Resumptivity in Mandarin Chinese: A Minimalist Account 9783110492385, 9783110487596

The use of resumptive pronouns is quite productive in Mandarin Chinese; however, their distribution has rarely been stud

170 83 2MB

English Pages 323 [324] Year 2016

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Resumptivity in Mandarin Chinese: A Minimalist Account
 9783110492385, 9783110487596

Table of contents :
Contents
List of abbreviations
1 Resumptivity: an overview
2 Syntactic derivation: two types of A'-dependencies
3 Reconstruction and internal structures of resumptive pronouns
4 Specificity effects
5 Concluding remarks
6 Appendix: General distribution of the resumptive pronouns in Mandarin Chinese
References
Subject index

Citation preview

Victor Junnan Pan Resumptivity in Mandarin Chinese

Trends in Linguistics Studies and Monographs

Editor Volker Gast Editorial Board

Walter Bisang Jan Terje Faarlund Hans Henrich Hock Natalia Levshina Heiko Narrog Matthias Schlesewsky Amir Zeldes Niina Ning Zhang Editor Responsible for this volume Volker Gast and Niina Ning Zhang

Volume 298

Victor Junnan Pan

Resumptivity in Mandarin Chinese

A Minimalist Account

978-3-11-048759-6 e-ISBN (PDF) 978-3-11-049238-5 e-ISBN (EPUB) 978-3-11-048975-0 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. 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. © 2016 Walter de Gruyter GmbH, Berlin/Boston Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com



for mom, Lily Pan, who raised me; for granny (1921–2014) who always cherished me like her Little Prince; for dad (1951–1990) who taught me how important it is to become independent.

Acknowledgements This monograph presents the results of one part of my research project “Resumptivity in the Minimalist Program” conducted since 2011. While writing earlier versions of each chapter, I have tremendously benefited from the detailed discussions with many colleagues and friends who helped me to sharpen the main ideas and proposals of this book. Alain Rouveret read every earlier version of my analyses and generously shared his ideas with me; his vision on the general architecture of minimalist machinery had an important influence on me. Waltraud Paul, not only a faithful reader but also a colleague who has challenging points of view, spent hours and hours discussing every single detail and every Chinese example with me during those years. As she often says, the real devil is in the detail. Dylan Wei-Tien Tsai greatly inspired me on interpretations of gaps and bound pronouns. During his stay in Paris in 2015, we had many occasions to talk about linguistics and I learned a lot from him. Haihua Pan and Jianhua Hu gave me many insightful comments on the restrictions of topicalization both in Paris and in Beijing. Paul Law always shared his ideas and concerns with me wherever we met. Our discussion on the bound variable reading of anaphora and Skolemized choice functions in the train from Brest to Lannion during a fieldwork for our project on resumptive pronouns in Breton helped me to clarify certain passages in this book. As the responsible editor of this book, Niina Ning Zhang not only encouraged me during the entire project but also provided me with many good ideas. She gave me a true thorough review going through page-by-page the near-final version of the manuscript, which saved me from unclear passages. I also express my gratitude to the two anonymous reviewers who helped me to make my point precise and accurate. Chapters 2, 3 and 4 have been presented in different conferences. The present version of analyses includes comments from the audience: The 6th International Spring Forum organized by The English Linguistic Society of Japan (ELSJ) (Tokyo University, Japan, 2013), The 21st annual conference of the International Association of Chinese Linguistics (IACL) (National Taiwan Normal University, Taiwan, 2013) and The 22nd annual conference of International Association of Chinese Linguistics (IACL) (University of Maryland, 2014). Special thanks go to Yen-hui Audrey Li, Shu-Ing Shyu, Hajime Ono and Koji Fujita for their questions, comments and suggestions during the question sessions and coffee breaks. Mandarin data presented in this monograph have been double-checked carefully with my native colleagues and students to whom I express my sincere gratitude: Hongyuan Sun, Yiqin Qiu, Xiaoshi Hu, Xiaoliang Huang, Yan Li, Xiaoqian Zhang, Chen Zhao and Xinyue Yu. I also thank the students who at-

VIII  Acknowledgements tended the weekly Reading Group of Formal Linguistics of Chinese that I have been organizing since 2011 for sharing their intuition and judgments on the sentences presented in this book. I used the draft of this book as teaching material for my Master’s and Doctoral seminars on formal syntax and semantics and on Chinese linguistics at the University of Paris Diderot-Paris 7 in fall semesters since 2013 and I am happy to express my gratitude to the students who attended to these courses for their helpful questions and feedback. I also thank Redouane Djamouri and Nicolas Guilliot for discussing the data from both Arabic dialects and French. Some important thoughts can be traced back to the research seminars of Hamida Demirdache that I attended during my PhD years at the University of Nantes. Earlier versions of different chapters were reread carefully by my dear colleagues and friends, especially Paul Boucher, Daniel Kwang Guan Chan and Aaron J. Danner. I highly appreciate their comments and questions. I always feel extremely useful to get feedback from readers of different domains, such as generative syntacticians, linguists of Chinese languages, mathematicians, computer scientists and so forth, since their comments made me clarify every single sentence I put in earlier drafts. The years during which I conducted this research project did not go by without obstacles; fortunately, my friends and colleagues were always there to support and encourage me unfailingly. I thank Blanche Chia-Ping Chiu, Chantal Seguy, Frédérique Gilbank, Damien Fleury, M.M. Jocelyne Fernandez-Vest, Hans Obenauer, Paul Boucher and Anne Boucher for their calls, afternoon walks, coffee and tea times, which gave me a lot of courage and motivated me to keep going forward in my career and life. In particular, Audrey Li and Dylan Tsai encouraged me and supported me morally during those years. Colleagues from my unit, Carla Soares-Jesel and Lucia Tovena, encouraged me during the most difficult time. Thanks to these great friends, I was never defeated by difficulties. I also thank my research institute LLF-UMR 7110 CNRS and the linguistic department of the University of Paris 7, especially, Anne Abeillé, Olivier Bonami, Caterina Donati and Claire Saillard for their support in this project. During the publication process, the editor of TiLSM, Volker Gast, was extremely encouraging and the De Gruyter editors, Julie Miesse and Birgit Sievert were very helpful. I also thank Nancy Christ for her technical supports. My dearest mum is always there to support me with her unfailing love. Victor Junnan Pan Paris Spring, 2016

Contents List of abbreviations

XV

Resumptivity: an overview 1 Introduction 1 Resumptivity 2 Resumptive pronouns 2 Two types of resumptive pronouns 3 Two different points of view 4 Syntactic level: to move, or not to move, that is the question! Period of GB 11 Period of MP 15 Semantic level: to move, or not to move, that is still the question! Main proposals in this study 19 Against the macro -variation hypothesis 19 Against the “spelled out traces” analysis 20 Main proposals 20 Organization of the argumentation 24

10

18

Syntactic derivation: two types of A'-dependencies 27 Introduction 27 Distribution of resumptive pronouns in relatives and in dislocation 30 structures Islands 30 Relatives 33 Dislocation structures 43 Summary 49 Crossover effects 51 Weak crossover effects 56 Relatives 56 Dislocation structures 61 Strong crossover effect 65 Relatives 65 Dislocation structures 67 Summary 69 Summary 71 Analyses 72 In the framework of Government and Binding theory 73

X  Contents .. ... ... .... .... ... .... .... .... .... .... .. .

In the Minimalist Program  76 Match, Agree and Move  76 Deriving resumptive constructions by Agree  81 Adger & Ramchand (2001, 2005)  82 Rouveret (2002, 2008, to appear)  87 Resumption in Chinese  92 Island effects  93 Crossover effects  96 Resumptive dislocationstructures  100 Differences between relatives and dislocation structures  105 Locality  112 Summary  122 Conclusion  124

 Reconstruction and internal structures of resumptive pronouns  129 . Introduction  129 . Reconstruction effects and movement  132 .. Two types of reconstruction effects  133 ... Reconstruction of anaphoric binding  133 ... Reconstruction of the scope of a quantifier  133 .. Reconstruction effects and movement  134 .. Two types of resumptivities  136 ... Two generalizations of Aoun et al. (2001)  136 ... Against the generalizations of Aoun et al. (2001)  140 .... Against the first generalization of Aoun et al.: absence of island  140 .... Against the second generalization of Aoun et al.: presence of islands  142 ... Conclusion  147 . Reconstruction effects and internal structures of resumptive pronouns  148 .. Two types of resumptive pronouns: strong pronouns/epithets and weak pronouns  149 .. Two types of copies: indefinite and definite (Guilliot 2006)  152 .. Two internal structures of resumptive pronouns: Rouveret (1994, 2002, 2008, to appear)  157 .. Choice of internal structure of a resumptive pronoun  162 .. Summary  163 . Reconstruction effects in Chinese  165 .. Reconstruction and quantifier scope  166

Contents  XI

... ... ... .. ... ... ... .. ... ... ... .. ... ... ... .. . .. .. .. ... .... .... .... .... .... ... ... .. .

Relatives  167 Dislocation structures  170 Summary  173 Possibility to take a quantificational antecedent  174 Relatives  174 Dislocation structures  175 Summary  177 Reconstruction of anaphoric binding  178 Relatives  178 Dislocation structures  181 Summary  182 Condition C effects under reconstruction  184 Relatives  184 Dislocation structures  186 Summary  189 Summary  190 Analysis  192 Correlation between types of syntactic constructions and forms of resumptives  192 Interaction of different components of grammar  198 Argument: two types of pronouns in Chinese  200 Distribution of the pronoun qi  201 Reconstruction of a quantifier scope  201 Possibility to take a quantificational antecedent  203 Anaphoric binding reconstruction  204 Condition C effects under reconstruction  206 Summary  209 Differences between ta and qi  209 Conclusion  211 A mysterious problem is solved!  211 Conclusion  213

 Specificity effects  217 . Introduction  217 . Derive individual reading and distributive reading  219 .. Two types of specificities  219 ... Non-specific reading (de dicto) vs. specific reading (de re)  219 ... Individual reading vs. pair-list reading  220 .. Resumptivity and semantic readings  221 ... Sharvit (1999)  221

XII  Contents ... ... ... . .. ... ... .. ... ... .... .... .. . .

Guilliot (2006, 2011)  223 Malkawi (2009): competition of the readings  224 Rouveret (2002, 2008, to appear)  227 General distribution of different readings in Chinese  230 Individual readings: non-specific (de dicto) vs. specific (de re)  230 Relatives  230 Dislocation structures  232 Distributive readings: natural function vs. pair-list  232 Relatives  232 Dislocation structures  235 Resumptive pronouns  235 Intrusive pronouns  236 Summary  238 Analysis  239 Conclusion  242

 Concluding remarks  245 . Syntactic level  245 . Semantic level  247 . Architecture of resumptivity  252 .. Interaction of different factors involved in the derivation of an A'dependency  252 .. Two predictions  253 . What Chinese tells us about resumptivity  255 

Appendix: General distribution of the resumptive pronouns in Mandarin Chinese  257 . Introduction  257 . Personal pronouns in Chinese  257 . Syntactic constructions involving A'-dependencies in Chinese  258 .. Wh-dependency  259 .. Relativization  260 .. Topicalization  262 .. Ex-situ cleft-focus structures  263 .. Wh-ex-situ constructions  275 .. ATB-constructions  280 .. Summary  284

Contents  XIII

. .. ..

General distribution of the resumptive pronouns in Chinese  285 Cases in which the presence of the resumptive pronouns is obligatory  285 Cases in which the presence of the resumptive pronouns is optional or prohibited  292

References  297 Subject index  303

List of abbreviations    ACC Agr C Cl Clv Exp F Fut Gen M Neg O Passive Perf Pl Prog Prt Rel SFP Sg Top

First person Second person Third person Accusative case Agreement marker Complementizer Classifier Verb classifier Experience aspect marker Feminine Future tense Genitive case Masculine Negative marker Object, inanimate Passive marker Perfective aspect Plural Progressive aspect Particle Relative complementizer Sentence-final particle Singular Topic marker

 Resumptivity: an overview . Introduction Resumptivity is one of the problems that have long history in the framework of generative grammar (since Ross 1967), and the reason why this phenomenon attracts so much attention from linguists is partially because it is not only a pure syntactic problem. Existing work explores this problem both in syntax and in semantics, which opens the possibility to treat resumptivity at the interface between syntax and semantics. Historically, Doron (1982) is one of the pioneering works developing a full semantic account on resumptive dependencies and on resumptive pronouns. Syntactically, the derivation of a resumptive chain is the center of the debate. In addition, the dual status of the mysterious resumptive pronouns makes this problem even more interesting. Resumptive pronouns display many similar properties of wh-traces; however, at the same time, they are different from wh-traces in that they are morphologically full pronouns. It is difficult to determine whether exploring the nature of resumptive pronouns is a syntactic problem or a semantic problem or both. Needless to say, resumptive pronouns have clear morphological forms which distinguish themselves from empty categories, such as wh-traces or gaps in certain types of A'-dependencies. On the other hand, resumptive pronouns occupy uniquely the variable position in an A'-chain, which suggests that semantically they are comparable with variables. Even if resumptivity is a rather general and common linguistic phenomenon observed in many languages cross-linguistically, it has been revealed that different languages do not always behave alike concerning the use of the resumptive pronouns, which of course makes a complicated problem even more complicated. In this study, I will explore the problems closely linked to resumptivity based on data from Mandarin Chinese. At the end of the monograph, I will not only present a clear detailed picture of the distribution of the resumptive pronouns but also present an interaction of several different factors involved in the derivation of resumption in an even more general sense. The research purpose in this study is not only about Chinese, but also an attempt to see what Chinese tells us about resumptivity in a broad sense and to what extend certain factors can be generalized cross-linguistically. I will concentrate on discussion of the following essential questions: (i)

What are possible syntactic mechanisms that can derive resumptive dependencies?

  Resumptivity: an overview (ii)

What is the nature of resumptive pronouns?

(iii)

Are resumptive pronouns merely phonologically spelled out traces?

(iv)

When different types of personal pronouns are used as resumptive pronouns, do they function in exactly the same way and give rise to exactly the same syntactic as well as semantic effects?

In this very first chapter, I will start by giving a brief introduction of the general research problems concerning resumptivity and the most important solutions provided in the previous research. I highly recommend the historical review of the problem in Rouveret (2011:1–62) in which the origin of the problem, the different solutions given during different periods of the development of the generative grammar, the advantages as well as the disadvantages of each solution and the remaining open questions have been clearly addressed in great detail. I will of course not repeat every single aspect of resumptivity in this chapter; however, I will single out and highlight several important points concerning resumptivity in Section 2. Again, Section 2 is greatly inspired by Rouveret (2011). In Section 3, I will present the results of my study and the most important generalizations based on the results of the relevant tests in this study as well as the answers that I tentatively suggest to the essential questions listed above.

. Resumptivity .. Resumptive pronouns Resumptive pronouns located on the variable site in an A'-chain function as variables bound by an operator in an A'-position. They are called resumptive pronouns because in most cases the elements that play the role of resumptives are pronominal ones. As will be detailed in Chapters 2, 3 and 4, certain languges, such as Arabic dialects, do use epithets resumptively and these epithets behave in a similar fashion as “strong pronouns” or “doubled pronouns” in these languages. Resumptive pronouns can only exist in A'-dependencies, for instance, in relative clauses, left-dislocation structures, cleft-focus structures, wh-questions, and Across-The-Board (ATB) constructions involving A'dependencies. In the Appendix, I present each type of A'-dependency existing both in English and in Chinese in great detail and I will not repeat all of them here for the sake of space.

Resumptivity  

... Two types of resumptive pronouns Existing work on resumption makes a distinction between two types of resumptive pronouns or two different uses of resumptive pronouns: the real resumptive pronouns and the intrusive pronouns. Some research work makes the same distinction by employing different terms: the real resumptive pronouns are also called grammatical resumptive pronouns or the grammatical use of the resumptive pronouns or the general/systematic use of the resumptive pronouns, whereas the intrusive pronouns are also referred to as intrusive use of the resumptive pronouns (Sells 1984). The general/grammatical/systematical use of the resumptive pronouns refers to the cases where resumptive pronouns appear systematically in an A'-bound position, i.e. the variable position generally occupied by gaps. This leads some scholars to think that resumptive pronouns and gaps are free alternatives (cf. Borer 1984, Koopman 1983, Engdahl 1980, 1985, Zaenen, Engdhal & Maling 1981). Contrary to the general use, resumptive pronouns can also be used in island contexts to prevent a sentence from eventually violating locality constraints such as Subjacency and ECP. Let us examine two frequently cited classical Hebrew examples from Sells (1984). (1) a.

raiti et ha-yeled šesaw-I Acc the-boy that ‘I saw the boy that Rina loves.’

rina Rina

ohevet loves

oto. him

b.

raiti et ha-yeled šesaw-I Acc the-boy that ‘I saw the boy that Rina loves.’

rina Rina

ohevet loves

___.

Hebrew, Sells (1984) In (1), the presence of the resumptive pronoun oto ‘him’ is optional in an islandfree context, which could suggest that this resumptive pronoun is in free alternation with gaps. In this case, the pronoun is in its systematic/ grammatical /general use and in other words, it is a “real” resumptive pronoun in the sense of Sells (1984). By contrast, the presence of the pronoun oto ‘him’ becomes obligatory in an island context such as the relative clause in (2), otherwise the sentence will be ungrammatical due to island effects. In this case, the resumptive pronoun oto ‘him’ is in its intrusive use and its presence will avoid potential violation of locality constraints. Therefore, the sentence in (2) remains grammatical with the presence of oto ‘him’.

  Resumptivity: an overview (2) a.

raiti saw-I

et ha-yeled ACC the-boy

še that

dalya Dalya

makira knows

et ACC

ha-iša še ohevet oto. the-woman that loves him ‘I saw the boyj [that Dalya knows the woman [that loves himj]].’ b. * raiti saw-I

et ha-yeled ACC the-boy

še that

dalya Dalya

makira knows

et ACC

ha-iša še ohevet _____. the-woman that loves (‘I saw the boyj [that Dalya knows the woman [that loves___j]].’) Hebrew, Borer (1984) In (2), there are two relative clauses, the outer one “the boy that Dalya knows…” and the inner one “the woman that loves”. The outer relative clause functions as an island that blocks the relativization of the boy due to violation of the Subjacency. However, since the boy is in the object position inside the inner island, ECP is not violated. In (2a), the insertion of the resumptive pronoun oto ‘him’ prevents the sentence from violating Subjacency. Thus, resumptive pronouns do not have the same function in (1) and in (2). ... Two different points of view The distinction between these two different uses of resumptive pronouns provokes a long-term debate on the typology of resumptivity. Certain previous studies try to argue that there is a macro-variation existing between languages that only use resumptive pronouns intrusively to avoid potential violation of locality constraints (i.e. as a last resort) and the languages that only use them as a general strategy to form A'-dependencies. In their sense, English and French belong to the former category of languages in that the presence of a resumptive pronoun in an island-free context will simply make a sentence ungrammatical (cf. 3a, b). However, the obligatory use of resumptive pronouns in certain possessive constructions is tolerated in island-free context as illustrated in (3c). (3) a.

* the bookj that I bought itj yesterday

b.

* le livrej que je lj’ai the book that I it-have (‘the book that I bought yesterday’)

acheté bought

hier yesterday

Resumptivity  

c.

The only kind of carj [which I can never seem to get itsj carburetor adjusted right] is them Stanley Steamers.

Let us take several often cited classical examples from Ross (1967) and Sells (1987) which illustrate that in English, resumptive pronouns can only be used in strong islands to avoid violation of island conditions or of the ECP. (4)

Island: complex NP (complement clause of noun) I just saw a girlj [who Long John’s claim that [shej was a Venusian]] made all the headlines. Ross (1967)

(5) a.

Island: wh-island created by an indirect question headed by if I’d like to meet the linguistj [that Mary couldn’t remember [if she had had seen (himj) before]].

b.

Island: if-conditionals Which of the linguistsj do you think [that [if Mary marries (himj) ] then everyone will be happy]? Sells (1987)

Postal (1998:111) also points out that resumptive pronouns can take morphologically invisible forms. In my approach, invisible resumptive pronouns are uniformly treated as pro. Previous works such as Engdahl (1980, 1985), Zaenen, Engdhal & Maling (1981), Koopman (1983) and Borer (1984) argue for a macro-variation typology on resumption. They reveal that in languages like Vata and Swedish, the use of resumptive pronouns merely represents a general strategy to establish an A'dependency and that resumptive pronouns can appear in every position where a gap appears in an A'-dependency. In other words, resumptive pronouns and the gaps created by wh-movement are free alternatives. In their view, resumptive pronouns are merely treated as spelled out traces. In fact, resumptive pronouns or gaps can only appear in several very restricted and specific positions in these languages. The macro-variation point of view has been quickly abandoned because more and more research has shown that most languages that dispose of the resumptive strategy use resumptive pronouns in both ways: the general/systematic/grammatical use and intrusive use. In particular, the strongest argument against the macro-variation point of view is based on the fact that in a non-trivial number of cases, resumptive pronouns and gaps can never be free

  Resumptivity: an overview alternatives; in other words, in certain specific positions, the presence of a resumptive pronoun is allowed but that of a gap in the same position is prohibited. I will cite two extreme cases from Aoun & Choueiri (2000) and from Rouveret (1994, 2011). Aoun & Choueiri (2000) show that in Lebanese Arabic, a relative clause without an island only permits the presence of a resumptive pronoun but not the presence of a gap, as illustrated in (6). (6)

S-Sabe yalli raħ tə∫ħaT*(o) Zeena mən the-boy that Fut-kick-3FSg him Zeena from l-madrase harab. the-school ran-away-3MSg ‘The boyj that Zeena will kick himj out of school ran away’ Lebanese Arabic, Aoun & Choueiri (2000)

Importantly, in (6), the resumptive pronoun o ‘him’ is not embedded within a strong island; however, the presence of o ‘him’ is still obligatory on the relativized site and the presence of the gap is excluded from the same position. Rouveret (1994, 2011) shows another extreme case of Welsh. In the position of a prepositional object (cf. 7) or a possessive object (cf. 8), the presence of a resumptive pronoun is always required and that of a gap is prohibited. (7)

a.

b.

(8)

Prepositional object y dyn y soniais the man that I-talked ‘the man I talked about’

amdan about-

y dyn y siaradasoch chwi the man that spoke you ‘the man that you spoke with’

*(o) [agr]

ag with

*(ef) [him]

Possessor y dyn yr oedd *(ei) fam gartref the man that was his mother at home ‘the man whose mother was at home’ Welsh, cited from Rouveret (2011)

Languages like Welsh does not permit prepositional stranding in that when the object NP of a preposition or of a possessor undergoes fronting (see Rouveret

Resumptivity  

2011 for a detailed presentation), it is obligatory to insert a resumptive pronoun in the object position, as illustrated in (7–8). A detailed description of this kind of data can be found in the Appendix and I will not repeat it here for the sake of space. Let me show an example of Hausa from Tuller (1986). Hausa behaves in a similar fashion with respect to Welsh. For instance, when the object NP of the preposition da ‘with’ in (9a) and that of the possessor in (9b) undergo whmovement, their base-positions must be respectively filled with the corresponding resumptive pronouns shii ‘him’ and nsa ‘of-his’ because gaps in the relevant positions would lead to ungrammatical sentences. (9) a.

b.

Prepositional object Waa ka nyi maganaa who 2SgM do talk ‘Who did you talk with?’ Possessor Waa ka karanta littaafi who 2SgM read book ‘Whose book did you read?’

da with

*(shii) ? [him]

*(nsa) of-[his] Hausa, Tuller (1986)

It is worthwhile pointing out that Mandarin Chinese behaves exactly like Welsh and Hausa when a prepositional object or a possessive object undergoes fronting. Again, the following examples have been discussed in great detail in the Appendix to which the reader can refer for a general distribution of resumptive pronouns in Mandarin. (10) a.

Relativization [我對 *(他 j) 很不錯]的那位朋友 j 1 [wo dui *(ta1j) hen bucuo] 1Sg to 3MSg very not.bad na-wei

de C

pengyouj

that-Cl friend ‘the friend to whom I am very kind’

 1 The Chinese characters used in this monograph are in their traditional style.

  Resumptivity: an overview b.

[由 *(她 j) 代表大家發言]的那位同事 j [you *(ta2j) daibiao dajia by 3FSg represent everyone na-wei

fa give

yan] talk

de C

tongshij

that-Cl colleague (Lit.) ‘the colleague that the talk by *(her) represented everyone’ = ‘the colleague who gave a talk representing everyone’ c.

[他們在 *(那兒 j) 渡過蜜月]的城堡 j [ta1-men zai *(narj) du-guo 3-Pl at there spend-Exp de

miyue] honeymoon

chengbaoj

C castle ‘the castle where they had their honeymoon’ d.

[當年以 *(那兒 j)作為公司總部]的大樓 j [dangnian yi *(narj) zuowei those.years with there used.as zongbu]

de

gongsi company

dalouj

headquarter C building ‘the building that was used as the headquarter of the company during those days’ After the fronting of the antecedent NPs for the reason of relativization, prepositions like dui ‘to (sb.)’, you ‘by (+ agent)’, zai ‘at (+ place)’ and yi ‘(using sth.) as’ can neither be stranded in-situ nor be fronted together with their NP-objects, and the only way to make the relevant sentences grammatical is to insert the corresponding resumptive pronouns on relativized sites. All of the above examples from different languages show that it is perfectly normal that a resumptive pronoun and a gap are not at all free alternatives. First, they are not alternatives in any position and second, natural languages show that in certain specific positions, the presence of a gap is never permitted whereas the presence of a resumptive pronoun is allowed. This observation based on the language facts constitutes a strong argument against the macrovariation point of view. Scholars like Rouveret (2011, to appear) argue for a micro-variation point of view in that languages cannot be different one from another in a radical way with respect to resumptivity. A clear-cut distinction between the languages that only use resumptive pronouns in a systematical way and those that only use them in an intrusive way should not be maintained. In

Resumptivity  

this sense, languages are different one from another in a trivial way concerning the use of the resumptive strategy. Scholars who believe that there is only a micro-variation concerning resumptivity claim that a real distinction should not be made between the languages that only use resumptive pronouns as a general strategy to form A'dependencies and those that only use them intrusively as a last resort; instead, a real distinction should be made between the languages that use resumptive pronouns in both ways (i.e. grammatical use and intrusive use) and those that only use them intrusively as a last resort. I will demonstrate through out Chapters 2–4 that Mandarin Chinese shows that the same language can perfectly use resumptive pronouns in both ways. In addition, a study based on Mandarin data suggests that these two uses of resumptive pronouns are determined by syntactic mechanisms/operations which derive different types of A'-dependencies where these resumptive pronouns appear, as will be detailed and outlined in this monograph. Another difference between resumptive pronouns and gaps is semantically oriented in that they do not allow the same interpretations even if they appear in exactly the same syntactic positions. It has been observed in Doron (1982) that in restrictive relative clauses, gaps in the direct object position give rise to two different types of individual readings: a specific individual reading (i.e. de re reading) and a non-specific individual reading (i.e. de dicto reading) (cf. 11a). By contrast, a resumptive pronoun in exactly the same object position in such restrictive relative clauses only allows the specific (de re) individual reading but not the non-specific (de dicto) reading (cf. 11b), as shown in the following examples. (11) a.

b.

Dani yimca et ha-išaj še hu mexapes ___j . Dani will-find ACC the-woman that he seeks ‘Dani will find the woman that he’s looking for.’ √ specific (de re) √ non-specific (de dicto) Dani yimca et ha-išaj še hu Dani will-find ACC the-woman that he ‘Dani will find the woman that he’s looking for.’ √ specific (de re)

-

mexapes seeks

otaj. her

?* non-specific (de dicto) Hebrew, Doron (1982)

Specifically, Doron (1982) shows that in Hebrew, when the trace is ccommanded by a universal quantified phrase, the relevant sentence will receive

  Resumptivity: an overview two readings: a singular individual reading and a multiple individual reading (i.e. pair-list reading), as shown in (12a). By contrast, if it is a resumptive pronoun that occupies the same position as the trace occupies, the pair-list reading is blocked, as illustrated in (12b). (12) a. ha-iša1 the-woman tišlax

še that lo2

kol every tmuna.

gever2 man

baxar chose

____1 ____

will-send him a-picture ‘The woman1 that every man2 chose ___1 will send him2 a picture.’ = (i) The same woman was chosen by all of the men. (Single individual reading) = (ii) For every man, there is a specific woman that he chose. (Multiple individual reading/ Pair-list reading) b. ha-iša1 the-woman tišlax

še that lo2

kol every tmuna.

gever2 man

baxar chose

ota1 her

will-send him a-picture ‘The woman1 that every man2 chose (*her1) will send him2 a picture.’ = (i) The same woman was chosen by all of the men. (Single individual reading) ≠ (ii) For every man, there is a specific woman that he chose. (Multiple individual reading/ Pair-list reading) Hebrew, Doron (1982) This is a very important contrast between gaps and resumptive pronouns, and I will come back to this point in Chapter 4. Importantly, these contrasts observed between the two types of readings confirm that semantically a resumptive pronoun behaves differently from a trace. Therefore, this can be taken as another strong argument against the point view according to which, resumptive pronouns are nothing more than the spelled out traces. .. Syntactic level: to move, or not to move, that is the question! Since a resumptive pronoun is A'-bound by an operator that is located in an A'position, it is crucial to know the mechanism that derives the A'-dependency established between these two elements. Historically, different analyses and solutions have been proposed during different periods of development of gen-

Resumptivity  

erative grammar: the period of the Government and Binding (GB) theory and the period of the Minimalist Program (MP). However, despite the change of theoretical model during each of these two periods, the center of the debate concerning resumptivity remains the same: it is important to know whether a resumptive dependency is derived by movement or not. This inquiry is also the crucial question that I will try to answer in this study. In the present section, I will go over briefly certain important analyses proposed previously in order to get a very general historical storyline of resumptivity. However, this brief presentation is neither complete nor exhaustive. ... Period of GB In the early 1980s, the analyses based on movement approach to resumptivity still considered the similarities between resumptive pronouns and the traces resulting from wh-movement as a crucial factor because these two types of elements demonstrate certain important common properties at the syntactic level. Since a wh-dependency is derived by an overt syntactic movement in languages such as English, if resumptive pronouns are comparable to wh-traces, then logically a resumptive dependency can also be derived by movement. It is precisely for this reason that many scholars argue for a movement-based approach to resumption. The two classical diagnostic tests for A'-movement are island effects and crossover effects. In GB framework, wh-movement is subject to locality conditions and gives rise to island effects and crossover effects. Following this line, if a resumptive dependency displays island effects and crossover effects, then one can conclude that the derivation of such a dependency involves movement. Relevant work in support of the movement approach to resumption is generally based on languages such as Vata and Swedish, in which resumptive pronouns and gaps are considered as free alternatives and they have exactly the same syntactic distribution in a very limited number of specifically restricted positions. For instance, Zaenen et al. (1981) claim that the relationship between a moved wh-element and a resumptive pronoun has the same nature as the relation existing between a moved wh-element and the trace that it left in-situ. The authors who argue that a resumptive pronoun is simply a spelled out trace defend of course the idea that it is possible to realize a trace resulting from movement phonologically/phonetically. However, it is important to point out that this kind of view immediately violates Economy Principle in the Minimalist Program. While still in the GB period, the movement approach received severe criticism because certain crucial properties demonstrated by wh-traces are not

  Resumptivity: an overview shared at all by resumptive pronouns in the same environment. Importantly, McCloskey (1990, 2001, 2002, 2005) distinguishes gaps from resumptive pronouns by claiming that gaps are always created by movement but resumptive pronouns are always created by binding. Concretely, the dependency established between a gap on the one hand and a moved wh-element or a relativized nominal head or a left-dislocated topic phrase on the other hand is always created by movement. The dependency established between a resumptive pronoun located on the variable site and its antecedent in relatives or in Left-Dislocation (LD) structures always involves a kind of binding either on the SurfaceStructure (S-S) or at Logical Form (LF). This generalization is based on Irish data. In Irish, a resumptive construction does not obey locality constraints and does not give rise to island effects, as illustrated in (13). (13)

ne dànta sin nach bhfuil fhios againn the poems these C-Neg is knowledge at-us cén àit ar cumadh iad what place C were-composed them ‘those poemsj [that we do not know [where theyj were composed]]’ Irish, McCloskey (1990)

A resumptive dependency does not give rise to crossover effects either, as shown in (14). (14)

an fear so ar mhairbh the man this C killed ‘this man that his own wife killed’

a bhean féin his-own-wife

é [him]

Irish, McCloskey (1990)

These two tests are considered as the strongest arguments in favor of an analysis based on a non-movement approach to the derivation of a resumptive dependency. However, as McCloskey points out himself, crossover effects are still observed in resumption in Irish if the crossed element is not a pronoun but an epithet. I will present McCloskey (1990)’s examples and the tests based on crossover effects in great detail in Chapter 2 and I will also apply such tests to Chinese data. Instead of going into detail here, let me just cite one of the crucial examples in McCloskey (1990) to illustrate the point. An epithet such as an bastard ‘the bastard’ in (15) refers to an fear ‘the man’. The binding relationship established between the resumptive pronoun sé ‘he’ and its antecedent an fear

Resumptivity  

‘the man’ crosses the epithet an bastard ‘the bastard’; all of these three elements bear the same index, and thus refer to the same individual. The ungrammaticality of the sentence in (15) is due to strong crossover effects, which suggests that movement must be involved within the resumptive dependency established between an fear ‘the man’ and sé ‘he’. (15)

*Sin an fearj ar dhuirt an bastardj go that the man C said the bastard C marodh séj muid. would-kill he us ‘That is the manj that the bastardj said hej would kill us.’ Irish, McCloskey (1990)

This test is then considered as one of the strongest arguments in favor of the analyses of resumption based on movement approach. Tallerman (1983) points out that in Welsh, weak resumptives are systematically excluded from island contexts, which is then treated as an argument in support of the idea that there are different types of resumptive pronouns and that they don’t behave uniformly. Here is a crucial example. (16)

?*Dyma’r dyn1 y cusanaist ti’r ddynes2 here the man that kissed you the woman a siaradodd amdano1 Rel talked about-[agr] ‘Here is the man1 [that you kissed the woman2 [who talked about him1]].’ Welsh, Tallerman (1983)

In (16), the agreement morpheme o attached to the preposition amdano ‘about’ is treated as the resumptive of the relativized antecedent NP dyn ‘the man’ (see below for why it has the status of agreement morpheme). The morpheme o is weak in the sense that it has no independent morphological form. In island contexts, this weak agreement resumptive o is excluded for exactly the same reason for which a gap is excluded. Therefore, (16) is treated as a strong argument in favor of the analyses based on movement approach. Authors during the golden period of GB in the early 1990s seem to agree with the idea that a resumptive construction is derived by movement; however, they were still debating the following questions: What is the element that has

  Resumptivity: an overview moved in a resumptive construction? At what level does the movement take place? Demirdache’s dissertation (1991) claims that in a restrictive relative clause, a null operator moves at LF in a resumptive chain and that a resumptive pronoun located on the variable site is comparable to the wh-pronoun staying insitu at the Surface-Structure in wh-in-situ languages. The only difference is that an operator-variable chain is created at LF in a resumptive dependency, while the same chain is created at S-S in the case of a wh-dependency.2 Rouveret (1994) still maintains the idea that a resumptive chain is only derived at syntax but not at LF; however, he proposes that it is not the entire antecedent NP that undergoes movement at syntax and that only a sub-part of that constituent is extracted from the relativized site, and this movement is called “sub-extraction”. Based on Welsh data, an important distinction has been made between resumptive pronouns that are base-generated and those derived by movement. The hypothesis behind this analysis is that the internal structure of a pronoun contains a head, which can be either a clitic or an agreement affix, and a pro as complement. In the derivation of a resumptive dependency, only the pro part undergoes movement to the operator position by leaving the head in-situ. The crucial point in this analysis is that the sub-extraction takes place in syntax (i.e. S-S) and that this movement is overt and subject to locality conditions such as Subjacency. This analysis can nicely account for the fact that weak resumptives in Welsh display island effects. By contrast, if the relativized site is located inside an island, then a strong resumptive pronoun must be used in order to save the sentence from violation of locality constraints. In the latter case, the binding relationship between the resumptive pronoun (i.e. strong pronoun) and the peripheral antecedent is established at LF without involving any movement at all. Let me cite a Welsh example. In (17), the relativized site is occupied by a strong pronoun represented by a full morphological form ef ‘him’. Even if this pronoun is embedded inside the island (i.e. the relative clause “the woman who…”), the sentence still remains grammatical. Following Rouveret’s solution, the A'-dependency established between the antecedent dyn (y…) ‘the man (that…)’ and the resumptive pronoun ef ‘him’ does not involve any movement and this binding relation is solely realized at LF and that is why no island effect is detected.

 2 In the framework of GB, locality constraints such as Subjacency, ECP and CED do not apply systematically both at surface-structure and at Logical Form.

Resumptivity  

(17)

Dyma’r here a Rel ‘Here is him1]].’

dyn1 the man siaradodd talked the man1 [that

y cusanaist ti’r ddynes2 that kissed you the woman amdano ef1 about-[agr] him you kissed the woman2 [who talked about Welsh, Tallerman (1983)

Historically, Rouveret’s solution based on a sub-extraction is the very first analysis of resumptive constructions that makes a link between the question of “movement or not” and internal structures of a pronoun. A strong pronoun is always base-generated in-situ on the variable site and inside its internal structure there is no such a component that can be a potential candidate for movement, which explains the fact that a strong pronoun never gives rise to island effects. By contrast, the pro part in the internal structure of a weak pronoun is a candidate for movement, which is the reason why island effects are only detected when a weak pronoun appears inside an island. This analysis has a very important influence on many of the later analyses on resumption in different languages; in particular, the hypothesis that a pronoun can have two different internal structures, an extended form and a reduced form, remains the core sprit of the later analyses in the framework of the Minimalist Program. During the period of MP, the contrast between strong pronouns and weak pronouns (Aoun et al. 2001 for Lebanese Arabic, Malkawi 2009 for Jordanian Arabic) and the contrast between two kinds of internal structures of a pronoun (Rouveret 2002, 2008, to appear and Guilliot 2006) have been extensively studied. As the reader will see in the next section, these ideas are extremely important to account for reconstruction effects observed in different types of resumptive constructions. ... Period of MP The debate on the question whether movement is involved in a resumptive chain continued in the Minimalist Program period. However, the operation Move in the framework of MP is defined quite differently from movement in the framework of GB. In Chapter 2, I will go over the minimalist devices and basic notions in a detailed way and in the present section I will only concentrate on the important representative analyses on resumption during this period. It is worthwhile pointing out that the analyses proposed in the framework of MP are very important for our discussion on Chinese data. In particular, analyses based on operation Agree are crucial for my analysis on resumption in Chinese pro-

  Resumptivity: an overview posed in this study. Again, in order to avoid unnecessary repetition, all of the explanations on derivational mechanisms in the Minimalist Program will be detailed in Chapter 2. I just want to give a very brief description on the most important notions used in MP here. Agree is an operation that establishes an agreement relationship between the head of a functional projection that works as Probe and another element that works as Goal. A Probe bearing a set of uninterpretable features F' searches for a Goal that bears the same set of features, say F, that matches the features F', in order to value F'. Agree is only activated when there are uninterpretable features (i.e. unvalued features) attached to the Probe and interpretable features (i.e. valued features) attached to the Goal. Interpretable features will value uninterpretable features and all of the valued features will be deleted before LF. Agree applies phase by phase in a cyclic fashion and solely at the Narrow Syntax level. Every well-established phase will be sent to interfaces for interpretation via operation Transfer. Once the relevant phase is transferred to LF and to PF, it becomes inaccessible for further operation. Therefore, Agree is subject to locality conditions and gives rise to island effects and crossover effects. The process of “searching for a Goal” is called “feature matching”. Matching is therefore defined as a kind of feature identity relation that exists between the features attached to the Probe and those attached to the Goal. When a lexical element enters into the process of a derivation, the features on this element can possess a value, but it is also possible that such an element does not possess any value at all. The crucial point is that Match does not require an absolute identical relation between the features on the Probe and those on the Goal but only a kind of identical Attribute, such as person, number and gender, of the features and in this sense, the Value of these matched features can be different on a Probe and on a Goal. The operation Move is only activated when the Probe not only bears the same features as the Goal does but also bears the EPP feature, which requires the relevant specifier position to be filled. The potential candidate for the Goal will thus move to the appropriate specifier position to fulfill the requirement of EPP. It is very important to highlight that in the Minimalist Program, the principal operation that establishes an A'-dependency is Agree not Move. In other words, Agree can derive almost all of the cases derived by movement in the framework of GB. Under such a reasoning, an A'-dependency can be established without any movement in MP. Two important analyses of resumption during that period, which are based on Agree, are proposed by Rouveret (2002, 2008, to appear) and by Adger & Ramchand (2005). These authors agree with the point that the highest complementizer functions as a Probe which will be translated as an operator in semantics and the resumptive pronoun downstairs functions

Resumptivity  

as a Goal which will be translated as a variable in semantics. Following this view, a resumptive construction is derived by phasal Agree. In the phase theory, vP and CP are strong phases and when the head of a higher phase (for instance, C°) is projected, the complement of the phasal head v° is sent to interfaces for interpretation via Transfer. In order to establish a connection between the resumptive pronoun located inside the complement of the phase vP and the antecedent or the closest complementizer which is located outside the phase vP, the derivation must be done phase by phase in a strict local fashion. Concretely, at the lowest vP phase, the first Agree relationship is established between the resumptive pronoun and the head of the phase vP before its domain VP is sent to the interfaces; then, the second Agree relationship is established between the phasal head v and the phasal head C. We should notice that in this version of analysis, Agree obeys two constraints: one is Condition on Phases and the other is locality constraints, such as island conditions. In Chapter 2, these two kinds of constraints will be elaborated and unified. One consequence of this analysis is to reduce all of the phenomena linked to resumptivity to the level of Narrow Syntax. Rouveret (2008) points out that such an analysis applies well in languages such as Welsh and that there is a morphological argument in favor of this analysis based on Agree. Let us examine his example, (18)

Beth yr ydych chwi yn ei ddisgwyl i what that are you Prog it expect for mi ei wneud? me it do ‘What do you expect me to do?’ Formal literary Welsh, Harlow (1981), Rouveret (2008, 2011)

In this example, the lower clitic ei ‘it’ attached to the verb-noun functions as a resumptive and it bears the same ϕ-features of the moved wh-phrase. The higher clitic ei ‘it’ in the mid-cycle is not a resumptive because it is not even bound by the antecedent in the A'-position. According to Rouveret, this middle ei is merely an agreement marker that recapitulates the ϕ-features of the relativized whphrase and it can be analyzed as a morphological reflex of the agreement relation established between the phasal head v of the matrix clause, the phasal head C and the phasal head v of the subordinate clause. This is a very important argument in favor of the analyses based on phasal Agree because it shows that ϕ-features can be involved directly during the establishment of the agreement relation in resumptive constructions. We should also notice that this agreement

  Resumptivity: an overview relationship does not require an absolute agreement of all of the ϕ-features, which confirms the idea that Agree does not necessarily apply to the entire set of ϕ-features, as proposed in Boeckx (2003). In Chapter 2, I will present in detail the analysis of Rouveret (2008, to appear) and the analysis of Adger & Ramchand (2005), both of which are based on the idea of Phasal Agree; however, they are different from one another in the relevant formal features associated with the C head and with resumptive pronouns respectively. .. Semantic level: to move, or not to move, that is still the question! As outlined at the beginning of this chapter, resumptivity does not only involve syntactic derivation, but also upon many complicated interpretation problems. One of the crucial semantic oriented problems is linked to reconstruction effects. I will examine this aspect of the problem in detail in Chapters 3 and 4. It has been observed that a resumptive construction gives rise to reconstruction effects (Aoun et al. 2001, Aoun & Li 2003). In the tradition of generative grammar, reconstruction effects constitute diagnostic tests for movement. The idea is that a constituent that has undergone movement can still get the same interpretation as it was in its initial position before the movement takes place, which gives an impression that after movement, such a constituent is reconstructed in its base-position and importantly, it is also interpreted in that position. In cases where a resumptive construction gives rise to reconstruction effects, this observation then constitutes an argument in favor of the analyses that support the idea that resumptive dependencies are derived by movement. However, as I will argue in Chapter 3 for Chinese, it is possible to disassociate reconstruction effects from the notion of movement. The presence of a copy on the variable site in an A'-dependency is enough to trigger reconstruction effects and this point has been extensively discussed in Rouveret (2002) and Guilliot (2006). Chapter 4 is devoted to a discussion of a very particular reading under reconstruction: the so-called Multiple Individual reading which is also named pair-list reading. Recall that in Section 2, I presented Doron (1982)’s semantic oriented argument against spelled out trace analysis of resumptive pronouns, which states that a resumptive construction systematically blocks reconstruction of the scope of a universal quantifier and that it also blocks a pair-list reading. Guilliot (2006)’s thesis also argues for this generalization based on data from Jordanian Arabic. By contrast, for Mandarin Chinese, I will take a different stance by adopting Rouveret (2002, 2008)’s analysis that postulates an ambiguity of the internal structure of resumptive pronouns in Welsh. The hypothesis is that a (resumptive) pronoun can have two different forms as its internal struc-

Main proposals in this study  

tures, an extended form [[D-ϕ] NP] and a reduced form [D-ϕ] and that only the extended form allows reconstruction. It is precisely the NP part in the extended form that gives rise to all reconstruction effects and in particular, to pair-list readings. By contrast, the reduced form lacking the NP part does not give rise to any of these effects. Different from Welsh, I will show that in Mandarin Chinese, different pronouns possess different internal structures; certain pronouns can have both forms but others can only have one of them. This hypothesis will explain the general distribution of reconstruction effects of different types of pronouns observed in Chinese. On the other hand, the study of Chinese also supports the claim that only the extended form can give rise to all of these different interpretative effects (including pair-list reading) under reconstruction.

. Main proposals in this study .. Against the macro-variation hypothesis The result of the survey on Chinese resumption in this monograph strongly rejects the macro-variation hypothesis. A macro-variation view cannot be on the right track in that, just like many other languages, Mandarin Chinese uses resumptive pronouns in both ways: the grammatical use and the intrusive use. This shows that the same language can perfectly use resumptive pronouns in different ways. In this study, in order to unify the terminology, “intrusive pronouns” have been replaced mostly by “the intrusive use of resumptive pronouns” which avoids potential violation of locality constraints in island contexts. In Chinese, there is no morphological distinction between so-called “real” resumptive pronouns and “intrusive” pronouns and it is precisely for this reason that I choose the terms based on “the use of”. Another important reason why I make such a choice in this study is that, as the reader will see, in the final conclusion, I claim that in Chinese, there are no two types of resumptive pronouns and instead, there is only one type of resumptive pronoun and that resumptive pronouns can never be used intrusively in Chinese. The repeatedly stated “fact” is that an intrusive pronoun can save a sentence from potential violation of locality constraints. As a matter of fact, it is not a correct view of the whole picture of resumption. The point of view that I will argue for is that it is not the case that there are so-called “intrusive pronouns” that can magically “prevent” a sentence from violating grammatical constraints as saving devices; the truth is quite another way around: those A'-dependencies (even if they contain an island) do not violate any locality constraint in the first place. This is because the syntactic mechanism/operation that derives these A'-dependencies

  Resumptivity: an overview is not subject to locality constraints and therefore does not give rise to island effects or crossover effects. .. Against the “spelled out traces” analysis In addition, Chinese shows that a resumptive pronoun has different syntactic and semantic properties compared with a trace. This study clearly illustrates that a gap, a resumptive pronoun in its general/grammatical use and a resumptive pronoun in its intrusive use exist independently in Chinese and that gaps and resumptive pronouns are not at all free alternatives anywhere. Gaps and resumptive pronouns are subject to different syntactic constraints and give rise to different semantic interpretative effects. Typologically, Chinese behaves differently with respect to languages like Vata and Swedish but similarly with respect to languages that make a clear distinction between resumptive pronouns and gaps. I will present previous work on languages such as Welsh (Rouveret 1994, 2002, 2008, to appear), Irish (McCoskey 1990, 2001) and different Arabic dialects (Aoun et al. 2001, Guilliot 2006, Malkawi 2009). These languages clearly reveal that a resumptive pronoun behaves differently from traces both in syntax and in semantics. .. Main proposals This study shows that there are different factors in the derivation of a resumptive construction. Two main factors are listed as the following: (I)

Syntactic mechanisms that derive the A'-dependencies containing resumptive pronouns

(II)

Internal structures of resumptive pronouns

Putting it generally, (I)

Syntactic mechanisms that derive different A'-dependencies (with a gap or with a resumptive pronoun);

(II)

Internal structures of A'-bound elements on variable sites (gaps or resumptive pronouns)

I will make two global generalizations based on these two factors respectively,

Main proposals in this study  

For (I): different types of A'-dependencies can be derived by different syntactic mechanisms; For (II): different types of A'-bound elements can have different internal structures. It is very important to highlight that there is no “cause-effect” relationship between these two generalizations, which means that each of them exists independently in languages. However, these two factors work together to derive different types of A'-dependencies: those with a gap as well as those with a resumptive pronoun. Resumptive constructions are only the result of one of the possible interactions between these two factors. I will show that the interaction of these two generalizations will enable us to account for all of the syntactic distribution and semantic properties observed in Chinese. Concretely, the following factors play important roles in the derivation process: (i)

Two types of A'-dependencies: relatives and dislocation structures (i.e. LD-structures and ex-situ cleft-focus structures);

(ii)

Two types of derivational mechanisms: Agree and Match;

(iii)

Two types of internal structures of A'-bound elements: extended form [[Dϕ] NP] and reduced form [D-ϕ];

(iv)

Three types of A'-bound elements: gaps, resumptive pronouns in their grammatical use and resumptive pronouns in their intrusive use;

(v)

Two types of pronouns of the third person: ta and qi;

These different factors interact in the following ways: (i)

Derivational mechanisms choose the types of A'-dependencies on which they operate. (a) Relatives (with a gap or with a resumptive pronoun) and dislocation structures with a gap are always derived by Agree. (b) Resumptive dislocation structures are derived by Match.

  Resumptivity: an overview (ii) Derivational mechanisms determine what type of resumptive pronoun appears in what type of A'-dependency. Agree applies phase by phase in a cyclic fashion; once the constructed phase is sent for interpretation, it will not be accessible again. Therefore, Agree is subject to locality constraints and gives rise to island effects and crossover effects. This is why the so-called intrusive pronouns (i.e. resumptive pronouns in the intrusive use) do not exist in relatives in Chinese. In an A'-dependency derived by Match alone (i.e. without Agree), each phase will not be sent immediately to interfaces before all of the phases (i.e. the entire sentence) are constructed. In other words, Match applies to the entire construction after each phase is well established. That is the reason why Match is not subject to locality constraints and does not give rise to island effects or crossover effects, which also explains why resumptive pronouns can only be used intrusively in dislocation structures in Chinese. Intrusive pronouns/ the intrusive use of resumptive pronouns can only exist in the A'-dependencies derived by Match. In other words, intrusive use cannot be found in the resumptive constructions derived by Agree. This is one of the crucial points that I will make in this study. In fact, the traditional description of the fact that it is intrusive pronouns that prevent an A'-dependency from violating locality constraints is simply an illusion. What I want to argue for Chinese is that there is no such a type of intrusive pronoun that can magically avoid any violation of locality constraints. Chinese data show that so-called intrusive pronouns are merely ordinary resumptive pronouns that exist in the A'dependencies (with an island) derived by Match. Since a resumptive dislocation structure is always derived by Match, which is not subject to locality constraints, a resumptive dislocation structure with an island is also derived by Match and that is why it does not display any island effect. In conclusion, it is not the presence of intrusive pronouns that saves the sentences; instead, the truth is that those sentences do not violate any locality constraint in the first place. Agree operates on gaps, which remains independent of the fact that these gaps are located on relativized sites or on dislocated sites. Gaps always give rise to island effects and crossover effects both in relatives and in dislocation structures. (iii) Derivational mechanisms also determine what type of feature is associated with what kind of A'-dependency.

Main proposals in this study  

There are two ways to satisfy the requirement of the EPP feature: either by movement or by base-generation via Merge.3 EPP feature is required to be activated in relatives (with a gap or with a resumptive pronoun) and in dislocation structures with a gap. There are different possible ways to look at this problem. We can assume that the NP antecedent in relatives and the topic in topicalization structures cannot be base-generated in their A'-positions and that these two types of elements are there to satisfy the requirement of EPP only via movement. The NP part in an extended form [[D-ϕ] NP] located on relativized sites or on dislocated sites is a potential candidate for movement in order to satisfy the requirement of EPP. Therefore, the EPP feature will require the extended form to be used as the internal structure of an A'bound element in an A'-dependency derived by Agree. In the case of resumptive dislocation structures, I assume that the hanging topic is always base-generated at TopP and the EPP feature is always satisfied automatically. As a result, EPP does not require the extended form of resumptive pronouns on dislocated sites to be used. Alternatively, another possible way to explain the fact is that the EPP feature is simply inactive in A'dependencies derived by Match. In other words, Match does not require EPP to be activated. There are two kinds of locality constraints: the conditions that derive intervention effects and island effects and conditions on phases. I will argue that in the case of resumption, only conditions on phases will be necessary and enough to account for all of the observed facts in Chinese. (iv) Internal structures of the elements located on relativized sites or dislocated sites are responsible for reconstruction effects. It is the presence of the NP part in an extended form [[D-ϕ] NP] that gives rise to reconstruction effects as well as different kinds of distributive readings without taking into consideration the types of A'-dependencies (i.e. relatives or dislocation structures) in which such an extended form appears. A reduced form [D-ϕ] never gives rise to any of the reconstruction effects or any of the distributive readings. (v)

Different types of pronouns possess different internal structures.

 3 In the case of expletives in English, there is base-generated (i.e. inserted) on [Spec, TP] to satisfy the requirement of EPP.

  Resumptivity: an overview (a)

(b)

Certain A'-bound elements possess only one specific form as their internal structures, for instance, gaps and the pronoun qi (3rd person) only have the extended form as their internal structures. They systematically give rise to reconstruction effects both in relatives and in dislocation structures. Other A'-bound elements have two internal structures, such as the pronoun ta (3rd person singular). This type of pronoun never determines to appear in what type of syntactic construction with what form as its internal structure. As I explained above, it is the EPP feature of the Probe that correlates with the specific form of an A'-bound pronoun or any A'-bound element in a general sense.

(vi) It is important to notice that in the interaction of all of these different factors, the presence or absence of islands in the relevant A'-dependencies does not play any significant role.

. Organization of the argumentation One of the main research purposes of this monograph is to systematically examine the general syntactic distribution and the distribution of reconstruction effects of gaps and of resumptive pronouns (including the intrusive use of resumptive pronouns) in two important types of A'-dependencies: relatives and dislocation structures (i.e. Left-Dislocation structures and ex-situ cleft-focus structures). Chapters 2, 3 and 4 focus on each aspect of the problems related to resumptivity; namely, Chapter 2 will concentrate on the syntactic derivation of resumptive dependencies; Chapters 3 and 4 on semantic interpretative effects of gaps and of resumptive pronouns. In Chapter 2, based on the two classical diagnostic tests for A'-movement – island effects and crossover effects (weak and strong) – I will examine the distribution of the gaps and of resumptive pronouns in relatives, LD-structures and ex-situ cleft-focus structures. With the results of the tests at hand, I will give a detailed presentation on the minimalist mechanisms that will be used to interpret these results. Certain previous analyses such as Rouveret (2008, to appear) and Adger & Ramchand (2005) will be presented in order to see whether they can be applied to account for the observed data in Chinese. Then, I will propose my own analyses of relatives and dislocation structures both in the framework of the Government and Binding (GB) theory and in the Minimalist Program (MP). I would like to underline that the analyses as well as the solutions proposed in this study will still mainly rely on the minimalist mechanisms. I will propose that an A'-dependency with a gap in relatives and in dislocation struc-

Organization of the argumentation  

tures are derived by operation Agree alone without Move and that a resumptive relative clause is also derived by Agree alone. By contrast, a resumptive dislocation structure is derived by Match alone without Agree. Chapter 3 will be focusing on reconstruction effects in resumptive constructions in Chinese. I will apply the following four tests, which have been used quite often in the literature, to Chinese: (i)

The possibility to reconstruct the scope of a universal quantifier in A'dependencies with gaps and with resumptive pronouns;

(ii)

The possibility for a gap and for a resumptive pronoun to take a quantified expression as antecedent;

(iii) The possibility to reconstruct a binding relationship when the antecedent contains an anaphoric element in A'-dependencies with gaps and with resumptive pronouns; (iv) The possibility to give rise to condition C effects of the binding theory under the reconstruction in A'-dependencies with gaps and with resumptive pronouns. Based on the results of these four tests, I will make the following generalizations: (i)

Resumptive constructions can give rise to reconstruction effects;

(ii) However, not all of the different types of resumptive constructions can systematically give rise to reconstruction effects. Following the original idea of Rouveret (2002, 2008), I will propose that in Chinese, the extended form [[D-ϕ] NP] systematically gives rise to all different reconstruction effects (including condition C effects) and the reduced form [D-ϕ] does not give rise to any of these effects. I will also carefully examine three types of A'-bound elements: gaps, the pronoun ta (3rd person) and the pronoun qi (3rd person) and I will conclude that these three types of elements do not have the same internal structures. This point of view will help us to account for the distribution of different reconstruction effects observed in this chapter. At the end of Chapter 3, I will discuss the core solution and the main analyses proposed in this study. Importantly, I will

  Resumptivity: an overview claim that the derivation of a resumptive dependency, or even generally, an A'dependency, is subject to an interaction of different factors. Chapter 4 examines in particular the so-called specificity effects in resumptive constructions. Based on the two important distinctions made in Doron (1982), I will examine the distribution of specific individual readings and nonspecific individual readings on the one hand and the distribution of natural functional readings and pair-list readings on the other hand. Languages like Hebrew show that resumptivity systematically blocks pair-list readings. I will present some important analyses in the previous literature concerning this point. However, Chinese data show that resumptivity never blocks pair-list readings and this observation is on the same direction as what observed in Welsh in Rouveret (2008, to appear). I will conclude that the availability of pairlist readings heavily relies on the choice of internal structures of resumptive pronouns. The extended form permits a reconstruction of the scope of a universal quantifier and gives rise to pair-list readings. In Concluding remarks, I will discuss the advantages of my analyses as well as the predictions made by them. The Appendix is also very important in that I present a general distribution of resumptive pronouns in Chinese as well as all of the semantic constraints on the use of resumptives. I will begin with an introduction to different types of A'dependencies that permit the presence of resumptive pronouns, such as relatives, left-dislocation structures and ex-situ cleft-focus structures, ATBconstructions and wh-ex-situ cases. Then, I will examine the distribution of resumptive pronouns in two different categories of contexts: the cases where their presence is obligatory and the cases where their presence is optional or even prohibited.

 Syntactic derivation: two types of A'dependencies . Introduction In the first chapter I presented several analyses that aim to account for all of the observed syntactic properties concerning resumption based on different theoretical approaches in the literature. Let me summarize these analyses below. A resumptive dependency can be derived by: (i)

Movement at Surface-Structure (S-S) or at Logical Form (LF) in the framework of Government and Binding Theory (Sells 1984, Tellier 1991, Demirdache 1991);

(ii)

Sub-extraction of a part of a resumptive pronoun (Rouveret 1994, Guilliot 2006);

(iii)

Match followed by Move in the Minimalist Program (Boeckx 2003);

(iv)

Agree alone without Move in the Minimalist Program (Adger & Ramchand 2005, Rouveret 2002, 2008, to appear)

In this chapter I will discuss how to determine which derivational mechanism is the one that we need to account for the syntactic properties displayed by the resumptive pronouns based on the empiric data from Mandarin Chinese that we will observe. Meanwhile, I will show what Chinese can tell us about different phenomena related to resumptivity from an even more general point of view. There are very little works that systematically study resumptive pronouns in Chinese and the existing work do not offer a clear, complete or a panoramic view of the general distribution of resumptive pronouns in Chinese. This chapter aims to give the reader an overall view of the distribution of the gaps as well as the distribution of the resumptive pronouns in different types of A'dependencies in Chinese.4 The argumentation in this chapter is organized as the following.  4 The PhD thesis of Gu (2001) argues for a base-generation approach to all of the types of resumptive constructions in different languages. The Master thesis of Ning (2007) discusses

  Syntactic derivation: two types of A'-dependencies Recall that in the first chapter I presented several earlier analyses of resumption; the debate is focused on the following two aspects of the problem: first, is a resumptive pronoun simply a phonologically spelled out trace? And second, is a resumptive construction necessarily derived by movement? In order to answer these two questions, in this chapter I will begin by presenting the original data collected from Mandarin Chinese. I will examine the distribution of resumptive pronouns in three different types of A'-dependencies: (i) relative clauses, (ii) LD-structures and (iii) ex-situ cleft-focus structures.5 In each of these three cases, I will examine the distribution of gaps on the one hand and that of resumptive pronouns and intrusive pronouns on the other. Meanwhile, in each type of A'-dependency I will study two types of contexts: those without the presence of islands and those involving islands. In order to see clearly whether the three types of A'-dependencies that I will examine, relatives, LDstructures and ex-situ clefts, are all derived by movement, I will apply two classical diagnostic tests for movement to Chinese data. These tests are based on island effects and crossover effects (strong and weak). I will start from the traditional consideration in the GB framework that island effects and crossover effects are reliable diagnostics for movement. Then, I will interprete the results based on these tests in the Minimalist Program by claiming that movement is not necessarily involved in resumptive dependencies that give rise to these two types of effects. The following table is presented as a map of the itinerary of the argumentation that will make reading easier and help the reader follow the logic of the reasoning more easily. My first task in this chapter is to complete each cell of the table with the results of the diagnostic tests.6  resumptive pronouns from the point of view of processing. However, neither of these two studies really applies the syntactic or semantic diagnostic tests to Chinese resumptive constructions in order to get an overview of the whole picture of the general distribution of resumptive pronouns. Aoun and Li (2003) discuss the link between movement and the reconstruction effects displayed by resumptive dependencies. Huang, Li and Li (2009) focus on the presence of resumptive pronouns in passive constructions and on the difference between a resumptive pronoun and a null object in different syntactic contexts. In the Appendix, I will give a detailed description of the distribution of resumptive pronouns in island-free contexts and the reader will see that the use of resumptive pronouns in Mandarin is still conditioned. 5 Please refer to the Appendix for a detailed presentation and several possible syntactic analyses on the so-called ex-situ cleft-focus structures in Mandarin Chinese. In the Appendix, I also justify my choice of the A'-dependencies to be examined in this study. As the reader will see, although a wh-question is treated as a typical A'-dependency, resumptive pronouns are excluded from wh-questions due to the wh-in-situ characteristic in Mandarin Chinese. 6 Partial results of these tests presented in this chapter have been discussed in Pan (2016).

Introduction   Tab. 1: Itinerary Relatives

---

Intrusive

Resumptive

--

Ex-situ clefts Gap

--

Intrusive

Resumptive

Gap

With an island

Intrusive

Resumptive

Gap

Without any island

Left-Dislocations

---

--

Weak Crossover effects

--

--

--

Strong Crossover effects

--

--

--

The table has three parts. If we look at it horizontally, there are three types of A'dependencies that I am going to examine: relatives, LD(-structures) and ex-situ cleft-(focus structures). Under each construction, there are three A'-bound elements to examine: gaps, resumptive pronouns and intrusive pronouns. If we look at the table vertically, there are four syntactic tests that I will apply: sentences without any island, sentences with an island, strong and weak crossover effects. Please also note that irrelevant cells are marked by “--”. Once the results of these tests are obtained, the distribution of gaps and of resumptive pronouns in different constructions will be clear, and then I will discuss the question: what is the best derivational mechanism to adopt in order to account for these results. I will interpret the results and try to answer this question in two different frameworks: Government and Binding Theory and Minimalist Program (cf. Section 2.3). I will argue that as far as the resumptive strategy goes, LD-structures and ex-situ cleft-focus structures are derived by the same mechanism in Chinese but relatives are derived differently. I will henceforth refer to LD-structures and ex-situ cleft-focus structures as “dislocation structures” when the distinction between them is not directly relevant to the discussion. Relatives (either with gaps or with resumptive pronouns) and dislocation structures with gaps are always derived by movement in the framework of Government and Binding or by operation Agree alone without Move in the framework of the Minimalist Program. All of these mechanisms work strictly at the level of Narrow Syntax. As mentioned briefly in Chapter 1 and as will be

  Syntactic derivation: two types of A'-dependencies detailed in Section 2.3 of the present chapter, Agree applies phase by phase in a strict cyclic fashion, which is the reason why it must obey locality conditions. By contrast, in resumptive dislocation structures (either resumptive LDstructures or resumptive ex-situ cleft-focus structures), the hanging topic (i.e. left-dislocated element) and the ex-situ focus element are always directly basegenerated in-situ in an A'-position in the left-periphery in Chinese, which implies that movement is not involved in the derivation of these two structures. In the Government and Binding theory, the link established between the topic or the focus in the left-periphery on the one hand, and the resumptive pronoun located on the dislocated site inside the TP on the other hand, can be realized after Logical Form, for instance, by a kind of linking or binding at the discourse level. However, in the Minimalist Program, I will propose that such a linking is established by the operation Match. Match aims at identifying a kind of nondistinctiveness relation between the Probe (i.e. the higher Top° or Foc°) and the Goal (i.e. the resumptive pronouns) in terms of the formal features attached to them respectively. Match is not any more considered as a pre-condition on Agree but as an independent syntactic mechanism that establishes an A'dependency. For some reasons which will be specified in detail in Section 2.3, in the case of the derivation of dislocation structures, Match will not be followed by Agree, and Match applies to the entire sentence after all of the different phases are constructed. This explains why Match is not subject to locality constraints.

. Distribution of resumptive pronouns in relatives and in dislocation structures .. Islands Ross (1967) discovered that certain syntactic structures block the movement of a wh-phrase from its base position to the sentence initial A'-position in English and the extraction of the relevant wh-phrase out of these constructions makes the relevant sentences ungrammatical. These blocking structures are called islands for A'-movement in general. In the Government and Binding framework, constraints called “locality conditions” require that A'-movement must take place in a cyclic fashion and that conditions such as Subjacency should not be violated. In the Minimalist Program, these different locality conditions have been reformulated under the theory of derivation by phase, for instance, movement must take pace phase by phase; I will come back to these notions in Section 2.3. Now let us concentrate on islands because island effects constitute the

Distribution of resumptive pronouns in relatives and in dislocation structures  

most important diagnostic tests for resumptive constructions in this chapter. Typical islands are listed below and I will use English and French examples.7 These two languages behave alike with respect to wh-extraction from islands. (1)

a.

Wh-island (island created by indirect wh-questions) * Whoj do you wonder [CP wheni John will meet tj ti]?

b.

* Quij

penses-tu

[CP quandi

Jean

rencontrera

tj ti]?

who think-you when Jean meet.Fut (‘For which person x, such that you think when Jean will meet x?’) (2)

a.

Complex-NP created by relative clauses * Whoj do you like [DP the books [CP that [TP tj wrote]]]?

b.

* Quij who

aimes-tu like-you

[DP les livres

[CP que

the books

[TP tj

that

a has

écrits]]]? written (‘For which person x, such that you like the books that x wrote?’) (3)

a.

Complex-NP created by complement clauses of noun * Whoj did you hear [DP the rumor [CP that [TP Mary insulted tj yesterday in a meeting]]]?

b.

* Quij who

as-tu

entendu

have-you

heard

[TP Marie a insulté Mary has insulted une réunion]]]?

[DP la

nouvelle

the tj

[CP que

news

hier yesterday

that dans in

a meeting (‘For which person x, such that you heard the rumor that Mary insulted x in a meeting yesterday?’) (4)

a.

Adjunct island * Which bookj do you want to go to France [CP after [TP you read tj ]]?

 7 I will not make a distinction between weak islands and strong islands à la Cinque in the present chapter.

  Syntactic derivation: two types of A'-dependencies

b.

* Quel which

livrej

veux

tu

aller

visiter

la

book

want

you

go

visit

the

France [CP après que [TP tu as lu tj ]]]? France after that you have read (‘For which book x, such that you want to visit France after you read x?’) (5)

a.

