [+Person] (DP) | [-Person] 1/2 personal pronouns > 3-person, pronoun > definite >specific > non-specific ← likelihood/obligatoriness of definiteness

58 A similar proposal for the Spanish a marker may be found in Leonetti (2004), RodriguezMondoñedo (2007).

3.4 On the featural make-up of KPs

121

These hierarchies enable us to draw a number of conclusions as to certain redundancy rules of the system. Firstly, it is only [+animate] NPs that may vary for Person: all inanimate NPs are 3rd person and do not require any specification in this respect. Secondly, only [+definite] nominals allow for Person variation, while indefinites, irrespective of whether they are specific or not, are automatically 3rd Person and as such do not need any [Person] specification. It seems that with indefinites and inanimates the [Person] feature is both interpretable and valued ([iPerson: 3]) and therefore is inactive, which means that it may go unrepresented. We will assume that [Person] is a syntactic property of definite and animate (+Human) nouns only. Cornilescu and Nicolae (2012) argue for Romanian that definiteness has become a syntactic feature marked on the nominal stem and valued by means of Agree with a corresponding interpretable feature on D. (148) shows how definiteness is resolved at the DP level by way of Agree. In Romanian both adjectives and nouns agree for definiteness – Drawing on examples such as (149), Cornilescu and Nicolae (2012) argue that Definiteness is spelled-out once in the highest [+N] position, of the DP: (148)

D ei D NP [uφ:__] N [idef__] [iφ:val] [udef: val] copil+ul

(149) a. bătrânul profesor old.DEF.ART. professor ‘the old professor’ b. profesorul bătrân professor.DEF.ART. old ‘the old professor’ With respect to the animacy scale, we notice that the critical cut off point is actually [± Human]. Thus, [+Human] denoting nominals may be marked DOs (KPs), as well as unmarked objects. When they are unmarked, these objects pattern with [-Human] nominals, and require no specification for a syntactic [person] feature. [+Human] is simply a lexical property of the NP. In line with Cornilescu and Tigău (ms.), we consider that when they are syntactically marked as denoting people, [+Human(like)] NPs incorporate an

122

3 Linking syntactic structure to interpretation

unvalued syntactic [iPerson] feature, which is copied in D and valued by merging K. K is endowed with a valued [uPerson] feature. The possibility of incorporating [Person] is open only to NPs that are provided with the lexical feature [+Human] or rather [+Human (like)], given that other types of nouns (e.g. pets etc.) may be treated like humans and may become marked objects (Mardale 2008, Tigău 2010). (150b) shows the structure for unmarked DOs, while (151b) presents the structure of marked DOs: unmarked DOs evince a standard DP structure, whereas their marked counterparts have an additional functional layer, the KP as well as an additional [iPerson] feature in N. The presence of this unvalued feature in N and then in D triggers the insertion of K. Note that their more complex structure associated with their scrambling syntax, makes marked DOs more “prominent.” (150) a. Văd un profesor. See.1.SG. INDEF.ART. professor ‘I see a profesor.’ b.

D ei D NP +d +N uφ :__ iφ:val idef:val udef:val +HUM un profesor

(151) a. Văd pe un profesor. See.1.SG. DOM INDEF.ART. professor ‘I see a profesor.’ b.

KP K

DP

+p uφ:__ u-person:val

D +d uφ: __ idef: __ iperson: ___

pe

un

NP +N iφ:val u-def i person: ___ +HUM profesor

3.4 On the featural make-up of KPs

123

As pointed out in Cornilescu and Tigău (ms.), the feature specification of marked DOs may explain the use of DOM with demonstratives, wh-forms, bare quantifiers a.o. Consider first the case of demonstratives. Cornilescu (2000) draws attention to an interesting difference of behaviour with respect to DOM between demonstratives when used as determiners as opposed to their use as pronouns. Thus, when demonstratives function as determiners heading a nominal expression, the use of DOM depends on the semantics of the noun, which gets marked only if it denotes a human referent. In (152a) the [+human] DP syntactically interprets the lexical [+human] feature by way of incorporating an [iPerson:___] feature and DOM is allowed, while in (152b) DOM is rejected as ungrammatical since the noun is semantically [-human]: (152) a. Ai ajuta (pe) acest om? Would.2.SG. help (DOM) this man ‘Would you help this man?’ b. Ai vrea *pe aceste cărți? Would.2.SG. want *DOM these.FEM. books ‘Would you like these books?’ When demonstratives are used pronominally, they necessarily require DOM irrespective of whether they denote a human referent or an inanimate one: (153) Nu aș vrea cartea asta, ci aș Not would.1.SG. want book.DEF.ART. this but would.1.SG. vrea(-o) pe aceea, dacă se poate. want (-CL.3.ACC.SG.) DOM that.FEM if REFL. possible ‘I would not like this book, but I would like that one, if possible.’ Unlike the demonstrative in its determiner use, which has a lexical NP restriction, the demonstrative in its pronominal use has a pro-NP restriction, where pro is a functional category which must be specified for definiteness agreement with the definiteness parameter. In this case, pro will be specified as [+definite].59 The presence of this definiteness syntactic feature on the pro-NP restriction of the

59 The neuter demonstrative asta (this) patterns differently in that it disallows DOM in its pronominal use: (1)

Ia *pe asta de aici. Take DOM this from here ‘Take this away.’

124

3 Linking syntactic structure to interpretation

demonstrative triggers the merger of an unvalued [iPerson___] feature. This, in turn, triggers the projection of K (contributing an [uPerson: val] feature) which will value the [iPerson___] of the pro-NP. Example (64) shows this at play. Note that pro also bears a complete ’ feature set:60 (154)

KP wp K +p uφ:__ u-person:val

pe

DP wp D +d uφ:__ i+def:__ i person:__ idem:val aceea

NP +N iφ:val u+def i person:___ pro

The case of strong quantifiers signalled above as problematic may also find an explanation along these lines. As seen, these expressions may optionally carry DOM if they function as determiners heading nominals which denote human referents (155 a,c), (156a,c). On the other hand, they require obligatory DOM when used pronominally (155b), (156b): (155) Context: Nu trebuie să dai acum un răspuns. ‘You don‘t have to give me a definite answer now.’ a. Citește fiecare document în parte și vorbim pe urmă. Read.2.SG every document in part and speak.1.PL afterwards ‘Read each document and we‘ll talk afterwards.’

Interestingly, asta disallows DOM even when used in a context where clitic resumption is allowed: (2)

*Pe asta am mai spus – (o). DOM this have.1.SG. more said – (CL.3.ACC.SG.) ‘I have already said this once.’

The analysis in this case starts from the observation that asta is an indefinite demonstrative and as such the indefinite pro-NP restricting it would not force the merger of an [iPerson] feature, which would account for the absence of DOM in these cases. 60 See Cornilescu and Nicolae (2012), Giurgea (2013) for a detailed discussion of demonstrative pronouns and determiners. Note that both papers agree that demonstrative pronouns build on the post nominal use of demonstratives.

3.4 On the featural make-up of KPs

125

b. Citește pe/*ø fiecare în parte și vorbim pe urmă. Read.2.SG DOM/*ø every in part and speak.1.PL afterwards ‘Read each (of them) and we’ll talk afterwards.’ c. Ion salută (pe) fiecare coleg pe care-l John greets (DOM) every colleague DOM whom-CL.3.ACC.SG. întâlnește în drum spre școală. meets in road towards school ‘John greets every colleague that he meets on his way to school.’ (156) a. Poți lua oricare carte din bibliotecă, dar trebuie să May.2.SG. take any book from library but must SUBJ. o aduci înapoi în maxim trei săptămâni. CL.3.ACC.SG. bring.2.SG. back in maximum three weeks ‘You may take any book from the library but you will have to bring it back in at most three weeks.’ b. Poți lua pe/*ø oricare carte din bibliotecă, dar May.2.SG. take DOM/*ø any book from library but trebuie să o aduci înapoi în maxim must SUBJ. CL.3.ACC.SG. bring.2.SG. back in maximum trei săptămâni. three weeks ‘You may take any book from the library but you will have to bring it back in at most three weeks.’ c. Poți invita (pe) oricare prieten vrei, doar să fie May.2.SG. invite (DOM) any friend want.2.SG only SUBJ. be de gașcă. cool ‘You may invite any friend you want, provided he is cool.’ Just like in the case of the demonstratives used pronominally, we might consider that the pro-NP restriction is specified as [+definite] in this case too, which triggers the merger of an unvalued [iPerson___] feature. This, in turn, will trigger the projection of K (contributing an [uPerson: val] feature) which will value the [iPerson___] of the pro-NP as shown in (157):

126

3 Linking syntactic structure to interpretation

(157)

KP wp K +p uφ:__ u-person:val pe

DP wp D +d uφ:__ i person:__

NP +N iφ:val i person:___

fiecare

pro

Cornilescu and Tigău (ms.) also mention the case of the definite determiner cel (the) and the possessive article al. The former heads definite descriptions as may be seen in (158) e.g., cardinals, and other quantifiers, while the latter heads possessives (159). Just like demonstratives and strong quantifiers, these determiners differ with respect to DOM function of whether they have a determiner or a pronominal use: in (158a) DOM is optional as the noun denotes a human referent, while in (158b) it is discarded, given that the DO points to an inanimate referent. (158c) shows the strong preference of the cel phrase for DOM when the DP has no lexical head. Thus, when the DP selected by cel has a lexical head, DOM is determined by the semantic properties of this head: in the absence of a lexical head, cel phrases strongly require DOM. (159) shows the DOM requirements for al phrases: since these DPs always lack a lexical head (and consequently the semantic features of the nominal head would not play a part), DOM is required: (158) a. Am văzut (pe) cele patru fete. Have.1.SG seen (DOM) DEF.ART. FEM.PL. four girls ‘I have seen the four girls.’ b. Am văzut cea de-a patra casă. Have.1.SG seen DEF.ART.FEM.SG. fourth house ‘I have seen the fourth house.’ c. Am construit(-o) pe/?ø cea de-a Have.1.SG built (-CL.3.ACC.FEM.SG) DOM/?ø DEF.ART.FEM.SG. patra. fourth ‘I built the fourth one.’ (159) Context: A cui mașină vrei să o cumpărăm? ‘Whose car do you want to buy?’ pe/*ø a Ioanei. DOM/*ø POSS.ART.FEM.SG. Ioana.GEN ‘Ioana’s’

3.4 On the featural make-up of KPs

127

These DPs may be quite well analysed on the pattern of demonstratives, with a pro restriction contributing definiteness, which translates syntactically into the insertion of an unvalued [iPerson] feature forcing the merger of DOM. Consider now the case of bare quantifiers (BQ), which also display an interesting behaviour with respect to DOM: while BQs denoting human referents are necessarily marked (160a), their counterparts pointing inanimate referents reject DOM (160b): (160) a. Am auzit pe/*ø cineva. Have.1.SG heard DOM/*ø somebody ‘I have heard someone.’ b. Am auzit *pe ceva. Have.1.SG heard *DOM something ‘I have heard something.’ Building on the morphology that these BQs evince in English, Cornilescu and Tigău (ms.) argue that they incorporate their restriction. This, in turn explains why these expressions may not exhibit a(nother) lexical restriction: their restriction amounts to an indefinite pro bearing an [udef:val] feature, which is compatible with the merger of a [iPerson: ____] feature. The incorporation of the [iPerson: ____] feature is, however, optional, with DOM obligatorily showing up only when this feature is present. (161) presents the analysis proposed for the DOMed variant: (161)

KP wp K +p uφ:__ u-person:val

pe

DP wp D NP ri tN N D +N +d iφ:val uφ:__ udef:val i def: __ i person:__ i person:___ pro

cineva

Finally, consider the case of indefinite pronouns and determiners (i.e. the weak determiners: the indefinite article un/o (a), the determiner vreun (any-some)

128

3 Linking syntactic structure to interpretation

functioning as a negative polarity item, the negative determiner nici un/nici o (none), the lexical quantifiers mulți (many), puțini (few) and the cardinals), which exhibit a different behaviour with respect to DOM in that marking depends on the semantic properties of the head noun, irrespective of whether this is overt or not: (162a) shows that DOM is optional with nominals denoting a human referent, while in (162b) DOM is discarded given that the DP points to an inanimate referent. The same is at stake in (163), where the lexical noun has been left out: (162) a. Am văzut (pe) câțiva profesori prin facultate Have.1.SG seen (DOM) some professors through faculty chiar dacă era sâmbătă. even if was Saturday ‘I saw some professors in the faculty, even though it was Saturday.’ b. Am văzut câteva mașini, dar nici una nu mi-a Have.1.SG seen some cars but none not REFL.-has plăcut așa de mult încât să o cumpăr. pleased so much that SUBJ. CL.3.ACC.FEM.SG buy.1.SG ‘I saw some cars but none of them pleased me enough as to buy it.’ (163) a. Am salutat (pe) câțiva, dar nu mi-au Have.1.SG greeted (DOM) some but not me.DAT-have.3.PL răspuns la salut. answered at greeting ‘I greeted some, but they did not answer back.’ b. Am cumpărat *pe câteva, dar îmi mai trebuie. Have.1.SG. bought *DOM some but me.DAT more need ‘I bought some but I need some more.’ The optional marking in (162a) and (163a) might be accounted for by considering that the [iPerson] feature incorporation with indefinites is not mandatory. DOM is triggered only when incorporation takes place. In (163a), where the indefinite shows up in its pronominal use, the restriction is an indefinite pro bearing [udef], which is compatible with the [iPerson] feature and which incorporates into D just like in the case of BQs (164):

3.5 Conclusion

(164)

129

KP wp K +p uφ:__ u-person:val

pe

DP wp NP D tN ri N +N iφ:val udef:val i person:__

D +d uφ :__ i def: __ i person:___

pro

câţiva

Thus, Romanian DOs seem to require the merger of K when they bear an unvalued [iPerson] feature in N and then in D. The feature itself is the syntactic reflex of the animacy and definiteness hierarchies to which DO marking is sensitive. The presence of an unvalued [iPerson] feature triggers the merger of K, itself being the spell out of a [uPerson: val] feature. In line with López (2012), the use of pe is determined by the local context and arises as a consequence of the interplay between the syntactic features of [Person] and [Definiteness].

3.5 Conclusion This chapter built on the analysis of Spanish DOs put forth in López (2012) and parametrized this account in such a way as to capture the differences observed with Romanian DOs. The main claim in this respect was that the parameter which sets Romanian and Spanish apart lies in the types of nominals that must scramble: Spanish distinguishes between KPs and DPs, only allowing the former (i.e., marked DOs) to scramble, while the latter incorporate. In Romanian, on the other hand, both KPs and DPs may scramble. The relevant division in this case is between NP/NumPs which incorporate and DPs/KPs which scramble. We furthermore put forth a proposal regarding the featural make-up of marked DOs in Romanian. Thus, the projection of K is argued to be determined by the presence of an unvalued person feature in N and then in D. The [Person] feature is the syntactic reflex of the animacy hierarchy. The DOM marker PE is the spell-out of an uninterpretable Person feature. The distribution of PE is determined by the local context, following from the interplay of the syntactic

130

3 Linking syntactic structure to interpretation

features of [Person] and [Definiteness]. The additional layer of structure of DOM-ed objects makes them “more prominent”, a property from which a number of pragmatic effects can be derived. The analysis of DOMed DOs will be further refined in the subsequent sections when the featural make-up of CDed+DOMed DOs will be touched upon.

4 Some considerations on the role of the clitic as a movement trigger 4.1 The pronominal clitic as a movement trigger As shown in the previous section, the experimental data seem to support the hypothesis that the pronominal clitic doubling direct objects does not have the initially hypothesized semantic role of inducing a specific reading on the marked nominals: CDed+DOMed indefinite DOs were found quite acceptable within the four tested contexts forcing a non-specific interpretation. Furthermore, examples containing CDed+DOMed indefinite DOs collected from the internet corroborate with this conclusion: these DPs allow for a non-specific reading. The specific reading remains, however, an (favoured) option with both DOMed and CDed+DOMed indefinite DOs and an explanation as to the availability of the specific interpretation was advanced in line with López (2012): marked DOs have an internal structure which forces movement to a position that is, in its turn, associated to a peculiar mode of semantic composition of the DO with the predicate, which enables the specific reading on the respective DO. Thus, marked DOs are KPs, with K acting as a function shifting the denotation of the nominal from to an object level one, e. As such, the DO may no longer stay in its merge position within the VP as this is a position where only those nominals with a smaller internal structure may be hosted. These defective nominals incorporate into the predicate and there is nothing to hinder incorporation. With marked DOs incorporation is, on the contrary, blocked by the KP layer. As a consequence, these nominals need to move to a position where their composition with the predicate would be allowed and where their case may get valued. DOMed DOs move to a position lying between v and V, i.e., SpecαP, in line with a marked DOs in Spanish (López 2012). CDed+DOMed DOs move even further, possibly reaching a position wherefrom they may c-command the subject DP. Since both DOMed and CDed+DOMed DOs pass trough SpecαP, the availability of a specific interpretation with these nominals is explained. This chapter is devoted to the syntax of CDed+DOMed DOs and aims at coupling the proposed account with some new experimental data confirming movement of these DPs to a high position in the T area. A proposal with respect to the appropriate cliticisation theory is advanced, in line with Preminger (2017). The chapter has the following structure: section one discusses some evidence put forth in the literature on Clitic Doubling supporting the hypothesis that clitic doubled DOs move out of their merge position inside the VP. https://doi.org/10.1515/9781501513657-004

132

4 Some considerations on the role of the clitic as a movement trigger

Section 2 presents some experimental evidence in support of the movement hypothesis with these DPs. In section 3 we put forth an account of clitic doubled DOs closely following in the steps of Preminger (2017). Section 4 contains the conclusions.

4.1.1 Background. Evidence for movement 4.1.1.1 Binding Romanian CDed DOs have been argued to occupy a different syntactic position than their unmarked counterparts, as they are supposedly forced to leave their position inside the VP and reach the T area (Dobrovie-Sorin 1994, Cornilescu 2002a, Tigău 2011). Binding dependencies with the subject DP have been used to show that this is, indeed, the case, with CDed DOs being able to bind into the Subject DP unlike their unmarked counterparts. In (155a) the DO un student (a student) may not bind the subject părinții lui (his parents), while its clitic doubled variant is argued to felicitously do that in (155b). This difference in behaviour shows that at some point during the derivation the clitic doubled direct object finds itself in a c-commanding position with respect to the subject. This is possible only if the direct object DP moves out of its merge position within the VP and reaches a landing site above the merge position of the subject DP: ajutat un studenti să (165) a. *Părinții luii au Parents.DEF.ART. his have helped INDEF.ART. student SUBJ. își ia mașină. REFL buy car ‘Hisi parents helped a studenti buy a himself a car.’ b. Părinții luii l-au ajutat pe Parents.DEF.ART. his CL.3.ACC.M.SG.-have helped DOM își ia mașină.61 un studenti să INDEF.ART. student SUBJ. REFL.DAT buy car ‘Hisi parents helped a studenti buy a himself a car.’

61 Notice, however, that the phenomenon of inverse binding may also find an explanation if the clitic (and not necessarily the associate DP) moves out of the VP to a position wherefrom it c-commands the subject in its merge position.

4.1 The pronominal clitic as a movement trigger

133

Two further tests supporting a movement hypothesis for CDed DOs are proposed by Cornilescu (2002a) and Cornilescu & Cosma (2014) regarding Parasitic Gaps and supine clauses respectively. 4.1.1.2 Parasitic gaps Cornilescu (2002b) starts out from the observation that Heavy NP Shift (HNPS) of undoubled DOs (irrespective of whether they are unmarked or DOMed) may license parasitic gaps (PG).62 In (166a, b) below the unmarked and the DOMed DO are shown to allow PGs, as opposed to the CDed+DOMed DOs in (167a) and (168a): (166) a. Am examinat t fără a întrerupe t fiecare concurent Have.1.SG. examined t without to interrupt t every competitor separat. separately ‘I examined without interrupting each candidate separately.’ b. Am examinat t fără a întrerupe t pe fiecare Have.1.SG. examined t without to interrupt t DOM every concurent separat. competitor separately ‘I examined without interrupting each candidate separately.’ (167) a. *L-am întîlnit fără CL.3.ACC.M.SG.-have.1.SG met without Ion. John ‘I met John without greeting him though.’ b. L-am întîlnit fără CL.3.ACC.M.SG.-have.1.SG met without însă pe Ion. though DOM John ‘I met John without greeting him though.’

a saluta însă pe to greet though DOM

a-l saluta to- CL.3.ACC.M.SG greet

62 Parasitic gaps are empty categories inside an island for extraction (an adjunct), which is rendered acceptable by another gap outside this island. The latter gap is known as the licensing gap. Both gaps are bound by the same constituent labelled as the antecedent. In (166a), the binder is fiecare concurent separat (each competitor separately) which has undergone HNPS, an A’ movement by means of which the licensing gap is created.

134

4 Some considerations on the role of the clitic as a movement trigger

(168) a. *L- am examinat t fără a intrerupe t CL.3.ACC.M.SG.-have.1.SG examined t without to interrupt t pe fiecare concurent separat. DOM every candidate separately ‘I examined every candidate separately without interrupting him.’ b. *L- am examinat t fără a-l CL.3.ACC.M.SG.-have.1.SG examined t without to- CL.3.ACC.M.SG. intrerupe t pe fiecare concurent separat. interrupt t DOM every candidate separately ‘I examined every candidate separately without interrupting him.’ (Cornilescu 2002b: 6, p.2) In order to understand why PGs are relevant for the movement of the clitic out of VP remember that HNPS is an A’ movement which targets a position inside vP: a DP inside v’ is A’ moved leaving behind a gap which licenses the parasitic gap. Adopting a Separate Antecedent63 proposal in line with Chomsky (1986) and Nissenbaum (2000), according to which the gap in the adjunct is bound by a null operator, Cornilescu (2006) argues that this operator should be a copy of the antecedent in the sense that it is structurally identical with it. Consequently, only DPs that are not clitic licensed may function as antecedents in a PG construction, given that these nominals stay inside the vP. When the antecedent is clitic licensed, the operator in the PG should also be licensed by a clitic. But the clitic moves out of the vP to T(ense) presumably hence the adjunct clause should be itself a TP, adjoined to the main TP, and should also contain a clitic (as the operator should be a copy of the antecedent). Indeed, HNPS is possible with CD-ed object DPs but these DPs may not license PGs: (169) I-am auzit fără a-i vedea însă CL.3.ACC.PL-have.1.SG heard without to- CL.3.ACC.PL see though pe hoți. DOM burglars ‘I heard the burglars without seeing them.’

63 Two types of theories have been proposed for the analysis of parasitic gaps. The Shared Antecedent hypothesis adopted in Chomsky (1982), Nunes (1995) posits that the PG is an empty category whose interpretation is given by the antecedent (the A’ moved constituent) (1a), while the Separate Antecedent supporters (Chomsky 1986) argue that the empty category within the adjunct is actually bound by a null operator (1b): (1)

a. What did Ann buy t [without paying for PG]? b. What did Ann buy t [Op without paying for t]?

4.1 The pronominal clitic as a movement trigger

135

(169) is felicitous on account of the fact that the adjunct clause (a TP itself) is adjoined to the main TP, as opposed to the examples above where the complex predicate forms at vP level. Note that in (169) the pronominal clitic surfaces both in the main as well as in the adjunct clause. However, this example is an instance of a TP level predicate, in direct contrast with examples such as (165) above, where the complex predicate is formed at the vP level. Thus, the fact that CD does not allow PGs constitutes an argument in support of the hypothesis that the clitic scrambles out of its merge position within the vP. 4.1.1.3 Supine clauses Cornilescu & Cosma (2014) provide further evidence in support of the hypothesis that the pronominal clitic moves out of the vP. They observe that in Romanian prepositional supine clauses, the verb may select a DOMed DO, but never a CDed+DOMed one, as shown in (170): (170) Nu pot să vin la tine azi am de Not can.1.SG. SUBJ. vin.1.SG. at you today have.1.SG. of vizitat pe cineva. Visited.SUPINE DOM somebody ‘I cannot pass by today, I have to visit someone.’ Note that the indefinite bare quantifier cineva (someone) ranging over humans is necessarily DOMed and never CDed. As such, it constitutes a perfect candidate for the verbal supine. On the other hand, the prepositional supine rejects personal pronouns as they are obligatorily clitic doubled. The reason why this is so is because the supine clause does not possess sufficient structure to accommodate a clitic: (171a) shows that DOs expressed by means of personal pronouns are obligatorily CDed – failure to double such a DP leads to ungrammaticality. The examples in (171b) show how the impossiblity of accommodating the clitic within the supine clause structure leads to ungrammaticality when a full pronominal DP is used. Other DPs which do not require CD are fine in these structures: consider the definite description which allows for optional CD (171c): (171) a. *(I-)am ajutat pe ei să reușească. CL.3.ACC.PL.-have.1.SG. helped DOM them SUBJ. succeed ‘I have helped them succeed.’

136

4 Some considerations on the role of the clitic as a movement trigger

b. *E ușor de ajutat pe ei Is easy of helped.SUPINE DOM them ‘They are easy to help.’ c. E ușor de ajutat elevii silitori. Is easy of helped.SUPINE pupils.DEF.ART. diligent.PL. ‘Hardworking pupils are easy to help.’ Thus, the prepositional supine in (b) seems to reject internal arguments that are necessarily CDed and to only accept DPs which do not require doubling or that disallow it. Cornilescu & Cosma (2014) account for the observed facts by arguing that the prepositional supine is a reduced clause lacking Agreement. If the clitic were to remain inside the vP, it would then have no problem in being licensed inside a prepositional supine. Examples from the internet strengthen this hypothesis: in (172a) there is a definite description within the supine: as known, these DPs optionally allow CD but do not necessarily require it. The bare quantifier oricine (anyone) in (172b) disallows CD and as such represents a perfect candidate for supines. Finally, the strong quantifier in (161c) allows optional CD and fares well within the supine: (172) a. Aici este vorba de voluntariat și nu de Here is about of volunteering and not of ajutat elevii să ia note bune. helped.SUPINE pupils.DEF.ART. SUBJ. take marks good.FEM.PL. ‘This is about volunteering work and not about helping pupils to take good grades’64 b. Ei erau buni de făcut cinste și de They were good.PL of made.SUPINE honour and of ajutat pe oricine. helped.SUPINE DOM anyone ‘They were only good at paying for drinks and helping anybody.’ (Ruse 2014: 250)

64 (https://ro.wikipedia.org/wiki/Discuție_Wikipedia:Ghid_de_coordonare_pentru_proiectele_ externe), Access date: 10.06.2018

4.1 The pronominal clitic as a movement trigger

137

c. Ar fi cazul de ajutat pe toți din Would be case.DEF.ART. of helped.SUPINE DOM all of această categorie. this category ‘It would be the case to help everybody in this category.’65 4.1.1.4 Focus projection Gierling (1997) contributes a further argument in support of a movement hypothesis for CDed DOs. As it seems, these DPs do not phonologically behave as arguments of the verb with respect to Focus Projection (Selkirk 1995, p. 555)66 but they rather evince a similar behaviour to adjuncts in the sense that they do not allow the projection of focus onto the verb: (173b) may function as an answer to both questions listed under (173a), whereas (174b) may only be used as an answer to a question inquiring about the argument i.e., Who are you looking for? but never as an answer to a broad focus question What are you doing?. (173) a. Ce cauți? Ce faci? ‘What are you looking for? What are you doing?’ b. Caut o CARTE. Search.1.SG. INDEF.ART. BOOK ‘I am looking for a book.’ (174) a. Pe cine cauți?/*Ce faci? ‘Who are you looking for?/* What are you doing?’ b. Îl caut pe ION. CL.3.ACC.M.SG. search.1.SG. DOM JOHN ‘I am looking for JOHN.’ In order for (b) to function as an answer to a broad focus question, it must contain an additional stress on the verb. Gierling (1997) accounts for this state of

65 (http://curentul.md/social/stimularea-simulata-a-familiilor-cu-gemeni.html), Access date: 10.06.2018 66 Selkirk (1995:555) Focus Projection: a. Focus marking of the head of a phrase licenses the F-marking of the phrase b. F-marking of an internal argument of a head licenses the F-marking of the head This licensing mechanism correctly ensures that F-projection is not possible from an accented adjunct. For a VP to be focused in this case, an accent on the verb is needed as well.

138

4 Some considerations on the role of the clitic as a movement trigger

affairs by arguing that focus cannot project from a DP that has been doubled by means of a clitic to a higher constituent (like the VP). (175) a. *Pe cine cauți?/Ce faci? ‘Who are you looking for?/* What are you doing?’ CAUT pe ION.] b. [F Îl CL.3.ACC.M.SG. SEARCH.1.SG. DOM JOHN ‘I am LOOKING FOR JOHN.’ But why should the focus feature instantiated by the accent on an argument not be able to license the F-marking of the verb and thus of the VP? That the argument in (164b) behaves like an adjunct is quite unexpected. Apparently, with CDed DPs focus cannot project from the DP. The answer lies in the fact that the doubled DP has left the VP. Thus, a number of arguments have been proposed, supporting a movement out of the VP for CDed DOs. The experimental data we gathered with respect to binding dependencies involving the DO and the Subject DP also point into this direction. The next section is devoted to advancing a detailed presentation of the experimental findings with Subject-DO binding dependencies meant to strengthen the hypothesis that CDed DOs leave their merge position reaching a landing site which allows them to c-command and hence to bind into the subject DP.

4.1.2 An experiment on Subject -DO binding dependencies 4.1.2.1 Design Our aim in this experiment was to check binding dependencies holding between the subject and the direct object. Since we tested the behaviour of three types of DOs i.e., unmarked, DOMed and CDed+DOMed DOs, we ended up having three experiments, each focusing on one DO type. In addition, each of the three experiments was comprised of two parts, a norming part checking whether the respondents understood the intended bound interpretation and another part probing for the acceptability of the tested items. The experiments were similar in design: we tested items wherein we varied: a) the surface word order between the subject and DO and b) the direction of binding. Four patterns in need of assessment thus arose as shown in Table (10) below:

139

4.1 The pronominal clitic as a movement trigger

Table 10: Parameters – word order and binding. Word order

Su before DO

Binding direction

Su binds DO

DO before Su DO binds Su

Su binds DO

DO binds Su

For each experiment we designed 36 sentences: 24 checked quantificational dependencies with the Romanian correspondent of each (fiecare . . . în parte) and 12 employed dependencies with nici un/o NP (no NP). The 36 sentences were then varied function of each of the four patterns such that 144 items obtained for each experiment. These items were then distributed into four different lists using the Latin square method for an even distribution. The lists were adapted to match the aim of each of the two component experiments: for the norming experiment probing whether the respondents understand the binding dependencies at stake, we designed for each tested item three different answer possibilities: a) one enclosing the binding dependency reading holding between Su and DO, b) another one stating lack of binding between Su and DO, and c) a third answer variant where both a) and b) would be ticked as possible or impossible (see example (178) below). For the acceptability experiment, we provided a seven-rung scale (1 being the lowest acceptability score and 7 being the highest) against which the tested items had to be assessed. To each list we added 32 fillers grouped into 8 expectedly unacceptable items (for the acceptability experiment)/with no possible answer (for the norming experiment), 8 completely acceptable items/two possible answers, and 16 expectedly average items/with one variant response (the fillers were previously tested for acceptability in a smaller, informal experiment). Each list thus ended up containing 68 items and was further assessed by at least 20 native speakers of Romanian such that more than 240 respondents took part in the norming experiments and just as many in the acceptability ones. The answers were then verified and the outliers removed i.e., questionnaires where more than 7 fillers had been answered incorrectly. The remaining questionnaires entered statistical analysis. (176) and (177) below contain examples with two sample items from the experiment checking binding with Su DPs and CDed+DOMed DOs. The items containing DOMed and unmarked DOs follow the same pattern: Su before DO; Su binds DO (176) a. Fiecare profesor în parte îl ajută pe Every professor in part CL.3.ACC.M.SG. help.3.SG DOM

140

4 Some considerations on the role of the clitic as a movement trigger

doctorandul său cu sfaturi și bibliografie. PhD student.DEF.ART. his with advice.PL and bibliograpy ‘Every professor helps his PhD student with advice and bibliography.’ Su before DO; DO binds Su b. Profesorul său îl ajută pe fiecare Professor.DEF.ART. his CL.3.ACC.M.SG. help.3.SG DOM every doctorand în parte cu sfaturi și bibliografie. PhD student in part with advice.PL and bibliograpy ‘His professor helps every PhD student with advice and bibliography.’

DO before Su; Su binds DO Pe doctorandul său îl ajută fiecare DOM PhD student.DEF.ART. his CL.3.ACC.M.SG. help.3.SG every profesor în parte cu sfaturi și bibliografie. professor in part with advice.PL and bibliograpy ‘Every professor helps his PhD student with advice and bibliography.’

DO before Su; DO binds Su d. Pe fiecare doctorand în parte îl ajută DOM every PhD student in part CL.3.ACC.M.SG. help.3.SG profesorul său cu sfaturi și bibliografie. professor.DEF.ART. his with advice.PL and bibliograpy ‘His professor helps every PhD student with advice and bibliography.’

Su before DO; Su binds DO (177) a. Nici un director nu îi va concedia No manager not CL.3.ACC.PL. will.3.SG. fire pe angajații săi pentru abateri minore. DOM employees.DEF.ART. his for offences minor.FEM.PL. ‘No boss will fire his employees for minor offences.’

4.1 The pronominal clitic as a movement trigger

141

Su before DO; DO binds Su b. Directorul lui nu îl va concedia pe Manager.DEF.ART. his not CL.3.ACC.M.SG. will.3.SG. fire DOM nici un angajat pentru abateri minore. no employee for offences minor.FEM.PL. ‘His manager will not fire any employee for minor offences.’ DO before Su; Su binds DO c. Pe angajații săi nu îi va concedia DOM employees.DEF.ART. his not CL.3.ACC.PL. will.3.SG. fire nici un director pentru abateri minore. no manager for offences minor.FEM.PL. Lit. His employees will not fire any business owner for minor offences. ‘No business owner will fire his employees for minor offences.’

DO before Su; DO binds Su d. Pe nici un angajat nu îl va concedia DOM no emplyoyee not CL.3.ACC.M.SG. will.3.SG. fire directorul lui pentru abateri minore manager.DEF.ART. his for offences minor.FEM.PL. Lit. No employee will fire his boss for minor offences. ‘His boss will not fire any employee for minor offences.’ In (178), one may see the format of the answers for the norming experiment: (178) Fiecare profesor în parte îl ajută pe Every professor in part CL.3.ACC.M.SG. help.3.SG DOM doctorandul său cu sfaturi și bibliografie. PhD student.DEF.ART. his with advice.PL and bibliograpy ‘Every professor helps his PhD student with advice and bibliography.’ Sample answer from the norming experiment: i.

From the fragment above I understand: Every PhD student is supported by his own supervisor with advice and bibliography.

yes

no

142

4 Some considerations on the role of the clitic as a movement trigger

ii. Every PhD student is supported with advice and bibliography yes by the supervisor of a certain person, who has not been mentioned in the text but to whom one seems to refer by means of the pronoun his. iii. Both (i) and (ii) yes

no

no

4.1.2.2 Expected results. Experimental hypotheses As seen above, the literature on Romanian DOs has distinguished between the behaviour of CDed+DOM DOs on the one hand and that of DOMed DOs and unmarked DOs on the other hand with respect to their binding potential. More specifically, it has been argued that, while the former DOs may bind into the Subject DP, the latter DOs may not do so (Cornilescu 2006, Dobrovie-Sorin and Cornilescu 2008). Based on these observations, Cornilescu (2006) advances the hypothesis that clitic doubled DOs must leave their merge position within the VP and move into a position wherefrom they c-command the subject in SpecvP and thereby bind it. The examples under (179) below show this at stake: in (a) the subject fiecare director (every manager) expectedly binds the possessive within the DO. The same happens with (179c) where a DO denoting an inanimate referent is at stake.67 The binding dependence has been argued to hold in the case of examples such as (b), where a clitic doubled DO may bind into the subject DP. Lack of CD as in (d) no longer allows for such a binding dependency. mai verifică (179) a. Fiecare directori în parte îi Every manager in part CL.3.ACC.M.PL. more checks din când în când pe subalternii săii. from time to time DOM employeess his ‘Every manager occasionally checks on his employees.’ mai verifică b. Directorul săui îl Manager.DEF.ART. his CL.3.ACC.M.PL. more checks din când în când pe fiecare subalterni în parte. from time to time DOM every employee in part ‘His manager occasionally checks every employee.’

67 As known DO pointing to inanimate referents may not be DOMed or CDed in Romanian.

4.1 The pronominal clitic as a movement trigger

143

c. Fiecare autori în parte recitește cărțile luii cu Every author in part re-reads books.DEF.ART. his with plăcere. pleasure ‘Every author in part rereads his books with pleasure.’ d. Autorul săui recitește fiecare cartei în parte cu Author.DEF.ART. its re-reads every book in part with plăcere. pleasure ‘Its author rereads every book in part with pleasure’ Cornilescu (2017) makes the same claims with respect to clitic doubled indirect objects: these too may seemingly bind into the subject DP, as opposed to their undoubled counterparts. Thus, while the dative DP in (180a) is argued to be able to bind the possessive within the Subject DP, as shown by co-indexation, its undoubled counterpart in (180b) may no longer do so. Likewise, the IO oricărui elev (any pupil) is claimed to be able to bind into the subject profesorii lui (his teachers) only in the clitic doubled variant (180c): ajută multorai. (180) a. Religia lori/j le Religion.DEF.ART. their CL.3.DAT.PL. helps many.DAT ‘Many people find help in their religion.’ b. Religia lor*i/j ajută multorai. Religion.DEF.ART. their helps many.DAT ‘Many people find help in their religion.’ plac oricărui elevi. c. Profesorii luii/j îi Professors.DEF.ART. his CL.3.DAT.SG. like any.DAT. pupil ‘Any pupil likes his teachers.’ d. Profesorii lui*i/j plac oricărui elevi. Professors.DEF.ART. his like any.DAT. pupil ‘Any pupil likes his teachers.’ Given these claims advanced in the literature on Romanian CD, our experimental endeavour started from the hypothesis that only CDed+DOMed DOs may bind into the subject DP irrespective of word order, given that these DOs always leave their merge position ending up in a landing site whereform they may ccommand the subject in SpecvP. DOMed and unmarked DOs on the other hand would be expected to bind into the subject only when having been dislocated. The following section presents the results we obtained.

144

4 Some considerations on the role of the clitic as a movement trigger

4.1.2.3 Experimental results and discussion The experiment checking binding dependencies between the Subject DP and a CDed+DOMed DO uncovered the fact that the surface word order of the two arguments has a say with respect to binding dependencies: respondents accepted the existence of binding interpretations more readily in the case of those instances where word order matched the binding direction. Nevertheless, reverse binding was considered possible by the vast majority of the participants. Importantly, 83% of the respondents assessed instances where a CDed+DOMed DO was interpreted as binding into a preceding subject DP as possible. These findings thus support our initial hypothesis that clitic doubled DOs reach a landing site wherefrom they may c-command the subject in its in-situ position (Graph 10).

CDed+DOMed DOs 100% 90%

96%

96% 89%

80%

83%

70% 60% 50% 40% 30% 20% 10% 0% Su binds into DO

DO binds into Su

Su before DO

Su binds into DO

DO binds into Su

DO before Su

Graph 10: Binding with CDed+DOMed DOs.

The experiment conducted on unmarked DOs revealed, again, that word order is an important factor when it comes to assessing binding dependencies between the subject and the DO, with 91% of the respondents accepting an interpretation according to which a preceding Subject DP would bind into the DO as possible. Only 66% of the participants accepted a reading where the DO would bind into a preceding subject (Graph 11).

4.1 The pronominal clitic as a movement trigger

145

Unmarked DOs 100% 90% 80%

91%

70% 60%

66% 60%

61%

Su binds into DO

DO binds into Su

50% 40% 30% 20% 10% 0% Su binds into DO

DO binds into Su

Su before DO

DO before Su

Graph 11: Binding with unmarked DOs.

Similarly, the respondents participating in the experiment verifying binding between the subject and a DOMed DO found instances with inverse binding, where a DO bound into a preceding subject as much less acceptable (71%) than their counterparts where the preceding subject bound into the DO (Graph 12).

DOMed DOs 100% 90%

99%

96%

80%

87%

70% 71%

60% 50% 40% 30% 20% 10% 0% Su binds into DO

DO binds into Su

Su before DO Graph 12: Binding with DOMed DOs.

Su binds into DO

DO binds into Su

DO before Su

146

4 Some considerations on the role of the clitic as a movement trigger

In fact, when comparing the differences of percetage holding between those participants who recognized a binding interpretation for the Su before DO, Su binds into DO instances and those who accepted the binding reading for the Su before DO, DO binds into Su ones, we noticed that they are roughly the same (25%- 28%): 91% vs 66% for unmarked DOs and 99% vs. 71%. Not the same may be maintained for the case of CDed+DOMed DOs where the difference of percentage is significantly smaller: 96% vs 83%.

Binding all DOs 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

99%

96%

91%

83% 71%

66%

Su binds into DO DO binds into Su Su binds into DO DO binds into Su Su binds into DO DO binds into Su Su before unmarked DO

Su before DOMed DO

Su before CDed+DOMed DO

Graph 13: Comparing Binding results with all DO types.

The respondents thus seemed to pattern alike with respect to the binding potential of unmarked and DOMed DOs and to consider a binding dependency where a CDed+DOMed DO bound into a preceding Subject as much more likely. Of course, one may wonder why the precentages per se are so high in view of the expectation that unmarked and DOMed DOs should not be able to bind into a preceding subject DP. An answer might come from the tested items – when verifying them individually, we noticed that some of the items got very high scores on the relevant binding interpretation irrespective of word order and DO type. Most probably this had to do with the fact that the situations described in these cases were such that the non-binding interpretation would be less likely. The pragmatic context thus seems to be important in this respect. Other items, where a binding interpetation was not that obvious, fared significantly worse, with scores of 40% and less. Indeed, when taking some of the problematic items out, the differences between CDed+DOMed DO and DOMed DOs with respect to their binding abilities in the Su before DO, DO binds into Su case increased dramatically. Consider Graph 14 below to see this at stake (the items containing unmarked DOs were left out given that they represented different lexicalisations):

4.1 The pronominal clitic as a movement trigger

99%

147

99%

100% 90%

80%

80% 66%

70% 60% 50% 40% 30% 20% 10% 0% Su binds into DO

DO bind into Su

Su before CDed+DOMed DO

Su binds into DO

DO bind into Su

Su before DOMed DO

Graph 14: CDed+DOMed DOs vs. DOMed DOs.

Even so, one may still identify a difference of behaviour with the respondents when it comes to accepting a binding interpretation for the Su before DO, DO binds into Su: when the DO has been clitic doubled, they accept such an interpretation much more readily than in the case of unmarked DOs and DOMed DOs. They also seem to group the latter two DO types together, which makes us infer that these DPs exhibit a similar binding potetial with respect to the subject. Interpreting the results in the context of the discussion in the previous sections dwelling on the syntax of the DO, we will posit that while CDed+DOMed DOs must leave their merge position and end up in a landing site wherefrom they may c-command the Subject DP in its in-situ position, unmarked and DOMed DOs, on the other hand, do not reach a c-commanding position with respect to the subject DP, although they may have to leave their merge position (López 2012). A question that naturally follows would address the rationality of such a movement of the CDed DO, as opposed to that of DOMed and unmarked counterparts. This is an issue we address in a subsequent section, where a theory of cliticisation as an instance of long head movement of D° out of the doubled DP along the lines of Preminger (2017) is proposed.

4.1.3 A theory of cliticisation Among the analyses which have been proposed for CD configurations crosslinguistically, the Big DP account put forth by Uriagereka (1995) seems to have

148

4 Some considerations on the role of the clitic as a movement trigger

gained ground in what concerns Romanian (see Cornilescu 2006, Cornilescu and Dobrovie-Sorin 2008, Tigău 2011 a.o.). Such an approach seems to have a better explanatory coverage for the problematic aspects evinced by CDed configurations: the interdependence between the doubling clitic and DOM (Cornilescu 2002b, 2006), the status of the clitic (Cornilescu (2002a,b, 2006), Săvescu (2009), Tigău (2011) a.o.), the position occupied by the clitic and the associate (Cornilescu (2002b, 2006), Cornilescu & Dobrovie-Sorin (2008) a.o.). Building on diachronic and synchronic similarities with determiners, Uriagereka (1995) puts forth an analysis of 3rd person accusative clitics as determiners heading their own projection and taking a NP pro complement. In CD instances, the specifier of this DP is occupied by a full DP. (181)

DP ei (double) D’ ei D NP Clitic pro

Thus, clitics are heads associated with a pro complement. As pointed out by Uriagereka, the structure closely mirrors that proposed by Postal (1966) for sequences of type ‘el/la que vino’ (the who came) or ‘el/la de Francia’ (the from France) in (182) where the determiner ‘el/la’ licenses a pro complement NP which has been modified by a relative clause (a) or a PP (b): (182) a. el/la que vino the who came ‘the one who came’ b. el/la de Francia the from France ‘the one from France’ Just like in the structures above, the clitic is generated as the head of a determiner phrase taking a pro as its complement. Unlike these sequences, however, where the determiner remains in situ, the pronominal clitic will have to move. This accounts for its surface position. In order to reach its Spell-Out position, the clitic undergoes head movement to a functional domain, in line with Kayne (1989). Uriagereka (1995) posits that what triggers movement of the 3rd person accusative clitics is the necessity of identifying their pro complement: being referential, specific expressions, clitics must move to a landing site outside the VP

4.1 The pronominal clitic as a movement trigger

149

so as to identify this pro. In view of the Mapping Hypothesis put forth by Diesing (1992), clitics need to occupy a position outside the VP at LF since specific elements need to be out of the nuclear scope at LF. One disadvantage of this approach is the fact that it posits the existence of an ‘unwieldy beast’ i.e., a ‘big DP’configuration where the associate DP occupies the specifier of a DP headed by the accusative clitic. The doubling aspect of CD is thus arrived at ‘by brute force’ (Preminger 2017) given that the derivation starts from an already doubled structure. More recent studies on Romanian CD (Cornilescu 2017, Cornilescu et al. 2017b) favour Preminger (2017) as a more suitable account of CD given the fact that it builds on a regular DP. Also, the account seems to have a better empirical coverage by way of being able to capture the behaviour of several other mechanisms besides CD i.e., predicate clefting, noun incorporation and V/v-toT-type head movement. Preminger (2017) analyzes CD as an instance of long head movement of D° out of the doubled DP, which involves syntactic φ agreement. Movement of D° is non-local, as D° leaps over at least one c-commanding head thus on its path, violating the Head Movement Constraint (Travis 1984). In (183), the clitic merges in the complement position of V and subsequently moves at least as far pffi pffi as v, skipping over /V. In (183) D has reached T, leaping over /V, as well as v (and possibly Asp): pffi (183) { DCl, Auxiliary/Tense/Aspect/Finiteness }, { /V (tDcl)}} (184) L’as-tu fait? CL.3.ACC.M.SG.-have.2.SG.-you done ‘Have you done it?’

(Preminger 2017: 27, 26; p. 14)

(185) shows the general structure of CD and simple cliticisation cases. The landing site for the clitic may be v or another higher head. (185)

vP ei D0-v0

VP ei V DP ei D0

NP 4

150

4 Some considerations on the role of the clitic as a movement trigger

One peculiar characteristic of CD is that both copies of D° get pronounced, as opposed to simple cliticisation cases where only the D° copy in the landing site is pronounced. This option comes as a consequence of certain conditions posed on phonological chain reduction on Head movement. Following Bošković and Nunes (2007) and Nunes (2004), Preminger (2017) starts from the idea that phonological chain reduction takes place when the system has recognized that the different instances of a certain syntactic object are copies of one another. This recognition process may, on the other hand, be hindered, a situation which obtains, for instance when one of the two copies occurs inside a larger morpho-syntactic unit whose internal make-up is not accessible to the linearization algorithm. Since the two occurrences of the same element cannot be recognized as such, they will both get pronounced at PF. Preminger further refines the account by proposing the following conditions on phonological chain reduction: (186) Let X° be a head that undergoes movement to Y° and let α be the lower copy of X°/ α will be phonologically deleted iff either of the following conditions is met: (i) α and Y° are not separated by a phasal maximal projection (incl. XP). (ii) X and Y are part of the same extended projection, and Y° c-commands α in the surface structure (i.e., no constituent containing α but not Y° has undergone subsequent movement to a position above Y°) (Preminger 2017: 33, p. 18) On examining CD configurations, one may notice that neither of the two conditions in (186) gets satisfied: (i) – the two copies of D° occur in different phases given that the movement site of D° is a DP i.e., a phasal domain; (ii) – the two copies of D° are not in the same extended projection as the lower D° occurs within the extended projection of N, while the landing site of the higher D° is to be found within the extended projection of V. As a consequence, both copies of D° get pronounced. One other peculiarity of CD is that the two copies of D° may be phonologically different. This is actually expected given that the two instances of the same element occur in different morphological environments i.e, as the head of the extended nominal projection and as a clitic adjoined to a verbal head: the morphological environment may trigger allomorphy, as well as suppletion. As it seems, projections hosting Romance clitics induce very little allomorphy, while the hosts of clitics in Basque trigger allomorphy to a considerably greater degree, causing pronominal clitics to be significantly different from the initial corresponding determiner (Preminger 2017).

4.1 The pronominal clitic as a movement trigger

151

Finally, CD is argued to involve an agreement operation targeting the maximal DP phrase before cliticization. The cliticization host will thus enter an agreement relation with the full DP and, by doing so, it will satisfy the Principle of Minimal Compliance and enable movement of D° to the host, by itself (see Preminger 2017: 20–30 for a complete discussion). In what follows, we will extend the CD account proposed in Preminger (2017) to Romanian CD structures. The analysis closely mirrors Cornilescu (2017)’s account for datives.

4.1.4 Deriving clitic doubling structures As seen above, the internal make up of marked DOs presupposes the existence of a KP layer, where K is triggered by an unvalued syntactic [iPerson] feature present in the NP which is then copied in D. The NP itself is a [+Human] denoting nominal and as such may incorporate the [iPerson] feature. The presence of the syntactic unvalued [iPerson] feature triggers the merger of K, bearing a valued [uPerson], which verifies the unvalued feature on D. The entire KP ends up bearing a valued [iPerson]. (187) below shows this at work: (187) a. Ajut pe un coleg. Help.1.SG. DOM INDEF.ART. colleague ‘I help a colleague.’ b. KP ep K DP ep +p D uφ:__ +d u-person:val uφ: __ idef: __ iperson: ___ pe

un

NP +N iφ:val u-def i person: ___ +HUM coleg

The K also shifts the denotation of its DP complement to e and as a consequence the entire construct moves to SpecαP where the KP check its [uCase] feature by way of Agreement with v:

152

(188)

4 Some considerations on the role of the clitic as a movement trigger

vP ri v´ Su ri v[Acc] αP ajut rp KP α´ ! iφ:val ri α ! VP idef:val ! iperson:val ri V tKP ! uCase t V z___ pe un coleg iφ:val idef: val iperson:val uCase

apud (López 2012: 33, p. 49) Unmarked DOs, on the other hand, pattern with [-Human] nominals and need no specification for a syntactic [Person] feature. Hence the projection of K is not required: (189) a. Ajut un student Help.1.SG. INDEF.ART. student ‘I am helping a student.’ b. D ei D NP +d +N uφ:__ iφ:val idef:val udef:val +HUM un student

Unlike, their Spanish counterparts, however, unmarked DOs in Romanian may scramble when they have e denotation. In this case, they undergo movement to SpecαP, just like their marked counterparts and check their [uCase] against v. The only thing that differs in this particular case is that these DOs merge as DPs and have no syntactic [Person] feature in their specification:

4.1 The pronominal clitic as a movement trigger

(190)

153

vP ri v´ Su ri v[Acc] αP ajut rp DP α´ iφ:val ri idef:val α VP uCase ri un coleg V tDP iφ:val tV

idef: val uCase

Conversely, unmarked DOs stay in-situ and incorporate into V when they have an denotation. They check case when V moves to v: (191) a. A luat (o) gripă și a Has.3.SG. taken a flu which has zăcut trei săptămâni. kept his bed three weeks ‘He caught a flu and was bed ridden for three weeks.’ v´ ei v VP ei ti v[Acc] V V´ ei ei V Vo DP D[uC] ei o D´ ei Do NP tD gripă

b.

With CD+DOMed DOs, the situation would be more complex as the DOM marker pe is further bleached and no longer carries a valued syntactic [uPerson] feature, but is merely specified as [uPerson:___] i.e., its [Person] feature is in need of valuation, just like the one on the nominal it precedes. The KP ends up

154

4 Some considerations on the role of the clitic as a movement trigger

having the feature specification [iPerson: ___] (192) and will have to find a way whereby to value this feature during the derivation. (192) a. Îl ajut pe un coleg. CL.3.ACC.M.SG. help.1.SG. DOM INDEF.ART. colleague ‘I help a colleague.’ b.

KP ep K +p uφ:__ uperson: ___ uCase__

DP ep D +d uφ: __ idef: __ i person: ___ uCase __

pe

un

NP +N iφ:val u-def i person: ___ uCase __ +HUM coleg

Just like regular KPs, the DO in this case will also undergo short scrambling to SpecαP and check its [uCase] against the v head. Unlike its DOMed counterparts, this DO will not stop here, however, and undergo further movement in the vicinity of a PersP in order to check its [iPerson: ___] In line with Săvescu (2009), we posit a PersonP at the vP periphery: the DO will move into a specifier of the vP and have its [iPerson __] feature valued against the Pers head. Nevertheless, having reached its position, the DO acts as an intervener between the T head and the Subject DP in the sense that T cannot access the DP Subject for case assignment. As shown in Preminger (2017) for Basque (but see also Anagnastopoulou 2003 a.o.), clitic doubled DPs act as non-interveners for other φ agreement and movement operations. On the other hand, undoubled DPs do count as interveners. Cliticisation would then be a way to remove the DP object blocking agreement between T and the Subject DP. The DP object will thus obligatorily cliticize on T, enabling the DP subject to get its case checked (193). (193) Maria îl vede pe Ion. Mary CL.3.ACC.M.SG. sees.3.SG. DOM John Mary sees John.

4.2 Conclusion

(194)

155

TP 3 T PersP eo Pers´ ei vP Persº [iPers] ri KP vP [iPers: __] rp v´ DPSu [Case: Nom] ri αP vº ri KP α´ tKP ri α VP ri V tKP tV iφ:val idef: val iperson: __ uCase

Incidentally, this mechanism also explains why DOs do not cliticize on v. Note also, that, as already pointed out above, CD is neither optional nor conditioned by the properties of the direct object DP. Cliticisation occurs as a means to remove an intervening DP blocking agreement between T and the DP subject. The fact that the DO has moved outside its merge projection and reached PersonP comes as a consequence of its featural make-up i.e., its having an unvalued [uPerson] feature in need of valuation.

4.2 Conclusion This chapter mainly focused on CDed+DOMed DOs and put forth an account of these nominals in lines with the proposal advanced in Preminger (2017), where CD is viewed as an instance of long head movement of D° out of the doubled DP. The account rested on some new experimental findings concerning binding dependencies between the subject and the DO DP showing that CDed+DOMed DOs may bind into the Subject DP irrespective of word order, as opposed to their DOMed and unmarked counterparts, which seem much less inclined to do so. This strengthened the hypothesis already advanced in the literature that CDed+DOMed DOs leave their merge position inside the VP and reach a landing site wherefrom they may c-command the Subject DP found in its merge position.

156

4 Some considerations on the role of the clitic as a movement trigger

It was then argued that CDed+DOMed DOs start out as KPs, just like their DOMed counterparts, but that the K within these nominals has undergone further bleaching in that the syntactic [Person] feature which shows up in their feature specification is unvalued i.e., [uPerson: __] (while with DOMed variants the feature is valued). As a consequence, the [iPerson:__] feature carried by the nominal DP complement may no longer be valued by way of agreement with the KP. The only consequence of the agreement mechanism taking place between K and the D head would be that the syntactic [Person] feature of the entire KP construct is [iPerson:__]. This feature needs further checking against an appropriate head, and the KP will have to scramble all the way to a PersonP above the vP to this end. Before reaching SpecvP, however, KP passes through SpecαP and checks its [uCase: _] against v, in line with DOMed KPs. Having checked case, KP undergoes further movement to SpecvP and checks its [iPerson:__] feature against the Pers head. Having reached this site, however, the DO acts as an intervener between the T head and the Subject DP in the sense that T cannot access the DP Subject for case assignment. As a consequence, the DO undergoes cliticisation (given that doubled nominals do not act as interveners for other φ agreement and movement operations cf. Preminger 2017). Cliticisation thus counts as a means to remove the DO blocking agreement between T and the Subject DP. The DO obligatorily cliticizes on T, enabling the DP subject to get its case checked.

5 On the feature specification of indirect objects 5.1 The featural make-up of the dative IO One further ingredient critical to our derivational account of ditransitive configurations has to do with the featural specification of the IO. This constitutes the main focus of the current chapter. Just like DOMed DOs, Goal DPs are sensitive to the animacy hierarchy. In what follows, we would like to build on this similarity and propose that IO also carries a [Person] feature (see also Cornilescu et al. 2017b). Before we arrive at this final proposal (section 2), we will first consider the relevant language facts in this respect (section 1).

5.1.1 Background Romanian allows for both inflectional datives (195) as well as prepositional ones (196). The prepositional dative is headed by the (functional) preposition la68 and both types of datives may be clitic doubled69: (195) Mihaela (îi) trimite o scrisoare Mariei. Mihaela (CL.3.DAT.SG.) sends INDEF.ART. letter Mary.DAT ‘Mihaela sends a letter to Mary.’ (196) Mihaela (îi) trimite o scrisoare la Maria. Mihaela (CL.3.DAT.SG.) sends INDEF.ART. letter to Mary.DAT ‘Mihaela sends a letter to Mary.’ The inflectional and the prepositional dative exhibit similar syntactic and semantic properties but differ one from the other at a stylistic level i.e., while the

68 La does not only represent a functional dative preposition, but it is also the core lexical preposition in location and movement frames as in (1). (1)

M-am dus la piață. REFL.1.SG.ACC-have.1.SG. gone to market ‘I went to the market.’

69 For a discussion of the history and the regional distribution of the prepositional dative, see Iorga (2013). https://doi.org/10.1515/9781501513657-005

158

5 On the feature specification of indirect objects

inflectional dative represents the standard variant, the prepositional one is rather restricted to dialectal speech. Nevertheless, there exist certain contexts in the standard language where an inflectional dative is disallowed and a prepositional one is required. This is the case of IOs headed by determiners which are morphologically invariable e.g., niște (some), cardinals a.o. In (197) below, for instance, only the prepositional dative is possible with the indefinite determiner niște or the cardinal trei. (197) Am împrumutat cărțile la niște/ trei colegi. Have.1.SG. lent books.DEF.ART. to some three colleagues ‘I lent the books to some colleagues.’ The IO in (198) exhibits accusative morphology. Even so, la should be analysed as a dative marker, given that the la marked IO may only be doubled by means of a dative clitic (198a); an accusative clitic is out (198b): (198) a. I-am dat apă CL.3.DAT.SG. give.1.SG. water ‘I gave water to the horse.’ b. *L-am dat apă CL.3.ACC.SG. give.1.SG. water ‘I gave water to the horse.’

la cal. to horse la cal. to horse

Note further that the fact that IOs may be clitic doubled suggests that laphrases are not PPs but rather DPs, given that pronominal clitics may only double DPs. Indeed, la has not been analysed as a preposition in the literature, but as a dative case marker (Diaconescu and Rivero, 2007, Tigău 2014). In this respect, the dative marker la has been bleached of the semantic import that the lexical la carries and has shifted to a more abstract interpretation. Given its semantic bleaching, the dative la may be used not only in ditransitive configurations to mark Goal DPs (where it may well be argued that it has retained at least some of its directional meaning, given that at least some of the verbs involved in these configurations denote caused motion); in fact, la may be used with all the other θ-roles, with which the inflectional dative has also been associated – Possessor-Goals (199), Beneficiaries (200), Source/Maleficiaries (201). Note also that it is only the dative DP (whether inflectional or prepositional) that may be clitic doubled. When lexical prepositions such as pentru (for) or de la (from) are employed, clitic doubling is no longer possible given the PP status of the nominal:

5.1 The featural make-up of the dative IO

(199) IO > Possessor-Goal a. Mama (le)-a dat Mother.DEF.ART. CL.3.DAT.PL-has given ‘Mother gave chocolate to the children.’ b. Mama (le)-a dat Mother.DEF.ART. CL.3.DAT.PL-has given ‘Mother gave chocolate to the children.’

159

ciocolată copiilor. chocolate children.DAT ciocolată la copii. chocolate to children

(200)IO > Beneficiary a. Mama le-a cusut fetelor rochia Mother CL.3.DAT.PL.-has sewn girls.DAT dress.DEF.ART. cea roșie. DEM.ART. red.FEM.SG. ‘Mother sewed the red dress for the girls.’ b. Mama (*le)-a cusut pentru fete rochia Mother CL.3.DAT.PL.-has sewn for girls dress.DEF.ART. cea roșie. DEM.ART. red.FEM.SG. ‘Mother sewed the red dress for the girls.’ c. Mama le-a cusut la fete rochia Mother CL.3.DAT.PL.-has sewn to girls dress.DEF.ART. cea roșie. DEM.ART. red.FEM.SG. ‘Mother sewed the red dress for the girls.’ (201) IO > Source/Maleficiary a. Copiii le-au furat vecinilor Children.DEF.ART CL.3.DAT.PL.-have stole neighbors.DAT cireșele din livadă. cherries.DEF.ART. from orchard ‘The children stole the cherries from the neighbors’ orchard.’ b. Copiii (*le)-au furat de la vecini niște Children.DEF.ART CL.3.DAT.PL.-have stole from neighbors some cireșe cherries ‘The children stole some cherries from the neighbors.’ c. Copiii le-au furat la vecini Children.DEF.ART CL.3.DAT.PL.-have stole to neighbors

160

5 On the feature specification of indirect objects

cireșele din livadă. cherries.DEF.ART. from orchard ‘The children stole the cherries from the neighbors’ orchard.’ When the subject is a Cause instead of an Agent, as is the case above, the dative is interpreted as an Affected Experiencer and substitution of an inflectional dative with la is possible: (202) a. Acest medicament (le) poate da dureri de This medicine CL.3.DAT.PL. can give pain of unora dintre bolnavi. some.DAT of sick ‘This medicine may give headaches to some sick people’ b. Acest medicament (le) poate da dureri de This medicine CL.3.DAT.PL. can give pain of la unii dintre bolnavi. to some of sick ‘This medicine may give headaches to some sick people’

cap head

cap head

IO does not usually denote inanimate referents, at least when used in the inflectional dative: (203) Am turnat apă *florilor/ la flori. Have.1.SG. poured water flowers.DAT to flowers ‘I watered the flowers’ Exceptionally, however, abstract nouns may function (metaphorically) as inflectional datives (204a), the prepositional dative is disallowed in this case (204b): (204) a. A supus proiectul atenției bordului. Has.3.SG. submitted project.DEF.ART. attention.DAT board.GEN Lit. ‘He submitted the project to the board’s attention.’ b. *A supus proiectul la atenția bordului. Has.3.SG. submitted project.DEF.ART. to attention board.GEN Lit. ‘He submitted the project to the board’s attention.’ (Cornilescu 2017: 11 c,d, p. 4) As seen, inflectional and prepositional datives may give rise to various interpretations (Goals, Possessors, Beneficiaries, Sources/Maleficiaries, Affected

5.1 The featural make-up of the dative IO

161

Experiencers). It thus seems that they generally denote human referents. Since only nouns which are prominent on the animacy hierarchy appear to function as possible realisations of the dative θ-roles, we will posit that dative DPs carry a [Person] feature. The subsequent section builds on this idea.

5.1.2 A [Person] feature for Goal DPs As shown in the previous section, an essential characteristic of the DPs which realize the dative θ-roles is the fact that they are sensitive to the animacy hierarchy i.e., it is only nouns which range high on this hierarchy that count as possible realizations of the dative θ-roles. Other categories of nouns are ruled out (e.g. abstract nouns), in principle, though metaphorical and idiomatic exceptions may arise. In this, IOs pattern with DOMed DOs which have also been shown to exhibit the same sensitivity so one natural proposal would be to consider that IO DPs are also syntactically marked for [+Person]. The two datives, however, seem to have different cut-off points regarding which nouns may be lexically marked [+Person] and which nouns may not. Consider the clear-cut grammaticality contrast in (205) where the inflectional dative is only felicitous with the Goal DP denoting a human referent (205a), while a prepositional dative may accept a DP complement denoting an inanimate referent (205b). In (205c), both an inflectional dative as well as a prepositional one may co-occur with an IO nominal denoting an animate, non-human referent: (205) a. Am turnat vin mesenilor/ la meseni. Have.1.SG. poured wine guests.DAT to guests ‘I have poured wine to the guests.’ b. Am turnat apă *florilor/ la flori. Have.1.SG. poured water flowers.DAT to flowers ‘I have poured water to the flowers.’ c. Am dat apă cailor/ la cai. Have.1.SG. given water horses.DAT to horses ‘I gave water to the horses’ Unlike inflectional datives, la phrases are allowed with non-human referents, probably on account of the fact that the functional la in ditransitives

162

5 On the feature specification of indirect objects

has a lexical counterpart in the locative preposition la wherefrom it has originated.70 However, as seen, the meaning of the functional la is more abstract since, in spite of its lexical/directional meaning, it is employed with any θ-role associated with the inflectional dative and which points to human referents. Thus, the essential property of the DPs that may realize the dative θ-roles is sensitivity to the animacy hierarchy. Only nouns prominent on this hierarchy are possible realizations of the dative θ-roles. Drawing on these observations, we will thus extend the same proposal for dative nominals as we did for DOMed DOs with respect to their featural load. More specifically, we will consider that both inflectional datives as well as their prepositional counterparts should be lexically specified for a binary grammatical [Person] feature. Following Richards (2008), we take [Person] to be the way in which scalar concepts such as the animacy and the definiteness hierarchies get incorporated into the system of the minimalist grammar. We further consider the la marker as the functional category labelled K similar to the DOM marker pe. Since K is external to definiteness and indefiniteness markers, it takes a DP complement acting as a phrasal affix as in (206): (206)

KP ei K[uC] la

DP ei D[uC] #P/NP niște copii

As already argued in the previous sections, [+Human(like)] NPs incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by merging K. K itself carries a valued [uPerson] feature. (207) shows this at work: the existence of an unvalued feature in N and then in D triggers the insertion of K. The entire KP will thus have the feature specification [iPerson: val]

70 See Tigău (2014) for an analysis of la in ditransitives as a dative marker.

5.1 The featural make-up of the dative IO

(207)

163

KP ep K +p uφ:__ u-person:val

DP ep D +d uφ:__ idef:__ iperson:___

la

nişte

NP +N iφ:val u-def i person:___ +HUM copii

Inflectional datives will follow the same pattern of analysis and evince the same feature specifications as KP, possessing a silent K head. When the IO is doubled by means of a pronominal clitic, the K constituent is further bleached and only carries an unvalued [Person] feature on the model of CDed+DOMed DOs i.e., [uPerson:___]. As such, the [Person] feature of the CDed IO needs checking, just like the one on the DP it selects. The entire KP ends up having the feature specification [iPerson: ___] (208) and has to find a way whereby to value this feature: (208) a. Le-am dat cartea la niște colegi. CL.3.DAT.PL.-have.1.SG. given book.DEF.ART. to some colleagues ‘I gave the book to some colleagues.’ b. KP ep K DP ep +p D NP uφ:__ +d +N uperson:___ uφ:__ iφ:val uCase__ idef: __ u-def iperson: ___ i person: ___ uCase __ uCase __ +HUM la nişte colegi Doubled inflectional datives pattern on a par with their prepositional counterparts, exhibiting a silent K. Note that undoubled IOs pattern with DOMed ones with respect to their feature specification, both expressions being KPs and bearing an [iPerson: val] feature specification and having real argumenthood. Drawing the parallelism even further, we assimilate CDed+DOMed DOs with CDed IOs with respect to their status as KPs and their feature specification as

164

5 On the feature specification of indirect objects

[iPerson: ___]. In this latter case, the K head has undergone further bleaching and only contributes an unvalued [uPerson: ___] feature, which does not suffice to value the [iPerson: ___] carried by the nominal expression. The feature of the KP will thus have to be valued at a later point during the derivation (most probably against the Appl head of the Applicative Projection proposed for ditransitives see Pylkkänen 2002, 2008 a.o. in this sense).

5.2 Conclusions This brief chapter focused on dative DPs and put forth the proposal to analyse these DPs on a par with DOMed DOs by allowing them to bear a [Person] feature. The idea underlying this similarity between IOs and DOMed DOs rests on the observation that dative DPs are sensitive to the animacy hierarchy, realising θ-roles which generally correspond to human individuals. In line with Richards (2008), we posited that sensitivity to animacy has a [+Person] specification as its syntactic counterpart. IOs were thus analysed as KPs bearing [iPerson: val]: the NP itself carries an unvalued [iPerson] which is copied on D and valued by K, which bears a valued [uPerson]. When clitic doubled, IOs have a slightly different feature specification, as the expression of K’s having undergone further bleaching only contributes [uPerson: ___]. When matching the unvalued [iPerson] of its DP complement, K can no longer value it such that further checking mechanisms would have to be sought for. The next section(s) will show this at work with ditransitive configurations.

6 Romanian ditransitive configurations This chapter focuses on Romanian ditransitive configurations and has a twofold aim: it first discusses the results of a series of experiments dwelling on binding dependencies holding between the two internal arguments. Its second aim is to provide an account of the experimental findings. Quantificational dependencies with possessives and with anaphors are checked in an effort to uncover the underlying configuration(s) with ditransitives. This undertaking finds a first justification in the literature on English twoobject constructions, where the Prepositional Object Construction (POC) and the Doubled Object Construction (DOC) have been shown to differ structurally, with the two internal objects evincing different c-command properties. Many studies focusing on Romance ditransitives have drawn a complete parallelism between these constructions and their English counterparts, positing the existence of a corresponding POC and DOC with the same configurational properties (Demonte 1995, Cuervo 2003a,b). Romanian has also been argued to closely mirror English in this respect (Diaconescu and Rivero 2007). More recent studies on Romance have, nevertheless, pointed out various empirical difficulties regarding such accounts of ditransitives, showing that the internal arguments within Romance ditransitives actually exhibit symmetric c-command (Georgala 2011, 2012, Georgala et al. 2008, Pineda 2012, 2013 a.o.). Our own intuitions on Romanian also seemed to favour the existence of bi-directional c-command within ditransitives so it was necessary to carry out an experiment meant to shed light on the facts. As already mentioned, a second aim of this chapter is to propose an account of Romanian ditransitives accommodating the experimental findings. The account is derivational, assuming one basic underlying configuration and explaining the experimentally uncovered symmetric binding potential of the two internal arguments by way of movement. The internal make-up of the two DPs is also shown to be instrumental with respect to their interaction and several problematic situations are shown to arise as a consequence of some intervention effects caused by the competition for the same probe between the two internal arguments that bear similar feature specifications. The chapter is structured as follows: in section 1 we discuss the main properties of ditransitive configurations in English and Romance with a focus on a seminal account on Romanian. The discussion about Romanian ditransitives further leads us to clarify the main reasons behind our experimental untertaking. Section 2 is devoted to the experiments proper, discussing their design and the results we obtained. Section 3 contains a proposal regarding what we consider to https://doi.org/10.1515/9781501513657-006

166

6 Romanian ditransitive configurations

be a more suitable syntactic analysis of Romanian ditransitives. Section 4 presents the conclusions.

6.1 Background 6.1.1 Preliminaries One of the main issues discussed with respect to ditransitive configurations is the relation between the Double Object Costruction (DOC) variant (209a) and the Prepositional Object Construction (POC) (209b): (209) a. Mary gave JohnGoal a bookTheme. b. Mary gave a bookTheme to JohnGoal. According to some studies, the two configurations are completely independent, while others stress the structural connectedness between them. The former line of analysis is known as the Alternative Projection Account, while the latter bears the name of the Derivational Account. There are two important arguments put forth to justify the hypothesis that the two configurations are actually unrelated and should be viewed as independent alternants: these are the lack of semantic uniformity of the alternating dative constructions and the asymmetric binding potential exhibited by the two internal arguments. a) Semantic uniformity or the lack thereof With respect to the ‘semantic argument’, Oehrle (1976) discusses for instance the examples under (210), arguing that they are not semantically equivalent in that the DOC may give rise to a causative reading, which the POC does not render available. More specifically, while (210a) allows both for a possession reading (a’) and a causative one (a’’), (210b) only permits the change of possession interpretation provided in (b’): (210) a. Nixon gave Mailer a book. a’: Nixon performed some action by means of which he caused a change of possession a’’: Mailer wrote a book which he wouldn’t have been able to write if it hadn’t been for Nixon = causative reading b. Nixon gave a book to Mailer.