Sentential subject island * Whatj are [TP you going to study tj ] will make your mum happy?

b.

* Quelle which

chansonj

est-ce

qu’

song

is-it

that

[TP entendre tj ] listen

te you

rend mal à la tête? render headache (‘For which song x, such that listening to x will give you a headache?’) In English, if a resumptive pronoun embedded within an island is in the direct object position, the relevant sentence is generally ungrammatical (cf. 6a) and similarly, French shows exactly the same pattern as illustrated in (6b). (6)

a.

?* the boyj [who John knows the girli [whoi kissed himj]]

b.

?* le the [quii who

garçonj boy lj’a him-has

[que

Jean

that Jean embrassé]] kissed

connaît

la

fillei

know

the

girl

(‘the boyj who Jean knows the girl who kissed himj.) By contrast, it has been observed that if the resumptive pronoun appears in the subject position inside an island, the sentences can be improved considerably. Let me cite one of the famous examples in English (cf. 7) provided in Ross (1967). (7)

I just saw a girlj [whoj Long John’s claim [that shej was a Venusian]] made all the headlines. Ross (1967)

Distribution of resumptive pronouns in relatives and in dislocation structures  

The same observation goes for French as illustrated in (8).8 (8) a.

b.

? le garçonj [que Jean connaît la fillei the boy that Jean knows the girl [à quii ilj a donné ce livre ti]] to whom he has given this book ? ‘the boyj [who Jean knows the girli [to whomi hej gave this book ti]]’ ? J’ai rencontré l’actricej [que Marie connaît I-have met the-actress who Mary knows le garçon [qu’ellej aime bien]]. the boy that-she likes much ? ‘I met the actressj [who Mary knows the boyi [whomi shej likes ti very much]].’

Each of the two sentences in (8) involves two relative clauses; the outer one constitutes an island that blocks the relativization of the antecedent NP from inside the inner relative clause. In both cases, it is the subject of the inner relative clause is relativized and in the relevant relativized site, a resumptive pronoun is inserted. The presence of resumptive pronouns in the subject position considerably improves the sentences. I must also admit that many native speakers that I consulted still do not like these sentences. ... Relatives As for Chinese, in a relative clause without an island, a resumptive pronoun and a gap can be free alternatives in certain positions and under certain specific conditions.9 Examples in (9) illustrate the cases in which a resumptive pronoun (either in object position or in subject position) is deeply embedded within subordinate clauses. However, none of these subordinate clauses are considered as islands for A'-movement.

 8 Guilliot (2006)’s dissertation (written in French) systematically applies these tests to relatives and LD-structures in French. I will not cite his examples here for the sake of space. 9 Please refer to the Appendix for a general distribution of resumptive pronouns in relatives and in dislocation structures in Chinese. Normally, the presence of a resumptive pronoun in relatives without island is optional. (i) [Wo zuotian da-le (taj) de] na-ge renj. 1Sg yesterday beat-Perf 3MSg C that-Cl person ‘the person who I beat yesterday’

  Syntactic derivation: two types of A'-dependencies (9) a.

Without island: direct object [一勤說[笑適相信[李燕會見到(他 j)/____j ]]]的那個人 j 是位醫生。 [Yiqin shuo [Xiaoshi xiangxin [Liyan hui jiandao Yiqin say Xiaoshi believe Liyan Fut see ta1j/____j]]] de na-ge renj shi wei yisheng. 3MSg C that-Cl person be Cl doctor ‘The personj [who Yinqin says [that Xiaoshi believes [that Liyan will see ___j]]] is a doctor.’

b.

Without island: subject [一勤說[李燕相信[(他 j)/____一定會及格]]]的那位學生 j 結果沒來參 加考試。 [Yiqin shuo [Liyan xiangxin [ta1j/____j yiding hui Yiqin say Liyan believe 3MSg certainly Fut jige ]]] de na-wei xueshengj jieguo mei lai pass.exam C that-Cl student finally Neg come canjia

kaoshi.

participate exam ‘The studentj [who Yiqin says [that Liyan believes [that hej/___j will certainly pass the exam]]] finally didn’t come for the exam.’ From this perspective, Chinese behaves like Hebrew in that the presence of a resumptive is optional in a relative clause without island, as shown in (10). (10)

raiti et ha-yeled šesaw-I Acc the-boy that ‘I saw the boy that Rina loves.’

rina Rina

ohevet loves

(oto). him Hebrew, Borer (1984)

In the same island-free context, French and English do not tolerate the presence of any resumptive pronoun in relative clauses, as shown in (11). (11) a.

J’ai I-have (*l’)

rencontré le met the aime bien.

garçon boy

que whom

Mary likes much ‘I met the boy whom Mary likes very much.’

Marie Mary

Distribution of resumptive pronouns in relatives and in dislocation structures  

b. I met the boy that Mary loves (*him). Importantly, in Chinese a resumptive pronoun cannot work intrusively as the last resort to save a sentence from potential violation of locality constraints in relative clauses. In other words, intrusive pronouns do not exist in relatives in Chinese. Here are some crucial examples in which I only mark the object gap since the subject gap is not relevant here. Sentences in (12–13) involve complex-NP islands constructed by relative clauses and the gaps created by relativization lead to ungrammatical sentences as shown in (a) cases. However, the presence of resumptive pronouns on relativized sites inside islands does not remedy the relevant sentences from strong island violation as illustrated in (b) cases. (12)

a.

b.

(13)

a.

Complex-NP island (relatives) *我碰到了[小倩認識[擁抱過___j 的]那位女同學的]法國影星 j。 * Wo pengdao-le [Xiaoqian renshi [yongbao-guo ____j ] 1Sg meet-Perf Xiaoqian know embrace-Exp de na-wei nü-tongxue de] Faguo yingxingj . C that-Cl female.student C French movie.star (‘I met the French movie starj [that Xiaoqian knows the female student [who embraced ___j]].’) *我碰到了[小倩認識[擁抱他 j 的]那位女同學的]法國影星 j。 * Wo pengdao-le [Xiaoqian renshi [yongbao-guo ta1j ] 1Sg meet-Perf Xiaoqian know embrace-Exp 3MSg de na-wei nü-tongxue de] Faguo yingxingj . C that-Cl female.student C French movie.star (‘I met the French movie starj [that Xiaoqian knows the female student [who embraced (himj)]].’) Complex-NP island (relatives) * 這是[我見到過[談論過____j 的] 那個女同學的]作家 j。 * Zhe shi [wo jiandao-guo [tanlun-guo ____j this be 1Sg meet-Exp talk-Exp de] na-ge nütongxue de] zuojiaj. C that-Cl female.student C writer (‘This is the writerj [whom I met the student [who talked about tj]].’)

  Syntactic derivation: two types of A'-dependencies b.

* 這是[我見到過[談論過他 j 的] 那個女同學的]作家 j。 * Zhe shi [wo jiandao-guo [tanlun-guo ta1j this be 1Sg meet-Exp talk-Exp 3MSg de] na-ge nütongxue de] zuojiaj. C that-Cl female.student C writer (‘This is the writerj [whom I met the student [who talked about (himj)]].’)

These two sets of examples illustrate the case where a gap and a resumptive pronoun are embedded within a complex-NP island respectively. In (12), there are two relative clauses: the inner one “the girl who knows…” constitutes the A'dependency that I will examine and the outer one “the French movie star that…” constitutes an island that blocks the relativization of the antecedent NP Faguo yingxing ‘French movie star’. The NP French movie star is relativized and is therefore extracted out of the island and violates locality condition, namely Subjacency here, which explains the ungrammaticality of this sentence. A crucial fact that we can observe here is that in relatives, island effects are always detected irrespective of whether it is a gap or a resumptive pronoun that occupies the relativized site. In other words, in relatives the insertion of a resumptive pronoun on the relativized site embedded within an island will not prevent a sentence from violating locality constrains. Intrusive pronouns cannot exist in relatives in Chinese. The examples in (13) illustrate exactly the same point. There are cases where the presence of a resumptive pronoun is obligatory, for instance, when the object NP of a preposition undergoes fronting in the case of relativization or left-dislocation, the preposition cannot be stranded in its base position and it cannot undergo pied-piping with the object NP either, and in this case a resumptive pronoun must be inserted in the object position. I illustrate this point with concrete examples in Section 6.4 of the Appendix. Recall that inside a relative clause with an island, the presence of a resumptive pronoun cannot save the sentence from violation of locality constraints in Chinese (cf. 12–13). The point concerned directly here is that even if in the cases where the presence of a resumptive pronoun is considered obligatory, if such a resumptive pronoun is embedded within an island in a relative clause, the sentence is always ungrammatical, which means that the obligatory resumptive pronoun cannot save a sentence from violation of the locality constraints either. Let us examine an example. In (14b), when the direct object na-ge huaping ‘that

Distribution of resumptive pronouns in relatives and in dislocation structures  

vase’ is fronted to a preverbal position, it is preceded by the marker ba.10 When the ba marked direct object undergoes relativization, ba cannot be stranded on the relativized site and in this case the insertion of the pronoun ta3 ‘it’ as resumptive is obligatory. (14)

a.

b.

Complex-NP island (relatives) *這就是[我罵了 [打碎了___j 的] 那個男孩兒的]那個花瓶 j。 * Zhe jiu shi [wo ma-le [da-sui-le ___j this exactly be 1Sg scold-Perf break-Perf de] na-ge nanhair de] na-ge huapingj. C that-Cl boy C that-Cl vase (‘This is exactly the vasej [that I scolded the boy [who broke ____j]].’) *這就是[我罵了 [把它 j 打碎了的] 那個男孩兒的]那個花瓶 j。 * Zhe jiu shi [wo ma-le [ba ta3j this exactly be 1Sg scold-Perf BA 3OSg [da-sui-le de] na-ge nanhair de] na-ge break-Perf C that-Cl boy C that-Cl huapingj. vase (‘This is exactly the vasej [that I scolded the boy [who broke itj]].’)

Such an observation also applies to the case of possessors.11 When the possessor NP is relativized a resumptive pronoun must be used on the relativized site; however, if such a resumptive pronoun is embedded within an island, for instance a complex-NP in (15), the sentence is ungrammatical due to island effects. The insertion of the resumptive pronoun ta1-de ‘his’ in the relativized position of the original possessor zuojia-de ‘writer’s’ inside the island does not prevent this sentence from violating strong island condition and therefore, the sentence remains ungrammatical.

 10 The ba-construction is one of the important research topics in Chinese both in traditional descriptive grammar and in generative grammar. Much of the previous literature extensively discusses the use of ba and the mainstream point of view is to treat ba as the head of an independent functional projection. 11 Please refer to the Appendix for the distribution of resumptive pronouns in the case of possessive constructions.

  Syntactic derivation: two types of A'-dependencies (15)

* 這是[我見到了[買下了他的 j 房子的]那個女同學]的作家 j。 * Zhe shi [wo jiandao-le [maixia-le ta1j-de fangzi this be 1Sg meet-Perf buy-Perf 3MSg-Gen house de] na-ge nütongxue de] zuojiaj. C that-Cl female.student C writer (‘This is the writerj [that I met the female student [who bought hisj house]].’)

In order to reinforce my arguments, I will apply the same test to other types of islands. There are two types of complex-NP islands: one is constructed by relative clauses and the other by complement clauses of noun. Extraction of a constituent out of a complement clause of a noun to an A'-position will make the relevant examples ungrammatical.12 For instance, (16)

a.

b.

Complex-NP: complement clause of noun * [[[瑪麗親了___ j 的]消息]傳遍了全醫院的]那個醫生 j * [[ NP [Mali qin-le ___ j de] xiaoxi] chuan-bian-le Mary kiss-Perf C rumor spread-Perf quan yiyuan de] na-ge yishengj entire hospital C that-Cl doctor (‘the doctorj [that [the rumor [that Mary kissed ___j ]] was spread everywhere in the hospital]’) * [[[瑪麗親了他 j 的]消息]傳遍了全醫院的]那個醫生 j * [[ NP [Mali qin-le ta1j de] xiaoxi] chuan-bian-le Mary kiss-Perf 3MSg C rumor spread-Perf quan yiyuan de] na-ge yishengj entire hospital C that-Cl doctor (‘the doctorj [that [the rumor [that Mary kissed himj ]] was spread everywhere in the hospital]’)

In (16), the subordinate clause “that Mary kissed the doctor” is the complement of the nominal head rumor. When the DP the doctor is relativized, it is extracted out of the complement clause, thus out of an island. In this situation, locality  12 The reader should pay special attention to the examples related to relatives in this chapter. Most of these examples are not full sentences but only NPs taking complex modifiers or complex complements. When reading these examples, please refer to the bracketing used to indicate the relevant syntactic structures.

Distribution of resumptive pronouns in relatives and in dislocation structures  

constraints are violated, which explains the ungrammaticality of the sentence in (16a). Even if the resumptive pronoun ta1 ‘he’ is inserted on the relativized site, violation of locality conditions cannot be avoided as illustrated in (16b). Therefore, this test based on the complement clause of noun confirms the result of the test based on complex-NP islands constructed by relatives. The following two examples, (17) and (18) illustrate the fact that the same result can be obtained with islands constructed by an adjunct clause and by a sentential subject respectively. When the relativized site is embedded within these two types of islands, regardless whether such a site is occupied by a gap (cf. 17b, 18a) or by a resumptive pronoun (cf. 17c, 18b) , the sentences are always ungrammatical. (17)

a.

b.

Adjunct clause island: extraction of a direct object [因為瑪麗親了那個醫生] 整個學校的男老師都很鬱悶. [Yinwei Mali qin-le na-ge yisheng] zheng-ge because Mary kiss-Perf that-Cl doctor entire-Cl xuexiao de nanlaoshi dou hen yumen. school DE male.teacher all very unhappy ‘All of the male teachers of the school are very unhappy [because Mary kissed the doctor].’ * [[因為瑪麗親了___j] 整個學校的男老師都很鬱悶]的那個醫生 j * [yinwei Mali qin-le ____j ] zheng-ge xuexiao because Mary kiss-Perf entire-Cl school de nanlaoshi dou hen yumen de] DE male.teacher all very unhappy C na-ge

yisheng

that-Cl doctor (‘the doctor j [that all of the male teachers of the school are very unhappy [because Mary kissed ___ j]]’) c.

* [[因為瑪麗親了他 j]整個學校的男老師都很鬱悶]的那個醫生 j * [yinwei Mali qin-le ta1j] zheng-ge xuexiao because Mary kiss-Perf 3MSg entire-Cl school de nanlaoshi dou hen yumen de] DE male.teacher all very unhappy C na-ge

yisheng

that-Cl doctor (‘the doctor j [that all of the male teachers of the school are very unhappy [because Mary kissed himj]]’)

  Syntactic derivation: two types of A'-dependencies (18)

a.

b.

Sentential subject island: extraction of an adjunct * [[王醫生____j 讀博士]使整個醫院的其他醫生都很嫉妒]的那個國家 j * [[Wang yisheng ___j du boshi] shi zheng-ge Wang doctor study PhD make entire-Cl yiyuan de qita yisheng dou hen jidu] hospital DE other doctors all very jealous de na-ge guojiaj C that-Cl country (‘the countryj [in which [[that Doctor Wang had his PhD ___j] makes the rest of the doctors in the entire hospital very jealous]]’) * [[王醫生在那裡 j 讀博士]使整個醫院的其他醫生都很嫉妒]的那個國 家j * [[Wang yisheng zai nalij du boshi] shi Wang doctor at there study PhD make zheng-ge yiyuan de qita yisheng dou entire-Cl hospital DE other doctors all hen jidu] de na-ge guojiaj very jealous C that-Cl country (‘the countryj [in which [[that Doctor Wang had his PhD therej] makes the rest of the doctors in the entire hospital very jealous]]’)

From this point of view, resumptive pronouns in relatives in Chinese behave similarly to the weak resumptives in languages like Welsh. Recall from Chapter 1, Welsh data show that weak resumptives can be either an agreement marker or a clitic, both of which are excluded from island contexts. Sentences in (19a–c) involve complex-NP islands; with the presence of either an agreement marker or a clitic pronoun on the relativized site inside islands, the relevant sentences still remain ungrammatical. (19) a.

?* Dyma’r dyn1 y cusanaist ti’r ddynes2 here the man that kissed you the woman a siaradodd amdano1 Rel talked about-[agr] ‘Here is the man1 [that you kissed the woman2 [who talked about him1]].’

Distribution of resumptive pronouns in relatives and in dislocation structures  

b.

?* Dyma’r dyn1 y cusanaist ti’r ddynes2 here the man that kissed you the woman a brynodd ei1 dy. Rel bought his house ‘Here is the man1 [that you kissed the woman2 [who bought his1 house]].’

c.

?* Dyma’r dyn1 y cusanaist ti’r ddynes2 here the man that kissed you the woman y gwn y cyfarfu --hi. that I-know that met-[agr] --her ‘Here is the man1 [that you kissed the woman2 [that I know [that he1 met her2]]].’ Welsh, Tallerman (1983)

By contrast, strong resumptives in Welsh can function as a last resort to save a sentence from violation of locality constraints, as shown in (20). (20)

Dyma’r dyn1 y cusanaist ti’r ddynes2 here the man that kissed you the woman a siaradodd amdano ef1. Rel talked about-[agr] him ‘Here is the man1 [that you kissed the woman2 [that talked about him1]].’ Welsh, Tallerman (1983)

From this perspective, Chinese behaves differently from the strong resumptives in Welsh. As for other Celtic languages such as Breton (cf. 21) and Irish (cf. 22), resumptive pronouns play the role of a last resort (i.e. the intrusive use) in relatives to prevent sentences from violating locality constraints. A general remark is that intrusive pronouns are not morphologically distinguished from ordinary resumptive pronouns in these languages. Let me cite several examples from Celtic languages (cf. 21, 22). (21) a.

* An den1 a anevez an dud2 the man prt you-know the people deus ____2 gwelet _____1 have seen (* ‘the man1 that you know the people who saw t1’)

o prt

  Syntactic derivation: two types of A'-dependencies b.

An den1 a anevez an dud2 o the man prt you-know the people prt deus ____2 gwelet anezzhañ1 have seen him ‘the man1 that you know the people who saw him1’ Breton, Guilliot (2006)

In (21b), the object resumptive pronoun anezzhañ ‘him’ is inserted within the relative clause island “the people who saw him” and the sentence is grammatical. By contrast, if a gap is left in the same object position inside the same island, as shown in (21a), the sentence is ungrammatical. Similarly, in each sentence of (22), an object resumptive pronoun is inserted on the relativized site inside the relevant islands, which avoids island effects. Sentence in (22a) involves a wh-island and the presence of the resumptive pronoun iad ‘them’ occupying the relativized site can remedy potential island violation. (22b) involves a complex NP island and the presence of the pronoun i ‘it’ also prevents the relevant sentence from violating strong island condition. One must bear in mind that the presence of gaps on the relativized site will make these sentences ungrammatical. (22) a.

b.

ne dànta sin nach bhfuil fhios the poems these C-Neg is knowledge againn cén àit ar cumadh iad at-us what place C were-composed them ‘those poemsj [that we do not know [where theyj were composed]]’ seanchasóg ar dócha go bhfuil an old-jacket Cpro probable C is the táilliúir a dhein i sa chré fadó tailor C made it in-the earth long-ago ‘an old jacketj [that the tailor [who made itj] has probably been in the grave for ages]’ Irish, McCkoskey (1990, 2001, 2002)

In Semitic languages such as Hebrew and different Arabic dialects, intrusive pronouns can exist in relatives as well, which is also different from Chinese. For instance, in (23), the presence of the intrusive pronoun oto ‘him’ embedded inside an island constructed by the relative clause “the woman that loves him” saves the sentence from violation of locality constraints.

Distribution of resumptive pronouns in relatives and in dislocation structures  

(23)

raiti saw-I

et ha-yeled ACC the-boy

še that

dalya Dalya

makira knows

et ACC

ha-iša še ohevet *(oto) . the-woman that loves him ‘I saw the boyj [that Dalya knows the woman [that loves himj]].’ Hebrew, Borer (1984) ... Dislocation structures In this section, I will examine the distribution of gaps and of resumptive pronouns in dislocation structures. Recall that there are two types of dislocation structures that I am going to discuss: one is Left-Dislocation structures where the dislocated element is a topic and the other is ex-situ cleft-focus structures where the dislocated element is marked by the copula shi ‘be’ and is interpreted as a clefted focus. Generally, resumptive pronouns are permitted in dislocation structures without any island, as illustrated in (24). (24) a.

b.

(25)

LD-structures 小倩啊,我覺得(她)很大方。 Xiaoqian a, wo juede (ta2) hen Xiaoqian Top 1Sg think 3FSg very ‘As for Xiaoqianj, I think that shej is generous.’

dafang. generous

這個古董花瓶呢,(它)已經被轉手了好幾次了。 Zhe-ge gudong huapingj ne, (ta3j) yijing bei this-Cl antique vase Top 3OSg already Passive zhuanshou hao ji-ci le. resell good several-times SFP ‘As for this antique vasej, itj has already been resold several times.’ Ex-situ cleft-focus structure 是這幅油畫,我們不好評估它的價值。 Shi zhe-fu youhuaj, women buhao pinggu be this-Cl painting 1Pl difficultly evaluate ta3j-de jiazhi. 3OSg-Gen value ‘It is this paintingj that we had trouble to estimate itsj value.’

In contrast to resumptive relatives, the presence of a resumptive pronoun can actually prevent a sentence from violating locality constraints in dislocation

  Syntactic derivation: two types of A'-dependencies structures, as shown in (26–30). That is to say, the intrusive use of resumptive pronouns is only permitted in dislocation structures but not in relatives in Chinese. Morphologically, there is no distinction between a resumptive pronoun and an intrusive pronoun. Sentences in (26–30) involve strong islands, i.e. complex NPs. The presence of resumptive pronouns inside these islands (in their so-called intrusive use) remedy violation of locality constraints. (26)

那位作家呢 j,我見到了[談論過 *(他 j) 的]那個女同學 。 Na-wei zuojiaj ne, wo jiandao-le [tanlun-guo that-Cl writer Top 1Sg meet-Perf talk-Exp *(ta1j) de] na-ge nütongxue. 3MSg C that-Cl female.student ‘As for that writerj, I met the female student [who talked about (*him)j].’

(27)

那位醫生 j 啊,我見到了[買下了*(他的 j) 房子的]那個女同學。 Na-wei yishengj a, wo jiandao-le [mai-xia-le that-Cl doctor Top 1Sg meet-Perf buy-Perf *(ta1j-de) fangzi de] na-ge nütongxue. 3MSg- Gen house C that-Cl female.student ‘As for that doctorj, I met the female student [who bought hisj house].’

(28)

那位法國影星 j,我碰到了[小倩認識[擁抱過*(他 j)] 的]那位女生。 Na-wei Faguo yingxingj, wo pengdao-le that-Cl French movie.star 1Sg [Xiaoqian renshi [yongbao-guo Xiaoqian know embrace-Exp de] na-wei nüsheng. C that-Cl female.student

meet-Perf *(ta1j)] 3MSg

‘As for that French movie starj, I met the female student [that Xiaoqian knows [who embraced himj]].’ (29)

那本字典 j,我剛剛看到了[ *(把它 j) 扔掉了的]那個男孩兒。 Na-ben zidian j, wo ganggang kandao-le [*(ba that-Cl dictionary 1Sg just.now meet-Perf BA ta3j) rengdiao-le de] na-ge nanhair. 3OSg throw-Perf C that-Cl boy ‘That dictionaryj, I met the boy [who threw itj away] just now.’

Distribution of resumptive pronouns in relatives and in dislocation structures  

(30)

那個花瓶 j,我把[ *(把它 j) 打碎了的]那個男孩兒罵了一頓。 Na-ge huapingj, wo ba [*(ba ta3j) da-sui-le that-Cl vase 1Sg BA BA 3OSg de] na-ge nanhair ma-le yi-dun. C that-Cl boy scold-Perf one-ClV ‘That vasej, I scolded the boy [who broke itj].’

break-Perf

Now let us examine other types of islands in Chinese. The following sentences contain islands created by complement clauses of noun, adjunct clauses and sentential subjects. The result of these tests also confirms the observation above: when it is a gap that occupies the extracted site inside an island, the relevant sentences are always ungrammatical, whereas when it is a resumptive pronoun, the relevant sentences are fully grammatical. Examples (31–32) show that a resumptive LD-structure containing an island constructed by a complement clause of noun is always grammatical. (31) a.

*那個醫生 j , [[瑪麗親了___j 的]消息]傳遍了全醫院。 * Na-ge yishengj , [ NP [Mali qin-le ___j

de]

that-Cl doctor Mary kiss-Perf DE xiaoxi] chuan-bian-le quan yiyuan. news spread-Perf entire hospital (‘As for that doctorj, the rumor [that Mary kissed ___j] spread everywhere in the hospital.’) b.

那個醫生 j , [[瑪麗親了他 j 的]消息]傳遍了全醫院。 Na-ge yishengj , [ NP [Mali qin-le ta1j

de]

that-Cl doctor Mary kiss-Perf 3MSg DE xiaoxi] chuan-bian-le quan yiyuan. news spread-Perf entire hospital ‘As for that doctorj, the rumor [that Mary kissed himj] spread everywhere in the hospital.’ (32)

那個學生 j,[ *(他 j 的)獎學金被取消了]的消息傳遍了學校。 Na-ge xueshengj, [*(ta1j-de) jiangxuejin bei that-Cl student 3MSg-Gen scholarship Passive qu-xiao-le] de xiaoxi chuanbian-le xuexiao. cancel-Perf C news spread-Perf school ‘As for that studentj, the rumor [that hisj scholarship was canceled] was spread everywhere at school.’

  Syntactic derivation: two types of A'-dependencies Let us continue to examine the cases of resumptive LD-structures containing an adjunct clause (cf. 33) or a sentential subject (cf. 34) as island. (33) a.

Adjunct clause *那個醫生 j ,[因為瑪麗親了____j ] 整個學校的男老師都很鬱悶。 * Na-ge yishengj, [yinwei Mali qin-le ____j] that-Cl doctor because Mary kiss-Perf zheng-ge xuexiao de nanlaoshi dou hen yumen. entire-Cl school DE male.teachers all very unhappy (‘As for that doctorj, all of the male teachers of the school are very unhappy [because Mary kissed ____j].’)

b.

那個醫生 j ,[因為瑪麗親了他 j] 整個學校的男老師都很鬱悶。 Na-ge yishengj, [yinwei Mali qin-le ta1j] that-Cl doctor because Mary kiss-Perf 3MSg zheng-ge xuexiao de nanlaoshi dou hen yumen. entire-Cl school DE male.teachers all very unhappy (‘As for that doctorj, all of the male teachers of the school are very unhappy [because Mary kissed himj].’)

(34) a.

Sentential subject * (在)巴黎 j,[王醫生____ j 讀博士]使整個醫院的其他醫生都很嫉妒。 * (Zai) Balij, [Wang yisheng ____ j du boshi] shi in Paris Wang doctor study PhD make zheng-ge yiyuan de qita yisheng dou hen jidu. entire-Cl hospital DE other doctors all very jealous (‘In Parisj, [that Doctor Wang had his PhD ____j] makes all of the other doctors in the hospital very jealous.’)

b.

(在)巴黎 j,[王醫生在那裡 j 讀博士]使整個醫院的其他醫生都很嫉 妒。 (Zai) Balij, [Wang yisheng zai nalij du boshi] shi in Paris Wang doctor at there study PhD make zheng-ge yiyuan de qita yisheng dou hen jidu. entire-Cl hospital DE other doctors all very jealous ‘(In) Parisj, [that Doctor Wang had his PhD therej] makes all of the other doctors in the hospital very jealous.’

Distribution of resumptive pronouns in relatives and in dislocation structures  

More interesting cases concerning island effects in topicalization structures can be found in Zhang, M. (2009). The situation will be different for ex-situ cleftfocus structures. As I show in the Appendix, there is an independent constraint on the extraction of an ex-situ cleft-focus, which is called the episodicality constraint (cf. Pan 2014). This constraint prohibits any extraction from a predicate that encodes episodic eventualities, such as action verbs, to an ex-situ cleft focus position. Instead, if such an extraction is effected from a predicate encoding non-episodic eventualities, such as psych verbs, the related sentences will be grammatical. Let us compare (35a) with (35b). (35) a.

是[你的態度]j,公司的老板不欣賞 tj。 Shi [ni-de taidu]j, gongsi-de

laoban

be your attitude company-DE boss bu xinshang tj. Neg appreciate ‘It is your attitude that the boss doesn’t appreciate.’ b.

* 是你的鑰匙,我在花園裡找到了 tj。 * Shi [ni-de yaoshi], wo zai

huayuan-li

be your key 1Sg at garden-in zhaodao-le tj. find-Perf (‘It was your key that I found in the garden.’) In (35a), ni-de taidu ‘your attitude’ originates in the direct object position of the psycho verb xinshang ‘appreciate’ which encodes a non-episodic eventuality, and such an extraction is licit in an ex-situ cleft-focus structure. By contrast, in (35b), when the resultative action verb zhaodao ‘find’ is used alone and marked by the perfective aspectual marker le, the predicate encodes an episodic eventuality (cf. Zhang 2002 for the semantics of episodicality). The extraction of ni-de yaoshi ‘your key’ from such a context to form an ex-situ cleft-focus structure is illicit. However, the episodicality constraint does not apply to topicalization in island-free contexts, as illustrated in (36). (36)

Topicalization 辭典 j(呢),她昨天买了 tj。 Cidianj (ne), ta2 zuotian

mai-le

dictionary Top 3FSg yesterday buy-Perf ‘As for the dictionaryj, she bought (itj) yesterday.’

tj.

  Syntactic derivation: two types of A'-dependencies The presence of a resumptive pronoun is permitted in ex-situ cleft-focus structures. Let us examine an island-free case (cf. 37). (37)

是[新來的秘書] j,大家都很喜歡她 tj。 Shi [xin lai de mishu]j,

dajia

be new come DE secretary everyone hen xihuan ta2j. very like 3FSg ‘It is the new secretary that everyone likes very much.’

dou all

The situation becomes more complicated when an island intervenes. At first sight, the presence of a resumptive pronoun is allowed in an ex-situ cleft-focus structure with an island, as illustrated in (38). (38)

是[這幅油畫]j,[懂得欣賞它 j 的]人不多。 Shi [zhe-fu youhua]j, [dongde

xinshang

ta3j

be this-Cl painting understand appreciate 3OSg de] ren bu duo. C people Neg many ‘It is this paintingj that the people [who know how to appreciate itj] are not many.’ However, surprisingly if I replace the resumptive pronoun ta3 ‘it’ inside the island with a gap, the sentence till remains grammatical, as shown in (39). (39)

是[這幅油畫]j,[懂得欣賞____j 的]人不多。 Shi [zhe-fu youhua]j, [dongde

xinshang

____j

be this-Cl painting understand appreciate de] ren bu duo. C people Neg many ‘It is this paintingj that the people [who know how to appreciate ___j] are not many.’ Actually, Zhang (2002) and Pan (2014) discover that only the extraction from a predicate encoding episodic eventualities embedded within an island displays island effects and the extraction from a predicate that encodes non-episodic eventualities embedded within an island does not display any island effect. This is even true with general topicalization cases, as shown in (40).

Distribution of resumptive pronouns in relatives and in dislocation structures  

(40)

Topicalization [這幅油畫]j,[懂得欣賞____j 的]人不多。 [Zhe-fu

youhua]j,

[dongde

xinshang

____j

this-Cl painting understand appreciate de] ren bu duo. C people Neg many ‘As for this paintingj, the people [who know how to appreciate ___j] are not many.’ The sentence in (40) should be a typical example to show an island violation; however, the sentence is fully grammatical. This is so because the topic NP zhefu youhua ‘this painting’ is extracted from a predicate that encodes non-episodic eventuality (i.e the psycho verb xinshang ‘appreciate’). Even though such a predicate is embedded within a relative clause, thus an island, the sentence remains fully grammatical. This contrast creates an enormous difficulty for my tests on ex-situ cleftfocus structures. On the one hand, if we want to test island effects, we must use action verbs or equivalents that create episodic eventualities; however, on the other hand, the extraction of an ex-situ cleft-focus from a predicate constructed by an action verb is already prohibited, irrespective of whether such a predicate is generated outside or inside an island. This contradictory situation makes it impossible to test island effects with ex-situ cleft-focus structures and this contradiction has been illustrated by the contrast between (38) and (39). In other words, the fact that the sentence in (38) is grammatical is not due to the presence of the resumptive pronoun ta3 ‘it’ inside the island. The sentence is grammatical because it does not violate island conditions at all in the first place, due to non-episodic predicate created by the psycho verb xinshang ‘appreciate’. Therefore, the pronoun ta3 ‘it’ is not at all an intrusive pronoun, nor does it play the role of the last resort in the present case. ... Summary In this section, I have shown the general distribution of resumptive pronouns in three types of A'-dependencies: relatives, LD-structures and ex-situ cleft-focus structures. It is very important to point out that a crucial contrast exists between a relative clause and an LD-structure. In the contexts with an island, so-called intrusive pronouns can only be inserted in dislocation structures but not in relatives, in order to prevent a sentence from violating locality constraints. Putting this differently, in relatives with an island, island effects are always observed, irrespective of whether it is the gap strategy or the resumptive strategy

  Syntactic derivation: two types of A'-dependencies that is adopted. By contrast, in dislocation structures with an island, island effects are only detected if the gap strategy is adopted. That is to say, resumptive dislocation structures do not give rise to any island effects at all. Such a contrast is very important because it will help us to identify the derivational difference between relatives and dislocation structures. As emphasized previously, island effects constitute one of the classical diagnostic tests for A'movement. Generally, if island effects are detected in an A'-chain, one can assume that there is a movement involved in the derivation of such a dependency. By contrast, if island effects are absent in an A'-chain, then one can presumably conclude that movement is not involved in the derivation of such a dependency. Let me summarize the result of the tests obtained in this section in the following table. In “without island” cases, “yes” means that the presence of the relevant items, gaps or pronouns, is permitted. In “island effects” cases, “yes” means that the relevant items give rise to island effects; “no” means that the relevant items do not give rise to any island effect; “--” means that the relevant tests are not applicable due to independent reasons. Tab. 2: Island effects

Gap

Resumptive

Ex-situ clefts

Resumptive

Resumptive

Gap Without island: the presence is possible?

Dislocations Gap

Relatives

yes

yes

yes

yes

yes

yes

– Relatives

yes

yes

yes

no

--

--

– Complement of noun

yes

yes

yes

no

--

--

– Adjunct clause

yes

yes

yes

no

--

--

– Sentential subject

yes

yes

yes

no

--

--

Island effects?

In order to check whether my reasoning is on the right track I will apply another classical diagnostic test for movement: crossover effects.

Distribution of resumptive pronouns in relatives and in dislocation structures  

.. Crossover effects As I pointed out previously, crossover effects are also considered as a test for all kinds of A'-movements. A description of the crossover effects is that movement of a wh-constituent cannot cross a pronoun that bears the same index. For instance, (41) a. SS: * Whoj does hej love tj? (= Who loves himself?) b. LF: * Whoj [hej loves tj ] ? (42)

* CP

TP

Whoj

hej

…… VP

loves

tj

(41a) is a typical example of a (strong) crossover effect; (41b) and (42) illustrate the structure of (41a) at LF. Descriptively, the movement of who crosses the pronoun he and both of them share the same index j. The configuration in (42) violates two different conditions. First, condition C of the Binding Theory is violated. Condition C requires that proper names and variables be A-free in the domain of their operators and A-free means that they cannot be c-commanded by an antecedent in an A-position. In (42) the trace tj resulting from the whmovement is directly c-commanded by the pronoun hej that bears the same index, which leads to the violation of condition C. Another way to describe such a violation is to say that the trace tj is A'-bound by the wh-phrase whoj located in an A'-position (i.e. [Spec, CP]) and at the same time it is A-bound by the pronoun hej located in an A-position. However, a trace cannot be A-bound and A'bound simultaneously (cf. Chomsky 2008, Abe 2015) and that is why the sen-

  Syntactic derivation: two types of A'-dependencies tence is ungrammatical. Thus, the real principle that filters such a configuration is condition C. Second, this configuration also violates the constraint on bound variable construal à la Reinhart (1983). This constraint states that a pronoun receives a bound variable reading only if it is bound by the trace of a quantifier. Apparently, in (42), the pronoun hej is not bound by the trace tj left by the wh-quantifier and that is why the bound variable reading is not available for the pronoun he. Topicalization is also an instance of A'-movement and it also gives rise to a strong crossover effect. For instance, in (43) the trace ti resulting from the movement (i.e. topicalization) of Paulj to the [Spec, TopP] position, which is also an A'-position, is directly c-commanded by the pronoun hej that shares the same index j. This derivation also violates the condition C as well as the constraint on the bound variable construal; thus the sentence is severely ungrammatical. (43) * [TopP Pauli [TP hei loves ti ]]. Another type of crossover effect, called the “Weak Crossover Effect”, is illustrated in (44). (44) a. ?* [CP Whoi [TP does [hisi mother] love ti ]] ? b. ?* CP

Whoi

TP

NP

hisi

...... mother love

VP

tj

The weak crossover effect describes the case where the crossed pronoun does not c-command the trace. If the trace is not c-commanded by the pronoun, then condition C will not be violated. In (44), the wh-movement of whoj, still crosses the pronoun hisj that bears the same index j; however, the first branching node that dominates the pronoun his is the NP and this NP does not dominate the

Distribution of resumptive pronouns in relatives and in dislocation structures  

trace tj, and therefore the pronoun his does not c-command the trace tj. From this point of view, condition C is not violated in a weak crossover configuration. Nevertheless, the constraint on bound variable construal is still violated in such a configuration because the pronoun his still cannot get a bound variable reading in that it is not bound by the trace tj resulting from the movement of the whquantifier. Compared with the strong crossover configuration, this type of crossover effect (cf. 44b) is called “weak” for two reasons. First, there is only one constraint that is violated and second, the degree of the ungrammaticality of the sentence is less severe with respect to the strong crossover effect. Generally, sentences such as (44a) sound odd if not totally ungrammatical. The following example illustrates the weak crossover effects observed in the case of topicalization. (45) ?*[TopP Johni, [TP [hisi mother] loves ti ]]. Crossover effects (strong or weak) are considered as diagnostics for A'movement and therefore, scholars working on resumptivity apply the tests based on crossover effects to resumptive dependencies. The idea is that if a resumptive dependency gives rise to crossover effects, then one can conclude that such a dependency is derived by movement; by contrast, if a resumptive dependency does not show any crossover effect, presumably movement is not involved. Historically, it has been observed that weak crossover effects are not detected in relatives with resumptive pronouns (Borer 1984, Sells 1984, McCloskey 1990). For instance, (46)

an fearj so ar mhairbh the man this C killed ‘the manj [that hisj own wife killed]’

aj bhean féin his-own-wife

éj [him]

Irish, McCloskey (1990)

In (46) the dependency established between the operator C (i.e. the complementizer ar) and the lower pronoun éj ‘him’, which is considered as a resumptive pronoun, crosses the pronoun aj ‘his’ that bears the same index j as the resumptive pronoun éj ‘him’. Despite the crossing effect, the fact that the sentence is still fully grammatical suggests that the weak crossover effect is not detected in (46). This test leads many scholars to assume that resumptive pronouns do not give rise to the weak crossover effect. However, McCloskey claims that it is not sure that (46) is a reliable test for movement derivation in resumption because it is always possible to establish a relation between the operator under C° and the higher pronoun (i.e. aj ‘his’)

  Syntactic derivation: two types of A'-dependencies without crossing the other one (i.e. éj ‘him’). The idea behind McCloskey’s reasoning is based on the two possibilities to establish a dependency between the antecedent and the two pronouns, as illustrated in (47). (47) a. Crossover NPj

Pronoun 1j

Pronoun 2j

b. No crossover NPj

Pronoun 1j

Pronoun 2j

(47a) and (47b) illustrate the two possible strategies to establish a dependency. In (47a), Pronoun 2 is treated as a resumptive pronoun that depends directly on the NP antecedent. By contrast, Pronoun 1 is an ordinary pronoun that is anaphorically related to the NP antecedent. In this case, it is Pronoun 1 that is crossed by the resumptive dependency established between the NP antecedent and Pronoun 2. On the contrary, in (47b) it is Pronoun 1 that is the resumptive pronoun that depends on the NP antecedent. Pronoun 2 is an ordinary pronoun that is anaphorically related to Pronoun 1. The resumptive dependency established between the NP antecedent and Pronoun 1 does not cross Pronoun 2 that bears the same index j, which explains the absence of the weak crossover effect. To conclude, if there are two different ways to establish an anaphoric dependency and only one of them triggers crossover effects whereas the other does not, crossover effects can be obviated. Based on this idea, McCloskey thinks that despite the grammaticality of the sentence in (46), it is still possible that resumptive dependencies give rise to crossover effects. His work (1990) shows that in Irish, strong crossover effect can be detected in resumptive constructions if the crossed element is an epithet instead of a pronoun, as illustrated in (48). (48)

*Sin an fearj ar dhuirt an bastardj go that the man C said the bastard C marodh séj muid. would-kill he us ‘That is the manj that the bastardj said hej would kill us.’ Irish, McCloskey (1990)

Distribution of resumptive pronouns in relatives and in dislocation structures  

In (48) McCloskey uses an epithet to replace the pronoun in an intermediate position, which excludes one of the two possibilities mentioned above to establish an anaphoric dependency. The chain is constructed by [the manj … the bastardj…hej]. The impossible way to establish such a dependency requires the epithet the bastard to be a resumptive, however it cannot be the case because an epithet cannot be interpreted as a bound variable. This leaves only one choice that treats the pronoun he as a resumptive pronoun. Then the resumptive dependency established between the antecedent NP the manj and the resumptive pronoun hej crosses the epithet the bastardj that shares exactly the same index j. The ungrammaticality of the sentence in (48) can therefore suggest that movement must be involved in the derivation. Crossover effects (weak or strong) are considered as an indication of movement. Sentences such as (48) are treated as arguments in support of the analyses according to which a resumptive chain is derived by movement. The movement based approach has been adopted by many previous analyses such as Rouveret (1987) and Demirdache (1991). A recent work of Demirdache & Percus (2011) also argue for a movement driven analysis of resumption based on Jordanian Arabic data. The observation is that configurations such as *Q/Wh […epithet…clitic…] are always illicit due to weak crossover effects (cf. 49). (49) a.

b.

* Q/Wh […epithet…clitic…] Weak crossover effect * [kull walad [[ʔum ħa-l-ħmar] fakkart every boy mother pro-the-donkey thought ʔinnu raħ yzittu-u bi-lħabs]] that they.will put.him in-prison (‘Every boy that this donkey’sj mother thought that they will put himj in prison’) Strong crossover effect * miin xabbartu ħa-l-ħmar who you.told pro-the-donkey ʔinnu raħ yzittu-u bi-lħabs ? that they.will put.him in-prison (‘Who did you tell this donkeyj that they will put himj in prison?’) Jordanian Arabic, Demirdache & Percus (2011:378)

In the above examples, the resumptive dependency established between the universal quantifier (cf. 49a) or the wh-phrase (cf. 49b) on the one hand and the

  Syntactic derivation: two types of A'-dependencies resumptive pronoun u ‘him’ on the other hand crosses the epithet DP the donkey that bears the same index. The ungrammaticality of these sentences illustrates that crossover effects are detected, which leads the authors to conclude that movement is involved in the derivation of these dependencies. ... Weak crossover effects .... Relatives Let us come back to Chinese. I will begin by examining relatives with gaps. Example (50) shows that relatives with gaps give rise to weak crossover effect. (50) a.

* [他 j 媽媽不喜歡_____j]的那個小孩 j * [ta1j mama bu xihuan ___ j] de his mother Neg like C (‘the kidj that hisj mother doesn’t like tj ’)

b.

na-ge

xiaohaij

that-Cl

kid

* [他 j 自己的老婆殺死了_____j]的那個律師 j * [ta1j-ziji de laopo sha-si-le ___ j] de himself DE wife kill-dead-Perf (‘the lawyerj that hisj own wife killed tj’)

C

na-ge

lüshij

that-Cl lawyer

Now let us turn to resumptive cases. At first glance, weak crossover effect is not detected in resumptive relatives if it is a pronoun that is crossed, as illustrated in (51). (51) a.

[他 j 自己的老婆把他 j 給殺死了]的那個人 j [ta1j-ziji de laopo ba ta1j gei himself DE wife BA 3MSg GEI de na-ge renj C that-Cl person ‘the manj that hisj own wife killed himj’ = ‘the manj who was killed by hisj own wife’

b.

[他 j 的媽媽不准他 j 吃飯]的那個小男孩 j [ta1j-de mama bu zhun

ta1j

sha-si-le] kill-dead-Perf

chi

3MSg-Gen mother Neg allow 3MSg eat de na-ge xiao nanhaij C that-Cl little boy ‘the little boyj that hisj mother does not allow himj to eat’

fan] food

Distribution of resumptive pronouns in relatives and in dislocation structures  

The explanation for the similar Irish cases provided by McCloskey also applies to Chinese. Each of the sentences in (51) contains two pronouns and according to McCloskey, there are two ways to establish the anaphoric dependencies. It is always possible to establish a dependency between the antecedent NP and one of the two pronouns without really crossing the other. When two strategies are both available, one triggering the weak crossover effect but not the other, the weak crossover effect can be obviated. Let us take (51a) for example, the resumptive chain is established between [the manj…hisj…himj]. In the first scenario, let us assume that him is the resumptive pronoun that depends on the antecedent DP the man and that his is an ordinary pronoun that is anaphorically related to the antecedent the man. The dependency established between the man and him will certainly cross the pronoun his that bears the same index j, and in this case the weak crossover effect is detected. This scenario is illustrated in (52). (52) Crossover the manj

hisj

himj

In the second scenario, let us assume that his is the resumptive pronoun that depends on the antecedent DP the man and that him is an ordinary pronoun that is anaphorically related to the resumptive pronoun his. In this case, two dependencies, [the manj…hisj] and [hisj…himj], do not overlap at all, and such a configuration does not display the weak crossover effect, as illustrated in (53). (53) No crossover the manj

hisj

himj

Since the two possibilities of establishing the anaphoric dependency are available simultaneously, one giving rise to the weak crossover effect whereas the other not, weak crossover effect can be obviated and that is why both (51a) and (51b) are grammatical sentences. Following the analysis of McCloskey, in the resumptive chain [the manj…hisj…himj] I will replace the pronoun his by an epithet in order to force the pronoun him to be the resumptive, which will help us to check whether the resumptive chain [the manj…epithetj…himj] displays weak crossover effects in Chinese.

  Syntactic derivation: two types of A'-dependencies Before applying this test based on epithets, it is very important to show that in Chinese, epithets can actually function as resumptives. Malkawi (2009)’s dissertation shows that in Jordanian Arabic, complex expressions like ha-lHmar ‘pro-the-donkey’ is treated as a resumptive epithet. In this expression, hais a pronominal morpheme or a demonstrative that is the equivalent to the definite article; l-Hmar is the DP the donkey. Similar composition exists for an epithet in Chinese. The DP in (54) has the same composition as in Jordanian Arabic: pro-D-NP. The only difference is that in Chinese there is always an agreement in terms of ϕ-features between pro and the NP. In addition, the pro part in Chinese is a morphologically full pronoun that can be used independently and from this view, it is not a morpheme like the one in Jordanian Arabic. In (54a), ta1 ‘he’ bears the feature [Masculine] and it agrees with the NP xiaozi ‘the guy’; in (54b), ta2 bears the feature [Feminine] and it agrees with guniang ‘the girl’. (54) a.

b.

他那小子 ta1-na-xiaozi 3MSg-that-guy ‘(hej) that guyj/ that chapj/ that boyj’ 她那姑娘 ta2-na-guniang 3FSg-that-girl ‘(shej) that girlj’

Normally, the pro part or the D part in these complex expressions can be omitted, as illustrated in (55). (55) a.

他(那)小子最會說謊。 Ta1-(na)-xiaozi zui

hui

shuo

3MSg-(that)-guy most can tell ‘(Himj) that guyj is good at telling lies.’ b.

(他)那小子最會說謊。 (Ta1)-na-xiaozi zui

hui

shuo

(3MSg)-that-guy most can tell ‘(Himj) that guyj is good at telling lies.’

huang. lie

huang. lie

Distribution of resumptive pronouns in relatives and in dislocation structures  

c.

哥哥姐姐們把(他)那倒霉鬼仍在了半路上。 Gegejiejie-men ba (ta1)-na-daomeigui

reng-zai-le

brother-sister-Pl BA 3MSg-that-poor-boy throw.at-Perf banlu-shang. half.way.on ‘Brothers and sisters threw that poor boy halfway.’ Now let us see how the presence of an epithet can help us with weak crossover in resumptive relatives. In (56a), I replace the pronoun his in [the manj…hisj…himj] by the epithet na-ge hundan ‘that bastard’ and the chain becomes [the manj…that bastardj…himj]. In (56b) I use the epithet na xiaogui ‘the little demon’ to do the same job. (56) a.

* [ 那個混蛋 j 自己的老婆把他 j 給殺死了]的那个個人 j * [na-ge hundanj-ziji de laopo ba ta1j that-Cl bastard-self DE wife BA 3MSg sha-si-le ] de na-ge renj kill-dead-Perf C that-Cl person (‘that manj that the bastardj’s own wife killed himj’)

b.

* [那小鬼 j 的媽媽不準他 j 吃飯]的那個小男孩 j * [na xiaoguij de mama bu

zhun

gei GEI

ta1j

that little.demon DE mother Neg allow 3MSg chi fan] de na-ge xiao nanhaij eat food C that-Cl little boy (‘the boyj that that little demonj’s own mother does not allow himj to eat’) These two examples show that when the epithet intervenes between the pronoun and the DP, the pronoun must function as resumptive pronoun. For instance, in (56a) anaphoric dependency can only be established between the manj and himj, which certainly crosses the epithet the bastardj that bears exactly the same index j. Such a configuration gives rise to crossover effects and as a result, the relevant sentence is ungrammatical. The result of these tests confirms that McCloskey’s analysis is correct. In Chinese, resumptive relatives display weak crossover effects only if the crossed element is an epithet. Through the English translation of these Chinese examples, repeated below, we can find that English behaves like Chinese concerning the weak crossover effects in relatives. In English, relatives with a gap display the weak crossover

  Syntactic derivation: two types of A'-dependencies effect automatically as shown in (57); however, resumptive relatives do not give rise to the weak crossover effect if the crossed element is a pronoun, as illustrated in (58). Resumptive relatives do display the weak crossover effect if the crossed element is an epithet, as shown in (59). (57) a. * the kidj that hisj mother doesn’t like tj b. * the lawyerj that hisj own wife killed tj (58) a. ? the manj that hisj own wife killed himj b. ? the little boyj that hisj mother does not allow himj to eat (59) a. * that manj that the bastardj’s own wife killed himj b. * the boyj that that little demonj’s own mother does not allow himj to eat In fact, French behaves in a similar fashion, as illustrated in the following examples, (60) a.

b.

(61) a.

b.

* l’hommej [que saj propre femme the.man that his own wife (*‘the manj that hisj own wife killed tj’)

a has

tué killed

tj ]

* le the tj

petit garçonj [que saj mère n’autorise pas little boy that his mother Neg.allow à manger] to eat (*‘the little boyj that hisj own mother doesn’t allow tj to eat’)

? l’hommej [que saj propre femme the.man that his own wife (? ‘the manj that hisj own wife killed himj’)

lj’a him-has

tué] killed

? le petit garçonj [que saj mère the little boy that his mother ne lj’autorise pas à manger] Neg.him.allow to eat (? ‘the little boyj that hisj own mother doesn’t allow himj to eat’)

Distribution of resumptive pronouns in relatives and in dislocation structures  

(62) a.

b.

* l’hommej [que la propre femme de the.man that the own wife of ce salaudj lj’a tué] this bastard him.has killed (*‘the manj that this bastardj’s own wife killed himj’) * le garçonj [que la mère de ce the boy that the mother of this petit diablej ne lj’autorise pas à manger] little demon Neg.him.allow to eat (* ‘the little boyj that this little demonj’s own mother doesn’t allow himj to eat’)

In French, when a relative clause contains a gap, it systematically gives rise to the weak crossover effect (cf. 60). When a relative clause contains a resumptive pronoun, even though the weak crossover effect is not observed when the crossed element is a pronoun (cf. 61), such an effect is always observed when the crossed element in an epithet (cf. 62). From this perspective, English, French and Chinese behave alike. .... Dislocation structures In this section, I will examine the weak crossover effect in dislocation structures. Let me begin by examining dislocations with gaps. (63) shows that LDstructures with gaps give rise to weak crossover effect. (63) a.

* 那個小孩 j, 他 j 媽媽不喜歡 tj. * Na-ge xiaohaij, ta1j mama

bu

xihuan

tj .

that-Cl kid his mother Neg like (‘As for that kidj, hisj mother doesn’t like tj. ’) b.

* 那個律師 j , 他 j 自己的老婆殺死了 tj. * Na-ge lüshij, ta1j-ziji de laopo that-Cl lawyer himself DE wife (‘As for that lawyerj, hisj own wife killed tj.’)

sha-si-le tj. kill-Perf

Ex-situ cleft-focus structures act in a similar way to topicalization cases. (64) shows that an ex-situ cleft focus structure with a gap gives rise to weak crossover effect.

  Syntactic derivation: two types of A'-dependencies (64)

*是那個小孩 j, 他 j 媽媽不喜歡 tj. * Shi na-ge xiaohaij, ta1j

mama

bu

be that-Cl kid his mother Neg (‘It is that kidj that hisj mother doesn’t like tj. ’)

xihuan tj . like

Let us turn to dislocation structures with resumptive pronouns. Exactly like relatives with a gap, if the crossed element is a pronoun, a resumptive LDstructure does not display the weak crossover effect, as illustrated in (65). (65) a.

b.

那個人 j,他 j 自己的老婆把他 j 給殺死了。 Na-ge renj, ta1j -ziji de laopo that-Cl man himself DE wife sha-si-le. kill-Perf ‘As for that manj, hisj own wife killed himj.’

ba BA

ta1j 3MSg

gei GEI

那個小男孩 j, 他 j 的媽媽不准他 j 吃飯。 Na-ge xiao nanhaij, ta1j de mama bu that-Cl little boy 3MSg DE maman Neg zhun ta1j chi fan. allow 3MSg eat food ‘As for that little boyj, hisj own mother does not allow himj to eat.’

Just like LD-structures, resumptive ex-situ cleft-focus structures do not give rise to the weak crossover effect when the crossed element is a pronoun, as shown in (66). As I explained, weak crossover effects can be obviated when there are two alternative ways to establish the relevant dependencies. (66)

是那個小孩 j, 他 j 的媽媽不喜歡他 j . Shi na-ge xiaohaij, ta1j

de

mama

be that-Cl kid his DE mother bu xihuan ta1j . Neg like 3MSg ‘It is that kidj that hisj mother doesn’t like himj. ’ However, what distinguishes dislocation structures from relatives is the fact that the weak crossover effects are never produced in dislocation structures even if the crossed element is an epithet, as illustrated in (67).

Distribution of resumptive pronouns in relatives and in dislocation structures  

(67) a.

那個人 j 啊, 那混蛋 j 自己的老婆把他 j 殺死了。 Na-ge renj a, nei hundanj-ziji de

laopo

that-Cl man Top that bastard-self DE wife ba ta1j gei sha-si-le. BA 3MSg GEI kill-dead-Perf ‘As for that manj, that bastardj’s own wife killed himj.’ b. 小寶 j 啊, 那小鬼 j 的媽媽不準他 j 吃飯。 Xiaobaoj a, na xiaoguij de

mama

Xiaobao Top that little.demon DE mother bu zhun ta1j chi fan. Neg allow 3MSg eat food ‘As for Xiaobaoj, that little demonj’s own mother does not allow himj to eat.’ From this perspective, ex-situ cleft-focus structures behave in exactly the same way as LD-structures in that neither type of them gives rise to the weak crossover effect even if the crossed element is an epithet (cf. 68). (68)

Ex-situ cleft-focus structure 是小寶 j, 那小鬼 j 的媽媽不喜歡他 j . Shi Xiaobaoj, na xiaoguij de mama be Xiaobao that little.demon DE mother bu xihuan ta1j. Neg like 3MSg ‘It is Xiaobaoj that the little demonj’s own mother doesn’t like himj. ’

So far, we can observe an important distinction between relatives and dislocation structures. Resumptive relatives display the weak crossover effect but resumptive dislocation structures do not. In order to confirm such a contrast, I will test the strong crossover effect in the next section. Let us come back to the comparison between Chinese, English and French. From the English translation for the Chinese examples, repeated below, we can see again that English behaves like Chinese concerning the weak crossover effect in dislocation structures. Dislocation structures with a gap always display the weak crossover effect, as illustrated in (69); whereas resumptive dislocation structures never give rise to the weak crossover effect irrespective of whether the crossed element is a pronoun as in (70) or is an epithet as in (71).

  Syntactic derivation: two types of A'-dependencies (69) a. * As for that kidj, hisj mother doesn’t like tj. b. * As for that lawyerj, hisj own wife killed tj. (70) a.

As for that manj, hisj own wife killed himj.

b. As for that little boyj, hisj own mother does not allow himj to eat. (71) a.

As for that manj, that bastardj’s own wife killed himj.

b. As for Xiaobaoj, that little demonj’s own mother does not allow himj to eat. Detailed discussion on the weak crossover effect in topicalization in English can be found in Lasnik & Stowell (1991). Let us compare some French data with Chinese. Dislocation structures with gaps always display the weak crossover effect as shown in (72); whereas resumptive dislocation structures never give rise to weak crossover effects irrespective of whether it is a pronoun (cf. 73) or it is an epithet that is crossed (cf. 74). (72) a.

b.

(73) a.

b.

* Cet enfantj, saj mère n’aime pas this kid his mother Neg.like (*‘As for this kidj, hisj mother doesn’t like tj.’)

tj.

* Cet avocatj, saj propre femme a this lawyer his own wife has (*‘As for this lawyerj, hisj own wife killed tj.’)

tué killed

Cet hommej, saj propre femme lj’a this man his own wife him-has ‘As for that manj, hisj own wife killed himj.’

tj.

tué. killed

Ce petit garçonj, saj mère ne lj’autorise pas this little boy his mother Neg.him-allow à manger. to eat ‘As for that little boyj, hisj own mother does not allow himj to eat.’

Distribution of resumptive pronouns in relatives and in dislocation structures  

(74) a.

b.

Cet hommej-là, la propre femme de this man-that the own wife of ce salaudj lj’a tué. this bastard him-has killed ‘As for that manj, that bastardj’s own wife killed himj.’ Quant à Xiaobaoj, la mère de ce petit as.for Xiaobao the mother of this little diablej ne lj’autorise pas à manger. demon Neg.him.allow to eat ‘As for Xiaobaoj, that little demonj’s own mother does not allow himj to eat.’

This set of tests shows that French behaves like Chinese and English with regard to dislocation structures. ... Strong crossover effect .... Relatives Let us now examine the strong crossover effect. I start with relatives with gaps. Example (75) shows that this type of relative clause gives rise to the strong crossover effect. (75)

* [那混蛋 j 揚言[我們一定要絕對服從____j]的]那個人 j * [na hundanj yangyan [women yiding

yao

that bastard claim 1Pl certainly must juedui fucong ____j] de] na-ge renj absolutely obey C that-Cl person (*‘the manj that the bastardj says that we must obey tj absolutely’) As expected, the strong crossover effect is not detected in resumptive relatives when the element that is crossed is a pronoun, as shown in (76). (76)

這就是[他 j 說[他 j 要把我們全都殺了]的]那個人 j。 Zhe jiu shi [ta1j shuo [ta1j yao

ba

this exactly be 3MSg say 3MSg will BA women quan dou sha-le] de] na-ge renj. 1Pl entirely all kill-Perf C that-Cl person ‘This is the personj who hej said that hej would kill all of us.’

  Syntactic derivation: two types of A'-dependencies Exactly as for the tests for the weak crossover effect, the strong crossover effect is detected in resumptive relatives when the crossed element is an epithet, such as na-ge hundan ‘that bastard’ in (77). (77)

* 這就是[那個混蛋 j 說[他 j 要把我们全都殺了]的]那個人 j。 * Zhe jiu shi [na-ge hundanj shuo this exactly be that-Cl bastard say [ta1j yao ba women quan dou 3MSg will BA 1Pl entirely all sha-le] de] na-ge renj. kill-Perf C that-Cl person (* ‘This is exactly the manj who the bastardj said that hej would kill all of us.’)

From the English translation of the Chinese examples repeated below we can see that English behaves like Chinese with respect to the strong crossover effect. The strong crossover effect is detected in relatives with a gap (cf. 78a) and in resumptive relatives when the crossed element is an epithet (cf. 78c). Again, such an effect disappears when the crossed element is a pronoun (cf. 78b). (78) a. * the manj that the bastardj says that we must obey tj absolutely b. ? This is the personj who hej said that hej would kill all of us. c. * This is exactly the manj who the bastardj said that hej would kill all of us. French shows exactly the same pattern. The strong crossover effect is detected in relatives with a gap (cf. 79a) or with a resumptive pronoun (i.e. il ‘he’ in 79b). In resumptive relatives, I replace the resumptive pronoun with an epithet le salaud ‘the bastard’ in (79c), the sentence is also ungrammatical due to the strong crossover effect. (79) a.

* l’hommei à qui ce salaudi dit que the.man to whom this bastard says that nous devons obéir ____i absolument we must obey absolutely (* ‘the mani that the bastardi says that we must obey ti absolutely’)

Distribution of resumptive pronouns in relatives and in dislocation structures  

b.

? C’est bien le typej qui ilj dit this.is exactly the guy who he said qu’ilj allait nous tous tuer. C-he would us all kill ‘? This is exactly the guyj who hej said that hej would kill all of us.’

c.

* C’est bien le typej qui le salaudj dit this.is exactly the guy who the bastard said qu’ilj allait nous tous tuer. C-he would us all kill (* ‘This is exactly the guyj who the bastardj said that hej would kill us all.’)

To conclude, English, French and Chinese behave alike with respect to the strong crossover effect in relatives. .... Dislocation structures Example (80) shows that LD-structures with a gap give rise to the strong crossover effect. (80)

* 張三 j 啊, 那混蛋 j 揚言[我們一定要絕對服從_____ j]. * Zhangsanj a, na-ge hundanj yangyan [women Zhangsan Top that-Cl bastard claim 1Pl yiding yao juedui fucong tj]. certainly must absolutely obey (‘As for Zhangsanj, that bastardj says that we must obey _____j absolutely.’)

By contrast, in the case of resumptive LD-structures, the strong crossover effect is never produced, irrespective of whether the crossed element is a pronoun (cf. 81a) or an epithet (cf. 81b, 82). (81) a.

那個人 j 啊,[他 j 說[他 j 要把我們全都殺了]]。 Na-ge renj a, [ta1j shuo [ta1j

yao

that-Cl person Top 3MSg say 3MSg will ba women quan dou sha-le ]]. BA 1Pl entirely all kill-Perf ‘As for that manj, hej said that hej would kill all of us.’

  Syntactic derivation: two types of A'-dependencies b.

那個人 j 啊, [那混蛋 j 說[他 j 要把我們全都殺了]]。 Na-ge renj a, [nei hundanj shuo

[ta1j

that-Cl person Top that bastard say 3MSg yao ba women quan dou sha-le ]]. will BA 1Pl entirely all kill-Perf ‘As for that manj, the bastardj said that hej would kill all of us.’ (82)

我兒子 j 啊,[那小子 j 說[他 j 再也不敢酒後駕駛了]]。 Wo erzij a, [nei xiaozij shuo 1Sg-(Gen) son Top that kid say zai ye bu gan jiu-hou jiashi again also Neg dare alcohol-after drive ‘My sonj, the kidj said that hej would never dare again drinking.’

[ta1j 3MSg le]]. SFP to drive after

Ex-situ cleft-focus structures behave in a similar way in that they do not display the strong crossover effect even if the crossed element is an epithet (cf. 83). (83)

Ex-situ cleft-focus structure 是張三 j, [那混蛋 j 說[他 j 要把我們全都殺了]]。 Shi Zhangsanj [nei hundanj shuo [ta1j yao be Zhangsan that bastard say 3MSg will ba women quan dou sha-le]]. BA 1Pl entirely all kill-Perf ‘It is Zhangsanj that the bastardj said that hej would kill all of us.’

Again, English behaves like Chinese in that the strong crossover effect is only detected in dislocation structures with a gap (cf. 84) but not in those with resumptive pronouns, irrespective of whether it is a pronoun (cf. 85) or an epithet (cf. 86) that is crossed. (84) (85)

* As for Johnj, that bastardj says that we must obey tj absolutely. As for that manj, hej said that hej would kill all of us.’

(86) a. As for that manj, the bastardj said that hej would kill all of us. b. My sonj, the kidj said that hej would never dare again to drive after drinking.

Distribution of resumptive pronouns in relatives and in dislocation structures  

French also behaves the same way as Chinese and English. Dislocation structures with gaps always give rise to the strong crossover effect (cf. 87) but resumptive dislocation structures do not, irrespective of whether the crossed element is a pronoun (cf. 88) or an epithet (cf. 89). (87)

* Quant à Jeanj, ce Salaudj dit que nous as.for Jean the bastard says that we devons absolument obéir à _____j. must absolutely obey to ‘As for Jeanj, that bastardj says that we must obey ____j absolutely.’

(88)

Ce typej-là, ilj a dit qu’ilj allait this guy.that he has said that.he would nous tous tuer. us all kill ‘As for that manj, hej said that hej would kill all of us.’

(89) a.

Ce typej-là, ce salaudj a dit qu’ilj this guy.that this bastard has said that.he allait nous tous tuer. would us all kill ‘As for that manj, the bastardj said that hej would kill all of us.’

b.

Mon filsj, le gaminj a dit qu’ilj n’oserait my son the kid has said that.he Neg.dare plus conduire après avoir consommé de l’alcool. more drive after have consumed of the.alcohol ‘My sonj, the kidj said that hej would never dare again to drive after drinking.’

Recall that the same contrast has been observed in English, Chinese and French. Epithets give rise to crossover effects in relatives but not in dislocation structures. ... Summary Weak crossover effects (WCO) and strong crossover effects (SCO) are classical tests for A'-movement. In this section I apply these tests to Chinese and the results are summarized in the following table.

  Syntactic derivation: two types of A'-dependencies Tab. 3: Crossover effects in Chinese

Gap

Resumptive

Ex-situ cleftfocus

Resumptive

Resumptive

Gap

LD-structures Gap

Relatives

Weak crossover

yes

yes

yes

no

yes

no

Strong crossover

yes

yes

yes

no

yes

no

From the table we can see that crossover effects (weak and strong) are always produced in relatives, irrespective of whether the gap strategy or the resumptive strategy is adopted. By contrast, in dislocation structures (LD-structures and exsitu cleft-focus structures), crossover effects are only detected when the gap strategy is adopted. Importantly, dislocation structures with resumptive pronouns never give rise to crossover effect. Since LD-structures and ex-situ cleftfocus structures behave in exactly the same way, I will use the term “dislocation structures” to refer to both of them in the rest of this chapter when their structural difference is irrelevant to the discussion. Let us turn to English and French. The results are given in the following table. Tab. 4: Crossover effects in English and in French

Gap

Resumptive

Ex-situ cleftfocus

Resumptive

Resumptive

Gap

LD-structures Gap

Relatives

Weak crossover

yes

--

yes

no

yes

no

Strong crossover

yes

--

yes

no

yes

no

Distribution of resumptive pronouns in relatives and in dislocation structures  

English and French behave in a similar fashion in the tests for movement. Relatives without any island do not tolerate the presence of any resumptive pronoun and only the gap strategy is allowed, which distinguishes themselves from languages such as Chinese where the resumptive strategy is allowed in island-free relatives. However, English and French relatives with gaps always give rise to crossover effects, and Chinese behaves in exactly the same way. In addition, in English, French and Chinese, dislocation structures always display crossover effects when the gap strategy is adopted and importantly, resumptive dislocation structures never give rise to the crossover effect. From this perspective, English, French and Chinese do behave in a similar fashion when the crossover effects are involved. .. Summary In this section two classical diagnostic tests for A'-movement, island effects and crossover effects, have been applied to three types of A'-dependencies in Chinese: relatives and dislocation structures (i.e. LD-structures and ex-situ cleftfocus structures). The results of these tests are presented in the following table. Tab. 5: Syntactic distribution

i) Without island

Relatives

Dislocations

Gaps

Gaps

Resumptives Intrusives

Resumptives Intrusives

yes

yes

-------

yes

yes

--------

yes

----

no (island effects)

yes

----

yes (no island effects)

iii) Weak crossover effect

yes

yes

------

yes

no

--------

iv) Strong crossover effect

yes

yes

------

yes

no

--------

ii) With island

Let us compare relatives with dislocation structures. First, there is no difference between relatives and dislocation structures when the gap strategy is adopted. Both relatives with a gap and dislocation structures with a gap give rise to island effects and crossover effects (weak and strong).