6.1 Background

167

b’: Nixon performed some action by means of which he caused a change of possession b’’: * Mailer wrote a book which he wouldn’t have been able to write if it hadn’t been for Nixon = causative reading (Oherle 1976: 1, p. 19) Lack of semantic equivalence between the POC and the DOC variants has also been signalled with examples such as the ones under (211), where it is argued that (211a) allows for an interpretation according to which Mary has learnt German so that she may now be considered a proficient speaker of the language. Example (211b), on the other hand, does not seem to allow such an entailment i.e., it may be quite possible for Mary not to have acquired German even though John did his best to teach her. (211) a. John taught Mary German. b. John taught German to Mary. Similarly, the reading in (212a) may only be that Mary has caught the ball thrown by John and, as such, the continuation provided in between brackets is not suitable. On the other hand, the POC in (212b) allows the continuation in the brackets since one no longer understands that Mary necessarily caught the ball thrown by John. (212) a. John threw Mary the ball (*but Bill jumped and caught it). b. John threw the ball to Mary (but Bill jumped and caught it). (Avram 2006:74, 225) In the same vein, Larson (2014) quotes Kayne (1975)’s observation regarding the examples in (213), where variant (b) seems to imply more strongly than its counterpart in (213a) that the baby referred to has already been given birth to. As a consequence, (213b) would be less natural than its counterpart (213a) if it were uttered by a woman who is still expecting or planning to become pregnant: (213) a. I knitted this sweater for our baby. b. I knitted our baby this sweater.

(Larson 2014: 9a,b, p. 38)

Building on examples such as those in (210) – (213), advocates of the alternative projection account assume that the semantic difference between the POC and the DOC is actually systematic, with the prepositional dative expressing obligatory caused movement and the DOC describing obligatory caused possession. The difference has been accounted for in various ways, either in terms of a

168

6 Romanian ditransitive configurations

different event structure (Krifka 2004), as a difference springing from the preposition relating the two internal arguments (Harley 2002) or between the light verbs go and cause present in the POC and the DOC respectively (Cuervo 2003a). Krifka (2004) for instance provides the event structures below arguing that the indirect internal argument in the POC has a salient directional meaning, given that it is part of an event of movement (MOVEðeÞ in (214a)). In contrast, the two internal arguments within the DOC enter a transfer of possession relation, resulting in the state where the referent denoted by the IO has something (HAVE ðPeter, bookÞ in (214b)):   (214) a. 9e9e′½AGENT ðe, MaryÞ ^ THEMEðe, bookÞ ^ CAUSE e, e′       ^ MOVE e′ THEME e′, book ^ GOAL e′, Peter b. 9e9s½AGENT ðe, MaryÞ ^ THEME ðe, bookÞ ^ CAUSE ðe, sÞ ^ s: HAVE ðPeter, bookÞ (Krifka 2004: 31, p. 265) Indeed, such an analysis seems suitable in the case of verbs of transmission and ballistic motion such as send or throw which appear to fit the two patterns of analysis: in (215a) Ann is interpreted as the possessor or recipient of the package, while in (215b) she acts as the goal of motion: (215) a. Mary sent Ann a package. ‘Mary CAUSED Ann to HAVE a package (by sending).’ b. Mary sent a package to Ann. ‘Mary CAUSED a package to GO to Ann (by sending).’ An analysis along these lines has several characteristics to recommend it: Firstly, we get the right distributional facts – while the oblique object is restricted to denoting animate referents as befits all possessors, the prepositional object is free from such constraints. The examples under (216) show this, with an ungrammatical DOC in (216b) and a grammatical POC in (216a): (216) a. Mary send a package to Bucharest. b. *Mary sent Bucharest a package. Larson (2014) also points to verbs such as ‘distribute’, ‘donate’, ‘release’, ‘lose’ etc. which denote release or loss of possession but which do not target a possessor or a recipient per se. As expected, the DOC is ungrammatical with these verbs: (217) a. John donated money (to/among those charities). b. *John donated those charities money.

6.1 Background

169

(218) a. John lost money to the stock market. b. *John lost the stock market money (on the (a) reading). (apud Larson 2014: 19, 20, p. 41) Finally, path modifiers such as halfway/all the way are only possible with the prepositional dative of send or throw and this is expected under the hypothesis that this configuration evinces the underlying caused motion pattern, which is not present in the DOC: (219) a. Jane threw the ball halfway to Sue. b. Jane threw (*halfway) Sue a ball (*halfway). Nevertheless, these semantic distinctions do not seem to carry over all classes of ditransitives and the hypothesis that the semantic differences professed to hold between the DOC and the POC are systematic has been questioned by a number of more recent studies such as Rappaport-Hovav and Levin (2008), Ormazabal and Romero (2012), or MacDonald (2015) a.o. Rappaport-Hovav and Levin (2008) show, for instance, that the caused motion meaning associated with the POC is not present with some of the verb classes which exhibit the dative alternation and hence allow the POC (e.g., verbs of future having such as grant, promise, allocate, verbs of communication tell, teach etc.). Moreover, the caused possession reading, linked to the DOC, is shown to surface with the POC variant of certain verb classes expressing caused motion. Rappaport-Hovav and Levin (2008) distinguish between two large classes of verbs (which further divide into subclasses), which may be roughly described as verbs of possession (give-verbs) and verbs of movement (throw-verbs). Giveverbs are shown to express caused possession with both the (English) POC and the DOC construction. This comes as a consequence of the fact that give verbs lack a path component, and as such may not express movement in any sense (even a metaphorical one), even if one of their arguments, i.e., the prepositional dative DP, suggests a Goal interpretation. Give-verbs thus have a single event schema expressing caused possession and their dative argument functions as a Possessor in both of their syntactic alternatives. On the other hand, throw-verbs and send-verbs support both readings due to the fact that they have a path component in their lexical make-up, which means that their Prepositional Dative is a genuine Goal. The caused possession reading, signalled by the DOC in English, occurs when there is an inference of successful transfer, so that the Goal becomes a Possessor-Goal. One further argument against an alternative approach is contributed by Larson (2014) who observes that the ‘extra’ elements of meaning which seem to

170

6 Romanian ditransitive configurations

differentiate the DOC in examples such as (211)-(213) above are actually conversational implicatures given the fact that they may be easily cancelled when suitable continuations are provided: (220) a. John taught Mary German for a whole semester but the final exam results showed that the little brat hadn’t learnt a damn thing. b. I knitted our baby this sweater. I just hope it fits him when he’s born. (apud Larson 2014: 10, p. 38) Larson further observes that such conversational implicatures rather than determining syntactic projection are actually calculated from the syntax and semantics. The semantic differences are further explained in terms of affectedness, which derives from the position occupied by the Goal argument: while in the DOC the Goal DP is inside the affectedness domain of the verb, in the POC configuration, the Goal DP is outside it, hence a speaker would be more prone to interpret the indirect object in (211a) and (213b) as affected by the action denoted by the verb. It follows that the contrasts in (211) and (213) are not actually relevant for the dative projection given that the extra element of meaning identified in the DOC variant is ‘read off the surface rather than figuring in its projection’ (Larson 2014, p. 38). A real difference with respect to the projection of arguments should actually boil down to a difference underlying the relations that entities bear to the event or state denoted by the verb i.e., a difference in theta roles, as is the case, for instance, of the two blame constructions in (221) below; blame functions as a dative verb in (221a) giving rise to a caused possession event structure involving the theta roles of i.e., ‘Job CAUSED God to HAVE the blame for his troubles’ and as a locative one in (221b) evincing an theta grid i.e., ‘Job CAUSED the blame for his troubles to GO on God’: (221) a. Job blames God for his troubles. b. Job blamed his troubles on God. Finally, the alternative projection account is further weakened by the semantic uniformity of alternating Dative constructions in Romance. Ormazabal and Romero (2010), Pineda (2012, 2013) a.o. argue against the existence of semantic differences between Spanish ditransitives with a clitic doubled IO and their counterparts with an undoubled IO; these have been claimed to correspond to the English DOC and POC respectively by the proponents of the Alternative Projection Account (Demonte 1995, Cuervo 2003a, b). Romanian goes along the same lines – in fact the descriptive grammars of Romanian never mention two semantically different structures for ditransitive give-verbs.

6.1 Background

171

b) Asymmetric binding Another important argument resorted to by the advocates of the alternative account in order to justify the view of two independent configurations revolves around the structural differences involving the DOC and the POC. Indeed, an issue noticed with respect to the two configurations in English concerns the different binding dependencies at stake: as early as Barss & Lasnik (1986), and Aoun & Li (1989) it is shown that the two internal arguments are not hierarchically equal and that their c-command potential differs, function of the configuration at stake: in the DOC, the IO seems to be hierarchically superior to DO, whereas the POC exhibits the opposite c-command relations between the two arguments. The different asymmetric c-command potential may be seen in (222) and (223) below: POC: DO c-commands IO (222) a. I showed Maryi to herselfi in the mirror. b. *I showed herselfi to Maryi in the mirror. DOC: IO c-commands DO (223) a. *I showed himselfi Johni in the mirror. b. I showed Johni himselfi in the mirror. (apud Barss & Lasnik 1986: 2, 3, p. 347) In (222) the DO Mary is shown to be able to bind the IO anaphor herself (a), while the DO anaphor in (b) may not be bound by the IO Mary, which is structurally lower than the DO in the POC configuration. In the DOC, the IO is hierarchically superior to the DO DP and it may consequently bind the DO anaphor (223b). Binding of the IO anaphor by the DO John in (223a) is impossible in the DOC, given that the DO is now hierarchically lower than the IO. The same binding dependencies between the two internal arguments is argued to hold for the examples under (224) and (225) involving possessives: (224) POC: DO c-commands IO a. I denied each paychecki to itsi owner. b. */?? I denied hisi paycheck to each worker. (225) DOC: IO c-commands DO a. * I denied itsi owner each paychecki. b. I denied each workeri hisi paycheck. (Barss & Lasnik 1986: 6, 7, p. 348)

172

6 Romanian ditransitive configurations

Thus, the DO expectedly binds the possessive within the IO in (224a), while the opposite is not possible (224b). In (225), the c-command relations are reversed and it is now the IO which may bind into the DO (225b), but not the other way around. One way to account for the binding asymmetries evinced by the two internal arguments in the DOC vs. the POC under the alternative projection framework was put forth by Harley (2002), who proposes that this difference boils down to the different prepositions which relate the two internal arguments. Double object verbs are argued to decompose into two heads, a vCAUSE which amounts to a predicate selecting the external argument and a PHAVE in the DOC vs. a PLOC in the POC. The differences in the hierarchical structures signaled in Barss and Lasnik (1986) are argued to arise in this way alongside the generalisation effects uncovered in Oherle (1976). Consider (226) below showing this at stake: (226) a. The double complement structure: vP ei v ei v PP ei CAUSE DP P´ ei a letter P PP Ploc to Mary b. The double object structure vP ei v ei v PP ei CAUSE DP P´ ei Mary PP P a letter Phave (Harley 2002: 3, p. 4) Numerous studies on Romance ditransitives have adopted the Alternative Projection account, assuming structural differences between ditransitive configurations containing clitic doubled DP Goals and their undoubled counterparts by

6.1 Background

173

grouping the former with the DOC and analyzing the latter as POCs (Demonte 1995, Cuervo 2003a, b a.o. for Spanish). Cuervo (2003a, b) claims, for instance, that the same syntactic and semantic differences found with DOC and POC in English are also at work in Spanish, where the two configurations are distinguished one from the other by means of a dative pronominal clitic doubling the IO: the configuration containing an undoubled IO functions as the equivalent of the English POC (227b), while the one containing a doubled IO patterns with the DOC (227a): (227) a. Pablo le mandó un diccionario a Gabi. Pablo CL.3.DAT.SG. sent INDEF.ART. dictionary to Gabi ‘Pablo sent Gabi a dictionary.’ b. Pablo mandó un diccionario a Gabi. Pablo sent INDEF.ART. dictionary to Gabi ‘Pablo sent Gabi a dictionary.’ (Cuervo 2003a: 28 p. 30) The same c-command configurations as with the English counterparts are argued to hold for Spanish ditransitives: in the construction containing a clitic doubled IO (227a), the DO DP is claimed to merge as the complement of a low ApplP which also hosts the IO in its specifier (228a). The clitic doubled IO thus c-commands the DO. The ditransitive configuration containing an undoubled IO is argued to exhibit the opposite c-command relation between the two internal arguments, given that IO is generated low in the complement position of a PP, while the DO merges above the PP (228b). Note that the Appl head is argued to have the dative clitic as a spell-out, unlike English or several other languages where it is null (see Pylkkänen 2002): (228) a.

VoiceP ru EA v ru Voice vP ru v ru Root ApplP ru Appl´ DPGoal ru Appl DPTheme

(Cuervo 2003a: 39, p. 35)

174

6 Romanian ditransitive configurations

b.

VoiceP ru EA v ru Voice vP ru v ru Root ru DPTheme PP ru P DPGoal

(Cuervo 2003a: 38, p. 34) The same binding dependencies as in English are argued to hold in the DOC and POC Spanish counterparts. In (229a) the DO el maestro is argued to bind the anaphor IO a sí mismo. (229b) shows that the opposite may not hold: (229) POC: DO c-commands IO a. Valeria mostró el maestroi a sí mismoi. Valeria showed DEF.ART. teacher to himself ‘Valeria showed the teacher to himself.’ maestroi. b. *Valeria mostró a sí mismoi al Valeria showed to himself to-DEF.ART teacher ‘* Valeria showed himself to the teacher.’ (Cuervo 2003a: 43, p. 36) The clitic doubled variants of the examples in (229) are claimed to give rise to the opposite binding dependecies with the anaphor: this time, the IO is arguably able to bind the DO anaphor (230b), while the DO may no longer bind an IO anaphor (230a): (230) DOC: IO c-commands DO a. *Valeria le mostró el maestroi a sí mismoi. Valeria CL.3.ACC.SG. showed DEF.ART. teacher to himself ‘Valeria showed himself the teacher.’ maestroi. b. ??Valeria le mostró a sí mismoi al Valeria CL.3.ACC.SG. showed to himself to-DEF.ART teacher ‘Valeria showed the teacher himself.’ (Cuervo 2003a: 44, p. 37)

6.1 Background

175

The same binding dependencies are argued to hold for binding dependencies with possessives: (231a) below is infelicitous on a quantificational dependence interpretation, given that the IO a los trabajadores (to the workers) may not bind the possessive within the DO sus cheques (their checks). On the other hand, the DO in (231b) may bind the possessive within the IO. (231) POC: DO c-commands IO trabajadoresi. a. *Entregamos susi cheques a los Gave.1.PL their checks to DEF.ART.PL workers ‘*We gave their checks to the workers.’ b. La policía entregó los bebési a susi DEF.ART.FEM police gave DEF.ART.PL babies to their (respectivos) padres. respective parents ‘The police gave the babies to their (respective) parents.’ (Cuervo 2003a: 47a, 48a, p. 38) The clitic doubled variants of the examples in (231), are argued to exhibit the opposed binding dependencies between the two internal arguments: as with the English DOC, the possessive within the DO may be bound by the clitic doubled IO (232a), while (232b) shows that the possessive within the IO may not be bound by the DO: (232) a. ?Les entregamos susi cheques a CL.3.DAT.PL gave.1.PL their checks to trabajadoresi. workers ‘We gave the workers their checks.’ b. */?? La policía les entregó DEF.ART.FEM police CL.3.DAT.PL gave susi (respectivos) padres. their respective parents ‘* The police gave their parents the babies.’

los DEF.ART.Pl

los bebési a DEF.ART.PL babies to

Romanian has also been claimed to pattern on a par with other Romance languages in this respect. Building on Cuervo (2003a), Diaconescu and Rivero (2007) distinguish between ditransitive configurations containing an undoubled IO (which they regard as similar to the English POC), and configurations where a dative clitic doubles the IO (which is assimilated to the English DOC). The former configuration obtains when the Goal DP is clitic doubled, while the

176

6 Romanian ditransitive configurations

latter holds with undoubled Goal DPs. (233) below contains the respective configurations (as was also proposed by Cuervo 2003a for Spanish): (233) a. POC: Theme c-commands Goal [VoiceP DPAgentVoice[vPv [PPDPTheme P DPGoal]]] b. DOC: Goal c-commands Theme (clitic doubling) [VoicePDPAgentVoice[vPv [ApplPDPGoal [clAppl] [VP V DPTheme (Diaconescu and Rivero 2007: 219–220) As in Cuervo (2003a), the Applicative head in the DOC configuration is not null as it is occupied by the dative clitic. The two distinct hierarchical structures are thus triggered by the presence of a dative clitic or the lack thereof: in (233a), the IO is merged low within a PP, while the DO occupies the specifier of this projection c-commanding the IO. In (233b), the IO merges in the SpecAppl, while the DO occupies the complement position of the lexical verb. Thus, DOC readings depend on CD since IO is introduced by the clitic in Appl. Diaconescu and Rivero further argue that the binding asymmetries between the two arguments pattern with the ones in English (Barss and Lasnik 1986) i.e., in the cliticless ditransitive, the DO c-commands and therefore may bind an anaphor/possessive pronoun inside the IO. Alternatively, the IO c-commands the DO in the DOC and binding relations are reversed. Their account makes a number of predictions to the effect that some configurations are discarded as ungrammatical. For instance, a DO should never be able to bind a clitic doubled IO since the latter DP is merged in a higher, c-commanding position. Similarly, an undoubled indirect object should not be able to bind into a direct object, since it merges in a lower position. Building on Barss and Lasnik (1986), Diaconescu and Rivero (2007) support their claims with respect to the c-command relations within ditransitives by resorting to tests resting on binding with anaphors and possessives. Here are some of their data: pe ea însășii. (234) a. ? Ion i-a descris feteii John CL.3.DAT.SG.-has described girl.DAT DOM her herself ‘John described the girl to herself. (Lit. John described to the girl herself.)’ b. * Ion i-a descris ei înseșii fatai. John CL.3.DAT.SG.-has described her.DAT herself girl.DEF.ART. ‘John described herself the girl.’ (Diaconescu and Rivero 2007: 25, p 27)

6.1 Background

177

The examples in (234) arguably indicate that, while a clitic doubled dative DP may bind the anaphor within a direct object (a), the reverse is not possible (b): a direct object may not bind the anaphor within a clitic doubled indirect object. Similarly, the undoubled indirect object in (235a) below is claimed to be unable to bind the anaphor in the direct object, while the opposite holds in (235b), i.e., the direct object may bind the anaphor within an undoubled indirect object: pe ea însășii. (235) a. * Ion a descris feteii John has described girl.DAT DOM her herself ‘John described the girl to herself. (Lit. John described to the girl herself.)’ b. Ion a descris ei înseșii fatai. John has described her.DAT herself girl.DEF.ART. ‘John described herself the girl.’ (Diaconescu and Rivero 2007: 27, p.30) The same c-commanding relations are argued to hold with possessives: a clitic doubled indirect object may bind the possessor within the direct object in (236a) below, while the direct object may not bind the possessor within the clitic doubled indirect object (236b): (236) a. I-am săui. dat muncitoruluii cecul CL.3.DAT.SG.-have.1.SG. given worker.DAT check.DEF.ART. his ‘I have given the worker his cheque.’ b. ?? Poliția i-a dat tatălui săui Police.DEF.ART. CL.3.DAT.SG.-has given father.DAT his pierdut. copiluli child.DEF.ART. lost ‘The Police has given the father his lost child.’ (Diaconescu and Rivero 2007: 28, p.30) Furthermore, Diaconescu and Rivero (2007) argue that an undoubled indirect object may not bind the possessor within the direct object in (237a) below, while the direct object may bind the possessor within the indirect object (237b): săui. (237) a. *Am dat muncitoruluii cecul have.1.SG. given worker.DAT check.DEF.ART. his ‘I have given the worker his cheque.’

178

6 Romanian ditransitive configurations

b. Poliția a dat tatălui săui copiluli pierdut. Police.DEF.ART. has given father.DAT his child.DEF.ART. lost ‘The Police has given the father his lost child.’ (Diaconescu and Rivero 2007: 30, p.33) The data presented in Diaconescu and Rivero (2007) show that the two internal arguments have an asymmetric binding potential as reflected by the configurations under (233) above, similarly to their Spanish and English counterparts. This lends support to the alternative projection account they offer, building on the analysis offered by Cuervo (2003a) for Spanish. The asymmetric binding potential of the two internal arguments in the POC and the DOC does not, however, represent an irrefutable argument that the two configurations function as independent structures. In fact, Larson (1988), in a seminal account, proposes a derivational analysis71 whereby the prepositional dative construction represents the basic configuration and the double object counterpart is derived from it by way of a mechanism similar to the passive derivation inverting the Agent and the Theme: in this particular case, the DOC was posited to obtain by raising the Goal to an empty, object-like position and realising the Theme as an adjunct (238). (238)

[vP John [V’ write [VP Mary [V’ [Vo write Mary] [DP a letter]]]]] (Larson 2014: 5c, p. 36)

A derivational account for ditransitives seems to be favoured in more recent studies on Romance (Ormazabal and Romero 2010, 2012, Pineda Pineda 2012, 2013), since it appears that the binding asymmetries signalled for English do not seem to arise in Spanish or Catalan. The results of our own experiments parallel these data showing that the two internal arguments have a symmetric c-command potential. Furthermore, the presence/absence of the pronominal clitic doubling the IO does not alter those properties of the construction which depend on the c-command relations between the two internal arguments, i.e. on binding and scope properties. These findings undermine the idea put forth by Diaconescu and Rivero that the dative clitic is the head of the Applicative projection licensing the IO in an English POC equivalent configuration. As will

71 Other proponents of the derivational account are: Baker 1988, 1996, den Dikken 1995, Ormazabal and Romero 2010, 2012 a.o., for whom the DOC is derived from the POC, while Dryer 1987, Aoun and Li 1989 a.o. argue that the POC obtains from the DOC.

6.1 Background

179

be shown below, the DO and IO exhibit symmetric binding abilities being able to bind into each other irrespective of word order, and of the presence/absence of the dative clitic. This suggests that a derivational account might be more adequate, given that it offers the possibility of reconstruction. It may nonetheless be possible for the alternative projection account to be implemented in a different way which might handle these configurational properties well. The following sections will dwell on some experimental data which seem to render a derivational account of Romanian ditransitives more desirable than an alternative one.

6.1.2 Motivation for the experiment and predictions Some of the examples supplied in Diaconescu and Rivero (2007) seemed problematic to us as native speakers. Some of the sentences which are discarded as ungrammatical appeared to only be infelicitous as a consequence of poor lexical choices, rather than because of ungrammatical structures. Thus, a DOC reading, where the IO binds the possessive within a DO, seemed quite possible in (239) below, even though the IO is not clitic doubled and the entire configuration should pattern with the English POC where IO is merged low in a position wherefrom it may not c-command the DO: turisti (239) Recepționerul arătă camera luii fiecărui Receptionist.DEF.ART. showed room.DEF.ART. his every.DAT tourist în parte, pentru a se asigura că nu sunt confuzii. in part for to REFL. ensure that not are confusions ‘The receptionist showed each tourist his room to make sure there are no confusions.’ Similarly, a sentence such as (240) below seemed quite felicitous to us under a reading pairing houses with owners, even though instances where a DO binds the possessive within a clitic doubled IO have been argued by D&R (2007) to be ungrammatical: (240) a. Statul nu i-a retrocedat proprietarului eii State.DEF.ART. not CL.3.DAT.SG.-has returned owner.DAT its de drept nici o casă naționalizată pe vremea comunismului. of right no house nationalized.FEM on time Communism.GEN ‘The State did not return to its rightful owner any house nationalised during the Communist regime.’

180

6 Romanian ditransitive configurations

Furthermore, these observations do not only hold for Romanian but extend to other Romance languages where the alternative projection account has been shown to face empirical difficulties. Ormazabal and Romero (2010) argue, for instance, that it is hard to find semantic differences between the cliticless and the clitic-doubled pattern(s) and that the two internal arguments show symmetric c-command (see also Georgala 2011, 2012, Georgala et al. 2008, Pineda 2012, 2013). Pineda (2012, 2013) argues against the view that the syntactic and semantic differences uncovered between the DOC and the POC in English are also apparent with Spanish ditransitives, where, it had been claimed, a dative clitic doubling the IO would distinguish between the two configurations (see Demonte 1995, Cuervo 2003a, b). Building on bidirectional c-command facts involving the IO and the DO, she shows that the dative clitic has no structural consequences: in (241a) the IO is shown to be able to bind the possessive within the DO, while the opposite binding relation also obtains in (241b). Both binding dependencies obtain irrespective of whether a dative pronominal clitic doubling the IO is present or not: (241) a.

ok/?

La profesora (le) DEF.ART.FEM professor.FEM (CL.3.DAT.SG) a cada niñoi. to each child ‘The teacher gave his drawing to each child.’ b. ok La profesora (le) DEF.ART.FEM professor.FEM (CL.3.DAT.SG) a sui autor. to its author ‘The teacher gave each drawing to its author.’

entregó sui dibujo gave his drawing

entregó cada dibujoi gave each drawing

(Pineda 2013: 26, p. 206)

The Spanish data are reinforced by similar facts uncovered in French and Italian, where ditransitives have been also shown to exhibit bidirectional c-command between the IO and the DO. Harley (2002) points out, for instance, that binding of both objects is possible in both directions as shown in (242) below: in (242a) the IO binds the possessive within the DO and in (242b) the DO binds the possessive in the IO: (242) a. Maria a donné soni crayon à chaque garçoni. Mary has given her pencil to every boy ‘Mary gave every boy his pencil.’

6.2 Experimental data on Romanian ditransitives

181

b. Jean a introduit chaque institutricei à sesi élèves. John has introduced every teacher.FEM to her pupils ‘John introduced every teacher to her pupils.’ (Harley 2002: 56, p. 34) Harley (2002) also quotes Giorgi and Longobardi (1991) on Italian ditransitives, which seem to pattern with the French examples in allowing bidirectional c-command. In (243a) the DO Maria binds the IO anaphor, while in (243b) the DO anaphor is bound by the IO a Maria: (243) a. Una lunga terapia psicoanalitica ha INDEF.ART.FEM long.FEM therapy psychoanalytic.FEM has restituito Mariai a se stessai. restored Mary to herself ‘A long psychoanalytic therapy has restored Mary to herself.’ b. Una lunga terapia psicoanalitica ha INDEF.ART.FEM long.FEM therapy psychoanalytic.FEM has restituito se stessai. a Mariai restored herself to Mary ‘A long psychoanalytic therapy has restored herself to Mary’ (Harley 2002: 55, p. 34) Given that our own intuitions occasionally disagreed with the data presented in Diaconescu and Rivero (2007) and inclined more towards the judgements expressed in Ormazabal and Romero (2010), Pineda (2012, 2013), Harley (2002), Giorgi and Longobardi (1991) a.o., and since the analysis seemed to be ‘data driven’, we thought that the only reasonable course of action was to get a better picture of the data by means of a comprehensive experiment. The next section extends upon the basic tenets of this experimental study.

6.2 Experimental data on Romanian ditransitives As already pointed out, the experiments we are about to present are necessary for two important reasons. Firstly, our intuitions differed from those in Diaconescu and Rivero (2007). Secondly, the more recent findings reported for ditransitives cross-linguistically question the appropriateness of an account distinguishing between two different ditransitive configurations as in English. Actually, Romanian seems to pattern with Spanish (Ormazabal and Romero 2010, Pineda 2012, 2013) in

182

6 Romanian ditransitive configurations

evincing one single ditransitive configuration, irrespective of whether there is a dative pronominal clitic doubling the IO or not. The general aim of the experiments was to capture the potential differences holding between those ditransitive configurations where the Goal DP had been clitic doubled and their undoubled variants. The (a)symmetric binding potential of the two internal arguments was thus at the core of our investigation and binding dependencies with possessives constituted the focus in the three experiments to be described. Two smaller experiments involving anaphor binding and reinforcing the results of the former three experiment(s) on quantificational dependencies with possessives will be presented at the end of this section: one discussed in Cornilescu et al. (2017a) and another, more comprehensive experiment on anaphor binding carried out for the current volume.

6.2.1 Quantificational dependencies with possessives 6.2.1.1 Design As already pointed out, the experiment on binding dependencies with possessives was comprised of three smaller experiments. These were similar in design and shared the same aim of testing binding possibilities holding between the two internal arguments in Romanian ditransitive constructions. We checked for the acceptability of items where we varied: a) the surface word order; b) the direction of binding c) the presence of a dative clitic doubling the IO. Two surface word orders were tested: one in which the DO preceded the IO and the reversed sequence. For each surface order, the direction of binding was varied with or without the addition of a clitic such that the following 8 patterns emerged: Table 11: surface word order, binding direction and clitic doubling. DO before IO

IO before DO

DO binds into IO

IO binds into DO

DO binds into IO

IO binds into IO

−cl

−cl

−cl

−cl

+cl

+cl

+cl

+cl

One further complication, which caused the division of the experiment into three parts, had to do with the possibility of differentially object marking (DOM) and

6.2 Experimental data on Romanian ditransitives

183

clitic doubling (CD) DOs in Romanian. To avoid the existence of too many experimental conditions, we decided to use the same type of DO throughout one experiment and to iterate the experimental efforts thrice, each time using a different type of DO i.e., unmarked DOs/ DOMed DOs/ CDed+DOMed DOs. Three different experiments thus emerged:72 Experiment 1: Binding with unmarked DO and IO Experiment 2: Binding with DOMed DO and IO Experiment 3: Binding with CDed+DOMed DO and IO For each experiment, we designed 32 sentences and varied them function of each of the eight patterns, by changing the surface word order between DO and IO, the direction of binding between the two internal arguments and the presence vs. absence of the dative clitic doubling the IO. 256 test items were thus obtained. The items were then distributed into 8 different lists using the Latin square method for an even distribution. Each questionnaire also included 32 fillers, grouped into 8 expectedly unacceptable items, 8 completely acceptable items and 16 average items with respect to acceptability (the fillers were separately checked for acceptability in a smaller, informal experiment). Each questionnaire thus ended up having a number of 64 items. The questionnaires were formatted as Google online forms in such a way that the potential respondent could only access one item at a time, without having the possibility of going back or forth. Each list was further assessed by at least 20 native speakers of Romanian such that more than 480 people took part in the three experiments. The items were to be evaluated on a seven-rung scale, with 1 representing the lowest acceptability value and 7, the highest one.

72 As known, Romanian DOs may be differentially object marked by means of the functional preposition pe, a similar preposition to the differential marker a in Spanish which, roughly, marks DOs that are proeminent on the animacy and definiteness scales (Aissen 2003). DOs thus marked may be further clitic doubled. What distinguishes between the three experiments is the type of DO employed: experiment 1 contained DOs denoting inanimate referents, which could not be pe marked, for experiment 2 we only selected pe marked DOs, while in experiment 3 only clitic doubled and pe marked DOs were employed.

184

6 Romanian ditransitive configurations

The results were then verified and outliers were removed i.e., questionaires in which more than 7 fillers had been wrongly assessed. Remaining questionnaires entered the statistical analysis (see Appendix 1).73 With respect to the general scores, one may notice their general tendency to cluster around the 3–5 values of the scale. No items got scores of 6 or 7 exhibiting a clearcut acceptability and this is probably due to the difficulty raised by the unexpected interpretation of a binding dependency between the two internal arguments.74 Note also that 3 was taken as a general, informal acceptability threshold in the sense that lower scores were deemed unacceptable for speakers. Informally, items getting scores of 3 and above were considered to fare as (rather) acceptable. As a result, the patterns DO before IO, DO binds into IO, clitic doubled IO and DO before IO, IO binds into DO, clitic doubled IO were identified as problematic, given the low scores below the acceptability threshold 3. On the other hand, items where the DO had been clitic doubled fared better than their counterparts in experiments 1 and 2, which made us think of DO CD as a saving mechanism75 that improved acceptability judgements. Considering acceptability intervals ranging below 3 as unacceptable (*), those ranging between 3 and 4 as rather acceptable (‒✔), those between 4 and

73

(Intercept) woIO before DO bindingIO binds into DO cliticwith CL woIO before DO:bindingIO binds into DO woIO before DO:cliticwith CL bindingIO binds into DO:cliticwith CL woIO before DO:bindingIO binds into DO:cliticwith CL

Estimate

Sth. Error t

t value

. −. −. −. . . . −.

. . . . . . . .

.* −.* −.* −.* .* .* .* −.*

* t-values greater than 2 and less than -2 are considered significant. 74 Luigi Rizzi (p.c.) discussing the general lower scores obtained on binding dependencies with ditransitives commented on the processing load for the speakers when it comes to decoding binding dependencies between the two internal arguments, especially since the usual expectation would be to have binding between the subject and the DO. Indeed, as a general observation, in the experiment checking binding dependencies between the Subject DP and the DO the general acceptability scores were much higher. 75 We will indeed see that according to the proposed account, the intervention effects uncovered for the problematic patterns do not arise in the items where the DO was clitic doubled.

185

6.2 Experimental data on Romanian ditransitives

5 as acceptable (✔) and those above five as quite acceptable (✔✔), the results in Table 12 above may be summarized as below:

Table 12: General results for all experiments. DO before IO DO binds into IO

IO before DO

IO binds into DO

DO binds into IO

IO binds into DO

−cl

+cl

−cl

+cl

−cl

+cl

−cl

+cl

Exp 

.

.

.

.

.

.

.

.

Exp 

.

.

.

.

.

.

.

.

Exp 

,

,

,

,

,

,

,

,

Table 13: General results for all experiments. DO before IO

IO before DO

DO binds into IO

IO binds into DO

DO binds into IO

IO binds into DO

−cl

+cl

−cl

+cl

−cl

+cl

−cl

+cl

Exp 



–✔

–✔

–✔

–✔

–✔





Exp 



*

–✔

*

–✔

–✔



–✔

Exp 



–✔

–✔

–✔

–✔

–✔

✔✔

–✔

The following sections elaborate on the results of each experiment in detail and also discuss the differences in the acceptability scores uncovered with the corresponding patterns between the three experiments. 6.2.1.2 Results 6.2.1.2.1 Experiment 1: unmarked DOs The first experiment focuses on ditransitives selecting inanimate DOs, which means that no effect of the differential marker pe might be expected here. Example (244) below contains a sample item with all its 8 tested variants:

186

6 Romanian ditransitive configurations

DO before IO; DO binds into IO; no clitic (244) a. Editorii au trimis fiecare cartei autorului eii Editors.DEF.ART. have sent every book author.DAT its pentru corecturile finale. for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent each book to its author for the final corrections.’ DO before IO; IO binds into DO; no clitic autori b. Editorii au trimis cartea sai fiecărui Editors.DEF.ART. have sent book.DEF.ART. his every.DAT. author pentru corecturile finale. for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent his book to every author for the final corrections.’ IO before DO; DO binds into IO; no clitic c. Editorii au trimis autorului eii fiecare cartei Editors.DEF.ART. have sent author.DAT. its every book pentru corecturile finale. for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent its author every book for the final corrections.’ IO before DO; IO binds into DO, no clitic sai d. Editorii au trimis fiecărui autori cartea Editors.DEF.ART. have sent every.DAT author book.DEF.ART. his pentru corecturile finale. for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent every author his book for the final corrections.’ DO before IO; DO binds into IO, + clitic e. Editorii i-au trimis fiecare cartei autorului eii Editors.DEF.ART. CL.3.DAT.SG.-have sent every book author.DAT its pentru corecturile finale. for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent each book to its author for the final corrections.’

6.2 Experimental data on Romanian ditransitives

187

DO before IO; IO binds into DO, + clitic f. Editorii i-au trimis cartea sai Editors.DEF.ART. CL.3.DAT.SG.-have sent book.DEF.ART. his fiecărui autori pentru corecturile finale. every.DAT. author for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent his book to every author for the final corrections.’ IO before DO; DO binds into IO, + clitic g. Editorii i-au trimis autorului eii Editors.DEF.ART. CL.3.DAT.SG.-have sent author.DAT. its finale. fiecare cartei pentru corecturile every book for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent its author every book for the final corrections.’ IO before DO; IO binds into DO, + clitic h. Editorii i-au trimis fiecărui autori Editors.DEF.ART. CL.3.DAT.SG.-have sent every.DAT author finale. cartea sai pentru corecturile book.DEF.ART. his for corrections.DEF.ART. final.FEM.PL. ‘The editors have sent every author his book for the final corrections.’ As a general observation, the general average acceptability scores ranged between 3 and 5 on the 7-rung acceptability scale.76 None of the binding dependencies is

76 As will be seen later, ditransitives containing unmarked DOs fare considerably better than their counterparts with DOMed DOs. This state of affairs might well be due to the absence of the differential object marker: the general improvement in acceptability might be due to the lack of the intervention effects between pe and the dative clitic in those cases where the IO has been clitic doubled. Also, the better scores of those items containing bare IOs and inanimate DOs as opposed to their counterparts with pe marked DOs point to a general degrading effect that a single pe marked DO seems to have on the sentence per se. Indeed, previous experiments in Avram (2014) show that single pe marked DOs are not readily accepted by all native speakers of Romanian and that more often than not Differential Object Marking is accompanied by Clitic Doubling. The two phenomena, which started out as independent mechanisms in Old Romanian (von Heusinger and Onea 2008, Stan 2009, Tigău 2010, 2016), seem to be so intertwined in modern Romanian, that few speakers still accept pe marking in the absence of CD. Avram (2014) even differentiates between two classes of speakers: one group always discard pe marked DOs in the absence of CD, while the other allow single pe marked instances and seem to be much more flexible in assessing the acceptability of items. In our first experiment, we also noticed that some of the speakers generally marked all items containing single pe marked DOs as unacceptable irrespective of the surface word order of the two arguments or the direction of binding.

188

6 Romanian ditransitive configurations

discarded as unacceptable so that ditransitives with inanimate DOs seem to evince bidirectional c-command77 (see Appendix 1 for statistics).

Table 14: GM ditransitives with inanimate DOs. DO before IO DO binds into IO −cl .

IO before DO

IO binds into DO

DO binds into IO

IO binds into DO

+cl

−cl

+cl

−cl

+cl

−cl

+cl

.

.

.

.

.

.

.

Consider also Graph (15) below, where the scores obtained for ditransitives may be compared to the ones assigned to the fillers. As it seems, items containing ditransitives fared generally better than the very bad fillers (GM: 2.13) and worse than the very good fillers (GM: 5.86).

Ditransitives and Fillers 7 6 5,86

5 4

4,57

4,56

3 3,08

4,35 3,71

3,64

3,40

3,00

3,22

2 2,13 Fillers (medium)

Fillers (v. good)

Fillers (v. bad)

with CL; IO before DO; IO binds into DO

with CL; IO before DO; DO binds into IO

with CL; DO before IO; IO binds into DO

with CL; DO before IO; DO binds into IO

No CL; IO before DO; IO binds into DO

No CL; IO before DO; DO binds into IO

No CL; DO before IO; IO binds into DO

no CL; DO before IO; DO binds into IO

1

Graph 15: GMs ditransitive configurations and fillers.

77 Note that the general lower acceptability scores obtained may be due to the difficulties of processing the binding dependencies that respondents encountered. Binding dependencies between the DO and the IO are rather unexpected, as respondents would actually expect binding with the subject (Luigi Rizzi p.c.).

6.2 Experimental data on Romanian ditransitives

189

In fact, the general acceptability scores obtained for the experimental items pattern with the ones obtained for those fillers assigned (informally, prior to the experiments) medium acceptability rates. Consequently, we assume that these structures were generally considered as acceptable by our respondents. 8

5.861742424

7 6 5

3.731710744

3.71969697

4

2.134469697

3 2 1 Ditransitives general

Fillers (v. bad)

Fillers (v. good)

Fillers (medium)

Graph 16: Ditransitives general vs. fillers.

Another important observation is that the surface word order of the two internal arguments had a significant impact on acceptability judgements: items where the surface word order matched the direction of binding generally obtained higher scores than instances where reverse binding was at stake:

7 6 5

4,45 4,10

4 3.04

3.31

3 2 1 DO binds into IO

IO binds into DO

DO before IO

DO binds into IO

IO binds into DO

IO before DO

Graph 17: Word order and binding (total).

These results hold irrespective of whether the IO was clitic doubled or not. This may be observed in Graph 18 below: configurations where the DO precedes the IO and binds into it have been rated as significantly more acceptable than their

190

6 Romanian ditransitive configurations

counterparts where the same word order is preserved but the binding direction is reversed (GM 4.57 vs. 3.08). Likewise, patterns where the IO precedes the DO were evaluated as more acceptable in those cases where the IO also binds the DO, as opposed to instances where the reversed binding relations obtain (GM 4.56 vs. 3.40). The same was observed with ditransitives containing a clitic doubled IO: DO before IO, DO binds into IO (+cl) vs. DO before IO, IO (+cl) binds into DO: 3.64 vs. 3.00; IO (+cl) before DO, IO (+cl) binds into DO vs. IO (+cl) before DO, DO binds into IO (+cl): 4.35 vs. 3.22.

7 6 5 4

4.57 3.64

3

3.08

3

no cl

with cl

3.4

3.22

no cl

with cl

4.56

4.35

no cl

with cl

2 1 no cl

with cl

DO binds into IO

IO binds into DO

DO binds into IO

DO before IO

IO binds into DO

IO before DO

Graph 18: Word order and binding (separate).

Finally, the presence of a dative clitic doubling the IO seems to trigger lower rates of acceptability rating. This state of affairs covers both surface word orders and directions of binding.

7 6 5 4

4.57

4.56 3.64 3.08

3.4

3

4.35

3.22

3 2 1

DO before IO; DO binds into IO

DO before IO; IO binds into DO Series1

IO before DO; DO binds into IO Series2

Graph 19: +cl vs. −cl; Series 1: -clitic; Series 2: +clitic.

IO before DO; IO binds into DO

6.2 Experimental data on Romanian ditransitives

191

The results of the first experiment seem to attest bidirectional c-command with the two internal arguments of ditransitives, just as it is argued for Spanish (Pineda 2012, 2013). As such, they discomfirm the existing analysis put forth in Diaconescu and Rivero (2007) which mainly builds on asymmetric binding dependencies between IO and DO.78 As already discussed above, Diaconescu and Rivero (2007) propose an alternative projection account for Romanian ditransitives arguing that the Theme-Goal construction should be associated with two different structures depending on whether the IO has been doubled by means of a dative clitic or not. Thus, ditransitives with single, undoubled IOs evince a configuration wherein the DO c-commands the IO, similarly to the Prepositional Dative Construction in English (245a). On the other hand, ditransitives containing a clitic doubled IO, are argued to rest on a configuration where the IO c-commands the DO and are assimilated to the English Double Object Construction (245b). In this configuration the Goal DP is no longer merged low within the VP but in an Applicative projection above it. (245) a. POC: Theme c-commands Goal VoiceP DPAgentVoice[vPv [PPDPTheme P DPGoal]]] b. DOC: Goal c-commands Theme (clitic doubling) [VoicePDPAgentVoice[vPv [ApplPDPGoal [clAppl] [VP V DPTheme]]]] (Diaconescu and Rivero 2007: 219–220) The two different hierarchical structures are thus triggered by the presence of a dative clitic or the lack thereof: in (a), the Goal DP is merged low within a PP, while the Theme DP occupies the specifier of this projection c-commanding the Goal DP. The Dative is inherently case licensed at merge. In (b), the Goal DP merges in the Specifier position of the ApplP, while the Theme DP is hosted below in a complement position. The Applicative head spells out as the dative clitic pronoun. Thus, DOC readings depend on CD given that the higher Goal is introduced by the clitic in the Appl head position. The same binding relations as in English are argued to obtain in Romanian, the account put forth by Diaconescu and Rivero (2007) giving rise to a number of predictions:79

78 Note also that Diaconescu and Rivero (2007) do not discuss surface word order but only focus on the direction of binding between the two arguments. Our discussion will, however, dwell on this parameter given its attested importance throughout the experiment. 79 Given that we did not test items containing clitic doubled DOs in this experiment, only the relevant predictions will be discussed.

192

6 Romanian ditransitive configurations

Prediction 1: IO cannot bind into DO Given that a bare IO is taken to merge low within the VP, in a position where it may be c-commanded by the DO, configurations where the IO binds into the DO are predicted by Diaconescu and Rivero (2007) to be ungrammatical. The experimental findings seem, however, to contradict these expectations as this binding relation turns out to be acceptable in both word orders, with a visible preference for instances where the word order matches the direction of binding: Table 15: Binding – undoubled IO binds into DO. Surface word order

Direction of Binding

DO before IO IO before DO

IO binds into DO IO binds into DO

Score . .

Prediction 2: DO binds into IO Given its low merge position, Diaconescu and Rivero (2007) posit that the undoubled IO may be bound by the DO. The experimental results show that this is indeed the case. Again, there is a marked preference for those instances where the word order matches the direction of binding.

Table 16: Binding – DO binds into undoubled IO. Surface word order

Direction of Binding

DO before IO IO before DO

DO binds into IO DO binds into IO

Score . .

The following two predictions concern patterns where the IO has been doubled by means of a dative clitic. As already discussed, these are cases where, according to Diaconescu and Rivero (2007), the IO is merged within an ApplP and thus finds itself in a position wherefrom it may c-command the DO. Prediction 3: IO(+cl) binds into DO For reasons extended upon above, Diaconescu and Rivero (2007) expect this pattern to always come out as grammatical. The results we obtained seem to

6.2 Experimental data on Romanian ditransitives

193

confirm these expectations for both word orders, with a visible preference for the situation where there is matching between the direction of binding and word order:

Table 17: Binding – doubled IO binds into DO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

IO(+cl) binds into DO IO(+cl) binds into DO

. .

Prediction 4: DO cannot bind into IO (+cl) According to the account proposed by Diaconescu and Rovero (2007), where the clitic doubled IO merges high within an ApplP, this pattern is predicted to come out as ugrammatical, given that the DO merged low within the VP may never c-command the IO. Contrary to these expectations, the binding of the clitic doubled IO seems to be possible, again with a slight preference for those instances where the word order matches the direction of binding:

Table 18: Binding – DO binds into doubled IO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

DO binds into IO (+cl) DO binds into IO (+cl)

. .

The predictions put forth by Diaconescu and Rovero (2007) are not in line with the experimental data presented here. Ditransitive configurations seem to exhibit bidirectional c-command, just as their counterparts in Spanish (Pineda 2012, 2013), French (Harley 2002) or Italian (Giorgi and Longobardi 1991). The results of the first experiment thus seem to disconfirm an analysis according to which Romanian ditransitives pattern with their English counterparts. The presence of the dative clitic doubling the IO does not seem to have structural import. In fact, there is no structural difference with respect to the c-command potential of the two internal arguments relative to whether the IO has been clitic doubled or not. The two arguments seem to retain the same binding potential both in ditransitive configurations containing undoubled IOs as well as in ditransitives with doubled IOs.

194

6 Romanian ditransitive configurations

The experimental findings uncovered in the first experiment, where inanimate (therefore unmarked) DOs were used, pattern with those obtained in a second experiment involving differentially object marked DOs, which will be dwelt on in the upcoming section. 6.2.1.2.2 Experiment 2: DOMed DOs The second experiment had a similar design to the first, the only difference being that the direct object DPs were all differentially object marked this time. (246) contains a sample item with all its 8 variants: DO before IO; DO binds into IO; no cl (246) a. Poliția a înapoiat pe fiecare copili pierdut pe Police.DEF.ART. has returned DOM every child lost on plajă părinților luii diperați. beach parents.DAT his desperate.PL ‘The police returned every child lost on the beach to his desperate parents.’ DO before IO; IO binds into DO; no cl b. Poliția a înapoiat pe copilul săui pierdut Police.DEF.ART. has returned DOM child.DEF.ART. his lost pe plajă fiecărui părintei, după îndelungi căutări. on beach every.DAT parent after long search.PL ‘The police returned his child lost on the beach to every parent, after a long search.’ IO before DO; DO binds into IO; no cl c. Poliția a înapoiat părinților luii disperați Police.DEF.ART. has returned parents.DAT his desperate.M.PL pe fiecare copili pierdut pe plajă. DOM every child lost on beach ‘The police has returned to his desperate parents every child lost on the beach.’

6.2 Experimental data on Romanian ditransitives

195

IO before DO; IO binds into DO; no cl d.

Poliția a înapoiat fiecărui părintei pe Police.DEF.ART. has returned every.DAT parent DOM copilul săui pierdut pe plajă. child.DEF.ART. his lost on beach ‘The police returned to every parent his child lost on the beach.’

DO before IO; DO binds into IO; +cl e. Poliția le-a înapoiat pe fiecare copili Police.DEF.ART. CL.3.DAT.PL-has returned DOM every child pierdut pe plajă părinților luii diperați. lost on beach parents.DAT his desperate.PL ‘The police returned every child lost on the beach to his desperate parents.’ DO before IO; IO binds into DO; +cl f. Poliția i-a înapoiat pe copilul Police.DEF.ART. CL.3.DAT.SG-has returned DOM child.DEF.ART. plajă fiecărui părintei, după îndelungi săui pierdut pe his lost DOM beach every.DAT parent after long căutări. search.PL ‘The police returned his child lost on the beach to every parent, after a long search.’ IO before DO; DO binds into IO; +cl g. Poliția le-a înapoiat părinților luii Police.DEF.ART. CL.3.DAT.PL-has returned parents.DAT his disperați pe fiecare copili pierdut pe plajă. desperate.M.PL DOM every child lost on beach ‘The police has returned to his desperate parents every child lost on the beach.’

196

6 Romanian ditransitive configurations

IO before DO; IO binds into DO; +cl h.

Poliția i-a înapoiat fiecărui părintei Police.DEF.ART. CL.3.DAT.PL-has returned every.DAT parent pe copilul săui pierdut pe plajă. DOM child.DEF.ART. his lost on beach ‘The police returned to every parent his child lost on the beach.’

The first thing to note is the generally lower acceptability scores assigned to all items in this experiment as compared to their counterparts from the first experiment. The test items containing DOMed DOs fared worse than the corresponding items with unmarked DOs irrespective of word order, direction of binding or the presence/absence of a dative clitic doubling the IO. Table 19 below allows us to compare the differences with respect to the acceptability scores between the first two experiments:

Table 19: GMs Experriment 1 and 2. DO before IO DO binds into IO

Exp  Exp 

IO before DO

IO binds into DO

DO binds into IO

IO binds into DO

−cl

+cl

−cl

+cl

−cl

+cl

−cl

+cl

. .

. .

. .

. .

. .

. .

. .

. .

Just as with the items containing unmarked DOs, the acceptability scores range between 3 and 5, which points to the fact that these sentences were generally found acceptable. We may consequently hypothesize that ditransitives with DOMed DOs exhibit bidirectional c-command, just like their counterparts containing unmarked DOs. Note, however, that there are two cases where the scores are very low, approaching the values assigned to those fillers expected to fare very badly (these fillers had been previously tested for acceptability in an informal experiment): both situations concern the DO before IO word order and the items in question contain clitic doubled IOs. Thus, as the table below shows, the scores assigned to these particular instances range below 3 (i.e., 2,64 and 2,36) and very close to the value assigned to the very bad fillers (2,15). The scores for the condition in which the binding direction matches word order are slightly improved as expected, given the general trend already observed in the first experiment:

197

6.2 Experimental data on Romanian ditransitives

Items and Fillers 7 6 5 4 3 2 1

5.66 4.43

4.32 3.051

2.64

3.27

3.68

3.53

3.36

2.36

2.15

DO binds IO binds DO binds IO(+cl) DO binds IO(+cl) DO binds IO binds very bad medium very good into IO into DO into binds into into binds into into IO into DO IO(+cl) DO IO(+cl) DO DO before IO

DO before IO(+cl)

IO (+cl) before DO

IO before DO

Fillers

Graph 20: Ditransitives and fillers.

As already noticed in the first experiment, there is a marked tendency to assess those instances where the word order matches the binding direction as more acceptable than their counterparts with the reversed binding dependence. As may be seen in Graph 21 below, structures where the DO precedes the IO and binds into it are strikingly assessed as more acceptable than their counterparts where the DO precedes the IO but the latter binds into the former (GM 3.53 vs. 2.705). Similarly, structures where the IO precedes the DO were evaluated as more acceptable in those cases where the IO also bound the DO, as opposed to instances where the reversed binding relations obtained (GM 4.00 vs. 3.315).

7 6 5 4 3

4

3.535

3.315 2.705

2 1 DO binds into IO

IO binds into DO DO before IO

DO binds into IO

IO binds into DO IO before DO

Graph 21: word order and binding (general).

The same tendency is generally observed if we separate the items with clitic doubled IOs from the items with undoubled IOs: the examples where the surface word order matches the direction of binding fare better than those with the reverse binding direction:

198

6 Romanian ditransitive configurations

7 6 5

4.43

4.32

4 3.05 2.64

3

3.68

3.36

3.27

no cl

with cl

2.36

2 1 no cl

with cl

DO binds into IO

no cl

with cl

IO binds into DO

DO binds into IO

DO before IO

no cl

with cl

IO binds into DO

IO before DO

Graph 22: Word order and binding (separate).

Finally, instances where a dative clitic doubles the IO are assigned lower acceptability scores than their counterparts containing undoubled IOs. This state of affairs carries over to both surface word orders as well as directions of binding. The results pattern with the ones discussed for experiment 1: 7 6 5 4

4.43

4.32

3 2

2.64

3.36 3.27

3.05

3.68

2.36

1 DO before IO; DO binds into IO

DO before IO; IO binds into DO Series2

IO before DO; DO binds into IO

IO before DO; IO binds into DO

Series3

Graph 23: +cl vs. −cl; Series 1: −clitic; Series 2: +clitic.

As in the first experiment, the results obtained with ditransitive configurations containing DOMed DOs attest the presence of bidirectional c-command with the two internal arguments of ditransitives. As such, they disconfirm the existing analysis put forth in Diaconescu and Rivero (2007) which mainly builds on asymmetric binding dependencies between the IO and DO. One difference between the two experiments has to do with the problematic cases where the DO

6.2 Experimental data on Romanian ditransitives

199

precedes a clitic doubled IO, which came out as severely degraded in both directions of binding. This result is all the more interesting in that, as will be seen, experiment 2 is the only experiment where these sequences turn out to be problematic. Ditransitives with unmarked DOs and CDed+DOMed DOs fare similarly with respect to these particular sequences (DO before IO, +cl). Furthermore, these instances do not seem to be special in any respect in the two other experiments when comparing their acceptability scores to the scores assigned to their counterparts with different word orders and binding directions. These results will be accounted for in a subsequent section dealing with the syntactic analysis of ditransitives. Before we get to that point, however, consider that, just like in the case of the first experiment, the results obtained with ditransitives containing DOMed DOs disconfirm the predictions put forth in Diaconescu and Rivero (2007). Remember that the expectation in Diaconescu and Rivero (2007) is that the two internal arguments exhibit asymmetric binding potential, which differs function of whether the IO has been clitic doubled or not i.e., an undoubled IO would never bind into a DO (given its low merge position), while its doubled counterpart will always do so (given that it merges inside the ApplP and thereby c-commands the DO). Consider how this analysis fares against the experimental findings: Prediction 1: IO cannot bind into DO According to Diaconescu and Rivero (2007), an undoubled IO is merged low within the VP, in a position where it is c-commanded by the DO. Consequently, configurations where the bare IO binds into the DO are rejected as ungrammatical. Our experimental findings disconfirm this prediction: both word orders are found acceptable by the respondents, with a clearcut preference for those cases where word order matches the direction of binding. Table 20: Binding – undoubled IO binds into DO. Surface word order

Direction of Binding

DO before IO IO before DO

IO binds into DO IO binds into DO

Score . .

Prediction 2: DO binds into IO Given that a bare IO is posited to merge low within the VP, structures where the DO binds into the IO are predicted to come out as grammatical by Diaconescu

200

6 Romanian ditransitive configurations

and Rivero (2007). Their expectations are confirmed by our experimental findings for both word orders. As expected, cases where the surface word order matches the binding direction emerged as significantly better (GM 4.43) than cases where inverse binding obtained (GM 3.36). Table 21: Binding – DO binds into undoubled IO. Surface word order

Direction of Binding

DO before IO IO before DO

DO binds into IO DO binds into IO

Score . .

Let us now focus on those cases where the IO has been doubled by a dative clitic. Remember that according to Diaconescu and Rivero (2007), such structures are deemed to be different from those configurations containing an undoubled IO, evincing inverse binding relations. Prediction 3: IO(+cl) binds into DO According to the underlying configuration in (245b) proposed by Diaconescu and Rivero (2007), a clitic doubled IO may always bind into a DO due to its merge position within the ApplicativeP, which allows it to c-command the DO. These expectations are, however, only met for the IO before DO word order (GM 3,68), but not for the opposite one (GM 2,36). Since this particular result differs from the one obtained with the counterpart situation involving unmarked DOs (in experiment 1) and, as will be seen from the one involving CDed+DOMed DOs (in experiment 3), we are led to believe that the low acceptability of these instances has to do with the internal structure of the DO itself and not with a configurational difference as will be shown. Table 22: Binding – doubled IO binds into DO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

IO(+cl) binds into DO IO(+cl) binds into DO

. .

6.2 Experimental data on Romanian ditransitives

201

Prediction 4: DO cannot bind into IO (+cl) As already pointed out above, a clitic doubled IO is posited to merge higher than a DO, within an ApplP wherefrom it is able to bind this DO. Consequently, Diaconescu and Rivero (2007) claim that the configuration DO binds into IO (+cl) is ungrammatical. Our experimental results seem to confirm their prediction for the DO before IO word order only. Again, when comparing this situation with its correspondents in Experiments 1 and 3, where these instances do not constitute a special case regarding acceptability judgements, one will be inclined to discard the hypothesis that a peculiar configuration would trigger the low scores here, but to consider that the low acceptability scores might have to do with the internal make-up of the DO. Indeed, as will be seen, the poor acceptability results obtain as a consequence of some intervention effects resulting from the competition between the DO and the IO for the same Applicative head as a consequence of having a similar featural load. Table 23: Binding – DO binds into doubled IO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

DO binds into IO (+cl) DO binds into IO (+cl)

. .

As it seems, the results obtained for DOMed DOs generally pattern with those obtained with ditransitives containing unmarked DOs: there is a general tendency to mark instances where the word order matches the direction of binding as more acceptable than their counterparts with reverse binding direction. The presence of a dative clitic doubling the IO lowers acceptability scores for instances in both experiments, irrespective of word order and binding direction. One important difference identified with respect to the second experiment has to do with the very low acceptability rates obtained for the DO before IO word order when the IO is clitic doubled. Experiment 2 is, however, the only one out of the three experiments where this problematic situation arises, which prompted us to hypothesize that the results have nothing to do with a special configuration, but rather with the internal structure of the argument DPs involved. In particular, we tentatively advanced the idea that the featural makeup of the IO and the DO is at the bottom of the low acceptability judgements obtained for these particular instances. Further discussion will be postponed

202

6 Romanian ditransitive configurations

for the section tackling the syntactic analysis of ditransitives. In the next section, we dwell on the results of the third experiment. 6.2.1.2.3 Experiment 3: CDed and DOMed DOs The third experiment was similar in design to the first two, the only difference being that the ditransitives employed contained CDed and DOMed DOs. (247) below contains a sample item with all its 8 variants: DO before IO; DO binds into IO; no clitic (247) a. Delegații au lăudat-o pe fiecare Delegates.DEF.ART. have praised-CL.3.ACC.FEM.SG. DOM every eii direct. secretarăi șefului secretary boss.DAT her direct ‘The delegates praised every secretary to her direct superior.’ DO before IO; IO binds into DO; no clitic b. Delegații au lăudat-o pe Delegates.DEF.ART. have praised-CL.3.ACC.FEM.SG. DOM șefi de departament. secretara luii fiecărui secretary.DEF.ART. his every.DAT chief of department ‘Lit. The delegates praised his secretaryi to every head of the departmenti.’ IO before DO; DO binds into IO; no clitic c. Delegații au lăudat-o șefului eii Delegates.DEF.ART. have praised-CL.3.ACC.FEM.SG. boss.DAT her direct pe fiecare secretarăi din departament. direct DOM every secretary from department ‘The delegates praised every secretary in the department to her direct superior.’ IO before DO; IO binds into DO; no clitic d. Delegații au lăudat-o fiecărui Delegates.DEF.ART. have praised-CL.3.ACC.FEM.SG. every.DAT secretara luii. șefi de departament pe chief of department DOM secretary.DEF.ART. his ‘Lit. The delegates praised to every head of the department his secretary.’

6.2 Experimental data on Romanian ditransitives

203

DO before IO; DO binds into IO; with clitic e. Delegații i-au lăudat-o pe Delegates.DEF.ART. CL.3.DAT.SG-have praised-CL.3.ACC.FEM.SG. DOM eii direct. fiecare secretarăi șefului every secretary boss.DAT her direct ‘The delegates praised every secretary to her direct superior.’ DO before IO; IO binds into DO; with clitic f. Delegații i-au lăudat-o pe Delegates.DEF.ART. CL.3.DAT.SG-have praised-CL.3.ACC.FEM.SG. DOM șefi de departament. secretara luii fiecărui secretary.DEF.ART. his every.DAT chief of department ‘Lit. The delegates praised his secretaryi to every head of the departmenti.’ IO before DO; DO binds into IO; with clitic g. Delegații i-au lăudat-o Delegates.DEF.ART. CL.3.DAT.SG-have praised-CL.3.ACC.FEM.SG. fiecare secretarăi din departament. șefului eii direct pe boss.DAT her direct DOM every secretary from department ‘The delegates praised every secretary in the department to her direct superior.’ IO before DO; IO binds into DO; with clitic h. Delegații i-au lăudat-o Delegates.DEF.ART. CL.3.DAT.SG-have praised-CL.3.ACC.FEM.SG. secretara luii. fiecărui șefi de departament pe every.DAT chief of department DOM secretary.DEF.ART. his ‘Lit. The delegates praised to every head of the department his secretary.’ Similarly to the first two experiments, the general average acceptability scores cluster around the mid area of the 7-rung scale. The respondents, however, found the sentences with CDed+DOMed DOs generally better than those containing unmarked or DOMed DOs.80 One instance i.e., IO before DO, IO binds into DO, no clitic even receives acceptability scores surpassing 5, with a GM of 5,52.

80 Only one sentence receives lower scores for the CDed+DOMed variant than for the DOMed and the unmarked one i.e., IO before DO, IO binds into DO, with clitic: unmarked DO: 4,35/ DOMed DO 3,68/ CDed+DOMed DO 3,51. And there are two other cases where the CDed+DOMed

204

6 Romanian ditransitive configurations

Table 24: GM ditransitives. DO before IO DO binds into IO

Exp  Exp  Exp 

IO before DO

IO binds into DO

DO binds into IO

IO binds into DO

−cl

+cl

−cl

+cl

−cl

+cl

−cl

+cl

. . ,

. . ,

. . ,

. . ,

. . ,

. . ,

. . ,

. . ,

Consider also Graph 24 below comparing the ratings for the three experiments: 7 5.52

6 5

4.57 4.51 4.43

4 3

4.56 4.32 3.64 3.52 2.64

3.47 3.08 3.05

3 3.18 2.36

with cl

no cl

with cl

3.67 3.43.36

3.42 3.27 3.22

no cl

with cl

4.35 3.68 3.51

2 1 no cl

DO binds into IO

IO binds into DO

DO before IO

DO before IO Exp. 1

Exp. 2

no cl

with cl

DO binds into IO

IO binds into DO

IO before DO

IO before DO

Exp. 3

Graph 24: General results with the three experiments.

As Table 23 shows, none of the binding dependencies is discarded as unacceptable, which further strengthens the hypothesis that Romanian ditransitives allow symmetric c-command. Graph 11 below compares the acceptability values assigned to ditransitives containing CDed+DOMed DOs with those assigned to the fillers (again organised into very bad/very good/medium as they turned out in an informal

variant fares slightly worse than its unmarked counterparts (i.e., DO before IO, DO binds into IO no clitic and the same configuration containing a clitic doubled IO). A comparison between the variants comparing unmarked DOs with those containing CDed+DOMed DOs is not entirely possible, given that the lexicalisations per se were completely different and the acceptability judgements may have been influenced by different factors. A relevant comparison would be the one between the variants containing DOMed DOs and variants with CDed+DOMed DOs, since the same lexicalisations were at stake and the only difference concerned the doubling of the DO. In this respect the CDed+DOMed variants are clearly more acceptable to the respondents.

205

6.2 Experimental data on Romanian ditransitives

experiment preceding the three experiments proper): the acceptability scores obtained for the ditransitive configurations in general are significantly higher than those assigned to the very bad fillers, and range on the same level as the scores for the medium fillers.

7 6 5.83 5 4 3.85

3.62

3 2 1.95 1 Diransitives general

Fillers; very bad

Fillers; medium

Fillers; very good

Graph 25: Items (general) and fillers.

The same holds for each of the instances separately: they all receive acceptability scores which are significantly higher than those assigned to the very bad fillers and range closely to the scores granted to the medium fillers:

7 6 5 4.51

3

3.47

3.52

2

3.18

3.67

3.42

3.62

3.51

Graph 26: items (separate) and fillers.

Fillers; very good

Fillers; medium

Fillers; very bad

IO before DO; IO binds into DO; with cl

IO before DO; DO binds into IO; with cl

IO before DO; IO binds into DO; no cl

IO before DO; DO binds into IO; no cl

DO before IO; DO binds into IO; with cl

DO before IO; IO binds into DO; no cl

1.95 DO before IO; DO binds into IO; no cl

1

5.83

5.52

DO before IO; IO binds into DO; with cl

4

206

6 Romanian ditransitive configurations

Another similarity with the first two experiments is the importance of the surface word order of the two internal arguments: test items where the direction of binding and the word order concurred received higher acceptability ratings than instances with reverse binding: 7 6 5

4.515 4.015

4

3.545

3.325

3 2 1 DO binds into IO

IO binds into DO

DO binds into IO

DO before IO

IO binds into DO IO before DO

Graph 27: Items and word order (general).

These results hold irrespective of whether the IO has been clitic doubled or not. As may be seen in Graph 28 below, the configurations where the DO preceedes the IO and binds it receive higher scores than their counterparts with the reverse binding directions (GMs: 4,51 vs. 3,47 / 3,52 vs. 3,18). Similarly, the patterns where the IO precedes the DO and binds it were assessed as much more acceptable than instances where the IO precedes the DO but the DO binds into it (GMs: 5,52 vs. 3,67/ 3,51 vs. 3,42): 7 6 5

5.52 4.51

4

3.52

3.47

with CL

no CL

3.67 3.18

3.51

3.42

3 2 1 no CL

DO binds into IO

with CL

IO binds into DO

DO before IO

Graph 28: Items and word order (separate).

no CL

with CL

DO binds into IO

no CL

with CL

IO binds into DO

IO before DO

6.2 Experimental data on Romanian ditransitives

207

Another similarity to the first two experiments has to do with the general lower acceptability of the sentences with clitic doubled IOs when compared to their undoubled IO counterparts: the presence of the dative clitic seems to trigger a general drop in the acceptability rates for both word orders and directions of binding:

7 6 5

5.52 4.51

4

3.52

3.47

with CL

no CL

3.67 3.18

3.51

3.42

3 2 1 no CL

with CL

no CL

with CL

no CL

with CL

DO before IO; DO binds into IO DO before IO; IO binds into DO IO before DO; DO binds into IO IO before DO; IO binds into DO

Graph 29: +cl vs. −cl.

Finally, as in the first two experiments, the results for ditransitive configurations containing CDed+DOMed DOs disconfirm the expectations of an alternative projection account such as the one put forth by Diaconescu and Rivero (2007). One important observation in this respect is that both directions of binding seem to be allowed for all word orders and irrespective of whether the IO has been clitic doubled or not. The attested symmetric binding with ditransitives, however, undermines an analysis according to which the dative clitic has structural import, and which distinguishes between two structures evincing different c-command relations between the two internal arguments. The predictions put forth in Diaconescu and Rivero (2007) are thus disconfirmed: Prediction 1: IO may not bind into DO Diaconescu and Rivero (2007) do not discuss the case of clitic doubled DOs in detail, but their analysis predicts that an undoubled IO, which is merged low within the VP, should not be able to bind into a CDed+DOM DO, just as it is not able to do so with unmarked DOs. The experimental findings seem to contradict

208

6 Romanian ditransitive configurations

this prediction as this binding dependency turns out to be quite acceptable in both word orders. As expected, there is a clearcut preference for those cases where the word order and the direction of binding concur. Table 25: Binding – undoubled IO binds into DO. Surface word order

Direction of Binding

DO before IO IO before DO

IO binds into DO IO binds into DO

Score . .

Prediction 2: DO binds into IO Since unmarked IOs merge below DOs, we expect in line with Diaconescu and Rivero (2007), that the DO should be able to bind into the IO. This is indeed the case, as shown in the table below. Again, there is a marked preference for those situations where the word order and the binding direction concur. Table 26: Binding – DO binds into undoubled IO. Surface word order

Direction of Binding

DO before IO IO before DO

DO binds into IO DO binds into IO

Score . .

Unlike bare IOs, their doubled counterparts are posited to merge higher than the DO, within an ApplP. As a consequence, the doubled IO is expected to always be able to bind into DO, given that it c-commands it. Prediction 3: IO (+cl) binds into DO The experimental findings confirm this prediction for both word orders with a slight preference for the situation where the direction of binding and the word order match: Table 27: Binding – doubled IO binds into DO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

IO(+cl) binds into DO IO(+cl) binds into DO

. .

6.2 Experimental data on Romanian ditransitives

209

Prediction 4: DO cannot bind into IO (+cl) Contrary to the expectations in Diaconescu and Rivero (2007), this binding dependency turns out to be acceptable for both word orders. In fact, the acceptability ratings range along the same values as those assigned to the opposite binding relation (predicted to be acceptable by Diaconescu and Rivero (2007), see prediction 3). The similar scores obtained for both directions of binding strengthen the hypothesis according to which ditransitives exhibit bidirectional c-command. Table 28: Binding – DO binds into doubled IO. Surface word order

Direction of Binding

Score

DO before IO (+cl) IO (+cl) before DO

DO binds into IO (+cl) DO binds into IO (+cl)

. .

The findings obtained in the third experiment disconfirm once more the predictions put forth by Diaconescu and Rivero (2007). Ditransitive configurations appear to allow bidirectional c-command, just as their counterparts in Spanish (Pineda 2012, 2013), French (Harley 2002) or Italian (Giorgi and Longobardi 1991). The results go against an analysis positing that Romanian ditransitives pattern with their English counterparts. The presence of the dative clitic doubling the IO does not seem to have structural import in this respect. 6.2.1.3 Conclusions and further steps In this section, attention was given to data not discussed for Romanian so far, namely the results of three experiments on ditransitives focusing on the binding potential of the two internal arguments. The three experiments were similar in design and checked the acceptability of sentences wherein we varied: a) the surface word order; b) the direction of binding c) the presence of a dative clitic doubling the IO. What distinguished between the three experiments was the type of DO employed: unmarked DO vs. DOMed DO vs. CDed+DOMed DOs. The following results obtained and need to be accounted for: Firstly, almost all acceptability ratings generally target the middle of the 7-rung scale, with most values concentrating around the 3–4 acceptability interval. The only exception in this respect is the case of DOMed DOs interacting with clitic doubled IOs in the DO before IO word order in both directions of

210

6 Romanian ditransitive configurations

binding, which get low scores of 2,64 GM and 2,36 GM respectively. Even so, all the sentences tested fare seemingly better than the ungrammatical fillers so they were deemed acceptable. Along the same lines, the acceptability scores show that all binding dependencies should be derivable. The DO and the IO thus show symmetrical binding abilities in that each of them can mutually bind a possessor, irrespective of word order or doubling. These results are in line with what has also been argued for other ditransitives in Spanish, French or Italian. Secondly, the surface word order seems to be crucial in determining the acceptability of binding relations: cases where the surface word order matched the direction of binding fare much better with respect to acceptability than those where inverse binding obtains. Thirdly, the presence of a dative clitic seems to have an impact on the overall acceptability of the tested items: the sentences where the IO is clitic doubled fare worse than their undoubled counterparts irrespective of the direction of binding and word order. Fourthly, the acceptability values assigned to test items containing DOMed DOs are generally lower than those given for ditransitives containing either unmarked DOs or CDed+DOMed DOs. One explanation that might be considered has to do with the general acceptability of sentences containing DOMed DOs which have not been clitic doubled. As it seems, the phenomenon of Differential Object Marking in present Romanian has reached a point in which the majority of speakers feel the need to always accompany it by CD. (Dobrovie- Sorin 1994, Tigău 2010, Avram 2014 a.o.); so much so that single DOMed DOs are judged by some speakers as infelicitous. Nevertheless, if this were the case, then all lexicalisations in Experiment 2 should receive very low scores given that only DOMed DOs are employed. As seen above, this is not the case: some instances containing undoubled DOMed DOs even pass the acceptability threshold 4 on the 7-rung acceptability scale. Another important observation had to do with the even lower acceptability of those patterns where the differential marker interacted with the dative clitic. This effect did not obtain with the counterparts of these patterns in Experiment 2, which further demonstrates that the lower acceptability has to do with the internal structure of the DPs involved. In this context, accusative CD seems to function as a repair strategy: configurations containing DOMed DOs and CDed IOs fare significantly better with respect to acceptability than their counterparts where the DOMed DO has not been doubled. Finally, the experimental data presented undermine an alternative projection account in line with Diaconescu and Rivero (2007), where Romanian ditransitives are argued to pattern with their English counterparts and evince two different

6.2 Experimental data on Romanian ditransitives

211

configurations where the two internal arguments have asymmetric c-command potential i.e., a POC-like configuration for those ditransitive constructions containing an undoubled IO and a DOC-like one for ditransitives with clitic doubled IOs. The experimental findings convincingly indicate that ditransitive configurations are based on a single underlying configuration irrespective of whether the IO has been doubled by means of a dative clitic or not, contra Diaconescu and Rivero (2007). An analysis along these lines will be proposed in the subsequent sections. Before we actually get to that point, however, let us consider some similar results described in Cornilescu, Dinu and Tigău (2017a) and another experiment we ourselves conducted on ditranstives testing binding dependencies between the two internal arguments with anaphors. As will be seen, the two internal arguments of ditransitives allow bidirectional binding with anaphors, just as in the case of the binding dependencies with possessives.