  Syntactic derivation: two types of A'-dependencies Second, there is no difference between relatives with a gap and relatives with a resumptive pronoun in that both strategies display island effects as well as weak and strong crossover effects. Third, there is a contrast between dislocation structures with a gap and those with resumptive pronouns. The gap strategy systematically gives rise to island effects as well as weak and strong crossover effects whereas the resumptive strategy never gives rise to any of these effects. Finally, there is a very important contrast between relatives and dislocation structures when the resumptive strategy is adopted. Resumptive relatives always display island effects and crossover effects (weak and strong); however resumptive dislocation structures never give rise to island effects nor to crossover effects. To conclude, the results of the tests based on island effects and crossover effects reveal that relatives with a gap, relatives with resumptive pronouns and dislocation structures with a gap behave in a similar fashion in that all of these constructions systematically give rise to island effects and crossover effects (weak and strong). By contrast, resumptive dislocation structures behave differently with regard to the three constructions mentioned above because resumptive dislocation structures never display island effects nor crossover effects. The most important observation in this section is that the crucial distinction should not be made simply between relatives and dislocation structures but between relatives (with gaps and with resumptive pronouns) and dislocation structures with gaps on the one hand, and dislocation structures with resumptive pronouns on the other! This distinction is highlighted by the color contrast in Table 5.

. Analyses During the different phases of the development of generative syntax, different analyses have been proposed to account for the syntactic properties observed in resumptive constructions. I presented several important representative analyses in the first chapter and I will not repeat them here. Each analysis is more or less based on one particular language. In the framework of the Government and Binding theory, analyses based on movement (Sells 1984, Tellier 1991) have been proposed for languages like Hebrew and Irish. Analysis based on LF-movement of a null operator has been proposed for languages like Egyptian (Demirdache 1991). Analysis based on a kind of sub-extraction of one part of a pronoun was proposed for languages like Welsh (Rouveret 1994, 2002, 2008). Moreover, one of the important questions

Analyses  

linked to the nature of resumptive pronouns has received different answers on the basis of different language facts. Importantly, this question is closely related to the syntactic derivation of resumption. A very close link between the internal structure of resumptive pronouns and the syntactic derivation of the resumptive dependencies has been pointed out by most scholars. Recall that in Vata, a resumptive pronoun was merely analyzed as a phonologically spelled out trace (Koopman 1983) because in a very few restricted positions, there is no difference in the syntactic distribution of a resumptive pronoun as opposed to wh-trace. Let me emphasize again that such a point of view is only based on Vata and that the syntactic positions allowing the free alternation between a resumptive pronoun and a wh-trace are very restricted. Interestingly, there is an analysis that takes a completely opposite position by claiming that resumptive pronouns always play the role of the last resort in all of the different cases to avoid potential violation of locality constraints (Shlonsky 1992). In this section, I will propose a minimalist analysis of the derivation of resumptive dependencies in Chinese and will also compare this analysis with a possible one in the framework of GB. .. In the framework of Government and Binding theory I emphasized in Chapter 1 that the essential question that I will try to answer at the end of this monograph is what Chinese can tell us about resumptivity in general. The two contrasts that we observed in the previous section suggest that besides some properties related to different languages, certain syntactic properties are specifically related to different structures.13 In other words, if the same type of resumptive pronoun behaves differently in two different types of A'dependencies, then it is not totally unreasonable to assume that certain differences must exist in the nature of the derivation process of these dependencies. This is precisely one of the hypotheses that I will make in this study. Concretely, I will make an assumption based on two different considerations: one is that relatives can be derived very differently with regard to dislocation structures; and the other is that A'-dependencies with a gap are derived differently with regard to resumptive A'-dependencies. Importantly, the hypothesis that I will develop must account for the four generalizations that I made on the basis of  13 I will discuss in Chapter 3 certain syntactic properties are in fact linked to the two types of pronouns in Chinese on the one hand and to the two internal structures of resumptive pronouns on the other hand. In the present chapter, I will only concentrate on the syntactic derivation of resumptive dependencies.

  Syntactic derivation: two types of A'-dependencies the result of the different tests presented in Table 5. Let us examine these four generalizations here. (i)

There is no difference between relatives and dislocation structures when the gap strategy is adopted. Both relatives with gaps and dislocation structures with gaps give rise to island effects and crossover effects (weak and strong).

What property (i) shows is that an A'-dependency with a gap is always derived by the same mechanism without making any distinction between these different types of dependencies. In the framework of the Government and Binding theory, island effects and crossover effects are reliable diagnoses for A'-dependency. In addition, what property (i) also shows is that the result of these tests suggests that movement must be involved in the derivation of these A'-dependencies: relatives with a gap and dislocation structures with a gap. An important implication of this assumption is that A'-dependencies with gaps are systematically derived by movement, which is stated as a general hypothesis. Along this line of the reasoning, the results that I get are in favor of the idea according to which there is a nominal head raising in the derivation of relatives (Vergnaud 1974, Kayne 1994) and also that in LD-structures, the topic phrase located in TopP is derived by movement. Both types of movements are A'-movements and are subject to locality conditions such as island constraints. The traces resulting from these movements must obey the constraint on the bound variable construal. Any violation of this constraint will give rise to crossover effects. (90) a. Relatives with a gap: [CP C0 [TP ……tj……]]] [DP DPj

b. Dislocation structures with a gap: [TopP DPj, [TP …… tj……]]

(ii)

There is no difference between relatives with gaps and relatives with resumptive pronouns in that both strategies give rise to island effects as well as to weak and strong crossover effects.

Property (ii) suggests that relatives with gaps and relatives with resumptive pronouns are derived in the same way and that the derivation of both structures involves A'-movement, which is illustrated in (91).

Analyses  

(91) a. Relatives with a gap: [CP C0 [TP ……tj…..]]] [DP DPj

b. Relative with a resumptive pronoun (RP): [DP DPj [CP C0 [TP ……RPj…..]]]

In other words, in relatives, resumptive pronouns and gaps do not display different syntactic properties and for the moment, in a very superficial way, we observe that a resumptive pronoun can be analyzed as a spelled out trace in relatives in Chinese. However, as will be explained in detail in chapters 3 and 4, this superficial observation is not at all correct and it is due to pure coincidence. (iii)

There is a contrast between dislocation structures with gaps and those with resumptive pronouns. Gap strategy systematically gives rise to island effects as well as to weak and strong crossover effects whereas the resumptive strategy never gives rise to any of these effects.

As the reader has seen, I suggested in property (i) that dislocation structures with gaps are derived by A'-movement which is subject to locality constraints and gives rise to island effects. What property (iii) suggests here is that in contrast to dislocation structures with gaps, dislocation structures with resumptive pronouns presumably do not involve any A'-movement. And from this perspective, a crucial generalization can be drawn: in dislocation structures, resumptive pronouns do not have the same properties as gaps and they are certainly not free alternatives. This is shown in (92). (92) a. Dislocation structures with a gap: [TP …… tj……]] [TopP DPj ,

b. Dislocation structures with a resumptive pronoun (RP): [TopP DPj , [TP …… RPj……]]

In (92b), we can assume that the topic DP is always base-generated in the TopP and the resumptive pronoun enters into the derivation from the beginning.

  Syntactic derivation: two types of A'-dependencies (iv)

There is a very important contrast between relatives and dislocation structures when resumptive strategy is adopted. Resumptive relatives always give rise to island effects and crossover effects (weak and strong); however resumptive dislocation structures never give rise to island effects nor to crossover effects.

The contrast revealed in (iv) specifies that resumptive relatives are systematically derived by movement whereas the derivation of resumptive dislocation structures does not involve any movement operation. Importantly, resumptive pronouns do not behave uniformly in relatives and in dislocations and such a contrast is crucial in this study because it shows that resumptive pronouns cannot be analyzed in the same way in different types of A'-dependencies. Moreover, resumptive pronouns can behave differently in the same language. (93) a. Relatives with resumptive pronouns (RPs) [TP ……RPj…..]]] [DP DPj [CP C0

b. Dislocation structures with resumptive pronouns (RPs) [TP ……RPj……]] [TopP DPj ,

.. In the Minimalist Program In this section, I will begin by giving a brief introduction to the essential notions and operations of the Minimalist Program (Chomsky 1995, 2000, 2001, 2004, 2008) that I will use in this study. Then, I will examine several important solutions proposed previously to account for the derivation of resumptive constructions in the Minimalist Program. And finally, based on the previous analyses, I will propose my own solution to account for the distribution of the gaps and of resumptive pronouns in the two different types of A'-dependencies that we observed in Section 2.2. ... Match, Agree and Move In this section, I will begin by introducing several important basic notions in the Minimalist Program.

Analyses  

The relevant notions and minimalist operations are listed below: – Merge – Agree – Match – Move – Feature checking – Valuation/ feature valuation It is worthwhile pointing out that the analyses that I will propose in this study particularly rely on operations such as Agree and Match. Merge is a fundamental operation that simply takes two objects α and β to form a new object, i.e. a set {α, β}. (94)

α

{α, β}.

β

Agree is an operation that establishes an agreement relationship between the head of a functional projection that works as Probe and another element that works as Goal. A Probe bearing a set of uninterpretable features F' searches for a Goal that bears the same set of features, say F, that matches the features F', in order to value F'. Match: The process of “searching for a Goal” is called “feature matching”. Matching is therefore defined as a kind of feature identity relation that exists between the features attached to the Probe and those attached to the Goal. Specifically, a feature can be considered as an Attribute-Value pair. For instance, the pronoun he has the following features: (95) he: [PERSON: 3rd] [NUMBER: singular] [GENDER: masculine] The items like “person, number and gender” are attributes and those like “3rd, singular and masculine” are values. For instance, the pronoun ta1 ‘he’ and tamen ‘they’ have the same attributes but different values. When a lexical element enters into the process of a derivation, the features on this element can possess a value, but it is also possible that such an element does not possess

  Syntactic derivation: two types of A'-dependencies any value at all. The crucial point is that Match does not require an absolute identical relation between the features on the Probe and those on the Goal but only a kind of identical Attribute (person, number and gender) of the features and in this sense, the Value of these matched features can be different on a Probe and on a Goal. For instance in (96), T° is a potential candidate for the Probe and imagine that he is the potential candidate for the Goal. Match checks whether the relevant features attached to these two items have the same Attributes without taking into consideration their value (cf. Rouveret 2015’s reformulation). (96)

T° ………. [PERSON: ___] [NUMBER: ___] [GENDER: ___]

he [PERSON: 3rd] [NUMBER: singular] [GENDER: masculine]

In (96), both the features of T° and those of he contain the same set of Attributes [PERSON, NUMBER, GENDER] and the Matching chain can be established between the two.14 Once the Probe finds the matched Goal, the features associated with the Goal matches those associated with the Probe. Then the process of valuation of these features begins. Valuation: The set of the features F' attached to the Probe is uninterpretable and underspecified. Uninterpretable features are unvalued, such as the set of the features of T° in (96) that only possesses the attributes but not the value. One technical detail is that an agreement relationship can be established between a Probe and a Goal only if both of them are active, which means that there are still unchecked uninterpretable features attached to them. Agree is an operation that only works on features; however, since an unvalued feature does not possess any value, it cannot be interpretable at interfaces. Such an unvalued feature must get a value via a sort of dependency relationship during the course of the derivation. Agree is the only operation that can provide uninterpretable features with a value. The Goal that bears a set of valued features will provide the set of the unvalued features F' attached to the Probe with a value. The operation Agree will copy the value of the interpretable features associated with the Goal to the uninterpretable features attached to the Probe.  14 T° also contains Case and Tense features (and EPP feature in English), which are irrelevant here.

Analyses  

Valuation precisely consists of feature checking. The result of feature checking is to delete all of the uninterpretable features between the Probe and the Goal. In this sense, Agree can also be considered as an operation that deletes the uninterpretable features attached to the Probe and to the Goal. Importantly, in the Minimalist Program, feature checking is realized by valuing the uninterpretable features. The schema in (97) illustrates how Agree functions. (97)

XP

Probe [uF']

YP

ZP

Goal [iF] - Feature Matching Agree: - Feature Valuation - Feature Checking (iF : interpretable features, uF : uninterpretable features) Let us come back to the concrete example in (96) where Match has checked the Attributes of the set of the values and the matching chain can establish a feature identical relationship between the Probe and the Goal. Once Matching is finished, Agree begins to copy the value of the interpretable features attached to the Goal, he, to the Probe T°, as illustrated in (98a) for ϕ features (i.e. Phi features, such as PERSON, NUMBER and GENDER). As we can see, all the unvalued and uninterpretable features have been valued in (98a). Once all of the unvalued features are valued, the former uninterpretable features attached to the Probe become interpretable and all of the uninterpretable features attached to the Probe can now be checked and deleted, as illustrated in (98b).

  Syntactic derivation: two types of A'-dependencies (98) a. Feature valuation T° ………. [PERSON: 3rd ]  [NUMBER: singular]  [GENDER: masculine] 

[PERSON: 3rd] [NUMBER: singular] [GENDER: masculine]

b. Feature checking T° ………. [PERSON: 3rd ]  [NUMBER: singular]  [GENDER: masculine] 

he [PERSON: 3rd] [NUMBER: singular] [GENDER: masculine]

he

It is very important to bear in mind that Agree not only applies to establish an Adependency but also applies to establish an A'-dependency. In the latter case, during the entire process of Agree, locality conditions must be obeyed. The potential candidate for the Goal must be the nearest one to the Probe. Operation Agree is therefore sensitive to island boundaries. Long distance Agree must apply cyclically and phase by phase. The notion of locality will be elaborated in detail later in this chapter. Move: In the Minimalist Program Move is defined quite differently from the operation movement used in the Government and Binding theory. In the first version of the Minimalist Program (Chomsky 1995), Move is defined as Copy + Merge, and in this sense, Move is a kind of re-Merge. Let us take wh-movement for example. First, the phonetic and phonological form of the wh-constituent and the entire set of features attached to it are copied. Then, this copy of wh-phrase will be combined with the head C° to form a CP. Finally, the lower copy left in its original argument position will be deleted. In recent versions of the Minimalist Program, Move is defined as: Agree + determine what is going to be moved with the features (i.e. Pied-piping of a lexical item) + Merge. Certain types of uninterpretable features, such as the EPP feature, can determine (i) whether the Probe is capable of providing an available position for movement and (ii) if such a position is available which category can move there. The crucial point in this view of Move is that movement is only triggered to satisfy the requirement of the EPP feature and that movement of a lexical category is realized by pied-piping the phonetic form of such a category with its features. From this point of view, most cases involving movement in the framework of Government and Binding theory can be now derived by Agree

Analyses  

alone without Move in the Minimalist Program.15 The two operations, Move and Agree, must obey locality conditions in exactly the same way and both of them give rise to island effects and intervention effects.16 (99)

Probe [uF'] [EPP] WP'

XP

YP

ZP

Move

Agree

Goal [iF] WP

... Deriving resumptive constructions by Agree The Minimalist Program provides us with the possibility of establishing A'dependencies without movement. An A'-dependency can be derived by Agree that works necessarily and exclusively at the Narrow Syntax level. Previous literature such as Rouveret (2002, 2008, to appear) and Adger & Ramchand (2005) argue that the derivation of a resumptive dependency involves Agree instead of Move. The idea behind these analyses is that the highest complementizer C° whose specifier position is occupied by the antecedent NP functions as a Probe and the resumptive pronoun downstairs functions as a Goal. The uninterpretable features that have not been valued yet attached to the highest C° will get a value from the interpretable features attached to the resumptive pronoun via Agree. Such a derivation by Agree is realized phase by phase. Accordingly, the core operation of these analyses is called “Phasal Agree”. In the phase theory, vP and CP are strong phases and when the head of a higher phase (for instance, C°) is projected, the complement of the phasal head v° is sent to inter 15 Richards (2001) discusses different versions of movement. 16 Boskovic (2007) has a different point of view of the locality that conditions Agree.

  Syntactic derivation: two types of A'-dependencies faces for interpretation. Every well-established phase will be sent to the interfaces for interpretation via the operation Transfer. Once the relevant phase is transferred to LF and to PF, it becomes inaccessible. The previous analyses seem to agree with the idea that it is the operation Agree that functions alone to derive a resumptive construction. However, the differences between these analyses are centered on the question: what kind of feature is attached to the complementizer and to resumptive pronouns respectively? Different authors have different points of view on this question. I will present two important analyses based on Agree that try to answer this question: Adger & Ramchand (2005) and Rouveret (2002, 2008, to appear) in the following sections. .... Adger & Ramchand (2001, 2005) The theoretical consideration behind the analysis of Adger & Ramchand (2001, 2005) is based on the Interface Legibility constraint requiring that all of the syntactic structures should be legible at interfaces. The technique that they use is feature valuation via phasal Agree. On the one hand, resumptive pronouns are base-generated in relatives and on the other hand, two different formal features are necessary for the interpretation of relatives at LF. Concretely, for a resumptive relative clause, the two features in question are [λ] and [ID]. [λ]:

[ID]:

Lambda is interpreted as an abstraction operator and it creates an open position inside a clause, which makes such a clause predicative. The CP bearing the [λ] feature will be interpreted as a predicate and the open position that it creates will be translated as a variable in semantics. the [ID] feature is interpreted as a variable occupying an open position created by λ-abstraction. In resumptive relative clauses, such a variable position is generally occupied by a resumptive pronoun. Importantly, for Adger & Ramchand, all pronouns are not only referentially dependent but also bear the [ID] feature. [ID] has two different values that are syntactic distinguishable. If the pronoun is identified via an assignment function controlled by an operator, the feature [ID] has [dependent] as value, written as [ID : dependent]. The feature [ID] will have [Phi] as value, written as [ID : phi] if such a pronoun is not only identified via an assignment function controlled by the context but also coherent with the ϕ-features of the pronoun. In this case, the pronoun will not depend on an λ-operator.

Analyses  

Typologically, these features with different values allow us to make a distinction between three types of languages with regard to the derivation of resumptive constructions. A) Cases where operation Agree applies cyclically In languages like Scottish Gaelic, when the complement of a preposition is relativized, the preposition always takes the masculine form under agreement. This is even true when the relativized element takes the feminine form, as illustrated in (100). (100)

Siud a' chaileag a dh'eisd that the girl C-Rel listen-Past ‘That's the girl that you listened to.’

thu you

ris to-3MSg

Scottish Gaelic, Adger & Ramchand (2001) This example shows that no movement is involved in the derivation of relatives in that the null pronoun pro located in the variable position does not move at all. The pro part bears [ID : dependent] but not [ID : ϕ] since there is no agreement with the ϕ-features. In order to account for the absence of the identity effects between the antecedent and the preposition in this A'-dependency, the [dependent] value must be selected. The process of agreement is realized as shown in (101) below. (101) = (100)

a … …. …. … … C-Rel [λ] [ID : dep.] Valuation

pro RP [ID : dep.]

In Scottish Gaelic, long distance relativization exists. The complementizer a always introduces all types of intermediate clauses in the case of long distance dependencies (cf. 102a). In the same context, the complementizer for declarative sentences, gum, is excluded (cf. 102b). (102) a.

an duine a thuirt e a the man Rel said he Rel ‘the man that he said that he will hit’

bhuaileas will-hit

e he

(* e) it

  Syntactic derivation: two types of A'-dependencies b.

* an duine a thuirt e gum the man Rel said he that (‘the man that he said that he will hit’)

bhuaileas will-hit

e he

__ [ __ ]

One should notice that in these two examples, the first complementizer a introduces the relative clause and the second a cannot be replaced by the declarative complementizer gum as indicated in (102b). It is important to point out that for Adger & Ramchand, movement is never involved in the derivation of this kind of relative clause. The variable position is occupied by the null pronoun pro bearing the unvalued ID feature. Since the pronoun in the variable position has only a null form, it does not possess the [ϕ] feature. Importantly, the sentence in (102a) shows convincingly that the pronoun e ‘it’ that bears the [ϕ] feature is excluded from the variable position. The complementizer of the relative clause a not only bears [ID : dependent] but also [λ] functioning as an abstraction operator that makes the complement of C° predicative. The null pronoun pro depends syntactically on the complementizer in order to get the relevant features in the agreement domain. Since the entire dependency is formed by Agree, locality constraints must be obeyed. Let us examine this derivation in detail as shown in (103). There are three different Agree operations involved in the derivation of the relative dependency. The pronoun that bears the unvalued [ID] feature will get [dependent] as value from the intermediate complementizer a, which is the nearest Probe to the null pro, via Agree 1. Then, the intermediate complementizer agrees with the highest complementizer that bears both [ID: dep.] and [λ] respectively through Agree 2 and Agree 3. The [ID: dep.] feature attached to the resumptive pronoun will be interpreted as a bound variable because, according to Adger & Ramchand, only the feature that is located at the lowest position will be interpreted at interfaces. (103) = (102a) a………… … C-Rel – [λβ] Agree 3 – [ID: dep.α]

a … …. …. … … C-Rel [λβ]

[ID: dep.α] Agree 2

↓ λx

Agree 1

pro RP

[ID: dep.α]

↓ x

Analyses  

B) Cases where semantic binding applies but Agree does not apply The analysis proposed by Adger & Ramchand (2005) can also account for the derivation of resumptive constructions in Modern Irish. In Irish, relatives take the complementizer with a special form aN that is always located in the highest position and a resumptive pronoun always has an explicit morphological form with overtly specified ϕ-features, as illustrated in (104). (104)

an scríbhneoir aN molann the writer C-Rel praised ‘the writer who the students praised’

na the

mic léinn students

é him

Importantly, in this case the interpretable syntactic dependencies are not necessary. The highest complementizer bearing the [λ] feature functions as λoperator which can create a predicate; and a resumptive pronoun that bears the interpretable [ID: ϕ] feature will be translated as a variable x in semantics. The binding relation between the λ-operator and the variable x can only be established via semantics without involving the narrow syntactic operation Agree. Since Agree does not apply in this case, locality effects are therefore not expected, as illustrated in (105). (105) = (104)

aN C-Rel [λ] ↓ λx

… …. …. … … é him [ID : ϕ] ↓ x

Movement is not at all an option for the derivation of relatives in Scottish Gaelic or in Modern Irish, irrespective of the choice of the specific complementizer. C) Cases where movement applies In Adger & Ramchand’s analysis, the derivation of relatives in English requires a movement option. The relative pronoun who bears the [λ] feature and the explicitly specified [ID : ϕ] feature. Who is introduced in an argument position where the [λ] feature cannot be interpreted and as a result, who moves to the highest complementizer position in order to be interpreted properly at LF. This movement is applied cyclically and obeys the locality condition. The [λ] feature interpreted in the highest C position functions as λ-operator and the [ID : ϕ] feature at the bottom of the chain is interpreted as a variable (cf. 106).

  Syntactic derivation: two types of A'-dependencies (106) Relatives in English [who C0 [I thought [ C0 [TP we moving]]]] [λ] [λ] [λ] [ID : ϕ] [ID : ϕ] [ID : ϕ] ↓ ↓ λx x Even if different combinations of the relevant features in the analysis of Adger & Ramchand can neatly account for the three different types of resumptive constructions in Scottish Gaelic, Irish and English, this analysis still runs into trouble when applied to Mandarin data. Take for example a resumptive relative clause that I examined, (107)

*我碰到了[一勤認識[擁抱他 j 的]那位女同學的]法國影星 j。 * Wo pengdao-le [Yiqin renshi [yongbao-guo ta1j ] 1Sg meet-Perf Yiqin know embrace-Exp 3MSg de na-wei nü-tongxue de] Faguo yingxingj . C that-Cl female.student C French movie.star (‘I met the French movie starj [that Yiqin knows the female student [who embraced (himj)]].’)

In (107), the relativization of the antecedent NP the French movie star from an island constructed by another relative clause leads to the ungrammaticality of the sentence. The insertion of the intrusive pronoun ta1 ‘him’ cannot save such a sentence from violating locality constraints and therefore, island effects are always detected. In Chinese, the complementizer de is a kind of structural particle that links a modifier to the head noun that it modifies. The particle de does not undergo any morphological change and the modifier that de links can be either an adjective or a nominal or a possessive or a relative clause. Therefore, we can assume that de bears the [λ] feature and it functions as a λ-operator in the sense of Adger & Ramchand. However, it is very difficult to imagine that de also bears the [ID : ϕ] feature since de never morphologically agrees with a resumptive pronoun in terms of ϕ-features. As for resumptive pronouns, it was made clear at the beginning of this chapter that there is always agreement between the antecedent and resumptive pronouns with regard to ϕ-features. Since in Chinese, a resumptive pronoun always takes a morphologically explicit form (in its character graphic writing form (cf. Appendix)), we can thus assume that it bears an interpretable [ID : ϕ] feature and therefore, it does not need to be valued by the C-Rel via a syntactic dependency. Semantically, the pronoun [ID : ϕ] will be

Analyses  

automatically translated as a variable x at LF. Along this line, (107) will have the following schema. (108) = (107)

de C-Rel [λ] ↓ λx

… …. …. … … ta1 3MSg [ID : ϕ] ↓ x

After a closer look at (108), one can see that the derivation in (108) has exactly the same configuration as in Irish (cf. 104, 105) where the binding between the operator and the variable x is exclusively established in semantics or at LF without involving any syntactic dependency. Since this dependency is neither established via the operation Agree nor via the operation Move, island effects are therefore not expected in such a configuration. However, the ungrammaticality of (107) shows that island effects are unmistakably detected. Therefore, this analysis does not correctly predict the Chinese data. Bianchi (2011) also discusses two difficulties that the analysis of Adger & Ramchand (2005) meets relative to semantics. First, their analysis predicts that a specific reading is always assigned to resumptive pronouns in all of the contexts. As a matter of fact, this prediction cannot account for all of the interpretative properties of resumptive constructions. Moreover, it is precisely these interpretative properties that distinguish the A'-dependencies with resumptive pronouns from A'-dependencies with gaps. In addition, A&R’s analysis cannot account for the reconstruction effects observed in resumptive constructions. I will examine all these different kinds of reconstruction effects in detail in chapters 3 and 4. .... Rouveret (2002, 2008, to appear) Let us turn to another Agree-based analysis proposed by Rouveret (2002, 2008, to appear) to account for the Welsh data. The main idea of his analysis is that resumptive constructions can be derived by Agree alone without involving Move at the phasal level. In relatives, a resumptive pronoun functioning as a Goal is A'-bound by the head C° functioning as a Probe. This point of view is shared by Rouveret and Adger & Ramchand (2005). However, these two analyses differ in the nature of the features attached to the Probe and to the Goal. I will present the first version of Rouveret (2002, 2008)’s analysis and then his second version of analysis (Rouveret to appear). The essential data from Welsh are cited below:

  Syntactic derivation: two types of A'-dependencies (109) a.

b.

y dyn y soniais the man that I-talked ‘the man I talked about’

amdano about-[agr]

y dynion y dywedodd Siôn the men C said Siôn y llyfr the book ‘the men that Sion said had read the book’

y C

darllenasant read-3Pl

Welsh, Rouveret (2011)

Rouveret (2002, 2008) presents his first version in which the Probe C° in Welsh relatives bears three features: [Rel]:

[ϕ]:

EPP:

When the highest C° takes a clause, containing a resumptive pronoun, as complement, this feature indicates that such a clause is a locus of an A'-relation. In this sense, [Rel] is interpretable when it is associated with C°. The Probe C° also bears the [ϕ] feature. An empirical argument is based on the fact that a clitic can attach to complementizers in Welsh. However, when the [ϕ] feature is associated with C°, it is uninterpretable and it will be valued once it enters into an agreement relation with a resumptive pronoun. C° also bears the EPP feature that requires the specifier position of the CP to be filled. In the analysis of Rouveret, a null operator is merged (i.e. base-generated) directly in [Spec, CP], which does not involve any movement operation.

In this analysis, resumptive pronouns bear an uninterpretable [Rel] feature and an interpretable [ϕ] feature. However, a resumptive pronoun cannot be directly linked to C° since Agree must take place phase by phase. Recall that in the phase theory, besides CP, vP is also a strong phase and that is why the agreement relation between a resumptive pronoun at the bottom of the chain and the C° must be constructed via v°. The phasal head v° will not only check the features attached to the resumptive pronoun but also check the features attached to the C° head. At the vP phase, the first agreement chain is established between the resumptive pronoun and the phasal head v° before the domain of the phase vP (i.e. VP) is sent to the interfaces. At the CP phase, the second agreement chain is established between the phasal head v° and the phasal head C°. This is

Analyses  

why, in the system proposed by Rouveret, the phasal head v° bears both features, [Rel] and [ϕ]; however, both of them are uninterpretable features when associated with the phasal head v°. Along this line, Rouveret proposes the following feature system for resumptive constructions in Welsh. (110) D NP [ y … [v0 …. [ y … [v0 … [y … pronoun … ]]] u-phi u-phi u-phi u-phi u-phi i-phi i-Rel u-Rel u-Rel u-Rel u-Rel u-Rel EPP (u : uninterpretable feature; i-interpretable feature) It is very important to point out a technical detail in Rouveret’s analysis. Recall that a pre-condition on the establishment of a dependency relation between a Probe and a Goal via Agree is that both elements must be syntactically active. Being active in this particular case means that there should still be uninterpretable features to be valued on each of these two elements. In the schema presented in (110), each phasal head (i.e. either C° or v°) and the resumptive pronoun bear at least one uninterpretable feature, which ensures that the relevant dependency is going to be successfully established by Agree. Technically, the [Rel] feature is crucial since without this feature, once the u-phi feature attached to C° or to v° at the bottom of the chain is valued by the resumptive pronoun, the derivation crashes. The need for such an extra uninterpretable feature ensures that this feature is always active without being valued before the derivation ends at the higher C° heads. Rouveret (2008) calls this extra feature the “activation feature”. Similarly, the [λ] feature in the system proposed by Agder & Ramchand (2005) has exactly the same function. Rouveret (to appear) investigates more precisely the status of the three features that he proposes in his previous work, which is motivated by theoretical considerations. Recall that in the derivation by Agree that I presented in Section 2.2, a c-commanding relationship exists between a Probe and a Goal. Concretely, the uninterpretable features that have not been valued attached to a Probe must c-command the valued interpretable features attached to a Goal. The operation Agree copies the value of the interpretable features from the Goal onto the uninterpretable features of the Probe that have not been valued yet (cf. Section 2.3.2.1). It is not difficult to see that in the schema based on Agree proposed in (110), the c-command requirement is not satisfied for the [Rel] feature. In order to develop this analysis, Rouveret (to appear) re-examines the variable status of resumptive pronouns. He proposes that since a resumptive pronoun is systematically interpreted as a bound variable and not as an ordinary referential pronoun, it must bear some kind of variable feature [var]. Since the variable

  Syntactic derivation: two types of A'-dependencies feature is capable of identifying the status of a resumptive pronoun as a bound variable, it is an interpretable feature and it possesses a value from the moment the pronoun is introduced into the derivation. As for the C° head in resumptive relative clauses, Rouveret assumes that the C° head also bears this variable feature [var] in order to identify the clause that it introduces as the domain of an A'-dependency, which is of course similar to the function of the [Rel] feature in the previous versions of his analyses. However, the crucial difference between the previous analyses and the present one is that when the [var] feature is associated with the C° head, it is an unvalued uninterpretable feature. The uninterpretable [var] attached to the C° head c-commands the interpretable [var] feature attached to the pronoun and the former feature will value the latter one. In this sense, the C° head that bears an uninterpretable [var] feature functions as a Probe and therefore triggers the Agree relation with a resumptive pronoun functioning as a Goal. Second, Rouveret also re-examines the interpretability of the [ϕ] feature attached to resumptive pronouns, which is a very important but complex problem in all of the Agree-based analyses of resumption. Recall that in the previous section, I presented a potential problem that the analysis proposed by Adger & Ramchand (2005) could encounter based on the data from Chinese. The essential problem is precisely linked to the hypothesis that the [ϕ] feature attached to a resumptive pronoun is an interpretable feature. Actually, doubt about the interpretability of the [ϕ] feature already exists in the previous analyses in favor of Agree. The observation is that the ϕ-features attached to a resumptive pronoun do not necessarily coincide with the same ϕ-features attached to the antecedent. Surprisingly, in Scottish Gaelic, non-agreement is possible (cf. 111a) but agreement is excluded (cf. 111b). (111) a.

Siud a' chaileag a dh'eisd that the girl C-Rel listen-Past ‘That's the girl that you listened to.’

thu you

ris to-3MSg

b.

* Siud a' chaileag a dh'eisd that the girl C-Rel listen-Past (‘That's the girl that you listened to.’)

thu you

rithe to-3FSg

Scottish Gaelic, Adger & Ramchand (2001) In the above sentences, the antecedent NP the girl bears a Feminin feature; however, the corresponding resumptive pronoun must take a masculin form and the feminin agreement is excluded.

Analyses  

Willis (2011) also shows that colloquial Welsh does not require agreement in terms of ϕ-features in resumptive structures and that the C° head takes the 3rd person singular form by default, as illustrated in the following examples.17 (112) a.

b.

Beth wyt ti 'n (ei) feddwl what are you Prog 3MSg think hyn yn ei olygu ___ this Prog 3MSg mean [___] ‘What do you think this means?’

mae is

Pa lyfrau wyt ti 'n ei feddwl which books are you Prog 3MSg think oedd Megan yn eu darllen ___? was Megan Prog 3Pl read [___] ‘Which books do you think Megan was reading?’ Colloquial Welsh, Willis (2011)

Rouveret (to appear) claims that ϕ-features on resumptive pronouns do not make any semantic contribution to the interpretation of a resumptive construction. His argument is based on the fact that the ϕ-features attached to resumptive pronouns are neither interpreted at LF nor in semantics and that such features only recapitulate the ϕ-features of its antecedent. In this version of analysis, the [ϕ] feature is always uninterpretable when attached either to a resumptive pronoun or to the C° head.18 The only grammatical function of the [ϕ] feature is to activate a potential candidate for Probe in an agreement relation between a Probe and a Goal. He also discusses in detail the fact that even if a Goal bears unvalued and incomplete ϕ-features, it can delete the unvalued uninterpretable features attached to a Probe via Agree under certain specific conditions. Importantly, Rouveret argues that valuation by Agree is possible be 17 Willis (2011) proposes that a resumptive construction involves a derivation by “Successive Move”. 18 The argument in support of the idea that the C° head bears the uninterpretable ϕ-features is based on the observation according to which in relatives, an ordinary pronoun can be realized as a enclitic affix attached to C°, such as ’ch in (i). (i) y dyn y 'ch gwelodd Mair yn siarad ag ef the man C.you saw Mair Prog talk with him ‘the man Mair saw you talk with’ According to the “non-selective attraction principle” proposed by Nash & Rouveret (2002), clitics can only attach to heads endowed with ϕ-features.

  Syntactic derivation: two types of A'-dependencies tween uninterpretable features. However, in this study, I will still take a traditional point of view to account for the Chinese data, according to which, only interpretable features attached to the Goal can value the uninterpretable features attached to the Probe. As the reader will see below, in my analysis based on Match, a resumptive dislocation chain can always be established even if the relevant dependency still contains unterpretable features attached to the Probe and to the Goal. As for the EPP feature, I will maintain the idea that this feature is the only motivation for movement. Under this view, Agree is undoubtedly licensed in the contexts in which Move is licensed; however, Agree is not considered as a precondition on Move. Rouveret proposes the following schema to represent the features associated with a resumptive construction. (113) C-relative u-[var] u-[ϕ] (EPP)

Resumptive pronoun i-[var] u-[ϕ] (incomplete)

This analysis makes a distinction among three different strategies that can form A'-dependencies: – Agree + Move – Agree alone – Neither Agree nor Move applies As for the derivation of a resumptive construction, analyses based on Agree work better for languages like Welsh and some Arabic dialects, in which resumption obeys island constraints, than languages like Irish, in which resumption does not give rise to island effects or to successive cyclicity effects. ... Resumption in Chinese In this section I will examine which minimalist mechanism applies best to the Chinese data observed in this chapter in order to account for the syntactic distribution of resumptive pronouns that I examined. Recall the main generalizations that I made based on the results of the syntactic tests: island effects and crossover effects. (i)

Relatives (either with a gap or with a resumptive pronoun) and dislocation structures with a gap systematically give rise to island effects and crossover effects.

Analyses  

(ii)

Dislocation structures with a resumptive pronoun do not give rise to any island effect or crossover effect.

.... Island effects Based on the two generalizations above, the hypothesis that I will make is that operation Agree applies to all of the cases specified in (i) but not to the one specified in (ii). Recall that the essential point that we have to bear in mind is that Agree applies phase by phase and cycle by cycle and this process takes place strictly at the level of Narrow Syntax and it obeys locality constraints. The difference between the solutions in the Government and Binding framework and those in the Minimalist Program is that the apparent violation of certain locality conditions, such as island constraints, are no longer the diagnostic tests exclusively reserved for movement. Again, the analyses proposed by Rouveret and those proposed by Adger & Ramchand show clearly that certain types of A'dependencies, such as relative clauses with a resumptive pronoun, can be established by Agree alone without involving any movement operation. This is the central idea that inspires my analysis of the data observed for Chinese. First, following Rouveret’s analysis of relative clauses in Welsh, I assume that resumptive pronouns in relatives bear the [var] feature in order to justify the bound variable status of resumptive pronouns and that this [var] feature should be an interpretable feature. Semantically, when a pronoun appears in an A'-bound variable position, it functions as a resumptive pronoun; in other words, a resumptive pronoun can only occupy an A'-bound variable position. The binding relation between the C-Rel and a resumptive pronoun in relatives is precisely defined as a relation between an operator and a variable. This relation will be interpreted properly at LF and in semantics. Since a resumptive pronoun needs to be interpreted obligatorily as a bound variable, the feature [var] attached to the pronoun must be an interpretable feature before the process of Agree (i.e. feature valuation) begins. By contrast, the complementizer C° of a relative clause can also bear the [var] feature but in this case, [var] is an uninterpretable feature. From this perspective, [var] associated with the C-Rel in Chinese plays the same role as Welsh in Rouveret’s system. The function of the [var] feature is to identify the relative clause that it introduces as a domain that licenses an A'-dependency. It is important to point out that this [var] feature works differently from the [λ] feature in the system of Adger & Ramchand (2005) in that [λ] attached to the C° can create an open position in order to render the relative clause predicative. In this sense, as we explained earlier, the C-[λ] functions as an abstractor. Then in their system, resumptive pronouns do not necessarily bear the same [λ] feature. It is thus difficult to see how the relationship between a Probe and a Goal can be established. In Rouveret’s system, the varia-

  Syntactic derivation: two types of A'-dependencies ble status of a resumptive pronoun is determined when it is located in a variable position and moreover, the [var] feature attached to a resumptive pronoun is endowed with a value and therefore it is a valued interpretable feature, which ensures its legitimate status as Goal. With this in mind, C-Rel in this system does not need to function as an abstractor because a variable position has already been created by the [var] feature attached to resumptive pronouns. As a result, the C-Rel in Chinese (and in Welsh) bears an u-[var] feature which means that it is uninterpretable. The relationship between a Probe and a Goal can then be established. The u-[var] (i.e. unvalued and uninterpretable) that functions as a Probe c-commands the i-[var] (i.e. valued and interpretable) that functions as a Goal. The u-[var] feature will be valued by the i-[var] feature via the dependency established by Agree. (114) Chinese C-Rel (de)… … … … … Pronoun u-[var] i-[var] Furthermore, a resumptive pronoun in Chinese always bears ϕ-features and it agrees systematically with its antecedent and this is the reason why I assume that resumptive pronouns bear the [ϕ] feature (i.e. the entire set of ϕ-features). Now the crucial question is whether this [ϕ] feature is an interpretable feature or an uninterpretable feature. As a matter of fact, whether this [ϕ] is interpretable or uninterpretable does not have any consequence for our analysis of Chinese. If we assume that the [ϕ] feature attached to a resumptive pronoun in Chinese is an interpretable feature, we do not really have any technical difficulty with the detail of the derivation via Agree. The only thing we need is that the complementizer C of the relative clause, occupied by the particle de, also bears the [ϕ] feature and that such a feature must be uninterpretable when attached to C-Rel. (115) C-Rel (de)… … … … … Pronom u-[ϕ] i-[ϕ] It is not surprising that the [ϕ] feature attached to the C-Rel is an uninterpretable feature in Chinese. The unvalued [ϕ] feature on the C head that functions as a Probe c-commands the interpretable [ϕ] feature attached to the resumptive pronoun that functions as a Goal. Under Agree, after feature valuation, the uninterpretable [ϕ] feature on C-Rel will be valued and then checked (i.e. deleted) and the derivation converges. It is precisely for this reason that there is no prob-

Analyses  

lem to assume that the [ϕ] feature attached to the resumptive pronoun is an interpretable feature. Under another possibility, if we assume that the [ϕ] feature attached to a resumptive pronoun is an uninterpretable feature, then we will have exactly the same situation as Welsh. Let us assume that the head C-Rel and a resumptive pronoun both bear the uninterpretable [ϕ] feature which has not been valued. There is a reason for which even if the [ϕ] feature attached to the pronoun is uninterpretable, operation Agree still applies. In this scenario, the resumptive pronoun already bears an interpretable i-[var] feature that will value the uninterpretable u-[var] feature on C-Rel. Agree can already apply to this dependency, irrespective of the fact that the other pair of features, u-[ϕ], attached to the Probe and to the Goal are uninterpretable. Put differently, assume that a Probe and a Goal bear several different formal features, [α], [β], [χ] and [δ]. Only the feature [α] remains uninterpretable on the Probe but interpretable on the Goal. Meanwhile [β], [χ] et [δ] are all uninterpretable features when they are attached the Probe and to the Goal respectively. Nothing can prevent Agree from establishing a resumptive dependency thanks to the feature [α], which suggests that [α] is enough to trigger Agree despite that the other features cannot trigger Agree. I have explained that both situations where resumptive pronouns bear either an interpretable [ϕ] feature or an uninterpretable [ϕ] feature can make the derivation successful. However, it is still preferable if the [ϕ] feature on resumptive pronouns is an uninterpretable feature because it is more reasonable to assume that only when a pronoun is not A'-bound, that is, when such a pronoun works as an ordinary referential pronoun, the [ϕ] feature attached to it is interpretable. By contrast, if such a pronoun is located in an A'-bound variable position, it functions as a bound variable and the [ϕ] feature attached to it does not give any concrete semantic interpretation at LF. These theoretical considerations seem to suggest that in relatives, the [ϕ] feature attached to resumptive pronouns cannot be interpretable. Let me summarize the different features associated with resumptive relatives in Chinese in the following schema. (116) C-Rel (de)… … … … Pronoun u-[var] i-[var] u-[ϕ] u-[ϕ] Till now, Chinese looks like Welsh on the interpretability status of the [ϕ] and [var] features, both of which are directly responsible for the island effects that we observed in Section 2.2 of this chapter. Island effects are unmistakably

  Syntactic derivation: two types of A'-dependencies expected when the concerned resumptive construction is derived by Agree. I will give a detailed description on how locality conditions apply in cases of Agree in Section 2.3.2.3.5. .... Crossover effects It was observed in Section 2.2 that crossover effects (strong and weak) have been produced in resumptive relative clauses in Chinese. As I pointed out, island effects and crossover effects are classical diagnostic tests for A'-movement in the framework of the Government and Binding theory. In the previous section, I showed that island effects are not exclusively reserved for movement since Agree, being a phasal bounded operation, also obeys locality conditions and gives rise to island effects. Now we still have crossover effects to explain and in this section, I will examine crossover effects in the Minimalist Program. The central question at this stage is whether crossover effects are still the exclusive diagnostic tests for A'-movement. In other words, we wonder whether or not an A'-dependency established via Agree also gives rise to crossover effects. Example (117) illustrates a strong crossover configuration in a relative clause with a resumptive pronoun. (117)

* 這就是[那個混蛋 j 說[他 j 要把我们全都殺了]的]那個人 j。 * Zhe jiu shi [na-ge hundanj shuo this exactly be that-Cl bastard say [ta1j yao ba women quan dou 3MSg will BA 1Pl entirely all sha-le] de] na-ge renj. kill-Perf C that-Cl person (* ‘This is exactly the manj who the bastardj said that hej would kill all of us.’)

The ungrammaticality of (117) is due to the fact that the resumptive pronoun ta1 ‘he’ is A-bound by the DP na-ge hundan ‘that bastard’ located in an argument position and that ta1 ‘he’ is also A'-bound by the head C-Rel. As a result, there are two dependencies of different natures that have been created for the pronoun he, which is an undesirable situation since it violates the constraint on the bound variable construal. The first one, A-dependency, is in fact an anaphoric dependency between the pronoun ta1 ‘he’ and the DP na-ge hundan ‘that bastard’. This A-dependency is established via ordinary co-reference with the DP that bastard c-commanding

Analyses  

the pronoun ta1 ‘he’. Since this is an anaphoric relationship, the dependency in question is a co-referential dependency, as illustrated in (118). (118) A-dependency * [DP the guyj [CP1 C°1 that [TP1 the bastardj said [CP2 C°2 that [TP2 hej …… ]]]]] co-reference The crucial question concerns the second one, A'-dependency, between the resumptive pronoun ta1 ‘he’ and the DP antecedent na-ge ren ‘the guy’. We wonder whether this dependency can only be established by movement operation. The answer that I suggest in this study is no and I will argue that such a dependency can also be established by Agree, as shown in (119). (119) A'-dependency * [DP the guyj [CP1 C°1 that [TP1 the bastardj said [CP2 C°2 that [TP2 hej …… ]]]]] u-[var] u-[var] i-[var] u-[ϕ] u-[ϕ] u-[ϕ] Agree 2

Agree 1

The resumptive pronoun ta1 ‘he’ bears an unvalued and uninterpretable [ϕ] feature that does not need to be interpreted properly in semantics on the one hand, and a valued and interpretable [var] feature that justifies the variable status of a resumptive pronoun on the other hand. This resumptive pronoun functions as a Goal. The complementizer of the relative clause, the higher C1 head, bears an uninterpretable [ϕ] feature as well as an uninterpretable [var] feature, neither of which has a value. This C1 head works as a Probe that ccommands the Goal, the resumptive pronoun. One must bear in mind that the Probe is not the NP antecedent na-ge ren ‘the guy’ occupying the [Spec, CP] position but the C° head since a Probe must be a functional head (i.e. X° category). Technically, the C-Rel will copy the relevant formal features, such as ϕfeatures, of the NP antecedent na-ge ren ‘the guy’ and then enters into the agreement relation with the resumptive pronoun ta1 ‘he’ to establish the relevant A'-dependency. Even if the NP antecedent shares the same index with the Goal (i.e. the resumptive pronoun), the former cannot function as a Probe. As a result, the A'-dependency can be established via Agree in (119). Since operation Agree takes place phase by phase, the function of the intermediate complementizer C2 becomes crucial. I assume that the intermediate C2 bears two features, [ϕ] and [var] and that both of them are uninterpretable features. Since all of the

  Syntactic derivation: two types of A'-dependencies three elements, C1, C2 and the resumptive pronoun, bear each at least one uninterpretable feature, they remain active at syntax.19 First, the first Agree applies between C2 and the resumptive pronoun ta1 ‘he’. The interpretable [var] feature of the pronoun he provides the uninterpretable [var] feature attached to the intermediate Probe C2 with a value; as a result, this [var] feature will be valued via the operation Feature Valuation. However, the C2 head still remains active thanks to the uninterpretable [ϕ] feature. Then, the second Agree applies between the two C heads, C1 and C2. The uninterpretable [var] feature will be valued under Feature Valuation and the uninterpretable [ϕ] does not play any role here. The point that I want to highlight is that the crossover configuration in an A'-dependency can very well be established by Agree alone without involving any movement strategy. Since the two different types of dependencies illustrated respectively in (118) and in (119) both apply to the same resumptive pronoun, the derivation crashes. Demirdache & Percus (2011) show that movement is still involved in resumptive dependencies in Jordanian Arabic. The contrast exists between two different orders. For instance, in dislocation structures, if the dislocated element is a quantified expression and the order between the clitic pronoun and the epithet is clitic > epithet, the sentence is grammatical (cf. 120). (120)

OK: Q/WH [CP…clitic... epithet...] kull walad ʔumm-oh fakkart

ha-l-ħmar

every boy mother his thought pro-the-donkey bi-l-bajat… at-the-house ‘Every boy that hisj mother thought that this donkeyj is at home’ Jordanian Arabic, Demirdache & Percus (2011:378)

 19 One anonymous reviewer points out that C2 is not a C-Rel and that unlike a C-Rel, there is no operator-variable relation for this C. Instead, it c-selects a clause and thus exhibits a selection dependency with the whole clause, rather than a relative dependency with any variable. In fact, as the reader will see, this C is active here for the pronoun-dependency, because it marks a phase for the pronoun to establish a syntactic relation to a higher C. The reviewer suggests that it is not necessary to assume that C2 also has u-[var] since that if every C that introduces a clause has such an uninterpretable feature, it is not clear how it is checked. I quite agree with the reviewer that the variable feature is not relevant in case of relatives; however, an uninterpretable feature is still needed here to activate Probe-Goal system. In Rouveret (to appear)’s analysis, a similar uninterpretable feature functioning as an activation feature is proposed. I leave the question open here.

Analyses  

By contrast, configurations such as *Q/Wh […epithet…clitic…] is always illicit due to weak crossover effects (cf. 121). (121) a.

b.

* Q/Wh […epithet…clitic…] Weak crossover effect * [kull walad [[ʔum ħa-l-ħmar] fakkart every boy mother pro-the-donkey thought ʔinnu raħ yzittu-u bi-lħabs]] that they.will put.him in-prison (‘Every boy that this donkey’sj mother thought that they will put himj in prison’) Strong crossover effect * miin xabbartu ħa-l-ħmar who you.told pro-the-donkey ʔinnu raħ yzittu-u bi-lħabs ? that they.will put.him in-prison (‘Who did you tell this donkeyj that they will put himj in prison?’) Jordanian Arabic, Demirdache & Percus (2011:378)

The solution that Demirdache & Percus (2011) propose to account for such a contrast is that resumptive pronouns move at LF to the scope position in order to create a binder for the trace left after the movement. For the order Q/WH [CP…cl... epithet...], the raising of the clitic pronoun will not cross the epithet and therefore, no crossover effect is detected and the relevant sentence is grammatical. By contrast, for the order Q/WH [CP… epithet... cl…], the raising of the clitic pronoun will certainly cross the epithet that bears the same index and this movement gives rise to the crossover effects and as a result, the sentence is ungrammatical. However, this contrast is not observed in Chinese shown in the following examples. Both orders, [CP … epithet... pro...] (cf. 122a) and [CP … pro... epithet] (cf. 122b), lead to ungrammatical sentences. (122) a.

* [CP … epithet... pro...] * [那個混蛋 j 自己的老婆把他 j 給殺死了]的每個人 j * [na-ge hundanj-ziji de laopo ba ta1j that-Cl bastard-self DE wife BA 3MSg gei sha-si-le] de mei-ge renj GEI kill-dead-Perf C every-Cl personne (*‘every manj that the bastardj’s own wife killed himj’)

  Syntactic derivation: two types of A'-dependencies b.

* [CP ... pro... epithet] * [ 他 j 自己的老婆把那个混蛋 j 給殺死了]的每個人 j * [ta1j -ziji de laopo ba na-ge hundanj 3MSg-self DE wife BA that-Cl bastard gei sha-si-le] de mei-ge renj GEI kill-dead-Perf C every-Cl person (*‘every manj that hisj own wife killed this bastardj’)

These examples show that in Chinese we do not really have an empirical argument in support of the movement approach and presumably movement is not an option for languages like Chinese. The reasoning in this section shows that crossover effects are no longer a diagnostic test exclusively reserved for movement. Operation Agree also displays crossover effects (weak and strong). Till now, I propose a unified analysis based on Agree for languages like Welsh. Chinese behaves like Welsh in that resumptive relatives can always be derived by Agree without necessarily involving Move. Agree is also subject to locality constraints and gives rise to crossover effects. In other words, island effects and crossover effects are no longer diagnostic tests for movement exclusively because Agree is also subject to the same condition. .... Resumptive dislocation structures Recall that we have dealt with two kinds of dislocation structures, LeftDislocation (LD) structures and ex-situ cleft-focus structures. The distribution of gaps and of resumptive pronouns is the same in both structures and therefore, I will discuss the derivation of “dislocation structures” as a general without making a distinction between them here.20 In dislocation structures, if it is the gap strategy that is adopted, the derivation is the same for relative clauses (with a gap or with a resumptive pronoun). Concretely, operation Agree applies to establish the relevant dependency in a dislocation structure, which explains why island effects and crossover effects are always detected. Let us take the derivation of LD-structures for example, (123) * [TopP NPj , [TP [CP-island … … _____ j … … ]]]

 20 More detailed analysis of the ex-situ cleft focus structures can be found in the Appendix.

Analyses  

In this type of LD-structure, the NP in the topic position is normally derived by movement in the literature on Information Structure and in the work on the leftperiphery (Rizzi 1997, 2004, Cinque 1999, Pan 2011a, b). However, I prefer to keep Agree as the only derivational mechanism without the need of Move when island and crossover effects are detected. This preference is based on two considerations. First, as I explained in great detail in the previous sections, Move in the Minimalist Program is defined very differently from movement in the Government and Binding theory. The second reason for this preference is related to the economy principle. It is always costly to derive by movement or by Move than to derive by Agree, irrespective of whether the movement is realized at narrow syntax or at LF. If an A'-dependency can be established without involving any movement operation, for instance via Agree, this non-movement approach should always be preferred by the computational system. Based on these considerations, I assume that the dislocated NP in an LD-structure is always base-generated in-situ in the topic position.21 Agree applies here to establish a dependency between the TopP and the gap that is traditionally translated as a bound variable. Since Agree is subject to locality conditions, island effects will be detected. However, the situation is completely different for the dislocation structures containing a resumptive pronoun (or an intrusive pronoun). Neither island effects nor crossover effects are detected in this case, which seems to suggest that Agree cannot apply to resumptive dislocation structures. The task now is to determine which one is the most appropriate mechanism for establishing such an A'-dependency in resumptive dislocation structures, as shown in (124). (124)

[TopP NP1, [TP [CP-island … … … intrusive pronoun1… ]]]

 21 In Chapter 3, I will show that in the case of dislocation structures with a gap, the topic NP can also undergo movement to [Spec, TopP] to satisfy the requirement of the EPP feature; however, whether such a movement takes place is determined by operation Agree. It is worthwhile highlighting that this is the only case where Agree can be followed by Move because the only reason for activating Move is to satisfy the EPP feature. By contrast, in the case of resumptive dislocation structures, the NP topic is always base-generated in the [Spec, TopP] position and, as the reader will see, it is operation Match that applies here. Since the topic NP is always in TopP, EPP is always satisfied and that is the reason why Move is never activated in this case. The result of the tests in Chapter 3 shows that such a distinction is linked to two different forms of the internal structures of the element (gaps or resumptive pronouns) on the dislocated site. The extended form containing an NP is always a potential candidate for movement; by contrast, the reduced form lacking the NP part cannot undergo movement.

  Syntactic derivation: two types of A'-dependencies Generally speaking, there are two ways to establish such a dependency between TopP and the resumptive pronoun. Either this dependency is a mere semantic linking that is established at the discourse level or it is a syntactic dependency as with all of the other types of A'-dependencies that I have examined. If it is the first choice that is made, the dependency in question only reflects a discourse relation that is naturally not subject to any syntactic constraint such as locality conditions, which explains of course the absence of island effects in resumptive dislocation structures. If it is the second choice that is adopted, such a dependency will be analyzed as syntactic linking. Then, what is needed here is a syntactic mechanism that not only allows us to establish such a dependency but also gives rise to neither island effect nor crossover effect. There is a potential candidate that can satisfy these double requirements: operation Match. Recall that Matching is defined as a kind of relation that is identified between a potential Probe and a potential Goal. Match does not require absolute identical relation between the Probe and the Goal but only a kind of nondistinctive relation. Such a non-distinction only requires that there are identical attributes of the relevant features attached to the Probe and attached to the Goal and that their values can be different. My proposal is that in dislocation structures containing a resumptive pronoun, the dependency between a resumptive pronoun and TopP can be established by Match without involving Agree. In fact, an analysis based on Match has been proposed by Boeckx (2003) to account for resumptive relatives in Irish. An asymmetry is observed between two kinds of A'-chains. (i)

A'-chains constructed by Agree22 are subject to island conditions because agreement relations are crucial for defining the extraction domains (i.e. strong islands);

 22 It is very important to make a distinction between the analysis proposed by Boeckx and the analysis presented in this section concerning operation Agree. For Boeckx, Agree means agreement in terms of ϕ-features in the strict sense. And it is precisely for this reason, in contrast to the analysis only based on Agree proposed by Adger & Ramchand (2001, 2005), Boeckx proposes another analysis based on Match + Move to establish resumptive chains for Scottish Gaelic. Examples in (i) show that there is no agreement in terms of the entire set of ϕ-features in Scottish Gaelic. (i) a. Siud a' chaileag a dh'eisd thu ris that the girl C-Rel listen-Past you to-3MSg ‘That's the girl that you listened to’

Analyses  

(ii)

Resumptive chains constructed by Match do not require any agreement and as a result, they do not give rise to any island effect.

In Irish, if it is a gap that is embedded within an island in a relative clause, island effects can be observed (cf. 125a); by contrast, if it is a an intrusive pronoun that is embedded within the same type of island context, island effects disappear, as shown in (125b). For Boeckx, it is exactly the operation Match that applies to (125b) and Match is not followed by Agree. (125) a.

* An

fear

a

phog



an

bhean

the man aL kissed I the woman phos ___ married [ __ ] ‘the man that I kissed the woman that married’ b.

An

fear

a

phog



an

bhean

a aL

a

the man aL kissed I the woman aL phos é married him ‘the man that I kissed the woman that married him’ Irish, cited by Boeckx (2003:69) from Sells (1984:200-201) A very important question that we need to ask is why an A'-dependency established by Match gives neither island effects nor crossover effects. The following example shows that a dislocation structure containing a resumptive pronoun in Chinese does not give rise to crossover effects even if the crossed element is an epithet. From this perspective, a resumptive dislocation chain in Chinese behaves like a resumptive relative chain in Irish with regard to island effects.

 b. * Siud that c. * Siud that

a' chaileag a dh'eisd thu rithe the girl C-Rel listen-Past you to-3FSg a' chaileag a dh'eisd thu ri___ the girl C-Rel listen-Past you to[___] Scottish Gaelic, Adger & Ramchand (2001) Concretely, Boeckx claims that a resumptive chain was derived by stranding the nominal head of an NP after A'-movement. In this sense, A'-movement is not motivated by ϕ-feature checking and the link between a resumptive pronoun and its antecedent is established by Match. Match is followed by Move and thus, in contrast to Agree, Match is considered as a pre-condition on Move.

  Syntactic derivation: two types of A'-dependencies (126)

我兒子 j 啊,[那小子 j 說[他 j 再也不敢酒後駕駛了]]。 Wo erzij a, [nei xiaozij shuo 1Sg-(Gen) son Top that kid say zai ye bu gan jiu-hou jiashi again also Neg dare alcohol-after drive ‘My sonj, the kidj said that hej would never dare again drinking.’

[ta1j 3MSg le]]. SFP to drive after

As I explained in Section 2.2, the essential point in a crossover configuration (strong crossover in the sentence above) is that such a configuration violates the constraint that prohibits a variable to be A-bound and A'-bound simultaneously. Let us examine the two different dependencies involved in (126). The anaphoric A-dependency between the pronoun ta1 ‘he’ and the epithet na xiaozi ‘the kid’ is established via co-reference and this anaphoric dependency is in fact a co-referential dependency, as illustrated in (127). (127) A-dependency [TopP My sonj Top°, [TP1 the kidj said

[CP C° that

[TP2 hej …… ]]]]

co-reference The A'-dependency between the pronoun ta1 ‘he’ and the C-Top (hosting wo erzi ‘my son’) was established via Match. Match only aims to check whether there is a non-distinctive relation between the features attached to the Probe and those attached to the Goal without examining whether these features are valued or not. In (128), all of the three elements, Top°, C° and the resumptive pronoun, bear the same features, i.e. [var] and [ϕ]. Once the feature identification ends, operation Match is finished. As a matter of fact, the relation based on Agree is not established between the C-Top and the resumptive pronoun he in a strict sense because the uninterpretable features attached to these two elements are not deleted (cf. 128). Recall that only the valued features can be deleted and only Agree can value features. (128) A'-dependency [TopP My sonj Top° , [TP1 the kidj said [CP C° that [TP2 hej …… ]]]] [var] [var] [var] [ϕ] [ϕ] [ϕ] Match 2 Match 1

Analyses  

Let me summarize. In sentence (126), A-dependency is well established, however, A'-dependency is not established by Agree. As the reader will see, only an A'dependency established by Agree gives rise to crossover effects. Therefore, the resumptive pronoun ta1 ‘he’ cannot be A-bound and A'-bound simultaneously in the strict sense. As a result, the constraint on the bound variable construal is not violated and therefore, no crossover effect is detected. Till now, my reasoning explains the absence of crossover effects in a matching chain. However, we still need to explain why the two elements, the topic phrase in TopP and the resumptive pronoun share the same index. Normally, if they share the same index, then they form a dependency. One possibility is to argue that in the case of topic, the dislocated NP and the epithet are assigned the same referent by the discourse and both of them form an equivalent of a kind of referential dependency instead of an operator-variable binding type of A'-dependency. In the next section, I will give empirical arguments to show that an operator-variable binding relationship never exists in resumptive dislocation structures in Chinese, which is a fundamental difference between resumptive relatives and resumptive dislocation structures. Along this line, the link between the epithet and the pronoun is established between two nominals, thus it is also a referential dependency. By contrast, the complementizer C° does not bind the resumptive pronoun as a variable and therefore, the A'-dependency cannot be established by Agree despite the fact that these three elements, NP-Top wo erzi ‘my son’, the epithet na xiaozi ‘the kid’ and the resumptive pronoun ta1 ‘he’, share the same index. A'dependencies in this situation are precisely established by Match alone without being followed by Agree. Match establishes an A'-dependency when the Probe is not necessarily interpreted as an operator and the Goal is not necessarily interpreted as a bound variable. However, the Probe and the Goal still share the same index once the A'-dependency is established by Match. This analysis, fundamentally different from previous Agree-based analyses, allows that Match functions alone as an independent syntactic mechanism that is capable of establishing an A'-chain. Crucially, as will be detailed, Match is not any more considered as a pre-condition on Agree or as a pre-condition on Move (contra Boeckx 2003). A matching chain is not sensitive to island boundaries nor to crossover configurations, contrary to an A'-chain established by Agree. As the reader will see in the next section, I will come back to the point that only the A'dependencies established via Agree give rise to crossover effects. .... Differences between relatives and dislocation structures When comparing the derivation of a relative clause with that of a dislocation structure, there is an extremely important problem, but for which it is technical-

  Syntactic derivation: two types of A'-dependencies ly difficult to find a satisfactory solution. Let us leave aside the cases involving gaps and only concentrate on resumptive pronouns in relatives and in dislocation structures. If I argue for the idea that it is Agree alone without Move that derives relatives and it is Match alone without Agree that derives dislocation structures, then the question is: What prevents Agree from applying to resumptive dislocation structures? Recall that in general, Agree follows Match and Match could be considered as a pre-condition for Agree, which is the case for relatives. What is the factor that avoids continuing the derivation with Agree once Match finishes feature identification? For Agree to work, it is necessary to have uninterpretable and unvalued features attached to the Probe and to have the same set of interpretable and valued features attached to the Goal. If Agree does not apply, then this leads us to think that all of these features are uninterpretable features both on the Probe and on the Goal and that being uninterpretable features, they do not contribute any semantic interpretation, such as the operator-variable type of relation, to the sentence. Another way to explain this assumption is that uninterpretable features are never activated in an established A'-chain. Once a relevant A'-dependency is established, for instance by Match, the unchecked uninterpretable features no longer play crucial roles at interfaces and in this sense, they are no longer activated. Again, I need to emphasize that for this technique to work, an important condition is to treat Match as an independent syntactic operation that is capable of establishing an A'-dependency alone without being followed necessarily by Agree. In other words, Match is no longer considered as a pre-condition on Agree. In this section, I will pursue this point of view by providing an empirical argument from Chinese. I will argue that the relationship between the C-Top and the resumptive pronoun in dislocation structures is not an operatorvariable relation in the strict sense and such a relationship is very different from the one existing between the C-Rel and the resumptive pronoun in relatives. One observation on Chinese is that the complementizer C is morphologically marked in two different ways in dislocation structures and in relatives. I will try to argue that C-Rel and C-Top bear exactly the same features but with different interpretabilities. In Chinese, in a dislocation structure or in a topicalization structure, it is socalled topic markers (TM), such as the particles ne, ba or a, that occupy the Top° position (Paul 2015, Pan 2011a, 2014, Pan & Paul 2016). The head of the TopP is considered as the equivalent of the C of CP. For instance, in (129), the particles a and ne are placed under the head position of the relevant TopP. Topic phrases are located in the specifier position of TopP and the topic head takes a TP as complement.

Analyses  

(129) a.

b.

張三 j 啊,我覺得(他 j)很風趣。 [TopP Zhangsanj [Top° a], [TP wo juede Zhangsan Top 1Sg think fengqu]]. funny ‘As for Zhangsanj, I think that hej is funny.’

(ta1j) 3MSg

hen very

這幅油畫 j 呢,(它 j)被轉手了好幾次了。 [TopP Zhe-fu youhuaj [Top° ne], [TP (ta3j) bei this-Cl painting Top 3OSg Passive zhuanshou-le hao ji-ci le]]. resell-Perf good several.times SFP ‘As for this paintingj, itj has been resold several times.’

In relatives, I will follow the traditional idea that it is the structural particle de that occupies the position C°. (130)

我昨天買的那本書 [NP [CP [TP wo zuotian

mai]

1Sg yesterday buy ‘the book that I bought yesterday’

[C° de]

na-ben

shu]]

C

that-Cl

book

The two complementizers, C-Rel and C-Top, are morphologically distinguished from one another in Chinese. Recall that in relatives, Rouveret (2002) proposes that it is the [Rel] feature that is attached to the complementizer head in relative clauses. As for the feature with the similar function, Adger & Ramchand (2005) suggest that it is the [λ] feature that is associated with C in relatives. These two variations of the same type of feature are proposed to account for the fact that a relative clause is translated as a predicate in semantics and the relevant operation is λ-abstraction. In Adger & Ramchand’s system, the [λ] feature is the λabstractor. The crucial point in this analysis is the [λ] feature can create a relation of predication in relative clauses and such a relation is identical to the one existing between an operator and its bound variable. Under such a consideration, the C [λ]/[Rel] binds a resumptive pronoun as a variable. This line of reasoning is also supported by the analysis of Rouveret (to appear) where the [var] feature is used to justify the variable status of resumptive pronouns. The core spirit of the analysis of Adger & Ramchand (2005) and that of Rouveret (to appear) show that relative clauses involve predication relationship where C-Rel

  Syntactic derivation: two types of A'-dependencies and a resumptive pronoun form an A'-chain and that C-Rel is translated as an operator and the resumptive pronoun as a variable in semantics. Based on these two analyses, it is possible to propose that in dislocation structures, C-Top (i.e. Top°) and the resumptive pronoun both bear the [λ] feature but [λ] is an uninterpretable and unvalued feature both on C-Top and on the resumptive pronoun. Another possibility is to say that these two elements bear both a [var] feature and similarly, [var] is an unvalued and uninterpretable feature. Since the two features associated with the Probe and with the Goal are both uninterpretable and unvalued features, the Probe and the Goal are not legitimated in an Agree relation because there is no interpretable feature that can value the uninterpretable feature attached to the Probe. As a result, after Match the derivation stops and Agree cannot apply here. It is not important that there are still unvalued uninterpretable features associated with the Probe and with the Goal since being an independent operation, Match does not look into the value of the relevant features and it can establish an A'-dependency alone by exclusively identifying the attribute of these features. One can assume that in this case, the value of these features simply do not participate into the derivation process. Once the relevant A'-dependency is established by Match, unchecked uninterpretable features no longer play crucial roles at interfaces since they do not provide any semantic interpretation to the dependency. From this perspective, unchecked uninterpretable features are no longer activated once the relevant A'-dependency has already been established. This is the reason why an A'-dependency established by Match is different from an A'-dependency established by Agree both in syntax and in semantics. Concretely, the former is not subject to locality conditions and therefore does not give rise to any island effect or crossover effect; however, the latter displays these effects. Semantically, as will be detailed immediately below, the former does not necessarily imply any operator-variable relation but the latter does. I will present an empirical argument in support of the idea that the relationship between C-Top and a resumptive pronoun is different from the one between C-Rel and a resumptive pronoun. In Chinese, nominal appositive structures are very frequently used and they can be placed either in subject position or in direct object position. (131) a.