6.2.2 Quantificational dependencies with anaphors 6.2.2.1 Cornilescu et al. (2017a) As part of their experiment based on binding dependencies between IO and DO within ditransitives, Cornilescu et al. (2017a) include a number of 4 items where binding with anaphors is also assessed. They only check the IO before DO, IO binds into DO, no clitic pattern as in (248) below: (248) Context: Ion este un tip foarte glumeț care mereu joacă feste colegilor de birou. ’John is a very funny guy who always plays pranks on his colleagues at the office’ Ieri, de pildă, Ion a înfățișat-o Dianei Yesterday for instance John has shown-CL.3.ACC.FEM.SG Diana.DAT pe nimeni altcineva decât pe ea însăși și s-a DOM nobody else than DOM her herself and REFL.-has amuzat teribil când aceasta nu s-a prins despre amused terribly when this.FEM not REFL.-has realised about cine era vorba who was discussion ‘Yesterday, for instance, John showed Diana nobody else but herself and was terribly amused when she did not realise who he was talking about.’ This binding dependency would be problematic for Diaconescu and Rivero (2007), due to the fact that the IO is posited to merge low, below DO, in a

212

6 Romanian ditransitive configurations

position wherefrom it may not c-command the DO.81The experimental findings presented in Cornilescu et al. (2017a) show, however, that all tested items of this type are assessed as acceptable. Note that Cornilescu et al. (2017a) use a four – rung acceptability scale, ranging from -I (completely unacceptable), +/−I (rather unacceptable), +/− A (rather acceptable), to +A (acceptable).

20 25 30 10

15

A ±A ±I I

0

5

Frequency

Acceptability frequencies for bindig with anaphora

Lex1

Lex2

Lex3

Lex4

Graph 30: Cornilescu et al. (2017a): Figure 12, p. 13.82

The data regarding binding dependencies with anaphors thus seem to corroborate with the findings concerning binding with possessives and disconfirm the account put forth in Diaconescu and Rivero (2007). 6.2.2.2 A further experiment on anaphor binding 6.2.2.2.1 Design In order to gather more evidence supporting the hypothesis that the two internal arguments in ditransitives actually evince symmetric c-command potential, we

81 Neither Cornilescu et al. (2017a), nor Diaconescu and Rivero (2007) discuss the word order of the two internal arguments within ditransitives. 82 The abbreviated term lex in Graph 17 stands for lexicalisation and refers to the experimental item(s) employed in the experiment.

6.2 Experimental data on Romanian ditransitives

213

extended the experiment initiated in Cornilescu et al (2017a) on the model of the three experiments revolving around binding dependencies with possessives above. Thus, 8 lexicalisations were constructed and varied according to the following principles: surface word order, direction of binding and the presence of a dative clitic doubling the IO. One sample item is listed in (249) below, with all its tested variants: (249) Context: Conducerea a propus demiterea celor doi angajați care nu depuseseră declarațiile fiscale la timp, cauzând prin aceasta mari prejudicii companiei. ‘The board proposed firing the two employees who had not submitted the fiscal declarations on time, thus causing serious prejudice to the company.’ DO before IO, DO binds into IO, no cl a. Când, însă am înfățișat-o pe When however have.1.SG. shown-CL.3.ACC.FEM.SG. DOM chiar ei înseșii făcând aceleași greșeli, directoarei manager.FEM even her.DAT herself making same.PL mistakes s-a răzgândit în privința lor. REFL.-has changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ DO before IO, DO binds into IO, with clitic b. Când, însă i-am înfățișat-o When however CL.3.DAT.SG.-have.1.SG. shown-CL.3.ACC.FEM.SG. chiar ei înseșii făcând aceleași pe directoarei DOM manager.FEM even her.DAT herself making same.PL greșeli, s-a răzgândit în privința lor mistakes REFL.-has changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ DO before IO, IO binds into DO, no clitic c. Când, însă am înfățișat-o chiar When however have.1.SG. shown-CL.3.ACC.FEM.SG. even făcând aceleași greșeli, ea însășii directoareii her.ACC herself manager.DAT. making same.PL mistakes

pe DOM s-a REFL.-has

214

6 Romanian ditransitive configurations

răzgândit în privința lor changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ DO before IO, IO binds into DO, with clitic d. Când, însă i-am înfățișat-o When however CL.3.DAT.SG.-have.1.SG. shown-CL.3.ACC.FEM.SG. făcând aceleași chiar pe ea însășii directoareii even DOM her.ACC herself manager.DAT. making same.PL greșeli, s-a răzgândit în privința lor mistakes REFL.-has changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ IO before DO, DO binds into IO, no clitic e. Când, însă am înfățișat-o chiar ei When however have.1.SG. shown-CL.3.ACC.FEM.SG. even her.DAT directoarei făcând aceleași greșeli, s-a înseșii pe herself DOM manager.FEM making same.PL mistakes REFL.-has răzgândit în privința lor. changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ IO before DO, DO binds into IO, with clitic f. Când, însă i-am înfățișat-o When however CL.3.DAT.SG.-have.1.SG. shown-CL.3.ACC.FEM.SG. directoarei făcând aceleași chiar ei înseșii pe even her.DAT herself DOM manager.FEM making same.PL greșeli, s-a răzgândit în privința lor. mistakes REFL.-has changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ IO before DO, IO binds into DO, no clitic g. Când, însă am înfățișat-o directoareii When however have.1.SG. shown-CL.3.ACC.FEM.SG. manager.DAT

6.2 Experimental data on Romanian ditransitives

215

chiar pe ea însășii făcând aceleași greșeli, s-a even DOM her.ACC herself making same.PL mistakes REFL.-has răzgândit în privința lor. changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ IO before DO, IO binds into DO, with clitic h. Când, însă i-am înfățișat-o When however CL.3.DAT.SG.-have.1.SG. shown-CL.3.ACC.FEM.SG. ea însășii făcând aceleași directoareii chiar pe manager.DAT even DOM her.ACC herself making same.PL greșeli, s-a răzgândit în privința lor. mistakes REFL.-has changed her mind about them ‘When, however, I described the manager to herself making the same mistakes, she changed her mind about them.’ 64 items thus resulted, which were distributed into 4 lists using the Latin square method for an even distribution. Each list was thus comprised of 18 test items in need of testing. To these, we added 22 fillers. Each list of items had two variants: one meant to serve for a norming experiment and to check whether the respondents understood the binding dependencies holding between the two internal arguments; and another one aiming at measuring acceptability rates with the items. In the norming experiment, each item was followed by relevant questions to which multiple answers were possible. (250) contains the questions and the format corresponding to the examples under (249) above. Variant (i.) probes for the reading according to which the reflexive in one of the internal arguments refers to the referent denoted by the other, whereas (ii.) probes for the reading according to which the reflexive in one of the two internal arguments refers to an antecedent other than the one contributed by the other internal argument; the antecedent itself may not surface within the sentence per se but be part of a larger context. Finally, variant (iii.) checks whether both readings suggested in (i.) and (ii.) are possible. The order of the test items was randomized. Each list was assessed by more than 20 native speakers of Romanian, and more than 80 people took part in the norming experiment alone:

216

6 Romanian ditransitive configurations

(250) From this sentence I understand the following: i. I described the manager to herself. ii. I described the manager to a certain person, who has not been mentioned in the text but to which the pronoun her refers. iii. (i) and (ii)

Possible Impossible Possible Impossible

Possible Impossible

The norming experiment was followed by another experiment probing for acceptability. The items were assessed on a seven-rung acceptability scale, with 1 representing the lowest score and 7, the highest. As with its norming counterpart, the lists in this experiment were each assessed by at least 20 respondents. 6.2.2.2.2 Results and discussion The norming experiment revealed that almost all respondents interpreted the items as evincing a binding relation between the two internal arguments irrespective of the relative word order holding between DO and IO, the binding direction and the presence of a dative clitic doubling IO. As may be seen in graph (31), more than 80% of our respondents interpreted the anaphor present in one of the arguments as being bound by the DP within the other argument.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% bare IO

CDed IO

bare IO

DO binds IO

CDed IO

IO binds DO DO before IO

bare IO

CDed IO

bare IO

DO binds IO

CDed IO

IO binds DO IO before DO

Graph 31: Norming with anaphor binding.

The acceptability experiment revealed that the respondents consider the assessed items to range quite high on the acceptability scale: as may be seen from table (29), the acceptability scores range between 4,61 GM and 5,52 GM for each tested pattern.

217

6.2 Experimental data on Romanian ditransitives

Table 29: Acceptability values with anaphor binding. word order

DO before IO

IO before DO

binding direction

DO binds IO

IO binds DO

DO binds IO

IO binds DO

CD of IO GM

−cl ,

−cl ,

−cl ,

−cl ,

+cl ,

+cl ,

+cl ,

+cl ,

The correspondence between word order and binding direction seems to influence acceptability judgements, as those instances where these two parameters match receive higher scores than their counterpars with inverse binding dependencies:

Acceptability anaphor binding 7 6 5 4 3 2 1 bare IO

CDed IO

bare IO

DO binds IO

CDed IO

bare IO

IO binds DO DO before IO

CDed IO

bare IO

DO binds IO

CDed IO

IO binds DO IO before DO

Graph 32: Acceptability values with anaphor binding.

The high acceptability rates obtained for all patterns, irrespective of word order, binding direction or dative CD, show that, just like in the case of binding with possessives dwelt upon in the previous sections, the two internal arguments within Romanian ditransitive configurations exhibit symmetric c-command. If the account put forth in Diaconescu and Rivero (2007) were on the right track then an undoubled IO should not be able to bind into the DO, nor would a DO be able to bind into a clitic doubled IO. As seen, these patterns receive high acceptability scores in our experiment for both word orders. These data thus further contradict the alternative projection account in Diaconescu and Rivero (2007) and further strengthen the hypothesis that Romanian ditransitives are based on a single underlying configuration irrespective of whether the IO has been doubled by means of a dative clitic or not.

218

6 Romanian ditransitive configurations

6.3 A syntactic analysis of Romanian ditransitive configurations This section aims at proposing a syntactic account of Romanian ditransitives in view of the experimental findings extended upon above. The account ranges with the derivational tradition and proposes one single basic configuration for Romanian ditransitives. It will be argued that this analysis is much more flexible than the alternative account existing in the literature on Romanian ditransitives and may, as a consequence, accommodate all the eight patterns found to be acceptable. The analysis thus does justice to the symmetric binding potential uncovered with the two internal arguments, and also explains the lower acceptability of those instances where a clitic doubled IO co-occurs with a DOMed DO.

6.3.1 Against two different configurations in Romanian ditransitives The experimental findings presented in the previous sections point to one important characteristic of Romanian ditransitives: the presence of a dative pronominal clitic doubling the IO or the absence thereof has no impact on those properties of ditransitive configurations depending on the c-command relation holding between the two internal arguments i.e., binding dependencies. This in turn undermines the idea of an alternative projection account along the lines of Diaconescu and Rivero (2007), who distinguish between a POC counterpart and a DOC one: while in the POC variant, obtained with ditransitives containing undoubled IOs, DO is claimed to c-command IO, in the DOC variant, IO is inserted within the Specifier position of an applicative phrase headed by the dative clitic and c-commands DO. Since the presence/absence of the dative clitic was experimentally shown not to engender any binding difference, we have no reason to expect two different alternants to give rise to different c-commanding dependencies between the two internal arguments. As seen, IO and DO exhibit symmetrical binding abilities, being able to bind into one another irrespective of word order or the existence of dative clitic doubling. Consequently, a derivational account of these constructions might seem a better variant in view of the fact that it allows reconstruction. An alternative projection account is further weakened by the lack of any semantic differences holding between ditransitive configurations with doubled IOs and their undoubled counterparts. The following section dwells on the semantic uniformity of Romanian ditransitives and puts forth a more comprehensive semantic description of these constructions along the lines of RappaportHovav and Levin (2008).

6.3 A syntactic analysis of Romanian ditransitive configurations

219

6.3.2 Further arguments: semantic uniformity As seen, one of the main arguments on which the proponents of the Alternative projection account base their proposal of two unrelated configurations for ditransitives has to do with the consistent semantic differences uncovered with respect to the POC and the DOC. Krifka (2004) accounts for these differences in terms of two different event structures (i.e., a caused movement one for POC and a caused possession one for DOC), Harley and Jung (2015) propose that the difference lies in the specific preposition relating the two arguments (i.e., a PGO in POC vs. a PHAVE in the DOC), while Cuervo (2003a) distinguishes between two light verbs i.e, go and cause, etc. In line with Spanish ditransitives, Romanian has also been argued to pattern with English in allowing a POC counterpart with a salient directional meaning, where the IO is part of a movement event, and a DOC counterpart involving a transfer of possession relation between IO and DO. The POC counterpart would be a ditransitive configuration containing an undoubled IO, while the DOC counterpart would amount to a ditransitive structure with a clitic doubled IO (Diaconescu and Rivero 2007). More recent studies on dative verbs, have undermined this semantic distinction between the POC and the DOC and have shown that the caused movement vs. caused possession readings boil down to the verb semantics and not to the type of construction as such. Thus, Rappaport-Hovav and Levin (2008) distinguish between two classes of verbs which may be roughly grouped as verbs of possession (i.e., give- verbs) and verbs of movement (i.e., throw – verbs) on account of the readings that they enable: while the former only express caused possession in both the POC and the DOC variants, the latter allow both a caused possession and a caused movement reading. The reason why Give verbs are restricted to a caused possession reading has to do with the fact that these verbs do not possess a path component and consequently may never express the idea of movement, even if their dative DP might suggest a Goal interpretation in the POC. Give verbs may thus only express caused possession, with the dative DP being interpreted as a Possessor in both the DOC and the POC variants. Throw verbs evince both readings. This is due to the fact that these verbs possess a path component, which allows for their dative DP to be interpreted as a genuine Goal in the POC variant. The caused possession reading is also available in the DOC, where the dative DP is interpreted as a Possessor-Goal and the construction is interpreted as successful transfer. As shown in Cornilescu, Dinu, Tigău (2017b) this distinction carries over to Romanian: give verbs only allow for the caused possession reading both in

220

6 Romanian ditransitive configurations

ditransitive configurations containig clitic doubled IOs as well as in their undoubled counterparts. As such, give verbs only evince one event schema irrespective of whether they show up in the doubled dintransitives or the undoubled variant. Consider (251) below, where the caused possession reading obtains for both thematic interpretations e.g., the Possessor-Goal in (a), and the Source in (b) irrespective of whether the IO is clitic doubled or not: (251) a. (Le-)am dăruit flori profesoarelor CL.3.DAT.PL.-have.1.PL given flowers teachers.DAT.FEM.PL. (#dar acestea nu au flori). but these not have flowers ‘I gave flowers to the teachers (but they don’t have flowers).’ b. Poliția (le-)a confiscat drogurile Police.DEF.ART. CL.3.DAT.PL-has.3.SG. confiscated drugs.DEF.ART. traficanților (# dar ei au droguri). smugglers.DAT but they have drugs ‘The police has confiscated the drug dealers their drugs.’ Thus, with give verbs the state of possession or the lack thereof is entailed and may not be cancelled. Give verbs do not exhibit the caused movement interpretation. The IO is always interpreted as a Possessor and this interpretation does not depend on the presence of a dative clitic. Another test proposed by Rappaport-Hovav and Levin (2008) to distinguish between the caused possession reading and the caused movement one is that regarding the (im)possibility of where questions referring to a path component and whose answer may be a Goal. As expected, give verbs do not allow such questions; this holds irrespective of whether the IO has been doubled or not. (252) a. (Le)-am oferit flori profesoarelor. CL.3.DAT.PL.-have.1.PL given flowers teachers.DAT.FEM.PL. ‘I have offered flowers to the teachers.’ a’*Unde ai oferit flori? Where have.2.SG given flowers ‘Where did you offer flowers?’ b. (Le-)am înapoiat mașina proprietarilor. CL.3.DAT.PL.-have.1.PL returned car.DEF.ART. owners.DAT.PL ‘I returned the car to the owners.’

6.3 A syntactic analysis of Romanian ditransitive configurations

221

b’.*Unde ai înapoiat mașina? Where have.2.SG returned car.DEF.ART. ‘Where did you return the car?’ Verbs of throwing and sending evince both a caused possession as well as a caused movement interpretation. Nevertheless, distinguishing between these two interpretations does not depend on the presence/absence of the dative clitic. (253) a. (I)-am aruncat cheile de la fereastră, (CL.3.DAT.SG.)-have.1.SG. thrown keys.DEF.ART. from window dar nu le-a prins. but not CL.3.ACC.PL.-has.3.SG. caught ‘I threw the keys to him from the window, but he did not catch them.’ b. (I)-am aruncat cheile de la fereastră, (CL.3.DAT.SG.)-have.1.SG. thrown keys.DEF.ART. from window și așa a putut intra în bloc. and so has.3.SG. could enter in building ‘I threw the keys to him from the window and this is how he could enter the building.’ Given that throw verbs possess a path component, they may also allow where questions. Again, these questions are felicitous both when the IO has been clitic doubled as well as when it is bare: (254) a. (I)-am aruncat mingea portarului. (CL.3.DAT.SG.)-have.1.SG. thrown ball.DEF.ART. goalkeeper.DAT ‘I threw the ball to the goal keeper’ a’ Unde ai aruncat mingea? Where have.2.SG thrown ball.DEF.ART. Where did you throw the ball? As it seems, Romanian ditransitive configurations do not differentiate themselves function of whether the IO has been clitic doubled or not. Both variants seem to allow for the same range of interpretations irrespective of dative clitic doubling and to distinguish between a caused possession and caused movement reading relative to the verb they employ: with give verbs only the caused

222

6 Romanian ditransitive configurations

possession interpretation is available due to the fact that these verbs only have one event structure, while with throw verbs both a caused possession and a caused movement reading are available given the existence of two event structures. The two readings available with throw verbs do not depend on the presence of a pronominal clitic doubling the IO.83 This synonymy which seems to hold between ditransitives containing a clitic doubled IO and their undoubled counterparts is expected with a derivational approach but less plausible under the alternative projection account which builds precisely on the existence of semantic differences between the two constructions. In light of this fact as well as the experimental findings suggesting the existence of symmetric c-command with the two internal arguments, we will endeavour to propose a derivational account for Romanian ditransitives.

6.3.3 A low position for datives In the previous sections we argued that an account positing two different configurations for Romanian ditransitives function of whether these constructions contain CD of the Goal DP does not fare well when faced with the data. In particular, the dative clitic does not seem to have the structural import assigned to it in Diaconescu and Rivero (2007). Hypothesizing one basic configuration thus seemed more in line with the experimental findings. In this section we would like to investigate what such a basic configuration should look like. Indeed, when opting for a derivational approach with ditransitives, one needs to establish the basic structural configuration of the lexical VP, i.e. whether the basic structure is a Theme-above-Goal configuration or a

83 In a similar undertaking for Greek, Michelioudakis (2012) discusses give verb classes vs. send/ throw classes reaching a similar conclusion, namely that there is one basic underlying configuration for ditransitives with both internal arguments originating within VP. The ApplP is also posited to be of the raising/expletive type attracting arguments for case reasons. However, the preferred underlying order of the two arguments is reversed, with the Goal IO originating low, as the complement of V. This conclusion is arrived at on the basis of several tests regarding the reconstruction of the lower copies with restitutive again and interchangeability with locative adverbs. In the following sections a possible DO before IO underlying order is considered for Romanian and shown that the opposite order would be more desirable when confronted with the data. We would like to thank an anonymous reviewer for pointing Michelioudakis’ work to us and the importance of several language phenomena which contributed to a better justification of the present account.

6.3 A syntactic analysis of Romanian ditransitive configurations

223

Goal-over-Theme one. Both proposals have been made in the literature, each claiming to represent a universal underlying structure. Deciding between them in particular cases should be an empirical matter, rather than an article of faith. One important ingredient to our account is the fact that Goal DPs seem to merge low, within the VP. This is shown experimentally by Tigău (to appear) which tests the acceptability of various sentences containing two dative DPs. Sentences containing a Goal DP dative and a high Possessive dative or an Ethical dative are felicitous in Romanian, since examples such as (255) below, where a possessive dative clitic co-occurs with a goal dative, were assessed as acceptable by native speakers. (255) a. Deși le promisesem părinților că Although CL.3.DAT.PL had promised.1.SG parents.DAT that mă voi ocupa personal de el, am REFL.1.SG will.1.SG. take care personally of him have.1.SG. sfârșit până la urmă prin a le încredința ended eventually through to CL.3.DAT.PL. entrust copilul unei bone. child.DEF.ART. INDEF.ART. nanny ‘Although I had promised his parents I would personally take care of him I ended up entrusting their child to a nanny.’ b. Ion și Maria și-au schimbat atitudinea John and Mary REFL.-have.3.PL changed attitude.DEF.ART. față de mine de când le-am recomandat towards me since CL.3.DAT.PL.-have.1.SG. recommended băiatul șefului meu spre a fi angajat. son.DEF.ART. boss.DAT. mine so that to be hired John and Mary changed their attitude towards me ever since I recommended their son to my boss for a job.’ c. Cum ei erau plecați din țară și îmi dăduseră As they were gone.PL. of country and me.DAT. had given.3PL mână liberă le-am vândut moșia hand free CL.3.DAT.PL-have.1.SG sold estate.DEF.ART. unui investitor interesat să facă agricultură bio INDEF.ART.DAT investor interested SUBJ. do agriculture bio în zonă. in area

224

6 Romanian ditransitive configurations

‘As they were out of the country and had given me full freedom of action, I sold their estate to an investor interested in making bio agriculture in the area.’ In the examples above, the higher dative is realized as a clitic while the second dative DP can no longer be so realised: the Goal DP may only surface as undoubled. This phenomenon might be understood as a case of competition between the clitic doubled Goal DP and the possessive dative for the same ApplP. In line with Pylkkänen (2002, 2008), the Possessive Dative DP is a non-core argument introduced by a high Appl(icative),84and is case and θ-licensed by Appl in whose specifier it has been merged. Given that the Possessive Dative occupies SpecAppl, the goal DP may only stay in its low merge position inside the VP and clitic doubling is no longer possible given that the position in which the Goal DP would surface is now occupied by the possessive.85 Given these considerations, we take the basic configuration with Romanian ditransitives to be one in which the Goal DP is part of the argument structure of the verb. In line with Larson (2010)’s view (and contra Pylkkänen 2002, 2008), the IO is actually part of the verb’s θ-grid. It is introduced by the lexical verb itself and it composes inside the lexical VP in a syntax very much like that of Larson (1988). Under this view, the Appl head selects the lower lexical VP as its complement. Having established that both the IO and the DO merge within VP, we would now need to take one step further and consider the basic configuration in which the two arguments are base generated. This is the aim of the subsequent sections.

6.3.4 On [Person] and more In the previous chapters we discussed the featural load of the two internal arguments within ditransitives and proposed that both the dative DP and the marked DO bear a [Person] feature.

84 While the High ApplP proposed in Pylkkänen (2002, 2008) has gone unchallenged, her low ApplP has been objected to (see Larson 2010, Georgala 2011 a.o.) 85 By drawing on experimental data investigating co-occurrence possibilities between core and non-core datives (possessives and ethical datives), Tigău (2018) reaches the conclusion that extra ApplP(s) may be posited for the ethical datives only, which allow co-occurrence with possessive datives and goal datives (both doubled and undoubled), as opposed to possessive datives.

6.3 A syntactic analysis of Romanian ditransitive configurations

225

Along the lines of Richards (2008), we consider [Person] to be the syntactic counterpart of animacy/definiteness at the (semantic) interface. Richards (2008) suggests that scalar concepts such as the animacy hierarchy may be incorporated into the discrete binary system of a minimalist grammar. These hierarchies are semantic and pragmatic in nature and should be conceived of as syntaxsemantics interface phenomena. Nouns exhibiting sensitivity to the animacy and definiteness hierarchies should thus be specified for a binary grammatical feature [Person]. The [Person] feature triggers an interpretation of the respective DP along the animacy hierarchy. Building on Richards’ account, we posit that nouns may come from the lexicon carrying an unvalued [Person] feature. The [Person] feature will thus account for all the theta roles associtated to datives in ditransitive configurations and perhaps in other dative constructions as well, given that all these roles entail a [+Human] feature. Indeed, as already shown in chapter 5, sensitivity to the animacy hierarchy seems to be a general characteristic of dative arguments: their theta roles standardly denote human individuals i.e., DPs marked for [Person] in this account (Possessor-Goal, Beneficiary, Maleficiary/Source). Note also that both the inflectional as well as the prepositional dative markers exhibit high sensitivity to the animacy hierarchy, though in slightly different ways: while both select [+Human] datives, the former may also select [+ abstract] nouns and the latter may allow nouns which are lower along the animacy hierarchy: (256) a. Chelnerul turna vin oaspeților/ la oaspeți. Waiter.DEF.ART. poured wine guests.DAT to guests ‘The waiter poured wine tot he guests.’ b. Maria turna apă *florilor/ la flori. Mary poured water flowers.DAT to flowers ‘Mary poured water to the flowers.’ c. Am supus cazul atenției Consiliului/ *la Have.1.SG. subjected case.DEF.ART. attention.DAT. Board.GEN to atenția Consiliului. attention.DEF.ART. Board.GEN ‘I subjected the case to the attention of the board.’ We thus ended up by positing that IO NPs incorporate an unvalued syntactic [iPerson] feature which is further copied onto Do and which may be valued by merging K. This is, for instance, the prepositional dative la (to) or the dative inflection. K itself bears a valued [uPerson] feature so that the specification of the entire KP ends up being [iPerson: val] as described in (257) below:

226

6 Romanian ditransitive configurations

(257)

Inflectional datives get a similar featural make-up, with the only difference that their K head is silent. The feature specification was further refined function of whether the dative DP is doubled by means of a pronominal clitic or not. (257) above depicts the undoubled IO variant. When the dative is clitic doubled we posited further bleaching of K, which ends up carrying an unvalued [Person] feature. The consequence of this slight change with respect to feature specification is that the clitic doubled IO will need to check its resulting [iPerson: __] against a suitable head, which was not the case for the undoubled variant, whose feature specification was [iPerson: val]. Marked DOs i.e., DOMed and CDed+DOMed DOs have also been argued to bear a [Person] feature in line with similar proposals for Spanish by Leonetti (2004) and Rodriguez-Mondoñedo (2007). As long claimed in the literature on Romanian DOM (Dobrovie-Sorin 1990, 1994), Cornilescu (2000) a.o., differentially object marked DOs are sensitive to the animacy and the definiteness scales in Aissen (2003) and a natural step for the account put forth was to analyse marked DO on a par with dative IOs as above. Briefly, we argued that, when DOs are syntactically marked as denoting people, [+Human(like)] NPs incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by merging K. As with datives, the K head of marked DOs is also endowed with a valued [uPerson] feature. The possibility of incorporating [Person] is open only to NPs that are provided with the lexical feature [+Human] or rather [+Human (like)], given that other types of nouns (e.g. pets etc.) may be treated like humans and may become marked objects (Mardale 2008, Tigău 2010). Example (258) shows the structure of marked DOs, which is similar to that of undoubled IOs. Unmarked DOs on the other hand will only exhibit a standard DP structure:

6.3 A syntactic analysis of Romanian ditransitive configurations

(258)

227

KP ep K +p uφ:__ u-person:val

DP ep D +d uφ: __ idef: __ iperson: ___

pe DOM

un a

NP +N iφ:val u-def i person:___ +HUM profesor professor

As in the constructions with doubled IOs, the CDed+DOMed DOs were assumed to have a slightly different feature specification, with K contributing an unvalued [uPerson:__] so that the entire KP ends up bearing [iPerson:__], which it will have to check against an appropriate head during the derivation. Let us now turn to the Applicative projection: remember that we have adopted Larson (1988)’s assumption that dative DPs are theta-marked by the verb within ditransitive configurations. Given that these DPs also need to check their case feature, we have also posited an expletive ApplP within vP, which takes VP as its complement. In line with Georgala (2011), we take the ApplP to be of the expletive type, functioning as a case assigner. Applo also introduces an [uPerson], which is the reflex of the essential property of dative arguments i.e., their sensitivity to the animacy hierarchy. Finally, besides [Person], both internal arguments carry an [uCase] feature which they would have to check against a suitable head during the derivation. In the following sections, we will try to show that the feature specifications of the two internal arguments plays an important part when it comes to accounting for the acceptability judgements of native speakers of Romanian on the various instances of binding dependences assessed. In particular, it will be shown that the unacceptability of some the patterns (i.e., DOMed DOs co-occuring with clitic doubled IOs) may be understood as a result of the competition between two DPs with the same feature specification for the same probe i.e, Appl.

6.3.5 One basic configuration In line with López (2012), and following the argumentation built on the behaviour of DOMed vs. unmarked DOs in Romanian, we start from the assumption that DO DPs are merged in the complement position of the lexical verb. Unmarked DOs,

228

6 Romanian ditransitive configurations

which undergo reanalysis to NP, incorporate into V and check their [uC] feature when V further incorporates into v through feature sharing; v values the [uC] feature of the D/# head incorporated into V and then feature sharing takes place as a consequence of the syntactic dependency between the incorporated D/# head and its copy. Marked DOs, on the other hand check their [uC] feature by moving into a position, SpecαP, wherefrom they may be probed by the accusative case assigning head v. Movement is triggered by the need to check Accusative case in a situation in which incorporation into V is blocked. Thus, in line with López (2012), we adopt the view that there is a single type of Case which is generally assigned by way of incorporation, unless the internal make-up of the nominal in question prevents it, in which case movement takes place as an alternative solution. We further take the IO as merging within the specifier of VP. This is in line with the idea that IO is related to the event denoted by the verb and hence part of its θ – grid; it is introduced by the lexical verb and it composes with it inside the VP along the lines of Larson (1988, 1990) for DOCs. The tree in (259) shows this at stake: (259)

vP ei John v ei gave VP ei Mary V´ ei gave Fido

(Larson 2014: 35, p. 19)

One argument in favour of this basic configuration comes from the realm of clitic clusters. Săvescu (2009) dwells on the order of pronominal clitics within clitic clusters and observes that, with singular clitics,86 the order is invariably one in which the Dative argument precedes the Accusative one both preverbally and postverbally:

86 Plural clitics may occur both in the Dative > Accusative and the Accusative > Dative word orders. The latter sequence arises only with those plural clitics that appear post-verbally and which have the accusative ‘default’ form in this case: (1)

Recomandându-vă-ne a zâmbit malițios. Recommending-2.ACC.PL.-1.DAT.SG. has.3.SG. smiled maliciously ’When he recommended you to us, he smiled maliciously.’ (Săvescu 2009: 43, p. 162)

6.3 A syntactic analysis of Romanian ditransitive configurations

(260) a. Mi l-a cumpărat Me.DAT CL.3.ACC.M.SG.-has.3.SG. bought ‘My mother bought it for me.’ b. Dăruindu-mi-l, mi-a Giving-me.DAT-CL.3.DAT.M.SG. me.DAT-has mare bucurie. big joy ‘By giving it to me, she made me very happy’

229

mama mother.DEF.ART. făcut o made INDEF.ART.

The order of pronominal clitics is argued to mirror the internal order of the two VP arguments: the direct and the indirect object clitics thus originate in a structure resembling the configuration of doubled object constructions along the lines of Anagnostopoulou (2003, 2005, 2006), Béjar and Rezac (2003) and move upwards first to Case projections and then to Person projections, maintaining their initial ordering (261). (261) a. Mi te văd. Me.DAT you.ACC see.1.SG ‘I see you’ b.

Person1P ei mi Person2P ei te TP ei KPDAT ei KPACC tIO ei tDO ............. V ei ApplP ei tIO Appl ei Appl tDO

(Săvescu 2009: 10, p. 96) In line with Săvescu (2009), we also posit that the basic order of the two argument DPs within ditransitives is Dative > Accusative. This is also in line with

230

6 Romanian ditransitive configurations

similar proposals made for Spanish ditransitives in López (2012), Pineda (2012, 2013) a.o. We further posit an expletive ApplP taking VP as its complement (Georgala 2011, Georgala and Whitman 2008), which may be understood as the αP of López (2012): (262)

ApplP/αP ei Appl/α ei Applo/α VP ei IO V´ ei V DO

As already announced, the Applicative projection is somewhat different from the one initially put forth by Pylkkänen (2002, 2008) in that it only participates in case checking and also carries a [uPerson] feature. Building on Marantz (1993), Pylkkänen (2002) posits an Applicative projection, which, just like the VoiceP of Marantz (1984), introduces an argument into the argument structure of the verb: while the VoiceP introduces the external argument, the ApplP introduces an internal argument i.e., the IO. As known, Pylkkänen (2002, 2008) proposes two types of applicative constructions. High applicatives merge above VP/vP and point to a thematic relation holding between the event denoted by the lexical verb and an individual.87 This is the case, for instance, of the Benefactive from the Chaga example in (263a) below with the corresponding derivation in (263b): (263) a. N-ä-ï-lyì-í-à m-kà k-élyà. FOC-1SG-PRES-eat-APPL-FV 1-wife 7-food ‘He is eating food for his wife.’

(Pylkkänen 2008: 2, p. 11)

87 High applicatives may merge quite high in the derivation given that they only need access to the event variable contributed by the verb. As opposed to high applicatives, low ones need to combine at an earlier stage in the derivation so that they may be in the vicinity of the DO with which they are related.

6.3 A syntactic analysis of Romanian ditransitive configurations

b.

231

VoiceP ei he Voice´ ei Voice ApplP ei wife Appl´ ei ApplBen VP ei eat food

(Pylkkänen 2008: 6a, p. 14)

Low arguments, such as the IO within ditransitive configurations, are argued to be introduced into the argument structure of the verb by means of a low applicative projection. Unlike high applicatives, which are analysed in a neoDavidsonian way as denoting relations between individuals and events [λx.λe Beneficiary ðe, xÞ], low applicatives are argued to denote relations between two individuals i.e., those denoted by the DO and the IO. Thus, the IO is claimed to bear no semantic relation to the verb, only evincing a relation of directional transfer of possession with the DO as shown in (264) below: (264) a. Mary bought John the book. b.