[張三這個學生]非常用功。 [Zhangsan zhe-ge xuesheng]

feichang

Zhangsan that-Cl student very ‘Zhangsan(,) this student, works very hard.’

yonggong. hardworking

Analyses  

b.

我非常喜歡[空乘這個行業]。 Wo feichang xihuan [kongcheng

zhe-ge

hangye].

1Sg very like steward this-Cl ‘I like the profession (working) as flight attendant.’

profession

In (131a, b), structures such as Zhangsan zhe-ge xuesheng ‘Zhangsan this student’ and kongcheng zhe-ge hangye ‘the profession (as) flight attendant’ each form an independent constituent. They are nominal appositive structures. It is quite possible to topicalize the two appositive NPs as a single one. (132)

[張三這個學生], 我非常欣賞。 [Zhangsan zhe-ge xuesheng]j,

wo

feichang

Zhangsan this-Cl student 1Sg very xinshang tj. appreciate ‘Zhangsan this studentj, I appreciate himj very much.’ In dislocation structures, Zhangsan can undergo topicalization alone by leaving its appositive this student in-situ. In this case, these two NPs are always co-referential and therefore they share the same index. (133)

張三 j 啊, 我非常欣賞這個學生 j 。 Zhangsanj a, wo feichang

xinshang

zhe-ge

Zhangsan Top 1Sg very appreciate this-Cl xueshengj. student ‘As for Zhangsanj, I appreciate this studentj very much.’ However, in relatives, it is impossible to relativize one of the two appositive NPs by leaving the other one on the relativized site as resumptive, as shown in (134). (134)

* [空乘 j 很危險]的這個行業 j * [Kongchengj hen weixian]

de

zhe-ge

hangyej

steward very dangerous C this-Cl profession (‘this professionj that (working as flight attendantj) is dangerous’) Following the analyses of Rouveret (2002) and Adger & Ramchand (2005), an operator-variable relation is involved in a relative clause; however, in the present case it is difficult to maintain the idea that the relation between zhe-ge

  Syntactic derivation: two types of A'-dependencies hangye ‘this profession’ and kongcheng ‘flight attendant’ as an operator-variable type of relation because the semantic interpretation of the NP kongcheng ‘flight attendant’ does not depend on the NP zhe-ge hangye ‘this profession’. Since in relatives the relationship between C-Rel and a resumptive pronoun is an operator-variable relation and the pair zhe-ge hangye ‘this profession’ and kongcheng ‘flight attendant’ does not construct as an operator-variable pair, the latter cannot exist in relatives. By contrast, this pair is tolerated in dislocation structures, which seems to suggest that dislocation structures do not imply an operatorvariable relation but a kind of co-referential relation between two simple NPs. To summarize, this contrast based on empirical observations suggests that the relation between C-Top and a resumptive pronoun does not imply any operatorvariable relation and from this perspective, C-Top is different from C-Rel. Based on the idea that the in-situ appositive NP in (133) is not a variable bound by the TopP, it is not necessary either for a resumptive pronoun in dislocation structures to be interpreted as a bound variable in semantics and as a result, such a resumptive pronoun can bear a [var] feature but it is an uninterpretable (i.e. unvalued) feature. Meanwhile, C-Top can also bear a [var] feature and it is also an uninterpretable and unvalued feature. This situation creates a problem for the application of Agree because the relation between a Probe and a Goal will not be established in this case. When both the Probe and the Goal bear the same set of features and both of them are uninterpretable and unvalued, the dependency cannot be established by Agree since there is no available interpretable features which can value and delete the uninterpretable [var] feature attached to the Probe. Another scenario based on the analysis of Adger & Ramchand (2005) will lead to the same conclusion. If C-Top does not bind resumptive pronouns as variables in semantics, it means that this Top° does not render its complement TP predicative. In this sense, C-Top can possibly bear a predicative [λ] feature but it is an uninterpretable and unvalued feature. As a potential candidate for Probe, C-Top cannot find a suitable Goal that bears the same [λ] feature which is an interpretable feature with a value. The eventually most suitable candidate for the Goal is the resumptive pronoun; however, the pronoun bears a [λ] feature that is also uninterpretable and unvalued. As a result, the dependency between a Probe and a Goal will not be able to be established between C-Top and the resumptive pronoun by Agree. In both scenarios, there is no available Goal that bears the same type of interpretable feature, either [var] or [λ], to value the uninterpretable and unvalued features attached to the Probe. This is the reason why after Match, Agree cannot continue to establish the A'-dependency (cf. 135).

Analyses  

(135) Dislocation structures [TopP NPj [Top° Top°] , either u-[var] or u-[λ]

[TP … … … … … RPj …… ]] u-[var] u-[λ] √ Match *Agree (u: uninterpretable feature)

(136) Relatives [NP NPj [CP [C° C-Rel] either u-[var] or i-[λ]

[TP … …… … … RPj …… ]]] i-[var] u-[λ] √ Match √Agree (u: uninterpretable feature; i: interpretable feature)

The derivational differences between a resumptive dislocation structure and a resumptive relative clause that I propose is illustrated in (135) and in (136). Let us concentrate on relatives here. In relatives, if I adopt the hypothesis of Adger and Ramchand (2005), the relation of Probe-Goal can be established between CRel and a resumptive pronoun. C-Rel bears the feature [λ] that is going to be interpreted as λ-operator/abstractor in semantics and therefore [λ] is an interpretable feature. A resumptive pronoun also bears the [λ] feature but it is an uninterpretable feature and as a result, it cannot function as an active Goal. The interpretable features attached to C-Rel will be able to value the uninterpretable features attached to resumptive pronouns. After Match, Agree continues to apply to check the relevant features on the Probe and on the Goal.23 If I adopt Rouveret (to appear)’s analysis, in relatives a resumptive pronoun will be translated as a variable in semantics and it will bear an interpretable and valued [var] feature. The potential candidate for the Probe is C-Rel that bears an uninterpretable and unvalued [var] feature. In this sense, the Probe is active for the derivation by Agree. The operator-variable relation will be established between C-Rel and a resumptive pronoun. The unvalued uninterpretable features

 23 Recall that in the system based on Agree proposed by Adger & Ramchand (2005), no ccommand relation is required between the Probe and the Goal. In Rouveret (to appear)’s analysis, the c-command requirement is explicitly stated. The uninterpretable features that are unvalued will c-command the interpretable features that already have a value attached to the Goal.

  Syntactic derivation: two types of A'-dependencies attached to the Probe c-command the valued interpretable features attached to the Goal. After Match, Agree applies. The interpretable [var] feature on the resumptive pronoun will value the uninterpretable [var] feature on C-Rel. All of these features will be checked and deleted after feature valuation. .... Locality Another difference between an A'-dependency derived by Agree and that derived by Match concerns locality conditions. I have shown earlier that in Chinese (i) relatives with a gap and with a resumptive pronoun and dislocation structures with a gap display island effects and (ii) resumptive dislocation structures do not give rise to any island effect. The distinction between these two cases is that structures in (i) are derived by Agree and structures in (ii) by Match. In this section, I will address their differences in terms of locality conditions. In the Minimalist Program, there are two types of locality conditions: the shortest link condition and the condition on phases. The shortest link condition covers two of the most often studied phenomena in the Government and Binding theory: island effects and intervention effects. The condition on phases requires that the derivation must be undergone phase by phase in a strict cyclic fashion. (137) [Phase 3 …… [Phase 2 …… [Phase 1 … …]]] In the Phase theory, vP and CP are strong phases; v° and C° are phasal heads. The domain of a phase consists of the complement of the phasal head. The edge of a phase contains the specifier and the adjuncts of a phasal head. In the present case, we have three phases: lower one (Phase 1), intermediate one (Phase 2) and higher one (Phase 3). After the next phasal head (for instance, Phase 2) is generated, the domain of the lower phase (i.e. Phase 1) will be sent to the interfaces via operation Transfer to get an interpretation. Once sent, the phasal domain becomes inaccessible. Only the edge of Phase 1 is accessible for further eventual operations. And this condition is called “Phase Impenetrability Condition (PIC)”. (138) Phase Impenetrability Condition (Chomsky 2000) In phase α with head H, the domain of H is not accessible to operations outside α, but only H and its edge. An important implication of this condition is that phases define impenetrable domains to movement. If XP is dominated by a complement of a phase α, XP cannot move out of α.

Analyses  

After the phase of a lower cycle is sent to the interfaces, it becomes inaccessible and the derivation continues to construct a higher phase. In the schema above, once Phase 1 is sent to the interfaces, one cannot go back to Phase 1 and must continue to construct Phase 2. Similarly, when Phase 2 is sent to the interfaces, it becomes inaccessible to any further operation and we must continue to construct Phase 3. In this sense, as will be detailed later, example in (137) illustrates the case where Multiple Transfers and Multiple Spell-Outs apply. As one can see, both the shortest link condition and condition on phase can account for locality constraints such as island effects and intervention effects. It seems that both of these conditions achieve the same result. Moreover, the condition on phase constraints the construction of phases and it is strong enough to filter all of the derivations that violate locality constraints. Along this line, the shortest link condition becomes redundant. The point of view that I will argue for in this study is that condition on phase is the only necessary condition that we need to constraint locality. The relevant intervention effects and island effects can be filtered by condition on phase alone without the need for the shortest link condition. Let us examine how the condition on phase applies to the derivations by Agree and by Match respectively. (i) Derivation by Agree Let us begin with the derivation by Agree. The step-by-step illustration of the derivation by Agree is presented in (139). In the lowest phase cycle, Match applies first. The Probe and the Goal of Phase 1 bear the same type of feature. Only the attribute of the feature (i.e. person, number and gender) but not the value of the feature is relevant to Match. If the attribute of the feature is identical for both the potential candidates for the Probe and for the Goal, a non-distinctive relation will be established between these two items. The Goal bears interpretable features and the Probe bears the same features but uninterpretable; in other words, the feature attached to the Probe and the one attached to the Goal do not possess the same value, which is indeed a necessary pre-condition on Agree. As a result, Agree can apply in this situation. Feature checking is realized under Feature Valuation. The interpretable value of the Goal will be copied on the Probe and once the feature valuation and checking process is finished, the derivation of Phase 1 by Agree ends. All of the relevant features will be deleted. Then, when the head of the next phase, Phase 2, is constructed, Phase 1 will be

  Syntactic derivation: two types of A'-dependencies sent to the interfaces by operation Transfer.24 Once this Transfer process is over, Phase 1 will be inaccessible and no operation will be possible any more inside Phase 1. During this derivation, Multiple Transfers and Multiple Spell-outs apply. Crucially, Agree is the only necessary condition on Transfer of a phase and Match is not. A phase can be transferred to interfaces for interpretation if it is constructed by Agree. (139) Derivation by Agree Step 1 :

[Phase 3 …… u-[F]

[Phase 2 …… [Phase 1 …… u-[F] u-[F]

]]] i-[F]

√ Match √ Agree SPELL-OUT 1

Step 2 :

[Phase 3 …… u-[F]

Transfer [Phase 2 …… u-[F]

[Phase 1 …… i-[F]

]]] i-[F]

√ Match √ Agree

SPELL-OUT 2

Step 3 :

[Phase 3 …… u-[F]

Transfer [Phase 2 …… i-[F]

SPELL-OUT 1 Transfer [Phase 1 …… i-[F]

]]] i-[F]

√ Match √ Agree

 24 In the literature on phases, a well-constructed phase can be sent to the interfaces at different moment. It can be sent immediately after it is constructed or it can only be sent after the head of the next higher phase is generated.

Analyses  

SPELL-OUT 3

Step 4 :

Transfer [Phase 3…… i-[F]

SPELL-OUT 2 Transfer [Phase 2 …… i-[F]

SPELL-OUT 1 Transfer [Phase 1 …… i-[F]

]]] i-[F]

(u-[F]: uninterpretable feature; i-[F]: interpretable feature) We can imagine that if Phase 1 contains an island, Agree will not be able to function between Phase 1 and Phase 2. It is precisely in this way that condition on phases filters strong islands. Step 2 contains the construction of Phase 2 under Agree. Once Phase 2 is sent to the interfaces, the construction of Phase 3 begins. Let us examine a concrete example, (140)

Complex-NP island (relatives) * 這是[我見到過[談論過他 j /____j 的] 那個女同學的]作家 j。 * Zhe shi [wo jiandao-guo [tanlun-guo ta1 / ____j this be 1Sg meet-Exp talk-Exp 3MSg de] na-ge nütongxue de] zuojiaj. C that-Cl female.student C writer (‘This is the writerj [whom I met the student [who talked about (him j )]].’)

Example (140) shows that a relative clause either with a resumptive pronoun or with a gap gives rise to island effects. The lowest phase, represented by Step 1, contains the inner relative clause, “the student who…” C-Rel and the resumptive pronoun him form a potential Probe-Goal configuration because both of them bear the variable [var] feature and the [ϕ] feature. Since the relevant A'dependency is a relative clause, the C-Rel and the resumptive pronoun located on the relativized site construct an operator-variable pair. At this phasal level, the resumptive pronoun will be interpreted as a bound variable at LF and therefore it bears a valued and interpretable [var] feature. It will value the uninterpretable [var] feature attached to the C-Rel. Therefore, an Agree chain will be established between these two and the relevant features will be valued and checked. As a result, the lowest phase is well established. One should bear in mind that relative clauses define strong islands and accordingly, the lower phase here is no longer transparent for any further Agree-based operation.

  Syntactic derivation: two types of A'-dependencies Step 1:

[CP1 who [C°] talked about (him j) ]] u-[var] i-[var] u-[ϕ] u-[ϕ] √ Match √ Agree

Step 2 constructs a higher phase that contains the outer relative clause, the writer whom… Once the phasal head of the second phase, [C0 whom], is constructed, Phase 1 [CP1who talked about (him j)] is sent to the interfaces by operation Transfer for interpretation and it will become inaccessible. The head of the Phase 2, [C0 whom], is a potential Probe that bears an uninterpretable [var] feature and an uninterpretable ϕ-feature. However, there is no available candidate for the potential Goal. Therefore, the derivation crashes. Step 2: [CP2 [C° whom] I met the student [CP1who talked about (him j)] ] u-[var] u-[ϕ] * Match * Agree

…..

The same derivation applies to dislocation structures with a gap and in this case, gap is still interpreted as a bound variable. (ii) Derivation by Match Before entering into the details of the derivation by Match, it is very important to point out one difference between Agree and Match. The fact that Agree is subject to locality conditions can be explained by the way how phases are constructed. Agree applies phase by phase and this derivation is realized strictly at the Narrow Syntax level. After the construction of each phase, the wellestablished phase will be sent immediately to the interfaces. Once the phase is transferred, it becomes inaccessible and the derivation of the next phase begins. From this point of view, in order for Match not to be subject to locality conditions, we must find a technical way to make sure that Match applies to the entire A'-dependency after the derivation of all of the phases is finished. In this case, the derivation mechanism must allow Match to be able to operate on the entire structure containing all of the well-constructed phases. In other words, such a mechanism must prevent each well-constructed phase from being sent immediately to the interfaces. We can imagine that after the construction of

Analyses  

each phase, it is not immediately transferred to the interfaces and the derivation continues to construct the next higher phase. Once all of the phases are well established, which means that the relevant structure is completely constructed, Match applies to the entire structure and it is at this moment that all of the phases are sent to the interfaces for interpretation at the same time, which ensures that island effects are not produced. Following this reasoning, we should admit that both Agree and Match apply precisely at the level of Narrow Syntax. Before the derivation is finished, only those constructed phases based on an agreement relation can be transferred to the interfaces. For instance, if the potential Goal and Probe of a phase bear the features that possess the same attribute but both features are unvalued and uninterpretable, the relation based on Matching can be established. However, the relation based on Agree cannot be established between the Probe and the Goal and as a result, the constructed phase will not be immediately transferred to the interfaces. An A'-dependency constructed by Match will not be sent to the interfaces until the derivation of the whole structure containing all of the phases is finished. Importantly, the cases where only Match applies are those in which Agree is blocked and the derivation cannot continue with Agree. From this perspective, Match applies strictly at Narrow Syntax and there is absolutely no reason to resort to an interface condition on a Matching chain. In other words, a resumptive dependency derived by Match is exclusively established at Narrow Syntax level and before Spell-Out. The derivation of Chinese resumptive dislocation structures can be regarded as an argument in favor of this view. As I explained earlier, my conclusion based on Chinese data echoes Rouveret (2008, 2011, to appear)’s view that resumptivity is after all not an interface phenomenon. (141) illustrates the general picture of a derivation by Match. First of all, in the lowest phase, Phase 1, the potential Probe and the potential Goal bear the same type of feature. Only the attribute of the feature is concerned in the derivation by Match. If the attribute of the feature is identical between the Probe and the Goal, a non-distinctive relation will be established between these two. The only problem is that the Probe and the Goal bear both uninterpretable features, which blocks the derivation by Agree. As a result, feature checking will not apply and the dependency will not be established by Agree for Phase 1. In this sense, Phase 1 that contains an A'-dependency established by Match will not be sent to the interfaces. Then, the derivation of Phase 2 begins, and so on. Once all of the phases are constructed and the derivation of the entire structure is finished, Match applies to the entire sentence containing all of the wellestablished phases. At the end, all of the phases will be sent to the interfaces by

  Syntactic derivation: two types of A'-dependencies Transfer at the same time (cf. Step 4). In this case, Multiple Transfers and Multiple Spell-outs do not apply. (141) Derivation by Match Step 1 : [Phase 3 …… [Phase 2 …… u-[F] u-[F]

[Phase 1 …… u-[F]

]]] u-[F]

√ Match * Agree

Step 2 :

[Phase 2 …… [Phase 1 …… u-[F] u-[F]

[Phase 3 …… u-[F]

]]] u-[F]

√ Match * Agree

Step 3 :

[Phase 3 …… u-[F]

[Phase 2 …… [Phase 1 …… u-[F] u-[F]

]]] u-[F]

√ Match * Agree

SPELL-OUT Transfer Step 4 : [Phase 3 …… u-[F]

[Phase 2 …… u-[F]

[Phase 1 …… ]]] u-[F] u-[F]

√ Match * Agree  (all of the phases are sent to the interfaces at the same time) Here is a concrete example,

Analyses  

(142)

那位法國影星 j,我碰到了[小倩認識[擁抱過*(他 j)] 的]那位女生。 Na-wei Faguo yingxingj, wo pengdao-le that-Cl French movie.star 1Sg [Xiaoqian renshi [yongbao-guo Xiaoqian know embrace-Exp de na-wei nüsheng. C that-Cl female.student

meet-Perf *(ta1j)]] 3MSg

‘As for that French movie starj, I met the female student [that Xiaoqian knows [who embraced himj]].’ For Step 1, the lowest phase contains the relative clause, who embraced him. CRel-who and the resumptive pronoun him located at the direct object position form a potential Probe-Goal pair, both of which bear the [ϕ] feature and the [var] feature. However, the dependency concerned here is that of a dislocation structure, the highest C-Top and the resumptive pronoun occupying the dislocated site do not construct an operator-variable pair. Recall that in dislocation structures, a resumptive pronoun is not interpreted as a bound variable at LF and therefore its [var] feature should be an uninterpretable feature. As for the [ϕ] feature, it should also be an uninterpretable feature associated with C-Rel-who and with the resumptive pronoun. C-Rel-who and the resumptive pronoun have the same set of features, [ϕ] and [var], which means that the attributes of these features are the same, and therefore, a matching chain can be established between C-Rel-who and the resumptive pronoun. After Match, the feature valuation under Agree is supposed to begin; however, since both C-Rel and the resumptive pronoun bear unvalued and uninterpretable [ϕ] and [var], the uninterpretable and unvalued features attached to the resumptive pronoun cannot value or check the uninterpretable features attached to C-Rel-who. Thus there is no suitable candidate for acting as a Goal that bears at least an interpretable feature that can eventually value the uninterpretable feature attached to the Probe. Therefore, the dependency cannot be established by Agree between C-Rel-who and the resumptive pronoun in this lowest phase cycle. Since the relevant features can neither be checked nor deleted, the lowest phase cannot be sent immediately to the interfaces by Transfer, and the derivation continues with the next phase cycle. The following schema illustrates Step 1 of the derivation.

  Syntactic derivation: two types of A'-dependencies Step 1:

[CP1 who [C0] embraced (him j)]] u-[var] u-[var] u-[ϕ] u-[ϕ] √ Match * Agree

In Step 2, the intermediate phase contains the relative clause, (the student) that Xiaoqian knows…, headed by C-Rel-that. In this phase cycle, C-Rel-that is a potential Probe that bears unvalued uninterpretable features and the lower phasal head C-Rel-who can potentially function as a Goal. The Matching chain can be established between these two heads since the attributes of the relevant features attached to each of them are the same. However, Agree cannot work at this stage since both the Probe and the Goal bear uninterpretable features that cannot be checked. With unvalued and unchecked features, this intermediate Phase cannot be transferred to the interfaces either and the derivation continues. Step 2:

[CP2 [C0 that] Xiaoqian knows [CP1who [C0] embraced (him j)]] u-[var] u-[var] u-[var] u-[ϕ] u-[ϕ] u-[ϕ] √ Match * Agree

In Step 3, the relevant phase is constructed between C-Top and C-Rel-that. Both of them possess the same attributes of the same features and a matching chain can be established between these two. However, the features attached to each of these two heads are unvalued and uninterpretable, the Agree chain cannot be established and the relevant phase will not be transferred immediately to the interfaces. Step 3: [TopP French star [Top0 ], … [CP2 [C0 that] Xiaoqian …[CP1who [C0] embraced (him j)]]] u-[var] u-[var] u-[var] u-[var] u-[ϕ] u-[ϕ] u-[ϕ] u-[ϕ] √ Match * Agree

Analyses  

After these three steps, the derivation of the entire sentence is finished and the highest Probe, C-Top, and the lowest Goal, the resumptive pronoun, form a dependency. Both of them bear the same features [ϕ] and [var] and of course the attributes of these features are the same. Therefore, a matching chain can be established. Since both features are uninterpretable features, Agree cannot work here. Even though the relevant dependency is not constructed by Agree, the whole sentence is still constructed by Match. After all of the phases are well established, the entire sentence is sent to the interfaces to be interpreted. In this case, even if Phase 1 contains a strong island, the derivation still converges because Match applies to the entire sentence, which can prevent the chain from violating locality constraints. Since neither Agree nor Move is involved in the derivation, island effects are not expected. Importantly, Multiple Transfers and Multiple Spell-outs do not apply here. This process is illustrated in Step 4. Step 4:

SPELL-OUT Transfer [TopP French star j [Top0 ], …, [CP2 [C0 that] X. knows [CP1who [C0] embraced (him j)]] u-[var] u-[var] u-[var] u-[var] u-[ϕ] u-[ϕ] u-[ϕ] u-[ϕ] √ Match * Agree  (all of the phases are sent to the interfaces at the same time)

Now let me use a metaphor to illustrate the derivational difference between Agree and Match. This metaphor is based on the installation of serial lamp bulbs. Imagine that our work is to install and to turn on six bulbs on the same electronic wire. Match is responsible for checking whether the connection between every pair of bulb sockets on the wire is good. If the connection is good, the electricity can pass freely from the first bulb socket to the last one and in this case, the relevant A'-dependency can be established by Match. Agree is responsible for checking if every bulb is compatible with its corresponding bulb socket. If there is no problem, every bulb can be turned on. If all of the bulbs on the wire can be turned on, the A'-dependency will be established by Agree. In this metaphor, we can see that Match and Agree look at different things and Match is an independent syntactic mechanism instead of a pre-condition for Agree. Now imagine that we have six bulbs A, B, C, D, E and F and six corresponding light sockets installed already on the wire.

  Syntactic derivation: two types of A'-dependencies A…..B…..C…..D…..E…..F Every two light sockets form a phase and so there are a total of five phases: (F, E); (E, D); (D, C); (C, B); (B, A). The function of Match is to check whether the connection between any two of the light sockets in each phase is good. Concretely, Match begins by checking the connection between light sockets A and B. If the connection (B, A) is good, the electricity can pass through A to B. Agree will check if the bulbs A and B are compatible with their corresponding sockets. And if they are compatible, the bulbs can be turned on, which means that phase (B, A) is well constructed and this phase can be sent to the interfaces. Then Match continues to check the connection between C and B and so on and so forth. Once all of the groups of two bulbs are checked and turned on, the A'dependency will be well established by Agree. This is the case for relatives (with a gap or with a resumptive pronoun) and the case for dislocation structures with a gap in Chinese. If the connection between each bulb in the first phase (B, A) is good, however, if the bulb is not compatible with its corresponding socket (for instance, their shapes or forms are not compatible), the two bulbs will not be able to be turned on, which means that Agree cannot establish the relevant dependency in this phase cycle. In this case, the connection between B and A can only be established by Match but not by Agree. Then, Match continues to check the two bulbs of the next group (C, B) and so on and so forth. We can imagine that none of these bulbs is turned on, which means that Agree cannot apply in this case. By contrast, the connection between each light socket on the wire is good and the electricity can pass from the first socket to the last one even if the corresponding bulbs cannot be turned on. This means that only Match works to establish the A'-dependency and Agree does not apply and such a situation corresponds to the derivation of resumptive dislocation structures in Chinese. .. Summary In this section, I propose an analysis in the Minimalist framework to account for the distribution of resumptive pronouns and of gaps in relative clauses and in dislocation structures respectively. The following table gives a summary of the results of the tests based on island effects and crossover effects as well as the relevant analyses that I propose.

Analyses   Tab. 6: Analysis Relatives

Dislocations

Gaps

Resumptives Intrusives

Gaps

Resumptives Intrusives

i) Without island

yes

yes

---------

yes

yes

--------

ii) With island

yes

-------

no (island effects)

yes

--------

yes (no island effects)

iii) Weak crossover effect

yes

yes

--------

yes

no

--------

iv) Strong crossover effect

yes

yes

---------

yes

no

--------

Syntactic derivation

GB

Movement

MP

Agree without Move

Binding in discourse Match without Agree

Essentially, relatives with a gap or with a resumptive pronoun and dislocation structures with a gap behave in a similar fashion in that all of these constructions give rise to island effects and crossover effects. In my analyses, A'dependencies involved in these structures are established by Agree without being followed by Move. Agree is an operation at the Narrow Syntax level and it only works on features. It applies phase by phase and cycle by cycle and that is why it is subject to locality conditions and displays island effects. Dislocation structures containing either resumptive pronoun in island-free contexts or an intrusive pronoun in island contexts are derived by Match without being followed by Agree. Match only identifies the attributes of the features attached to the Probe and to the Goal and it does not examine the value of these features. A dependency established by Match alone does not give rise to island effects or crossover effects. My analyses aim to explain all of these phenomena linked to the derivation of an A'-dependency (with a gap or with a resumptive pronoun) at Narrow Syntax. Typologically, relatives and dislocation structures with a gap in Chinese behave like relatives in languages such as Welsh and Scottish Gaelic since it is Agree that derives all of these structures. By contrast, resumptive dislocation structures in Chinese behave in a similar fashion to relatives in Irish and to dislocation structures in French where it is Match that applies (cf. Table 7).

  Syntactic derivation: two types of A'-dependencies Tab. 7: Comparison with other languages Mechanisms

Structures and languages

Agree

Match

– Relatives in Welsh25

– Relatives in Irish

– A'-dependencies in Scottish Gaelic

– Dislocation in French

– Relatives (with a gap or with RP) in Chinese – Dislocation with a gap in Chinese

– Resumptive dislocation in Chinese

. Conclusion This chapter is devoted to a detailed examination of the general distribution of resumptive pronouns in Mandarin Chinese. The essential question that I try to answer is whether a resumptive construction is derived by movement in Chinese. In order to answer this question, I apply two diagnostic tests for A'movement on different types of resumptive constructions: island effects and crossover effects. Based on the results of these tests, I try to analyze the observed generalizations in the framework of the Government and Binding theory as well as of the Minimalist Program respectively. (i)

Relatives with a gap, relatives with a resumptive pronoun and dislocation structures with a gap behave in a similar fashion in the two tests. Island effects and crossover effects are both detected in these three types of structures mentioned above. In this sense, intrusive pronouns do not exist in relatives and therefore, the eventual violation of the locality constraint cannot be avoided. In the framework of the Government and Binding theory, I conclude that there is movement involved in the derivation of these structures. By contrast, in the Minimalist Program, I argue that it is operation Agree that applies here and that Agree can work alone to establish the relevant A'-dependencies without requiring any movement operation. Agree operates at the strict narrow syntax level

 25 According to Rouveret (2002, 2008, to appear), two different internal structures can be assigned to resumptive pronouns in Welsh. One of them triggers Move and the other does not. In the former case Agree is followed by Move. I will discuss this analysis in detail in the next chapter.

Conclusion  

phase by phase in a cyclic fashion, which is subject to locality constraints. This is the reason why both island effects and crossover effects have been observed. The essential work of Agree is to check and delete the uninterpretable features before LF. In the system proposed by Adger & Ramchand, for instance, the complementizer of a relative clause, C-Rel, bears an interpretable [λ] feature that functions as a Probe and the resumptive pronoun that bears an uninterpretable [λ] feature functions as a Goal. The interpretable features attached to the Probe can value the uninterpretable feature attached to the Goal and once the feature valuation under Agree is finished, all of the uninterpretable features will be deleted. In the system of Rouveret (to appear), the [var] feature attached to the Probe (i.e. C-Rel) is uninterpretable, and the same feature attached to the Goal (i.e. the resumptive pronoun) is interpretable. Uninterpretable features that are not valued yet on the Probe c-commands the interpretable features on the Goal. After Agree, all of the uninterpretable features will be checked and deleted. Actually, both of these feature systems are applicable to Chinese relatives. (ii)

Dislocations with a resumptive pronoun do not give rise to any island effect or crossover effect, which makes possible for intrusive pronouns to exist in this type of construction. In this case, the eventual violation of locality conditions can be avoided. Such an observation suggests that movement is not involved in the derivation of resumptive dislocation structures. In the Minimalist Program, I propose that it is Match that applies alone without being followed by Agree in the derivation of this construction. The essential function of Match is not to check the relevant features but merely to identify the attributes of the features associated with the potential candidate for the Probe and for the Goal respectively. What prevents Agree from applying to this type of structure is that the features attached to the Probe and to the Goal are both uninterpretable features. For this reason, the Agree relation cannot be established between the Probe and the Goal since there is no suitable Goal that bears the same type of features that are interpretable and are capable of valuing and checking the uninterpretable features on the Probe. I also provided an empirical argument that suggests that the relation between the C-Top and the resumptive pronoun in a dislocation structure is not necessarily an operator-variable type of relation. By contrast, the relationship between the C-Rel and the resumptive pronoun located on the relativized site is indeed an operator-variable type of relation. As a result, the [λ] feature is uninterpretable when it is as-

  Syntactic derivation: two types of A'-dependencies sociated with C-Top but interpretable when it is associated to C-Rel. The [var] feature is interpretable on the resumptive pronoun in relatives but uninterpretable on the resumptive pronoun in dislocation structures. The second question that I addressed in this chapter concerns the role of gaps and of resumptive pronouns in these structures: is a resumptive pronoun merely analyzed as a phonologically spelled out trace resulting from wh-movement? Recall that based on languages like Vata, Koopman (1983) claims that resumptive pronouns and the wh-traces have exactly the same syntactic distribution in relatives and in interrogative sentences and that resumptive pronouns are subject to locality conditions in the same way as traces. Concretely, resumptive constructions can be blocked by islands in these languages in the same way as gap constructions are blocked by islands. As a result, Koopman concludes that resumptive pronouns are nothing more than phonologically spelled out traces. Asudeh (2011) also follows this idea. In other languages such as Swedish, island effects do not show up for resumptive pronouns in subject position, which is different from Vata. Engdahl (1985) concludes later that resumptive pronouns are not phonologically spelled out traces in Swedish. With the results of the different tests in Section 2.2, Chinese seems superficially to suggest that a resumptive pronoun can be analyzed as a spelled out trace in relatives but it can never be analyzed as a spelled out trace in dislocation structures. In fact, this asymmetry is only a superficial illusion due to pure coincidence. I will come back to the detail of this contrast in Chapter 3. In my final conclusion, I will state clearly that a resumptive pronoun never functions as a spelled out trace in relatives or in dislocation structures. The apparent asymmetry illustrated here is only a first superficial observation of a partial picture of resumptivity in Chinese. Moreover, when a dislocation structure contains an island, the intrusive pronoun clearly behaves differently from a gap in the same type of context. In the next chapter, I will continue to discuss the differences between a gap and a resumptive pronoun in terms of semantic interpretation. The third question that I asked at the beginning of this chapter concerns what Chinese can tell us about the phenomenon of resumptivity in a more general sense. (i)

Sells (1984) adopts a typological point of view in the basis of a macrovariation that divides languages into two categories: (a) those that only use resumptive pronouns in the grammatical/systematic/general way to form an A'-dependency and (b) those that only use resumptive pro-

Conclusion  

nouns as a last resort to save the sentence from potential violations of locality constraints. Category (a) also corresponds to “apparent resumption” and category (b) corresponds to “real resumption” in the sense of Aoun, Choueiri & Hornstein (2001). What we observe in Chinese suggests that such a distinction can also exist within the same language, which confirms the proposal of Boeckx (2003). As a result, the point of view based on macro-variation cannot be maintained. Chinese also shows that resumptive pronouns cannot be analyzed uniformly in different types of A'-dependencies nor do they behave in exactly the same way in the same language. (ii)

The strategies used to establish resumptive dependencies are specific to the types of structures but not specific to different languages. It is perfectly possible for a language to only use a single strategy, for example Move or Agree, in all of the resumptive constructions such as interrogatives, relatives and dislocation structures. However, Chinese shows that it is also perfectly possible to use different strategies in the same language and that these different strategies can be used to establish A'-dependencies in different languages in a specific way. For instance, in Chinese, resumptive relatives are derived by Agree without necessarily involving any movement operation and resumptive dislocation structures are derived by Match without being followed by Agree. In these structures, some specific features with different interpretabilities are associated with different complementizers (i.e. Probes) and with resumptive pronouns (i.e. Goals) simultaneously. For instance, CRel in a relative clause bears an interpretable [λ] feature since it functions as λ-operator in semantics, whereas the C-Top in a dislocation structure bears an uninterpretable [λ] feature because it does not bind the resumptive pronoun as variable. The resumptive pronoun in a dislocation structure bears an uninterpretable [var] feature since it is not translated as a variable in semantics, whereas the same resumptive pronoun in a relative clause bears an interpretable [var] feature since it is bound by C-Rel. The interpretabilities of these features are parametric with respect to the properties of the relevant functional elements in the language, rather than with respect to the language as a whole.

 Reconstruction and internal structures of resumptive pronouns . Introduction One of the traditions in the Government and Binding theory is to associate reconstruction effects exclusively with movement. Besides island effects and crossover effects, reconstruction effects also became another diagnostic test to determine whether an A'-dependency is derived by movement. Certain research work on resumption is based on this test. The logic of the reasoning is that if a resumptive construction displays reconstruction effects, it is derived by movement. In this chapter I will begin by presenting the work done by Aoun, Choueiri & Hornstein (2001) that offers an analysis of resumptive constructions with the aim of establishing a close relationship between reconstruction effects and movement. Their analysis is based on the observation from Lebanese Arabic. Along the same line, Aoun & Li (2003) also provide a detailed analysis for Chinese. Another point of view is to disassociate movement from reconstruction effects since there is not necessarily a cause-consequence relation between the two. One of the analyses that follow this view is proposed by Guilliot (2006). Based on the observation from Jordanian Arabic, he claims that the presence of a copy on the extracted site is enough to give rise to reconstruction effects. This analysis is in the same vein as the hypothesis according to which reconstruction effects are not conclusive for the presence of movement in the derivation of an A'-chain. Recall that in Chapter 1, I briefly presented the work of Doron (1982) which argues that gaps and resumptive pronouns do not give rise to the same readings. This observation is also confirmed by Guilliot (2006) who states that a copy is either analyzed as an indefinite description or as a definite description and that definite description and indefinite description do not permit the same interpretations. Namely, a definite copy allows neither pair-list reading nor reconstruction of a quantifier scope; by contrast, an indefinite copy permits both of them. Traces resulting from wh-movement can be interpreted either as definite copies or as indefinite copies; on the contrary, resumptive pronouns in Jordanian Arabic can only be interpreted as definite copies, which leads the author to conclude that resumptive pronouns are analyzed as definite descriptions, and therefore they can be treated as E-type pronouns (Evans 1980, Heim 1990, Heim & Kratzer 1998, Elbourn 2001).

  Reconstruction and internal structures of resumptive pronouns In contrast to what is observed in Jordanian Arabic, Rouveret (2008) shows that resumptive constructions in Welsh not only allow the reconstruction of a quantifier scope but also allow a pair-list reading. Based on this comparison, in this chapter, I will show that Chinese behaves like Welsh in that both of the reconstruction of the scope of a universal quantifier and the pair-list reading can be observed in resumptive constructions. In Section 3.2, I will present data from Chinese based on four different tests related to reconstruction effects. These tests aim to find out: (i)

whether both a gap and a resumptive pronoun can take a universal quantified phrase as antecedent;

(ii)

whether both an A'-dependency with a gap and an A'-dependency with a resumptive pronoun allow the reconstruction of a quantifier scope;

(iii)

whether both an A'-dependency with a gap and an A'-dependency with a resumptive pronoun allow anaphoric binding reconstruction;

(iv)

whether condition C effects are detected under reconstruction in both an A'-dependency with a gap and an A'-dependency with a resumptive pronoun.

I will show how to account for the facts observed in Chinese partially based on the original idea of Rouveret (2002, 2008) according to which, reconstruction effects are not linked to derivational mechanisms but to the internal structure of a gap and that of a resumptive pronoun (or an intrusive pronoun). Some elements located on variable sites can only have one specific form as their internal structure; however, the others can have different forms as their internal structures. Namely, Rouveret’s hypothesis based on Welsh is that a pronoun can have either a reduced form containing only a D head that does not give rise to reconstruction effects or an extended form containing a full NP that triggers reconstruction effects. Crucially, these two forms do not have the same status in that the reduced form, [D ϕ], is the default form for a pronoun but the extended form, [[D ϕ] NP], is a marked one in the sense that such a form is only triggered to satisfy interpretive requirements. Based on Mandarin data, I will show that the entire distribution of reconstruction effects is subject to an interaction between derivational mechanisms, internal structures of a resumptive pronoun and the syntactic constructions in which these pronouns appear. This establishes a logical relation between Chapter 2 and Chapter 3. In Chapter 2 we observed that derivational mechanisms

Introduction  

determine which type of resumptive pronoun exists in which type of syntactic construction. For instance, Agree requires that intrusive pronouns should not exist in relatives; whereas Match requires that intrusive pronouns can exist in dislocation structures. Concretely, a relative with an island is always derived by Agree, irrespective of whether an intrusive pronoun is present in such an island or not. Agree always obey locality constraints and this is the reason why with the presence of an intrusive pronoun, an A'-dependency derived by Agree, such as a relative clause with an island, automatically violates locality constraints and gives rise to island effects. This explains an observation that we had in Chapter 2: the presence of an intrusive pronoun cannot prevent a sentence from violating locality conditions. By contrast, since a resumptive dislocation with an island is always derived by Match and Match is not at all subject to locality conditions, the derivation does not display any island effect. As pointed out, the illusion that we had is that it is the presence of an intrusive pronoun in this case that prevents the sentence from violating locality constraints in spite of the presence of an island. I also show that derivational mechanisms determine which type of feature is associated with which type of syntactic construction. Specifically, I discussed in great detail the activation and deactivation of the EPP feature in the two types of derivations (Agree and Match). For instance, imagine that a complementizer C bears the EPP feature and there are two ways to satisfy the requirement of EPP. An XP can be base-generated in [Spec, CP] via operation Merge; it can also move there via operation Move. EPP is activated in relatives (with a gap or with a resumptive pronoun) and in dislocation structures with a gap. In these constructions, EPP always requires the [Spec, CP] position to be filled but the question that we ask is how. Actually, the EPP feature will require the presence of an NP in the internal structure of a gap and of a resumptive pronoun and this NP is a perfect candidate for movement in order to satisfy the requirement of EPP in [Spec, CP]. In this chapter, I will show that in the same type of structure, it is precisely the presence of an NP that gives rise to different reconstruction effects, which then explains why reconstruction effects are only observed in this type of structure. By contrast, topic in a resumptive dislocation structure is always base-generated in-situ via Merge. In other words, the [Spec, CP] position is always occupied and if the EPP feature is activated in these structures, then EPP is always satisfied. Another possibility is to assume that EPP is not at all activated in this case and as a result, it does not require the presence of any NP in the internal structure of a resumptive pronoun. This is the reason why in a resumptive dislocation structure, a resumptive pronoun can either have an extended form with the presence of an NP or a reduced form without any NP as its internal structure. Since it is the presence of an NP that triggers all of the interpreta-

  Reconstruction and internal structures of resumptive pronouns tive effects, reconstruction effects are not observed in the case of resumptive dislocations. In this chapter, I will show that different types of pronouns possess different internal structures. For the pronouns that possess several possible internal structures, it is always the EPP feature of the Probe that correlates with the internal structure of these pronouns.

. Reconstruction effects and movement When talking about reconstruction effects, two different syntactic constraints are generally examined: the binding condition (Chomsky 1982) and the constraint on bound variable construal (Reinhart 1983). These two types of conditions are distinguishable one from the other on the basis of the nature of the antecedent that appears in different syntactic constructions. An antecedent must be a quantified phrase if it is the constraint on the bound variable reading that is going to be tested. As for the binding conditions, they are listed below: (1) Condition A: An anaphor should be locally bound by its antecedent. Condition B: A pronoun cannot be locally bound by its antecedent. Condition C: A referential expression must be free everywhere. (2) Constraint on bound variable construal An anaphoric expression can get a bound variable reading if and only if it is bound by the trace of a quantifier. Certain syntactic operations such as wh-movement or topicalization will lead to a structural modification of a sentence. Concretely, the position of an antecedent and the position of an anaphoric expression (or a trace) will be inversed after movement, which will necessarily lead to a change of the relation between these two elements in terms of c-command. However, the violation of relevant conditions such as the Binding condition and the constraint on bound variable construal will not be systematically detected, which gives rise to so-called “reconstruction effects”. Reconstruction effects describe the situation where a displaced constituent is always interpreted in its base-position (i.e. its original argument position) before movement takes place. I will present two types of reconstruction effects in this section.

Reconstruction effects and movement  

.. Two types of reconstruction effects ... Reconstruction of anaphoric binding According to the original formulation of the constraint on bound variable construal proposed by Reinhart (1983), a pronoun can receive a bound variable reading if and only if it is bound by the trace of a quantifier. (3)

[Every manj [IP tj thinks [CP that hej is smart]]].

In the above example, the interpretation of the pronoun hej depends on the value of the quantifier phrase every manj, which gives rise to so-called bound variable reading of the pronoun hej. The pronoun hej is bound by the trace tj resulting from the raising of the universal quantified phrase every manj. Recall that I explained in Chapter 2 that crossover configurations (strong or weak) precisely violate the constraint on bound anaphora, as shown in (4). (4) ?? Whoj does hisj mother like tj? The pronoun hisj cannot get a bound variable reading from the quantifier who because his is not bound by the trace tj resulting from the movement of who. However, it has been observed that topicalization gives rise to bound variable reading in spite of its apparent violation of the constraint on bound anaphora. Here is an example from Guilliot (2006). (5) a. A picture of himself1, everyman1 tore. b. Everyman1 tore a picture of himself1. Although c-command relation is inversed in (5a), the anaphor himself can always get a bound variable reading, which means that the constraint on bound anaphora is not necessarily violated. It seems that the topicalized NP a picture of himself in (5a) is still interpreted in its base-position (i.e. its original argument position) at the syntactic level as shown in (5b). This phenomenon illustrates reconstruction effects. ... Reconstruction of the scope of a quantifier Syntactically, the order ∀ > ∃ gives rise to a wide scope reading of the universal quantifier yielding a distributive reading. For instance,

  Reconstruction and internal structures of resumptive pronouns (6) a. Every doctor will examine a patient. b. Which patientj will every doctor examine tj ? In (6a), when the universal quantified phrase every doctor scopes over the indefinite, the existential quantified phrase a patient, the distributive reading in which each doctor examines a potentially different patient is available. Syntactically, the existential quantifier ∃ is under the scope of the universal quantifier ∀ and therefore, the interpretation of the indefinite will depend on the universal quantifier. By contrast, in the wh-question (6b) after the movement of which patient to the [Spec, CP] position, which patient is no longer under the scope of every doctor. The reversed order now is ∃>∀; however, this reversed c-command relation does not change the original reading of the sentence. The universal quantified phrase always takes a wide scope yielding the reading where every doctor examines a different patient. In addition, (6b) also shows that which patient is always interpreted in its original argument position occupied by the trace tj, which illustrates a reconstruction of the scope of a universal quantifier. Similar distinction has been made in Pan (2014) where wh-movement can but topicalization cannot resolve scope ambiguities of a quantifier. .. Reconstruction effects and movement It has been pointed out that in the framework of the Government and Binding theory, reconstruction effects are particularly detected in constructions derived by A'-movement such as topicalization and wh-movement as illustrated in the previous section. As a result, the relationship between reconstruction effects and movement becomes a kind of cause-consequence relation. The presence of reconstruction effects naturally signals that movement is involved in the derivation. Obviously, a trace resulting from an A'-movement satisfies neither the binding condition nor the constraint on bound variable construal because these constraints are structurally defined in terms of c-command at the level of Surface-Structure. In order to account for this contrast, different analyses that have been proposed previously assign a derivational ambiguity to a given A'dependency. One of the frequently cited examples involves two ways to derive a relative clause. Two possible solutions on the basis of reconstruction effects have been proposed for deriving relative clauses in English. The first analysis is called “raising analysis” which proposes that an Op operator or its equivalent which raises from inside TP to [Spec, CP], as shown in (7). (7) a. the book1 [CP [Op/which t1]2 John likes t2 ]

Reconstruction effects and movement  

b.

DP NP

D the 0

CP

NP book1

C’

Spec [Op/which t1]2 C0 [+ Rel]

TP John likes t2

The second possibility to derive a relative clause is called “matching analysis” (Sauerland 1998, 2004, Hulsey and Sauerland 2006). In this analysis, there are two heads in a relative clause: one is an external head and the other is an internal head. The external head, i.e. the NP antecedent, is base-generated directly in [Spec, CP] outside the relative clause via operation Merge. The internal head joins the embedded verb in order to get a θ-role and then this internal head will be phonologically deleted under an identity condition with the external head. (8) a. the book1 [CP [Op/which book1 ]2 John likes t2 ] b.

DP NP

D0 the NP book1

CP

C’

Spec [Op/which book1]2 C0 [+ Rel]

TP John likes t2

  Reconstruction and internal structures of resumptive pronouns Husley & Sauerland (2006) argue that both strategies of the derivation of relatives are available in the Universal Grammar and that they are licensed by different conditions. In other words, independent syntactic and semantic conditions exist to make a distinction between them. Sauerland (1998) claims that a relative clause derived by the raising of an NP gives rise to anaphoric binding effects, reconstruction effects and bound variable effects. Under reconstruction effects, it is also possible to detect condition C effects of the Binding Theory. Crucially, in this analysis, all of these detected effects are exclusively related to derivation of A'-dependencies by movement, and as a result, these effects naturally become reliable diagnostic tests for movement. Along this line, if a relative clause displays these effects, one can conclude that there is NP-raising, thus a movement, involved in the derivation. By contrast, if a relative clause is derived by Matching, and thus no movement is involved in the derivation, then all of the interpretative effects, reconstruction effects and condition C effects under reconstruction will not be detected. Of course, there are other possible analyses of the derivation of relative clauses with resumptive pronouns. Namely, Salzmann (2006) proposes an analysis based on German and different varieties of German and he examines the relation between movement, ellipsis and reconstruction effects.

.. Two types of resumptivities ... Two generalizations of Aoun et al. (2001) Much previous work on resumptivity uses reconstruction effects as diagnostic tests to determine whether a resumptive construction is derived by movement. The idea is that if resumption displays different kinds of reconstruction effects, movement is involved in the derivation of such a resumptive construction; by contrast, the absence of these reconstruction effects will suggest that such a resumptive construction is not derived by movement. Based on this reasoning, Aoun et al. (2001) proposes an important analysis that studies resumptive constructions in Lebanese Arabic. Crucially, they make a distinction between a “true resumption” and an “apparent resumption”. A true resumption refers to the type of resumptive construction that does not give rise to reconstruction effects, such as definite relative clauses with an island and indefinite relative clauses. In addition, this type of resumptive construction is not sensitive to island boundaries. In a true resumptive construction, the mechanism that establishes the dependency between a resumptive pronoun (or an epithet) embedded within an island and its antecedent is the operation called Bind. It is worthwhile pointing out that resumptive pronouns (or epithets) in a true resumptive con-

Reconstruction effects and movement  

struction in the sense of Aoun et al. corresponds to intrusive pronouns in the sense of Sells (1984) because the presence of an intrusive pronoun is licensed in island contexts. This type of resumptive pronoun is analyzed as E-type pronoun (Evans 1980, Heim 1990, Heim & Kratzer 1998, Elbourn 2001). An E-type pronoun behaves as a definite expression and it cannot take any quantificational expression as its antecedent. In contrast to a true resumption, an apparent resumptive construction systematically gives rise to reconstruction effects. Typical constructions containing apparent resumptives are definite relatives without any island. An apparent resumptive construction is derived by the operation Move and thus it obeys locality constraints. Apparent resumptive constructions roughly correspond to “grammatical resumptions” or “real/true resumptions” or “the general use of resumptive pronouns” in the sense of Sells (1984). We should also notice that the term “true resumption” is used in opposite ways in Aoun et al. (2001) and in Sells (1984). The following table summarizes the differences between these two types of resumptions in these two works. Tab. 1: Terminology in Aoun et al. (2001) True resumption

Apparent resumption

Strategies

Bind

Move

Structures

– Definite relatives with an island

– Definite relatives without any island

– Indefinite relatives Reconstruction effects

Equivalents in Sells ()

NO

YES

“intrusive pronouns”

Grammatical resumption/ true resumption

– Can appear in island – Cannot take any quantificational antecedent – Treated as E-type pronoun

Following examples are cited from Aoun et al. (2001), which illustrate the two types of resumptive constructions that appear in dislocation structures.

  Reconstruction and internal structures of resumptive pronouns (9) a.

təlmiiz-[a]i student-her

l-kəsleen the-bad

ma NEG

baddna want.1P

nχabbir tell.1P

[wala mʕallme]i ʔənno huwwe zaʕbar no teacher that he cheated.3SM b-l-faħʂ. in-the-exam ‘Herj bad student, we don’t want to tell any teacherj that he cheated on the exam.’ b.

* təlmiiz-[a]i student-her

l-kəsleen the-bad

ma NEG

badkun want.2P

tχabbro tell.2P

[wala mʕallme]i ʕan l-bənt yalli huwwe no teacher about the-girl that he zaʕbar maʕ-a b-l-faħʂ. cheated.3SM with-her in-the-exam (‘Herj bad student, you don’t want to tell any teacherj about the girl with whom he cheated on the exam.’) Lebanese Arabic, Aoun et al. (2001) Sentence in (9a) illustrates an apparent resumptive construction that is established by Move. The embedded clause is a complement clause of a verb that does not construct any island. Reconstruction effects have been observed in this sentence in that the distributive reading of the resumptive pronoun her depends on the negative quantified phrase no teacher. In this case, in spite of the apparent order in syntax between these two elements, no teacher still takes a wide scope over the pronoun her. This reading will force the dislocated NP, her bad student, to be interpreted on the variable site occupied by the resumptive pronoun he. Since this type of resumptive construction is derived by Move and it gives rise to reconstruction effects, Aoun et al. conclude that reconstruction effects can only be detected in the case where a resumptive construction is derived by movement. Under the consideration that movement is in fact movement of copies, the copy of the dislocated DP is present on the variable site occupied by a resumptive pronoun in an apparent resumptive construction. (9a’) [TopP [DP herj student]i…[TP any teacher… [DP herj student hei]]] Move

Reconstruction effects and movement  

In (9a’), the copy her student at the bottom of the chain is under the scope of any teacher and her student can actually get a distributive reading, which gives rise to reconstruction effects. By contrast, (9b) contains a true resumptive construction in that there is a complex-NP island constructed by a relative clause. According to Aoun et al., this type of resumptive construction is derived by Bind. With the presence of an island, the distributive reading of the pronoun her disappears. We should also notice that the ungrammaticality of this sentence is not due to violation of locality conditions because the presence of a resumptive pronoun in island contexts plays the role of a last resort that can avoid this potential violation. The sentence is filtered by the unavailability of the distributive reading of the pronoun her. Based on this contrast, Aoun et al. (2001) conclude that the presence of an island blocks reconstruction of the scope of a quantifier. In the schema illustrated below for (9b), the pronoun her does not appear under the scope of the universal quantified phrase any teacher due to absence of the lower copy. There is no movement involved in the derivation of this dependency because according to the authors, the linking between the dislocated DP in TopP position and the resumptive pronoun he is established via Bind. As a result, there is no copy of this DP present in the chain. This configuration violates the constraint on the bound variable construal and that is why distributive reading for her is not available. (9b’) * [TopP [DP herj student]i…. [TP any teacher…

[Island ….. [DP hei]]]]

Bind The contrast observed by Aoun et al. (2001) is clear-cut: the presence of reconstruction effects with an absence of islands is considered as an argument in support of the hypothesis that an apparent resumptive construction is derived by movement; by contrast, the absence of reconstruction effects in a resumptive construction with an island is considered as an argument in support of the idea that such a dependency does not involve any movement. In other words, if reconstruction effects show up in a context without islands, then the dependency in such a resumptive construction is established by Move and such a construction is an apparent resumptive construction. If reconstruction effects are absent in a context with an island, then the dependency in such a resumptive construction is established by Bind, and this construction is treated as a true resumptive construction in the sense of Aoun et al.

  Reconstruction and internal structures of resumptive pronouns ... Against the generalizations of Aoun et al. (2001) It is also observed that there is not necessarily a cause-consequence type of relation between movement and reconstruction effects. Guilliot (2006) claims that the presence of a copy on the variable site occupied by a resumptive pronoun is sufficient to trigger reconstruction effects and this copy is not necessarily derived by movement. He provides two counter-examples to invalidate the generalizations of Aoun et al. Namely, on the one hand, the presence of an island does not necessarily block reconstructions and on the other hand, the absence of islands do not systematically give rise to reconstruction effects either. In this section, I will begin by presenting these counter-examples in Guilliot (2006) and then I will provide Chinese data to confirm that structures derived by movement do not coincide with structures that display reconstruction effects. More importantly, the presence or the absence of reconstruction effects does not construct a boundary between apparent resumptive constructions and true resumptive constructions in the sense of Aoun et al. (2001). .... Against the first generalization of Aoun et al.: absence of island First, it is not really true that reconstruction effects systematically show up in all of the apparent resumptive constructions. In other words, reconstruction effects are not always observed in resumptive constructions without islands. In Welsh, for instance, a construction containing a weak resumptive does not give rise to condition C effects. Here is an example from Rouveret (2008) that shows condition C effects are not detected under reconstruction in an apparent resumptive dependency (i.e. without islands). (10)

Yn ddiweddar, dygwyd darlun o Siôn yr recently was-stolen picture of Siôn that oedd ef wedi ei roddi i Mair. was he Perf [it] give to Mair ‘Recently was stolen a picture of Siôn which he had given to Mair.’

If the anaphoric expression a picture of Siôn is reconstructed in the variable position occupied by the resumptive pronoun ef ‘he’, the reconstructed structure will give rise to condition C effects. The offended configuration would be something like the following: (10’)

Condition C effects: * Hei had given a picture of Siôni to Mair.

Reconstruction effects and movement  

In this configuration, the proper name Siôn inside the DP a picture of Siôn is directly c-commanded by the pronoun ef ‘he’ (in the subject position of the relative clause) that shares the same index i, which leads to the violation of condition C. However, the grammaticality of the sentence in (10) indicates that this co-referential relationship is properly licensed between the pronoun ef ‘he’ and the proper name Siôn. As a result, what we can conclude is that the DP a picture of Siôn is not reconstructed at all, which leads to the absence of condition C effects. Data from Jordanian Arabic also confirms this observation. In (11) (my translation), the null subject, pro, of the complement clause is interpreted as a coreferent of Karim and this binding relationship does not give rise to reconstruction effects. (11)

ʕalamit Karim gultu ʔinnu lazim ʔiggayyar-ha note Karim say-2Pl that should pro change-[la] ‘The note1 of Karim2, you say that he2 should change it1.’ Jordanian Arabic, Malkawi (2009)

The absence of condition C effects in (11) also shows that no copy of the antecedent DP is present on the extracted site. Similar observation has also been made for Breton in Guilliot (2006). In the following two relative clauses, no condition C effect is detected, which confirms that the DP antecedent the photo/brother of Yann is not reconstructed on the relativized site. (12) a.

b.

poltred2 Yann1 a lâres en deus (pro1) photo Yann Prt you-say Prt has he en2 gwelet it seen ‘the photo2 of Yann1 that you say that he1 has seen it2’ breur2 Yann1 en deus (pro1) komzet gant-añ2 brother Yann Prt has he spoken with-him ‘the brother of Yann1 with whom he1 has spoken’ Breton, Guilliot (2006)

Importantly, in the same context, Chinese examples show exactly the same point. A resumptive pronoun in a dislocation structure does not display any condition C effect.

  Reconstruction and internal structures of resumptive pronouns (13) a.

一勤 1 的論文 2 , 我覺得她 1 可以把它 2 發表在國家一級期刊上。 Yiqin1 de lunwen2, wo juede ta21 keyi ba Yiqin DE article 1Sg think 3FSg ta32 fabiao zai guojia yi-ji 3OSg publish at national first.rate ‘As for the article2 of Yiqin1, I think that she1 can first-rate national journal.’

b.

我弟弟 1 的新手錶 2,他 1 覺得它 2 很酷! Wo didi1 de xin shoubiao2,

can BA qikan-shang. journal-on publish it2 in the

ta11

juede

1Sg-Gen brother DE new watch 3MSg think ta32 hen ku ! 3OSg very cool ‘As for the new watch2 of my little brother1, he1 thinks that it2 is so cool!’ c.

小倩 1 的小狼狗 2,她 1 非常喜歡它 2。 Xiaoqian1 de xiao langgou2,

ta21

feichang

Xiaoqian DE little wolf-dog 3FSg very xihuan ta32. like 3OSg ‘As for Xiaoqian1’s little wolf dog2, she1 likes it2 very much.’ The grammaticality of (13a) suggests that after the reconstruction, the anaphoric binding between Yiqin1 and the subject pronoun she1 is not established and therefore, the pronoun she1 does not c-command the antecedent NP. The same observation goes for (13b) and (13c). All of these examples in different languages constitute counter-arguments for the generalizations made by Aoun et al. (2001). Our conclusion is that an apparent resumptive construction in the sense of Aoun et al. (2001), without islands, does not systematically give rise to condition C effects. .... Against the second generalization of Aoun et al.: presence of islands Recall that the second generalization of Aoun et al. (2011) claims that a true resumptive construction (i.e. with the presence of an island), is always derived by operation Bind and that since such a derivation does not involve any movement, reconstruction effects are never observed. However, examples from different languages clearly show again that reconstructions are not at all systematically excluded in true resumptive constructions. For instance, Guilliot (2006)

Reconstruction effects and movement  

shows that in Jordanian Arabic, reconstruction effects are indeed detected in sentences where a weak resumptive is embedded within an island, as shown in (14–15) (my translation in English). (14)

Dislocation out of an adjunct island [ Talib-[ha]1 l-kassul ] 2 ma

ziʕlat

student-her the bad Neg upset-3FSg [wala mʕalmih]1 laʔannuh l-mudiirah no teacher because the-principle kaħ∫at oh2 mn l-madrase. expelled-3FSg 3MSg from the-school ‘Her1 bad student2, no teacher1 was upset [because the principle expelled him2 from school].’ (15)

Dislocation out of a complex-NP island [ Talib-[ha]1 l-kassul ] 2 ma badku student-her the bad Neg want-2 tχabbru [wala mʕalmih]1 ʕan l-bint say-2 no teacher of the-girl saʕadat oh2 b-l-faħiʂ. helped-3Sg 3MSg in-the-exam

illi who

‘Her1 bad student2, you don’t want to talk to any teacher1 about the girl [who helped him2 in the exam].’ In these two examples, the antecedent NP Talib-ha ‘her student’ can always receive a distributive reading where the pronoun oh ‘him’ is interpreted as a variable bound by the universal quantifier (even if it is under negation). The only way to get such a reading is by reconstructing the pronoun ha ‘her’ on the extracted site occupied by the resumptive pronoun oh ‘him’ and as a result, ha ‘her’ is under the scope of the quantified phrase wala mʕalmih ‘no teacher’. Both examples show that this distributive reading can be obtained if and only if the dislocated NP is reconstructed by crossing an island boundary. The related resumptive pronoun is in its intrusive use here and the presence of oh ‘him’ embedded inside each island (i.e. an adjunct island in (14) and a complex-NP in (15)) prevents relevant sentences from violating locality constraints. Importantly, the presence of islands in these two sentences does not block reconstruction of a quantifier scope. The configuration after the reconstructions is given below: (14’/15’) [TP no teacher1….[island her1 bad student]]

  Reconstruction and internal structures of resumptive pronouns We should notice that the configuration presented in (14’/15’) is precisely the case that should be filtered by grammar in Aoun et al. (2001)’s hypothesis. Recall that their generalization claims that the presence of islands will necessarily block reconstruction effects because a resumptive dependency containing an island is constructed by Bind without involving any movement operation. However, Jordanian Arabic data show that this is not necessarily true. As reconstruction effects go, Lebanese Arabic and Jordanian Arabic behave quite differently one from the other. French data in Guilliot (2006)’s work can also invalidate the generalization of Aoun et al. (2001). French shows that it is perfectly possible to reconstruct a dislocated constituent on an extracted site occupied by a resumptive pronoun embedded inside a strong island. Let me cite two examples (my translation): (16) a.

Dislocation out of an adjunct island La photo1 de sa2 classe,

tu

es

fâché

the picture of his class you are angry [parce que chaque prof2 l1’a déchirée]. because every teacher it-has torn ‘The picture1 of his2 class, you are angry because every teacher2 tore it1.’ b.

Dislocation out of a complex-NP island La photo1 de sa2 fille,

tu

connais

la

the picture of his daughter you know the personne [à qui chaque prof2 l1’a montrée]. person to whom every teacher it-has shown ‘The picture1 of his2 daughter, you know the person2 to whom every teacher has shown it1.’ In the above examples, the resumptive pronoun co-referring with the dislocated DP the picture is embedded inside strong islands: an adjunct island in (16a) and a relative clause in (16b). It is important to observe that both sentences can get a distributive reading of the pronoun his in the antecedent NP. As I showed above, such a distributive reading is based on the bound variable reading of the pronoun his which is interpreted as a bound anaphor. This reading can only be available after the reconstruction of the antecedent DP containing the pronoun his on the dislocated site occupied by the resumptive clitic pronoun l(a) ‘it’, as illustrated below.

Reconstruction effects and movement  

(16a’/16b’) [TP …. [island every teacher1 … the photo of his1 class]] After the reconstruction, the pronoun his is under the scope of the universal quantified phrase every teacher, which ensures the bound variable construal of his. It is important to bear in mind that the binding between these two elements under reconstruction can perfectly be established. However, there is still an important distinction of the way to construct the relevant examples between Jordanian Arabic (cf. 14–15) and French (cf. 16a–b). In Jordanian examples, quantified phrases are located outside islands whereas in French examples, quantified phrases appear inside islands. In order to compare real minimal pairs, I will construct two French examples (cf. 17–18) in which quantified phrases and the reconstruction site are clearly separated by island boundaries. (17)

La photo2 de sa1 fiancée, aucune personne1 ne the picture of his fiancée no person Neg veut pardonner celui [qui l2’a déchirée]. want forgive the.one who it-has torn ‘The picture2 of his1 fiancée, nobody1 will forgive the one who tore it2.’

(17’) [TP nobody1….[island the picture of his1 fiancée]] (18)

La bande-annonce2 de son1 nouveau film, the trailer of his new movie chaque réalisateur1 est très content every director is very happy [parce qu’elle2 a été bien diffusée]. because.3FSg has been well distributed ‘The trailer2 of his1 latest movie, every director1 is very happy because it2 has been well distributed.’

(18’)

[TP every director1….[island the trailer of his1 latest movie]]

In (17), the DP the picture of his fiancée can receive a distributed reading bound by the quantified phrase nobody. The only way that allows the pronoun his to be interpreted as a bound anaphor is to reconstruct the antecedent DP inside the island (i.e. relative clause), as indicated in (17’). Crucially, the quantificational antecedent is very well separated from the pronoun his by an island boundary, which gives exactly the same configuration as the one in the original examples in Aoun et al. (2001). Similarly, (18’) illustrates the same configuration but with

  Reconstruction and internal structures of resumptive pronouns a universal quantified phrase, every director, and with an island constructed by an adjunct clause. These two examples confirm the point that the presence of islands in a given resumptive dependency does not necessarily block reconstruction, which goes against the generalization made by Aoun et al. The observation is that a true resumptive construction in the sense of Aoun et al. can also give rise to reconstruction effects. I would like to point out that Chinese data also confirm that the generalization of Aoun et al. is too strong. Examples (19–20) show that it is perfectly possible to reconstruct the pronoun his inside an island and that this pronoun can be interpreted as a anaphor bound by universal quantified phrases, thus the distributive reading is available. (19)

自己 1 的身份證 2,每位公民 1 必須要*(將其 2)更新的指定機構在漢口。 Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen ID.card *(jiang qi2) gengxin JIANG it replace zai Hankou. in Hankou

every-Cl citizen de zhiding C designated

must jigou service

‘His1 own ID card2, the official service [where every citizen1 is supposed to replace *(it2) with a new one] is in Hankou District.’ (19’)

[island every citizen1 … his1 own ID card]

In (19), the pronoun ziji ‘self’ in the dislocated antecedent DP ziji-de shenfenzheng ‘his own ID card’ can receive a bound variable reading depending on the universal quantified phrase mei-wei gongming ‘every citizen’. It is clear that the antecedent DP must be reconstructed on the dislocated site occupied by the resumptive pronoun qi ‘it’ as indicated in (19’), which also shows that reconstructions by crossing island boundaries are possible. Example (20) shows exactly the same point. The possessive use of the resumptive pronoun qi ‘their’ in NP, qi yongxin ‘their intention= the parents’ good intention’ is embedded inside the strong island constructed by a complex-NP; however, the sentence can receive a reading in which parents vary according to every child and this shows that the topic NP phrase ziji-de fumu ‘his own (= self) parents’ is fully reconstructed on the dislocated site occupied by the resumptive pronoun qi ‘their’ inside the island and therefore, his own is under the scope of the universal quantified phrase every child yielding a distributive reading.

Reconstruction effects and movement  

(20)

自己 1 的父母 2,每位子女 1 都能完全理解*(其 2)用心的家庭還是不多 的。 Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen wanquan entirely jiating family

parents every-Cl child all lijie *(qi2) yongxin understand 3-Gen intention haishi bu duo de. yet Neg many DE

de DE

can

‘His1 own parents2, the families [in which every child1 can perfectly understand them2] are not many yet.’ (20’) [island every child1 … his1 own parents] ... Conclusion In this section, I presented an important analysis of Aoun et al. (2001) that makes a distinction between two types of resumptive constructions: true resumptions and apparent resumptions. A definite relative clause containing an island constitutes an apparent resumptive construction and this type of resumption is derived by operation Bind since it does not give rise to any reconstruction effect. A relative clause without any island is considered as a true resumption that is derived by operation Move. Aoun et al. claim that only true resumptive constructions give rise to reconstruction effects. They make the following generalizations: i)

The absence of islands makes it possible to reconstruct an antecedent on the relativized site occupied by a resumptive pronoun.

ii)

The presence of islands blocks the reconstruction of an antecedent on the relativized site occupied by a resumptive pronoun.

However, data from Welsh, Breton, Jordanian Arabic, French and Mandarin Chinese show that these two generalizations are not universal principles but simply some generalization based on one particular language. Specifically, data from Mandarin Chinese confirm the observation made in Guilliot (2006) according to which, on the one hand, the absence of islands do not force reconstructions and on the other hand, the presence of an island will definitely not block reconstructions. From this point of view, our tests in Chinese also confirm that resumptive dependencies derived by Move and resumptive dependencies in which reconstructions are possible certainly overlap but one does not necessari-

  Reconstruction and internal structures of resumptive pronouns ly coincide with the other. Based on this line, one question that we can pose at this stage is what factor can account for the presence or the absence of reconstruction effects in resumptive dependencies.

. Reconstruction effects and internal structures of resumptive pronouns In the previous section, I presented the analysis of Aoun at al. (2001) that tries to establish a close relationship between reconstruction effects and movement. I also presented counter-arguments based on different languages to invalidate the main generalization of their analysis. If the presence or the absence of reconstruction effects is not due to movement, then the question is what is the factor that is responsible for reconstruction effects in resumptive constructions. In this section, I will present several analyses on the basis of internal structures assigned to resumptive pronouns. The observation is that not all of different types of resumptive pronouns can systematically display reconstruction effects, and this leads some authors to propose a hypothesis according to which presumably, it is the internal structure of a resumptive pronoun itself that triggers reconstruction effects. What Aoun et al. (2001) discover is that Lebanese Arabic possesses two types of pronouns, strong pronouns and epithets on the one hand and weak pronouns on the other hand. These two types of pronouns behave differently when reconstruction effects are taken into consideration. Aoun et al. assign a specific structure to each of these two types of pronouns in order to account for the observed interpretative properties. The analysis proposed in Guilliot (2006) is based on the idea that the presence of a syntactic copy on the reconstructed site is sufficient to trigger reconstruction effects and that a copy can have two different forms: a definite form and an indefinite form. A resumptive pronoun in Jordanian Arabic is analyzed as a definite description, and thus as an E-type pronoun which allows neither pair-list reading nor reconstructions of quantifier scope. However, as the reader will see, this analysis cannot be applied to Chinese because Chinese behaves like Welsh in that resumptive constructions in both languages not only allow pair-list reading but also reconstruction of a quantifier scope. Finally, I will present the analysis of Rouveret based on his serial work on this topic (2002, 2008, to appear). His analysis tries to reduce all of the interpretative properties of a resumptive construction to a structural ambiguity of a given resumptive pronoun. The main idea is that a resumptive pronoun can have two different forms as its internal structures: an extended form and a re-

Reconstruction effects and internal structures of resumptive pronouns  

duced form. The extended form can and the reduced form cannot give rise to reconstruction effects. I will also discuss how grammar chooses the appropriate internal structure, between the two available ones, to use in a given resumptive construction. Rouveret (to appear) proposes that between the two available forms, one is a marked form and the other is a default form and that the marked form (i.e. the extended form) is only activated when there are interpretative requirements to satisfy. .. Two types of resumptive pronouns: strong pronouns/epithets and weak pronouns Following Aoun and Benmamoun (1998), Aoun et al. (2001) discover a distinction between strong pronouns and epithets on the one hand and weak pronouns (i.e. clitic pronouns) on the other hand in Lebanese Arabic. Strong pronouns and epithets behave as independent morphemes and namely, an epithet can be used as a resumptive pronoun and interpreted as a bound variable, as illustrated in (21). (21)

S-Sabej yalli raħ tə∫ħaT Zeena ha-l-majduubj the-boy that Fut-kick-3FSg Zeena this-the-idiot mən l-madrase harab from the-school ran-away-3MSg ‘The boyj that Zeena will kick this idiotj out of school ran away’ Lebanese Arabic, Aoun et al. (2001)

Aoun et al. propose that strong pronouns and epithets have a structure, which is parallel to that of a full DP, when they are used with the pronominal morpheme ha ‘this/that’. The pronominal ha occupies the specifier position of DP and ϕfeatures such as person, number and gender are provided by epithets, as illustrated in (22). (22) a. [DP ha- D [NP epithet]] b. [DP ha- [D morpheme ϕ]] By contrast, weak pronouns (clitics) are affixes attached to lexical heads such as V°, N° or P°. When a weak pronoun is merged with a full DP, the former is considered as the head of the full projection and therefore, occupies the head position of the DP. The specifier position can be occupied by an NP that will potentially undergo movement, as illustrated in (23).

  Reconstruction and internal structures of resumptive pronouns (23) [DP DP (lexical) [D weak pronoun (clitic)]] There is a crucial difference between the structure of a weak pronoun and the structure of a strong pronoun/an epithet. When the head position of a DP is occupied by a weak pronoun, its specifier position will be available for an eventual movement; by contrast, when the head position of a DP is occupied by a strong pronoun, its specifier position will be occupied by the pronominal morpheme ha ‘this/that’ which is not a potential candidate for movement. One of the advantages of this analysis is that it successfully accounts for the fact that not both of these two types of pronouns can take quantificational antecedents. It has been observed that a strong pronoun or an epithet can take a quantified phrase as its antecedent if and only if this pronoun is embedded within an island. (24a) shows that the pronoun hiyye ‘she’ that is not embedded within an island cannot take the universal quantified phrase kəll muttahame ‘every suspect’ as its antecedent ; by contrast, hiyye ‘she’ can take kəll muttahame ‘every suspect’ as its antecedent when she is embedded within an island, as shown in (24b). (24) a.

Without island * kəll muttahame

arəfto

ʔənno

hiyye

every suspect-F know-2Pl that [she] nhabasit. imprisoned-3FSg (‘Every suspect, you know that she was imprisoned.’) b.

Wh-island kəll muttahame

badkun

taarfo

[miin

every suspect-F want-2Pl know-2Pl who bifakkir ənno hiyye harabit]. think-3MSg that [she] ran-away-3FSg ‘Every suspect, you want to know [who thinks that she ran away].’ Lebanese Arabic, Aoun et al. (2001) By contrast, a weak pronoun can always systematically take a quantificational antecedent, irrespective of whether this pronoun is embedded within an island or not, as illustrated in (25).

Reconstruction effects and internal structures of resumptive pronouns  

(25) a.

Without island kəll məʒrim

fakkarto

ʔənno

every criminal-M thought-2Pl that l-bolisiyye laʔatu-u the-police-Pl caught-3Pl-[him] ‘Every criminal, [you thought [that the police caught him]].’ b.

Wh-island kəll muttahame

badkun

taʕrfo

every suspect-F want-2Pl know-2Pl [miin ħabas-a]. who imprisoned-3SgM-[her] ‘Every suspect, you want to know [who imprisoned her].’ Lebanese Arabic, Aoun et al. (2001) Aoun et al. also point out that a quantified phrase can be the antecedent of a resumptive pronoun in a true resumptive construction established by Bind (cf. 24b, 25b); by contrast, a quantified phrase cannot be the antecedent of a resumptive pronoun in an apparent resumptive construction established by Move (cf. 24a). Let us come back to reconstruction effects. Guilliot & Malkawi (2007, 2009) inquire further differences between strong pronouns/epithets and weak pronouns based on Jordanian Arabic data. Their generalization is that only the structure of a weak pronoun gives rise to reconstruction effects and the structure of a strong pronoun and an epithet does not. The presence of islands blocks reconstruction in the case of strong pronouns, such as hu ‘he’ in (26), and in the case of epithets, such as (ha)-l-gabi ‘that idiot’ in (27). However, the presence of islands allows reconstructions in the case of weak pronouns such as ho ‘he’, as shown in (28). (26)

Strong pronoun in an adjunct island * [ Talib-[ha]1 l-kassul]2 ma ħakjan

maʕ

student-her the-bad Neg talked.1Pl with [wala mʕalmih]1 gabl-ma hu2 yesal. no teacher before he arrive.3MSg (‘Her1 bad student2, we didn’t talk to any teacher1 before he2 arrived.’)

  Reconstruction and internal structures of resumptive pronouns (27)

Epithet in an adjunct island * [ Talib-[ha]1 l-kassul]2

ma

ħakjan

maʕ

student-her the-bad Neg talked.1Pl [wala mʕalmih]1 gabl (ha)-l-gabi2 no teacher before the.idiot.3MSg (‘Her1 bad student2, we didn’t talk to any teacher1 arrived.’) (28)

Weak pronoun in an adjunct island [Talib-[ha]1 l-kassul]2 ma ziʕlat student-her the-bad Neg upset-3FSg laʔannuh l-mudiirah kaħ∫at because the-principle expelled-3FSg l-madrase. the-school

with yesal. arrive.3MSg before this idiot2

[wala

mʕalmih]1

no oh2 3MSg

teacher mn from

‘Her1 bad student2, no teacher1 was upset because the principal expelled him2 from school.’ Jordanian Arabic, Guilliot & Malkawi (2011) The generalization made by Guilliot & Malkawi is that weak resumption allows reconstructions within an island and strong resumption blocks such reconstructions. This generalization is based on the observation that all of the cases which permit reconstructions inside an island involve either a weak pronoun or a clitic pronoun and no reconstruction effect is detected in the case of strong pronouns and epithets. Crucially, for both authors, a real distinction that one should make exists between weak resumptive constructions and strong resumptive constructions but not between apparent resumptive constructions (derived by Move) and true resumptive constructions (derived by Bind without any movement). .. Two types of copies: indefinite and definite (Guilliot 2006) According to Guilliot (2006) and Malkawi (2009), data from Jornadian Arabic and from French confirm that there should not be a cause-consequence type of correlation between movement and reconstruction, contrary to what Aoun et al. (2001) propose. One of the important generalizations made by Guilliot (2006) is that resumptivity blocks the reconstruction of a quantifier scope and as a result, it does not allow any pair-list reading. The following Jordanian Arabic examples cited from Guilliot (2006) (my translation) show that a resumptive pronoun does

Reconstruction effects and internal structures of resumptive pronouns  

not allow any pair-list reading (cf. 29) whereas a trace resulting from whmovement allows a pair-list reading (cf. 30). (29)

ʔaya Surah1 il-uh2 kul zalamih2 mazaʕ-ha1? which picture of-him every man tore-it ‘Which picture of him did every man tear (it)?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading ≠ (ii) Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading

(30)

ʔaya Surah1 il-uh2 kul zalamih2 which picture of-him every man mazaʕ ____1? tore ‘Which picture of him did every man tear?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading = (ii)

Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading

These examples thus confirm that resumptivity blocks pair-list reading and as a result, resumptive constructions indeed permit reconstruction of anaphoric binding but do not permit reconstruction of a quantifier scope. In order to account for such an asymmetry, Guilliot (2006) proposes an analysis of resumption based on a theory of copy. According to him, a syntactic copy can result either from movement or from ellipsis. Copies resulting from movement must obey all of the locality constraints but copies resulting from ellipsis do not have

  Reconstruction and internal structures of resumptive pronouns to. Syntactic copies are interpreted either as indefinite descriptions or as definite descriptions and these two forms do not give rise to the same interpretative effects. Concretely, reconstructions resulting from movement contain either indefinite copies or definite copies (i.e. traces) whereas reconstructions resulting from ellipsis (i.e. resumption) contain only elided definite copies. Only indefinite copies give rise to reconstruction of the scope of a quantifier yielding a pair-list reading.26 Since resumptive constructions in Jordanian Arabic do not give rise to a pair-list reading and they do not allow any reconstruction of a quantifier scope either, Guilliot claims that resumptive pronouns are analyzed as definite descriptions. In this sense, resumptive pronouns are treated as copies resulting from ellipsis and therefore behave as E-type pronouns (cf. Elbourne 2005). Two different forms can be associated with resumptive pronouns: (31) a. [DP [D resumptive ] NP ] b. [DP resumptivei ] The structure in (31a) represents a form containing a definite copy in which a definite determiner takes an NP as argument and this NP will be deleted later under identity condition with its antecedent. The presence of a nominal restriction in this structure triggers a covariant reading and a bound anaphor reading. By contrast, the structure in (31b) represents a form containing a definite determiner that bears an index. This form gives rise to a distributive reading yielding a bound variable construal when the antecedent is a quantified phrase and to an individual reading when the antecedent is not quantified. This analysis can explain several observed facts in resumptive constructions. First, it has been observed that reconstructions are possible in contexts where a resumptive construction is located inside an island, which means that islands do not block reconstructions. The assumption of Guilliot is that reconstructions do not depend on movement but on the presence of a copy on variable sites in a given syntactic construction. In addition, in contrast to movement, ellipsis is not subject to island conditions and in this case, the form of such a resumptive pronoun is the one represented in (31a). Second, it has also been observed that all of the different types of reconstruction effects have not been detected systematically in resumptive constructions. Lebeaux (1992) makes a distinction between two types of conditions: positive conditions are those which either license a deriva-

 26 See Guilliot (2006) for a detailed description of the mechanism called Skolem(ized) Function that generates indefinite copies.

Reconstruction effects and internal structures of resumptive pronouns  

tion or allow a certain reading, such as condition A of the Binding theory and the constraint on the bound anaphora; on the contrary, negative conditions are those which filter a derivation or certain readings, such as condition C of the Binding theory. In the previous sections, I have shown that certain resumptive dependencies give rise to reconstruction effects in order to satisfy positive conditions and that certain resumptive dependencies give rise to reconstruction effects to satisfy negative conditions. The solution proposed by Guilliot is that when positive conditions are involved, the form in (31a) is selected and it permits reconstructions of anaphoric binding whereas when negative conditions are involved, such as condition C, the form in (31b) is selected and it blocks reconstructions. In the latter case, the absence of an NP under reconstruction will not trigger a potential violation of condition C. Let me just point out that in this analysis an important relation is established between the definiteness of a copy and derivational mechanisms that produce such a copy. For instance, a resumptive pronoun resulting from ellipsis is treated as a definite copy. In fact, this type of analysis based on a structural ambiguity of a pronoun has already been proposed a long time ago in Rouveret (1994) and I will present his analysis in the next section. Through a close survey on Chinese data concerning the permitted readings in resumptive constructions, we can find that Chinese behaves very differently from Jordanian Arabic. In Chinese, reconstruction of a quantifier scope is possible and more importantly, a pair-list reading is also available in resumptions. For instance, (32)

[每個導演都會將它 j 看好幾遍]的一個電影預告片 j [mei-ge daoyan dou hui jiang ta3j

kan

each-Cl director all will JIANG 3OSg watch hao-ji-bian de yi-ge dianying-yugaopianj several.times C one-Cl movie.trailer ‘a movie trailerj [that each director will watch itj several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃) (i) Shi ta-de xin pian. be his new movie (ii)

‘It is his latest movie.’ Functional reading Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

  Reconstruction and internal structures of resumptive pronouns In (32), the direct object resumptive pronoun ta3 ‘it’ occupies the dislocated site that is under the scope of the universal quantified phrase mei-ge daoyan ‘every director’ in the subject position. The fact that the sentence can have a reading where each director watches a different movie trailer shows that the antecedent indefinite NP a movie trailer is fully reconstructed on the dislocated site occupied by the resumptive pronoun ta3 ‘it’. The following example shows the same point. (33)

每個導演 1 都會將它 2 拿去參賽的他自己 1 的新片 2 [mei-ge daoyan1 dou hui jiang ta32

na

qu

each-Cl director all will JIANG 3OSg take to can-sai ] de ta31-ziji-de xin-pian2 participate.to.competition C 3MSg-self-DE new-movie ‘the latest movie2 of his1 own [that every director1 will take it2 to participate to a competition]’  √ Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading Recall that reconstruction of the scope of a quantifier is also possible in resumptive constructions in Welsh, as is illustrated in Rouveret (2008). Here is an example, (34)

Mae gan Siôn farn ar ei lyfr y is with Siôn opinion about his book that mae pob awdur yn ei pharchu. is each author Prog [it] respect ‘Siôn has an opinion about his book that each author respects.’ Welsh, Rouveret (2008)

The pronoun ei ‘his’ can be interpreted as a variable bound by the universal quantified phrase pob awdur ‘every author’ embedded within a relative clause, which gives rise to a distributive reading and therefore a pair-list reading: ‘every author respects a different point of view.’ Data from Chinese and from Welsh confirm that a copy resulting from resumption is not necessarily interpreted as a definite description and resumptive constructions can unmistakably give rise to a pair-list reading and they also allow the reconstruction of a quantifier scope. In order to account for all of these

Reconstruction effects and internal structures of resumptive pronouns  

observations, I will present the analysis of Rouveret (1994, 2002, 2008, to appear) below. .. Two internal structures of resumptive pronouns: Rouveret (1994, 2002, 2008, to appear) One of the important properties concerning reconstruction effects is that not all of the resumptive dependencies can give rise to reconstruction effects. Example (34) above shows that in Welsh, a resumptive construction can give rise to reconstruction of the scope of a universal quantifier. By contrast, the following example shows that resumptive constructions do not systematically give rise to the same type of reconstruction effects. (35)

barn yr athro ar ei mab [y gŵyr opinion the teacher on her son that knows pob mam [y mae ef yn ei chuddio]] each mother that is he Prog [it] conceal # ‘the teacher1’s opinion on her2 son [that each mother2 knows [that he1 conceals]]’ Welsh, Rouveret (2008)

The sentence in (35) cannot get a distributive reading which can simultaneously satisfy the following two conditions: a) each mother is linked to a different son (i.e. under pair-list reading); b) a co-referential relation is established between yr athro ‘the teacher’ and ef ‘he’. However, an analysis based on movement incorrectly predicts that such a reading exists since movement gives rise to the following LF representation: 27 (35’) LF: each mother2 > the teacher1 > her2 son > he1 Recall that this phenomenon leads Rouveret (2008) to conclude that in Welsh a resumptive construction behaves as a base-generated structure and that movement does not intervene in the case of resumption. His analysis is also in support of the idea that a pronoun has an internal structure and that different types of pronouns have different types of internal structures (Cardinaletti & Starke 1999, Déchaine & Wiltschko 2002, et al.).

 27 This type of example has been extensively discussed in Aoun and Li (2003).

  Reconstruction and internal structures of resumptive pronouns Now let me cite the four crucial examples that show the most important four properties of reconstruction effects in resumptive relatives in Welsh summarized in Rouveret (2008, 2011). First, it is possible to have a pronominal variable binding under reconstruction, as demonstrated in (36). (36)

Mae gan Siôn farn ar ei lyfr y is with Siôn opinion about his book that mae pob awdur yn ei pharchu. is each author Prog [it] respect ‘Siôn has an opinion about his book that each author respects.’ Welsh, Rouveret (2008)

Quantified phrase pob awdur ‘each author’ embedded within the relative clause can scope over the pronoun ei ‘his’ located in the NP antecedent. The pronoun his is interpreted as a bound anaphor, which gives rise to a distributive reading. In order to get a bound variable construal, the quantificational phrase each author must c-command the pronoun his and as a result, NP farn ar ei lyfr ‘opinion about his book’ should be reconstructed on the relativized site occupied by the resumptive pronoun ei ‘it’, as shown in (36’). (36’) that eachi author respects an opinion about hisi book It is also possible to have an anaphoric binding under reconstruction, as illustrated in (37). (37)

Fe Prt

'm me

hysbyswyd was-reported

am about

y the

clecs gossips

amdano about

ei hun y mae Siôn wedi eu clywed himself that is Siôn Perf them hear yn y cyfarfod. at the party ‘The gossips about himselfj [that Siônj heard at the party] were reported to me.’ In order for ei hun ‘himself’ to be anaphorically related to Siôn, the NP y clecs amdano ei hun ‘the gossips about himself’ must be reconstructed on the relativized site occupied by the resumptive pronoun eu ‘them’, as shown in (37’). (37’) Siônj heard the gossips about himselfj at the party…

Reconstruction effects and internal structures of resumptive pronouns  

Under reconstruction, condition C effects can be observed in resumptive constructions, as illustrated in (38). (38)

* barn yr athro ar ei opinion the teacher on her ef y mae pob mam he that is each mother (‘the teacheri’s opinion on her son that respects (it)’)

mab y gŵyr son that knows yn ei pharchu. Prog it respect hei knows that each mother

The anaphoric relationship is established between the pronoun ef ‘he’ and the NP yr athro ‘the teacher’. After the reconstruction, the configuration is represented as the following: (38’) * hei knows that each mother respects the teacheri’s opinion on her son The NP yr athro ‘the teacher’ is directly bound by the pronoun ef ‘he’, which violates condition C of the Binding theory. The ungrammaticality of this sentence suggests that condition C effects are detected in this case. As a result, the NP the teacher’s opinion on her son is properly reconstructed on the relativized site occupied by the resumptive pronoun ei ‘it’ in this sentence. Surprisingly enough, Rouveret (2008) also shows a case where condition C effects are not at all observed in a resumptive construction, as shown in (39). (39)

Yn ddiweddar, dygwyd darlun o Siôn yr recently was-stolen picture of Siôn that oedd ef wedi ei roddi i Mair. was he Perf [it] give to Mair ‘Recently was stolen a picture of Siôn which he had given to Mair.’

In this example, the co-referential relationship is licensed between the pronoun ef ‘he’ in the subject position of the relative clause and the proper name Siôn embedded as a prepositional complement in the nominal head of the relative clause. Crucially, if the antecedent darlun o Siôn ‘picture of Siôn’ is really reconstructed in the position occupied by the resumptive pronoun ei ‘it’, the resulting configuration should be like the following: (39’) * hei had given a picture of Siôni to Mair

  Reconstruction and internal structures of resumptive pronouns Apparently, such a reconstruction violates condition C because the proper name Siôn is directly c-commanded by the pronoun he. However, the fact that this sentence is grammatical suggests that the relevant reconstruction cannot take place. Before introducing the solution to this paradox proposed by Rouveret, I would like to present several analyses of internal structure of a pronoun. According to Postal (1966) and Freidin & Vergnaud (2001), personal pronouns can be analyzed as definite descriptions, as illustrated in (40). (40) Representation of definite pronouns [DP [+def] ϕ NP ], ϕ : agreement feature NP : null NP component Rouveret points out that such a representation satisfies The Inclusiveness Condition (Chomsky 1995) which prohibits any new item to be introduced during the course of a derivation. Under such a consideration, index becomes redundant. Importantly, in the structure presented in (40), it is the definite determiner that is responsible for syntactic and semantic properties of the relevant pronoun. At LF, a definite pronoun is interpreted as a definite description and at PF, a definite pronoun is always interpreted as a definite description in which the NP component is not spelled out. If the idea that what is left after movement is not a trace but a copy of such a moved element, then the notion of “reconstruction” becomes redundant. Rouveret (2008) applies this analysis of definite pronouns to resumptive pronouns in Welsh. Already proposed in his previous work, the author establishes a structural ambiguity of resumptive pronouns: an extended form as shown in (41a) and a reduced form as shown in (41b). (41) Extended form: [[D ϕ] NP] (the null NP will be deleted at LF) Reduced form:

[D ϕ]

Recall that the original idea proposed in Rouveret (1994) is that a sub-extraction movement applies in a resumptive chain. The sub-part NP of the extended form of a resumptive pronoun will be extracted in syntax and this movement is subject to locality constraints. The co-existence of these two forms as internal structures for a resumptive pronoun is always conserved in later work of Rouveret (2002, 2008) in order to account for the presence and the absence of reconstruction effects in resumptive constructions. When a resumptive construction gives rise to certain interpretative effects under reconstruction, it is these interpretations that require the extended form with the presence of an NP to be used.

Reconstruction effects and internal structures of resumptive pronouns  

First, in the case of reconstruction of a quantifier scope illustrated in example (36), the extended form must be adopted. The presence of an NP will ensure that opinion about his book will be entirely reconstructed on the relativized site occupied by the resumptive pronoun ei ‘it’. After this reconstruction, the existential expression an opinion will be under the scope of the universal quantified expression each author, which yields a distributive reading on DP an opinion. (36’’) = (36) Quantifier scope reconstruction effects ei ‘it’ = [DP D ϕ (3rd, Sg., Masc.) [an opinion about his book]]  [DP [each (∀) ]2 author ] ... [DP D1 ϕ [an opinion (∃) about his2 book]] (∀ > ∃) As for the binding reconstruction in (37), it is always the extended form that is selected. The NP the gossips about himselfj will be entirely reconstructed in the position occupied by the resumptive pronoun, which makes the anaphor himself be c-commanded, thus bound, directly by the proper name Siôn. The presence of an NP under reconstruction ensures that the binding is established between Siôn and himself. (37’’) = (37) Anaphoric binding reconstruction effects eu ‘them’ = [DP D ϕ (3rd, Pl., Masc.) [the gossips about himself]]  [IP Siônj heard [DP D ϕ [the gossips about himselfj]] at the party] As for the presence of condition C effects under reconstruction in (38), it is always the extended form that will be selected. As shown in (38”), after the reconstruction of the NP opinion of the teacher on the site occupied by the resumptive pronoun, the NP teacher is directly c-commanded and thus bound by the pronoun he bearing the same index. The resulting configuration violates condition C. (38’’) = (38) The presence of condition C effects under reconstruction ei ‘it’ = [D ϕ(3rd, Sg., Masc.) [opinion of [teacher]]]  * [IP hej knows…[D ϕ [opinion of [teacherj]]]] In the case with the absence of condition C effects as shown in (39’), Rouveret assumes that the reconstruction still takes place in the relevant resumptive construction; however it is the reduced form without the presence of an NP that is selected. Since the NP is absent, the DP a picture of Siôn will not be reconstructed in the position of the resumptive pronoun and as a result, the pro-

  Reconstruction and internal structures of resumptive pronouns noun he does not bind any proper name that potentially bears the same index, which does not violate condition C. (39’’) = (39) The absence of condition C effects under reconstruction ef ‘it’ = [DP D [ϕ (3rd, Sg., Masc.)]] (reduced form)  [IP he had given [D ϕ] to Mair]. In his analysis, Rouveret establishes a linking between the possibility to have a reconstruction and the presence of a certain type of copy in an argument position. Crucially, in this analysis, movement option is completely excluded. An element (or a sub-part of this element) that is relativized, questioned or dislocated is represented as a copy of the trace or as a covert definite description left on the original site, which causes that constructions which are not derived by movement still give rise to similar reconstruction effects. .. Choice of internal structure of a resumptive pronoun Assuming a structural ambiguity of internal structures of a resumptive pronoun encounters a difficulty; the question that one can ask is in what way grammar selects a suitable structure to use in a given A'-dependency. This question is totally justified in the sense that none of the previous analyses predicts that a suitable structure of a pronoun triggered by the need of satisfying interpretative requirements in a given resumptive construction is completely random. The reason is simple: if such a choice is completely free and random, then grammar will become unpredictable, which is of course not true. In the analysis offered by Guilliot (2006), resumption is only derived by ellipsis and a resumptive pronoun is interpreted as a definite description. However, such a definite description always remains ambiguous between two different internal structures. Guilliot & Malkawi (2007) try to suggest that a given type of pronoun has only one particular internal structure: strong pronouns and epithets possess a different structure from weak pronouns and clitics. As illustrated in the previous section, such a division based on strong resumption and weak resumption does not apply to all of the languages. For instance, in Welsh, it is always the same resumptive pronoun that appears in different resumptive constructions; however, such a pronoun does not systematically give rise to the same reconstruction effects in different syntactic structures. More importantly, a resumptive con-

Reconstruction effects and internal structures of resumptive pronouns  

struction in Welsh28 and in Mandarin allows both the reconstruction of quantifier scope and a pair-list reading, which is completely different from Jordanian Arabic. Rouveret (2008) argues that between the two different forms associated with a given resumptive pronoun, one is a marked form and the other is a default form. The marked form is selected if and only if there are interpretative requirements to satisfy, which means that the extended form with the presence of an NP is the marked form because NP can satisfy interpretative properties and give rise to reconstruction effects. The unmarked form is reduced to D-ϕ and the NP is not present in such a form. As a result, reconstruction effects are not present when the reduced form is selected. In Section 3.4, I will propose, based on Mandarin data, another solution to this problem related to the choice of an appropriate form of a resumptive pronoun. This solution suggests that all of the reconstruction effects observed in Chinese can be explained by an interaction of at least four different factors, namely derivational mechanisms, syntactic constructions in which resumptive pronouns appear, different types of pronouns and internal structures of a pronoun. On the one hand, a given construction can have a preference on one of the two available forms of a pronoun for some specific reasons; on the other hand, some pronouns can have two different forms as their internal structures but others have only one form. In the first case, if a given pronoun has two available forms, it is the syntactic mechanism that derives the construction in which such a pronoun appears that selects a suitable form of the pronoun in order to satisfy interpretative properties required by formal features associated with this construction. However, in the second case, when a resumptive pronoun has only one form as its internal structure, the syntactic construction has no choice but to take the only available form. .. Summary In this section, I have presented several promising analyses that try to assimilate reconstruction effects to internal structures of resumptive pronouns. The observed facts are presented as the following, (i)

A resumptive construction can give rise to reconstruction effects;

 28 In Welsh, the choice between a strong pronoun and a weak pronoun is syntactically determined, irrespective of resumption (p.c. Rouveret).

  Reconstruction and internal structures of resumptive pronouns (ii)

Not all of the resumptive constructions systematically give rise to reconstruction effects.

In order to account for these two properties, some related work tries to argue that a given language has different types of resumptive pronouns which give rise to different kinds of reconstruction effects. Under such a consideration, in these languages, it is ordinary referential pronouns that function as resumptive pronouns. For instance, strong pronouns and weak pronouns do not behave uniformly with regard to reconstruction effects in Lebanese Arabic (Aoun et al. 2001) and in Jordanian Arabic (Guilliot 2006, 2011, Malkawi 2009, Guilliot & Malkawi 2006, 2007, 2009, 2011). Some work also tries to assimilate different types of reconstruction effects to different derivational mechanisms. For Guilliot (2006), since ellipsis leaves definite copies and resumption is derived by ellipsis, resumptive pronouns are better analyzed as definite copies. This analysis ties syntactic representations and semantic interpretations together and it can account for the fact that in certain languages, resumptive constructions allow neither the reconstruction of a quantifier scope nor a pair-list reading. Unfortunately, this analysis is limited to languages like Jordanian Arabic. In fact, as I showed earlier, Mandarin Chinese and Welsh resumptive constructions do allow the reconstruction of a quantifier scope and a pair-list reading without any problem. In order to account for this aspect of the fact, another possible analysis proposed in a serial work of Rouveret suggests that some languages allow a structural ambiguity in terms of internal structures of a resumptive pronoun. For instance, a resumptive pronoun in Welsh can either have an extended form or a reduced form as its internal structure. The reduced form is a default structure and the extended form is only triggered in the case of need to satisfy interpretative properties. As Rouveret shows, the crucial point implied in this analysis is that two types of internal structures are present in the derivation from the beginning of the numeration process, which satisfies the Inclusiveness condition that prohibits the introduction of any new item during the course of a derivation. This is how indices are eliminated in the computational system in the Minimalist Program. In the next section, I will systematically examine Chinese data in order to get a general distribution of reconstruction effects in resumptive dependencies. Based on the result of the tests, I will show that the hypothesis of Rouveret partially applies to Chinese and the complexity of resumption in Chinese suggests that derivational mechanisms, syntactic constructions, internal structures of pronouns and different types of pronouns work together to account for the general distribution of reconstruction effects observed.

Reconstruction effects in Chinese  

. Reconstruction effects in Chinese In this section I will apply reliable diagnostic tests for reconstruction effects to Chinese in a systematic way. I will examine gaps, resumptive pronouns and intrusive pronouns in two different types of A'-dependencies: relatives and dislocation structures. These tests will help us to see whether the three types of elements, gaps, resumptives and intrusives, behave in exactly the same way concerning reconstruction effects. The four tests that I will apply are the following: (i)

Whether a distributive reading is available after the reconstruction of the scope of a universal quantifier;

(ii)

Whether a gap, a resumptive pronoun and an intrusive pronoun can take a quantified NP as their antecedent;

(iii)

Whether an anaphoric binding can be established under reconstruction;

(iv)

Whether condition C effects are produced under reconstruction.

I will account for the results of these four tests with the previous analysis based on an ambiguity of internal structures of a pronoun. The idea is that if an A'dependency gives a “yes” as answer to all of the listed four questions, then the related element (gaps or resumptive pronouns) in this dependency will have an extended form [[D-ϕ] NP] as its internal structure. It is precisely the presence of an NP in this structure that satisfies the interpretative properties and gives rise to condition C effects, as listed in (i-iv) on the above. By contrast, if an A'dependency gives a “no” as answer to the above four questions, then the gap or the resumptive pronoun in such a dependency will have a reduced form [D-ϕ] as its internal structure. The lacking of an NP part will be responsible for the absence of these interpretative properties as well as the presence of condition C effects. The following table serves as the itinerary of our tests and argumentation in this section.

  Reconstruction and internal structures of resumptive pronouns Tab. 2: Reconstruction effects itinerary Relatives

Dislocations

Gaps

Resumptives

Gaps

Resumptives

Intrusives

(a)

(b)

(d)

(e)

(g)

(i) Quantifier scope reconstruction? (ii) Quantificational antecedent? (iii) Reconstruction of anaphoric binding? (iv) Condition C effects? Internal structure

In this section, my tests will only be realized on pronouns of the third person in the series constructed by ta, such as ta1 ‘he/him’, ta2 ‘she/her’, ta3 ‘it’ (for objects and animals) as well as their plural forms.29 .. Reconstruction and quantifier scope In this section, I will apply the first test for reconstruction of a quantifier scope to Chinese. I will construct my examples based on the following configuration to see whether an antecedent DP can have a distributive reading. (42) [CP DPi (x), [TP [∀ (x) ... [DP RPi /____i (x)]]]]

 29 I will examine another pronoun of the third person, qi, in the next section for a comparative study.

Reconstruction effects in Chinese  

If a distributive reading is available for such a DP antecedent, then it means that reconstruction effects are produced in the relevant A'-dependency. This DP will be reconstructed on the extracted site occupied either by a resumptive pronoun or by a gap. The universal quantifier binds the reconstructed DP as variable. Following the internal structure hypothesis proposed by Rouveret (2002, 2008), I will claim that if the reconstruction of a quantifier scope is possible, a resumptive pronoun or a gap in this situation takes an extended form [[D-ϕ] NP] as its internal structure; by contrast, if this reconstruction does not take place and distributive reading is not available, then I will conclude that a gap or a resumptive pronoun in this situation takes a reduced form [D-ϕ] without the presence of NP as its internal structure. ... Relatives In example (43), a distributive reading of the indefinite antecedent yi-ge dianying yugaopian ‘a movie trailer’ in which each movie director watches several times a potentially different movie trailer is available. This implies that the NP a movie trailer can be reconstructed on the relativized site occupied by a gap and that the sentence can get an interpretation where the universal quantified phrase mei-ge daoyan ‘each/every director’ scopes over the existential quantified phrase a movie trailer. (43)

[每個導演都會看好幾遍_____j]的一個電影預告片 j [mei-ge daoyan dou hui kan

hao-ji-bian

each-Cl ____ j]

director all will watch several.times de yi-ge dianying-yugaopianj C one-Cl movie.trailer ‘a movie trailerj [that each director will watch _____j several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃) Example (44) below can also receive a distributive reading in which the interpretation of the anaphoric expression qi xinzuo de yugaopian ‘the trailer of his latest movie’ depends on the universal quantified phrase mei-ge daoyan ‘each/every director’. The only way to get such a reading is to reconstruct the trailer of his latest movie on the relativized site occupied by a gap located in the direct object position and this will make the anaphoric expression fall under the scope of the universal quantified expression every director located in the subject position.

  Reconstruction and internal structures of resumptive pronouns (44)

[每個導演 2 都會看好幾遍_____1]的關於其 2 新作的預告片 1 [mei-ge daoyan2 dou hui kan hao-ji-bian ____ 1] each-Cl director all will watch several.times de guanyu qi2 xinzuo de yugaopian1 C about 3-Gen new.product DE trailer ‘the trailer1 of his2 latest movie [that every director2 will watch _____1 several times]’

If reconstruction of the scope of the universal quantifier is possible in both examples, then we can claim that a gap, like the ones in the above examples, takes an extended form as its internal structure, i.e. a full DP form [[D-ϕ] NP]. Only the extended form with the presence of an NP can be bound as variable. As stated earlier, the presence of an NP is crucial for a bound variable since a variable must have a morphologically explicit nominal restriction; otherwise, the universal quantifier will bind no variable in its scope, which leads to a vacuous quantification. (43’) [____] = [DP D ϕ (3rd, Sg., inanimate) [NP a movie trailer]]  [DP [eachi (∀)] director ] ... [DP D ϕ [NP a (∃) movie traileri]] In (43’), the indefinite a movie trailer is located inside the NP. If this NP is reconstructed and only in this case, the existential quantified expression a movie trailer located in the direct object position will fall under the scope of the universal quantified phrase each director in the subject position, which yields a distributive reading. This analysis also applies to (44’) where the anaphoric expression his latest movie falls under the scope of the universal quantified expression each director, which also yields a distributive reading. (44’) [____] = [DP D ϕ (3rd, Sg., inanimate) [NP the trailer of his latest movie]]  [DP [ each2 (∀)] director ] ... [DP D ϕ [NP the trailer of his2 latest movie]] Let us turn to resumptive pronouns in relative clauses. Examples (45) and (46) show that a resumptive pronoun can also have a distributive reading in the same context containing a universal quantified phrase.30

 30 We must also note that the result of these tests is different from that discussed in Aoun & Li (2003:169), and such a difference could be due to different language regions and the personal preference of the different speakers.

Reconstruction effects in Chinese  

(45)

[每個導演都會將它 j 看好幾遍]的一個電影預告片 j [mei-ge daoyan dou hui jiang ta3j

kan

each-Cl director all will JIANG 3OSg watch hao-ji-bian de yi-ge dianying-yugaopianj several.times C one-Cl movie.trailer ‘a movie trailerj [that each director will watch itj several times]’  √‘each director watches a different movie trailer’ (∀ > ∃) (46)

每個導演 1 都會將它 2 拿去參賽的他自己 1 的新片 2 [mei-ge daoyan1 dou hui jiang ta32

na

qu

each-Cl director all will JIANG 3OSg take to can-sai ] de ta31-ziji-de xin-pian2 participate.to.competition C 3MSg-self-DE new-movie ‘the latest movie2 of his1 own [that every director1 will take it2 to participate to a competition]’  √ ‘each director will take his own new movie to participate to a competition’ (∀ > ∃) We can assume that in this case, a resumptive pronoun also has an extended form as its internal structure identical to that of a gap. The extended form contains a D° head with certain specified ϕ-features and an NP. In (45), since the indefinite a new movie trailer is a part of the NP, when this NP is reconstructed, the indefinite will fall under the scope of the universal quantified expression each director and this is how the indefinite gets a distributive reading.31 (45’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate) [NP a movie trailer]]  [DP [eachj (∀)] director ] ... [DP D ϕ [NP aj (∃) movie trailer]] (∀ > ∃) Similar reasoning also applies to the case in (46). What is special in (46) is that the use of the anaphor ziji ‘self’ will force a covariant reading between the anaphoric expression the latest movie of his own and the universal quantified phrase each director, which is the only available reading of the sentence. The grammaticality of this sentence indicates that the related binding relation is properly established, which leads me to assume that the pronoun ta3 ‘it’ takes an extended form in this context. After the reconstruction of the full NP, the

 31 Indefinites are analyzed as pure variables without any inherent quantificational force (cf. Heim 1982).

  Reconstruction and internal structures of resumptive pronouns anaphoric expression the latest movie of his own falls under the scope of the universal quantified expression each director and as a result, a distributive reading is available in this sentence. (46’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate) [NP his latest movie]]  [DP [eachj (∀)] director ] ... [DP D ϕ [hisj latest movie]] (∀ > ∃) ... Dislocation structures In dislocation structures, a gap can always get a distributive reading under the scope of a universal quantifier after the reconstruction (cf. 47). (47)

自己 1 的新片 2 , 每個導演 1 都會拿____2 去參賽。 Ziji1-de xin-pian2 , mei-ge daoyan1

dou

self-DE new-movie each-Cl director all hui na ____2 qu can-sai. will take to participate.to.competition ‘The latest movie2 of his1 own, every director1 will take ______2 to participate to a competition.’ In (47), the use of the anaphor ziji ‘self’ forces the distributive reading of the NP the latest movie of his own and this reading depends on the universal quantified expression every director. The grammaticality of this sentence leads me to assume that the latest movie of his own falls under the scope of every director after the reconstruction takes place. Recall that the only way to ensure that the establishment of the binding relation between these two elements is to reconstruct the full NP on the dislocated site occupied by the gap. In this case, I assume that the gap in this situation takes an extended form [[D-ϕ] NP], as illustrated in (47’). (47’) [____] = [DP D ϕ (3rd, Sg., inanimate) [NP the latest movie of his own]]  [DP [each1 (∀)] director] ... [DP D ϕ [NP the latest movie of his1 own]] In the same context with a resumptive pronoun, the judgment on the grammaticality and acceptability of the relevant sentences change from one native speaker to another. However, most informants (i.e. higher than 80%) that I consulted think that sentences like (48) are comprehensible but not fully acceptable.

Reconstruction effects in Chinese  

(48)

?* 自己 1 的新片 2 , 每個導演 1 都會把它 2 拿去參賽。 ?* Ziji1-de xin-pian2 , mei-ge daoyan1 dou

hui

self-DE new-movie each-Cl director all will ba ta32 na qu can-sai. BA 3OSg take to participate.to.competition (‘The latest movie2 of his1 own, every director1 will take it2 to participate to a competition.’) If this sentence is unacceptable for most native speakers that I consulted, it is because for them, the latest movie of his own cannot be bound by the universal quantified expression every director, which means that the NP part is not reconstructed on the dislocated site occupied by the resumptive pronoun ta3 ‘it’. Logically, in this case, I assume that the pronoun ta3 ‘it’ takes a reduced form lacking an NP as its internal structure. NP here is considered as the nominal restriction of a variable; lacking such a restriction will lead to a vacuous quantification for the universal quantifier every, as illustrated in (48’). (48’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate)]  * [DP [everyi (∀)] director ] ... [DP D ϕ]

(vacuous quantification)

The presence of the anaphor ziji ‘self’ forces a distributive reading that is the only available reading of this sentence. Therefore, if this only reading is not possible, the sentence becomes ungrammatical. The situation is exactly the same for an intrusive pronoun. I begin by presenting two sentences with islands to which I will apply the test. The ungrammaticality of the sentences in (49) and (50) are due to island effects. (49)

*自己 1 的身份證 2,每位公民 1 必須要更新_____2 的指定機構在漢 口。 * Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen gengxin replace zai in

ID.card _____2 Hankou. Hankou

de C

every-Cl citizen must zhiding jigou designated service

(‘His1 own ID card2, the official service [where every citizen1 is supposed to replace _____2 with a new one] is in Hankou District.’)

  Reconstruction and internal structures of resumptive pronouns (50)

* 自己 1 的父母 2,每位子女 1 都能完全理解_____2 的家庭還是不多的。 * Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen parents every-Cl child all wanquan lijie _____2 de entirely understand DE haishi bu duo de. yet Neg many DE

can jiating family

(‘His1 own parents2, the families [in which every child1 can perfectly understand _____2] are not many yet.’) Now let us turn to the cases of intrusive pronouns. I replaced the gaps in the original sentences (49–50) with intrusive pronouns in (51–52) respectively. (51) confirms that his own ID card cannot be bound by the universal quantified expression every citizen and therefore, the former cannot be interpreted as a bound anaphor nor can it get a distributive reading. (51)

?*自己 1 的身份證 2,每位公民 1 必須要把它 2 更新的指定機構在漢口。 ?* Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen ba BA zai in

ID.card ta32 gengxin 3OSg replace Hankou. Hankou

every-Cl de C

citizen zhiding designated

must jigou service

(‘His1 own ID card2, the official service [where every citizen1 is supposed to replace *(it2) with a new one] is in Hankou District.’ (52) also confirms the unavailability of distributive reading of intrusive pronouns in island contexts. (52)

?* 自己 1 的父母 2,每位子女 1 都能完全理解他們 2 的家庭還是不多的 ?* Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen wanquan entirely haishi yet

parents every-Cl lijie understand bu duo Neg many

child ta1-men2 3Pl de. DE

all de DE

can jiating family

(‘His1 own parents2, the families [in which every child1 can perfectly understand them2] are not many yet.’)

Reconstruction effects in Chinese  

In both cases, intrusive pronouns take a reduced form without the presence of any NP as internal structure and therefore there is no NP which is reconstructed, as illustrated in (51’) and (52’) respectively. As a result, distributive readings are not available. (51’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate)]  * [DP [ everyi (∀)] citizen ] ... [DP D ϕ]

(vacuous quantification)

(52’) ta1-men ‘them’ = [DP D ϕ (3rd, Pl., human)]  * [DP [everyi (∀)] child ] ... [DP D ϕ]

(vacuous quantification)

... Summary In this section, I examined the possibility to reconstruct a relativized or a dislocated antecedent under the scope of a universal quantifier. The relevant result is presented in the following table. Tab. 3: Quantifier scope reconstruction Relatives

yes

(g)

yes

no

no

[[D-ϕ] NP]

[[D-ϕ] NP]

(e)

[D-ϕ]

yes

Intrusives

(d)

[D-ϕ]

Gaps

(b)

Resumptives

Resumptives

(a)

[[D-ϕ] NP]

Internal structure

Gaps

(i) Quantifier scope reconstruction?

Dislocations

A gap always takes an extended form [[D-ϕ] NP] in relative clauses and in dislocation structures. Such a form ensures that the NP will always be reconstructed with the head D° and that the universal quantifier binds this NP as a variable by providing it with a distributive reading. The presence of an NP in the internal structure of a gap or of a resumptive pronoun is crucial for the availability of a distributive reading.

  Reconstruction and internal structures of resumptive pronouns A resumptive pronoun also takes an extended form [[D-ϕ] NP] in relatives, which makes a distributive reading available. After reconstruction, the NP will fall under the scope of a universal quantifier. From this point of view, in relatives, a gap behaves exactly like a resumptive pronoun. In dislocation structures, resumptive pronouns and intrusive pronouns always take a reduced form [D-ϕ] as their internal structures. This form does not allow them to get any distributive reading with regard to the universal quantifier. Without the presence of any NP on the reconstructed site, the head D° alone cannot be bound by the universal quantifier as variable due to the lack of a nominal restriction. My generalization based on the survey in this section is that a resumptive pronoun can possess two different internal structures according to the syntactic constructions in which it appears. This generalization is crucial for my core analysis in this study and I will come back to it later in this chapter.

.. Possibility to take a quantificational antecedent The second test is to check whether or not a gap, a resumptive pronoun and an intrusive pronoun are able to take a quantified expression as their antecedents. The configuration of the test is the following: (53) [CP QPi [TP … … [DP RPi /____i ]]] The hypothesis is that if a gap or a resumptive pronoun can take a quantified phrase as its antecedent, then it possesses an extended form [[D-ϕ] NP] as its internal structure. In (53) if there is an NP that is present on the variable site, occupied either by a resumptive pronoun or by a gap, the QP can bind this NP as a variable. Note again that NP is treated as a nominal restriction of a bound variable. ... Relatives In relative clauses, a gap can take a quantified phrase as its antecedent, as shown in (54). Example (54) can have a reading in which the interpretation of the gap, xuexiao hui zhunxu bu-yong bukao de xuesheng ‘the student who would be exempted from the make up exam by the school’, varies according the interpretation of the universal quantified expression mei-ge bu-jige de xuesheng ‘every student that failed the exam’. This means that a gap can take a universal quantified expression as its antecedent and therefore, it takes an extended form [[D-ϕ] NP] as its internal structure. The presence of an NP in this structure triggers a distributive reading.

Reconstruction effects in Chinese  

(54)

[張三以為[學校會准許[___j 不用補考]]]的每個不及格的學生 j 結果都 留級了。 [Zhangsan yiwei [xuexiao hui zhunxu [ ____j Zhangsan bu-yong no.need xueshengj student

think school bu-kao ]]] make.up.exam jieguo dou finally all

will permit de mei-ge bujige C every-Cl failed liuji le. stay.grade SFP

de DE

‘Every studentj [whoj failed the exam [that Zhangsan thought that school would permit [that ___j is exempted from the make up exam]]] will finally not go up to the next year.’ Let us examine another example. (55)

[張三以為[學校會准許[他 j 不用補考]]]的每個不及格的學生 j 結果都 留級了。 [Zhangsan yiwei [xuexiao hui zhunxu [ ta1j Zhangsan bu-yong no.need xueshengj student

think school bu-kao ]]] make.up.exam jieguo dou finally all

will permit 3MSg de mei-ge bujige de C every-Cl failed DE liuji le. stay.grade SFP

‘Every studentj [whoj failed the exam [that Zhangsan thought that school would permit [that hej is exempted from the make up exam]]] will finally not go up to the next year.’ Similar situation can be found in (55), which shows that a resumptive pronoun can also take a quantified expression as its antecedent in relatives. As a result, the resumptive pronoun in (55) also takes an extended form [[D-ϕ] NP] as its internal structure. With the presence of an NP, the quantificational antecedent every student that failed the exam can bind the resumptive pronoun ta1 ‘he’ as variable. ... Dislocation structures In dislocation structures, a gap can always take a quantificational antecedent, as shown in (56). The interpretation of the gap, which was occupied by taofan ‘escaped criminal’, depends on the interpretation of the universal quantified phrase mei-ge ‘every’. This implies that a gap in a dislocation structure can always take the extended form [[D-ϕ] NP] as its internal structure.

  Reconstruction and internal structures of resumptive pronouns (56)

每個逃犯 j,張三還天真地以為警察都抓住了_____j。 Mei-ge taofanj, Zhangsan hai tianzhen-de every-Cl escaped.criminal Zhangsan still naively yiwei jingcha dou zhua-zhu-le _____j. think policeman all arrested.Perf ‘As for every escaped criminalj, Zhangsan still thought naively that policemen had arrested _____j.’

By contrast, a resumptive pronoun in a dislocation structure cannot take a quantificational antecedent as illustrated in (57) and (58). Concretely, the interpretation of the personal pronoun ta1 ‘him’ takes neither the universal quantified phrase every escaped criminal nor every student who failed his exam as its antecedent in these two examples respectively. As a result, in this context, a pronoun takes a reduced form [D-ϕ] as its internal structure. (57)

?* 每個逃犯 j,張三還天真地以為警察把他 j 逮捕歸案了。 ?* Mei-ge taofanj, Zhangsan hai tianzhen-de every-Cl escaped.criminal Zhangsan still naively yiwei jingcha ba ta1j daibu-gui’an le. think policeman BA 3MSg arrested SFP (‘As for every escaped criminalj, Zhangsan still thought naively that policemen had arrested himj.’)

(58)

?*每個不及格的學生 j,張三還天真地以為學校會准許他 j 不用補考。 ?* Mei-ge bujige de xueshengj, Zhangsan hai every-Cl failed DE student tianzhen-de yiwei [xuexiao hui naively thnik school will [ta1j bu-yong bu-kao]]. 3MSg no.need make.up.exam

Zhangsan zhunxu permit

still

(‘As for every studentj who failed the exam, Zhangsan still thought naively [that the school will permit [that hej is exempted from the make up exam]].’) Example (59) illustrates the case of intrusive pronouns used in a sentence containing a complex-NP island. The insertion of the resumptive pronoun ta1 ‘he’ seems to save the sentence from violation of locality conditions; however, as I explained in great detail in Chapter 2, this is not true. Recall that it is Match that derives such a dependency here and Match is an island-free operation. The

Reconstruction effects in Chinese  

derivation of a resumptive dislocation structure itself is never subject to locality constraints. Instead, the problem of (59) is that the interpretation of he does not depend on the universal quantified phrase every escaped criminal. Since this interpretation is the only possible reading of this sentence according to the logic of the context, when the only possible reading is not available, the sentence becomes ungrammatical. Crucially, this sentence is not filtered by island condition but by the impossibility of establishing a distributive reading of the relevant intrusive pronoun. (59)

* 每個逃犯 j,追捕他 j 所需的時間是一個月。 * Mei-ge taofanj, [zhui-bu

ta1j

every-Cl excaped.criminal pursue.capture 3MSg suo xu de] shijian shi yi-ge yue. SUO need DE time be one-Cl month (‘As for every escaped criminalj, the time needed for capturing himj is a month.’) Example (59) confirms that an intrusive pronoun takes a reduced form [D-ϕ] that cannot trigger any distributive reading. ... Summary The result of the test on the possibility to take a quantificational antecedent is presented in the following table. Tab. 4: Quantificational antecedent Relatives Gaps

Resumptives

Gaps

Resumptives

Intrusives

(ii) Quantificational antecedent?

Dislocations

(a)

(b)

(d)

(e)

(g)

yes

yes

yes

no

no

  Reconstruction and internal structures of resumptive pronouns Comparing this result with that of the previous section, it is not difficult to find a coherence between these two. The two tests lead to the same conclusion with regard to internal structures of a gap and of a resumptive pronoun, as shown in the following table. Tab. 5: Interim results Relatives

Resumptives

Intrusives

(d)

(e)

(g)

yes

yes

yes

no

no

yes

yes

yes

no

no

[D-ϕ]

[D-ϕ]

[[D-ϕ] NP]

Gaps

(b)

[[D-ϕ] NP]

(a)

[[D-ϕ] NP]

Internal structure

Resumptives

(ii) Quantificational antecedent?

Gaps

(i) Quantifier scope reconstruction?

Dislocations

.. Reconstruction of anaphoric binding In this section, I will continue with a test based on the possibility to reconstruct an anaphoric binding. ... Relatives In relatives, it is possible to establish an anaphoric binding between a relativized antecedent and a proper name embedded within a relative clause, as shown in (60). The pronoun ta3-ziji-de ‘his own’ in the antecedent NP, ta3-zijide huaping ‘his own vase’, can be bound by Zhangsan in spite of the fact that the latter does not c-command the former at the Surface-Structure.

Reconstruction effects in Chinese  

(60)

[張三 1 昨天晚上打碎了____2]的那個他 1 自己的花瓶 2 [Zhangsan1 zuotian wanshang da-sui-le

]

________2

Zhangsan yesterday evening break-Perf de na-ge ta11-ziji de huaping2 C that-Cl 3MSg-self DE vase ‘his1 own vase2 that Zhangsan1 broke _______2 last night’ The only way to establish this binding relation is by reconstructing the antecedent NP, his own vase, on the relativized site occupied by the gap, as illustrated in (60’). The gap takes an extended form [[D-ϕ] NP], which allows the antecedent NP to be reconstructed properly. As a result, Zhangsan anaphorically binds the pronoun his after the reconstruction. (60’) [____] = [DP D ϕ (3rd, Sg., inanimate) [NP his own vase]]  [TP Zhangsan1 ... [DP D ϕ [NP his1 own vase]]] The same type of reconstruction also applies to the case of intrusive pronouns embedded inside a relative clause, as shown in (61). (61)

[張三 1 昨天晚上把它 2 給打碎了]的他 1 自己的那個寶貝花瓶 2 [Zhangsan1 zuotian wanshang ba ta32 gei Zhangsan da-sui-le] break-Perf huaping2 vase

yesterday evening de ta11-ziji de C 3MSg-self DE

BA 3OSg GEI na-ge baobei that-Cl treasured

‘his1 own treasured vase2 that Zhangsan1 broke it2 last night’ In this example, the pronoun ta1-ziji-de ‘his own’ in the antecedent NP, ta1-zijide baobei huaping ‘his own treasured vase’, can be anaphorically bound by the proper name Zhangsan, which means that the pronoun his own is properly reconstructed on the relativized site occupied by the resumptive pronoun ta3 ‘it’. In this case, we can assume that the resumptive pronoun takes an extended form. Since the NP his own treasured vase is entirely reconstructed, the pronoun his own will be c-commanded by Zhangsan. As a result, the anaphoric binding relation can be established between these two. (62) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate) [NP his own treasured vase]]  [TP Zhangsan1 ... [DP D ϕ (3rd, Sg., inanimate) [NP his1 own treasured vase1]]]

  Reconstruction and internal structures of resumptive pronouns The following two examples also confirm this observation with regard to gaps and resumptive pronouns in the same context. (63)

[張三 1 一直____2 珍藏在皮夾裡]的他 1 過世的太太的照片 2 [Zhangsan1 yizhi ____2 zhencang zai pijia-li] Zhangsan always keep.preciously at purse-in de ta11 guoshi-de taitai de zhaopian2 C 3MSg-Gen pass.away-DE wife DE picture ‘the picture2 of his1 wife who passed away that [Zhangsan1 always keeps ____2 preciously in his purse]’

(63’) [____]= [DP D ϕ (3rd, Sg., inanimate) [NP the picture of his wife who passed away]]  [TP Zhangsan1 ... [DP D ϕ (3rd, Sg., inanimate) [NP the picture2 of his1 wife who passed away]]] In (63), the gap takes an extended form that is fully reconstructed on the relativized site, which gives rise to anaphoric binding effects between the pronoun ta1 ‘he’ in the antecedent NP and Zhangsan inside the relative clause. The same binding relation can also be established between Zhangsan and his in (64) when the resumptive pronoun ta3 ‘it’ takes an extended form after the reconstruction. (64)

[張三 1 一直把它 2 珍藏在皮夾裡]的他 1 過世的太太的照片 2 [Zhangsan1 yizhi ba ta32 zhencang Zhangsan always zai pijia-li] de at purse-in C taitai

de

BA

3OSg ta11 3MSg-Gen

keep.preciously guoshi-de pass.away-DE

zhaopian2

wife DE picture ‘the picture2 of his1 wife who passed away that [Zhangsan1 always keeps it2 preciously in his purse]’ (64’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate) [NP the picture of his wife who passed away]]  [TP Zhangsan1 ... [DP D ϕ (3e, Sg., inanimate) [NP the picture2 of his1 wife who passed away]]]

Reconstruction effects in Chinese  

... Dislocation structures In a dislocation structure such as (65), the anaphoric expression ta1-ziji-de ‘his own’ can be fully reconstructed in the position occupied by the gap. Under the reconstruction, Zhangsan can anaphorically bind the pronoun his. (65)

他 1 自己的花瓶 2 , 張三 1 昨天晚上(給)打碎了____2。 Ta11-ziji de huaping2 , Zhangsan1 zuotian 3MSg-self DE vase Zhangsan yesterday wanshang (gei) da-sui-le _____2. evening GEI break-Perf ‘As for his1 own vase2, Zhangsan1 broke _____2 last night.’

We can assume that the gap here always takes an extended form [[D-ϕ] NP] and that the anaphoric binding relation is established after the reconstruction of the antecedent NP ta1-ziji-de huaping ‘his own vase’, as illustrated in (65’). (65’) [____] = [DP D ϕ (3rd, Sg., inanimate) [NP his own vase]]  [TP Zhangsan1 ... [DP D ϕ (3rd, Sg., inanimate) [NP his1 own vase]]] However, it is very difficult to establish the same type of anaphoric binding relation in the case of resumptive pronoun. My informants do not have the same judgment on the sentences of this type; however, they are definitely uncomfortable with these sentences. Therefore, it is possible to assume that the pronoun in this case takes a reduced form [D-ϕ]. Without the reconstruction of NP his own vase, anaphoric binding relation cannot be established between the pronoun his and Zhangsan, as illustrated in (66). (66)

? 他 1 自己的花瓶 2 , 張三 1 昨天晚上把它 2 給打碎了。 ? Ta11-ziji de huaping2 , Zhangsan1 zuotian 3MSg-self DE vase Zhangsan yesterday wanshang ba ta32 gei da-sui-le. evening BA 3OSg GEI break-Perf (‘As for his1 own vase2, Zhangsan1 broke it2 last night.’)

(66’) ta3 ‘it’ = [DP D ϕ (3rd, Sg., inanimate)]  * [TP Zhangsan1 ... [DP D ϕ(3rd, Sg., inanimate)]] Example (67) confirms this observation.

  Reconstruction and internal structures of resumptive pronouns (67)

? 對她 1 孩子不公正的評語 2,王麗 1 已經把它 2 告訴了學校校長。 ? Dui ta21-haizi bugongzheng-de pingyu2, Wangli1 to 3FSg-kid unfair comment Wangli yijing ba ta32 gaosu-le xuexiao xiaozhang. already BA 3OSg tell-Perf school principal (‘As for the unfair comment2 on her1 kid, Wangli1 has told it2 to the principal of the school.’)

In dislocation structures, an intrusive pronoun also resists the same type of reconstruction of anaphoric binding relation. In (68) the pronoun ta3 ‘it’ is inserted inside an adjunct island. The sentence does not violate locality constraints because the relevant dependency is established by Match which is an island-free operation. However, the anaphoric binding relation cannot be established between Wangli and the pronoun her in the comment on her kid. Therefore, we can assume that the pronoun ta3 takes a reduced form [D-ϕ] and the lacking of an NP in this structure makes it impossible to reconstruct the comment on her kid on the dislocated site, which leads to the impossible anaphoric binding. (68)

? 對她 1 孩子不公正的評語 2,[在王麗 1 把它 2 告訴學校校長之前],學 校還很平靜。 ? Dui ta21-haizi bugongzheng-de pingyu2, [zai to 3FSg-kid unfair comment at Wangli1 ba ta32 gaosu xuexiao xiaozhang Wangli BA 3OSg tell school principal zhiqian], xuexiao hai hen pingjing. before school still very peaceful (‘As for the unfair comment2 on her1 kid, [before Wangli1 told it2 to the principal of the school], the school was still in peace.’)

... Summary The result of the third test on anaphoric binding is shown in the table 6. Based on the result, we will draw the same conclusion concerning the internal structure of a gap and of a resumptive pronoun as in the previous section, which is given in the table 7.

Reconstruction effects in Chinese   Tab. 6: Reconstruction of anaphoric binding Relatives Gaps

Resumptives

Gaps

Resumptives

Intrusives

(iii) Reconstruction of anaphoric binding?

Dislocations

(a)

(b)

(d)

(e)

(g)

yes

yes

yes

no?

no?

Tab. 7: Interim results Relatives

Dislocations Gaps

Resumptives

Intrusives

(d)

(e)

(g)

yes

yes

yes

no

no

(ii) Quantificational antecedent?

yes

yes

yes

no

no

(iii) Reconstruction of anaphoric binding

yes

yes

no?

no?

[D-ϕ]

[D-ϕ]

(i) Quantifier scope reconstruction?

[[D-ϕ] NP]

Internal structure

yes

[[D-ϕ] NP]

Resumptives (b)

[[D-ϕ] NP]

Gaps (a)

  Reconstruction and internal structures of resumptive pronouns .. Condition C effects under reconstruction In this section, I will examine whether effects due to violation of condition C are observed under reconstruction. ... Relatives In relatives with a gap (cf. 69), after the reconstruction of the NP Yiqin de lunwen ‘Yiqin’s article’ on the relativized site occupied by a gap, the ungrammaticality of the sentence implies that condition C effects show up. (69)

* [我覺得[她 1 可以____2 發表在國家一級期刊上]]的一勤 1 的論文 2 * [wo juede [ta21 keyi ____2 fabiao zai 1Sg think 3FSg guojia yi-ji national first-rate de lunwen2 DE article

can qikan-shang]] journal-on

publish at de Yiqin1 C Yiqin

(‘Yiqin1’s article2 [that I think [that she1 can publish ____2 in firstrate national journals]]’) If condition C effects are observed, then we can assume that NP Yiqin’s article is entirely reconstructed and under the reconstruction, the pronoun ta2 ‘she’ directly c-commands the proper name Yiqin and both of them share the same index, which violates condition C of the Binding theory. As a result, we can assume that gaps always take an extended form [[D-ϕ] NP] in this case, as illustrated in (69’). (69’) [____]= [[D ϕ (3rd, Sg., inanimate)] [NP Yiqin’s article]] Condition C effects: * [TP she1 can publish [DP [D ϕ] [NP Yiqin1’s article]]] Example (70) confirms that condition C effects are always observed when it is a resumptive pronoun that occupies the relativized site. This implies that the pronoun ta3 ‘it’ has an extended form [[D-ϕ] NP] in this sentence. After the reconstruction of the NP Yiqin’s article on the site occupied by the resumptive pronoun ta3 ‘it’, the NP containing the proper name Yiqin will be directly bound by the subject pronoun ta2 ‘she’ that shares the same index, which of course triggers condition C effects.

Reconstruction effects in Chinese  

(70)

* [我覺得[她 1 可以把它 2 發表在國家一級期刊上]]的一勤 1 的論文 2 * [wo juede [ta21 keyi ba ta32 fabiao zai 1Sg think 3FSg guojia yi-ji national first-rate de lunwen2 DE article

can BA 3OSg qikan-shang]] de journal-on C

publish Yiqin1 Yiqin

at

(‘Yiqin1’s article2 [that I think [that she1 can publish it2 in first-rate national journals]]’) (70’) ta3 ‘it’ = [[D ϕ(3rd, Sg., inanimate)] [NP Yiqin’s article]] Condition C effects: * [TP she1 can publish [DP [D ϕ] [NP Yiqin1’s article]]] In order to check whether our observation is on the right track, here is another pair of examples. (71)

* [她 1 已經_____2 交到簽證處]的一勤 1 的照片 2 * [ta21 yijing _____2 jiao-dao qianzheng-chu] 3FSg already give-to service.of.visa de Yiqin1 de zhaopian2 C Yiqin DE picture (‘Yinqin1’s picture2 [that she1 has already given ____2 to the service of visa]’)

(71’) [____]= [[D ϕ (3rd, Sg., inanimate)] [NP Yiqin’s picture]] Condition C effects: * [TP she1 has already given [DP [D ϕ] [NP Yiqin1’s picture]]] (72)

* [她 1 已經把它 2 交到簽證處]的一勤 1 的照片 2 * [ta21 yijing ba ta32 jiao-dao qianzheng-chu] 3FSg already BA 3OSg give-to service.of.visa de Yiqin1 de zhaopian2 C Yiqin DE picture (‘Yinqin1’s picture2 [that she1 has already given it2 to the service of visa]’)

  Reconstruction and internal structures of resumptive pronouns (72’) ta3 ‘it’ = [[D ϕ (3rd, Sg., inanimate)] [NP Yiqin’s picture]] Condition C effects: * [TP she1 has already given [DP [D ϕ] [NP Yiqin1’s picture]]] In all of the above sentences, both the gap in (71) and the resumptive pronoun ta3 ‘it’ in (72) take an extended form and the NP antecedents are always fully reconstructed on relativized sites. After the reconstruction, proper names are bound by the pronouns that bear the same indices, which leads to the violation of condition C. ... Dislocation structures Let us turn to dislocation structures and observe the following example, (73)

?? 小倩 1 的論文 2 , 我覺得她 1 可以____2 發表在國家一級期刊上。 ?? Xiaoqian1 de lunwen2, wo juede ta21 keyi Xiaoqian DE article 1Sg think 3FSg can ____2 fabiao zai guojia yi-ji qikan-shang. publish at national first.rate journal-on (‘As for Xiaoqian1’s article2, I think that she1 can publish ____2 in firstrate national journals.’)