VoiceP ei Voice´ Mary ei Voice VP ei buy ei John ApplP ei Appl book (Pylkkänen 2008: 6b, p. 14)

c. λx.λy.λf .λe. f (e,x) & theme (e,x) & to-the-possession (x,y)

232

6 Romanian ditransitive configurations

An analysis of Romanian ditransitives along the lines of Pylkkänen (2002, 2008) would thus presuppose resorting to a low applicative account. This analysis, however, has been shown to face several problems. Larson (2010) signals, for instance a semantic problem: in Pylkkänen (2002, 2008), the IO is uncoupled from the event structure of the verb and this gives rise to several logical consequences. Given that applied arguments bear no semantic relation with the verb and only a transfer of possession relation with the DO, the applied argument is not part of V’s θ-grid. As such, the semantics put forth by Pylkkänen for ditransitives is mono-eventive contrary to the bieventive neo-Davidsonian traditional analysis (Pesetsky 1995). Larson (2010) points out that this departure from the standard neoDavidsonian semantics according to which all participants are event properties causes a number of undesirable inferences, and draws on examples based on conjunction to show this. Thus, (265a), with the DO pointing to the same entity in both conjuncts, does not entail (265b). The fact that John wrote the letter and that the letter in question ended up in the possession of Mary does not necessarily entail that John wrote the letter to Mary. While an analysis along the lines of standard neo-Davidsonian semantics successfully hinders such an inference, Pylkkänen’s account has the undesirable result of allowing it. (266a) represents the analysis of (265a) along the lines proposed in Pylkkänen: (266a) actually entails (266b) under this account (see (267)): (265) a. John wrote that letter and Bill gave Mary that letter. b. John wrote Mary that letter. (Larson 2010: 3, p. 702) (266) a. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter)] & 9 e’[giving (e’) & Agent (e’, Bill) & Theme (e’, that_letter) & to-the-possession-of (that_letter, Mary)] b. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter) & to-thepossession-of (that_letter, Mary)] (Larson 2010: 4, p. 702) (267) a. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter)] & 9 e’[giving (e’) & Agent (e’, Bill) & Theme (e’, that_letter) & to-the-possession-of (that_letter, Mary)] b. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter)] & 9 e’[giving (e’) & Agent (e’, Bill) & Theme (e’, that_letter) & to-the-possession-of (that_letter, Mary)] c. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter)] & to-thepossession-of (that_letter, Mary) & 9 e’[giving (e’) & Agent (e’, Bill) & Theme (e’, that_letter)]

6.3 A syntactic analysis of Romanian ditransitive configurations

233

d. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter) & to-thepossession-of (that_letter, Mary)] e. 9e [writing (e) & Agent (e, John) & Theme (e, that_letter) & to-the(Larson 2010: 5, p. 702) possession-of (that_letter, Mary)] Given the undesirable inferences that Pylkkänen’s semantic account gives rise to, Larson (2010) proposes that the applied argument is in fact related to the event denoted by the lexical verb and as such part of the verb’s θ- grid. Georgala, Paul and Whitman (2008) point to morphological problems arising with low applicatives. Thus, under the view that head-movement presupposes uniform raising, the account in Pylkännen (2002, 2008) predicts a structural difference between the position of high vs. low applicative heads. Adhering to Baker’s (1988) Mirror image principle, if head movement involves uniform raising and adjunction to the left, then high applicative morphemes should be suffixed on the verb, since the verb raises to Appl. This prediction is confirmed in Bantu languages, which have a rich morphological system of Appl heads. In contrast, for low applicatives, the Appl should be left-adjoined to the higher lexical verb and as such end up as a prefix. This prediction is, however, disconfirmed crosslinguistically, and Georgala, Paul and Whitman (2008) propose that Appl morphemes uniformly occupy the same suffixal position. In face of the discussion above, we adopt the view that, in ditransitive configurations, both internal arguments pertain to the thematic structure of the main verb. Consequently, Appl may merely have a case-licensing role in line with Georgala (2010, 2011) a.o. Indeed, Harada and Larson, (2009), Georgala (2011), Wood and Maratz (2015) a.o. distinguish between thematic and expletive (raising) applicative heads, as in (268). Thematic applicatives are endowed with θ-features, which get valued by the non-core arguments they introduce, as in (257a). They also case-license the non-core arguments they introduce. Expletive (raising) applicatives, on the other hand, merely case-license an argument which is introduced by the lexical verb. The lexical verb also assigns a θ-role to this argument. (268) a. Thematic applicatives [vP SUBJ [v’v [ApplP IO Benf/Loc/ Instr. . . [Appl’ Appl [VP V DO]]]]] b. Raising applicatives [vP SUBJ [ v’v [ApplP IOrec . . . [Appl’ Appl [VP tIOrec V DO]]]]]

234

6 Romanian ditransitive configurations

As already announced, we further posit that Appl may also introduce a supplementary semantic feature, thus enriching, or even changing the thematic interpretation assigned by the verb to the applied DP.88 This would be a [uPerson].89 The next section offers an analysis of Romanian ditrasitives putting to work all the ingredients extended upon so far. 6.3.6 A derivational analysis for Romanian ditransitives The current analysis is reminiscent of the Featural Relativized Minimality put forth by Rizzi (2001) and later work, according to which the acceptability of a configuration where intervention arises varies function of the degree of featural overlap between the extracted element and the intervener i.e., configurations presupposing a lower degree of featural overlap are more acceptable than sentences exhibiting a higher degree of featural overlap. Note, however, that this system is further corroborated with some priority criteria for movement: the featural load of the two internal arguments seems to have a say with respect to the order in which the two DPs are allowed to move. We thus posit that the two internal arguments within ditransitive configurations need to undergo checking of their features by obeying certain priority criteria. More specifically, the DO will have general priority over the IO. Priority may, however, change function of the feature specification of the two objects if both of them exhibit sensitivity to [Person]. The following cases arise: 1. Unmarked DOs only bear [uC] and have no specification with respect to [Person] in syntax; the IO will always have both [uC] and [Person], (irrespective of whether this latter feature is [iPerson: val] as with undoubled IOs or [iPerson: ___] as with their doubled counterparts). Given that DO has no [Person] feature to verify, it will simply undergo scrambling first. 2. DOMed DOs bear [uC] and [iPerson: val]. In this case, both DO and IO are sensitive to [Person] so a prioritization as to which values their [Person] feature first needs to occur. Two situations may arise: a. IO has the same feature specification i.e., [uC] and [iPerson: val] (as it is undoulbed): DO will be given priority for movement.

88 The proposal is similar to that of Michelioudakis (2012), who proposes that Appl contributes a [+ mental state] component along the lines of Reinhart (2002), which is assigned to event participants that are intended to be consciously affected and which correlates with an unvalued [uParticipant] feature which probes for a suitable IO. 89 See Cornilescu and Tigău (2018) for further considerations on the feature specification of Appl.

6.3 A syntactic analysis of Romanian ditransitive configurations

235

b. IO is doubled and as such has more features to verify i.e., [uC] and [iPerson: __]: in this particular case the IO will gain priority over the DO, which only needs to verify one feature. 3. CDed+DOMed DOs bear [uC] and [iPerson: __] and will always have priority over the IO: a. If the IO is undoubled its feature specification is [uC] and [iPerson: val]: DO has priority because of its DO status and also because it will have more features to verify. b. If the IO is doubled, then it will have the same feature specification as the DO i.e., [uC] and [iPerson: __]: DO has priority according to the initial criterion. The next sections will account for each of the patterns found to be acceptable.

6.3.6.1 DOMed DOs and CDed IOs As seen in the previous section, the co-occurrence of DOMed DOs with CDed IOs received very low acceptability scores in the DO before IO word order for both directions of binding (DO>IO: 2,64; IO>DO: 2,36). The IO before DO word order received scores found within the range of values obtained for the other cases (DO>IO: 3,27; IO>DO: 3,68). Note first that, in this particular case there is a high featural overlap between the IO and DO and that the two are both sensitive to [Person]. The priority criteria proposed above thus need to apply. Consider the derivation at work: we start from a VP where the DO is merged in the complement position and has the feature specification: [iPers: val], [uCase]. (Remember that the NP in this case bears an [iPers: __], which triggers the KP layer with the specification [uPers: val]. By way of agreement, the [Person] specification of the KP is [iPers: val]). The DO thus only needs to verify its [uCase] feature. The IO, on the other hand, has the feature specification [iPerson__], [uCase]. (Remember that, just like CDed DOs, CDed IOs were argued to have their K further bleached, so this functional projection only contributes a [uPerson: __]. As such the NP bearing [iPers: __] cannot verify its feature and the specification for the entire KP is [iPers: __]). The IO will thus have to find a way in which to value both these features. Note that both objects are specified for Person, but that the IO has more features to verify and will thus gain priority over the DO. The IO enters an Agreement relation with the α head (specified as [uPerson:val]) and checks both its case and its [iPers: _] feature. The [uPerson:val] feature of α is EPP and

236

6 Romanian ditransitive configurations

the IO moves to Spec αP. As such, it acts as an intervener for the DO, which may no longer move to a Spec α in order to get its case feature valued by v. Thus, movement of the DO out of VP is not possible and the derivation crashes. Hence the low results with the DO before IO word order: the DO may not get out of the VP. One way to save the situation is by scrambling the IO out of Spec αP, into a specifier of v. As a consequence, the IO will no longer act as an intervener for the DO, which may scramble to a specifier of α and get its case feature valued by v. This explains why the order IO before DO (involving optional scrambling) was found to be acceptable. (269)

vP ei v’ ei αP v ei IOcl α’ iPers:val ei uC VP αcase uPers:val ei IOcl V’ iPers__ ei uC V DOKP iPerson: val uC

6.3.6.2 CDed+DOMed DOs and CDed IOs This combination is allowed in both directions of binding and word orders: a. DO before IO: DO binds into IO: 3,52/ IO binds into DO: 3,18 b. IO before DO: DO binds into IO: 3,42/ IO binds into DO: 3,51 In this particular situation, the DO and IO have the same feature specification: [iPerson:__] and [uCase]90 and the DO will have priority over the IO. As a consequence, it will move to Spec αP and value its case feature against v. Given that the DO also needs to value its [iPerson:__] feature, this KP will move further, to a specifier of v and enter an agreement relation with the Person head. The IO

90 The NP bears an [iPers: __], which triggers the KP layer with the specification [uPers: __]. Note that in this particular case K has undergone further bleaching and the [Person] feature it contributes is unvalued. As a consequence of agreement the [Person] specification of the KP is [iPers: ___]). The same reasoning applies for the IO.

6.3 A syntactic analysis of Romanian ditransitive configurations

237

will be probed by the α head and will have both its [iPerson:__] and case features valued against this head. The [uPerson: val] of α will also be checked as a consequence. (270) below shows this at work: (270)

Pers’ ei Perso vP [iPers:val] ei DOcl v’ iPers:__ ei vcase αP ei DOcl α’ iPers: ei uC VP αcase uPers:val ep IOcl V’ iPers__ ei uC V DOcl iPerson: __ uC

The resulting surface order is DO before IO. This derivation shows that both binding directions are possible, given that the DO may occupy a position wherefrom it may c-command the IO and the other way round. The IO before DO surface order, also attested as acceptable, may be easily obtained by scrambling the IO to a position above the landing site of the DO. 6.3.6.3 Unmarked DOs and CDed IOs These structures have also been found fairly acceptable in both directions of binding and word orders: a. DO before IO: DO binds into IO: 3,64/ IO binds into DO: 3,00 b. IO before DO: DO binds into IO: 3,22/ IO binds into DO: 4,35 In this particular case, there is less featural overlap between the two internal arguments: it is only the IO which evinces sensitivity to [Person], bearing [iPerson:__] and [uCase]. The DO91 has no [Person] specification and only needs

91 Remember that unmarked DOs have DP status in Romanian and may optionally scramble or stay in-situ and incorporate into the V. The latter case has not been discussed here.

238

6 Romanian ditransitive configurations

to check case. In line with the priority criteria posited above, it will be the first to enter the derivation and move into the specifier of αP where it values its case feature against v. The IO will verify both case and [Person] against the α head: (271)

vP ei v´ ei vcase αP ei α´ DO ei uC– VP αcase uPers:val ei IOcl V´ iPers__ ei uC V DO uC

The derivation above gives rise to the DO before IO surface order and accounts for both directions of binding, given that DO may reconstruct. The IO before DO surface order obtains by scrambling the IO above the landing position of the DO. Again, both directions of binding are accounted for and found acceptable by the respondents, given the reconstruction possibilities of the IO. The dependencies obtained by way of reconstruction, however, get slightly lower acceptability scores than the items where the surface word order matches the binding direction, as generally seen for all patterns. 6.3.6.4 Unmarked DOs and unmarked IOs These configurations have been assessed as acceptable by the Romanian respondents in both directions of binding and surface word orders: a. DO before IO: DO binds into IO: 4,57/ IO binds into DO: 3,08 b. IO before DO: DO binds into IO: 3,40/ IO binds into DO: 4,56 In this particular situation, the featural overlap is only partial: DO only bears [uCase] and is insensitive to [Person]. The IO, on the other hand, bears [iPerson: val] and [uCase]. As in the previous case, the DO will be the first to move, reaching Spec αP where it has its case feature valued against v. IO will enter agreement with the α head and value its case feature against this head. In turn, the IO will value the [uPerson: val ] of α.

6.3 A syntactic analysis of Romanian ditransitive configurations

(272)

239

vP ei v´ ei vcase αP ei DO α´ uC ei VP αcase uPers:val ei IO V´ iPers:val ei uC V DO uC

The surface DO before IO order thus obtains. Both directions of binding find an explanation with this derivation, given the reconstruction possibilities of the DO in its merge position below IO. As expected, the DO binds into IO binding dependency is considered more acceptable than the reverse, given that the DO scrambles out of the VP and reaches a c-commanding position with respect to the IO. The opposite binding dependency is not as felicitous as it presupposes the reconstruction of the DO in its merge position. The dependencies arising from the surface IO before DO word order may also be accounted for if the IO undergoes scrambling to a specifier of vP above the landing site of DO. Again, as expected, the IO binds into DO dependency fares much better in this case. 6.3.6.5 DOMed DOs and unmarked IOs These sequences follow the pattern of most ditransitive configurations in that both directions of binding and word orders are found to be fairly acceptable: a. DO before IO: DO binds into IO: 4,43/ IO binds into DO: 3,05 b. IO before DO: DO binds into IO: 3,36/ IO binds into DO: 4,32 In this particular case, there is a high featural overlap between the two objects: they both bear [iPerson:val] and [uCase]. As a consequence, DO will have priority over IO and will be the first to move. DO thus undergoes scrambling to Spec αP and values its case feature against v. IO will be probed by α and have its case feature valued against this head. α itself will value its [uPerson:val] feature by way of agreement with IO.

240

(273)

6 Romanian ditransitive configurations

vP ei v´ ei αP vcase ei α´ DO uC ei iPers:val VP αcase uPers:val ei IO V´ iPers:val ei uC V DO uC iPers:val

The derivation accounts for the DO before IO surface order and explains both binding directions given the possibility for the DO to reconstruct below IO. Note also the expected preference for the DO binds into IO: 4,43 pattern. The opposite surface word order may be arrived at by means of IO optional scrambling to a position c-commanding the landing site of the DO. Again, the higher scores obtained for the matching binding direction, i.e., IO binding into DO, is expected. 6.3.6.6 CDed+DOMed DOs and unmarked IOs These sequences received similar acceptability scores to most of the ditransitive configurations regarding both directions of binding and word orders: a. DO before IO: DO binds into IO: 4,51/ IO binds into DO: 3,47 b. IO before DO: DO binds into IO: 3,67/ IO binds into DO: 5,52 In this particular situation, the DO has the feature specification [uCase] and [iPerson: ___] (Remember that K is bleached in this case and only contributes [uPerson: ___]. This feature will enter agreement with the [iPerson: ___] feature provided by the NP and the result is [iPerson: ___]. Hence the [Person] feature on DO needs to be valued against an appropriate head). The IO bears [uCase] and [iPerson: val]. Given that the DO generally has priority over the IO and that it also has more features to check, the DO will be the first to enter the derivation and undergo movement to Spec αP. Having reached this position, the DO will value its case feature against v and will have to undergo further scrambling to a specifier of v so that it may also verify its [iPerson: ___] against the Person head.

6.3 A syntactic analysis of Romanian ditransitive configurations

241

The IO gets probed by α, values its case feature by way of agreement and allows α itself to value its own [uPerson: val]. (274)

Pers´ ei Perso vP [iPers:val] ei DOcl v´ iPers:__ ei αP vcase ei α´ DOcl iPers: ei uC VP αcase uPers:val ei IO V´ iPers:val ei uC V DOcl iPerson: __ uC

The derivation gives rise to the surface order DO before IO and also explains the acceptability of both binding directions under the hypothesis that DO may reconstruct into its merge position. Again, the DO binds into IO dependency comes out as more acceptable, as expected: by undergoing short scrambling to Specα, DO finds itself in a position wherefrom it c-commands IO. The reverse binding dependency presupposes reconstruction of the DO in its merge position. The IO before DO surface word order obtains by way of optionally scrambling the IO out of VP, into a specifier of vP c-commandig the landing site of DO. As a consequence, the IO is now in a c-commanding position with respect to DO and the binding dependency IO binds into DO fares quite well, getting acceptability scores of 5,52 GM. The opposite binding dependency is less felicitous as it presupposes reconstruction of the IO in its initial merge position. 6.3.6.7 Conclusions In this section we have put forth a derivational analysis of Romanian ditransitive constructions. This account has proved more adequate in that it was shown to accommodate all the structures which were shown to obtain experimentally by varying the word order of the two internal arguments, the direction of binding, and presence/absence of the clitic doubling the IO. The analysis accounts for the experimentally uncovered symmetric potential of the DO and IO as well as for

242

6 Romanian ditransitive configurations

the lack of semantic differences holding between ditransitive configurations containing undoubled IO and their counterparts involving doubling of the IO. Romanian was shown to pattern with English and distinguish between a class of give verbs which only allow for a caused possession interpretation and a class of throw verbs which exhibit both a caused possession as well as a caused motion reading. The two different readings do not depend, however, on the presence/ absence of a dative clitic doubling the IO within ditransitive constructions but rather on the type of verb employed. In this respect ditransitives containing doubled IOs pattern with their undoubled counterparts. Finally, this section also accounts for the problematic instances where a DOMed DO interacts with a clitic doubled IO. In this respect, the account draws on the internal featural make-up of the two internal arguments and also proposes a priority rule between these DPs when it comes to the order in which they check their relevant features. Both IO and DO are part of the θ-grid of the verb and are merged within VP. They further move in order to check their case and Person features. Appl and v represents heads against which case may be valued. The Appl head also carries a [uPerson] feature. In this respect, this section implicitly argues for a constructuivist approach to θ-role interpretation: Appl introduces a θ feature in the derivation i.e., [Person] modifying the interpretation of an initially assigned role e.g., Possessor may thus be seen as the sum of [Location, Person] (see Cornilescu and Tigău 2018 for an extensive discussion on the topic).

6.3.7 Experimental evidence in favor of the proposed syntactic analysis Scope dependencies and the existence of scope freezing effects with ditransitive constructions have been long employed as a diagnostic test for the identification of the underlying order of the two internal arguments. Hoji (1985), for instance, proposes that the basic order within Japanese ditransitives is Goal-overTheme, given that, in this particular configuration the Goal asymmetrically outscopes the Theme. The opposite order, where the Theme precedes the Goal is argued to be derived by means of scrambling due to the existence of scope ambiguities, which may be explained by way of DO reconstruction into its merge position. Consider (275): (275) a. Taroo-ga dareka-ni dono-nimotu-mo okutta. Taro.NOM someone.DAT every-package sent ‘Taro sent someone every package.’ some > every, *every > some

6.3 A syntactic analysis of Romanian ditransitive configurations

243

b. Taroo-ga dono-nimotu-mii dareka-ni ti okutta. Taro.NOM every-packagei someone.DAT ti sent ‘Taro sent someone every package.’ some > every, every > some (Miyagawa and Tsujioka 2004: 11, 12, p. 5)

In (275a) the only available reading is the one in which the dative DP dareka-ni (someone) has wide scope i.e., ‘there is someone to whom Taro sent every package’. Example (275b), on the other hand, allows both for the wide scope interpretation of the Goal DP and a narrow scope reading according to which ‘each package was sent to a (possibly) different person’. As already seen, Romanian also allows for a Goal – over– Theme as well as a Theme –over – Goal configuration and under the derivational approach adopted the former was claimed to be the basic one. The current section provides further evidence in favor of this hypothesis by presenting the results of three experiments probing for scope dependencies within ditransitives. Unlike their counterparts in Japanese, Romanian ditransitives do not exhibit scope freezing effects in either of the two orders. Nevertheless, it will be shown that the experimental data seem to be better accounted for if a Goal-over-Theme order is adopted as the underlying configuration. This section has the following structure: we begin by describing the basic aims of the three experiments in section 6.3.7.1 and continue with a discussion of the results in 6.3.7.2. Section 6.3.7.3 contains the conclusions. 6.3.7.1 Three experiments on scope dependencies with ditransitives The three experiments we carried out aimed at verifying scope dependencies holding between the two internal arguments of Romanian ditransitive constructions. The experiments had a similar design and only differed with respect to the type of DO employed: a) Experiment 1 verified scope dependencies between unmarked DOs and IOs b) Experiment 2 investigated scope dependencies between DOMed DOs and IOs c) Experiment 3 dealt with scope dependencies holding between CDed+DOMed DOs and IOs. For each experiment we varied: the surface order of the two internal arguments with respect to each other, doubling of the IO by means of a dative pronominal clitic or the lack thereof, and the DP type i.e., indefinite vs. universal quantifier. Table (30) below shows this:

244

6 Romanian ditransitive configurations

Table 30: Experimental conditions. DO before IO

IO before DO

DOUQ before IOind

DOind before IOUQ

IOUQ before DOind

IOind before DOUQ

−clIO  items

−clIO  items

−clIO  items

−clIO  items

+clIO  items

+clIO  items

+clIO  items

+clIO  items

We built 8 items and then varied them function of the three aforementioned conditions coming up to a total of 64 items. The items were further split into four different lists using the Latin square method and 14 fillers were added to each list. The items and the fillers in each list were then randomized and then entered into Google forms. The four resulting questionnaires were assessed by at least 20 native speakers of Romanian each, so more than 80 people took part in each experiment, with a rough overall participant score of 240 for the three experiments. For each item, the respondents were required to perform an acceptability task, assigning scores on a 7-rung acceptability scale (1 being the lowest acceptability score and 7 the highest), and a norming one selecting either the wide scope interpretation for the indefinite DP in the item, the narrow scope one or both readings. Example (276) below is a test item with the accompanying multiple choice question: (276) Comisia a repartizat unui conducător de Committee.DEF.ART. has assigned INDEF.ART.DAT supervisor of doctorat fiecare student. PhD every student ‘The committee has assigned every student to a PhD supervisor.’ One may interpret this sentence as follows: a) There are several PhD supervisors and several students and each student was assigned to a (possibly) different supervisor. b) All the students were assigned to the same PhD supervisor. c) (a) and (b) As already pointed out, each item was further modified by changing the surface word order of the two internal arguments, the DP type and the presence/absence of a dative clitic doubling the IO, each with its own suitable interpretation variants:

6.3 A syntactic analysis of Romanian ditransitive configurations

245

(277) DOUQ before IOind, no IO clitic a. Comisia a repartizat fiecare student Committee.DEF.ART. has assigned every student unui conducător de doctorat. INDEF.ART.DAT supervisor of PhD The committee has assigned every student to a PhD supervisor. DOUQ before IOind, IO clitic b. Comisia i-a repartizat fiecare student Committee.DEF.ART. CL.3.DAT.SG-has assigned every student unui conducător de doctorat. INDEF.ART.DAT supervisor of PhD The committee has assigned every student to a PhD supervisor. DOind before IOUQ, no IO clitic c. Comisia a repartizat un student fiecărui Committee.DEF.ART. has assigned INDEF.ART. student every.DAT conducător de doctorat. supervisor of PhD The committee has assigned a student to every PhD supervisor. DOind before IOUQ, IO clitic d. Comisia i-a repartizat un Committee.DEF.ART. CL.3.DAT.SG-has assigned INDEF.ART. student fiecărui conducător de doctorat. student every.DAT supervisor of PhD The committee has assigned a student to every PhD supervisor. IOUQ before DOind, no IO clitic e. Comisia a repartizat fiecărui conducător de Committee.DEF.ART. has assigned every.DAT supervisor of doctorat un student. PhD INDEF.ART. student The committee has assigned a student to every PhD supervisor.

246

6 Romanian ditransitive configurations

IOUQ before DOind, IO clitic f. Comisia i-a repartizat fiecărui Committee.DEF.ART. CL.3.DAT.SG-has assigned every.DAT conducător de doctorat un student. supervisor of PhD INDEF.ART. student The committee has assigned a student to every PhD supervisor. IOind before DOUQ, no IO clitic g. Comisia a repartizat unui conducător Committee.DEF.ART. has assigned INDEF.ART.DAT supervisor de doctorat fiecare student. of PhD every student The committee has assigned every student to a PhD supervisor. IOind before DOUQ, no IO clitic h. Comisia i-a repartizat unui Committee.DEF.ART. CL.3.DAT.SG-has assigned INDEF.ART.DAT conducător de doctorat fiecare student. supervisor of PhD every student The committee has assigned every student to a PhD supervisor. The current section will only dwell on the results regarding the norming part of the experiments and leaves the aspects related to item acceptability aside for ease of presentation.

6.3.7.2 Results and discussions 6.3.7.2.1 Experiment 1: Scope dependencies with ditransitives (unmarked DO) One important observation with respect to the two possible scope readings is that neither is completely excluded but that speaker preferences were tied up with the surface word order, the DP type i.e., indefinite vs. universal quantifier and the presence/absence of IO clitic doubling (Table 31): In the DOUQ – IOIndef order, the participants prefer the narrow scope interpretation on the indefinite IO (57.63%) but also allow for a wide scope reading (25.54% + 16.88%). The presence of a dative clitic doubling the IO seems to tilt the balance towards a more prominent wide scope reading of the indefinite DP (38.80%). Nonetheless, the narrow scope reading for the IO remains an option,

6.3 A syntactic analysis of Romanian ditransitive configurations

247

Table 31: Scope dependencies with ditransitives (unmarked DO). Nr. crt.

Word order

DP type

+/− IO clitic

Results

Narrow scope DPindef

Wide scope DPindef

Both readings

.

DO before IO

UQDO & IndefIO

no IO clitic

.%

.%

.%

.

DO before IO

UQDO & IndefIO

with IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO

no IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO

with IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO

no IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO

with IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO

no IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO

with IO clitic

.%

.%

.%

GM

DO (UQ) & IO (IND)

DO (IND) & IO (UQ)

DO BEFORE IO

Graph 33: Scope dependencies with ditransitives.

WITH IO CLITIC

IO (UQ) & DO (IND)

18.48% 65.50% 16.03%

NO IO CLITIC

26.17% 60.42% 13.38%

WITH IO CLITIC

75.30%

68.71% 13.80% 17.49%

NO IO CLITIC

11.30% 13.40%

65.33% 16.50% 18.15%

WITH IO CLITIC

Series3

75.43%

38.95% 39.80% 21.23%

NO IO CLITIC

Series2

10.93% 13.63%

57.63% 25.54% 16.88%

Series1

NO IO CLITIC

WITH IO CLITIC

IO (IND) & DO (UQ)

IO BEFORE DO

248

6 Romanian ditransitive configurations

with 38.95% of the respondents selecting it as the only possible interpretation and 21.23% accepting both the wide and the narrow scope readings. When an indefinite DO is employed in this configuration, the same marked preference for its narrow scope interpretation may be observed both when the IO has been clitic doubled and when it has been left undoubled (65.33% and 68.71% respectively). Likewise, the wide scope interpretation is also available, with a small percentage of respondents selecting it as the only available reading (16.50%/13.80%), while another group consider both readings acceptable (18.15% and 17.49%). In the IOUQ before DOIndef surface order the indefinite DO is assigned a narrow scope interpretation by most of the respondents (75.43% and 75.30% respectively). In the case of an indefinite IO within the same word order configuration respondents tend to assign the wide scope reading (60.42%/ 65.50). Thus, an unmarked indefinite DO tends to be assigned a narrow scope reading irrespective of the surface word order. The wide scope interpretation remains, nevertheless, an option. In what follows, we will try to account for the experimental data by considering both the DO before IO and the IO before DO orders in turn as possible underlying configurations for Romanian ditransitives. If DO before IO were the underlying order of all surface realisations with ditransitives, then the matching DO before IO surface order in the first four patterns in Table 31 should actually be arrived at by way of scrambling both the DO and the IO outside the VP. Lack of IO scrambling would be conducive to the impossibility of explaining its wide scope interpretation attested as available: according to (278), if IO is merged below DO one may only expect a narrow scope reading, unless there is movement of IO above the DO. (278)

VP ei DO V' ei Vº IO

As with the derivational account presented in the previous sections, it may be posited that the IO moves to Spec α/Appl for reasons of case. The DO before IO surface order is then derived by DO scrambling above the landing site of IO. Let us consider how a derivation would unfold in a situation in which an unmarked DO co-occurs with an undoubled IO: Being unmarked, DO will lack a [Person] feature and will only need to check [uCase]. The IO, on the other hand, will need to check case and will also

6.3 A syntactic analysis of Romanian ditransitive configurations

249

be sensitive to [Person], given that it bears a [iPerson: val] feature. DO checks its case against the α head, while IO will have to move to Spec αP so as to get its case checked by vo. By so doing, IO will also check the [uPers:val] of the applicative head: (279)

vP ei v´ ei vcase αP o IO α´ uC ei iPers:val αcase VP uPers:val ei DO V´ ei uC V IO uC iPers:val

The further scrambling of the DO into a position c-commanding the landing position of the IO ensures the surface DO before IO order. Note that movement of the DO seems to be optional in this case. Indeed, if DO scrambling does not take place, the IO before DO word order obtains. The proposed derivation may also account for the facts uncovered experimentally: in this particular configuration, an indefinite DO has been widely assigned a narrow scope interpretation (by 75.43% of the respondents in the undoubled variant and by 75,30% of them in the sequence where IO has been clitic doubled). Only 10,93% and 11.30% of the participants respectively accept a wide scope reading for the DO and 13,63%/13,40% accept both readings. Given that the IO moves to Spec αP, which is a c-commanding position with respect to the merge position of the DO, we would expect IO to outscope DO. The wide scope interpretation of DO remains an option, however, as IO may reconstruct into its low merge position. Let us now consider how the opposite underlying configuration i.e., IO before DO fares in accounting for the same data. The feature specification remains unchanged, with DO in need of verifying [uC] and IO bearing [uC] and [iPerson:_]:

250

(280)

6 Romanian ditransitive configurations

vP ei v’ ei vcase αP ei DO α’ ei uC– VP αcase uPers:val ei V’ IO ei iPers:val V uC DO uC

The DO moves to Spec αP and gets its case feature valued against the v head, while IO enters an agreement relation with αo, checking its own case feature and valuing the [uPerson:val] of α. The DO before IO surface order is arrived at as a consequence of case checking, and no further scrambling needs to be posited as in the case of the previous underlying IO before DO order. The derivation also accounts for the experimentally uncovered scope readings: DO may evince a wide scope interpretation along a narrow scope one, given its possibility to reconstruct in its low merge position. The IO before DO surface word order obtains in this case as a consequence of having scrambled the IO to the vP edge. Interestingly, respondents have a clear cut preference for a wide scope interpretation of the IO in this particular case (both when IO is headed by a universal quantifier as well as when it has been expressed by means of an indefinite DP). An explanation we might want to entertain in this case, might go along the lines of Diesing (1992) i.e., having left the Nuclear Scope domain, the IO acquires a strong, presuppositional reading, hence its strong propensity towards a wide scope interpretation. So far, both accounts seem to fare equally well at capturing the experimental data on scope dependencies. Experiment 2: Scope dependencies with ditransitives (DOMed DO) The norming data regarding the co-occurrence of DOMed DOs and IOs within ditransitive configurations differ to some extent from those uncovered in the first experiment, where only unmarked DOs were employed, in that the respondents have a general preference for the wide scope reading of the DOMed DO. The narrow scope interpretation, however, remains an option.

6.3 A syntactic analysis of Romanian ditransitive configurations

251

As shown in Table 32, in the DO before IO surface order most respondents selected a narrow scope interpretation for the IOIndef irrespective of whether this DP had been clitic doubled or left unmarked (52.67% and 52.64% respectively). When the DO is expressed by means of an indefinite DP, the participants seem to favor a wide scope interpretation (50.21% and 43.68%), which differentiates this context from its counterpart with an unmarked DO from Experiment 1 (where the narrow scope interpretation was generally favoured, 65.33% of the respondents selecting it in the undoubled variant and 68.71% in the doubled condition). The same difference may be noted regarding the IO before DO surface order, where the respondents seem to agree that the DOMed DO should have a wide scope interpretation, both when expressed by means of an indefinite DP (51,5% and 50,86% respectively), as well as when headed by a universal quantifier (45,64% in the undoubled variant vs. 28,59% of respondents who prefer a wide scope reading on the IO). As mentioned above, in the first experiment, the DO was assigned a narrow scope interpretation by most speakers (75,43%/ 75,3% when headed by an indefinite determiner and 60,42%/65,5% when expressed by means of a universal quantifier).

Table 32: Scope dependencies with ditransitives (DOMed DO). Nr. crt.

Word order

DP type

+/− IO clitic

Results

Narrow scope DPindef

Wide scope DPindef

Both readings

.

DO before IO

UQDO & IndefIO no IO clitic

.%

.%

.%

.

DO before IO

UQDO & IndefIO with IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO no IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO with IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO no IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO with IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO no IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO with IO clitic

.%

.%

.%

252

6 Romanian ditransitive configurations

GM DOM DO

DO (UQ) & IO (IND)

DO (IND) & IO (UQ)

DO BEFORE IO

59.75%

WITH IO CLITIC

NO IO CLITIC

WITH IO CLITIC

IO (UQ) & DO (IND)

18.22%

54.09% NO IO CLITIC

22.01%

WITH IO CLITIC

28.13% 17.78%

40.31% 43.68% 16.01%

11.58%

38.21% 50.21% NO IO CLITIC

Series3 29.83% 52.01% 18.71%

WITH IO CLITIC

Series2

50.73% 34.06% 15.21%

NO IO CLITIC

27.25% 20.07%

20.55% 26.83%

52.64%

52.67%

Series1

IO (IND) & DO (UQ)

IO BEFORE DO

Graph 34: Scope dependencies with ditransitives.

Let us first see how a basic DO before IO configuration would fare in accounting for the experimental facts: the DOMed DO is now sensitive to [Person] and bears [iPerson:val] and [uCase]. Depending on whether it has been doubled or not, the IO carries either [iPerson:val] and [uCase] (the undoubled variant) or [iPerson:_] and [uCase] (when clitic doubled – remember that, just like CDed DOs, CDed IOs were argued to have their K further bleached such that this functional projection only contributes a [uPerson: __]. As such the NP bearing [iPers: __] cannot verify its feature and the specification for the entire KP is [iPers: __]). DOMed DO & undoubled IO The α head probes its c-commanding domain in order to check its [uPers:val] feature and DO is the closest suitable goal in this respect. α thus enters an agreement relation with DO and values its [uPers:val]; α also values [uCase] on DO. The IO will be forced to move outside the VP in order to check case. Finally, the DO before IO sequence obtains by scrambling the DO at the vP periphery:

6.3 A syntactic analysis of Romanian ditransitive configurations

(281)

253

vP ei v´ ei vcase αP ep α´ IO uC ei iPers:val VP αcase uPers:val ei V´ DO uC ei iPers:val V IO uC iPers:val

If we were now to consider the IO before DO order as the underlying configuration, the derivation would proceed as follows: IO is probed by α and thereby has its [uCase] valued. α itself values its own [uPerson:val] feature by way of agreement with IO. DO will move to Spec αP and check its case against v. The surface DO before IO order thus obtains as a consequence of case assignment prerequisites. The opposite surface order may obtain by scrambling the IO to the vP edge. (282)

vP ei v´ ei vcase αP ep DO α´ uC ei iPers:val VP αcase uPers:val ei IO V´ iPers:val ep DO uC V uC iPers:val

Both derivations seem to fare equally well in capturing the available scope readings of the two internal arguments.

254

6 Romanian ditransitive configurations

DOMed DO & doubled IO The derivation roughly proceeds as in the previous case up to a certain extent: the DO checks the [uPers:val] feature of the applicative and values its own [uCase] against this head. The IO is compelled to move to Spec αP in order to have its case feature valued by vo. Note, however, that due to its different feature specification with respect to [Person], the IO would have to move even further so as to get its [iPerson:__] valued by Perso. In this particular case, we will have to posit DO scrambling above the ultimate landing site of the IO in order to obtain the DO before IO word order. (283)

Pers' ei Perso [iPers:val]

vP eo IO v´ uC ei αP iPers:__ vcase eo α´ IO ei uC iPers:__ VP α case uPers:val ep DO V´ uC ep iPers:val V IO uC iPers:__

Let us now see whether positing a IO before DO underlying configuration would allow us to provide a more economical account for this particular case: the DOMed DO will thus be merged as a complement of Vo and bear [iPerson: val] and [uCase]. The doubled IO needs to verify both its [iPerson:__] and [uCase] for reasons already extended upon above. The derivation might proceed as follows: α probes its c-commanding domain in search for a suitable goal to check its [uPerson:val] against and, to this end will enter an agreement relation with the IO. The IO will check its [iPerson:__] and [uCase]. The DO needs to move into Specα in order to have its [uCase] feature valued against v.92 The DO before IO surface order obtains as a consequence of case assignment requirements and accounts for the preferred wide scope reading of

92 Note that there is a difference between this derivation and the one proposed in the previous section discussing acceptability results on binding dependencies with ditransitives. For ease of exposition we chose to leave the discussion on the acceptability of certain patterns aside.