Native speakers do not give the same judgment on this sentence and some of them find it extremely difficult to establish an anaphoric relationship between the subject pronoun ta2 ‘she’ inside the TP and the reconstructed NP Xiaoqian. I assume that such a difficulty is due to the fact that reconstruction of the NP Xiaoqian’s article gives rise to condition C effects. Along this line, a gap always takes an extended form as its internal structure and the proper name Xiaoqian, which is fully reconstructed in the position occupied by the gap, is ccommanded by the subject pronoun she, and the resulting binding leads to the violation of condition C. (73’) [____]= [[D ϕ (3rd, Sg., inanimate)] [NP Xiaoqian’s article]] Condition C effects: * [TP she1 will publish [DP [D ϕ] [NP Xiaoqian1’s article]]] By contrast, in the same context, a resumptive pronoun does not give rise to the same condition C effects, as illustrated in (74).

Reconstruction effects in Chinese  

(74)

小倩 1 的論文 2 , 我覺得她 1 可以把它 2 發表在國家一級期刊上。 Xiaoqian1 de lunwen2, wo juede ta21 keyi ba Xiaoqian DE article 1Sg think 3FSg can BA ta32 fabiao zai guojia yi-ji qikan-shang. 3OSg publish at national first.rate journal-on ‘As for Xiaoqian1’s article2, I think that she1 can publish it2 in first-rate national journals.’

The grammaticality of (74) suggests that after the reconstruction, the anaphoric binding relation between the proper name Xiaoqian and the subject pronoun she is not established and as a result, the pronoun ta2 ‘she’ does not c-command the proper name. In this case, it is possible to assume that the resumptive pronoun ta3 ‘it’ takes a reduced form [D-ϕ] without the presence of any NP in its internal structure. The absence of an NP cannot trigger condition C effects under reconstruction. (74’) ta3 ‘it’ = [D ϕ (3rd, Sg., inanimate)] Absence of condition C effects: [TP she1 can publish [DP D ϕ]] Here is another example in support of my assumption. (75)

張教授 1 的孩子們 2 , 我發覺他 1 不認為他們 2 作風有問題。 Zhang jiaoshou1 de haizi-men2, wo fajue Zhang ta11 3MSg you have

professor DE bu renwei Neg think wenti. problem

kid-Pl 1Sg think ta1-men2-de zuofeng 3MPl-Gen behavior

‘As for Professor Zhang1’s kids2, I feel that he1 does not think that their2 behavior is problematic.’ (75’) ta1-men ‘their’= [D ϕ (3rd, Pl., human)] Absence of condition C effects [TP he1 does not think [DP D ϕ (3rd, Pl., human)] behavior problematic] Let us turn to cases of intrusive pronouns now. Here are two sentences involving strong islands. The ungrammaticality of these sentences is due to the presence of a gap inside islands.

  Reconstruction and internal structures of resumptive pronouns (76)

* 張三 1 的學生們 2,[他 1 罵了_____2 一頓]的那一天正好是學校的校 慶日。 * Zhangsan1 de xuesheng-men2 , [ta11 ma-le Zhangsan DE ____2 yi-dun] one-Cl.v shi xuexiao be school

student-Pl 3MSg de na yi tian C that one day de xiaoqingri. DE anniversary

scold-Perf zhenghao just

(‘Zhangsan1’s students2, the day [when he1 scolded _____2] was school’s foundation anniversary.’) (77)

* 張三 2 的未婚妻 1,他 2 明天必須要見到_____1 的地方是天主教 堂。 * Zhangsan2 de weihunqi1, ta12 mingtian bixu-yao Zhangsan DE fiancée 3MSg tomorrow must jiandao ____1 de difang shi tianzhu-jiaotang. voir DE place be catholic.church (‘As for Zhangsan2’s fiancée1, the place [where he2 must meet ___1 tomorrow] is the church.’)

(78)

張三 1 的學生們 2,[我看到他 1 罵了他們 2 一頓]的那一天正好是學校 的校慶日。 Zhangsan1 de xuesheng-men2 , [wo kan-dao ta11 Zhangsan DE student-Pl 1Sg ma-le ta1-men2 yi-dun] de scold-Perf 3MPl one-Cl.v C tian zhenghao shi xuexiao de day just be school DE

see

3MSg na yi that one xiaoqingri. anniversary

‘Zhangsan1’s students2, the day [when I saw that he1 scolded them2] was the foundation anniversary of the school.’ (79)

張三 2 的未婚妻 1,他 2 明天必須要見到她 1 的地方是天主教堂。 Zhangsan2 de weihunqi1, ta12 mingtian bixu-yao Zhangsan DE fiancée 3MSg tomorrow must jiandao ta21 de difang shi tianzhu-jiaotang. voir 3FSg DE place be catholic.church ‘As for Zhangsan2’s fiancée1, the place [where he2 must meet her1 tomorrow] is the church.’

Reconstruction effects in Chinese  

Based on (76–77), I examine the condition C effects with the presence of intrusive pronouns inside islands in (78–79). The grammaticality of these two examples in (78–79) suggests that on the one hand, the presence of intrusive pronouns does not trigger condition C effects under reconstruction and accordingly, I assume that intrusive pronouns take a reduced form as their internal structures in both sentences. In (78), since the antecedent NP Zhangsan de xuesheng ‘Zhangsan’s students’ is not reconstructed, it is not c-commanded by the subject pronoun, ta1 ‘he’, inside the island, and this leads to an absence of condition C effects (cf. 78’). Similarly, in (79), the antecedent NP Zhangsan de weihunqi ‘Zhangsan’s fiancée’ is not reconstructed either and as a result, it cannot be bound by the subject pronoun ta1 ‘he’. Therefore, no condition C effect is detected (cf. 79’). (78’) ta3-men ‘them’= [D ϕ (3rd, Pl., human)] Absence of condition C effects: [TP he1 scolded [DP D ϕ]] (79’) ta2 ‘her’= [D ϕ (3rd, Sg., human, F)] Absence of condition C effects: [TP he1 must meet [DP D ϕ] tomorrow…]

... Summary The following table summarizes the results in this section. Tab. 8: Condition C effects Relatives Gaps

Resumptives

Gaps

Resumptives

Intrusives

(iv) Condition C effects?

Dislocations

(a)

(b)

(d)

(e)

(g)

yes

yes

yes

no

no

  Reconstruction and internal structures of resumptive pronouns .. Summary Let me put all of the results based on the four tests in this section in the following table. The word “Internal structure” is understood as proposed internal structure even if Chinese does not have morphological evidence for these proposed forms. Therefore, one has to keep in mind that in this table, the first four rows show the observations, and the last row shows the proposed theory to explain the observations. Apparently, there is coherence in these results. Tab. 9: Results Relatives

Dislocations Resumptives

Intrusives

(e)

(g)

yes

yes

yes

no

no

(ii) Quantificational antecedent?

yes

yes

yes

no

no

(iii) Reconstruction of anaphoric binding?

yes

yes

yes

no?

no?

(iv) Condition C effects?

yes

yes

yes

no

no

[D-ϕ]

[D-ϕ]

(i) Quantifier scope reconstruction?

Internal structure

[[D-ϕ] NP]

Gaps (d)

[[D-ϕ] NP]

Resumptives (b)

[[D-ϕ] NP]

Gaps (a)

The tests in this section are based on the hypothesis that there is a close relationship between reconstruction effects and internal structure of resumptive

Reconstruction effects in Chinese  

pronouns. Only an extended form [[D-ϕ] NP] can give rise to different interpretative effects under reconstruction, such as bound variable reading and anaphoric binding construal. Similarly, it is also this extended form that triggers condition C effects under reconstruction. All of these effects are due to the presence of an NP in the internal structure. By contrast, the reduced form lacking an NP, [D-ϕ], does not give rise to any of these effects. Let us have a comparative survey on different categories in the table above. Observe the table vertically column by column. Each column represents result of the four tests for the same element in a specific type of A'-dependency. For instance, columns (a) and (b) represent gaps and resumptive pronouns in relatives respectively; (d), (e) and (g) represent gaps, resumptive pronouns and intrusive pronouns in dislocation structures. Recall that intrusive pronouns do not exist in relatives, and that is why there is no corresponding column in this table. First, let us begin by comparing column (a) with column (b). After the reconstruction in relatives, both a gap and a resumptive pronoun are, on the one hand, capable of satisfying interpretative requirements with regard to the scope of a universal quantifier and the anaphoric binding relation, and on the other hand, capable of displaying condition C effects. This coherence leads me to assume that gaps and resumptive pronouns in this context must have an extended form [[D-ϕ] NP] as their internal structures. Observing the result of these tests, no difference is detected between these two, and this suggests that in relatives, a gap and a resumptive pronoun behave in exactly the same way. Again, maybe this superficial resemblance will lead to an incorrect generalization according to which, a resumptive pronoun can be analyzed as a spelled out trace. As will be detailed later in this chapter, this superficial generalization is due to a pure coincidence and in fact, a resumptive pronoun can never be analyzed as a spelled out trace despite that it can sometimes have a similar behavior to a trace in certain syntactic constructions. Secondly, if we compare column (a) with column (d), we find that there is no difference between a gap in relatives and a gap in dislocation structures, both of which give rise to interpretative effects and condition C effects under reconstruction. This leads me to assume that both of them have the same internal structure, i.e. an extended form. More importantly, this observation implies that gaps have only one available form, [[D-ϕ] NP], without making any difference between syntactic structures in which they appear. This generalization confirms what we traditionally say about traces: traces left after A'-movement are essentially NPs. Furthermore, comparing (e) with (g) we find that there is no difference between a resumptive pronoun and an intrusive pronoun in a dislocation struc-

  Reconstruction and internal structures of resumptive pronouns ture in that neither of them displays any reconstruction effect. Neither interpretative effect nor condition C effect is detected. This observation leads to the generalization according to which, both types of resumptive pronouns have a reduced form [D-ϕ] as their internal structures. Let us then compare column (d) with columns (e/g). We find that a gap does not have the same behavior as a resumptive pronoun in dislocation structures. A gap has an extended form as its internal structure, whereas a resumptive pronoun or an intrusive pronoun cannot be analyzed as a spelled out trace in dislocation structures in Chinese. This difference highlights a contrast existing between a relative clause and a dislocation structure. As I explained above, resumptive pronouns do not behave uniformly in relatives and in dislocation structures. The most important contrast shown in this table is between (b) and (e/g). A resumptive pronoun has an extended structure in relatives but a reduced structure in dislocations. This implies that resumptive pronouns constructed on the basis of ta (3rd person) potentially have two different forms as their internal structures. One question that I will discuss is whether this type of pronoun is naturally ambiguous between two different syntactic structures. If the answer is a “yes”, then we should find out how to choose the correct structure to use in a given syntactic dependency. If the answer is a “no”, then we will follow the idea that between the two available forms, one is a marked form while the other is a default form. Recall that this second option is exactly what Rouveret (to appear) argues for. Deviating from Rouveret’s original proposal, I will argue for the first hypothesis based on the observation on Chinese data. I assume that resumptive pronouns of ta-type naturally have two different forms; it is precisely derivational mechanisms that derive the relevant syntactic constructions in which these pronouns appear that select one of the two available internal structures. I will explain this point in detail in the next section.

. Analysis .. Correlation between types of syntactic constructions and forms of resumptives In this section, I will concentrate on the correlation between internal structures of pronouns and of gaps on the one hand and reconstruction effects on the other hand. One of the generalizations that I presented at the end of the previous section is very surprising: pronouns of ta-series can have two different forms as internal structures. When they take an extended form, all of the interpretative

Analysis  

effects under reconstruction can be detected; by contrast, when they take a reduced form, none of these effects shows up. As I said earlier, a crucial question, which is also a very general one, is how to select one of the two available forms of the pronoun in a given A'-dependency. Logically, I do not think that such a choice is completely arbitrary. The position that I take in this study is that it is grammar that has precise constraints on the choice of these two structures. In Rouveret (2002, 2008)’s analysis, the same question was posed. He suggests that between these two structures, one is a marked form and the other is a default form. However, facts observed in Chinese lead us to pursue another possibility to explain such a choice. A pronoun like ta3 is only ambiguous when we examine its behavior globally; however, it is never ambiguous in a given syntactic construction. Concretely, it always takes an extended form in relatives and a reduced form in dislocation structures. This contrast makes us assume that types of syntactic constructions constitute precise criteria that determine the choice of the form of a pronoun to use. Relatives (with a gap or with a resumptive pronoun) and dislocation structures with a gap require an extended form [[D-ϕ] NP], whereas resumptive dislocation structures prefer a reduced form [D-ϕ]. However, it is possible that dislocation structures do not exclude an extended form of a pronoun either since the truth is that they simply do not need the presence of an NP in the internal structure of a pronoun. This is the central hypothesis that I argue for in the rest of this section. It seems that the problem becomes even more complicated if we pursue this type of reasoning. The hypothesis that I make here implies that there is eventually a kind of interaction between syntactic constructions and internal structures of pronouns. Before continuing the discussion on this problem, it is crucial to establish a link between the conclusion that I made with regard to the derivation of relatives and dislocation structures in Chapter 2 on the one hand, and the general distribution of internal structures of pronouns presented in this chapter on the other hand. In order to make the comparison more direct, I combine two essential recapitulative tables of these two chapters into one in order to get a panoramic and synthetic view.

  Reconstruction and internal structures of resumptive pronouns Tab. 10: Final results Relatives

Dislocations Resumptive

Intrusive

---------

yes

yes

--------

ii) With island

yes

------

no (island effects)

yes

--------

yes (no island effects)

iii) Weak crossover effect

yes

yes

--------

yes

no

--------

iv) Strong crossover effect

yes

yes

---------

yes

no

--------

v) Quantificational antecedent

yes

yes

-----

yes

no

no

vi) Scope reconstruction

yes

yes

-----

yes

no

no

vii) Binding reconstruction

yes

yes

-----

yes

no?

no?

viii) Condition C effects

yes

yes

-----

yes

no

no

Gap

yes

(II) Internal structures

Intrusive

yes

(I) Derivational mechanisms

Resumptive

Gap

i) Without island

GB

Movement

Binding in discourse

MP

Agree without Move

Match without Agree

[[D-ϕ] NP]

[D-ϕ]

The tests from (i) to (iv) are syntactic ones in Chapter 2 and those from (v) to (viii) are semantic ones based on reconstruction effects. Conclusion (I) is for Chapter 2 based on the tests (i) to (iv) and conclusion (II) is for the present chapter based on the tests (v) – (viii). It is extremely important and interesting to

Analysis  

notice that the boundary of the two types of analyses for conclusion (I) and that for conclusion (II) coincide perfectly; however, I do not think that it is a pure coincidence here. Concretely, the boundary is located between the two categories of constructions: (A) relatives (with a gap or with a resumptive pronoun) and dislocation structures with a gap on the one hand and (B) dislocation structures with resumptive pronouns or with intrusive pronouns on the other hand. The generalizations that I make are the following: (i)

In the framework of the Government and Binding theory, structures in category (A) are derived by movement and those in (B) are derived by a discourse link without involving any movement.

(ii)

In the Minimalist Program, structures in category (A) are derived by Agree alone without necessarily involving Move and those in (B) are derived by Match without involving Agree.

(iii)

Pronouns and gaps in (A) have an extended form [[D-ϕ] NP] whereas pronouns in (B) have a reduced form [D-ϕ].32

Generalizations (i) and (ii) are about syntactic constructions where resumptive pronouns and gaps appear. Generalization (iii) is about internal structures of pronouns and gaps. A crucial question is to know what is the relationship between (i/ii) and (iii). Let me begin with the relationship between (i) and (iii). Recall one of the two possible analyses of relative clauses is based on a raising of the internal head of relatives. It is not difficult to see that the NP part in an extended form is a potential candidate for raising proposed in this type of analysis. NP raises from its base argument position inside a relative clause to the CP position. The nature of this raising is an A'-movement and it is subject to locality conditions, which is the reason why such a movement gives rise to island effects. If dislocation structures with a gap are also derived in the same way, we will be able to find a kind of coherence.

 32 An alternative way to understand such a generalization, as a reviewer correctly suggests, is that pronouns and gaps in (A) are phonological realizations of the structure [[D-ϕ] NP], whereas pronouns in (B) are phonological realizations of the structure [D-ϕ]. Since different structures may have the same phonological realization, it is thus not surprising that the same form is used in two ways.

  Reconstruction and internal structures of resumptive pronouns The extended form [[D-ϕ] NP] is generated inside a TP and the NP part in this structure undergoes topicalization and arrives at TopP. This topicalization also gives rise to island effects. Since the derivation of these constructions requires movement, they will certainly need an extended form with precisely the presence of an NP in the internal structure of the raised element. Following this reasoning, it is the derivation of relatives and dislocation structures with a gap that requires an extended form. In other words, it is precisely syntactic derivations of different A'-dependencies that select a specific form of an NP that occupies the relativized or dislocated site, as illustrated in (80). (80) a. Relatives (with a gap or with a RP): movement * NPi [CP …. … [island … [DP D ϕ NPi ]]]

b. Dislocation with a gap: movement [TopP … [island …… [DP D ϕ NPi ]]] * NPi,

On the other hand, in resumptive dislocation structures, topic is always basegenerated in [Spec, TopP] via Merge and the link between the NP topic and a resumptive pronoun can be established at the discourse level beyond the Narrow Syntax level, as illustrated in (81). (81) Dislocation with resumptives or intrusives: link at the discourse level NP, [TopP ………[island ……. [DP D ϕ]]] In the Minimalist Program, the relationship between generalizations (ii) and (iii) is even less obvious since movement is no longer an option in our analysis. In order to solve this problem, it is necessary to recall three crucial features that I discussed in great detail in Chapter 2: [ϕ], [λ] and EPP. We can assume that structures in category (A) that are subject to locality conditions bear the EPP feature and EPP requires that the specifier position of a CP should not be empty. A potential candidate that is able to satisfy the requirement of EPP is certainly the NP part present in internal structures. In other words, it is precisely the extended form of a resumptive pronoun or of a gap that can satisfy the requirement of the EPP feature.

Analysis  

By contrast, in dislocation structures with a resumptive pronoun or with an intrusive pronoun, we can assume that the EPP feature is not activated and that the presence of an NP is not required in internal structures of a pronoun. However, this assumption does not imply that the extended form is necessarily excluded from constructions in (B). Then we are wondering why relatives and dislocation structures with a gap bear the EPP feature and resumptive dislocations do not. Recall what I proposed concerning the EPP feature in these two constructions in Chapter 2. In the case of resumptive dislocations, positions like [Spec, CP] or [Spec, TopP] are always filled and if EPP is activated in these structures, EPP is always satisfied. In this situation, movement is not involved at all and more importantly, there is no potential candidate for movement and this is why island effects are not detected. This assumption is in the same line with the proposition of Chomsky (1995) that EPP is not introduced if there is no effect on the output in the computational system. From this perspective, there is no need to activate EPP in resumptive dislocations and as a result, the presence of an NP in the internal structure of a resumptive pronoun is not required. This is the reason why in resumptive dislocations, a resumptive pronoun can either have an extended form or a reduced form as its internal structure. In other words, there is no specific requirement on the form of internal structures of pronouns used in resumptive dislocations; both an extended form and a reduced form are permitted in such constructions. It seems that choice of the form of a pronoun in a given resumptive construction is determined by syntactic mechanisms that derive different types of A'-dependencies. An A'-dependency established by Agree always bears the EPP feature but that established by Match does not bear such a feature. In a resumptive construction derived by Match, the [Spec, CP] position is always occupied by a base-generated XP via Merge and the EPP feature is not activated in this case. (82) and (83) illustrate the relevant contrast. (82) a. Relatives (with a gap or with a RP): Agree [CP …… [island ……. [DP D ϕ NPi ]]] * NPi, (EPP)

b. Dislocation with a gap: Agree [Top’ ……[island ……. [DP D ϕ NPi ]]] * NPi, (EPP)

  Reconstruction and internal structures of resumptive pronouns (83) Resumptive dislocations: Match NP, [Top’ …… [island …….

[DP D ϕ]]]

.. Interaction of different components of grammar The analysis that I argue for can account for the boundary between structures in category (A) and those in category (B) in a logical way. – Structures in (A): (i) are derived by movement in the framework of GB; (ii)

are derived by Agree alone without Move in the framework of MP;

(iii)

bear the EPP feature;

(iv)

require an extended form [[D-ϕ] NP];

(v)

systematically give rise to island effects.

– Structures in (B): (i) are derived by a link at the discourse level without involving any movement operation in the framework of GB; (ii)

are derived by Match alone without Agree in the framework of MP;

(iii)

do not bear the EPP feature;

(iv)

do not require an extended form [[D-ϕ] NP];

(v)

do not give rise to island effects.

Now let us consider the problem from the point of view of resumptives. – A gap has only one internal structure: the extended form [[D-ϕ] NP]. As a result, (i)

it gives rise to interpretative effects under reconstruction in structures listed in category (A);

Analysis  

(ii)

it also gives rise to interpretative effects under reconstruction in structures listed in category (B).

– A resumptive pronoun based on ta-series has two internal structures: an extended form [[D-ϕ] NP] and a reduced form [D-ϕ], As a result, (i)

it can give rise to interpretative effects under reconstruction in structures in category (A) but only with its extended form [[D-ϕ] NP];

(ii)

it cannot give rise to interpretative effects under reconstruction in structures in category (B) because these structures prefer a reduced form [D-ϕ].

From this perspective, I tentatively suggest that there is an interaction of different factors involved in a resumptive construction: First, it is derivational mechanisms that select syntactic constructions on which they operate. Agree operates on structures in category (A) and Match on those in category (B). Second, it is not the specific form of a pronoun or of a gap that chooses a syntactic construction in which it appears. On the contrary, it is syntactic mechanisms which derive this A'-dependency that require specific forms of a DP occupying relativized/dislocated sites. This requirement remains independent of whether it is a gap or a resumptive pronoun or an intrusive pronoun that appears in this dependency. Concretely, derivational mechanisms determine which type of formal feature exists in which type of A'-dependency. For instance, complementizers in different A'-dependencies bear different features and namely, structures in category (A) derived by Agree bear the EPP feature and those in category (B) derived by Match do not bear EPP feature. EPP in structures listed in category (A) requires an extended form of the element located on variable sites. It is the presence or the absence of an NP that determines whether interpretative effects are triggered under reconstruction. My analysis is crucially different from the one proposed by Rouveret (2002, 2008) in that it is not the need of interpretation under reconstruction that requires the presence of an NP. My reasoning implies an interaction of the following factors:

  Reconstruction and internal structures of resumptive pronouns (i)

Derivational mechanisms;

(ii)

Types of A'-dependencies;

(iii)

Formal features;

(iv)

Specific forms of pronouns and of gaps;

(v)

Types of elements occupying the relativized/dislocated site do not play any role;

(vi)

Syntactic islands do not play any role.

In the next section, I will present an empirical argument in support of my analysis. .. Argument: two types of pronouns in Chinese In the previous section, I introduced my own hypothesis concerning an interaction of different factors involved in a resumptive construction in Chinese. The logic of my reasoning is the following. Derivational mechanisms in grammar choose types of constructions on which they operate; these mechanisms also determine which type of construction bears which type of formal feature. The presence or the absence of the EPP feature determines the presence or the absence of an NP in the internal structure of the DP located on the relativized/dislocated site. The nature of elements (i.e. gaps, resumptives or intrusives) in the variable position in a given A'-dependency does not determine anything. I also argue for a correlation between internal structures of a gap or of a resumptive pronoun and interpretative properties of different constructions. A gap can only have an extended form that gives rise to reconstruction effects in structures listed in category (A) as well as in those listed in category (B). A pronoun such as ta (3rd person) has two different forms, an extended form and a reduced form, and it gives rise to reconstruction effects in structures in (A) but not in (B). Here I will present another type of pronoun in Chinese, which does not have two internal structures, contrary to ta. It is the pronoun qi (其). I will examine properties of qi in this section.

Analysis  

... Distribution of the pronoun qi The pronoun qi is frequently used in classical Chinese and it is also used in literal contexts in modern Chinese. It is essentially a personal pronoun and a possessive pronoun of the third person (masculine, feminine, object, singular and plural). For instance, (84)

他這次沒有成功,其原因很複雜。 Ta1 zhe-ci meiyou chenggong, qi yuanyin 3MSg this.time Neg succeed 3.Gen reason hen fuza. very complicated ‘He has not succeeded this time and the reason (for which he has not succeeded) is complicated.’

(85)

她們這次得金牌與其一直以來努力訓練密不可分。 Ta2-men zhe-ci de jinpai yu qi 3FPl this.time obtain gold-medal with 3 yizhi yilai nuli xunlian mibukefen. always since make.effort train close.relation ‘That they got the gold medal is closely related to their hard daily training.’

The pronoun qi can also be used as a resumptive pronoun in Chinese. Let me begin by a detailed survey on the distribution of reconstruction effects with qi. .... Reconstruction of a quantifier scope In relatives, reconstruction of the scope of a universal quantifier is possible with qi. Examples (86–87) confirm that after the reconstruction, qi can get a distributive reading with regard to the universal quantified phrase mei ‘every’. (86)

[每個導演都會將其 j 看好幾遍]的一個電影預告片 j [mei-ge daoyan dou hui jiang qij

kan

each-Cl director all will JIANG 3 watch hao-ji-bian de yi-ge dianying-yugaopianj several.times C one-Cl movie.trailer ‘a movie trailerj [that each director will watch itj several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃)

  Reconstruction and internal structures of resumptive pronouns (87)

每個導演 1 都會將其 2 拿去參賽的他自己 1 的新片 2 [mei-ge daoyan1 dou hui jiang qi2

na

qu

each-Cl director all will JIANG 3 take to can-sai ] de ta31-ziji-de xin-pian2 participate.to.competition C 3MSg-self-DE new-movie ‘the latest movie2 of his1 own [that every director1 will take it2 to participate to a competition]’ In a dislocation structure, scope reconstruction is always possible, as shown in (88). The sentence can get a reading in which the NP antecedent his own (= self) latest movie varies according to the change of each director yielding a distributive reading. (88)

自己 1 的新片 2 , 每個導演 1 都會將其 2 拿去參賽。 Ziji1-de xin-pian2 , mei-ge daoyan1 dou hui self-DE new-movie each-Cl director all will jiang qi2 na qu can-sai. JIANG 3 take to participate.to.competition ‘The latest movie2 of his1 own, every director1 will take it2 to participate to a competition.’

Similarly, scope reconstruction is also possible for intrusive pronouns, as illustrated in (89–90). In these sentences, the resumptive pronoun qi is used intrusively since it is embedded inside complex-NP islands. With the presence of qi inside strong islands, the sentences do not give rise to any island effect. Also note that the relevant sentences can receive a distributive reading. (89) 自己 1 的身份證 2,每位公民 1 必須要*(將其 2)更新的指定機構在漢口。 Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen ID.card *(jiang qi2) gengxin JIANG 3 replace zai Hankou. in Hankou

every-Cl de C

citizen zhiding designated

must jigou service

‘His1 own ID card2, the official service [where every citizen1 is supposed to replace *(it2) with a new one] is in Hankou District.’

Analysis  

(90)

自己 1 的父母 2,每位子女 1 都能完全理解*(其 2 用心)的家庭已經不多了 Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen wanquan entirely jiating family

parents every-Cl child all lijie *(qi2 yongxin) understand 3 intention yijing bu duo le. already Neg many SFP

de DE

can

‘His1 own parents2, the families [in which every child1 can perfectly understand them2] are no longer many.’ .... Possibility to take a quantificational antecedent The pronoun qi can take a quantified expression as its antecedent both in relatives (cf. 91–92) and in dislocation structures (cf. 93–94). In (91–92), qi is reconstructed respectively as full quantified expressions, every escaped criminal and every student who failed the exam. (91)

[張三以為[警察把其 j 已經逮捕歸案]]的每個逃犯 j 都成功越獄了。 [Zhangsan yiwei [jingcha ba qij yijing Zhangsan daibu-gui’an ]] arrest chenggong successfully

think policemen de mei-ge C every-Cl yue yu break prison

BA 3 already taofanj escaped.criminals le. SFP

dou all

‘Every criminalj [that Zhangsan thought [that policemen had already arrested himj]] has successfully escaped from prison.’ (92)

[張三以為[學校會准許[其 j 不用補考]]]的每個不及格的學生 j 結果都留級 了。 [Zhangsan yiwei [xuexiao hui zhunxu [ qij Zhangsan bu-yong no.need xueshengj student

think school bu-kao ]]] make.up.exam jieguo dou finally all

will permit 3 de mei-ge bujige C every-Cl failed liuji le. stay.grade SFP

de DE

‘Every studentj [whoj failed the exam [that Zhangsan thought that school would permit [that hej is exempted from the make up exam]]] will finally not go up to the next year.’

  Reconstruction and internal structures of resumptive pronouns (93)

每個逃犯 j,張三還天真地以為警察把其 j 逮捕歸案了。 Mei-ge taofanj, Zhangsan hai tianzhen-de every-Cl escaped.criminal Zhangsan still naively yiwei jingcha ba qij daibu-gui’an le. think policeman BA 3 arrested SFP ‘As for every escaped criminalj, Zhangsan still thought naively that policemen had arrested himj.’

(94)

每個不及格的學生 j,張三還天真地以為學校會准許其 j 不用補考。 Mei-ge bujige de xueshengj, Zhangsan hai every-Cl failed DE student tianzhen-de yiwei [xuexiao hui naively thnik school will [qij bu-yong bu-kao]]. 3 no.need make.up.exam

Zhangsan zhunxu permit

still

‘As for every studentj who failed the exam, Zhangsan still thought naively [that the school will permit [that hej is exempted from the make up exam]].’ It is also possible for qi in its intrusive use to take a quantificational antecedent, as shown in (95). (95)

每個逃犯 j,追捕其 j 所需的時間是一個月。 Mei-ge taofanj, [zhui-bu

qij

every-Cl excaped.criminal pursue.capture 3 suo xu de] shijian shi yi-ge yue. SUO need DE time be one-Cl month ‘As for every escaped criminalj, the time needed for capturing himj is a month.’ .... Anaphoric binding reconstruction In relatives, anaphoric binding reconstruction is also possible for qi, as shown in (96–97). In (96), the pronoun ta1-ziji ‘himself’ in the NP antecedent of the relative clause, na-ge ta1-ziji de huaping ‘the vase of himself’, can be anaphorically related to the subject inside the relative clause, which means that such an antecedent NP is fully reconstructed on the relativized site occupied by the direct object resumptive pronoun qi ‘it’.

Analysis  

(96)

[張三 1 昨天晚上將其 2 打碎了]的那個他 1 自己的花瓶 2 [Zhangsan1 zuotian wanshang jiang Zhangsan yesterday evening JIANG da-sui-le] de na-ge ta11-ziji de break-Perf C that-Cl 3MSg-self DE ‘his1 own vase2 that Zhangsan1 broke it2 last night’

(97)

qi2 3 huaping2 vase

[王麗 1 已經將其 2 告知學校校長]的老師對她 1 孩子不公正的評語 2 [Wangli1 yijing jiang qi2 gaozhi xuexiao Wangli already JIANG xiaozhang] de laoshi principal C teacher bugongzheng-de pingyu2 unfair comment

3 dui to

tell ta21-haizi 3FSg.kid

school

‘the unfair comment2 from the teacher on her1 kid [that Wangli1 has already told it2 to the principal of the school]’ Such an anaphoric binding reconstruction is also possible for qi when it is used both as resumptive pronoun (cf. 98–99) and as intrusive pronoun (cf. 100) in dislocation structures. In this case, the topic NP phrase is fully reconstructed in the dislocated site occupied by qi. In both examples, qi takes an extended form with the presence of an NP as its internal structure. (98)

他 1 自己的花瓶 2 , 張三 1 昨天晚上將其 2 打碎了。 Ta11-ziji de huaping2 , Zhangsan1 zuotian 3MSg-self DE vase Zhangsan yesterday wanshang jiang qi2 da-sui-le. evening JIANG 3 break-Perf ‘As for his1 own vase2, Zhangsan1 broke it2 last night.’

(99)

對她 1 孩子不公正的評語 2,王麗 1 已經將其 2 告知了學校校長。 Dui ta21-haizi bugongzheng-de pingyu2, Wangli1 to 3FSg-kid unfair comment Wangli yijing jiang qi2 gaozhi-le xuexiao xiaozhang. already JIANG 3 tell-Perf school principal ‘As for the unfair comment2 on her1 kid, Wangli1 has already told it2 to the principal of the school.’

  Reconstruction and internal structures of resumptive pronouns (100)

對她 1 孩子不公正的評語 2,[在王麗 1 將其 2 告知學校校長之前],學 校還很平靜。 Dui ta21-haizi bugongzheng-de pingyu2, [zai to 3FSg-kid unfair comment at Wangli1 jiang qi2 gaozhi xuexiao xiaozhang Wangli JIANG 3 tell school principal zhiqian], xuexiao hai hen pingjing. before school still very peaceful ‘As for the unfair comment2 on her1 kid, [before Wangli1 told it2 to the principal of the school], the school was still in peace.’

.... Condition C effects under reconstruction Condition C effects are also observed in relatives with qi (cf. 101–102), which leads me to assume that qi always takes an extended form [[D-ϕ] NP] as its internal structure. (101)

* [我覺得[她 1 可以將其 2 發表在國家一級期刊上]]的一勤 1 的論文 2 * [wo juede [ta21 keyi jiang qi2 fabiao zai 1Sg think 3FSg guojia yi-ji national first-rate de lunwen2 DE article

can JIANG qikan-shang]] journal-on

3 de C

publish Yiqin1 Yiqin

at

(‘Yiqin1’s article2 [that I think [that she1 can publish it2 in first-rate national journals]]’) (101’) qi ‘it’ = [[D ϕ (3rd, Sg., inanimate)] [NP Yiqin’s article]] Condition C effects: * [TP she1 can publish [DP [D ϕ] [NP Yiqin1’s article]]] (102)

* 他 1 已經將其 2 交到簽證處的笑適 1 的照片 2 * [ta11 yijing jiang qi2 jiao-dao

qianzheng-chu]

3MSg already JIANG 3 give-to service.of.visa de Xiaoshi1 de zhaopian2 C Xiaoshi DE picture (‘Xiaoshi1’s picture2 [that he1 has already given it2 to the service of visa]’)

Analysis  

(102’) qi ‘it’ = [[D ϕ (3rd, Sg., inanimate)] [NP Xiaoshi’s picture]] Condition C effects: * [TP he1 has already given [DP [D ϕ] [NP Xiaoshi1’s picture]]] In dislocation structures, condition C effects can also be observed. However, my informants do not have a uniform judgment on the following data (cf. 103–104). Most of my informants do not like these sentences; however, they do not judge them fully ungrammatical. (103)

?* 小倩 1 的論文 2 , 我覺得她 1 可以將其 2 發表在國家一級期刊上。 ?* Xiaoqian1 de lunwen2, wo juede ta21 keyi jiang Xiaoqian DE article 1Sg think 3FSg can JIANG qi2 fabiao zai guojia yi-ji qikan-shang. 3 publish at national first.rate journal-on (‘As for Xiaoqian1’s article2, I think that she1 can publish it2 in firstrate national journals.’)

The fact that this sentence is not fully grammatical can be due to the binding relation established between the pronoun ta2 ‘she’ and the NP Xiaoqian in the Topic Xiaoqian’s lunwen ‘Xiaoqian’s article’ that is fully reconstructed on the dislocated site occupied by the resumptive pronoun qi ‘it’. Therefore, I assume that the pronoun qi ‘it’ takes an extended form as internal structure in which the NP Xiaoqian’s article is present and that such a configuration yields a violation of condition C. (103’) qi ‘it’ = [[D ϕ (3rd, Sg., inanimate)] [NP Xiaoqian’s article]] Condition C effects: * [TP she1 can publish [DP [D ϕ] [NP Xiaoqian1’s article]]] (104)

?* 張教授 1 的孩子們 2 , 我發覺他 1 不認為其 2 作風有問題。 ?* Zhang jiaoshou1 de haizi-men2, wo fajue Zhang ta11 3MSg you have

professor DE bu renwei Neg think wenti. problem

kid-Pl qi2 3-Gen

1Sg zuofeng behavior

think

(‘As for Professor Zhang1’s kids2, I feel that he1 does not think that their2 behavior is problematic.’)

  Reconstruction and internal structures of resumptive pronouns (104’) qi ‘their’= [D ϕ (3rd, Pl., human) [NP the behavior of the kids of Professor Zhang]] Condition C effects: * [TP he1 does not think [DP [D ϕ] [NP the behavior of the kids of Professor Zhang1]]] Condition C effects can also be detected in the case of intrusive pronouns (cf. 105–106). (105)

?* 張三 1 的學生們 2,[我看到他 1 將其 2 罵了一頓]的那一天正好是學 校的校慶日。 ?* Zhangsan1 de xuesheng-men2 , [wo kan-dao [ta11 Zhangsan DE jiang qi2 JIANG 3 tian zhenghao day just

student-Pl 1Sg see 3MSg ma-le yi-dun] de na yi scold-Perf one-Cl.v C that one shi xuexiao de xiaoqingri. be school DE anniversary

(‘Zhangsan1’s students2, the day [when I saw that he1 scolded them2] was the foundation anniversary of the school.’) The observed condition C effects suggest that the pronoun qi takes an extended form in this sentence, as illustrated in (105’). The reconstructed NP Zhangsan is directly bound by the pronoun he that shares the same index and this leads to condition C effects. (105’) qi ‘them’= [D ϕ (3rd, Pl., human) [NP Zhangsan’s students]] Condition C effects: * [TP he1 scolded [DP [D ϕ] [NP Zhangsan1’s students]]] A similar observation goes for (106). (106) ?* 張三 2 的未婚妻 1,他 2 明天必須要將其 1 帶去的地方是天主教堂。 ?* Zhangsan2 de weihunqi1, ta12 mingtian bixu-yao Zhangsan DE jiang qi1 JIANG 3 tianzhu-jiaotang. catholic.church

fiancée dai-qu take.to

de DE

3MSg tomorrow difang shi place be

must

(‘As for Zhangsan2’s fiancée1, the place [to which he2 must take her1] is the church.’)

Analysis  

(106’) qi ‘her’= [D ϕ (3rd, Sg, human) [NP Zhangsan’s fiancée]] Condition C effects: * [TP he1 must take [DP [D ϕ] [NP Zhangsan1’s fiancée]]] .... Summary In this section, I tested the pronoun qi with different reconstruction effects and the relevant results are presented in the following table. Tab. 11: Pronoun qi Relatives

Dislocations

Resumptives

Resumptives

Intrusives

qi

qi

qi

(c)

(f)

(h)

(i) Quantificational antecedent

yes

yes

yes

(ii) Scope reconstruction

yes

yes

yes

(iii) Binding reconstruction

yes

yes

yes

(iv) Condition C effects

yes

yes

yes

Conclusion: Internal structure

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

The results lead to the conclusion that the pronoun qi has only one specific form as its internal structure, which is an extended form. In different A'constructions, relatives or dislocation structures, qi always gives rise to interpretative effects under reconstruction. ... Differences between ta and qi In this section, I will compare qi with ta. Let me combine relevant tables on the pronouns based on ta and the pronoun qi in order to give a complete view of the distribution of these two types of pronouns.

  Reconstruction and internal structures of resumptive pronouns Tab. 12: Reconstruction effects of the two types of the pronouns Relatives

Dislocation structures Resumptives

Resumptives

Intrusives

qi

ta

qi

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i) Quantificational antecedent

yes

yes

yes

yes

no

yes

no

yes

(ii) Reconstruction of quantifier scope

yes

yes

yes

yes

no

yes

no

yes

(iii) Reconstruction of anaphoric binding

yes

yes

yes

yes

no

yes

no

yes

(iv) Condition C effects

yes

yes

yes

yes

no

yes

no

Yes

Internal structures

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

ta

[D-ϕ]

gaps

[[D-ϕ] NP]

qi

[D-ϕ]

ta

[[D-ϕ] NP]

gaps

First, let us compare column (e) with column (g) on the one hand, and column (f) with column (h) on the other hand. We can see from the table that these two pronouns, ta and qi, always keep the same internal structure in the same syntactic construction, irrespective of whether they are used as resumptive pronouns or intrusive pronouns. The pronoun ta always has a reduced form in dislocation structures. By contrast, the pronoun qi always has an extended form in relatives and in dislocation structures. Second, let us compare column (e) with column (f) on the one hand, and column (g) with column (h) on the other hand. In dislocation structures with a resumptive pronoun or with an intrusive pronoun, there is a clear difference between ta and qi. With qi, different interpretative effects under reconstruction are observed, whereas with ta, none of these effects is detected. Based on this observation, qi has only one form as its internal structure, which is an extended form. By contrast, ta has two different forms: the extended form is only used in relatives and the reduced form is only used in dislocation structures. Another important observation is that relatives require an extended form to be used but dislocation structures do not require this. Dislocation structures

Analysis  

tolerate two internal structures, whereas relatives permit only an extended form. ... Conclusion The survey in this section shows that the pronoun qi possesses only one possible form as its internal structure. This observation is crucial to my analysis because it confirms that not all of the pronouns can systematically have two different forms as their internal structures. From this point of view, Chinese behaves differently from Lebanese Arabic, Jordanian Arabic and Welsh. In these two Arabic dialects, strong resumptives and weak resumptives possess one specific structure respectively. In Welsh, a resumptive pronoun is intrinsically ambiguous between two different forms and Rouveret (2002, 2008) claims that it is interpretative requirements that select the suitable form of a given pronoun. In Mandarin Chinese, certain pronouns such as ta, have two possible forms but others such as qi have only one specific form. In the former case, it is derivational mechanisms that select the suitable form and in the latter case, no selection takes place. When a derivational mechanism fails to choose the right form of a pronoun in cases where such a pronoun does not possess the suitable form at all, the sentence becomes ungrammatical. By contrast, when the only available form of this pronoun coincides with the form required by derivational mechanisms, the relevant sentence is grammatical, and this is precisely the case where the pronoun qi appears in relatives. This is so because dislocation structures do not select the form of a resumptive pronoun due to deactivation of the EPP feature.

.. A mysterious problem is solved! Recall that in Chapter 2, I presented the fact that in relatives, a resumptive pronoun behaves in exactly the same way as gaps, which leads us to a very superficial assumption that in relatives, a resumptive pronoun can be analyzed as a spelled out trace. However, we have also observed that in dislocation structures, a resumptive pronoun and a gap behave differently. This asymmetry still remains a mystery. Now, the analysis that I argue for in this chapter can help us to solve this mysterious problem. Here is what we get concerning the distribution of internal structures of gaps and of resumptive pronouns in relatives and in dislocation structures.

  Reconstruction and internal structures of resumptive pronouns Tab. 13: Internal structures Relatives

Dislocation structures Resumptives

Intrusives

ta

qi

gaps

ta

qi

ta

qi

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h) [[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

Internal structures

Resumptives

gaps

The reason why a resumptive pronoun seems to be analyzed as a spelled out trace in relatives is very simple: this is due to a pure coincidence. In the interaction that I proposed earlier in this chapter, I explained that it is derivational mechanisms that choose the internal structure of an element located on variable sites in a given A'-construction. Agree determines that in a relative clause, an element located on a relativized site is in its extended form in order to satisfy the requirement of the EPP feature. This requirement remains absolutely independent of whether it is a gap or a pronoun that is used in this relative clause. Such a requirement also remains independent of type of pronoun (i.e. ta-type or qi-type). The only thing that EPP requires is to enter into the derivation with an extended form. As a pure coincidence, a gap has only an extended form as its internal structure and the pronoun qi has only an extended form as its internal structure as well. This coincidence leads to assume that when the pronoun qi is used as resumptive pronoun, it behaves as a trace. The pronoun ta has eventually two available forms; under the requirement of EPP, ta enters into the derivation of a relative clause with its extended form, and this gives an illusion that ta behaves like a trace. Therefore, the combination of these two independent facts leads to the superficial assumption that a resumptive pronoun can be analyzed as a spelled out trace in relatives. With these explanations in mind, it will be very easy to explain what happens in the case of dislocation structures. Since a dislocation structure is derived by Match which does not require activation of the EPP feature, such a construction does not require that an element on the dislocated site should absolutely have an extended form. In addition, a gap and the pronoun qi only have an extended form as their internal structures, they can only enter into the derivation of a dislocation structure with an extended form [[D-ϕ] NP]. By contrast, since the pronoun ta has two different forms, when the EPP feature is not activated in a dislocation structure, the extended form will not be required. As a

Conclusion  

result, the pronoun ta enters into the derivation of a dislocation structure with a reduced form as its internal structure. The asymmetry that we have observed initially between a relative clause and a dislocation structure in Chinese does not represent a typological phenomenon. It is not true that a resumptive pronoun can be analyzed as a spelled out trace nor is it true that a resumptive pronoun can function as a spelled out trace in certain types of A'-dependencies not in others.

. Conclusion Chapters 2 and 3 constitute the two most important parts linked to a complex problem on resumptive constructions that I study in this monograph. The starting point is that this complex problem has two aspects: syntactic properties and interpretative effects. The complexity of the study on resumptivity comes from the fact that these two aspects are inseparable. However, the position that I took in this study is to treat them separately in the first place and then study the correlation between them. Chapter 2 is focused on syntactic properties such as island effects and crossover effects. Chapter 3 is focused on interpretative effects under reconstruction. The reason why I decide to treat them separately is that the syntactic part and the semantic part are linked to different aspects of the problem respectively. Syntactic properties are linked to the derivation of syntactic constructions where resumptive pronouns can appear; whereas, reconstruction effects are linked to internal structures of pronouns. This chapter studies the relation between reconstruction effects and internal structures of a gap and of a resumptive pronoun. I examined two kinds of internal structures: an extended form [[D-ϕ] NP] that systematically gives rise to all of the reconstruction effects (including condition C effects) and a reduced form [D-ϕ] that gives rise to none of these effects. The presence of an NP in the extended form [[D-ϕ] NP] is crucial to trigger different interpretative effects. It is this NP that allows the extended form to function as a variable bound by a universal quantifier and it is also this NP that gives rise to reconstruction effects of anaphoric binding and triggers condition C effects. I examined internal structures of three types of elements (i.e. gaps, resumptive pronouns and intrusive pronouns) in two types of A'-dependencies (i.e. relatives and dislocation structures). A gap has only an extended form as its internal structure in different syntactic constructions. A resumptive pronoun and an intrusive pronoun behave in the same way in dislocation structures. I also examined two types of pronouns: the pronoun ta that has two different

  Reconstruction and internal structures of resumptive pronouns internal structures (i.e. an extended form and a reduced form) and the pronoun qi that has only an extended form as its internal structure. Based on reconstruction effects in resumptive constructions, I try to show that different factors intervene in reconstruction effects in Chinese. The problem of resumptivity in Chinese is particularly complicated because there are at least five factors that function together. (i)

two types of constructions : relatives and dislocation structures

(ii)

two types of derivational mechanisms: Agree and Match

(iii)

three types of elements: gap, resumptive and intrusive

(iv)

two types of pronouns: ta and qi

(v)

two types of internal structures: extended and reduced forms

I established an interaction that shows how these different factors work together. (i)

It is derivational mechanisms that choose types of A'-dependencies on which they operate. Relatives (with a gap and with a resumptive pronoun) and dislocation structures with a gap are derived by Agree, whereas resumptive dislocation structures are derived by Match.

(ii)

It is also derivational mechanisms that determine which type of resumptive pronoun appears in which type of syntactic construction. Agree is subject to locality conditions, which does not enable intrusive pronouns to exist in relatives. By contrast, Match is not subject to locality conditions, which makes it possible for intrusive pronouns to exist in dislocation structures. Independent of the presence or absence of an intrusive pronoun, an A'-dependency derived via Agree always gives rise to island effects, whereas an A'-dependency derived by Match does not give any of these effects.

(iii)

Derivational mechanisms also determine which type of feature is associated with which type of A'-dependency. For instance, EPP feature is activated in relatives (with a gap or with a resumptive pronoun) and in dislocations with a gap. By contrast, EPP is not activated in a Matching chain.

Conclusion  

(iv)

EPP requires the presence of an NP in internal structures of a gap and of a resumptive pronoun in an A'-dependency.

(v)

It is the presence of an NP in an extended structure that gives rise to different reconstruction effects.

(vi)

Different types of resumptive pronouns possess different internal structures. For pronouns that have two different forms as their internal structures, they can never decide by themselves to exist in a given type of A'-dependency with a given form of internal structure. As what I explained above, it is EPP that selects specific forms of pronouns to be used.

 Specificity effects . Introduction Syntactically, different analyses that have been proposed to account for resumptivity are generally based on island effects and crossover effects. Studies such as Zaenen et al. (1981), Koopman (1983) and Engdahl (1985), argue for the idea that a resumptive pronoun is analyzed as a spelled out trace because in languages that they study, a resumptive construction displays island effects. This leads these authors to conclude that a resumptive pronoun behaves exactly as a trace resulting from wh-movement. A challenge for this type of analysis comes from the semantic side of the problem. Authors who point out semantic differences between a trace resulting from movement and a resumptive pronoun are Doron (1982) and Sharvit (1999). It has also been observed that a gap displays two types of individual readings (i.e. specific reading (de re) and nonspecific reading (de dicto)) and two types of distributive readings (i.e. natural functional reading and pair-list reading). By contrast, a resumptive construction seems to block non-specific readings and the pair-list reading.33 This observation constitutes an important counter-argument for analyses based on the idea of “spelled out traces”. In this chapter, I will begin by presenting different semantic readings based on the work of Doron (1982) and that of Sharvit (1999). Then, I will present three different points of view on the relationship between resumptivity and nonavailability of the pair-list reading. Guilliot (2006, 2011) establishes a close relationship between definite descriptions and non-availability of the pair-list reading by suggesting that this reading is only generated with an indefinite copy but not with a definite copy. Since a resumptive pronoun is essentially a definite description, it blocks the pair-list reading. Based on Jordanian Arabic, Malkawi (2009) and Guilliot & Malkawi (2006, 2007) show that in fact, availability or unavailability of the pair-list reading is exclusively associated with a morphological competition between gaps, weak resumptives (clitic pronouns) and

 33 Bianchi (2004, 2011) discusses explicitly the relation between these two kinds of readings: specific individual reading and natural functional reading. She thinks that these two readings represent two different forms of specificity. Based on syntactic analyses of Boeckx (2003) and of Adger & Ramchand (2005), Bianchi (2011) provides a solution to specificity effects observed in resumptive constructions.

  Specificity effects strong resumptives (doubled pronouns).34 The analysis of Rouveret (to appear), based on Welsh, confirms that a resumptive pronoun does not systematically block the pair-list reading and that this reading exclusively depends on an extended form of resumptive pronouns. In his analysis, optional resumptive pronouns take a reduced form [D-ϕ] and obligatory resumptive pronouns take an extended form [[D-ϕ] NP]. In addition, I will systematically examine the availability of pair-list reading for gaps, the pronoun ta and the pronoun qi in relatives and in dislocation structures when these elements are located on variable sites in a given A'dependency. Let me present these tests in a table below: Tab. 1: Itinerary Relatives

Dislocations

Resump-

Resump-

tives

tives

Intrusives

Gaps ta

qi

Gaps ta

qi

ta

qi

(a)

(c)

(d)

(e)

(f)

(g)

(h)

(b)

Individual

Non-specific reading

--

--

--

--

--

reading

Specific reading

--

--

--

--

--

--

--

--

--

--

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

Distributive reading

Functional reding Pair-list

Indefinite

reading

Binding [[D-ϕ] NP]

[D-ϕ] NP]

[[D-ϕ] NP]

Structure interne

In Chapter 3, I examined the reconstruction of the scope of the universal quantifier, and this reconstruction generally gives rise to the pair-list reading. Such a reading can be obtained when the antecedent of a relative clause (i.e. the head noun) contains either an indefinite or an anaphoric element. After the reconstruction, these two types of antecedents will be under the scope of a universal quantifier, which gives rise to a distributive reading. However, since a topic  34 A similar difference also exists between two classes of pronouns in Japanese: N-type pronoun and D-type pronoun (Noguchi 1997).

Derive individual reading and distributive reading  

cannot be indefinite, the antecedent of a dislocation structure (i.e. the dislocated element in the position of topic) cannot be indefinite either. As a result, a dislocation structure can never have non-specific readings (de dicto) and we can only apply such a test concerning distributive reading to antecedents that contain an anaphoric element in dislocation structures. Results of these tests show a direct relationship between the availability of the pair-list reading and the extended form of the element located on the relativized/dislocated site. The conclusion that I will draw is that only the extended form allows reconstruction of the scope of a quantifier and the pair-list reading, irrespective of whether the element that has this extended form is a gap or a resumptive pronoun or an intrusive pronoun. From this perspective, Chinese behaves differently from languages such as Hebrew, Swedish, French and Jordanian Arabic but similarly to Welsh with several minor differences. As the reader will see in this chapter, Chinese data also confirm what Guilliot (2011) proposes in that the presence of an anaphoric element in the antecedent of a relative clause will force the pair-list reading.

. Derive individual reading and distributive reading In this section, I will begin by presenting two types of readings that will be examined and then I will introduce different analyses proposed previously in order to account for the relationship between these readings and resumptivity. .. Two types of specificities ... Non-specific reading (de dicto) vs. specific reading (de re) Doron (1982) observes for the first time a distinction existing between a gap and a resumptive pronoun concerning two types of individual readings, i.e. specific reading (de re) and non-specific reading (de dicto). If the relativized site is occupied by a gap, as illustrated in (1), both readings are available, whereas if the relativized site is occupied by a resumptive pronoun, as illustrated in (2), only specific reading (de re) is allowed. (1)

Dani yimca et ha-išaj še hu mexapes ___j . Dani will-find ACC the-woman that he seeks ‘Dani will find the woman that he’s looking for.’ √ specific (de re) √ non-specific (de dicto)

  Specificity effects (2)

Dani yimca et ha-išaj še hu Dani will-find ACC the-woman that he ‘Dani will find the woman that he’s looking for.’ √ specific (de re)

-

mexapes seeks

otaj. her

?* non-specific (de dicto) Hebrew, Doron (1982)

... Individual reading vs. pair-list reading The second contrast observed by Doron (1982) concerns two types of individual readings: single individual reading and multiple individual reading. In fact, the multiple individual reading is a sort of pair-list reading. In Hebrew, in relatives with a gap, when the trace is c-commanded by a quantified expression, the sentence will have both readings, as shown in (3). By contrast, when it is a resumptive pronoun that occupies the same position, the pair-list reading is blocked, as shown in (4). (3)

ha-iša1 the-woman tišlax

še that lo2

kol every tmuna.

gever2 man

baxar chose

____1 ____

will-send him a-picture ‘The woman1 that every man2 chose ___1 will send him2 a picture.’ = (i) The same woman was chosen by all of the men. (Single individual reading) = (ii) For every man, there is a specific woman that he chose. (Multiple individual reading/ Pair-list reading) (4)

ha-iša1 the-woman tišlax

še that lo2

kol every tmuna.

gever2 man

baxar chose

ota1 her

will-send him a-picture ‘The woman1 that every man2 chose (*her1) will send him2 a picture.’ = (i) The same woman was chosen by all of the men. (Single individual reading) ≠ (ii) For every man, there is a specific woman that he chose. (Multiple individual reading/ Pair-list reading) Hebrew, Doron (1982) These two contrasts lead to an assumption that semantically, resumptive pronouns possess certain properties of bound pronouns whereas gaps do not.

Derive individual reading and distributive reading  

.. Resumptivity and semantic readings ... Sharvit (1999) Sharvit (1999) makes another distinction among three possible interpretations: individual reading, functional reading and pair-list reading. For instance, in English, a question like the one in (5) allows three types of answers: individual answer, functional answer and Pair-list answer. (5)

Q: A: a. b. c.

Which woman did every man invite? Mary. Individual answer His mother. Functional answer John, Mary; Bill, Sally. Pair-list answer

Guilliot (2006) claims that the same type of question in French also gives rise to these three types of answers. (6)

Q:

Quelle femme est-ce que chaque homme a embrassée ? ‘Which woman did which man kissed?’ A: a. Marie. Individual answer b. Sa mère. ‘His mother’ Functional answer c. Jean, Marie; Fred, Justine ; Benoît, Valérie. Pair-list answer

Sharvit (1999) shows that in the case of Hebrew, if the element located on the variable site is a gap, all of these three types of answers are detected (cf. 7). By contrast, if such an element is a resumptive pronoun, only individual reading (cf. 8a) and functional reading (8b) are possible and pair-list reading is unavailable. (7)

Q:

ezyo iša kol gever himin ____? which woman every man invited ‘Which woman did every man invite?’ A: a. et Gila. ‘Gila.’ Individual answer b. et im-o. ‘His mother.’ Functional answer c. Yosi et Gila; Rami et Rina. Pair-list answer

  Specificity effects ezyo iša kol gever himin ota? which woman every man invited [her] ‘Which woman did every man invite?’ A: a. et Gila. ‘Gila.’ Individual answer b. et im-o. ‘His mother.’ Functional answer c. * Yosi et Gila; Rami et Rina. Pair-list answer

(8)

Q:

Meanwhile, contrary to what is observed by Doron (1982), Sharvit shows that in Hebrew resumptive pronouns do not systematically resist multiple individual reading. Both singular individual reading and multiple individual reading are displayed in specificational relatives with a gap (cf. 9a) or with a resumptive pronoun (cf. 9b). (9) a.

ha-iša the-woman hayta

še that išt-o.

kol every

gever man

hizmin invited

____

was wife-his ‘The woman that every man invited was his wife.’ = (i) the same woman was invited by all of the men. = (ii) for every man, there is a different woman that he invited. b.

ha-iša the-woman hayta

še that išt-o.

kol every

gever man

hizmin invited

ota her

was wife-his ‘The woman that every man invited was his wife.’ = (i) the same woman was invited by all of the men. = (ii) for every man, there is a different woman that he invited. The distribution of these readings in different contexts in Hebrew based on Sharvit’s (1999) work is presented in the following table.

Derive individual reading and distributive reading  

Tab. 2: Hebrew Readings Relatives

Questions

Gap

Resumptive pronoun

Singular individual

yes

yes

Multiple individual

yes

Individual

yes

yes

Functional

yes

yes

Pair-list

yes

no

Yes: specificational sentences No: non-specificational sentences

In non-specificational relatives, a pronoun can have an individual reading but not a pair-list reading; in questions, a pronoun can have a natural functional reading but not a pair-list reading. Sharvit claims that the only reading that a pronoun allows is natural functional reading. However, a natural function is not systematically available in all of the contexts. Functional reading and pair-list reading cannot be treated as two special cases of a distributive reading. According to the constraint on the bound anaphora that I extensively discussed in Chapter 3, a resumptive pronoun needs a discourse-linked antecedent. Specificational sentences correspond to natural functions that can provide discourse-linked antecedents. By contrast, non-specificational sentences correspond to arbitrary pair-lists. This is the reason why a resumptive pronoun is only licensed in specificational sentences. ... Guilliot (2006, 2011) Guilliot’s (2006, 2011) analysis claims that multiple individual reading/pair-list reading in relatives depends on reconstruction of the antecedent via the presence of a copy of this antecedent on the relativized site. As already presented in Chapter 3, in this analysis, a copy can be interpreted as a definite description analyzed as an E-type pronoun, which gives rise to specific individual reading and natural functional reading. A copy can also be interpreted as an indefinite description that gives rise to pair-list reading. A resumptive pronoun can only be interpreted as a definite, which explains why it blocks pair-list reading. However, the presence of a reconstructed bound variable in a given sentence will permit distributive reading. Let me cite a Jordanian Arabic example and a French example from Guilliot (2011).

  Specificity effects (10)

S-Surah2 the.picture

tabaʕat of

ʔibin-ha1 illi son-his that

kul every

mwazaf1 employee

ʒab-ha2 riʒʕat l-uh1. bring-Past.3s-it give.back-Passif to-him ‘The picture2 of his1 son that every employee1 brought (it2) was given back to him1.’ (11)

Marie a vu une photo de lui1 que chaque homme1 avait choisie. ‘ Mary has seen a photo of him1 that every man1 had chosen.’

In these two examples, antecedents contain anaphoric elements, his in (10) and of him in (11). After the reconstruction of the antecedent, the anaphor will be under the scope of the universal quantifier, which gives rise to pair-list reading. In fact, the description that resumptivity systematically blocks the pair-list reading in Jordanian Arabic is not exact. Malkawi (2009) shows that the readings permitted in a resumptive construction are licensed by an independent grammatical factor. ... Malkawi (2009): competition of the readings By studying functional readings in Jordanian Arabic, Malkawi (2009) claims that different readings that a resumptive pronoun can have depend on whether the variable position occupied by this resumptive pronoun is in competition with a gap or with another type of pronoun. Concretely, if there is a pair of sentences, one contains a gap and the other contains a weak resumptive (i.e. clitic pronoun) or one contains a weak pronoun and the other contains a strong resumptive (doubled pronoun), between these two sentences, the one that is less morphologically specified permits a reading that the other one that is more specified does not permit. This extra reading is always the semantically least specified one among all of the possible readings. (12) a.

ʔaya Surah1 il-uh2 kul zalamih2 which picture of-him every man mazaʕ ____1? tore ‘Which picture of him did every man tear?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading

Derive individual reading and distributive reading  

= (ii)

b.

Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading

ʔaya Surah1 il-uh2 kul zalamih2 mazaʕ-ha1? which picture of-him every man tore-it ‘Which picture of him did every man tear (it)?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading ≠ (ii) Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading

(12a) and (12b) show that in a context where a gap and a clitic resumptive pronoun are alternatives in the same position, the gap allows a pair-list reading that the clitic pronoun does not since generally gaps are morphologically less specified than clitic pronouns. The distribution of these readings are presented below: Tab. 3: Jordanian Arabic Natural functional reading

Pair-list reading

Gap

yes

yes

Clitic pronoun

yes

no

Now let us continue with another pair.

  Specificity effects (13) a.

b.

ʔaya Surah il-uh illi kul zalamih which picture of-him that every man mazaʕ-ha? tore-it ‘Which picture of him is it that every man tore it?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading = (ii) Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading ʔaya Surah il-uh illi kul zalamih which picture of-him that every man mazaʕ-ha hi? tore-it 3FSg ‘Which picture of him is it that every man tore it?’ = (i) Surit zawaʒ-uh. picture marriage-his ‘The picture of his marriage.’ Natural functional reading ≠ (ii) Karim, Surit ʔibn-uh; Redouan, Karim picture son-his Redouan Surit zawaʒ-uh; … picture marriage-his ‘Karim, the picture of his son; Redouan, the picture of his marriage; …’ Pair-list reading

(13a) and (13b) show a context where the presence of a gap is prohibited and the presence of a resumptive pronoun is obligatory. Such a context is constructed by a clefted question. The clitic pronoun in (13a) is morphologically less specified than the doubled pronoun in (13b). Clitics allow pair-list reading and doubled pronouns generally do not. The distribution of these readings is summarized below.

Derive individual reading and distributive reading  

Tab. 4: Jordanian Arabic Natural functional reading

Pair-list reading

Clitic pronoun

yes

yes

Doubled pronoun

yes

no

Malkawi’s generalization is that on the one hand, the readings permitted by a doubled pronoun are more restricted than those permitted by a clitic pronoun and on the other hand, a resumptive pronoun allows exactly the same readings as a gap but only in contexts where it is not in competition with a gap. Based on these generalizations, the author concludes that constructions containing an A'dependency are subject to a scale of morphological specification. (14) Morphological specification scale trace < clitic < clitic + doubled pronoun In a given context, a morphologically less specified structure gives rise to two possible readings and the more specified structure permits only one reading and this reading is the less marked one among other readings. From the point of view of morphological competition, Chinese behaves drastically different from Jordanian Arabic. I will show in the next section that morphological specification does not play any role in resumption in Chinese. Chinese data confirm that a gap and a pronoun give rise to exactly the same readings in a given context if and only if these two elements take the same form as their internal structures. ... Rouveret (2002, 2008, to appear) Recall that in Chapter 3, I discussed the idea of a structural ambiguity of the internal structure of a pronoun proposed by a serial work of Rouveret. Different forms give rise to different reconstruction effects. Only the extended form with the presence of an NP [[D-ϕ] NP] gives rise to reconstruction effects. Importantly, since pair-list reading is only obtained after the reconstruction, thus it is always the extended form that gives rise to pair-list reading. Then, Rouveret assumes that so-called specificity effects are associated with a contextual optionality. A resumptive pronoun whose presence in a given context blocks reconstruction of a quantifier scope and as a result, it does not give rise to pair-list reading. By contrast, a resumptive pronoun whose presence is obligatory in a given context allows reconstruction of the scope of a universal quantifier and it gives rise to pair-list reading. As a result, a resumptive pronoun always takes a

  Specificity effects reduced form in an optional context and an extended form in an obligatory context. If in a given language, a pronoun and a gap are never free alternatives in any position, in other words, if optional resumptive pronouns do not exist in this language, then both a pronoun and a gap allow pair-list reading. Rouveret provides the following Welsh example. In (15), individual reading and pair-list reading are both observed in relatives, irrespective of whether it is a gap (cf. 15a) or a pronoun (cf. 15b) that occupies the relativized site. (15) a.

Diolchodd thanked

y the

wraig1 woman

a C

wahoddodd invited

pob every

dyn2 ___1 iddo2. man [ __ ] to-him ‘The woman1 who every man2 invited ____1 thanked him2.’ = (i) The same woman was invited by all of the men. Individual = (ii) For every man, there is a different woman who was invited. Pair-list b.

Diolchodd thanked

y the

wraig1 woman

yr C

oedd was

pob every

dyn2 man

yn meddwl ei2. fod yn ei1 hoffi Prog think CLmasc be Prog CLfem like orau iddo2. best to-him ‘The woman1 [who every man2 thought [he2 preferred her1]] thanked him2.’ = (i) The same woman was preferred by all of the men. Individual = (ii) For every man, there is a different preferred woman. Pair-list Welsh, Rouveret (2002, 2008) In interrogative sentences, both a gap (cf. 16a) and a resumptive pronoun (cf. 16b) can give rise to three different readings: individual, natural functional and pair-list. (16) a.

Pa wraig a wahoddodd which woman aL invited ‘Which woman did every man invite?’

pob every

dyn man

__ ? [ __ ]

Derive individual reading and distributive reading  

b.

Pa which

lun picture

o’r ferch of the daughter

dych are

chi you

’n Prog

meddwl y mae pob dyn wedi ei gymryd ? think c is every man Perf CL take ‘Which picture1 of his2 daughter do you think that every man2 has taken?’ Welsh shows that being definite description, a pronoun does not systematically block pair-list reading. Rouveret also shows that in a Swedish dialect, either in a context where the presence of a resumptive pronoun is optional or in a context where its presence is obligatory, individual reading and pair-list reading are both permitted. For instance, (17) a.

Den the

kvinna1 woman

som that

varje every

man man

trodde thought

att that

(hon1) hade skrivit dikten, blev förstås (she) had written the.poem, became obviously otroligt smickrad. extremely flattered. ‘The woman1 that every man thought (she1) had written the poem was extremely flattered.’ = (i) Only one woman was flattered. Individual = (ii) For every man, there is a different woman who was flattered. Pair-list b.

Den the

kvinna1 woman

som that

varje every

man2 man

frågar asks

sej himself

om *(hon1) kommer att beundra honom2, if (she) will admire him blir vanligtvis besviken. becomes usually disappointed ‘The woman1 that every man2 asks himself whether she1 will admire him2 becomes usually disappointed.’ = (i) Only one woman is disappointed. Individual = (ii) For every man, there is a woman who was disappointed. Pair-list

  Specificity effects In (17a), even if the presence of the pronoun hon ‘she’ is not obligatory and a gap is also licensed in this position, the sentence can always give rise to an individual reading and a pair-list reading. This observation shows that this Swedish dialect behaves differently from Hebrew and Jordanian Arabic. Example (17b) illustrates the case where the pronoun hon ‘she’ plays the role of a last resort as intrusive pronoun. It gives rise to a pair-list reading. Welsh and this Swedish dialect clearly show that it is not the case that a resumptive pronoun systematically blocks pair-list reading. In the next section, I will show that Chinese behaves as Welsh and as the Swedish dialect mentioned in that a resumptive pronoun gives rise to pair-list reading and that this reading is permitted if and only if the extended form of the pronoun is selected.