6.3 A syntactic analysis of Romanian ditransitive configurations

255

the DO (50.21% and 43.68% respectively). A narrow scope reading of the DO remains an option, given the possibility of this DO to reconstruct in its merge position. The IO before DO surface word order obtains by scrambling IO to the edge of the vP: (284)

vP ei v´ eo v αP eo DO α´ iPers:val ei uC VP αcase uPers:val eo V´ IOcl iPers__ ei uC V DOKP iPerson: val uC

In this particular case the underlying IO before DO order seems to fare better at accounting for the experimental data. The respondents’ preference for the wide scope interpretation of the DOMed DO may also be explained if we posit that a DOMed DO resists reconstruction in its merge position. Remember that, along the lines of López (2012), an account which we also adopted for DOMed DOs, these DPs necessarily move out of VP given that their K layer prevents incorporation of the DO into V. IOs, which may never acquire their case by way of incorporation, may reconstruct into their merge position more easily hence their larger flexibility in acquiring a narrow scope interpretation in the IO before DO order. Experiment 3: Scope dependencies with ditransitives (CDed+DOMed DO) Just like in the case of DOMed DOs, respondents seem to prefer the wide scope interpretation of CDed+DOMed DOs over the narrow scope one: in the DO before IO surface order the wide scope reading of the indefinite DO is selected by 62.27% (in the undoubled variant) and 44.72% (in the variant where the IO has been clitic doubled) of the participants respectively. The same preference for the wide scope interpretation may be observed for the QP DO (59.52%/54.85). The same holds for the IO before DO surface order: the indefinite DO is said to outscope the IO by 51.5%/50.86% of the respondents. The QP DO seems to be

256

6 Romanian ditransitive configurations

preferred on a wide scope interpretation only when there is no dative clitic doubling the IO (45.64%). In the doubled variant, the indefinite IO is assigned a wide scope interpretation by more than half of the respondents (42.82%). Table 33: Scope dependencies with ditransitives (CDed+DOMed DO). Nr. crt.

Word order

DP type

+/− IO clitic

Results

Narrow scope DPindef

Wide scope DPindef

Both readings

.

DO before IO

UQDO & IndefIO

no IO clitic

,%

.%

.%

.

DO before IO

UQDO & IndefIO

with IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO

no IO clitic

.%

.%

.%

.

DO before IO

IndefDO & UQIO

with IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO

no IO clitic

.%

.%

.%

.

IO before DO

UQIO & IndefDO

with IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO

no IO clitic

.%

.%

.%

.

IO before DO

IndefIO & UQDO

with IO clitic

.%

.%

.%

CDed+DOMed DOs & undoubled IOs Let us first consider how an account based on a DO before IO underlying configuration captures the experimental data. Remember that in this particular case, DO has the feature specification [uCase] and [iPerson: ___] and that IO bears [uCase] and [iPerson: val]. The derivation proceeds as follows: α probes

6.3 A syntactic analysis of Romanian ditransitive configurations

257

GM CD + DOM DO

DO (UQ) & IO (IND)

DO (IND) & IO (UQ)

DO BEFORE IO

11.87%

36.61% 51.50% NO IO CLITIC

38.48% 42.82% 18.65%

WITH IO CLITIC

Series3

45.64% 28.59% 25.75%

NO IO CLITIC

14.50%

10.65%

27.06%

40.76% 44.72%

62.27%

WITH IO CLITIC

Series2

36.78% 50.86% 12.36%

NO IO CLITIC

27.15% 17.97%

17.83% 22.61%

54.85%

59.52%

Series1

WITH IO CLITIC

NO IO CLITIC

WITH IO CLITIC

IO (UQ) & DO (IND)

IO (IND) & DO (UQ)

IO BEFORE DO

Graph 35: Scope dependencies with ditransitives.

its c-command domain and finds DO to be a suitable goal. It will thus check its [uPerson: val]; DO itself will value its [uC]. IO needs to check case and will move to Spec αP and have its [uC] feature valued by v. The DO before IO surface order obtains by scrambling the DO to the edge of vP. This derivation accounts for all the scope reading possibilities uncovered experimentally: both IO and DO may exhibit wide and narrow scope interpretations due to the fact that they may both reconstruct to lower merge positions if need be. (285) v´ ei vcase αP ei IO α´ iPers:val ep αcase VP uC uPers:val ei DO V´ iPers:__ eo IO uC V iPerson: val uC If we were to adopt an IO before DO underlying configuration, the derivation would proceed in the following way: the α head checks its [uPerson] against IO,

258

6 Romanian ditransitive configurations

given that this DP is now closer than DO. The IO itself will value its case feature against α. DO will have to move to SpecαP and value its case feature against v. Note that this DP will have to scramble even further given that it still needs to value its [iPerson: __] feature. This is how the surface DO before IO obtains. For the opposite surface order, one needs to further scramble IO into a position c-commanding the landing site of the DO: (286)

Pers´ ep vP Perso [iPers:val] ei v´ DOcl iPers:__ ei vcase αP ei DOcl α´ eo iPers:__ αcase VP uC uPers:val eo IO V´ ei iPers:val V DOcl uC iPerson: __ uC

CDed+DOMed DOs & doubled IOs In this particular case, both the DO and the IO bear [uCase] and [iPerson: ___] as feature specification. α agrees with the DO and checks its [uPerson:val], also valuing the [uCase] feature of this DP. The IO will have to move to Spec αP in order to value its case feature and even further, to the edge of vP to also have its Person feature valued by the Person head.

6.3 A syntactic analysis of Romanian ditransitive configurations

(287)

259

PersP ei Pers’ ep Perso vP [iPers:val] eo IO v’ iPers:___ ei vcase αP eo IO α’ iPers:___ ei uC αcase VP uPers:val wu DO V’ iPers:__ eu IO V uC iPerson: __ uC

The derivation just described enables us to obtain the surface order IO before DO. The narrow and the wide scope readings of the IO and DO may obtain by interpreting the IO either in its landing site or in its merge position. The DO before IO surface order obtains by scrambling the DO past the final landing site of the IO. Consider now the IO before DO underlying configuration under the same feature specification: α probes its c-commanding domain and finds IO as a suitable goal, checking its own [uPers] feature as well as the [uCase] of the IO. The DO will then move to Spec αP to get its case feature valued by v but will have to move even further so that it may also check its unvalued [Person] feature. The DO before IO order obtains as a consequence of feature checking requirements and the availability of both a wide and a narrow scope interpretation with the two internal arguments is accounted for in the usual way i.e. DO may reconstruct. Furthermore, the observed tendency of the participants to assign wide scope readings to the CDed+DOMed DO may be explained along the lines of López (2012), in terms of resistance to reconstruction as above:

260

6 Romanian ditransitive configurations

(288)

Pers’ ep Perso vP [iPers:val] eo DOcl v’ iPers:___ ei αP vcase eo IO α’ iPers:___ ei uC αcase VP uPers:val wu IOcl V’ iPers:__ eu uC IO V iPerson: __ uC

6.3.7.3 Conclusions Unlike Japanese, where the existence of scope freezing with the Goal-overTheme configuration accounts for this order as being the basic one, Romanian allows for wide and narrow scope interpretations with both internal arguments irrespective of their ordering one with respect to the other. The DO type employed i.e., whether it is unmarked vs. DOMed or CDed+DOMed seems to have a say in the matter, in that participants tend to assign a wide scope reading to the marked DOs irrespective of the surface word order. It was argued that in the case of these DPs, exiting their merge position within VP is mandatory, given that their K layer hinders case checking via incorporation (along the lines of López 2012). This was further linked to the resistance of these DPs to getting reconstructed in their merge position. Unmarked DOs, on the other hand, are not forced to move out of VP. Remember that, unlike their Spanish counterparts, which seem to always incorporate into V for case reasons, Romanian unmarked DOs were shown to act either as DPs and move or as NPs, in which case they incorporate. Importantly, they evince no K layer, which might account for the fact that they may easily reconstruct in their merge position. Although both the DO before IO and the IO before DO sequences were examined as possible candidates for the underlying configuration with ditransitives, the latter order was claimed to fare better in this respect given the somewhat more econominal derivation to which it gives rise and the fact that it may also offer an explanation for the propensity of respondents to select the wide scope interpretation for the marked DOs.

6.3 A syntactic analysis of Romanian ditransitive configurations

261

6.3.8 Secondary objects Some Romanian ditransitives subcategorize for two Accusative objects (289). One peculiar characteristic of this construction is that DP2 is always DOMed and denotes referents which rank very high on the animacy / definiteness scales (Aissen 2003), while DP3 points to inanimate referents and never gets DOMed. DP3 has also been labelled as ‘a secondary object’ (Pană Dindelegan 1974, GA 2005, Pană Dindelegan 2016 a.o.). The DOMed DP usually functions as a Goal or Patient, while the secondary object is interpreted as a Theme. (289) DP1 V pe DP2 (Acc) DP3 (Acc) The verbs allowing this configuration form a relatively small group, including: a asculta (‘listen/examine’), a examina (‘examine’), a învăţa (‘teach/learn’), a sfătui (‘advise’), a întreba (‘ask a question’), a ruga (‘ask somebody to do something’), a trece (‘pass’), a traversa (‘to cross’), a vesti (‘inform’), a anunţa (‘announce’): (290) a. Profesorul nu l-a trecut clasa Professor.DEF.ART. not CL.3.ACC.M.SG.-has passed grade.DEF.ART. pe Ion. DOM John ‘The profesor did not allow John to pass into the next grade.’ b. I-am anunțat pe prieteni ora CL.3.DAT.PL.-have.1.SG. announced DOM friends hour.DEF.ART. plecării. departure.GEN ‘I announced the departure time to my friends.’

6.3.8.1 Properties The double accusative configuration evinces several properties: firstly, only the pe marked DP may be clitic doubled (291a). This is expected, since the secondary object always denotes inanimate referents. In addition, the secondary object may not be cliticized (291b) nor clitic left dislocated. (291) a. Maria l-a învățat germană pe Mihai. Mary CL.3.ACC.M.SG.-has taught German DOM Michael ‘Mary taught Mihai German.’

262

6 Romanian ditransitive configurations

b. *Maria a învățat-o pe Mihai. Mary has taught-CL.3.ACC.FEM.SG. DOM Michael ‘Mary taught it to Mihai.’ c. *Germana, a învățat-o Maria pe German.DEF.ART. has taught-CL.3.ACC.FEM.SG. Mary DOM Mihai. Michael ‘As for German, Mary has taught it to Mihai.’ The secondary object may be substituted by [-animate] demonstratives, interogative pronouns and indefinite pronouns: asta (‘this’), aia (‘that’), ce (‘what’), ceva (‘something’), orice (‘anything’), etc.:93 (292) a. Îl învăț asta. CL.3.ACC.M.SG teach.1.SG. this ‘I am teaching him this.’ b. Ce l-au anunțat? What CL.3.ACC.M.SG.-have.3.PL announced ‘What did they announce him?’ c. Vrea să te întrebe ceva. Wants SUBJ. you.ACC ask something ‘He wants to ask you something.’ Secondary objects may take wide scope with respect to another scope bearing expression and may also be read specifically: (293) a. L-am învățat pe fiecare elev o CL.3.ACC.M.SG.-have.1.SG. taught DOM every pupil INDEF.ART. poezie. poem ‘I have taught every pupil a poem.’ a. ∀ > 9 b. 9 > ∀

93 For a more extensive discussion of the properites of the secondary object consider Gramatica Academiei (2005) or Pană Dindelegan (2016).

6.3 A syntactic analysis of Romanian ditransitive configurations

263

b. L-am întrebat pe fiecare student CL.3.ACC.M.SG.-have.1.SG. asked DOM every student o întrebare. INDEF.ART. question ‘I asked every student a question.’ a. ∀ > 9 b. 9 > ∀ The data with respect to binding fare differently, however, showing that the DOMed DO occupies a c-commanding position with respect to the secondary object: while the examples in (294) are felicitous under a reading accoding to which there is pairing between the two internal arguments, the examples in (295) are infelicitous under a similar interpretation: (294) a. Nu l-am traversat pe nici un bătrâni Not CL.3.ACC.M.SG.-have.1.SG. crossed DOM no old man strada luii din fața casei. street.DEF.ART. his from front house.GEN ‘I have not helped any old man cross the street in front of his house.’ b. Am ascultat pe fiecare elevi în parte poezia Have.1.SG. listened DOM every pupil in part poem.DEF.ART. acasă. luii, învățată de his learned from home ‘I examined every pupil in part his lesson learnt from at home.’ autorul eii, fără (295) a. *Am ascultat fiecare poeziei pe Have.1.SG. listened every poem DOM author its without să mă plictisesc. SUBJ. REFL.1.SG. get bored Lit. ‘I listened its author every poem without getting bored.’ b. *Comisia a ascultat pe interpretul luii Commission.DEF.ART. has listened DOM interpreter.DEF.ART. its a decis la final câștigătorul. fiecare șlagări și every sond and has decided at end winner.DEF.ART. ‘Lit. The commission listened its interpreter every song and decided on the winner at the end.’ The data with respect to the binding dependencies between the two objects prompt us to hypothesize that the secondary object merges low within the VP and that it is c-commanded by the DOMed DO. In the following section we will

264

6 Romanian ditransitive configurations

propose an analysis of the double accusative structure by drawing on the similarities between these configurations and ‘regular’ ditransitives. 6.3.8.2 Deriving Ditransitives with two Accusatives We take ditransitives with two accusative objects to pattern similarly to ‘regular’ ones i.e., ditransitives subcategorizing for an accusative DO and a dative IO. As with ‘regular’ ditransitives and in line with Larson (1988, 2010 ff), we take both internal arguments to merge within the VP and be part of the verb’s argument structure. Given the binding data presented in the previous section, we take the secondary object to merge as the complement of the VP, while the DOMed DO will be hypothesized to merge in the specifier of VP, similarly to the IO in ‘regular’ ditransitives. (296) shows this at stake: (296)

VP eo V´ DOMed DO eo V DO

An expletive Appl is also posited, taking VP as its complement. As already established, the Appl may assign case and also bears [uPerson]: (297) ApplP 3 Appl´ ei Appl° VP [uPers] eo [Case: _] DOMedDO V´ [uC] 3 DO V° [uC] As to the feature specification of the two DOs, the following difference is at stake: given that it may only denote inanimate referents, the secondary object will not carry a [Person] feature; it will thus only have to value its [uCase] feature. The DOMed DO, on the other hand does carry a [Person] feature, given its sensitivity to the animacy/definiteness hierarchy. As already established there will be a difference with respect to the featural make-up of the DOMed DO depending on whether it has been clitic doubled or not:

6.3 A syntactic analysis of Romanian ditransitive configurations

265

– undoubled DOMed DOs: the internal structure of DOMed DOs presupposes the existence of a KP layer, where K is triggered by an unvalued syntactic [iPerson:__] feature present in the NP which is then copied in D. The NP itself is a [+Human] denoting nominal and as such may incorporate the [iPerson] feature. The presence of the syntactic unvalued [iPerson] feature triggers the merger of K, bearing a valued [uPerson], which verifies the unvalued feature on D. The entire KP ends up bearing a valued [iPerson] in the end. Example (298) below shows this: (298) KP ep DP ep D NP +p +N +d uφ:__ uφ:__ u-person:val idef: __ u-def iperson: ___ i person: ___ +HUM pe un elev K

– CDed+ DOMed DOs: the DOM marker is further bleached and no longer carries a valued syntactic [uPerson] feature, but is merely specified as [uPerson:___]; its [Person] feature is in need of valuation, just like the one on the nominal it precedes. The KP has the feature specification [iPerson: ___] (299) and will have to find a way whereby to value this feature. (299) KP ep K +p uφ:__ uperson: ___ uCase__

DP ep D +d uφ: __ idef: __ iperson: ___ uCase __

pe

un

NP +N iφ:val u-def i person: ___ uCase __ +HUM coleg

266

6 Romanian ditransitive configurations

We are now in a position to proceed with the syntactic account of these constructions. Let us start from example (300a) below, where the DOMed DO pe Ion bears no clitic doubling: in (300b), the secondary object lesson has been merged as a complement of V and bears [uC] in need of checking. The DOMed DO pe Ion occupies the SpecVP position and carries [iPerson:val] and [uCase]. The Appl head itself, carries [uPerson:val]. (300) a. Am învățat pe Ion lecția. Have.1.SG. taught DOM John lesson.DEF.ART. ‘I have taught John the lesson.’ b.

vP ep DPSu v’ ei [Case: Nom] v° ApplP [Case__] ep DODOM Appl’ [uC] ro [iPerson] Appl° VP eo [uPers] [Case: _] DODOM V’ [uC] ei [iPerson] V° DO [uC]

Appl will probe for a suitable Goal so as to value its [uPerson]. It encounters the DOMed DO and checks this feature against it. The DOMed DO will move to the specifier of ApplP and check case against the v head. The secondary object will value its case by incorporating into V along the lines of López (2012). Consider now the case of an example containing a CDed+DOMed DO and a secondary DO as (301a) below: as in the previous examples, the secondary DO merges as the complement of V and needs to value its [uC] feature. The CDed+DOMed DO carries the feature specification [iPerson: ___] this time, as discussed above, and needs to also verify case. (301) a. L-am învățat pe Ion lecția. CL.3.ACC.M.SG.-have.1.SG. taught DOM John lesson.DEF.ART. ‘I have taught John the lesson.’

6.3 A syntactic analysis of Romanian ditransitive configurations

b.

267

PersP eu Pers´ eo vP Perso [iPers] tp v´ DODOM [iPerson:__]1 ep pe Ion DPSu v´ [Case: Nom] ro v ApplP [Case__] eo DODOM Appl´ [uC] ri iPerson:__]1 pe Ion

Appl° [uPers]1 [Case: _]

VP to DODOM V´ tu [uC] [iPerson:__] V° DO [uC] pe Ion lecția

The derivation also has to capture the essential difference between CDed and non-CDed DPs, in that only the former pass through a vP external position on the way to the T-field (Dobrovie Sorin 1994, see agument from Binding of the CDed DO into the Subject DP in the preceding sections). Building on Preminger (2017), we take CD to be an instance of long head movement, with the DP moving into a specifier of the vP from whence cliticisation arises. The Person head carries a [iPerson] and serves as an escape hatch for constituents which need to check Person in T so that they are not left inside the lexical phase, when the vP spells-out. The DOMed DP will be attracted to the SpecvP and will check its [iPerson:___] feature against the Person head. The derivation proceeds in the following way: Appl probes its c-command domain in search for a suitable Goal against which it might value its [uPerson]; it finds the DOMed DO, and agrees with it, but this time its [uPerson] does not get deleted, since the corresponding feature on the targeted DO is also unvalued. The two features are related by agreement and count as instances of the same feature (Pesetsky and Torrego 2007). The DO further moves to the specifier of Appl where it checks case against the v head. The secondary DO checks case against Appl. Crucially, after agreement with Appl, the DOMed DO is still left with an unvalued person feature and this will force movement to the PersonP, at the vP periphery. Having reached SpecvP, the DP checks its [iPerson:__] feature against Person°. The [uPerson] of Appl° also gets valued, as it has been related by

268

6 Romanian ditransitive configurations

means of agreement with the [Person] feature of the DPDOM and counts as an instance of the same feature. Nevertheless, the undoubled DPDOM now counts as an intervener between the T head and the subject DP, which cannot get case licensed unless the DPDOM is removed. CD is thus resorted to in line with Preminger (2017). Cliticisation is thus a way to remove the DP object blocking agreement between T and the Subject DP. The DP object obligatorily cliticize on T, enabling the DP subject to get its case checked. As already discussed in the previous sections, the mechanism of CD is neither optional nor conditioned by the properties of the direct object DP but occurs as a means to remove an intervening DP blocking agreement of a lower DP with a higher functional head. The fact that the DP object has moved outside its merge projection and reached SpecPersonP comes as a conquence of its featural make-up i.e., its having an unvalued [iPerson] feature in need of valuation.

6.4 Conclusions This chapter was devoted to Romanian ditransitive configurations and had a two-fold aim. On the one hand, we presented the findings of several experiments checking binding dependencies with these constructions, and, on the other hand we provided a derivational account of ditransitives in light of these experimental results. The analysis we proposed goes counter to the alternative projection account established in the literature, but seems to offer more flexibility in accommodating for the experimentally uncovered facts. The following aspects proved relevant regarding Romanian ditransitive configurations: 1. Surface word order is crucial in determining the acceptability of binding relations: cases where surface word order concurred with the direction of binding fared much better with respect to acceptability than those where inverse binding obtained. 2. Dative clitic: Sentences where the IO is clitic doubled fare worse than their undoubled counterparts irrespective of the direction of binding and word order. 3. Generally average acceptability rates: ratings generally target the middle of the acceptability scale, with most values concentrating around the 3–4 acceptability interval. 4. Symmetric binding: both directions of binding have generally received similar acceptability scores. They seem to be acceptable irrespective of presence/absence of IO clitic doubling.

6.4 Conclusions

269

5.

Ditransitives with DOMed DOs have lower acceptability rates than their counterparts containing unmarked DOs or CDed DOs. 6. There is even lower acceptability rate of those patterns where the differential marker interacts with the dative clitic. This effect was not visible with the counterparts of these patterns in Experiment 1 or Experiment 3, which strengthens the hypothesis of an intervention effect between pe and the dative clitic. 7. Accusative CD seems to function as a repair strategy: configurations containing DOMed DOs and CDed IOs fare much better with respect to acceptability than their counterparts where the DOMed DO has not been doubled.

These experimentally uncovered facts were accounted for by building on the initial intuitions in Cornilescu et al. (2017a) concerning the basic configuration of ditransitives. More specifically, we adopted a derivational account according to which dative DPs merge low within VP. Thus, we take the basic configuration with Romanian ditransitives to be one in which the Goal DP is part of the verb’s argument structure. In line with Larson (2010)’s view, IO is actually part of the verb’s θ-grid. It is introduced by the lexical verb itself and composes inside VP in a syntax similar to that of Larson (1988). Applo is required to have the lower lexical VP as a complement. The DO also merges within the VP in complement position. We further proposed that there exists a priority for feature verification between the two objects. More specifically, the DO generally has priority over the IO but that priority may change function of the feature specification of the two objects. This allowed us to account for all the existing patterns found to be acceptable experimentally. Note also that this chapter also accounted for the problematic instances where a DOMed DO interacts with a clitic doubled IO. The analysis offered built on the internal featural make-up of the two internal arguments and showed that the interaction between DOMed DOs and clitic doubled IOs boils down to a locality issue: VAppl, which may match both nominals in its c-commanding domain in what the valuation of its [uPerson] feature is concerned, may only do so with the higher object, in our case the IO. Finally, an analysis of ditransitives containing two internal arguments surfacing in the accusative case was also provided along the same lines.

7 General conclusions This book has had a two-fold aim: it has put forth a syntactic analysis of Romanian ditransitive configurations and, moreover, it has done so by drawing on a wide range of data uncovered experimentally. In this respect, it counts as the first extensive experimental study on these constructions in Romanian, thus enabling a more accurate perspective on various aspects related to these configurations and considered to be crucial in the literature cross-linguistically: the hierarchical dependencies between the two internal arguments, the existence of dative alternation, the structural import of the dative clitic doubling the IO a.o. Our investigation started with an inquiry into the syntax of one of the two internal arguments within ditransitive configurations, i.e., the DO. We differentiated between three types of DOs given the fact that Romanian allows for both Differential Object Marking as well as Clitic Doubling with DO DPs: unmarked DOs, DOMed DOs and CDed+DOMed DOs. In this, Romanian patterns with other Romance languages and in particular with Spanish, which also relies on DOM and CD to mark more prominent DOs. Furthermore, marked DOs may acquire a specific interpretation and outscope other scope bearing expressions in both languages. They are further disallowed in contexts selecting property denoting nominals (e.g., the language specific counterparts of have) and surface in a number of environments where only nominals with real argumenthood are allowed: small clauses, object control and clause union in causative configurations. Romanian, however, differentiates itself from Spanish with respect to unmarked DOs in that it seems to group marked and unmarked indefinite objects together when it comes to the aforementioned contexts and to distinguish between these DPs on the one hand and bare plurals on the other. Romanian unmarked DOs may thus acquire specific and wide scope readings and may also surface as arguments in small clauses, object control and clause union in causative configurations. Spanish unmarked DOs pattern differently from their marked counterparts and on a par with bare nominals: they do not acquire a specific interpretation, they do not outscope other scope bearing expressions and never surface within small clauses, object control and clause union in causative configurations. These observations were backed up by two experiments checking the ability of the three DO types to acquire specific and wide scope interpretations. The first experiment tested the behaviour of marked and unmarked indefinite DOs when co-occurring with other scope bearing expressions i.e., phrases headed by the extensional quantifier fiecare (every), two intensional operators, the conditional https://doi.org/10.1515/9781501513657-007

272

7 General conclusions

and the modal verb trebuie (must), and negation. As a general conclusion, we observed that while marking seemed to render the respective DO a more suitable candidate for outscoping the other scope bearing expressions, the possibility of acquiring wide scope also remained an option for unmarked DOs. Similarly, the second experiment uncovered the fact that both marked and unmarked DOs pattern somewhat alike with respect to specificity: all DO types lose their otherwise possible specific interpretation when used within contexts forcing a non-specific reading. Even though marking seems to render the respective DO a less suitable candidates for these contexts, neither DOMed DOs nor CDed+DOMed DOs are actually excluded as unacceptable. This shows that both types of marked DOs may read non-specifically and that Romanian does not possess a category of objects which are necessarily interpreted as specific, a conclusion also reached for Spanish by López (2012). Building on the uncovered experimental results, we proposed an analysis of Romanian DOs in line with López (2012) by parametrizing this account in such a way as to accommodate for the differences holding between Romanian and Spanish. Thus, just like their Spanish counterparts, marked DOs are analysed as nominals evincing real argumenthood: the DOM marker guarantees an e type reading on the nominal it marks, thereby securing argumenthood and individuation. Consequently, marked DPs may not stay in their merge position as complements of V, since this is a position that only property denoting nominals may occupy, which compose with the predicate via incorporation. Marked DPs have to scramble to a landing position (i.e., Spec αP, a landing site situated between vP and VP) where they may compose with the predicate via Function Application, as expected for nominals functioning as real arguments. The landing site secures a peculiar mode of semantic composition and a mode of interpretation of the scrambled DO (via a choice function). The specific interpretation will thus obtain as a possible reading deriving from this mode of semantic composition and not as a consequence of DOM marking. CDed+DOMed DOs scramble even further than their DOMed counterparts, passing through the intermediate landing site reached by DOMed DOs. These DOs also get composed with the predicate in the same way as their DOMed counterparts, which allows us to expect the same readings with both marked variants. Note that what differentiates between Romanian and Spanish at this point is the behaviour of their unmarked DOs. Spanish counterparts evince a behaviour which prompts López (2012) to posit that these DPs never leave their merge position, incorporating into V. As such these DPs get composed by means of Restrict and hence never acquire specific or wide scope readings. López (2012) links DO behaviour to the status of the respective nominal: unmarked DOs have DP status and as such incorporate; marked DOs have an

7 General conclusions

273

extra KP layer and must scramble given that K blocks incorporation into V. The parameter differentiating between Romanian and Spanish in this respect then lies in the types of nominals which have to scramble: Spanish draws a line between KPs and DPs, only allowing the former to scramble, while the latter incorporate, whereas Romanian allows both KPs and DPs to scramble, as opposed to bare nominals which incorporate just like their Spanish counterparts. The relevant division in this case is between NP/NumPs which incorporate and DPs/KPs which scramble. One last point of discussion with respect to Romanian DOs has to do with the featural load of the three DO types: marked DOs are analysed as KPs, where the projection of K is triggered by the presence of an unvalued [Person] feature in N, which further projects in D. The [Person] feature itself represents the syntactic reflex of the animacy and definiteness hierarchies in line with Richards (2008). The DOM marker is the spell out of an uninterpretable [Person] feature and its distribution is determined by the local context, following from the interplay of the syntactic features of [person] and [definiteness]. Thus, [+Human] denoting nominals may surface as marked DOs (KPs) or unmarked objects. In the latter case, these objects pattern with their [-Human] nominals, eliciting no specification as to a syntactic [Person] feature. [+Human] is simply a lexical property of the NP. On the other hand, when these nominals are syntactically marked as denoting people, they incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by a merging K. As pointed out, K bears a valued [uPerson] feature. The possibility of incorporating [Person] is only available to NPs that have the lexical feature [+Human] or rather [+Human (like)], given that other types of nouns (e.g. pets etc.) may be treated like humans and may become marked objects. Marked DOs are thus KPs with an additional [iPerson] feature in N. The presence of this unvalued feature in N and then in D triggers the insertion of K. Note that their more complex structure associated with their scrambling syntax, makes marked DOs more “prominent.” Unmarked DOs, on the other hand, exhibit a standard DP structure. Just like DOMed DOs, CDed+DOMed DOs also exhibit a KP layer. Nevertheless, the K within these nominals we argued to have undergone further bleaching in that the syntactic [uPerson] feature it contributes is unvalued (as opposed to DOMed DOs where K contributes a valued uninterpretable [Person] feature). As a consequence, the [iPerson:__] feature carried by the nominal DP complement may no longer be valued by way of agreement with the KP. An immediate result for these DOs is that they need to find a way whereby to check their [Person] feature and will thus have to scramble even further than DOMed DOs, up to an outer specifier of vP where they would be able to enter agreement

274

7 General conclusions

with the Person head carrying [iPerson: val]. Experimental results concerning binding with the Subject DP support this hypothesis given that CDed+DOMed DOs may bind into the Subject irrespective of word order, while undoubled counterparts seem not to fare as well with respect to this binding dependency. Having discussed the case of Romanian DOs, we focused next on the IO within Romanian ditransitive configurations with the aim of uncovering the featural make-up of these nominals. Thus, in line with DOMed DOs, Goal DPs were shown to exhibit high sensitivity to the animacy hierarchy in both their inflectional and prepositional form: only nouns ranging high on this hierarchy count as possible realizations of the dative θ-roles. Other categories of nouns are ruled out (e.g. abstract nouns), in principle, though metaphorical and idiomatic exceptions may arise. Given that IOs pattern with DOMed DOs in this respect, which have also been shown to evince the same sensitivity to animacy, one natural proposal was to posit a [Person] feature for IO DPs on a par with DOMed DOs. Also, IOs were also granted KP status: when surfacing as prepositional datives, the la marker functions as the functional category labelled K in line with the DOM marked pe; inflectional datives are also KPs, with a silent K. Goal nominals, which are [+Human(like)] NPs, thus incorporate an unvalued syntactic [iPerson] feature, which is copied in D and valued by merging K. K itself carries a valued [uPerson] feature. The entire KP will have the feature specification [iPerson: val]. Clitic doubling of IO has the same effect as with the doubling of DOMed DOs: K is further bleached only carrying an unvalued [Person] feature [uPerson:___]. The entire KP ends up having the feature specification [iPerson: ___] in this particular case and needs to undergo scrambling up to the Person field on a par with CDed+DOMed DOs. Establishing the featural make-up of the two internal arguments, enabled us to further investigate the syntax of ditransitive configurations and provided us with a useful tool when it came to account for the language facts uncovered experimentally. The theory of cliticisation in line with Preminger (2017) we adopted also proved useful in this respect. The proposal we advanced with respect to the syntax of Romanian ditransitives relied on a series of experiments checking binding dependencies with the two internal arguments. To this end, both quantificational dependencies with possessives and with anaphors were tested. Furthermore, the impact of parameters such as clitic doubling of both the DO and the IO, differential object marking of the DO, surface word order of the two internal arguments, and the direction of binding also came under scrutiny. This experimental undertaking aimed at uncovering the underlying configuration(s) with Romanian ditransitives, which had been previously argued to

7 General conclusions

275

pattern with their English counterparts in allowing both for a configurational equivalent of the English Prepositional Object Construction (where the DO c-commands the IO), as well as for a structural counterpart of the English Double Object Construction (where the IO c-commands the DO) (Diaconescu and Rivero 2007). Note that this complete parallelism with English has not been only maintained for Romanian, but has been also claimed to hold for other Romance languages as well (Spanish – Demonte 1995, Cuervo 2003a). Contrary to the existing discussions on Romanian ditransitives (Diaconescu and Rivero 2007), the experimental findings revealed that the two internal arguments evince symmetric binding abilities in the sense that each of them may bind a possessor or an anaphor into the other. It was, moreover, discovered that this binding potential is not affected by the presence of a dative clitic doubling the IO as formerly claimed. Surface word order, on the other hand, was shown to influence binding: higher acceptability scores were assigned to those items where binding direction matched word order. Reverse binding, however, was also proved to be possible. The symmetric binding potential of the two internal arguments together with the fact that the dative clitic appears to have no structural import in this respect encouraged us to posit one single underlying configuration for ditransitives, contra the alternative account proposed in Diaconescu and Rivero (2007) which differentiates between two configurations function of whether the IO has been clitic doubled or not (302): (302) a. POC: Theme c-commands Goal [VoiceP DPAgentVoice[vPv [PPDPTheme P DPGoal]]] b. DOC: Goal c-commands Theme (clitic doubling) [VoicePDPAgentVoice[vPv [ApplPDPGoal [clAppl] [VP V DPTheme]]]] (Diaconescu and Rivero 2007: 219–220) Further problematic aspects with respect to the interaction between DOMed DOs and CDed IOs prompted us to look into the featural make-up of these DPs as a possible explanation for the lower acceptability of the respective items. Indeed, the problem seemed to boil down to a locality issue: DOMed DOs seem to act as interveners, hindering CDed IOs from checking their [uPerson] feature against the VAppl. The similar featural make-up of the two nominals rendered them as suitable goals for the Appl, with the DO entering agreement with this head given its closer proximity. As a final part of our investigation, we adapted the proposed account to ditransitive configurations containing two internal arguments surfacing in the accusative.

276

7 General conclusions

While drawing away from the initial proposals for Romanian ditransitives, the derivational account we put forth in this book fares better at accommodating the experimentally uncovered data and pattern with more recent analyses provided for these constructions in other Romance languages (Pineda 2012, 2013 for Spanish a.o.; see also Giorgi and Longobardi 1991 and Harley 2002 arguing in favour of bi-directional c-command within ditransitives in Italian and French respectively).

Appendices

Appendix 1: Three experiments on quantificational dependencies with ditransitives (binding with possessives) Experiment 1 (Datenset: input-ditransitives1; ditransitives1.out) Statistical analyses were conducted in R version 1.0.136 using the lme4 package (Bates et al., 2015) to perform linear mixed-effect models (LMEM) with the score as outcome variable. As fixed effects, we entered word order, Binding and Clitic Marking into the model. As random effects, we had intercepts for subjects and items. The word order DO before IO condition, Binding DO binds into IO condition and the Clitic Marking no clitic condition were mapped onto the intercept. To identify the best model fit we performed likelihood ratio tests. This revealed that the full model with a three-way interaction affected the acceptance rate ( χ2 (4) = 130.33, p < .001).94

Estimate

Std. Error t

.

.

.*

woIO before DO

−.

.

−.*

bindingIO binds into DO

−.

.

−.*

cliticwith CL

−.

.

−.*

woIO before DO:bindingIO binds into DO

.

.

.*

woIO before DO:cliticwith CL

.

.

.*

bindingIO binds into DO:cliticwith CL

.

.

.*

−.

.

−.*

(Intercept)

woIO before DO: bindingIO binds into DO:cliticwith CL

*t-values greater than 2 and less than −2 are considered significant.

94 Formulars for model comparison: (1) full model: lmerModLmerTest (score ~ wo: binding: clitic + (1|subj) + (1|item)) (2) reduced model: lmer(score ~ wo + binding + clitic + (1|subj) + (1|item)) https://doi.org/10.1515/9781501513657-008

t value

280

Appendices

Experiment 2 (Datenset: input-ditransitives2; ditransitives2.out) [. . .] This revealed that the full model with a three-way interaction affected the acceptance rate (χ2 (4) = 36.21, p < .001). Estimate

Std. Error t

t value

.

.

.*

woIO before DO

−.

.

−.*

bindingIO binds into DO

−.

.

−.*

cliticwith CL

−.