. General distribution of different readings in Chinese In this section, I will give a systematic survey on the readings permitted by gaps, resumptive pronouns and intrusive pronouns in relatives and in dislocation structures in Chinese. Based on two types of specificities proposed by Doron (1982), I will test two types of individual readings, de re and de dicto, on the one hand and two types of distributive readings, natural functional reading and pair-list reading, on the other hand. In order to compare two types of pronouns, I will examine the pronouns ta and qi. .. Individual readings: non-specific (de dicto) vs. specific (de re) ... Relatives In relatives, no difference is observed between a gap and a resumptive pronoun concerning non-specific reading (de dicto) and specific reading (de re). Both a gap and a resumptive pronoun permit these two types of individual readings. Examples in (18) and in (19) illustrate that a gap permits a non-specific reading and a specific reading and examples in (20) and in (21) show that a resumptive pronoun also permits these two readings. (18)

[張三找了_____ j 一個月]的人 j 不知道在哪裡。 [Zhangsan zhao-le _____ j yi-ge

yue]

Zhangsan search-Perf one-Cl month de ren j bu zhidao zai nali. C person Neg know at where ‘The personj [that Zhangsan looks for _____j for a month] cannot be found.’ (non-specific/de dicto)

General distribution of different readings in Chinese  

(19)

[張三找了_____ j 一個月]的人 j 是他的親弟弟。 [Zhangsan zhao-le _____ j yi-ge

yue]

Zhangsan search-Perf one-Cl month de ren j shi ta-de qin-didi. C person be his brother ‘The personj [that Zhangsan looks for _____j for a month] is his brother.’ (specific/de re) (20)

[張三找他 j 找了 一個月]的人 j 不知道在哪裡。 [Zhangsan zhao ta1j zhao-le

yi-ge

yue]

Zhangsan search 3MSg search-Perf one-Cl month de ren j bu zhidao zai nali. C person Neg know at where ‘The personj [that Zhangsan looks for (himj) for a month] cannot be found.’ (non-specific/de dicto) (21)

[張三找他 j 找了一個月]的人 j 是他的親弟弟。 [Zhangsan zhao ta1j zhao-le

yi-ge

yue]

Zhangsan search 3MSg search-Perf one-Cl month de ren j shi ta-de qin-didi. C person be his brother ‘The personj [that Zhangsan looks for (himj) for a month] is his brother.’ (specific/de re) The same result was observed for the pronoun qi that also allows two types of individual readings in relatives (cf. 22–23). (22)

[張三對其 j 失去信心]的人 j 很少。 [Zhangsan dui qij shiqu

xinxin]

de

Zhangsan on 3 lose trust C renj hen shao. person very few ‘The peoplej [on whomj Zhangsan loses his trust] are not many.’ (non-specific/de dicto)

  Specificity effects (23)

[張三對其 j 失去信心]的人 j 是他的親弟弟。 [Zhangsan dui qij shiqu xinxin]

de

Zhangsan on 3 lose trust C renj shi ta-de qin-didi. person be his brother ‘The peoplej [on whomj Zhangsan loses his trust] is his brother.’ (specific/de re) ... Dislocation structures Unfortunately, it is impossible to apply the same test to dislocation structures, because a topic/dislocated element can only have a specific reading. Topic can never be a non-specific indefinite. Therefore, the unavailability of non-specific reading (de dicto) is due to a particular property of dislocation constructions themselves.

.. Distributive readings: natural function vs. pair-list Examples in this section have already been used in Chapter 3 to examine whether a gap, a resumptive pronoun and an intrusive pronoun can all have a distributive reading under reconstruction of the scope of a universal quantifier. Normally, if a distributive reading is allowed in an A'-dependency, this means that such a construction gives rise to pair-list reading and natural functional reading. ... Relatives A gap (cf. 24), the pronoun ta3 (cf. 25), the pronoun qi (cf. 26), can have both a natural functional reading and a pair-list reading. (24)

[每個導演都會看好幾遍_____j]的一個電影預告片 j [mei-ge daoyan dou hui kan

hao-ji-bian

each-Cl director all will watch several.times ____ j] de yi-ge dianying-yugaopianj C one-Cl movie.trailer ‘a movie trailerj [that each director will watch _____j several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃) (i) Shi ta-de xin pian. be his new movie ‘It is his latest movie.’

Functional reading

General distribution of different readings in Chinese  

(ii)

(25)

Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

[每個導演都會將它 j 看好幾遍]的一個電影預告片 j [mei-ge daoyan dou hui jiang ta3j

kan

each-Cl director all will JIANG 3OSg watch hao-ji-bian de yi-ge dianying-yugaopianj several.times C one-Cl movie.trailer ‘a movie trailerj [that each director will watch itj several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃) (i) Shi ta-de xin pian. be his new movie (ii)

(26)

‘It is his latest movie.’ Functional reading Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

[每個導演都會將其 j 看好幾遍]的一個電影預告片 j [mei-ge daoyan dou hui jiang qij

kan

each-Cl director all will JIANG 3 watch hao-ji-bian de yi-ge dianying-yugaopianj several.times C one-Cl movie.trailer ‘a movie trailerj [that each director will watch itj several times]’  √ ‘each director watches a different movie trailer’ (∀ > ∃) (i) Shi ta-de xin pian. be his new movie (ii)

‘It is his latest movie.’ Functional reading Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

If an anaphor is present in the antecedent, pair-list reading will be forced as predicted in Guilliot (2011). Example (27) shows that in an A'-dependency with a

  Specificity effects gap whose antecedent contains the pronoun his, the only possible reading is distributive reading, hence a pair-list reading. (27)

[每個導演都會看好幾遍_____j]的關於其新作的預告片 j [mei-ge daoyan dou hui kan hao-ji-bian

____ j]

each-Cl director all will watch several.times de guanyu qij xinzuo de yugaopianj C about 3-Gen new.product DE trailer ‘the trailer1 of his2 latest movie [that every director2 will watch _____1 several times]’ = Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading The same result goes for the pronouns ta and qi (cf. 28–29). (28)

每個導演 1 都會將它 2 拿去參賽的他自己 1 的新片 2 [mei-ge daoyan dou hui jiang ta32

na

qu

each-Cl director all will JIANG 3OSg take to can-sai ] de ta31-ziji-de xin-pian2 participate.to.competition C 3MSg-self-DE new-movie ‘the latest movie2 of his1 own [that every director1 will take it2 to participate to a competition]’ = Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading (29)

每個導演 1 都會將其 2 拿去參賽的他自己 1 的新片 2 [mei-ge daoyan dou hui jiang qi2

na

qu

each-Cl director all will JIANG 3 take to can-sai ] de ta31-ziji-de xin-pian2 participate.to.competition C 3MSg-self-DE new-movie ‘the latest movie2 of his1 own [that every director1 will take it2 to participate to a competition]’ = Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

General distribution of different readings in Chinese  

... Dislocation structures Let us turn to dislocation structures. As I pointed out earlier, a dislocated element cannot be indefinite and as a result, I can only apply such a test to topics that contain an anaphoric element. .... Resumptive pronouns The presence of the anaphor ziji ‘-self’ in a dislocated element always forces distributive reading, as demonstrated in (30). This confirms the observation in Guilliot (2011). (30)

自己 1 的新片 2 , 每個導演 1 都會拿____2 去參賽。 Ziji1-de xin-pian2 , mei-ge daoyan1 dou self-DE new-movie each-Cl director all hui na ____2 qu can-sai. will take to participate.to.competition ‘The latest movie2 of his1 own, every director1 will take ______2 to participate to a competition.’ = Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

However, in the same context, not all of the native speakers accept distributive reading when it is the pronoun ta3 that is used as resumptive. (31)

?* 自己 1 的新片 2 , 每個導演 1 都會把它 2 拿去參賽。 ?* Ziji1-de xin-pian2 , mei-ge daoyan1 dou

hui

self-DE new-movie each-Cl director all will ba ta32 na qu can-sai. BA 3OSg take to participate.to.competition ?* ‘The latest movie2 of his1 own, every director1 will take it2 to participate to a competition.’ = ?*Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading By contrast, if it is the pronoun qi that occupies the same position in the same context, distributive reading becomes more natural.

  Specificity effects (32)

自己 1 的新片 2 , 每個導演 1 都會將其 2 拿去參賽。 Ziji1-de xin-pian2 , mei-ge daoyan1 dou hui self-DE new-movie each-Cl director all will jiang qi2 na qu can-sai. JIANG 3 take to participate.to.competition ‘The latest movie2 of his1 own, every director1 will take it2 to participate to a competition.’ = Wang dao, Chuntian; Li dao, Aiqing; Zhang dao, Shenghuo… ‘Director Wang, Spring; Director Li, Love; Director Zhang, Life, etc.’ Pair-list reading

Such a contrast is not surprising at all after having examined the possibility to reconstruct the scope of a universal quantifier in resumptive dislocation structures in Chapter 3. What we observed is that it is difficult to reconstruct the scope of a universal quantifier when the resumptive pronoun is ta3; by contrast, such a reconstruction becomes much easier to get when the resumptive pronoun is qi. Since a pair-list reading is essentially a distributive reading, the observed contrast in this chapter further confirms that the results that we get from Chapter 3 are reliable. .... Intrusive pronouns The same contrast between the pronoun ta and the pronoun qi concerning the pair-list reading is detected when both of them are used as intrusives, as shown in (33–36). (33) ?*自己 1 的身份證 2,每位公民 1 必須要把它 2 更新的指定機構在漢口。 Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen ba BA zai in

ID.card ta32 gengxin 3OSg replace Hankou. Hankou

every-Cl de C

citizen zhiding designated

must jigou service

(‘His1 own ID card2, the official service [where every citizen1 is supposed to replace it2 with a new one] is in Hankou District.’) ?*‘Zhangsan, the ID card of Zhangsan ; Lisi, the ID card of Lisi ; Wangwu, the ID card of Wangwu, etc.’ Pair-list reading

General distribution of different readings in Chinese  

(34) ?* 自己 1 的父母 2,每位子女 1 都能完全理解他們 2 的家庭還是不多的 ?* Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen wanquan entirely haishi yet

parents every-Cl lijie understand bu duo Neg many

child ta1-men2 3Pl de. DE

all de DE

can jiating family

(‘His1 own parents2, the families [in which every child1 can perfectly understand them2] are not many yet.’) ?* ‘Zhangsan, the parents of Zhangsan ; Lisi, the parents of Lisi ; Wangwu, the parents Wangwu, etc.’ Pair-list reading (35) 自己 1 的身份證 2,每位公民 1 必須要*(將其 2)更新的指定機構在漢口。 Ziji1-de shenfen-zheng2, mei-wei gongmin1 bixu-yao self-Gen *(jiang JIANG zai in

ID.card qi2) gengxin 3 replace Hankou. Hankou

every-Cl de C

citizen zhiding designated

must jigou service

‘His1 own ID card2, the official service [where every citizen1 is supposed to replace *(it2) with a new one] is in Hankou District.’ ‘Zhangsan, the ID card of Zhangsan ; Lisi, the ID card of Lisi ; Wangwu, the ID card of Wangwu, etc.’ Pair-list reading (36)

自己 1 的父母 2,每位子女 1 都能完全理解*(其 2 用心)的家庭已經不多了 Ziji1-de fumu2, mei-wei zinü1 dou neng self-Gen wanquan entirely jiating family

parents every-Cl child all lijie *(qi2 yongxin) understand 3 intention yijing bu duo le. already Neg many SFP

de DE

can

‘His1 own parents2, the families [in which every child1 can perfectly understand them2] are no longer many.’ ‘Zhangsan, the parents of Zhangsan ; Lisi, the parents of Lisi ; Wangwu, the parents Wangwu, etc.’ Pair-list reading

  Specificity effects The above sentences confirm again that the reconstruction of the scope of a universal quantifier is only possible for the pronoun qi but not for the pronoun ta3 when both are used intrusively. This explains why a pair-list reading is only available in the former case but not in the latter case. .. Summary The result of these tests is recapitulated in the following table. Tab. 5: Specificity in Mandarin Chinese Relatives

Dislocations

Resumptives Gaps

ta

qi

(a)

(b)

(c)

(e)

(f)

(g)

(h)

Non-specific reading

yes

yes

yes

----

---

----

----

----

Specific reading

yes

yes

yes

----

---

----

----

----

Functional reading

yes

yes

yes

yes

no

yes

no

yes

Indefinite reading

yes

yes

yes

----

---

----

----

----

Binding

yes

yes

yes

yes

no

yes

no

yes

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

(d)

qi

[[D-ϕ] NP]

ta

[[D-ϕ] NP]

Gaps

[D-ϕ] NP]

Distributive reading

qi

Intrusives

[[D-ϕ] NP

Individual reading

ta

Resumptives

Pairlist

Structure interne

Let us compare (a) with (d). A gap can always give rise to pair-list reading, irrespective of whether it is located in a relative clause or in a dislocation structure. A gap always takes an extended form. As for columns (c), (f) and (h), no difference is observed. The pronoun qi systematically gives rise to pair-list reading both when it is used as a resumptive pronoun and when it is used as an intrusive pronoun. This observation is also true when it is used either in relatives or in dislocation structures. By contrast, the pronoun qi always takes a reduced form. Let us compare (b) with (e/g). When the pronoun ta is used in a relative clause, it takes its extended form and gives rise to pair-list reading. By contrast,

Analysis  

when it is used in a dislocation structure, it takes its reduced form and does not give rise to pair-list reading. Now, let us compare the distribution of the pair-list reading with the distribution of internal structures of a gap and of a pronoun. In other words, I will compare columns (a), (b), (c), (d), (f) and (h) with columns (e) and (g). It is not difficult to see that pair-list reading is available when it is the extended form [[D-ϕ] NP] that is adopted. On the other hand, pair-list reading is not available when it is the reduced form [D-ϕ] that is adopted.

. Analysis In this section, I will begin by interpreting the results that we obtained based on the tests in the previous section. Then, I will present my analysis and conclusion concerning different factors that influence the unavailability of the pair-list reading. The results of the tests that I showed at the end of the previous section illustrate an interesting but superficial coincidence. It seems that the boundary between the availability or the unavailability of a specific reading and of a pair-list reading coincides with the boundary between the extended form and the reduced form of the element located on the variable site in a given construction. When it is the extended form that is adopted, both readings are observed; when it is the reduced form that is adopted, neither of these readings is observed. This coincidence reminds us of another coincidence observed in Chapter 3 concerning the relationship between the possibility of reconstruction of the scope of a universal quantifier and the choice of the internal structure of gaps and of resumptive pronouns. The generalization is that the extended form always gives rise to reconstruction of a quantifier scope, whereas the reduced form never displays this effect. A pair-list reading is a kind of distributive reading and a distributive reading is observed if and only if an indefinite antecedent or an anaphoric antecedent is reconstructed under the scope of a universal quantifier. It is clear that such a reconstruction exclusively depends on internal structures of the element located on the variable site in a given A'-dependency. Reconstruction is only possible when the extended form is adopted as the internal structure of this element. The Chinese data above confirm that a resumptive construction does not necessarily exclude pair-list reading. They also confirm Rouveret’s (to appear) generalization on Welsh. His generalizations are the following.

  Specificity effects (i)

Specificity associated with a reading in a resumptive construction depends on the optionality of resumptive pronouns that appear in such a construction.

(ii)

An optional resumptive pronoun blocks reconstruction and an obligatory resumptive pronoun gives rise to reconstruction. An optional resumptive pronoun takes a reduced form and an obligatory resumptive pronoun takes an extended form.

(iii)

Definite personal pronouns do not resist at all multiple individual reading (i.e. pair-list reading).

(iv)

Pronouns in this type of language (i.e. Welsh) are bound pronouns.

My observation on Chinese gives the following generalizations: (i)

Resumption does not systematically block a distributive reading (i.e. pair-list reading).

(ii)

Availability of a pair-list reading exclusively depends on the form of a pronoun. Only the extended form gives rise to reconstruction of a quantifier scope and to a distributive reading (i.e. pair-list reading).

(iii)

If a resumptive construction gives rise to pair-list reading and if there is no other independent factor, such as the specificity associated with topics in dislocation structures, that excludes a natural functional reading, then a distributive reading can be observed.

(iv)

The interpretation of a resumptive pronoun depends on its derivational origin.

The difference between my generalizations and those of Rouveret’s lies in the fact that in Chinese, the extended form is not systematically associated with an obligatory resumptive pronoun and likewise, the reduced form is not systematically associated with an optional resumptive pronoun either. Chinese shows that (i) derivational mechanisms determine the choice of the form of the element located on the variable site in a given A'-dependency, irrespective of whether it is a gap or a resumptive pronoun or an intrusive pronoun that occupies such a variable position and (ii) different types of pronouns can possess

Analysis  

different internal structures, irrespective of whether they are used as optional resumptive pronouns or as obligatory resumptive pronouns or even as intrusive pronouns. My generalizations can account for several important observations in Chinese: Generalization 1: a gap, a resumptive pronoun and an intrusive pronoun can give rise to pair-list reading in relatives. This is so because Agree necessarily requires the extended form of the element in the variable position to be used in relatives where the EPP feature is activated. On the one hand, a gap and the pronoun qi possess only the extended form, whereas the pronoun ta obligatorily enters into the derivation of a relative clause with its extended form. On the other hand, the extended form gives rise to pair-list reading. As a result, all of these three elements can give rise to pairlist reading in relatives. Generalization 2: in dislocation structures, a gap, a resumptive pronoun and an intrusive pronoun do not systematically give rise to pair-list reading. The operation Match does not necessarily require a specific form of the element located on the variable site in a dislocation structure because the EPP feature is not activated. Since a gap and the pronoun qi can only enter into the derivation of a dislocation structure with an extended form, both elements give rise to pairlist reading. The pronoun ta3 enters into the derivation with its reduced form and as a result, it permits neither reconstruction of a quantifier scope nor pairlist reading. From this perspective, these three elements do not behave uniformly in a dislocation structure when a pair-list reading is concerned. Generalization 3: a given pronoun does not systematically give rise to a pair-list reading in different A'-dependencies. This generalization is true if and only if this pronoun possesses at the same time an extended form and a reduced form as its internal structures. When this pronoun enters into a syntactic derivation that requires an extended form, it gives rise to a pair-list reading; by contrast, when it enters into a syntactic derivation with a reduced form, it does not give rise to pair-list reading.

  Specificity effects These three generalizations lead to the conclusion according to which, only the extended form allows both reconstruction of quantifier scope and pair-list reading, irrespective of whether the element that possesses such a form is a gap or a resumptive pronoun or an intrusive pronoun.

. Conclusion This chapter examined the distribution of the pair-list reading in relatives and in dislocation structures in Chinese. I began by a brief review of the previous work on this topic. It has been observed that a resumptive pronoun blocks nonspecific (de dicto) reading and pair-list reading in languages such as Hebrew (Doron 1982, Sharvit 1999). Different authors tried different ways to account for such a phenomenon. I presented several important analyses on this topic. Based on Jordanian Arabic and French, Guilliot (2006, 2011) claims that being definite description, a resumptive pronoun systematically excludes pair-list reading when there is no binding relation involved in a relative clause. Working on Jordanian Arabic, Malkawi (2009) proposes a description of the phenomenon based on a competition between gaps, weak resumptives and strong resumptives. Such a description is subject to a morphological specification scale. Chinese behaves differently from the above languages in that pair-list reading is only observed when the extended form of the element located in the variable position in a given A'-dependency is adopted. If and only if a gap or a resumptive pronoun enters into a derivation with an extended form as its internal structure, it gives rise to pair-list reading and at the same time, to reconstruction of the scope of a universal quantifier. From this point of view, the crucial factor that determines the availability of the pair-list reading is the syntactic mechanisms that derive the concerned A'-dependency. This conclusion can be confirmed by the previous discussion in Chapters 2 and 3. The operation Agree works on relatives (with a gap or with a resumptive pronoun) and on dislocation structures with a gap; it gives rise to island effects and crossover effects. Agree determines that the EPP feature should be activated in relatives. EPP requires an extended form of the element located on the relativized site to be used and this choice remains independent of whether such a relativized site is occupied by a gap or by a resumptive pronoun. The extended form gives rise to different reconstruction effects including reconstruction of binding relation and quantifier scope, to condition C effects under reconstruction and it also gives rise to distributive reading (i.e. pair-list reading). The operation Match works on resumptive dislocation structures. Dislocated elements are always base-generated in the left-periphery. Since Match does not

Conclusion  

obey locality conditions, neither island effects nor crossover effects is detected. Match does not require EPP to be activated; elements on the dislocated site do not obligatorily enter into the derivation with an extended form. When an element that enters into the derivation possesses only an extended form, such as gaps or the pronoun qi, it gives rise to reconstruction effects and to pair-list reading. When an element enters into a derivation with its reduced form, for instance the pronoun ta, these reconstruction effects and the pair-list reading are not detected.

 Concluding remarks Resumptivity is a rather complex problem involving two inseparable components: syntax and semantics. In this study, I have examined systematically the general distribution of gaps, resumptive pronouns (i.e. the general/grammatical /systematic use of the resumptive pronouns) and so-called intrusive pronouns (i.e. the use of resumptive pronouns in islands) in two different types of A'dependencies in Mandarin Chinese: relatives and LD-structures. Chapters 2, 3 and 4 treat each of the core problems linked to resumptivity. Chapter 2 examines the syntactic properties of the resumptive chains in different A'dependencies and Chapters 3 and 4 the different interpretative effects of resumptive pronouns and gaps in those dependencies. However, there is a close logical correlation between the observed syntactic derivational properties and semantic interpretative properties. The solution that I gave in this monograph is syntactically oriented and the main problems linked to resumptivity, syntactic as well as semantic, can be solved in Narrow Syntax. In this sense, my analysis is “syntactic” in that it accounts for all of the observed semantic interpretative properties only in syntax. This is sharply different from the traditional “syntax-semantics” interface account. Importantly, the correlation between syntax and semantics when resumptivity is concerned is subject to an interaction of different factors involved in the resumptive chains.

. Syntactic level As I showed in Chapters 1 and 2, one of the most important questions at the heart of the debate on resumptivity is whether a resumptive chain is derived by movement. Two different points of view exist. One of them defends the idea that a resumptive chain is derived by movement either at syntax (i.e. surfacestructure) (cf. McCloskey 1990) or at LF (Demirdache 1991). Under such an approach, two analyses have been proposed: either the full form of the antecedent NP moves to the A'-position or only a part of the NP moves by leaving the pronoun part in-situ in the variable site and the latter analysis is called “subextraction” (cf. Rouveret 1994). These two analyses based on movement have been proposed in the framework of GB. Apparently, under the Minimalist Program, the operation Move has been proposed to account for the derivation of resumptive dependencies (Aoun et al. 2001, Boeckx 2003). The other point of view argues that a resumptive chain can be derived without movement (Adger & Ramchand 2005, Rouveret 2008, 2011). Such a point of view has been proposed

  Concluding remarks based on the Minimalist Program in which Move is defined quite differently from “movement” in the Government and Binding theory. Based on Mandarin data, the results of the tests in Chapter 2 suggest that different types of resumptive dependencies can be established by different syntactic derivational mechanisms. I examined the derivation of relatives and that of LD-structures in both the GB framework and the Minimalist Program. I am only concentrating here on the conclusions drawn from my minimalist analyses. Relatives (with gaps or with resumptive pronouns) and LD-structures with gaps are derived by Agree, which is subject to locality constraints and gives rise to island effects and crossover effects. Essentially, Agree establishes a checking relation between a Probe and a Goal in order to check and delete all of the uninterpretable features. The [var] feature on Probe (i.e. C-Rel), is uninterpretable and the same feature on Goal (i.e. the resumptive pronoun) is interpretable. After Agree, the uninterpretable feature has been deleted. Therefore, the A'dependency derived by Agree can be established. Resumptive LD-structures are derived by operation Match which is not subject to the locality constraint. Neither island effects nor crossover effects are detected. As I pointed out in Chapter 2, the essential function of Match is identifying the attributes of the relevant formal features attached to the potential candidates for Probe and for Goal respectively. Importantly, Agree does not apply to the cases of resumptive LD-structures due to the fact that the formal features attached to the Probe and those attached to the Goal are all uninterpretable. In this case, there is no appropriate Goal having the matched interpretable features that can eventually check and delete the same types of features that are uninterpretable when attached to the Probe. Therefore, the relation between Probe and Goal cannot be established by Agree. Match is not considered as a precondition on Agree but as an independent syntactic mechanism that establishes an A'-dependency. Agree functions phase by phase and it is subject to locality constraints. Match does not obey locality constraints but it also works at Narrow Syntax. Match applies to the entire syntactic structure containing all of the wellestablished phases that have not been transferred yet to the interfaces for interpretation. Only phases that are derived by Agree containing no uninterpretable features can be sent immediately to interfaces for interpretation. From this perspective, multiple Transfers and multiple Spell Outs only apply to Agree chains but not to Matching chains. It is precisely for this reason that Match does not give rise to island effects nor to crossover effects. Agree operates unselectively on structures with gaps, which is independent of the question whether these gaps are located on a relativized site or on a dislocated site. Both cases give rise to island effects and crossover effects.

Semantic level  

An important difference between relatives and LD-structures must be highlighted when resumptive pronouns are involved. The relationship between CTop and the resumptive pronoun in an LD-structure is not necessarily comparable to that which exists between an operator and a variable. By contrast, the relationship between C-Rel and the resumptive pronoun located on a relativized site is obligatorily an operator-variable type of relation. Therefore, the [λ] feature is uninterpretable when it is associated with C-Top and the same feature becomes interpretable when it is associated with C-Rel. The [var] feature is interpretable when it is associated with a resumptive pronoun in a relative clause but uninterpretable when it is associated with a resumptive pronoun in an LDstructure. The following table summarizes the distribution of the gaps and of the resumptive pronouns in different syntactic tests. Tab. 1: Syntactic distribution Relatives

Dislocations

Gaps

Resumptives Intrusives

Gaps

Resumptives Intrusives

i) Without island

yes

yes

---------

yes

yes

--------

ii) With island

yes

-------

no (island effects)

yes

--------

yes (no island effects)

iii) Weak crossover effect

yes

yes

--------

yes

no

--------

iv) Strong crossover effect

yes

yes

---------

yes

no

--------

Syntactic derivation

GB

Movement

MP

Agree without Move

Binding in discourse Match without Agree

. Semantic level In Chapters 3 and 4, I showed that previous work also investigates resumptivity in terms of semantics. One of the most studied phenomena is that of resumptive dependencies which give rise to reconstruction effects. In order to account for this phenomenon, recall the two generalizations that I have made based on Mandarin data (cf. Chapter 3).

  Concluding remarks (i) (ii)

Only certain resumptive constructions can give rise to reconstruction effects; Not all types of resumptive constructions give rise to reconstruction effects systematically.

Following Rouveret (2002, 2008, to appear), I examined the close relationship between the reconstruction effects and the internal structures of the elements located on the variable site in an A'-dependency. The extended form [[Dϕ] NP] systematically gives rise to reconstruction effects (including condition C effects); the reduced form [D-ϕ] does not give rise to any of these effects. The NP part makes it possible for the extended form [[D-ϕ] NP] to function as a variable bound by the universal quantifier. It is also this NP part that gives rise to reconstruction effects of anaphoric binding as well as triggers eventually the condition C effects. In Chinese, gaps can only have the extended form [[D-ϕ] NP] as their internal structure, which is independent of the type of A'-chain they appear in. In other words, in all of the A'-dependencies, relatives or LD-structures, a gap cannot have a reduced form [D-ϕ] as its internal structure. From this perspective, the pronoun qi behaves in a similar fashion in that it can only have an extended form [[D-ϕ] NP] as its internal structure. By contrast, the pronoun ta3 is ambiguous between two possible internal structures, i.e. the extended form [[D-ϕ] NP] and the reduced form [D-ϕ]. This reasoning explains several important observed facts. Gaps and the pronoun qi give rise to reconstruction effects systematically in both relatives and in LD-structures. As I explained, it is the NP part in their extended structure that triggers the reconstruction effects. By contrast, the pronoun ta enters into the derivation of a relative clause with its extended form, which triggers of course the reconstruction effects; it enters into the derivation of an LD-structure with its reduced form, which does not give rise to any of the reconstruction effects. It was argued in Chapter 3 that pronouns themselves do not determine which type of A'-dependency goes with which form of internal structure. Instead, I argue that it is the derivational mechanisms (i.e. Agree or Match) that determine it. Agree determines that the EPP feature has to be activated in relatives (with a gap or with a resumptive pronoun) and in LD-structures with gap. The EPP requires the extended form of the element that is located on the relativized site to be used. This requirement remains independent of whether this relativized site is occupied by a gap or by a resumptive pronoun. Systematically, the extended form gives rise to reconstruction effects. Crucially, Agree is not considered as a precondition on Move.

Semantic level  

Match operates on resumptive dislocation structures. I assume that the dislocated elements in this case are systematically base-generated in the leftperiphery. Match does not require the EPP feature to be activated. Accordingly, elements on the dislocated site do not have to enter into the derivation with their extended forms, which makes two situations possible. If an element that enters into the derivation of a resumptive dislocation structure has only the extended form as its internal structure, such as gaps or the pronoun qi, then it gives rise to reconstruction effects. If an element having both extended and reduced forms as its internal structures enters into the same derivation with its reduced form, such as the pronoun ta3, none of the reconstruction effects are detected. The following table summarizes the discussion of this part. Tab. 2: Reconstruction effects Relatives

Dislocation structures Resumptives

Resumptives

Intrusives

qi

ta

qi

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i) Quantificational antecedent

yes

yes

yes

yes

no

yes

no

yes

(ii) Reconstruction of quantifier scope

yes

yes

yes

yes

no

yes

no

yes

(iii) Reconstruction of anaphoric binding

yes

yes

yes

yes

no

yes

no

yes

(iv) Condition C effects

yes

yes

yes

yes

no

yes

no

Yes

Internal structures

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

ta

[D-ϕ]

gaps

[[D-ϕ] NP]

qi

[D-ϕ]

ta

[[D-ϕ] NP]

gaps

The crucial argument that supports such an analysis is based on the fact that when a pronoun such as ta has two different forms, an extended one and a reduced one, it enters into the derivation of the A'-dependencies established by Agree with its extended form only and it enters into those established by Match with its reduced form only. The fact that such a pronoun does not enter into the

  Concluding remarks derivation of the first type of A'-dependency with the reduced form and it does not enter into that of the second type of A'-dependency with its extended form either leads me to assume that the choice between these two forms cannot be completely free. I also examined the so-called specificity effects in Chapter 4. Previous studies show that resumption blocks the pair-list reading in languages like Hebrew. Later work reveals that such a blocking effect is only apparent and it is subject to several independent factors, for instance, the morphological specification scale in Jordanian Arabic proposed by Malkawi (2009). Still, we must admit that many other languages do not display this blocking effect at all. Rouveret (to appear) shows that Welsh and one particular Swedish dialect are completely free from the blocking effect and that they allow the pair-list reading in resumptions. My test in this study shows that Mandarin Chinese belongs to the latter category of languages: those that do not block the pair-list reading in resumptive dependencies. The conclusion that we drew from these tests is that only the extended form [[D-ϕ] NP] allows the reconstruction of the quantifier scope and it gives rise to the distributive reading, i.e. the pair-list reading in particular. Under reconstruction, the NP part in the extended form will be within the scope of the universal quantifier and yields the pair-list reading, which is independent of the issue whether the element that possesses the extended form is a gap or a resumptive pronoun or even an intrusive pronoun. In relatives, Agree requires that the element located on the variable site (for instance, relativized site) enters into the derivation with its extended form and that the EPP feature is activated. Gaps and the pronoun qi only possess the extended form and they have no choice but to enter into the derivation with this extended form. However, the pronoun ta possesses both the extended and the reduced forms and has to enter into the derivation of a relative clause in its extended form due to the requirement of the EPP feature. Therefore, these three elements give rise to the pair-list reading in relatives. By contrast, in a resumptive LD-structure, Match does not require a specific form of the internal structure of the element located on the variable site to be used because the EPP feature is not activated in this case. Since gaps and the pronoun qi have no choice but enter into the derivation with the extended forms, both of them give rise to the pair-list reading. By contrast, the pronoun ta enters into the derivation with its reduced form and therefore, it does not permit the reconstruction of the quantifier scope nor any of the distributive readings. One of the advantages of my analysis is that it explains why the same pronoun does not give rise to the pair-list reading systematically in different types of A'-dependencies. When a pronoun possesses at the same time the extended form and the reduced form as internal structures, it only gives rise to the pair-

Semantic level  

list reading in the A'-dependencies derived by Agree in which the EPP feature is required. By contrast, when it enters into the derivation of an A'-dependency derived by Match in which the EPP feature is not required, it does not give rise to the pair-list reading. The reader can also see the logical coherence between Chapter 3 and Chapter 4 in that it is always the NP part in the extended form that triggers all of the interpretative effects under reconstruction (cf. Chapter 3) and gives rise to the pair-list reading (cf. Chapter 4). If a resumptive construction does not display any reconstruction effects, it does not give rise to the distributive readings either. Under such a reasoning, the crucial factor that determines the availability of the pair-list reading in an A'-dependency is still the syntactic mechanism that derives such an A'-dependency. Derivational mechanisms determine whether or not the EPP feature is activated. The (de)activation of the EPP feature determines whether or not the element located on the variable site enters into the derivation with the extended form. Then it is the extended form that triggers the reconstruction effects as well as the pair-list reading. The following table summarizes the distribution of these readings. Tab. 3: Specificity in Mandarin Chinese Relatives

Dislocations

Resumptives Gaps

ta

qi

Gaps

ta

qi

ta

qi

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Non-specific reading

yes

yes

yes

----

---

----

----

----

Specific reading

yes

yes

yes

----

---

----

----

----

Functional reading

yes

yes

yes

yes

no

yes

no

yes

Indefinite reading

yes

yes

yes

----

---

----

----

----

Binding

yes

yes

yes

yes

no

yes

no

yes

[D-ϕ] NP]

[[D-ϕ] NP]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

Distributive reading

Intrusives

[[D-ϕ] NP

Individual reading

Resumptives

Pairlist

Structure interne

  Concluding remarks

. Architecture of resumptivity .. Interaction of different factors involved in the derivation of an A'dependency Chapters 2, 3 and 4 show the entire picture of resumptivity in Chinese from a comparative perspective, which enables us to come up with a coherent analysis of resumptivity. I showed earlier that different factors intervene during the course of the derivation of a resumptive dependency and that these factors are subject to an interaction. In the following paragraphs, these factors are categorized according to two different aspects of the problem. (A) Concerning A'-dependencies (i) two types of A'-dependencies: relatives and LD-structures; (ii) two types of derivational mechanisms: Agree and Match; (B) Concerning A'-bound elements (iii) three types of A'-bound elements: gaps, the grammatical/systematic use of the resumptive pronoun and the intrusive use of the resumptive pronoun; (iv) two types of internal structures of an A'-bound element: extended form and reduced form; (v) two types of pronouns: ta and qi. Let us examine these elements one by one. (A) Concerning A'-dependencies Derivational mechanisms choose the type of A'-dependency on which they operate. Relatives (with a gap or with a resumptive pronoun) and LD-structures with a gap are derived by Agree; resumptive LD-structures are derived by Match. These mechanisms determine what type of resumptive pronoun appears in what type of syntactic construction. Agree is subject to locality constraints and it displays island effects and crossover effects, which means that the intrusive use of the resumptive pronoun cannot exist in relatives. By contrast, Match is not subject to locality constraints and it does not display island effects nor crosso-

Architecture of resumptivity  

ver effects, which determines that resumptive pronouns can be used as intrusives in LD-structures. Crucially, I also propose that locality conditions can be reduced to phasal locality. Derivational mechanisms also determine what kind of feature is associated with what kind of A'-dependency. For instance, the EPP feature is activated in relatives (with a gap or with a resumptive pronoun) and in LD-structures with a gap. By contrast, EPP is not necessarily activated in resumptive dislocation constructions because topics in these structures are always base-generated directly in [Spec, TopP]. Logically, the presence of the EPP feature requires the presence of an NP in the internal structure of the A'-bound element (i.e. gaps or resumptive pronouns) on the variable site in the relevant A'-dependencies. As we have seen, it is the presence of the NP in the extended form of the internal structure that triggers different kinds of reconstruction effects and gives rise to the distributive reading. This is the reason why only the extended form can display reconstruction effects and give rise to the par-list readings in all kinds of A'-dependencies. On the contrary, the reduced form lacking the presence of the NP does not trigger any reconstruction effect and does not give rise to any distributive reading either. (B) Concerning A'-bound elements Different types of pronouns have different internal structures. Certain elements possess only one specific form as internal structure. For instance, gaps and the pronoun qi only have the extended form, which explains why both of them systematically display reconstruction effects in different types of A'dependencies: either relatives or LD-structures. As for the pronouns that have two different forms as internal structures, such as the pronoun ta, they never decide to enter into any type of A'-dependency with any sort of internal structure. It is the EPP feature of the Probe that correlates with the specific form of the pronouns to use. .. Two predictions In addition to all of the reasoning and conclusions that I made previously, we can make two more interesting predictions. First, my analysis would be stronger if I could find a third type of A'-bound variable (i.e. an A'-bound pronoun) that only possesses the reduced form [D-ϕ] but not the extended form [[D-ϕ] NP] as internal structure. We can imagine for the moment that such an element, say X, exists. If X exists, it will have a complementary distribution with respect to the pronoun qi concerning the reconstruction effects. The expected result of the tests is presented in the following table.

  Concluding remarks The imagined element X enters into the derivation of a relative clause or of an LD-structure with its reduced form. It does not give rise to any reconstruction effect nor to pair-list reading. Tab. 4: Complementary distribution of qi and of the element X Relatives

Dislocations

Resumptives

Resumptives

Intrusives

X

qi

X

qi

X

qi

(i) Quantificational antecedent

no

yes

no

yes

no

yes

(ii) Reconstruction of the quantifier scope

no

yes

no

yes

no

yes

(iii) Reconstruction of the anaphoric binding

no

yes

no

yes

no

yes

(iv) Condition C effects

no

yes

no

yes

no

yes

(v) Pair-list reading

no

yes

no

yes

no

yes [[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

[[D-ϕ] NP]

[D-ϕ]

Internal structure

Secondly, my analysis also predicts that an A'-bound pronoun that has two internal structures, an extended form and a reduced form, will not enter into the derivation of the A'-dependencies established by Agree with its reduced form [D-ϕ]; nor will it enter into that of the A'-dependencies established by Match with its extended form [[D-ϕ] NP]. In other words, my analysis also predicts the existence of a pronoun that is only [D-φ] waiting to be discovered. If this type of pronoun exists, then it means that derivational mechanisms will finally not require a specific form of the resumptive pronoun that enters into the derivation and that the EPP feature is not required either in the A'-dependencies derived by Agree. Thus, the choice between two available forms will be free. Future work is needed.

What Chinese tells us about resumptivity  

. What Chinese tells us about resumptivity What makes a resumptive pronoun special is that on the one hand, it behaves like a wh-trace in that it is located in a variable position bound by an operator in an A'-position; and on the other hand, it differs from a wh-trace in that it is a morphologically overt pronoun. A crucial research question that all of the existing work on resumptivity poses is how to understand the nature of the resumptive pronouns. In order to get a clear idea on this question, a comparative approach has usually been adopted by comparing a resumptive dependency with an A'-dependency with gap. First, from the typological point of view, the existing work on languages like Vata (Koopman 1983) and Swedish (Endahl 1985) argues that a resumptive pronoun is simply treated as a spelled out trace in that it is subject to all of the syntactic constraints that a wh-trace obeys. Unfortunately, in these languages the syntactic distribution of the resumptive pronouns is highly restricted. Furthermore, resumptive constructions in these languages display island sensitivity and crossover effects, which are traditionally treated as classical diagnostic tests for A'-movement in the GB framework. There is also research on resumption in languages like Welsh (Rouveret 1994, 2002, 2008, to appear), Irish (McCloskey 1990, 2001) and different Arabic dialects (Aoun et al. 2001, Demirdache 1991, Guilliot 2006 and Malkawi 2009) which reveals that resumptive pronouns behave quite differently from wh-traces and importantly, these differences are not only syntactic but also semantic. My survey of the resumption based on Chinese data in this study illustrates clearly that gaps and resumptive pronouns (in their grammatical/systematic use and in their intrusive use) both exist independently. More importantly, this study shows that there is no such context in which a gap and a resumptive pronoun are free alternatives. Of course, this is only true for languages like Chinese but not for languages like Irish (cf. Chapter 2). Secondly, still from the typological point of view, it is interesting and also important to know whether there is a so-called macro-variation within the languages that use the resumptive strategy. It has been shown that resumptive pronouns quite often play the role of the last resort to prevent a sentence from eventually violating locality constraints. Thus, it is important to know whether one particular langue uses only one specific resumptive strategy. The point of view based on macro-variation defends the idea that generally there are two kinds of languages concerning the use of the resumptive strategy. Certain languages only use a resumptive pronoun grammatically/systematically in order to spell out phonologically a trace left after an A'-movement and these resumptive pronouns are highly restricted in certain specific syntactic positions. Vata and

  Concluding remarks Swedish represent this kind of languages. At the same time, there are also other languages that only use resumptive pronouns as a last resort to avoid the potential violation of locality constraints, i.e. the intrusive use. However, much existing work shows that the point of view based on a macro-variation is not on the right track because an important number of natural languages use resumptive pronouns in both ways: in general use and in intrusive use. The study of Chinese also confirms this critical point of view. Chinese data suggest that the factor that plays a crucial role in determining the function of a resumptive pronoun in an A'-dependency is in fact the derivational mechanism that operates on this dependency. This echoes the observed generalization that a relative clause does not allow the intrusive use (i.e. last resort use) of a resumptive pronoun but a resumptive LD-structure allows it. In my analysis, there are not two kinds of resumptive pronouns: grammatical resumptives and intrusives. Concretely, it is not the insertion of a resumptive pronoun (in its intrusive use) that prevents a sentence from eventually violating locality constraints; instead, it is the case that the derivation of those sentences is not subject to locality constraints in the first place. Otherwise, it will be extremely difficult to explain why an intrusive pronoun functions as a last resort only in dislocation structures but not in relatives. This is one of the main proposals that I try to make in this study. In other words, the boundary between a true resumptive pronoun, i.e. the grammatical/systematic use of a resumptive pronoun, and the intrusive use of a resumptive pronoun, simply does not exist. The potential violation (or not) of locality constraints is totally under the responsibility of the derivational mechanisms. Relatives and LD-structures with gaps are derived by Agree that is an operation subject to locality constraints. The nature of Agree makes it impossible to use an intrusive pronoun in the above structures. By contrast, a resumptive dislocation structure is derived by Match which is not subject to locality constraints and naturally, it allows the intrusive use of a resumptive pronoun in island contexts. Therefore, Chinese shows that there is no problem for a given language to use resumptive pronouns in two different ways and accordingly, the macro-variation point of view cannot be maintained.

 Appendix: General distribution of the resumptive pronouns in Mandarin Chinese . Introduction This appendix provides the necessary information on the use of resumptive pronouns in Mandarin Chinese. I will begin by giving a description of the basic use of personal pronouns in Chinese in Section 6.2. Before introducing the general syntactic distribution of resumptive pronouns, I will introduce several main types of A'-dependencies in which the presence of resumptive pronouns is legitimized. Concretely, I will examine wh-dependency, relativization, topicalization, ex-situ cleft-focus structures, wh-fronting, and Across-The-Board (ATB)constructions (cf. Section 6.3). In Section 6.4, I will examine the general distribution of the personal pronouns used as resumptives in different contexts. Specifically, their syntactic distribution will be discussed in two different situations: the cases in which the use of a resumptive pronoun is obligatory and those in which its use is optional or prohibited.

. Personal pronouns in Chinese In this section I will give a brief introduction on the basic use of the personal pronouns in Chinese. Pronouns in Chinese do not display systematic morphological change, i.e. there is no change of grammatical case. The same morphological form of a pronoun can be used as a subject, object or prepositional object. It is worthwhile noting that the suffix –men is considered as a marker of plurality of personal pronouns in traditional grammar. Despite the absence of gender markings in the phonetic representation of these pronouns, gender distinction can be sometimes represented in their graphic forms (i.e. character writing). In the traditional style of characters, gender is visibly marked in the radical of pronouns of the second person. For instance, “你” ni (2MSg) contains the radical “亻” which means “person” and represents the masculine gender. By contrast, “妳” ni (2FSg) contains the radical “女” which refers to the feminine gender. As for the pronouns for the third person, gender markings are also represented in graphic form of the characters. The three pronouns of the third person have exactly the same pronunciation ta, but they do not have the same written form in their graphic representations. In order to make a distinction between

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese these three, I indicate them as ta1, ta2 and ta3. The following table presents the paradigm of the pronouns in Chinese. Tab. 1: Pronoun system in Chinese

 person st

nd person

Singular

Feature

Gloss.

Plural

Feature

Gloss.

我 wo

 person, singular

Sg

我們 wo-men

 person, plural

Pl

你 ni

妳 ni

st

nd person, masculine, singular nd person, feminine, singular

他 ta

 person, Masculine, singular

她 ta

rd person, Feminine, singular

MSg

FSg

你們 ni-men

妳們 ni-men

rd

rd person

它 ta

rd person, [- human] (object, animal), singular

MSg

FSg

OSg

他們 ta-men

她們 ta-men 它們 ta-men

st

nd person, masculine, plural

MPl

nd person, mixed-gender, plural

Pl

nd person, feminine, plural

FPl

rd person, Masculine, plural ;

MPl

rd person, Pl mixed-gender, plural rd person, Feminine, plural rd person, [- human] (object, animal), plural

FPl

OPl

. Syntactic constructions involving A'-dependencies in Chinese In this section I will present several important syntactic constructions that contain A'-dependencies in Chinese. An A'-dependency is formed by an antecedent located in a non-argument (i.e. A') position and a variable located in an argument (i.e. A) position bound by this antecedent. In the generative framework, positions such as [Spec, VP], [Spec, vP] and [Spec, TP] are considered as A-

Syntactic constructions involving A'-dependencies in Chinese  

positions; while [Spec, CP], [Spec, TopP] and [Spec, FocP] are A'-positions. In fact, specifier position of any functional projection which is located in the leftperiphery is an A'-position. Now let us examine each type of A'-dependency one after another. .. Wh-dependency In languages with wh-movement, wh-words move from their original argument position to the [Spec, CP] position, which is an A'-position, in order to get a scope and to be interpreted properly. The head C° of CP is encoded with different illocutionary forces; in the case of wh-questions, the C is responsible for the interrogative force. The C-head attracts the morphological form of the relevant wh-phrase to move to the the [Spec, CP] position. The moved wh-phrase, which is translated as an operator at LF, and the trace that it leaves behind form a whchain which is considered as an A'-dependency. In languages like English and French, resumptive pronouns cannot appear in the wh-trace position (i.e. a gap position), as illustrated in (1). (1) a. What1 did you buy yesterday e1/*it1 ? lu b. Quel livre1 (*l’) avez-vous what book it have-you read ‘What book did you read yesterday?’

e1

hier ? yesterday French

In (1a), only the gap represented by e is allowed in the base position of the moved wh-word what and the resumptive pronoun it is prohibited in that position. The same situation is also found in French (cf. 1b): only the gap e is allowed in the base position of quel livre ‘what book’ and the preverbal direct object clitic pronoun le ‘it’ is banned. By contrast, in Irish, the presence of a resumptive pronoun is allowed in a wh-question. Let us examine two examples in McCloskey (1990). (2) a.

d’inis siad cén turas a told they what journey COMPpro said air pro. they on-3SNG-MASC ‘They told what journey they were on (it).’

raibh be-PAST

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese b.

cén t-oigigeach ar shíl tú which officer COMPpro thought you go mbeadh sé i láthair ? COMP would-be he present ‘Which officer did you think would be present?’ Irish, McCloskey (1990)

(2a) contains an indirect wh-question with the wh-word what moving from its base position to the embedded [Spec, CP] position. The original argument position of what is occupied by a null pro and the verb also takes a special morphological form which is considered as a resumptive pronoun. (2b) is a root whquestion with the wh-phrase cén t-oigigeach ‘which officer’ moving from its base embedded subject position to the matrix [Spec, CP] position. Instead of a trace, the base position of which officer is occupied by a resumptive pronoun sé ‘he’. Mandarin Chinese is known as a strict wh-in-situ language; wh-words stay in their base position without moving to the sentence-initial scope position, contrary to their counterparts in French and in English. (3)

你喜歡吃什麼? Ni xihuan chi

shenme ?

2Sg like eat what ‘What do you like eating?’ Huang (1982) argues that all of the in-situ wh-words undergo movement at LF; Tsai (1994) argues that only wh-adverbs undergo LF-movement and that whnouns are unselectively bound by a null interrogative operator Op located in the [Spec, CP] position. Lacking overt wh-movement in Chinese, there is no available empty position to host a resumptive pronoun in wh-questions, which is the reason why it is impossible to test the distribution of resumptive pronouns in wh-questions in Chinese. .. Relativization Relative clauses constitute another class of constructions that involve A'dependencies. Generally, resumptive pronouns cannot stay in simple relative clauses which do not contain any island in English and in French. (4) a.

the movie1 [that I saw (*it1)]

Syntactic constructions involving A'-dependencies in Chinese  

b.

la

fille1

[que

Paul

the girl that Paul ‘the girl that Paul met.’

(*l1’)

a

rencontrée]

her

has

met French

However, Ross (1967) shows that resumptive pronouns can only be used intrusively in relative clauses with an island in English. (5) I just saw a girlj [whoj Long John’s claim [that shej was a Venusian]] made all the headlines. Ross (1967) In (5), there are two subordinate clauses, both of which are considered as Complex-NP islands for A'-movement. The outer one is a relative clause who Long John’s claim that… and the inner one is a complement clause of noun that she was a Venusian. The complement clause of noun is considered as an island that blocks the relativization of the antecedent NP a girl from the subject position inside the complement clause due to the Subjacency condition and ECP. When the pronoun she is inserted in the trace position, the sentence becomes grammatical. In this sense, the resumptive pronoun she functions intrusively to avoid eventual violation of locality conditions. It is also worthwhile noting that the relative pronoun who in the [Spec, CP] position, translated as an operator, binds the in-situ resumptive pronoun she as a variable; they share the same index j and they form an A'-dependency. Similarly, French also allows the presence of an intrusive pronoun in the subject position of an embedded clause, as shown in (6). (6)

La fille1 [que Jean m’a dit [qu’ elle1 the girl that Jean me.has told that she viendrait aujourd’hui]] n’est pas encore arrivée. would.come today Neg.is yet arrived ‘The girl1 who [Jean told me that [she1 would come today]] has not arrived yet.’

The syntactic distribution and the different properties of the resumptive pronouns in relative clauses have received much attention in the relevant literature. Here are two frequently cited examples from Hebrew that have been described in detail in Chapter 1.

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese (7) a.

b.

raiti et ha-yeled šesaw-I Acc the-boy that ‘I saw the boy that Rina loves.’

rina Rina

ohevet loves

(oto). him

raiti

et

še

dalya

makira

et

saw-I

ACC the-boy

that

Dalya

knows

ACC

ha-yeled

ha-iša

še

ohevet

the-woman

that

loves

*(oto). him

‘I saw the boyj [that Dalya knows the woman [that loves himj]].’ Hebrew, Borer (1984) Different from English and French, Hebrew allows resumptive pronouns to appear optionally in island free relative clauses as illustrated in (7a). At the same time, the insertion of an intrusive pronoun in a relative clause with an island will save a sentence form violating locality constraints, as shown in (7b). Like Hebrew, Mandarin Chinese allows optional resumptive pronouns in relatives as shown in (8). However, as will be detailed in Section 6.4, the use of resumptive pronouns is highly restricted by different factors. (8)

[院長想趕(他 j)出醫院]的那個小流氓 j 不見了。 [Yuanzhang xiang gan ta1j/ ___ chu

yiyuan]

president want evict 3MSg out hospital de na-ge xiao liumangj bu-jian-le. C that-Cl little rogue disappear-Perf ‘The little roguej [that the president wanted to kick (*himj) out of the hospital] has disappeared.’ In (8), when the relativized site is either occupied by a gap or by a resumptive pronoun, the sentence is always grammatical. As for the intrusive use of resumptive pronouns in relative clauses with an island, I have shown enough examples in Chapter 2 and I will not repeat them here due to the lack of space. Please recall that a resumptive dislocation structure involving an island never gives rise to island effects. .. Topicalization The third type of A'-dependency in Chinese is topicalization (Ning 1993; Qu 1994; Xu 2004; Paul 2002, 2005, 2014, 2015; Pan 2011b, 2014, 2015a, b, c). The

Syntactic constructions involving A'-dependencies in Chinese  

most classical analysis on topicalization argues for a movement approach in that a topic constituent moves from its base A-position to the [Spec, TopP] position (i.e. A'-position). In topicalization cases, the gap position can be optionally occupied by a resumptive pronoun, and in this case, the structure is called a Left-Dislocation structure (LD-structure). Resumptive pronouns in this type of structure agree obligatorily with the constituent in the topic position in terms of ϕ-features. The following are several examples from English, French and Chinese which show that resumptive pronouns always agree with their antecedents in person, number and gender. (9) a. b.

John1, I like him1 very much. Valérie1,

je

la1

connais

Valery 1Sg 3FSg-Acc know ‘Valery, I know her very well.’ c.

瑪麗 1,我很瞭解她 1。 Mali1, wo hen liaojie Mary 1Sg very know ‘Mary1, I know her1 very well.’

très

bien.

very

well

ta21. 3FSg

.. Ex-situ cleft-focus structures Another type of A'-dependency is ex-situ cleft-focus structures. In Chinese, the shi ‘be’…de (Comp) construction is used to form a standard cleft-focus structure (Paris 1979, Cheng 2008, Paul & Whitman 2008, Paul 2015). Generally, shi ‘be’ is placed immediately before the focused element and this focused element does not move to the sentence-initial position. (10)

我是昨天到柏林的。 Wo shi zuotian

dao

Bolin

1Sg be yesterday arrive Berlin ‘It was yesterday that I arrived in Berlin.’

de. DE

It has also been pointed out that under certain specific conditions, a constituent can be fronted and, at the same time, be marked by the copula shi ‘be’, as shown in (11a) (cf. Pan 2014). In the same environment, de in the original cleft shi ‘be’…de construction must be dropped; otherwise the sentence will be ungrammatical (cf. 11b). The patterns like (11a) is called “ex-situ cleft-focus struc-

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese ture” and the one in (10) “in-situ cleft-focus structure”. In (10), the focus NP zuotian ‘yesterday’ stays in its original position without moving to the sentenceinitial position. (11) a.

是[你的態度]j,公司的老板不欣賞 tj。 Shi [ni-de taidu]j, gongsi-de

laoban

be your attitude company-DE boss bu xinshang tj. Neg appreciate ‘It is your attitude that the boss doesn’t appreciate.’ b.

*是[你的態度]j,公司的老板不欣賞 tj 的。 *Shi [ni-de taidu]j, gongsi-de

laoban

be your attitude company-DE boss bu xinshang tj. de. Neg appreciate DE (‘It is your attitude that the boss doesn’t appreciate.’) The NP ni-de taidu ‘your attitude’ is originated in the object position of the verb xinshang ‘appreciate’ and it is fronted to the sentence-initial cleft-focus position and is marked by the copula shi ‘be’ (cf. 11a). However, such a fronting is not tolerated in cleft-constructions with shi ‘be’…de (cf. 11b). This is a crucial contrast between the two different cleft-constructions. There are several important specific constraints on the use of this ex-situ cleft-focus and from this perspective, topicalization and ex-situ cleft-focus structures behave differently in that they are not subject to the same constraints. For instance, Pan (2014) shows that the extraction of a constituent from a predicate encoding an episodic eventuality in order to form an ex-situ cleftfocus structure will render the sentences ungrammatical (cf. 12b); while a topicalization from the same type of predicate is possible, as shown in (12a). (12) a.

Topicalization 你的鑰匙,我在花園裡找到了。 [Ni-de yaoshi], wo zai your key 1Sg at zhaodao-le tj. find-Perf ‘Your key, I found (it) in the garden.’

huayuan-li garden-in

Syntactic constructions involving A'-dependencies in Chinese  

b.

Cleft-focus structure *是你的鑰匙,我在花園裡找到了。 *Shi [ni-de yaoshi], wo

zai

be your key 1Sg at zhaodao-le tj. find-Perf (‘It was your key that I found in the garden.’)

huayuan-li garden-in

In the above sentences, when the resultative action verb zhao-dao ‘find’ is used and marked by the perfective aspectual marker -le, the predicate encodes an episodic eventuality (cf. Zhang 2002 for the semantics of episodicality). From this type of predicate, topicalization of the NP ni-de yaoshi ‘your key’ is allowed but the fronting of the same NP to form an ex-situ cleft-focus structure is banned. Here are a couple of more examples with different action verbs such as mai ‘buy’ (cf. 13) and qu ‘go’ (cf. 14). These verbs combined with the perfective aspectual marker -le also create predicates encoding episodic eventualities from which an extraction for topicalization purpose is licit (cf. (a) cases in the following examples) and an extraction to form an ex-situ cleft-focus structure is illicit (cf. (b) cases). (13) a.

Topicalization 辭典(呢),她昨天买了。 Cidian (ne), ta2

zuotian

mai-le.

dictionary Top 3FSg yesterday buy-Perf ‘As for the dictionary, she bought (it) yesterday.’ b.

Ex-situ cleft-focus structure *是辭典,她昨天买了。 *Shi cidian, ta2

zuotian

be dictionary 3FSg yesterday (‘It was a dictionary that she bought.’) (14) a.

Topicalization 巴黎(啊),他去年去了。 Bali (a), ta1 qunian

qu-le.

Paris Top 3MSg last.year go-Perf (‘As for Paris, he went (there) last year.’)

mai-le. buy-Perf

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese b.

Ex-situ cleft-focus structure *是巴黎,他去年去了。 *Shi Bali, ta1 qunian

qu-le.

be Paris 3MSg last.year go-Perf (‘It was Paris that he went to last year.’) The majority of the native speakers that I consulted totally rejects sentences in (12b, 13b and 14b) and automatically corrects them by using pseudo-cleft constructions, as shown in (15). (15) a.

我在花園裡找到的是你的鑰匙。 [DP [Wo zai huayuan-li zhaodao]

de

1Sg at garden-in find shi ni-de yaoshi. be your key ‘(What) I found in the garden is your key.’ b.

她昨天买的是辭典。 [DP [Ta zuotian

mai]

de

Ø]

DE

shi

3FSg yesterday buy DE be ‘(What) she bought yesterday was a dictionary.’ c.

他去年去的是巴黎。 [DP [Ta qunian

qu]

de

Ø]

Ø]

shi

3FSg last.year go DE be ‘(The place) where he went to last year is Paris.’

cidian. dictionary

Bali. Paris

The sentences above contain the so-called pseudo-cleft constructions that are of course fundamentally different from the real cleft-focus structures. Importantly, pseudo-cleft structures contain a null head noun Ø and they are roughly paraphrased as free relatives in English. This type of sentence is called pseudo-cleft only because there are copula shi ‘be’ and the particle de in the sentence; however, shi ‘be’ and de do not have any focalization function in pseudo-cleft structures. Shi ‘be’ here functions as a real copula and de serves to link the relative clause to the null head noun. Pseudo-cleft constructions are very often used when the foci are objects because it is well documented that with the original cleft-structure shi ‘be’…de, it is impossible to focus on direct objects.

Syntactic constructions involving A'-dependencies in Chinese  

All of the offending cases involve an extraction of a focus NP from the predicates that encode episodic eventualities. Verbs in the resultative compound form, such as zhao-dao ‘find’ (cf. 12) or verbs marked by the perfective aspectual marker -le, such as mai-le ‘(have) bought’ (cf. 13) and qu-le ‘went’ (cf. 14), create typical predicates that encode episodic eventualities. However, when there is an extra component that can make an episodic eventuality non-episodic-like the sentence will be acceptable again. For instance, the insertion of the durational post-verbal complements such as san-ge xiaoshi ‘three hours’ in (16a) and yi zheng ge xiawu ‘a whole afternoon’ in (16b) or the insertion of the post-verbal complements of frequency such as san ci ‘three times’ in (16c) will render the relevant sentences grammatical by creating non-episodic eventualities. (16)

a.

Ex-situ cleft-focus structures in non-episodic eventualities 是你的鑰匙,我昨天在花園裡找了三个小时。 Shi [ni-de yaoshi], wo zuotian zai be your key 1Sg yesterday at huayuan-li zhao-le san-ge xiaoshi. garden-in look-for-Perf three-Cl hours ‘It was your key that I was looking for (it) during three hours yesterday in the garden.’

b.

是本辭典,她昨天居然买了一整個下午。 Shi [ben cidian], ta2 zuotian

juran

be Cl dictionary 3FSg yesterday unexpectedly mai-le yi zheng ge xiawu. buy-Perf one entire Cl afternoon ‘It was (for buying) a dictionary that it took her a whole afternoon yesterday.’ c.

是巴黎,他去年去了三次。 Shi Bali, ta1 qunian

qu-le

san

be Paris 3MSg last.year go-Perf three ‘It was Paris that he went to three times last year.’

ci. times

One should note that it is not the action verbs alone that determine whether episodic eventualities are created; instead, it is the entire predicate that should be considered as a whole that creates such eventualities. There are also many other ways to create non-episodic eventualities. For instance, when the verb chi ‘eat’ is marked by the perfective aspectual marker -le,

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese it creates an episodic eventuality predicate from which the fronting of zuotian na-dun fan ‘the dinner of yesterday’ was banned in an ex-situ cleft-focus structure, as illustrated in (17a). However, if a post-verbal complement hen kaixin ‘very happy’ introduced by the structural particle de is inserted in the sentence containing the same action verb chi ‘eat’, the fronting of the NP zuotian na-dun fan ‘the dinner of yesterday’ is perfectly licit in an ex-situ cleft-focus structure, as shown in (17b). In traditional grammar of Mandarin, the post verbal complement de hen kaixin ‘very happy’ headed by de describes a state of eating, which makes the relevant sentence stative like and therefore, it creates a predicate entailing a non-episodic eventuality. Instead, the original action verb chi ‘eat’ will not be able to create a predicate entailing an episodic eventuality. As a result, the extraction of NP zuotian na-dun fan ‘the meal of yesterday’ to an exsitu cleft-focus position is licit. (17) a.

*是昨天那頓飯,我吃了。 *Shi zuotian na-dun

fan,

wo

be yesterday that-Cl diner 1Sg (‘It was the dinner of yesterday that I had.’) b.

是昨天那頓飯,我吃得很開心。 Shi zuotian na-dun fan,

chi-le. eat-Pert

wo

be yesterday that-Cl diner 1Sg chi de3 hen kaixin. eat DE very happy ‘It was (for) the dinner of yesterday that I was very happy.’ There are two different existing analyses on ex-situ cleft-focus structures. One is proposed by Xu (2004) who analyzes the copula shi ‘be’ as a focus marker on par with the particles ne or a as topic makers. However, configurationally, this analysis encounters many difficulties when applied to ex-situ cleft-focus structures presented here. Let us examine the following pairs. (18) a.

Topicalization 你的態度 j 呢,公司的老板不欣賞 tj。 [Ni-de taidu]j, ne, gongsi-de

laoban

your attitude Top company-DE boss bu xinshang tj. Neg appreciate ‘As for your attitude, the boss doesn’t appreciate.’

Syntactic constructions involving A'-dependencies in Chinese  

b.

Ex-situ cleft-focus structure 是[你的態度]j,公司的老板不欣賞 tj。 Shi [ni-de taidu]j, gongsi-de

laoban

be your attitude company-DE boss bu xinshang tj. Neg appreciate ‘It is your attitude that the boss doesn’t appreciate.’ (18a) is derived by fronting the NP ni-de taidu ‘your attitude’ to the sentenceinitial topic position and the topic can be optionally followed by so-called topic markers, such as ne or a. Paul (2002, 2005, 2014, 2015), Pan (2015), Pan & Paul (2016), Paul & Pan (to appear) treat these topic markers as heads projecting TopicP in Chinese. Topic phrases occupy the [Spec, TopP] position and TopP takes the TP from which the topic NP is extracted as complement. Accordingly, (18a) has the following structure. In (19), the so-called topic marker ne is directly generated under the head Top° and the topic phrase your attitude moves from its base-position inside the TP to [Spec, TopP].35 (19) Topicalization

TopP

Your attitudej Top° ne

Top'

TP

the boss doesn’t appreciate tj However, if I follow Xu (2004)’s view to treat the copula shi ‘be’ as a focus marker on par with the particles ne or a as topic makers, configurationally, there is a serious problem due to the fact that shi ‘be’ precedes the focused element and not follow it. Obviously, shi ‘be’ cannot take the focus NP as comple-

 35 An alternative way is to say that the topic phrase your attitude is generated directly under the specifier position of TopP. Since island effects are always absent from a context involving an episodic eventuality predicate, it is not certain that (19) is necessairily derived by movement.

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese ment because otherwise, shi ‘be’ cannot take the TP as another complement, which is an undesirable situation as illustrated in (20). (20) Impossible tree 1: * FocP

Complement the boss doesn’t appreciate. Foc° Shi ‘be’

Complement your attitude

One cannot put ni-de taidu ‘your attitude’ in the [Spec, FocP] position either because it will create a linearization problem, as shown in (21). (21) Impossible tree 2: *FocP

Complement the boss doesn’t appreciate.

Foc° Shi ‘be’

Spec your attitude

For these technical reasons, it does not seem a good idea to treat the copular shi ‘be’ as a focus marker occupying the head position of the focus projection. Paul (2015) has a different analysis in which the copula shi ‘be’ is treated as the main verb of the entire sentence and it takes the following clause as its complement. In this analysis, the focused constituent is located in the left periphery of the complement clause, not in the periphery of the matrix clause. Her conclusion is that there is no focus cleft projection in the matrix left periphery in Chinese. Following these lines (18b) will have a structure as shown below in (22). The reader should pay attention to the position of the XP in (22); XP is considered as a peripheral position of the TP2 that is roughly treated an equivalent of subordinate clause.

Syntactic constructions involving A'-dependencies in Chinese  

(22)

TP1

T° Shi ‘be’

XP

your attitude

TP2

the boss doesn’t appreciate. There are two potential problems in this analysis. Treating the copula shi ‘be’ as the main predicate will necessarily imply that it scopes over the entire sentence; however, semantically, it doesn’t seem to be quite the case. In ex-situ cleft-focus cases, it is the clefted NP that receives a contrastive focus interpretation but not the entire XP. For instance, (23) a.

是[你的態度],公司的老板不欣賞,不是[你的衣著]。 Shi [ni-de taidu], gongsi-de laoban

bu

be your attitude company-DE boss Neg xinshang, bu shi [ni-de yizhuo]. appreciate Neg be your dressing ‘It is [your attitude] that the boss doesn’t appreciate but not [your way of dressing].’ b.

*是[你的態度,公司的老板不欣賞],不是[你的衣著,客戶很不滿 意]。 *Shi [ni-de taidu, gongsi-de laoban bu be your attitude xinshang], bu shi appreciate Neg be hen bu manyi]. very Neg satisfied

company-DE boss [ni-de yizhuo, your dressing

Neg

kehu client

(*‘[It is your attitude that the boss doesn’t appreciate] but not [it is your way of dressing that the client is not satisfied at all].’) In (23a), the contrastive reading is only realized on the NP ni-de taidu ‘your attitude’; the sentence is still grammatical when we put a contrastive focus NP

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese ni-de yizhuo ‘your way of dressing’ at the end of the sentence. However, (23b) shows that the TP cannot be contrasted since adding another contrastive TP it is your way of dressing that the client is not satisfied at all will make the entire sentence ungrammatical. This pair of examples shows that in an ex-situ cleft-focus structure, the copular shi ‘be’ only scopes over the NP that immediately follows it and that shi ‘be’ cannot scope over the entire TP. In the tree presented in (21), it seems difficult to argue why the contrastive focus reading must be realized on ni-de taidu ‘your attitude’ but not on the entire XP that is also under the scope of shi ‘be’. Another problem is that it is somehow difficult to determine the structural position of ni-de taidu ‘your attitude’ in (21). In other words, it is difficult to see the nature of XP whose specifier is occupied by ni-de taidu ‘your attitude’. Recall that XP is located in the leftperiphery of the embedded TP. Logically, we have two options: one is to say that XP is actually TopP and the other is to say that XP is a FocP. Clearly, XP cannot be TopP because if it is so there is no way to explain why it will get an exclusive focus reading when preceded by the copula shi ‘be’. Semantically, it will be difficult to argue that the copula shi ‘be’ can transform a topic structure into an ex-situ cleft-focus structure. Alternatively, if we argue that XP is a FocP, such an assumption will lead to a false prediction that the sentence in (24) is ambiguous between two structures: one is a topic structure and the other one is a focus structure. As we know, ni-de taidu ‘your attitude’ in (24) can only be interpreted as topic but not focus and therefore, (24) cannot imply a focus structure. Even if we can successfully prove that XP is FocP, then it will still be difficult to argue why (24) needs the main copula shi ‘be’ at all since under such an assumption, ni-de taidu ‘your attitude’ is already interpreted as a contrastive focus. (24)

你的態度 j,公司的老板不欣賞 tj。 [Ni-de taidu]j, gongsi-de

laoban

your attitude company-DE boss bu xinshang tj. Neg appreciate ‘Your attitude, the boss doesn’t appreciate.’ In this section, I will propose my own analysis of ex-situ cleft-focus structures in Chinese. I assume that there is a matrix Focus projection headed by a null Foc°. The clefted focus phrase marked by shi ‘be’ is a TP occupying the specifier position of FocP. FocP takes TP2 the boss doesn’t appreciate as complement, as shown in (25).