.

−.*

woIO before DO:bindingIO binds into DO

.

.

.*

woIO before DO:cliticwith CL

.

.

.*

bindingIO binds into DO:cliticwith CL

.

.

.*

−.

.

−.*

(Intercept)

woIO before DO:bindingIO binds into DO:cliticwith CL

*t-values greater than 2 and less than −2 are considered significant.

Experiment 3 (Datenset: input-ditransitives3; ditransitives3.out) [. . .] This revealed that the full model with a three-way interaction affected the acceptance rate (χ2 (4) = 100.35, p < .001). Estimate

Std. Error t

t value

.

.

.*

woIO before DO

−.

.

−.*

bindingIO binds into DO

−.

.

−.*

cliticwith CL

−.

.

−.*

woIO before DO:bindingIO binds into DO

.

.

.*

woIO before DO:cliticwith CL

.

.

.*

bindingIO binds into DO:cliticwith CL

.

.

.*

−.

.

−.*

(Intercept)

woIO before DO:bindingIO binds into DO:cliticwith CL

*t-values greater than 2 and less than −2 are considered significant.

Appendix 2: Quantificational dependencies with ditransitives

281

Appendix 2: Quantificational dependencies with ditransitives (binding with anaphors) (Datenset: input-anaphor; anaphor.out) Statistical analyses were conducted in R version 1.0.136 using the lme4 package (Bates et al., 2015) to perform linear mixed-effect models (LMEM) with the score as outcome variable. As fixed effects we entered word order, binding direction and clitic doubling into the model As random effects, we had intercepts for subjects and items. The word order DO before IO condition, Binding DO binds into IO condition and the Clitic Marking no clitic condition were mapped onto the intercept. To identify the best model fit we performed likelihood ratio tests. This revealed that the full model with a three-way interaction affected the acceptance rate (χ2 (4) = 11.01, p < .05).95 Estimate

Std. Error t

Intercept

.

.

.***

woIObef

−.

.

−.

bindingIObin

−.

.

−.*

cliticCLyes

−.

.

−.

woIObef:bindingIObin

.

.

.**

woIObef:cliticCLyes

.

.

.

bindingIObin:cliticCLyes

.

.

.

−.

.

−.

woIObef:bindingIObin:cliticCLyes

*t-values greater than 2 and less than −2 are considered significant.

95 Formulars for model comparison: (1) full model: lmer(score ~ wo: binding: clitic + (1|subj) + (1|item)) (2) reduced model: lmer(score ~ wo + binding + clitic + (1|subj) + (1|item))

t value

282

Appendices

Appendix 3: DOs in contexts inducing non-specificity (Datenset: input-specificity; specificity.out) Statistical analyses were conducted in R version 1.0.136 using the lme4 package (Bates et al., 2015) to perform linear mixed-effect models (LMEM) with the score as outcome variable. As fixed effects we entered context and DO type into the model As random effects, we had intercepts for subjects and items. The context cel mult/cel puțin and DO type unmarked DO were mapped onto the intercept. To identify the best model fit we performed likelihood ratio tests. This revealed that the full model with a two main effects affected the acceptance rate (χ2 (4) = 6.34, p < .05).96

Estimate

Std. Error

.

.

.*

contextcate (cel)

−.

.

−.

contextmood (cel)

−.

.

−.

contextoarecare (cel)

−.

.

−.

contextmood (cate)

.

.

.

contextoarecare (cate)

.

.

.

contextmood (oarecare)

.

.

.

DOtypeC (U)

−.

.

−.*

DOtypeD (U)

−.

.

−.*

DOtypeD (C)

−.

.

−.*

Intercept

t value

*t-values greater than 2 and less than −2 are considered significant. Factors in brackets indicate the intercept.

96 Formulars for model comparison: (1) full model: lmerHE(score ~ context + DO type + (1|subj) + (1|item)) (2) reduced model: lmerIN(score ~ cotext * DO type + (1|subj) + (1|item))

Appendix 4: An experiment on scope with indefinite DOs

283

Appendix 4: An experiment on scope with indefinite DOs (Datenset: input-scope; scope.out) Statistical analyses were conducted in R version 1.0.136 using the lme4 package (Bates et al., 2015) to perform linear mixed-effect models (LMEM) with the score as outcome variable. As fixed effects we entered context and DO type into the model As random effects, we had intercepts for subjects and items. The context Universal QP and DO type unmarked DO were mapped onto the intercept. To identify the best model fit we performed likelihood ratio tests. This revealed that the full model with a two-way interaction affected the acceptance rate ( χ2 (6) = 18.21, p < .01).97 Estimate

Std. Error t

Intercept

.

.

contextConditional

.

.

.

contextModal

.

.

.

contextNegation

−.

.

.

DOtypeCD

−.

.

−.***

DOtypeDOM

−.

.

−.***

ContextConditional:DOtypeCD

−.

.

−.

ContextModal:DOtypeCD

−.

.

−.

.

.

ContextConditionalDOtypeDOM

−.

.

−.

ContextModalDOtypeDOM

−.

.

−.

.

.

.

ContextNegation:DOtypeCD

ContextNegationDOtypeDOM

*t-values greater than 2 and less than −2 are considered significant.

97 Formulars for model comparison: (1) full model: lmerHE3(score ~ context + DO type + (1|subj) + (1|item)) (2) reduced model: lmerIN3(score ~ cotext * DO type + (1|subj) + (1|item))

t value .***

.**

284

Appendices

The context Conditional and DO type CDed+DOMed DO were mapped onto the intercept. To identify the best model fit we performed likelihood ratio tests. This revealed that the full model with a two-way interaction affected the acceptance rate.98

Estimate

Std. Error t

.

.

.***

−.

.

−.

contextNegation

.

.

.

contextUniversal

−.

.

−.**

DOtypeDO

−.

.

−.

DOtypeUN

.

.

.***

ContextModal:DOtypeDO

.

.

.

ContextNegation:DOtypeDO

−.

.

−.

ContextUniversal:DOtypeDO

.

.

.

ContextModalDOtypeUN

−.

.

−.

ContextNegationDOtypeUN

−.

.

−.***

ContextUniversalDOtypeUN

−.

.

−.

Intercept contextModal

*t-values greater than 2 and less than −2 are considered significant.

98 Formulars for model comparison: (1) full model: lmerHE3(score ~ context + DO type + (1|subj) + (1|item)) (2)

reduced model: lmerIN3(score ~ cotext * DO type + (1|subj) + (1|item))

t value

Bibliography Abney, Steven Paul. 1987. The English noun phrase in its sentential aspect. Cambridge, MA:MIT dissertation Agouraki, Georgia. 1994. Spec-Head Licensing: The Cases of Foci, Clitic Constructions and Polarity Items. A Study of Modern Greek. London: University College London Dissertation. Aissen, Judith. 2003. Differential Object Marking: Iconicity vs. Economy. Natural Language and Linguistic Theory 21(3): 435–483. Anagnostopoulou, Elena. 1994. Clitic Dependencies in Modern Greek. Salzburg: Salzburg University dissertation. Anagnostopolou Elena. 2003. The syntax of ditransitives. Evidence from Clitics. Berlin/ NewYork: Mouton de Gruyter. Anagnostopolou, Elena 2005. Strong and Weak Person Restrictions: A feature checking analysis. In Lorie Heggie and Francisco Ordónez (eds.), Clitics and Affix Combinations. 199–235. Amsterdam/Philadelphia: John Benjamins. Anagnostopoulou, Elena. 2006. Clitic Doubling. In Martin Everaert and Henk van Riemsdijk (eds.), The Blackwell companion to syntax, 519–581. Oxford: Blackwell. Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds.). 1997. Materials on Left Dislocation. Amsterdam: Benjamins. Aoun, Joseph and Audrey Yen-Hui Li. 1989. Scope and constituency. Linguistic Inquiry 20(2): 141–172. Ariel, Mira. 1988. Referring and accessibility. Journal of Linguistics 24 (1): 65–87. Ariel, Mira. 1991. The function of accessibility in a theory of grammar. Journal of Pragmatics 16 (5): 443–464. Avram, Larisa. 2006. English Syntax: The Syntax of Root Clauses. Bucharest: Oscar Print. Avram, Larisa. 2014. Differential object marking in Romanian: the view from acquisition.Paper presented at the Annual Conference of the English Department 16, University of Bucharest, 5–7 June. Avram, Larisa. & Martine. Coene. 2009. Null objects and accusative clitics in Romanian. Bucharest Working Papers in Linguistics 11 (1): 233–252. Baker, Mark. 1988. Incorporation: A Theory of Grammatical Function Changing. Chicago: University of Chicago Press. Baker, Mark. 1996. On the structural position of Themes and Goals. In Johan Rooryck and Laurie Zaring, (eds.), Phrase Structure and the Lexicon, 7–34. Dordrecht: Kluwer. Baker, Mark, and Chris Collins. 2006. Linkers and the internal structure of vP. Natural Language and Linguistic Theory 24 (2), 307–354. Barss, Andrew and Howard Lasnik 1986. A note on anaphora and double objects. Linguistic Inquiry 17 (2): 347–35. Bates, D., Machler, M., Bolker, B. and Walker, S. 2015. Fitting Linear Mixed- Effects Models Using lme4. Journal of Statistical Software, 67, 1–48. Béjar, Suzana & Rezac, Milan. 2003. Person Licensing and the Derivation of PCC effects. In Ana Teresa Pérez-Leroux and Yves Roberge (eds.), Romance Linguistics. Theory and Acquisition: Selected Papers from the 32nd Symposium on Romance Languages (LSRL), Toronto, Arpil 2002. Amsterdam: John Benjamis, 49–62. Belletti, Adriana. 1988. The case of unaccusatives. Linguistic Inquiry 19 (1), 1–34.

https://doi.org/10.1515/9781501513657-009

286

Bibliography

Bittner, Maria and Ken Hale. 1996. The structural determination of Case and agreement. Linguistic Inquiry 27 (1), 1–68. Bleam, Tonia. 1999. Leísta Spanish and the Syntax of Clitic Doubling. Newwark, Delaware: University of Delaware dissertation Bleam, Tonia. 2005. The role of semantic type in differential object marking. In Vogeleer, S. (ed.), Bare Plurals, Indefinites, and Weak-Strong Distinction. Belgian Journal of Linguistics. 19: 3–27. Amsterdam: John Benjamins. Bošković, Željko and Nunes, Jairo. 2007. The copy theory of movement: a view from PF. In Norbert Cover and Jairo Nunes (eds.), The Copy Theory of Movement, 13–74. Amsterdam/ Philadelphia: John Benjamins. Bresnan, Joan and Nikitina, Tatiana. 2007. The Gradience of the Dative Alternation. In Linda Uyechi & Lian Hee Wee (eds.), Reality Exploration and Discovery: Pattern interaction in Language and Life, 1–23. Stanford: Center for the Study of Language and Information Publications. Bruening, Benjamin. 2001. QR obeys Superiority: Frozen Scope and ACD. Linguistic Inquiry 32 (2): 233–273. Bruening, Benjamin. 2010. Double object constructions disguised as prepositional datives. Linguistic Inquiry 41 (2): 287–305. Brugè, Laura. 2000. Categorie funzionale del nome nelle lingue romanze. Milan: Cisalpino. Brugè, Laura, and Gerhard Brugger. 1996. On the accusative a in Spanish. Probus 8 (1), 1–51. Cardinaletti, Anna, and Michal Starke. 1999. The typology of structural deficiency: A case study of the three classes of pronouns. In Henk van Riemsdijk (ed.), Clitics in the languages of Europe, 145–235. Berlin: Mouton de Gruyter. Carlson, Gregory Norman. 1977. Reference to kinds in English. Massachusetts, Amherst: University of Massachusetts dissertation. Published 1980 New York: Garland. Carlson, Gregory Norman. 2003. Weak indefinites. In Martine Coene and Yves D’hulst (eds.), From NP to DP. Vol. 1, The syntax and semantics of noun phrases, 195–210. Amsterdam: John Benjamins. Cechetto, Carlo. 2000. Doubling Structures and Reconstruction. Probus 12 (1): 93–126. Chiriacescu, Sofiana and von Heusinger, Klaus. 2011a. Definite NPs and DOM-marking in Romanian ms. University of Stuttgart Chiriacescu, Sofiana and von Heusinger, Klaus. 2011b. The Discourse Structuring Potential of Definite Noun Phrases in Natural Discourse. In António Branco, Iris Hendrickx,Ruslan Mitkov, Sobha, Lalitha (eds.), Proceedings of the 8th Discourse Anaphora and Anaphor Resolution Colloquium, 151–162. Lisbon: Edições Colibri. Chomsky, Noam. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Linguistic Inquiry Monograph 6. Cambridge, MA: MIT Press. Chomsky, Noam. 1986. Barriers, Linguistic Inquiry Monograph 13. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–156. Cambridge, MA: MIT Press. Chung, Sandra, and Ladusaw, William. 2004. Restriction and saturation. Cambridge, MA: MIT Press. Chung, Sandra, and Ladusaw, William. 2006. Chamorro evidence for compositional asymmetry. Natural Language Semantics 14 (4), 325–357. Cinque, Guglielmo. 1977. The Movement Nature of Left Dislocation. Linguistic Inquiry 8 (2): 397–412.

Bibliography

287

Cinque, Guglielmo.1990. Types of A’-Dependencies. Cambridge: MIT Press. Cornilescu, Alexandra. 2000. Notes on the Intepretation of the Prepositional Accusative in Romanian. Bucharest Working Papers in Linguistics, 2 (1): 91–106. Cornilescu, Alexandra. 2002a. Direct object at the left periphery. Bucharest Working Papers in Linguistics 4 (1): 1–15. Cornilescu, Alexandra. 2002b. Clitic doubling and parasitic gaps in Romanian. Paper presented at Going Romance Groningen, November 2002. Cornilescu, Alexandra. 2006. On Clitic Doubling and Parasitic Gaps in Romanian. Revue Roumaine de Linguistique, LI (1), 23–42. Cornilescu, Alexandra. 2017. Landscaping Romanian Datives, ms. Univ. of Bucharest Cornilescu, Alexandra and Dobrovie-Sorin, Carmen. 2008. Clitic doubling, complex heads, and interarboreal operations. In Kallulli, Dalina and Liliane Tasmowski (eds.). Clitic Doubling in the Balkan Languages. Vol. 30 Linguistik aktuell, 289–319. Amsterdam: John Benjamins. Cornilescu, Alexandra and Cosma Ruxandra. 2014. On the Functional Structure of the Romanian de supine. In Ruxandra Cosma, Stefan Engelberg, Susan Schlotthauer, Speranța Stănescu and Gisela Zifonun (eds) Komplexe Argumentstrukturen. Kontrastive Untersuchungen zum Deutschen, Rumänischen und Englischen. (Konvergenz und Divergenz Series), 283–336. Berlin/Boston: de Gruyter. Cornilescu, Alexandra, Dinu Anca and Tigău, Alina. 2017a. Experimental Data on Romanian Double Object Constructions, Revue Roumaine de Linguistique LXII (2). 157–177. Cornilescu, Alexandra, Dinu Anca and Tigău, Alina. 2017b. Romanian Dative Configurations: DitransitiveVerbs, A Tentative Analysis, Revue Roumaine de Linguistique, special volume Investigating Romanian Datives / Recherches autour du datif roumain, LXII(2). 179–206 Cornilescu, Alexandra, Nicolae, Alexandru. 2012. Nominal ellipsis as definiteness and anaphoricity: the case of Romanian. Lingua 122 (10): 1070–1111. Cornilescu, Alexandra and Tigău, Alina. 2018a. Towards a drivational account of Romanian binding ditransitive constructions. In Teresa Parodi (ed.), Proceedings of the VIII Nereus International Workshop “Referential Properties of the Romance DP in the Context of Multilingualism”. Arbeitspapier 129. Fachbereich Sprachwissenschaft, 121–140. Universität Konstanz. Cornilescu, Alexandra and Tigău, Alina. 2018b. On the syntax of Romanian DOM. ms. University of Bucharest. Cuervo, Maria Cristina. 2003a. Datives at large. Cambridge, Mass: MIT dissertation. Cuervo, Maria Cristina. 2003b. Structural asymmetries but same word order: the dative alternation in Spanish. In Anna Maria Di Sciullo (ed.), Asymmetry in Grammar. Volume 1: Syntax and semantics, 117–144. Amsterdam/Philadelphia, John Benjamins. David, Oana. 2015. Clitic doubling and differential object marking: A study in diachronic construction grammar. Constructions and Frames 7(1): 103–135. David, Oana. 2014. Subjectification in the Development of Clitic Doubling: A Diachronic Study of Romanian and Spanish. Proceedings of the Annual Meeting of the Berkeley Linguistics Society, 40: 42–62. de Hoop, Helen. 1996. Case configuration and noun phrase interpretation. New York: Garland. Déchaine, Rose-Marie, and Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33 (3), 409–442. Demonte, Violeta. 1995. Dative alternation in Spanish. Probus 7(1),5–30. http://dx.doi.org/ 10.1515/prbs.1995.7.1.5

288

Bibliography

Dikken, Marcel den. 1995. Particles: On the syntax of verb-particle, triadic, and causative constructions. Oxford: Oxford University Press. Diaconescu, Constanţa Rodica and María Luisa Rivero. 2007. An applicative analysis of double object constructions in Romanian, Probus, 19(2),209–233. http://aix1.uottawa.ca/~rom lab/pubs/RiveroDiaconescu.2006.pdf Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Dobrovie-Sorin, Carmen. 1990. Clitic Doubling, Wh-Movement, and Quantification in Romanian. Linguistic Inquiry 21 (3): 351–397. Dobrovie-Sorin, Carmen. 1994. The Syntax of Romanian. Studies in Generative Grammar, vol 40, Berlin: Mouton de Gruyter. Dryer, Matthew. 1987. On primary objects, secondary objects, and antidative, Language 62 (4): 808–845. Edmonds, Joseph. 1985. A unified theory of syntactic categories. Dordrecht: Foris. Fábregas, Antonio. 2013. Differential Object Marking in Spanish: State of the Art. Borealis II (2): 1–80. Farkas, Donka. 1987. Direct and indirect object reduplication in Romanian. Chicago Linguistics Society 14: 88–97. Farkas, Donka. 1982. Intensionality and Romance Subjunctive Relatives. Pennsylvania: Pennsylvania State University dissertation. Farkas, Donka. 1994. Specificity and scope. In Léa Nash and Georges Tsoulas (eds.) Actes du premier Colloque Langues et grammaires: 119–137. Farkas Donka. 1995. Specificity and Scope. ms. University of Santa Cruz. Farkas, Donka. 2001. Dependent indefinites and direct scope. In Condoravdi, Cleo and Renardel, Gerard (eds.), Logical Perspectives on Language and Information, 41–72. Stanford, CA: CSLI. Farkas, Donka, and Henriëtte de Swart. 2003. The semantics of incorporation. Stanford, CA: CSLI Publications. Farkas, Donka and Klaus von Heusinger. 2003. Stability of reference and object marking in Romanian. Ms., Universität Stuttgart, (Paper given at the XV European summer school in Logic, Language and Information (ESSLLI) in Vienna). GALR (2005) = Gramatica limbii române. Academia Română, Institutul de Lingvistică “Iorgu Iordan – Al. Rosetti”. Coord.: Valeria Guţu Romalo. Bucharest: Editura Academiei Române. Georgala, Effi. 2010. Why German is not an exception to the universal base order of double object constructions. Handout for 28th West Coast Conference on Formal Linguistics, University of Southern California, 19–21 February. Georgala, Effi. 2011. Applicatives in their Structural and Thematic Function: A Minimalist Account of Multitransitivity. Ithaca, New Yok: Cornell University dissertation. Georgala, Effi. 2012. Short object shift and ditransitive structure in Greek. In Sumru Özsoy, and Ayșe Gürel (Eds.), Current Issues in Mediterranean Syntax. Leiden: Brill. http://conf. ling.cornell.edu/georgala/MSM2_georgala.pdf. Georgala, Effi, Waltraud Paul and Whitman, John. 2008. Expletive and Thematic Applicatives. In Charles B. Chang & Hannah .J. Haynie (eds.). Proceedings of the 26th West Coast Conference on Formal Linguistics (WCCFL 26), 181–189. Somerville, Mass.: Cascadilla Proceedings Project. Ghomeshi, Jila. 2008. Markedness and bare nouns in Persian. In Simin Karimi, Vida Samiian, and Donald Stilo (eds.), Aspects of Iranian linguistics, 85–111. Newcastle upon Tyne, UK: Cambridge Scholars Press.

Bibliography

289

Gierling, Diana. 1997. Clitic Doubling, Specificity and focus in Romanian. In Black, James. & Motapanyane, Virginia (eds.) Clitic, pronouns and movement, 63–85. Amsterdam: John Benjamins. Giorgi, Alessandra and Longobardi, Giuseppe. 1991. The syntax of noun phrases: configuration, parameters and empty categories. Cambridge: Cambridge University Press. Giurgea, Ion. 2013. Originea articolului posesiv-genitival al şi evoluţia sistemului demonstrativelor în română. [The origin of the possessive-genitival article al and the evolution of the system of demostratives in Romanian] Bucureşti: Editura Muzeului Naţional al Literaturii Române. Gramatica limbii române [The Grammar of the Romanian Language]. 2005. Bucureşti: Editura Academiei Române. Grohmann, Kleanthes. 2003. Prolific Domains: On the Anti-Locality of Movement Dependencies. Amsterdam/Philadelphia: John Benjamins. Harada, Naomi and Richard K. Larson. 2009. Datives in Japanese. In Ryosuke Shibagaki and Reiko Vermeulen (eds.) Proceedings of the 5th Workshop on Altaic Formal Linguistics (WAFL5). MIT Working Papers in Linguistics Vol. 58, 3–17. Cambridge, MA: MIT Working Papers in Linguistics. Harley, Heidi. 2002. Possession and the DOC. Linguistic Variation Yearbook 2 (1). 31–70. http://dx.doi.org/10.1075/livy.2.04har Harley, Heidi and Hyun Kyoung Jung. 2015. In support of the PHAVE analysis of the double object construction. Linguistic Inquiry 46 (4): 703–730. Heim, Irene and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell. von Heusinger, Klaus. 2002. Specificity and definiteness in sentence and discourse structure. Journal of Semantics 19 (3), 245–275. von Heusinger, Klaus and Onea, Edgar. 2008. Triggering and Blocking Effects in the Diachronic Development of DOM in Romanian. Probus 20 (1): 67–110. Hoji, Hajime. 1085. Logical Form constraints and configurational structures in Japanese. PhD dissertation University of Washington, Seattle. Hornstein, Norbert, Sara Rosen and Uriagereka, Juan. 1995. Integral Predication. In Camacho, José, Choueiri, Lina and Watanabe, Maki (eds.), Proceedings of WCCFL 14, 169–184. Chicago: University of Chicago Press. (Reprinted in J. Uriagereka, 2002) Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Kayne, Richard. 1989. Facets of Romance Past Participle Agreement. In Paola Benincà (ed.), Dialect Variation and the Theory of Grammar, 85–104. Dordrecht: Foris. Kayne, Richard. 1975. French syntax. Boston MA: MIT Press. Koizumi, Masatoshi. 1995. Phrase structure in the Minimalist Program. Cambridge, Mass: MIT dissertation. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Johan Rooryck and Laurie Zaring (eds.) Phrase structure and the lexicon, 109–137. Dordrecht: Kluwer. Krifka, Manfred. 2004. Semantic and pragmatic conditions for the dative alternation. Korean Journal of English Language and Linguistics 4: 1–32. Krivochen, Gabriel Diego. 2014. Clitics, Procedural Elements and Spanish Syntax. ReVEL, 12 (22): 43–79 [www.revel.inf.br/eng]. Iatridou, Sabine. 1991. Clitics and island effects. ms. MIT, Cambridge Iatridou, Sabine. 1994. Clitics and Island Effects. In Roumyana Izvorski and Victoria Tredinnick (eds.), University of Pennsylvania Working Papers in Linguistics 2 (1). 11–30. Philadelphia: University of Pennsylvania Linguistics Club.

290

Bibliography

Iorga, Ana. 2013. O Tipologie a dativului românesc [A typology of the Romanian Dative]. Bucharest: University of Bucharest dissertation. Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19 (3), 335–391 . Larson, Richard K. 1990. Double objects revisited: Reply to Jackendoff. Linguistic Inquiry 21 () 4: 589–632. Larson, Richard K. 2010. On Pylkkänen’s semantics for low applicatives. Linguistic Inquiry 41 (4):701–704. Larson, Richard. 2014. On shell structure. Routledge: London. Leonetti, Manuel. 2004. Specificity and differential object marking in Spanish. Catalan Journal of Linguistics 3: 75–114. Lidz, Jeffrey. 2006. The grammar of accusative case in Kannada. Language 82 (1), 1–23. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and Logical Form. Linguistic Inquiry 25 (4): 609–665. López, Luis. 2009. A derivational syntax for Information Structure. Oxford Studies in Theoretical Linguistics 23, Oxford: Oxford University Press. López, Luis. 2012. Indefinite Objects. Scrambling, Choice Functions, and Differential Marking. Cambridge, Massachusetts: MIT MacDonald, J., 2015, “A Movement Analysis of Some Double Object Constructions”, in: U. Steindl, et al. (eds), Proceedings of the 32nd West Coast Conference on Formal Linguistics, Somerville, MA, Cascadilla Proceedings Project, 276–285. Marantz, Alec.1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT Press (= Linguis- tic Inquiry Monograph 10). Marantz, Alec. 1993. Implications of asymmetries in double object constructions. In Mchombo, Sam A. (ed.). Theoretical Aspects of Bantu Grammar. Stanford, CA: CSLI Publications (= CSLI lecture notes 38). 113–150. Mardale, A. 2008. Microvariation within Differential Object Marking: Data from Romance. Revue roumaine de linguistique, LIII (4): 448–467. Martí, Luisa. 2008. The semantics of plural indefinite noun phrases in Spanish and Portuguese. Natural Language Semantics 16 (1). 1–37. Michelioudakis, Dimitris. 2012. Dative arguments and abstract case in Greek. Cambridge, UK: University of Cambridge disseration. Miyagawa, Shigeru and Takae Tsujioka. 2004. Argument structure and ditransitive verbs in Japanese. Journal of East Asian Linguistics 13: 1–38. Mondoñedo, Miguel Rodriguez. 2007. The syntax of objects: Agree and differential object marking. Storrs, Connecticut: University of Connecticut dissertation. Munteanu, Oana. 2016. Te urăsc – nu mă părăsi [I hate you don’t leave me]. București: Editura Trei (Translation from English: Jerold J. Kreisbahn and Tal Straus. 1991. I hate you don’t leave me. New York: Avon Books) Nissenbaum Jonathan. 2000. Explorations in covert phrase movement. Cambridge, MA: MIT dissertation. Nunes, Jairo. 1995. The copy theory of movement and linearization of chains in the Minimalist Program. Maryland: University of Maryland, College Park dissertation. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement. Cambridge, MA: MIT Press. Oehrle, Richard Thomas. 1976. The grammatical status of the English dative alternation. Cambridge, MA: MIT dissertation. Ormazabal, Javier and Romero, Juan. 2002. Agreement Restrictions. Manuscript, University of the Basque Country and University of Alcalá de Henares

Bibliography

291

Ormazabal, Javier and Romero, Juan. 2007. The Object Agreement Constraint, Natural Language and Linguistic Theory 25 (2): 315–347. Ormazabal, Javier and Romero, Juan. 2010. The derivation of dative alternations. In Maia Duguine, Susana Huidobro, and Nerea Madariaga (eds.), Argument structure and syntactic relations: A cross-linguistic perspective, 203–232. Amsterdam: John Benjamins. Ormazabal, Javier and Romero, Juan. 2012. PPs without Disguises: Reply to Bruening. Linguistic Inquiry 43 (2). 455–474. Pană Dindelegan, G. 1974. Sintaxa limbii române. Bucureşti: T.U.B. Pesetsky, David. 1995. Zero Syntax: Experiencers and Cascades. Cambridge, MA: The MIT Press. Pesetsky, David and Torrego, Esther. 2007. The Syntax of Valuation and the Interpretability of Features. In Simin Karimi, Vida Samiian, Wendy K. Wilkins (eds), Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation, 262–194. Amsterdam: John Benjamins. Pineda, Anna. 2012. Transitividad y afectación en el entornolingüístico romance y eusquérico. X. Viejo Fernández (ed.), Estudios sobre variación sintáctica peninsular, 31–73. Oviedo: Seminariu de Filoloxía Asturiana de la Universidad de Oviedo. Pineda, Anna. 2013. Double object constructions in Spanish (and Catalan) revisited. In Baauw, Sergio, Drijkoningen, Frank, Meroni Luisa, Pinto, Manuela (ed.), Romance Languages and Linguistic Theory 2011, 193–216. Amsterdam: John Benjamins. Postal, Paul. 1966. On so-called “pronouns” in English. In David Reibel and Sanford Schane (eds.), Modern studies in English 201–224. Englewood Cliffs, N.J.: Prentice-Hal. Preminger, Omer. 2017. What the PCC tells us about “abstract” agreement, head movement and locality. https://ling.auf.net/lingbuzz/003221. Pylkkänen, Liina. 2002. Introducing Arguments. Cambridge, MA: MIT dissertation. Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, MA: MIT Press. Ramchand, Gillian. 2008. Verb Meaning and the Lexicon: A First Phase Syntax. Cambridge: Cambridge University Press. Rappaport Hovav, Malka and Beth Levin. 2008. The English dative alternation: The case for verb sensitivity. Journal of Linguistics 44 (1): 129–167. Reinhart, T. 2002. The Theta System: an Overview. Theoretical Linguistics 28(3): 229–290. Reinhart, Tanya. 1997. Quantifier scope: How labor is divided between QR and choice functions. Linguistics and Philosophy 20 (4), 335–397. Richards, Marc. 2008. Defective Agree, Case Alternations, and the Prominence of Person. In Marc Richards and Andrej Malchukov (eds.), Scales. Linguistische Arbeits berichte vol 86, 137–161. Universität Leipzig. Rivero, Maria Luisa. 1979. Mood and presupposition in Spanish. Foundations of Language 7 (3): 305–36. Rizzi, Luigi. 2001. On the position “Int(errogative)” in the left periphery of the clause. In Cinque, Guglielmo, Salvi, Giampaolo. (Eds.), Current Studies in Italian Syntax, Essays offered to Lorenzo Renzi, 287–296. Amsterdam: Elsevier. Ruse, Andrei. 2014. Zaraza. București: Polirom. Sportiche, Dominique. 1989. Le Mouvement Syntaxique: Constraintes et Parametres. Langages 95, 35–80. Sportiche, Dominique. 1996. Clitic constructions. In Rooryck, Johan and Zaring, Laurie (Eds.), Phrase Structure and the Lexicon, 213–276. Dordrecht: Kluwer. Săvescu, Oana. 2009. A Syntactic Analysis of Pronominal Clitic Clusters in Romance: The View from Romanian. New York: New York University dissertation.

292

Bibliography

Selkirk, E. 1984. Phonology and Syntax: The Relation between Sound and Structure. Cambridge: MIT Press. Selkirk, Elizabeth O. 1995. Sentence prosody: intonation, stress and phrasing. In Goldsmith, John (ed.) The Handbook of Phonological Theory, 550–569. London: Blackwell. Stan, Camelia. 2009. Complementul direct. ms. University of Bucharest. Steriade, Donca. 1980. Clitic doubling in the Romanian wh-constructions and the analysis of topicalization. Proceedings of the Chicago Linguistic Society 16: 282–297. Suñer, Margarita. 1988. The Role of Agreement in Clitic-Doubled Constructions. Natural Language and Linguistic Theory 6: 391–434. Suñer, Margarita. 2006. Left dislocation with or without epithets. Probus 18 (3): 127–158. Tănase-Dogaru, Mihaela. 2009. The category of Number. Bucharest: Editura Universității din București. Tigău, Alina. 2010. Syntax and Interpretation of the Direct Object in Romance and Germanic Languages with an Emphasis on Romanian, German, Dutch and English. Bucharest: University of Bucharest dissertation. Tigău, Alina. 2011. Syntax and Interpretation of the Direct Object in Romance and Germanic Languages with an Emphasis on Romanian, German, Dutch and English. Bucharest: Editura Universităţii Bucureşti. Tigău, Alina 2014. The Two-Object Construction in Romanian and German. In Ruxandra Cosma, Engelberg, Stefan, Schlotthauer, Susan, Stănescu, Speranța and Gisela Zifonun (eds) Komplexe Argumentstrukturen. Kontrastiv Untersuchungen zum Deutschen, Rumänischen und Englischen. (Konvergenz und Divergenz Series), 85–141. Berlin/Boston: de Gruyter. Tigău, Alina. 2015. On the syntax of Romanian Clitic Doubling Constructions. Generative Grammar in Geneva IX: 1–25. Tigău, Alina. 2016. A DRT analysis of Clitic Doubling. Bucharest: Editura Universităţii Bucureşti. Tigău, Alina. 2018. A higher applicative: the case of the Romanian ethical dative. Revue Roumaine de Linguistique. LXIII (4) : 361–378. Tigău, Alina and K., von Heusinger. 2018. Romanian Ditransitives. ms. University of Cologne. Tigău, Alina and K., von Heusinger. 2018. Specificity with DOM and CD. ms. University of Cologne Torrego, Esther. 1998. The dependencies of objects. Cambridge, MA: MIT Press. Travis, Lisa. 1984. Parameters and effects of word order variation. Cambridge, MA: MIT dissertation Travis, Lisa. 1992. Inner aspect and the structure of VP. Cahiers de Linguistique de l’UQAM 1, 130–146. Travis, Lisa. 2010. Inner aspect. Dordrecht: Springer Verlag. Uriagereka, Juan. 1987. Variables in Basque and Governance. Ms University of Connecticut, Storrs Uriagereka, Juan. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26 (1): 79–123. Uriagereka, Juan. 2001. Doubling and Possession. In Gerlach, Birgit and Grijzenhout, Janet (eds.), Clitics in Phonology, Morphology and Syntax, 405–431. Amsterdam: John Benjamins. Uriagereka, Juan. 2002. Derivations, London: Routledge. Uriagereka, Juan. 2005. On the Syntax of Doubling. In Heggie, Lorie and Francisco Ordóñez (eds.) Clitic and Affix Combinations: Theoretical perspectives, 343–374. Amsterdam: John Benjamins. Villalba, Xavier. 2000. The syntax of sentence periphery. Barcelona: Universitat Autònoma de Barcelona dissertation. Wood, Jim and Alec Marantz. 2015. The interpretation of external arguments. In Roberta D’Alessandro, Irene Franco and Ángel Gallego (eds.) The Verbal Domain. Oxford: Oxford University Press. Zagona, Karen. 2002. The Syntax of Spanish. Cambridge: Cambridge University Press.

Index Acceptability scale 54 Alternative Projection Account 166 Anaphors 106 Animacy hierarchy 5 Applicative 102 Asymmetric binding 166 Bare nominals 14 Bare plurals 43 Bare singulars 111 Bi-directional c-command 165 Binding dependencies 10, 88 C-command 91 Case valuation 114 Case-valuation 8 Causative-permissive structure 44 Caused motion 169 Caused possession 169 Choice functions 5, 80 Clause union structures 45 Clitic doubling 4 Clitic left dislocated 66 Clitic resumption 68 Definiteness 9 Derivational account 166, 179 Direct object 13 Distributive câte 66 DOM 4 Doubled Object Construction 165 Epistemic and partitive specificity 15 Event Identification 97 Existential Closure 97 External Argument 87 Feature specification 6 Feature valuation 11 Generalized quantifiers 103

https://doi.org/10.1515/9781501513657-010

Heavy NP Shift 133 High applicatives 230 Incorporation 8, 95 Indefinites 20 Indicative mood 61 Indirect object 13 Individuation 7, 116 Inflectional or prepositional dative case 10 Integrals 58 Intensional operators 55 Internal arguments 3 Intervention effects 6 KP 103 Mapping Hypothesis 149 Marked and unmarked objects 14 Narrow scope 49 Object control configurations 41 Parasitic gaps 133 Partitive reading 61 Permissive constructions 43 Person feature 4 Prepositional Object Construction 165 Property denoting 37 Quantifier-variable reading 105 Raising to object 44 Real argumenthood 7 Reconstruction 179 Referential anchoring 98 Restrict 5, 86 Satisfy 101 Scope 38

294

Index

Scope dependencies 13, 85 Short scrambling 96 Small clause complements 38 Specificity 8 Stability of reference 117

Subjunctive 39 Symmetrical binding 10 Weak indefinites 112 Wide scope 7, 49