Syntactic constructions involving A'-dependencies in Chinese  

(25)

FocP TP1

T° Shi ‘be’

NP your attitude

Foc'

Foc°

TP2

the boss doesn’t appreciate. One of the advantages of this analysis is that it can nicely account for the fact that only the clefted NP ni-de taidu ‘your attitude’ receives a contrastive focus interpretation but not the entire sentence. Thus this analysis confirms the intuition that there is an adjacency relation between shi ‘be’ and the clefted NP. A second advantage is that it clearly points out that the nature of [shi ‘be’ + NP] is focus and avoids the awkward assumption that (24) is ambiguous between a topic structure and an ex-situ cleft-focus structure. In my sense, the fronting NP ni-de taidu ‘your attitude’ can only be interpreted as topic but not as focus. The sequence [shi ‘be’ + NP] is placed in the [Spec, FocP] position headed by a null Foc° element, which thus ensures its focus status. Importantly, I put shi ‘be’ under the head of TP and in this sense, shi ‘be’ is analyzed as a normal verb but not as a so-called focus marker. This is so because shi ‘be’ can be modified by different adverbs, such as jiu ‘exactly’ or negative adverb bu ‘not’. In both cases, the adverbs only modify the sequence [shi + NP] but not the entire TP. (26)

[就是昨天那頓飯],大家都吃得很不開心。 [FocP [TP1 Jiu shi zuotian na-dun exactly

fan],

be yesterday that-Cl dinner dou chi de hen bu kaixin]]. everyone all eat DE very Neg happy (i) ‘It was exactly for the dinner of yesterday that everyone was not very happy (, not for the dinner of the day before yesterday).’ (ii) (* ‘It is exactly the case that it was for the dinner of yesterday that everyone was not very happy(, but not the case that it was for the service of yesterday that everyone was very happy.’)

[TP2 dajia

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese (27)

[不是你的態度 j ],公司的老板不欣賞 tj。 [FocP [TP1 Bu shi ni-de taidu j],

[TP2 gongsi-de

laoban

not be your attitude company-DE boss bu xinshang tj]]. Neg appreciate (i) ‘It is not your attitude that the boss doesn’t appreciate (, it is your way of dressing).’ (ii) (* ‘It is not the case that it is your attitude that the boss doesn’t appreciate (, but the case that it was your way of dressing that the client likes very much.’) In (26), the adverb jiu ‘exactly’ can only scope over TP1 shi zuotian de na-dun fan ‘it was the dinner of yesterday’ but not over TP2 and therefore in the paraphrase (26i), the contrastive meaning can only be realized on TP1. If the adverb jiu ‘exactly’ also scopes over TP2, then the paraphrase with the contrastive meaning on the entire sentence should be correct, however, this is not the case as illustrated in (26ii). Similarly, the negation only scopes over TP1 in (27) but not over TP2, which is demonstrated by the fact that the paraphrase in (27i) with the contrastive reading on TP1 is the only possible interpretation. Paraphrase in (27ii) with the contrastive focus on the entire sentence is illicit. Ex-situ cleft-focus structures allow the presence of a resumptive pronoun and the following is an example from Irish. In (28), the NP Tig beag caol ‘a narrow little house’ is clefted to the sentence initial position and its original position is marked by an agreement marker on the preposition ann ‘in-it’. (28)

Tig beag caol ar mhaireamar house little narrow COMPpro live-PAST-1PL ann pro in-3SNG-MASC ‘It was a narrow little house that we lived in (it)’ Irish, McCloskey (1990)

The same observation holds for Chinese as shown in (29-30). In (29), the NP Shanghai-Lu de tianzhu-jiaotang ‘the catholic church located on the Road Shanghai’ is extracted from inside the TP2; this NP and the copular shi ‘be’ form TP1 that occupies the specifier position of FocP. Similar derivation goes for (30) where Yiqin and Xiaoqian undergoes cleft-fronting.

Syntactic constructions involving A'-dependencies in Chinese  

(29)

是上海路的天主教堂 j,他們倆昨天在那兒 j 舉行了婚禮。 [FocP [TP1 Shi [NP Shanghai-Lu de tianzhu-jiaotang]j], be Shanghai-road DE catholic.church [TP2 tamen-lia zuotian zai narj juxing-le hunli]]. 3Pl-two yesterday at there hold-Perf marriage ‘It was in the catholic church on the Road Shanghai that they had their marriage.’

The resumptive pronoun na’er ‘there’ in (29) occupies the gap position in TP2 and it shares the same index with the fronted cleft NP. In the following example (30), ta2-men ‘they-Feminine’ is also linked to the antecedent Yiqin and Xiaoqian located in an A'-position. (30)

是一勤和小倩 j,我很欣賞她們 j 刻苦的精神。 [FocP [TP1 Shi [NP Yiqin he Xiaoqianj]], [TP2 wo

hen

be Yiqin and Xiaoqian 1Sg very xinshang ta2-menj keku de jingshen]]. appreciate 3FPl diligent DE spirit ‘It is Yiqin and Xiaoqian that I highly appreciate (their) hardworking.’ In this section, I have shown that there are topicalization structures and exsitu focus structures, and they behave differently both on the syntactic and semantic level. They are subject to different constraints, for instance, in island free contexts, episodicality works for ex-situ cleft-focus but not for topicalization. Pan (2014) shows that topic and cleft-focus target different projections in the left-periphery in Chinese and his analysis is also in support of the idea according to which topics are always higher than ex-situ cleft-foci. .. Wh-ex-situ constructions It is well known that Chinese is a wh-in-situ language in that in order to form a wh-question, the relevant wh-word stays in their base position instead of moving to the [Spec, CP] position on overt syntax. Different analyses have been proposed to account for the wh-dependencies in Chinese (Huang 1982, Cheng 1991, Aoun and Li 1993, Tsai 1994, Shyu 1995). It has also been proposed that under certain specific conditions, a wh-phrase can undergo overt syntactic fronting to the sentence-initial position in the left-periphery. The phenomenon is referred to as “wh-fronting” or “wh-ex-situ” and has been studied in the generative framework (Tang 1988, Wu 1999, Cheung 2008, 2014, Pan 2011a, b,

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese 2014). Although these authors propose different solutions to wh-ex-situ in Chinese, they agree that this fronting process cannot be analyzed as a standard whmovement. Tang (1988) and Wu (1999) claim that the nature of the wh-fronting in Chinese is topicalization; Cheung (2008, 2014) argues that all of the cases involving wh-fronting must be analyzed as a cleft-focus structure. Based on these two different views, Pan (2014) claims that wh-topicalization and wh-exsitu cleft-focus structure both exist and that they are two completely different structures which are subject to different syntactic and semantic constraints. Therefore these two structures cannot be analyzed in the same way and one cannot be assimilated to another either. Any unified analysis, such as Cheung (2008, 2014), cannot be on the right track. In this section, I will make a very brief introduction of wh-ex-situ in Chinese. In the previous section, I argued that (31a) and (31b) are completely different structures: one is topicalization and the other one is ex-situ cleft-focus structure. (31) a.

Topicalization 這部電影,我很不喜歡。 [TopP [Zhe-bu dianying]j,

[TP wo

hen

this-Cl movie 1Sg very ‘As for this movie, I don’t like at all.’ b.

Ex-situ cleft-focus structure 是這部電影,我很不喜歡。 [FocP [TP Shi [zhe-bu dianying]j], bu

be this-Cl xihuan tj]].

movie

bu

xihuan

Neg

like

[TP wo

tj]].

hen

1Sg

very

Neg like ‘It is this movie that I don’t like at all.’ One can question the fronted NP in (31) to get the patterns in (32). Accordingly, (32a) is analyzed on par with (31a) as a case of wh-topicalization and (32b) is analyzed on par with (31b) as a case of wh-ex-situ cleft-focus structure. (32) a.

Topicalization 哪部電影,你很不喜歡? [TopP [Na -bu dianying]j,

[TP ni

which-Cl movie 2Sg ‘Which movie, you don’t like at all?’

hen

bu

xihuan

very

Neg

like

tj]]?

Syntactic constructions involving A'-dependencies in Chinese  

b.

Ex-situ cleft-focus structure 是哪部電影,你很不喜歡? [FocP [TP Shi [na-bu dianying]j], bu

be which-Cl xihuan tj]]?

[TP ni

movie

2Sg

hen very

Neg like ‘Which movie is it that you don’t like at all?’ Exactly like (31), the two sentences in (32) demonstrate different syntactic and semantic properties. Pan (2014) shows that sentences like (32b) are subject to the episodicality constraint but sentences like (32a) do not. This is one of the crucial arguments against the unification analysis proposed in Cheung (2008, 2014) who insists that (32a) and (32b) can be analyzed uniformly as a case of exsitu cleft-focus structure by denying the existence of wh-topicalization. For her, the presence of the copula shi ‘be’ is optional in wh-fronting. As the reader can see, this analysis cannot be on the right track since the two structures do not have the same properties. It is precisely the presence or absence of the copula shi ‘be’ that makes the crucial difference between the two structures in (32). In addition, another important difference existing between a whtopicalization and a wh-ex-situ cleft-focus structure is based on the exhaustivity test (Zubizarreta & Vergnaud 2006). It is observed that in French, a contrastively focused wh-question gives rise to an exhaustivity effect in that a list answer is excluded (cf. 33), while the normal wh-question in English (considered as an information focus), does not (cf. 34). (33) French Speaker A:

C’est [qui]C-FOC qui a écrit un livre sur les rats? ‘It is who that wrote a book about rats?’

Speaker B: # C’est [DP le chat]C-FOC qui a écrit un livre sur les rats, et c’est aussi [DP la chauve-souris]C-FOC ‘It is the cat that wrote a book about rats, and also the bat.’ (Zubizarreta & Vergnaud 2006: 525 (8, 9)) (34) Speaker A: Who wrote a book about rats? Speaker B: [DP The cat]I-FOC wrote a book about rats, and [DP the bat]I-FOC did too.

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese Pan (2014) applies such a test to Chinese. For the sake of space, I have slightly shortened his original examples. (35) shows that a question with whtopicalization (cf. 35A) allows a list of answers (cf. 35B), whereas (36) shows that a question with a wh-ex-situ cleft-focus structure (cf. 36A) only permits a unique answer (36C) but not a list of answers (cf. 36B) and thus gives rise to the socalled “exhaustivity effect”. (35) A:

B:

(36) A:

B:

Wh-topicalization: list-answer (no exhaustivity effect) 哪部電影,很多人都看過? [Na-bu dianying]i, hen duo ren which-Cl movie very many people dou kan-guo ti ? all see-Exp ‘Which moviei, many people have seen iti?’ 《愛在黎明破曉前》,很多人都看過;《愛在日落餘暉時》,很多人 也都看過。 Ai Zai Liming Poxiao Qian hen duo ren dou kan-guo; Ai Zai Riluo Yuhui Shi hen duo ren ye dou kan-guo. ‘Before Sunrise, many people have seen; Before Sunset, many people have seen as well.’ Wh-ex-situ cleft-focus structure: single answer (exhaustivity effect) 是哪部電影,很多人都看過? Shi [Na-bu dianying]i, hen duo ren be which-Cl movie very many people dou kan-guo ti ? all see-Exp ‘Which moviei is it that many people have seen iti?’ # 是《愛在黎明破曉前》,很多人都看過;也是《愛在日落餘暉時》, 很多人都看過。 # Shi Ai Zai Liming Poxiao Qian, hen duo ren dou kan-guo; ye shi Ai Zai Riluo Yuhui Shi, hen duo ren dou kan-guo. (*‘It is Before Sunrise that many people have seen; it is also Before Sunset that many people have seen.’)

Syntactic constructions involving A'-dependencies in Chinese  

C:

是《愛在黎明破曉前》,很多人都看過。 Shi Ai Zai Liming Poxiao Qian, hen duo ren dou kan-guo. ‘It is Before Sunrise that many people have seen.’

Importantly, (35A) patterned with (32a) is a case of topicalization which does not show any exhaustivity effect, whereas (36A) patterned with (32b) is an exsitu cleft-focus structure which does give rise to an exhaustivity effect. Again, wh-topicalization and wh-ex-situ cleft-focus structure are two different constructions and they cannot be analyzed in the same way. Wh-topic in (37a) and wh-ex-situ cleft-focus in (37b) target different positions in the left-periphery and they form A'-dependencies with the gaps that they left in-situ. For instance, (37) a.

b.

哪一位老師,學生們都很喜歡? [TopP [Na-yi-wei laoshi]j, [xuesheng-men duo which-one-Cl teacher student-Pl many hen xihuan ___j]]? very like ‘Which teacherj, every student likes ___j very much?’ 是哪一位老師,學生們都很喜歡? [FocP [TP1 Shi [na-yi-wei laoshi]j, [xuesheng-men be which-one-Cl teacher student-Pl duo hen xihuan ___j]]? many very like ‘Which teacherj is it that every student likes ___j very much?’

Gaps in the above sentences can be easily replaced by resumptive pronouns, as shown in (38). Therefore, resumptive pronouns can be present in topic structures and in cleft-focus structures and they are directly linked to fronted NP (contra Cheung 2014). (38) a.

哪一位老師,學生們都很喜歡他? [TopP [Na-yi-wei laoshi]j, [xuesheng-men duo which-one-Cl teacher student-Pl many hen xihuan taj]]? very like 3MSg ‘Which teacherj, every student likes himj very much?’

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese b.

是哪一位老師,學生們都很喜歡他? [FocP [TP1 Shi [na-yi-wei laoshi]j, [xuesheng-men be which-one-Cl teacher student-Pl duo hen xihuan taj]]? many very like 3MSg ‘Which teacherj is it that every student likes himj very much?’

.. ATB-constructions In this section, I will introduce the so-called Across-The-Board (ATB) constructions in Chinese. Different analyses have been proposed to account for ATBconstructions in English, such as the parasitic gap analysis (Munn 1993, 1999), the Parallel Merge analysis (Citko 2005, 2006), the Ellipsis analysis (Ha 2007) and etc. As for Chinese, Zhang (2010) extensively discussed ATB-movement in Chinese and Pan (2011b) takes a different view based on a topicalization approach. Pan (2011b) shows that ATB-movement in Chinese is in fact a kind of topicalization in that it demonstrates all of the properties observed in standard topicalization structures. (39)

哪部電影,你最喜歡看而妳弟弟最不喜歡看? [TopP3 Na-bu dianyingj, [&P [TopP2 t2j' [TP2 ni which-Cl xihuan kan like see zui bu most Neg

movie t2j]]

[&' er but xihuan kan like see

zui

2Sg most [TopP1 t1j' [TP1 ni didi 2Sg brother t1j ]]]]?

‘For which movie x, is it true that you like x most and your brother dislikes x most?’ The shared object NP na-bu dianying ‘which movie’ firstly moves to the embedded TopP positions (i.e. TopP1 and TopP2) respectively in each of the conjuncts of the conjunction projection &P by leaving two traces t1j and t2j. Then, the NP continues to move up and lands at the matrix TopP3 and two intermediate traces, t1j' and t2j' are left in the intermediate sites. Pan assumes that there is an intersective operator generated at the matrix topic position, which ensures that the ATB-extracted NP is a common variable of TP1 and TP2. The answer to this kind of ATB-question must be the common value of the each conjunct. A general schema of ATB-movement in Chinese that Pan (2011b) gives is presented below.

Syntactic constructions involving A'-dependencies in Chinese  

(40)

CP Spec C°

C' TopP3 (∩)

Spec

Top' Top°

TopP1 Spec

Top°

&P

&'

&° Top'

Spec TP1

TopP2 Top' Top°

… α 1…

TP2

… α2…

Pan (2011b) does not talk about ATB movement in ex-situ cleft focus cases and in fact the same analysis also applies to ATB-constructions involving ex-situ cleft-foci, as shown in (41). (41)

是哪部電影,你最喜歡看而妳弟弟最不喜歡看? [FocP3 [TP Shi [NP na-bu dianyingj, [&P [FocP2 t2j' [TP2 ni be

which-Cl movie 2 xihuan kan t j]] [&' er [FocP1 t1j' [TP1 ni like see but 2Sg bu xihuan kan t1j ]]]]? Neg like see

zui

2Sg most didi zui brother most

‘For which movie x, is it true that you like x most and your brother dislikes x most?’

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese In this case, the shared object NP na-bu dianying ‘which movie’ firstly moves to the embedded FocP positions in each of the conjuncts then to the matrix FocP3 to join the TP headed by the copula shi ‘be’. Following Pan (2011b), I assume that the intersective operator is generated at the matrix FocP3 and ensures that the ATB-extracted NP is a common variable of TP1 and TP2. The general schema for the derivation of an ATB-extracted ex-situ cleft-focus is given below. (42)

CP Spec



C'

TP3 Shi ‘be’

FocP3 (∩) Foc' Foc°

FocP1 Spec Foc°

&P

&'

&° Foc'

Spec

TP1

FocP2 Foc' Foc°

TP2

… α2…

… α1…

In English and in French, resumptive pronouns are generally banned from ATB-movement cases. The French sentence in (43b) is the equivalent of (43a) in English. (43) a. Which moviej do you like (*itj) but your brother doesn’t like (*itj) ? b. Quel filmj est-ce que tu (*lj’)aimes bien mais ton frère ne (*lj’)aime pas ?

Syntactic constructions involving A'-dependencies in Chinese  

By contrast, resumptive pronouns are permitted in ATB-constructions in Chinese, as illustrated in (44). The two pronouns in the extracted variable sites agree with the antecedent wh-nominal in terms of ϕ-features. (44a) is an ATBtopicalization case and (44b) an ATB-ex-situ cleft-focus structure. (44) a.

ATB-topicalization 這本書 j,[[我只(把它 j)讀過一遍]而[我弟弟卻至少(把它 j)讀過 三遍]]。 Zhe-ben shuj, [[wo zhi (ba ta3j) du-guo this-Cl book 1Sg only BA 3OSg read-Exp yi-bian er [wo didi que zhishao one-time but my brother however at.least (ba ta3j) du-guo san-bian]]. BA 3OSg read-Exp three-times ‘As for this bookj, I only read itj once but my younger brother read itj at least three times.’

b.

ATB-ex-situ clefted wh-question 是哪位體育老師 j,[[所有的男生都喜歡(他 j)]而[所有的女生都不喜 歡(他 j)]]? Shi na-wei tiyu laoshij, [[suoyou-de nansheng be dou all dou all

which-Cl sport teacher all xihuan (ta1j)] er [suoyou-de like 3MSg but all bu xihuan (ta1j)]]? Neg like 3MSg

boys nüsheng girls

‘Which sport teacherj is it that all of the boys like himj very much and none of the girls likes himj at all?’ Based on the reasoning of this section, a resumptive ATB-topicalization and a resumptive ATB-ex-situ focus structure are given below. (45) a. ATB-topicalization [TopP Topicj, [&P [TP1 …. RPj … ] & [TP2 …. RPj … ]]]. b. ATB-ex-situ cleft-focus structure [FocP [TP Shi ‘be’] + Focusj, [&P [TP1 …. RPj … ] & [TP2 …. RPj … ]]].

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese .. Summary In this section, I have introduced several main types of A'-dependencies in Chinese: wh-dependency, relative clauses, topicalization and ex-situ cleft-focus structures. Topicalization includes ordinary topicalization, wh-topicalization, ATB-topicalization and wh-ATB-topicalization; ex-situ cleft-focus structures include ordinary ex-situ cleft-focus structures, wh-ex-situ cleft-focus questions, ATB-ex-situ cleft-focus structures and ATB-ex-situ clefted wh-questions. All of these types of A'-dependencies except for wh-in-situ questions permit the use of resumptive pronouns. This is the main reason why I only focused on relatives, left-dislocation structures and ex-situ cleft-focus structures in Chapters 2, 3 and 4. Tab. 2: A'-dependencies in Chinese A'-dependencies

RP

Wh-question

no

Relatives

yes

Topicalization

Ex-situ cleftfocus structure

Ordinary topicalization

yes

Wh-topicalization

yes

ATB-topicalization

yes

ATB-topicalized wh-question

yes

Ordinary ex-situ cleft-focus structure

yes

Wh-ex-situ cleft focus structure

yes

ATB-ex-situ cleft-focus structure

yes

ATB-ex-situ clefted wh-question

yes

The general conclusions that I draw and highlight are the following: (i)

Topicalization and ex-situ cleft-focus structures are two different constructions and they must be analyzed differently (cf. also Pan 2014);

(ii)

Wh-fronting is a general term that covers two different cases: whtopicalization and wh-ex-situ cleft-focus questions. These two constructions display different syntactic and semantic properties and are subject to different constraints and therefore, they cannot be analyzed in the same way (cf. also Pan 2014).

General distribution of the resumptive pronouns in Chinese  

(iii)

The ATB-construction also corresponds to two sub-cases: ATBtopicalization (cf. Pan 2011b) and ATB-cleft focus structures.

Topicalization and ex-situ cleft-focus structures, being completely two different constructions, cannot be treated uniformly. The results of the syntactic tests in Chapter 2 and that of the semantic interpretative tests in Chapter 3 reveal that they display similar properties when resumptive pronouns are involved. Resumptive pronouns have exactly the same distribution and interpretations in both structures.

. General distribution of the resumptive pronouns in Chinese In Chinese, resumptive pronouns must always agree with the antecedent NP in terms of ϕ-features, as illustrated in (46). The pronoun ta2-men (2FPl) agrees with the antecedent Xiaoqian and Liyan (i.e. two female individuals). (46)

[一勤和李燕]j,我昨天在學校見到過她們 j 倆。 [Yiqin he Liyan]j, wo zuotian zai

xuexiao

Yiqin and Liyan 1Sg yesterday at school jiandao-guo ta2-men-lia/___ . meet-Exp 3FPl-both ‘As for Yiqin and Liyanj, I met both of themj at school yesterday.’ In this section, I will examine the general distribution of the resumptive pronouns in different contexts and I will make a distinction between two cases: those in which the presence of the resumptive pronouns are obligatory and those in which their presence is optional or even prohibited. .. Cases in which the presence of the resumptive pronouns is obligatory In many languages, when the object NP of a preposition undergoes fronting, the preposition cannot be fronted together with the NP nor can it stay alone in its base position. These languages do not permit pied-piping or preposition stranding. In this case, the presence of a resumptive pronoun is required in the gap position in a relative clause or in a wh-question when the prepositional object NP is relativized or questioned. For instance, Rouveret (1994) shows that in Welsh, when an object of a preposition is relativized, the preposition cannot be

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese stranded in its base position. In order to avoid the potential violation of this ban on preposition stranding, a resumptive pronoun is required in the object position. In (47a), a morphological agreement o that functions as a resumptive is realized on the preposition and o agrees with the verb in terms of ϕ-features. In (47c), a pronoun with a full morphological form, ef, is inserted in the object position and it also has a resumptive function. The ungrammaticality of (47b) and (47d) shows that a gap is not allowed in these positions. (47) a.

y dyn y soniais the man that I-talked ‘the man I talked about’

amdano about-[agr]

b.

*y dyn y soniais the man that I-talked (‘the man I talked about’)

amdan ____ about

c.

y dyn y siaradasoch the man that spoke ‘the man that you spoke with’

chwi you

ag with

ef [him]

d.

*y dyn y siaradasoch the man that spoke (‘the man that you spoke with’)

chwi you

ag with

_____

Welsh, Rouveret (1994, 2011)

The following example of Hausa from Tuller (1986) shows that the preposition da ‘with’ cannot undergo fronting with the nominal wh-word waa ‘who’ to [Spec, CP] nor can it be stranded in-situ because Hausa is also a language that does not permit any preposition stranding. The only strategy to save the sentence from the violation of these constraints is to insert a resumptive pronoun shii ‘him’ in the prepositional object position. (48)

Waa ka nyi maganaa who 2SgM do talk ‘Who did you talk with?’

da with

*(shii) ? [him] Hausa, Tuller (1986)

Rouveret (1994, 2011) also shows that in languages like Welsh and Hausa, a genitive construction requires the use of resumptive pronouns in relativization.

General distribution of the resumptive pronouns in Chinese  

For example, in (49), the possessor yr is extracted and the nominal head fam ‘mother’ cannot be extracted together with the possessor; the resumptive pronoun ei ‘his’ is inserted to save the structure. Similarly, in the example of Hausa (cf. 50), the insertion of -nsa ‘of-his’ serves the same function of saving device. (49) a.

b.

(50) a.

b.

y dyn yr oedd ei fam the man that was his mother ‘the man whose mother was at home’

gartref at home

* y dyn yr oedd ____ fam gartref the man that was mother at home (‘the man whose mother was at home’) Welsh, Rouveret (1994, 2011) Waa ka karanta who 2SgM read ‘Whose book did you read?’

littaafinsa book-of-[his]

* Waa ka karanta littaafi__ who 2SgM read book (‘Whose book did you read?’)

Hausa, Tuller (1986)

From this point of view, Chinese behaves exactly like Welsh and Hausa in that none of these three languages permits preposition stranding or preposition pied-ping. In case of the prepositional object NP-fronting, the insertion of a resumptive pronoun is needed to prevent the sentence from violating these constraints. I will examine the distribution of the resumptive pronouns in the position of prepositional object in relatives (cf. 51) and in LD-structures (cf. 52) respectively. (51) a.

Relativization [我對 *(他 j) 很不錯]的那位朋友 j [wo dui *(ta1j) hen bucuo] 1Sg to 3MSg very not.bad na-wei

pengyouj

that-Cl friend ‘the friend to whom I am very kind’

de C

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese b.

[由 *(她 j) 代表大家發言]的那位同事 j [you *(ta2j) daibiao dajia by 3FSg represent everyone na-wei

fa give

yan] talk

de C

tongshij

that-Cl colleague (Lit.) ‘the colleague that the talk by *(her) represented everyone’ = ‘the colleague who gave a talk representing everyone’ c.

[他們在 *(那兒 j) 渡過蜜月]的城堡 j [ta-men zai *(narj) du-guo 3-Pl at there spend-Exp de

miyue] honeymoon

chengbaoj

C castle ‘the castle where they had their honeymoon’ d.

[當年以 *(那兒 j)作為公司總部]的大樓 j [dangnian yi *(narj) zuowei those.years with there used.as zongbu]

de

gongsi company

dalouj

headquarter C building ‘the building that was used as the headquarter of the company during those days’ Resumptive pronouns function as prepositional objects in the above sentences and their presence is obligatory. (52)

Left-Dislocation a. 那位朋友 j, 我對 *(他 j) 很不錯。 Na-wei pengyouj, wo dui that-Cl friend 1Sg to

*(ta1j) 3MSg

hen very

bucuo. not.bad

‘That friendj, I am very kind to himj.’ b. 那位同事 j, 這次由 *(她 j) 代表大家發言。 Na-wei tongshij, zhe-ci you *(ta2j) that-Cl colleague this-time by 3FSg dajia

fa

daibiao represent

yan.

everyone give talk ‘That colleaguej, shej will represent everyone to give a talk this time.’

General distribution of the resumptive pronouns in Chinese  

c. 樓下那家餐廳,我們曾經在 *(那兒)吃過一次飯。 Lou-xia na-jia cantingj, women cengjing downstairs that-Cl restaurant 1Pl ever zai

*(narj)

chi-guo

yi-ci

fan.

at there eat-Exp one-time dinner ‘The restaurant downstairsj, we had our dinner therej once.’ We can see from these examples that prepositions like dui ‘to sb.’, you ‘by + Agent’, zai ‘at + Place’ and yi ‘using…as’ must take a noun as object and they cannot be stranded. If their nominal objects are relativized or left-dislocated as shown in (51–52), resumptive pronouns must be inserted into the relativized or left-dislocated site. Otherwise, the relevant sentences will be ungrammatical. Again, similar to Welsh and Hausa, Chinese permits neither prepositional stranding nor prepositional pied-piping. In all of the above Chinese examples, the fronting of the relevant prepositions together with the relativized or leftdislocated NPs is not at all an option and therefore, the only way to save the relevant sentences is to insert a resumptive pronoun in the original prepositional object position. Tang (1979) states clearly that when the object of a verb followed by a complement or by a prepositional object undergoes fronting, a pronominal copy of this object must be inserted in its base-position (cf. 53-54). (53) a.

我們請張先生幫你設計。 Women qing Zhang

Xiansheng

bang

1Pl invite Zhang Mr. help ‘We invited Mr. Zhang to help you to design.’ b.

張先生我們請 *(他) 幫你設計。 Zhang Xiansheng women

qing

*(ta1)

ni

sheji.

2Sg

design

bang

Zhang Mr. 1Pl invite 3MSg help ni sheji. 2Sg design ‘Mr. Zhang, we invited *(him) to help you to design.’ (54) a.

我們跟那個人毫無關係。 Women gen na-ge

ren

haowuguanxi.

1Pl with that-Cl person have.no.relation ‘We have nothing to do with that person.’

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese b.

那個人我們跟 *(他)毫無關係。 Na-ge ren women gen

*(ta1)

haowuguanxi.

that-Cl person 1Pl with 3MSg have.no.relation ‘That person, we have nothing to do with *(him).’ Tang (1979:89-90)36 In Chinese, a direct object (DO) can undergo movement to a preverbal and postsubject position and usually, this DO can be marked by ba, as shown in (55). (55) a.

我看完了那本小說。 Wo kan-wan-le

na-ben

1Sg read-finish-Perf that-Cl ‘I have finished reading that novel.’ b.

我把那本小說看完了。 Wo ba na-ben xiaoshuo

xiaoshuo. novel

kan-wan-le.

1Sg BA that-Cl novel read-finish-Perf ‘I have finished reading that novel.’ The nature and the use of ba have been extensively discussed in different frameworks. The use of ba is subject to many specific syntactic and semantic constraints. For instance, Li and Thompson (1981) claim that when the verb has a disposal nature, its DO can be fronted and marked by ba and that the bamarked DO has a definite reading. In generative framework, despite different analyses and approaches, recent work, such as Whitman & Paul (2005), Li (2006), Huang, Li & Li (2009), still agree that ba must head an independent functional projection. However, differences still exist among these analyses and, more importantly, the following questions are still unclear: what is the position occupied by the NP immediately following ba? What is the position of the projection baP with regard to vP and VP in Chinese? Of course, I will not go into the details of these questions that are not concerned directly in this study. However, one important characteristic of ba is that, like all of the other prepositions, it cannot be stranded in-situ if its object is fronted in the case of relativization or left-dislocation; on the other hand, pied-piping is not an option either. In this case, inserting a resumptive pronoun in the DO position is obligatory to

 36 The original examples are only in Chinese.

General distribution of the resumptive pronouns in Chinese  

make the sentence grammatical. Here are several examples with the ba-marked resumptive pronouns in LD-structures and in relatives. (56)

你的女兒 j, 我把 *(她 j)給你帶過來了。 Ni-de nü’erj, wo ba *(ta2j)

gei

your daughter 1Sg BA 3FSg to dai-guo-lai-le. bring-Dir-Perf ‘As for your daughter j, I bring her j here to you.’ (57)

[我弟弟把 *(它 j)弄壞了]的那個玩具 j [Wo didi ba *(ta3j) my brother BA 3OSg na-ge wanjuj that-Cl toy ‘the toy that my brother broke’

(58)

ni 2Sg

nong-huai-le

de]

break-Perf

C

[我們二十年前把 *(它 j)轉手給老李]的那個古董花瓶 j [Women ershi nian qian ba *(ta3j) 1Pl zhuanshou resell gudong antique

twenty years gei Laoli] to Laoli huapingj vase

ago de C

BA 3OSg na-ge that-Cl

‘the antique vase that we resold to Laoli twenty years ago’ The passive marker bei (cf. 59) and the indirect object (IO) marker gei ‘to’ (cf. 60) behave in exactly the same way as ba. In (60), the resumptive pronoun ta1-men ‘them’ referring to the homeless people was used as the indirect object of the preposition gei ‘to’ and its presence is obligatory here. Similarly, the presence of ta3 ‘it’ in (59) is also obligatory. (59)

那隻黑貓 j,小倩曾經被 *(牠 j)咬過。 Na-zhi hei maoj, Xiaoqian

cengjing

that-Cl black cat Xiaoqian ever *(ta3j) yao-guo. 3OSg bite-Exp ‘That black catj, Xiaoqian was bitten by itj.’

bei Passive

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese (60)

那些來自歐洲各國的流浪漢 j,我們在聖厄斯塔什教堂前給*(他們 j)發熱 湯和食物。 Na-xie lai-zi Ouzhou ge guo de those come.from Europe each countries C liulang-hanj, women zai Sheng Esitashi jiaotang homeless.people 1Pl in Saint Eustache church qian gei *(tamenj) fa re-tang he shiwu. front to 3pl distribute hot.soup and food ‘Those homeless peoplej from different European countries, we distributed hot soup and food to themj in front the Saint Eustache church.’

Although, ba, bei and gei are not analyzed as genuin prepositions, when they take resumptive pronouns as direct objects, they behave exactly in the same way as any other preposition. .. Cases in which the presence of the resumptive pronouns is optional or prohibited In island free contexts, action verbs such as zhua ‘arrest’ in (61a) and gan…chu ‘chase…out of’ in (61b) can optionally take a resumptive pronoun as DO and stative verbs or psycho verbs, such as xihuan ‘like’, taoyan ‘dislike’ in (62a), xinshang ‘appreciate’ in (62b) and chongbai ‘admire’ in (62c) generally resist resumptive pronouns in DO position, instead, they prefer null objects. (61) a.

[警察想抓(他 j)的] 那個壞人 j 最後跑掉了。 [Jingcha xiang zhua (ta3j) de]

na-ge

police want arrest 3MSg C that-Cl huai-renj zuihou pao-diao-le. bad.guy finally run-away-Perf ‘The bad guy whom the police wanted to arrest ran away at last.’ b.

警衛想趕(他 j)出醫院的那個小混混 j 不见了。 [Jingwei xiang gan (ta3j) chu yiyuan guard want chase na-ge xiao-hunhun j that-Cl little.hooligan ‘The hooligan that the guard disappeared.’

de]

3MSg out hospital C bujian-le. disappear-Perf wanted to chase out of the hospital

General distribution of the resumptive pronouns in Chinese  

(62) a. 我認識[小倩很喜歡/討厭(*她 j)的]那個女生 j。 Wo renshi [Xiaoqian hen xihuan/ taoyan 1Sg know Xiaoqian very like (*ta2 j) de] na-ge nüsheng j. 3FSg C that-Cl girl ‘I know the girl that Xiaoqan likes/doesn’t like.’

dislike

b. [同學們都很欣賞(*他 j)] 的那位作家 j [tongxue-men dou hen xinshang classmate-Pl all very appreciate (*ta1 j) de] na-wei zuojiaj 3MSg C that-Cl writer ‘the writer that all of the classmates appreciate very much’ c. [小孩子們都很崇拜(*他 j)] 的那位好萊塢影星 j [xiaohaizi-men dou hen chongbai (*ta1 j) child-Pl all very admire 3MSg de] na-wei Haolaiwu ying-xingj C that-Cl Hollywood movie.star ‘the Hollywood star that all of the children admire’ Normally, when an experiencer subject of predicates like bing ‘be sick’, si ‘be dead’ and zhu ‘live’ is relativized, the original subject position cannot host any resumptive pronoun. For example, na-wei lao-taitai ‘that old lady’ in (63) is relativized and the insertion of the resumptive pronoun ta2 ‘she’ is illicit. (63)

[ (*她 j) [(*ta2j)

病了/ 死了/ 住在倫敦] 的那位老太太 j bing-le / si-le/ zhu zai Lundun]

3FSg sick-Perf die-Perf live in London de na-wei lao-taitaij C that-Cl old.lady ‘the old lady who was sick/ was dead/ lived in London’ By contrast, when the experiencer subject is left-dislocated, the presence of a resumptive pronoun in the subject dislocated site is optional. In (64), the topic phrase na-wei laotaitai ‘that old lady’ is generated in or moved to the specifier position of TopP, the resumptive pronoun ta2 ‘she’ can be inserted in the specifier position of TP.

  Appendix: General distribution of the resumptive pronouns in Mandarin Chinese (64)

那位老太太 j,(她 j) 病了/死了/ 住在倫敦。 Na-wei lao-taitaij, [(ta2j) bing-le /

si-le/

that-Cl old.lady 3FSg sick-Perf die-Perf zhu zai Lundun]. live in London ‘As for that old lady, she was sick/ was dead/ lived in London.’ If the relativized subject is not an experiencer, then the insertion of a resumptive pronoun in the subject position is permitted. For instance, in (65), the subject na-ge haizi ‘that kid’ is the agent of the verbal construction rang…yanmian saodi ‘make someone lose face/ humiliate someone’, and when the subject is relativized, it is possible to insert the pronoun ta1 ‘he’ as a resumptive in the original subject position. (65)

[(他 j)讓父母在親朋好友面前顏面掃地]的那個孩子 j [(ta1j) rang fumu zai qinpenghaoyou

mianqian

3MSg make parents in relatives.friends front yanmian saodi] de na-ge haizij face sweep.floor C that-Cl kid ‘the kid who humiliated his parents right in front of all of the family members and friends’ In Chinese, if a resumptive pronoun functions as a possessor itself, its presence is optional, as illustrated in (66). (66) a.

[(他 j 的)失業金被停發了]的那個工人 j [(taj-de) shiyejin

bei

ting

3MSg-Gen unemployement.pension Passive stop fa le] de na-ge gongrenj distribute SFP C that-Cl worker ‘the workerj whose pension for hisj unemployment was cancelled’ b.

[(他們 j 的)爸爸都在高中教書]的那些孩子們 j [(ta3-men-de) baba dou zai gaozhong 3MPl-Gen fathers all at high.school jiao shu] de na-xie haizi-men teach book C that-Cl kid-Pl ‘the kids whose fathers all teach in high schools’

General distribution of the resumptive pronouns in Chinese  

c.

唯一一種[我好像都一直都不能正確調整(它 j 的)控油器的] 車 j 是 神龍公司生產的。 Weiyi yi-zhong [wo haoxiang yizhi dou bu only neng can chej car

one-kind 1Sg seem always all Neg zhengque tiaozheng (ta3j-de) kongyouqi de] correctly adjust 3OSg-Gen carburetor C shi Shenlong gongsi shengchan de. be Shenlong company produce DE

‘The only type of car that I can never adjust its carburetor correctly is that produced by Shenlong company.’ Generally, the presence of a genitive pronoun is not always obligatory in relative clauses if their complement nouns are not relativized.

References Abe, Jun. 2015. The EPP and subject extraction. Lingua 159: 1–17. Adger, David and Gillian Ramchand. 2001. Phases and interpretability. Proceedings of WCCFL 20: 1–14. Somerville, Mass.: Cascadilla Press. Adger, David and Gillian Ramchand. 2005. Merge and Move: Wh-dependencies revisited. Linguistic Inquiry 36: 161–193. Aoun, Joseph and Elabbas Benmamoun. 1998. Minimality, Reconstruction and PF-movement. Linguistic Inquiry 29: 569–597. Aoun, Joseph, and Yen-hui Audrey Li. 1993. Wh-Elements in Situ: Syntax or LF? Linguistic Inquiry 24.2:199–238. Aoun, Joseph, and Yen-hui Audrey Li. 2003. Essays on the Representational and Derivational Nature of Grammar. The Diversity of Wh-Constructions. Cambridge, Mass.: MIT Press. Aoun, Joseph and Lina Choueiri. 2000. Epithets. Natural Language and Linguistic Theory 18: 1–39. Aoun, Joseph Lina Choueiri and Norbert Hornstein. 2001. Resumption, Movement and Derivational Economy. Linguistic Inquiry 32: 371–403. Asudeh, Ash. 2011. Toward a unified theory of resumption. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. by A. Rouveret), 121–188. John Benjamins Publishing Company. Bianchi, Valentina. 2004. Resumptive relatives and LF chains. Oxford: Oxford University Press. Bianchi, Valentina. 2011. Some notes on the ‘specificity effects’ of optional resumptive pronouns. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. by A. Rouveret), 319-342. John Benjamins Publishing Company. Boeckx, Cedric. 2003. Islands and Chains: Resumption as Stranding. Amsterdam: John Benjamins. Borer, Hagit. 1984. Restrictive Relatives in Modern Hebrew. Natural Language and Linguistic Theory 2: 219–260. Boskovic, Zeljko. 2007. On the Locality and Motivation of Move and Agree: an even more minimal theory. Linguistic Inquiry 38: 589–644. Cardinaletti, Anna, and Michal Starke. 1999. The Typology of Structural Deficiency. Clitics in the Languages of Europe (ed. by H. van Riemsdijk), 145–233. Berlin: Mouton de Gruyter. Cheng, Lisa L.S. 1991. On the Typology of WH-Questions. PhD dissertation, MIT. Cheng, Lisa L.S. 2008. Deconstructing the shi…de construction, The Linguistic Review 25: 413– 451. Cheung, C.-H. Candice. 2008. Wh-fronting in Chinese. PhD dissertation, USC. Cheung, C.-H. Candice. 2014. Wh-fronting and the left periphery in Mandarin. Journal of East Asian Linguistics 23 (4): 393–431. Chomsky, Noam. 1982. Some concepts and consequences of the theory of Government and Binding. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2000. Minimalist Inquiries, the Framework. In Step by step, Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David Michaels, & Juan Uriagereka, 89–156. Cambridge, Mass.: MIT Press.

  References Chomsky, Noam. 2001. Derivation by Phase. In Ken Hale. A Life in Language, ed. by Michael Kenstowicz, 1–52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 2004. Beyond Explanatory Adequacy. Structures and Beyond. The Cartography of Syntactic Structure (ed. by Adriana Belletti), Vol 3, 104–131. Oxford: OUP. Chomsky, Noam. 2008. On Phases. In Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud (ed. by Robert Freidin, Carlos P. Otero and Maria Luisa Zubizarreta), 133–166. Cambridge, MA: The MIT Press. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford, New York: Oxford University Press. Citko, Barbara. 2005. On the Nature of Merge: External Merge, Internal Merge, and Parallel Merge,’ Linguistic Inquiry 36: 475–497. Citko, Barbara. 2006. On the Interaction Between ATB Wh-Movement and Left Branch Extraction. Syntax 9 (3): 225-247. Déchaine, Rose-Marie and Martina Wiltschko. 2002. Decomposing pronouns. Linguistic Inquiry 33: 409–422. Demirdache, Hamida. 1991. Resumptive Chains in Restrictive Relatives, Appositives, and Dislocation Structures. PhD dissertation, MIT. Demirdache, Hamida and Orin Percus. 2011. Resumptives, movement and interpretation. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. by A. Rouveret), 367–394. John Benjamins Publishing Company. Doron, Edit. 1982. On the Syntax and Semantics of Resumptive Pronouns. Texas Linguistic Forum 19: 1–48. Elbourne, Paul. 2001. E-type anaphora as NP-deletion. Natural Language Semantics 9: 241– 288. Elbourne, Paul. 2005. Situations and individuals. Cambridge, Mass: MIT Press. Engdahl, Elisabet. 1980. The Syntax and Semantics of Questions in Swedish. PhD dissertation, University of Massachusetts, Amherst. Engdahl, Elisabet. 1985. Parasitic gaps, resumptive pronouns and subject extractions. Linguistics, 23: 3–44. Evans, Gareth. 1980. Pronouns. Linguistic Inquiry 11: 337–362. Freidin, Robert and Jean-Roger Vergnaud. 2001. Exquisite connections: some remarks on the evolution of linguistic theory. Lingua 111: 639–666. Gu, Gang. 2001. A study of resumptive pronouns. PhD dissertation, The Chinese University of Hongkong. Guilliot, Nicolas. 2006. La Reconstruction à l'Interface entre Syntaxe et Sémantique. PhD dissertation, Université de Nantes. Guilliot, Nicolas. 2011. Reconstructing functional relatives. Empirical Issues in Syntax and Semantics 8 (ed. by Bonami, Olivier & P. Cabredo Hofherr), 97–121. Guilliot, Nicolas and Nouman Malkawi. 2006. When Resumption determines Reconstruction. Proceedings of WCCFL 25 (ed. by Donald Baumer, David Montero, and Michael Scanlon), 168–176. Somerville, MA: Cascadilla Proceedings Project. Guilliot, Nicolas and Nouman Malkawi. 2007. Reconstruction without Movement. Coreference, Modality and Focus, (ed. by Luis Eguren & Olga Fernandez Soriano), 113–131. Amsterdam: John Benjamins. Guilliot, Nicolas and Nouman Malkawi. 2009. When movement fails to reconstruct. In Merging Features (ed. by J. M. Brucart, A. Gavarró and J. Solà), 159–174. Oxford University Press.

References   Guilliot, Nicolas and Nouman Malkawi. 2011. Weak versus strong resumption: Covarying differently. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. by A. Rouveret), 367–394. John Benjamins Publishing Company. Ha, Seungwan. 2007. Contrastive Focus: Licensor for Right Node Raising. Proceedings of the 37th North East Linguistic Society (ed. by E. Elfner and M. Walkow), 247–260. GSLA. Amherst, Umass. Harlow, Stephen. 1981. Government and relativization in Celtic. In Binding and filtering (ed. by Franck Heny), 213–254. London: Croom Helm. Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases. PhD Dissertation. Amherst, MA: University of Massachusetts. Heim, Irene. 1990. E-type Pronouns and Donkey-Anaphora. Linguistics and Philosophy 13: 137–177. Heim, Irene and Angelika Kratzer. 1998. Semantics in Generative Grammar. Malden: Blackwell. Huang, James C.-T. 1982. Logical relations in Chinese and the theory of grammar. PhD dissertation, MIT. Huang, James C.-T., Yen-Huei A. Li and Yafei Li. 2009. The Syntax of Chinese, Cambridge University Press. Husley, Sarah and Uli Sauerland. 2006. Sorting out relative clauses. Natural Language Semantics 14: 111–137. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Koopman, Hilda. 1983. Control from COMP and Comparative Syntax. The Linguistic Review 2: 365–391. Lasnik, Howard and Tim Stowell. 1991. Weakest crossover. Linguistic inquiry 22.4: 687–720. Lebeaux, David. 1992. Relative Clauses, Licensing, and the Nature of the Derivation. Perspectives on Phrase Structure: Heads and Licensing (ed. by Susan Rothstein & Margaret Speas), 209–239. Syntax and Semantics 25, San Diego, Calif.: Academic Press. Li, Yen-Huei Audrey 2006. Chinese ba. In The Blackwell companion to syntax (ed. by Martin Everaert and Henk van Riemsdijk), vol. 1: 374–468. Malden, Mass.: Blackwell. Li, Charles and Sandra Thompson. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley and Los Angeles: University of California Press. Malkawi, Nouman. 2009. Sur la syntaxe de quelues expressions anaphoriques: épithètes et pronoms résomptifs. PhD dissertation. University of Nantes. McCloskey, James. 1979. Transformational Syntax and Model Theoretic Semantics. A CaseStudy in Modern Irish. Dordrecht: Reidel. McCloskey, James. 1990. Resumptive Pronouns, A'-binding and Levels of Representation in Irish. The Syntax of the Modern Celtic Languages (ed. by Randall Hendrick), 199–248. Syntax and Semantics 23, New York & San Diego: Academic Press. McCloskey, James. 2001. The Morphosyntax of Wh-extraction in Irish. Journal of Linguistics 37: 67–100. McCloskey, James. 2002. Resumption, Successive Cyclicity, and the Locality of Operations. Derivation and Explanation in the Minimalist Program, ed. par Samuel David Epstein and Daniel Seely, 184–226. Oxford: Blackwell. McCloskey, James. 2005. Resumption. The Blackwell Companion to Syntax (ed. by Martin Everaert and Henk van Riemsdijk), volume 3, chapter 55: 94–117. Oxford: Blackwell. Munn, Alan. 1993. Topics in the syntax and semantics of coordinate structures. PhD dissertation, University of Maryland, College Park.

  References Munn, Alan. 1999. On the identity requirement of ATB movement. Natural Language Semantics 7: 421–425. Nash, Lea and Alain Rouveret. 2002. Cliticization as Unselective Attract. Catalan Journal of Linguistics 1: 157–199. Ning, Chunyan. 1993. The Overt Syntax of Relativization and Topicalization in Chinese. PhD dissertation. University of California. Ning, Li-Hsin. 2007. The grammar and processing of resumptive pronouns in Chinese relative clauses. Master’s thesis, National Chengchi University. Noguchi, Tohru. 1997. Two types of pronouns and variable binding. Language 73: 770–797. Pan, Victor Junnan. 2011a. Interrogatives et quantification: une approche générative. Rennes: Presses Universitaires de Rennes. Pan, Victor Junnan. 2011b. ATB-topicalization in Mandarin Chinese : an Intersective Operator Analysis, Linguistic Analysis, Volume 37: 231–272. Pan, Victor Junnan. 2014. Wh-ex-situ in Chinese: Mapping Between Information Structure and Split CP. Linguistic Analysis, Volume 39 : 371–413. Pan, Victor Junnan. 2015a. Mandarin Peripheral Construals at Syntax-Discourse Interface, The Linguistic Review, Volume 32, Issue 4: 819–868. Pan, Victor Junnan. 2015b. Contre le point de vue “macro-variation” sur les résomptions. Studii de lingvistică, Volume 5: 203–220. Pan, Victor Junnan. 2015c. Syntactic and Prosodic Marking of Contrastiveness in Spoken Chinese, in Fernandez-Vest, M. M. J., & Van Valin Jr., R. D. (Eds.). Information structure and spoken language in a cross-linguistics perspective. Berlin: Mouton de Gruyter, p. 191– 210. Pan, Victor Junnan. 2016. Resumptivity and Two Types of A′-dependencies in the Minimalist Program, International Journal of Chinese Linguistics, Volume 3, Issue 1: 46–79. Pan, Victor Junnan. To appear. Minimaliser les dépendances A' : résomptifs vs lacunes, Revue Canadienne Linguistique/Canadian Journal of Linguistics. Pan, Victor Junnan and Waltraud Paul. 2016. Why Chinese SFPs are neither optional nor disjunctors, Lingua, Volume 170, p. 23–34. Paris, Marie-Claude. 1979. Nominalization in Mandarin Chinese. The morpheme de and the ‘shi…de’ constructions. Paris: Université Paris 7, Département de Recherches Linguistiques. Paul, Waltraud. 2002. Sentence-internal topic in Mandarin Chinese : The case of object preposing. Language and Linguistics 3 (4): 695–714. Paul, Waltraud. 2005. Low IP area left periphery in Mandarin Chinese. Recherches linguistiques de Vincennes vol. 33: 111–134. Paul, Waltraud. 2014. Why particles are not particular: Sentence-final particles in Chinese as heads of a split CP. Studia Linguistica 68 (1): 77–115. Paul, Waltraud. 2015. New perspectives on Chinese syntax. Berlin: De Gruyter Mouton. Paul, Waltraud and Victor Junnan Pan. To appear. What you see is what you get: Chinese sentence-final particles as head-final complementisers, in J. Bayer & V. Struckmeier (eds.) Discourse Particles – Formal Approaches to their Syntax and Semantics, [Linguistiche Arbeiten], Berlin : Mouton De Gruyter. Postal, Paul. 1966. On so-called ‘pronouns’ in English. Report of the 17th Annual Roundtable Meeting on Linguistics and Language Studies, 177–206. Washington, D.C.: Georgetown University Press. [Reprinted in David Reibel & Sandford A. Schane (eds.), Modern Studies in English, Englewood Cliffs, N.J.: Prentice-Hall, 1969.]

References   Postal, Paul. 1998. Three investigations of extraction. MIT Press. Qu, Yanfeng. 1994. Object Noun Phrase Dislocation in Mandarin Chinese. PhD dissertation, University of British Columbia. Reinhart, Tanya. 1983. Anaphora and semantic interpretation. University of Chicago Press. Richards, Norvin. 2001. Movement in Language. Oxford: Oxford University Press. Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery, in Elements of Grammar (ed. by Liliane Haegeman), 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2004. Locality and Left Periphery. In Structures and Beyond. The Cartography of Syntactic Structures, vol.3 (ed. by A.Belletti), 223–251. New York: Oxford University Press. Ross, John Robert. 1967. Constraints on Variables in Syntax. PhD dissertation. MIT. Rouveret, Alain. 1987. Syntaxe des dépendances lexicales. Identité et identification dans la théorie syntaxique. Thèse de Doctorat d’Etat, Université de Paris VII, Paris. Rouveret, Alain. 1994. Syntaxe du gallois. Principes généraux et typologie. Paris: CNRS Editions. Rouveret, Alain. 2002. How are Resumptive Pronouns Linked to the Periphery? Linguistic Variation Yearbook 2: 123–184. Rouveret, Alain. 2008. Phasal Agreement and Reconstruction. In Foundational Issues in Linguistic Theory (ed. by Robert Freidin, Carlos P. Otero and Maria Luisa Zubizarreta), 167–195. Cambridge, Mass: The MIT Press. Rouveret, Alain. 2011. Some issuses in the theory of resumption: A perspective on early and recent research. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. A. Rouveret), 1–62. John Benjamins Publishing Company. Rouveret, Alain 2015. Arguments minimalistes : Une présentation du Programme Minimaliste de Noam Chomsky. Lyon : ENS Editions. Rouveret, Alain. to appear. Computational and semantic aspects of resumption, in Interfaces of Grammar, Jianhua Hu & Haihua Pan eds., [Language Faculty and Beyond (LFAB)], John Benjamins. Salzmann, Martin. 2006. Resumptive Prolepsis. A Study in indirect A'-dependencies. PhD dissertation. University of Leiden. Sauerland, Uli. 1998. The Meaning of Chains. PhD dissertation, MIT. Sauerland, Uli. 2004. The interpretation of traces. Natural Language Semantics 12: 63–127. Sells, Peter. 1984. Syntax and Semantics of Resumptive Pronouns. PhD dissertation. University of Massachusetts at Amherst. Sells, Peter 1987. Binding resumptive pronouns. Linguistics and Philosophy 10: 261–298. Sharvit, Yael. 1999. Resumptive Pronouns in Relative Clauses. Natural Language and Linguistic Theory 17: 587–612. Shlonsky, Ur. 1992. Resumptive Pronouns as Last Resort. Linguistic Inquiry 23: 443–468. Shyu, Shu-ing. 1995. The Syntax of Focus and Topic in Mandarin Chinese. PhD dissertation, University of Southern California. Tallerman, Maggie. 1983. Island Constraints in Welsh. York Papers in Linguistics 10: 197–204. Tang, C.-C. Jane. 1988. Wh-topicalization in Chinese. Ms, Cornell University, Ithaca. Tang, Ting-Chi. 1979. Guoyu Yufa Yanjiu Lunji [Studies in Chinese Syntax]. Taipei: Student Book Co., Ltd. Tellier, Christine. 1991. Licensing Theory and French Parasitic Gaps. Dordrecht : Kluwer Academic Publishers. Tsai, Wei-Tien Dylan. 1994. On Economizing the Theory of A'-Dependencies. PhD Dissertation, MIT.

  References Tuller, Laurice. 1986. Bijective Relations in Universal Grammar and the Syntax of Hausa. PhD dissertation, UCLA. Vergnaud, Jean-Roger. 1974. French Relative Clauses. PhD Dissertation, MIT. Whitman, John. and Wautraud Paul. 2005. Reanalysis and conservancy of structure in Chinese. In Grammaticalization and Parametric Variation (ed. by M. Batllori et al.): 82–94. Oxford: Oxford University Press. Willis, David. 2011. The limits of resumption in Welsh wh-dependencies. In Resumptive Pronouns at the Interfaces [Language Faculty and Beyond, 5] (ed. A. Rouveret), 189–222. John Benjamins Publishing Company. Wu, Jian-Xin. 1999. Syntax and semantics of quantification in Chinese, PhD Dissertation, University of Maryland at College Park. Xu, Liejiong. 2004. Manifestation of Informational Focus. Lingua 114 (3): 277–299. Zaenen, Annie, Elisabet Engdahl and Joan M. Maling. 1981. Resumptive pronouns can be syntactically bound. Linguistic Inquiry 12: 679–682. Zhang, Min. 2009. Hànyǔ huàtí huà jiégòu xiànzhì zhōng de línjiē tiáojiàn: Rènzhī chǔlǐ jiǎodù dì lùnzhèng [A revisit of the subjacency constraint in Chinese topicalization from the perspective of cognitive processing]. In Yuyanxue Luncong [Essays on Linguistics] Vol. 39, 523–572. Beijing: Commercial Press. Zhang, Niina Ning. 2002. Island Effects and Episodic Eventualities in Chinese Topicalization. In Linguistics by Heart: in honor of Horst-Dieter Gasde (ed. by D. Hole, P. Law, and N. Zhang). ZAS-Berlin. Zhang, Niina Ning. 2010. Coordination in Syntax. Cambridge Studies in Linguistics Series 123, Cambridge: Cambridge University Press. Zubizarreta, Maria Luisa and Jean-Roger Vergnaud. 2006. Phrasal stress and syntax. In The Blackwell companion to syntax (ed. by Martin Everaert and Henk van Riemsdijk), vol. III, 522–568. Blackwell.

Subject index A'-bound 3, 11, 23, 25, 26, 25, 29, 54, 92, 99, 101, 102, 111, 112, 253, 254, 255, 256 A'-chain 1, 2, 52, 113, 115, 129, 248 A'-dependency – A'-dependencies 2, 6, 11, 16, 18, 20, 24, 25, 27, 26, 28, 30, 37, 77, 78, 84, 86, 88, 95, 99, 102, 103, 104, 107, 108, 110, 111, 112, 113, 115, 116, 118, 120, 124, 125, 126, 130, 131, 133, 136, 129, 130, 131, 135, 165, 169, 170, 196, 198, 203, 205, 206, 222, 218, 228, 234, 235, 241, 243, 245, 246, 248, 249, 250, 252, 253, 254, 255, 256, 258, 259, 262, 263, 264 A'-movement 12, 27, 31, 34, 52, 54, 55, 73, 74, 78, 79, 101, 102, 109, 134, 135, 197, 201, 256, 261 A'-position 2, 11, 19, 30, 31, 39, 54, 245, 256, 259, 263, 276 A-bound 54, 102, 111, 112 abstractor 99, 115, 119 Across-The-Board 2, 257, 281 activation 94, 104, 131, 219, 252 A-dependency 84, 102, 103, 111, 112 Adjunct island 32 A-free 53 agree 15, 17, 18, 19, 20, 23, 24, 25, 27, 30, 80, 81, 82, 83, 84, 85, 86, 87, 89, 90, 92, 93, 94, 95, 97, 98, 99, 100, 101, 102, 103, 104, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 137, 131, 200, 203, 204, 205, 219, 221, 243, 245, 246, 247, 248, 249, 250, 251, 253, 254, 256, 257, 263, 277, 284, 287, 292 agreement 15, 17, 19, 41, 60, 81, 82, 87, 88, 89, 91, 93, 95, 96, 97, 103, 109, 110, 125, 163, 275, 287 anaphoric binding 130, 133, 137, 144, 156, 157, 161, 168, 170, 183, 184, 185, 186,

187, 188, 192, 195, 196, 211, 216, 220, 248, 250, 255 anaphoric dependency 57, 60, 62, 102, 111 antecedent 9, 13, 14, 15, 18, 19, 25, 37, 53, 56, 57, 59, 60, 86, 88, 91, 95, 97, 100, 103, 109, 130, 132, 133, 136, 137, 138, 142, 144, 145, 146, 147, 148, 149, 152, 153, 157, 158, 161, 162, 168, 170, 171, 177, 178, 179, 180, 182, 183, 184, 186, 188, 194, 195, 199, 208, 209, 210, 211, 215, 216, 219, 224, 225, 235, 241, 245, 250, 255, 258, 261, 276, 284, 287 A-position 53, 263 apparent resumption 136, 137 appositive 116, 117, 118 assignment function 87 attributes 81, 82, 109, 128, 129, 133, 135, 246 base-generation 25, 28 base-position 20, 133, 134, 291 Bind 137, 138, 140, 141, 144, 145, 149, 153, 155 binding 13, 14, 16, 25, 30, 89, 90, 92, 99, 112, 132, 135, 142, 146, 160, 164, 173, 174, 183, 185, 186, 187, 191, 211, 213, 244, 245 Binding condition 133 bound anaphor 146, 147, 157, 161, 176 bound variable 54, 55, 57, 78, 95, 99, 101, 102, 108, 112, 115, 118, 124, 125, 128, 132, 133, 134, 135, 137, 140, 146, 148, 151, 157, 161, 172, 178, 196, 224 Breton 42, 43, 142, 143, 149 Celtic languages 42 checking 81, 83, 84, 109, 122, 126, 130, 131, 135, 246 clitic 15, 19, 41, 57, 58, 93, 104, 105, 146, 151, 152, 154, 218, 225, 226, 227, 228, 259 complement clause of noun 39, 40, 47, 261

  Subject index complementizer 18, 56, 86, 88, 89, 90, 91, 99, 100, 103, 112, 114, 115, 134, 131 complex-NP 31, 32, 36, 38, 39, 123, 261 condition A 157 Condition B 133 condition C 25, 53, 54, 55, 130, 133, 137, 141, 142, 143, 144, 157, 161, 162, 164, 165, 169, 188, 189, 191, 192, 193, 194, 196, 197, 213, 214, 220, 245, 248 Condition on Phases 19 contrastive focus 272, 273, 274, 275 copy 20, 83, 84, 85, 103, 129, 140, 141, 142, 150, 156, 157, 159, 163, 165, 217, 224, 291 co-reference – co-referential dependency 102, 103, 111 C-Rel 87, 88, 89, 90, 92, 95, 96, 99, 100, 101, 102, 103, 104, 109, 114, 115, 116, 117, 119, 120, 124, 127, 128, 129, 134, 135, 137, 246, 247 crossover effects 29, 53, 55, 57, 73, 74, 101 C-Top 111, 114, 115, 116, 118, 128, 129, 135, 137, 247 cyclic 18, 24, 30, 31, 120, 134 de dicto 10, 217, 219, 220, 231, 232, 233, 244 de re 10, 217, 220, 231, 232, 233 deactivation 131, 218 default form 131, 151, 166, 197, 198 definite description 129, 150, 159, 163, 165, 217, 224, 230, 244 derivational mechanism 27, 29, 107, 218, 256 derivational mechanisms 17, 23, 130, 131, 157, 166, 167, 168, 197, 205, 218, 219, 221, 243, 246, 249, 253, 256, 257 discourse 30, 108, 112, 132, 200, 202, 204, 224, 248 dislocated sites 25, 26, 205 dislocation 2, 23, 24, 25, 26, 27, 26, 30, 31, 34, 37, 44, 45, 51, 64, 65, 66, 67, 68, 71, 73, 74, 75, 77, 78, 79, 80, 98,

104, 107, 108, 109, 110, 112, 113, 114, 115, 117, 118, 119, 120, 125, 126, 128, 131, 132, 133, 134, 135, 136, 137, 131, 132, 139, 143, 168, 174, 177, 178, 180, 181, 185, 186, 191, 196, 197, 198, 199, 200, 201, 202, 208, 209, 211, 213, 216, 217, 218, 219, 220, 221, 222, 218, 219, 231, 233, 236, 238, 240, 242, 243, 244, 245, 254, 257, 263, 285, 292 distinctiveness 30 distributive reading 134, 139, 140, 145, 146, 147, 148, 157, 159, 160, 161, 164, 168, 170, 171, 172, 173, 174, 175, 176, 178, 179, 181, 207, 208, 219, 224, 234, 235, 236, 237, 238, 241, 242, 245, 251, 254 economy principle 107 ECP 3, 4, 5, 15, 261 English 3, 4, 5, 12, 25, 31, 33, 35, 62, 64, 66, 67, 68, 69, 70, 71, 73, 74, 82, 90, 91, 135, 144, 221, 259, 260, 261, 262, 263, 267, 278, 281, 284 episodic eventualities 48, 50, 51, 266, 267, 268 episodicality 48, 49, 265, 276, 278 epithet, 14, 57, 58, 60, 61, 62, 64, 65, 66, 67, 69, 70, 71, 72, 105, 106, 110, 111, 112, 137, 151, 152 EPP 18, 25, 26, 27, 82, 85, 93, 94, 97, 107, 131, 202, 203, 204, 205, 206, 218, 219, 222, 243, 245, 249, 251, 252, 254, 255, 256 E-type pronoun 138, 150, 224 exhaustivity 278, 279, 280 existential quantified phrase 134, 171 ex-situ cleft-focus 23, 27, 26, 28, 30, 44, 48, 49, 50, 51, 65, 66, 73, 74, 107, 257, 264, 265, 266, 268, 269, 272, 273, 274, 276, 277, 278, 279, 280, 283, 284, 285, 286 extended form 17, 21, 23, 25, 26, 25, 26, 108, 131, 132, 151, 163, 164, 166, 167, 169, 170, 172, 173, 174, 177, 178, 179, 180, 183, 184, 185, 186, 189, 191, 196, 197, 198, 201, 202, 204, 205,

Subject index   206, 212, 213, 214, 216, 217, 219, 220, 218, 219, 228, 231, 240, 241, 242, 243, 244, 245, 248, 249, 250, 251, 252, 253, 254, 255, 256 extraction 15, 16, 27, 31, 40, 41, 48, 49, 50, 51, 76, 110, 163, 245, 265, 266, 267, 269 feature identity 18, 81 Feature Matching 83 formal features 20, 30, 87, 101, 103, 166, 246 French 5, 31, 33, 35, 36, 37, 46, 63, 64, 66, 67, 68, 69, 70, 72, 73, 74, 91, 127, 129, 130, 133, 145, 146, 149, 155, 219, 222, 224, 244, 259, 260, 261, 262, 263, 278, 284 functional reading 155, 156, 217, 221, 222, 224, 225, 226, 227, 228, 231, 234, 242 gap 6, 7, 9, 13, 15, 22, 23, 24, 25, 27, 25, 34, 35, 37, 40, 43, 46, 50, 52, 62, 64, 66, 69, 70, 71, 73, 74, 75, 77, 78, 79, 98, 107, 108, 110, 120, 123, 125, 131, 132, 133, 134, 136, 130, 132, 168, 169, 170, 171, 172, 173, 174, 177, 178, 179, 180, 182, 183, 185, 186, 187, 189, 191, 192, 196, 197, 198, 200, 201, 202, 203, 204, 205, 206, 216, 218, 219, 220, 221, 222, 217, 219, 220, 222, 223, 225, 226, 227, 228, 229, 231, 232, 234, 235, 240, 241, 243, 244, 245, 249, 250, 251, 254, 256, 259, 263, 275, 276, 281, 287 generative grammar 1, 2, 12, 20, 38 Goal 17, 18, 30, 81, 82, 83, 84, 85, 86, 92, 94, 97, 99, 100, 101, 103, 104, 109, 111, 112, 113, 115, 118, 119, 120, 121, 124, 125, 126, 127, 128, 129, 133, 134, 135, 246 Government and Binding 12, 27, 29, 30, 31, 76, 77, 78, 84, 85, 98, 102, 107, 120, 134, 129, 135, 200, 246 grammatical resumptions 138 grammatical resumptive pronouns 3 grammatical use 3, 6, 9, 21, 22, 23

Hausa 7, 8, 288, 289 Hebrew 3, 4, 10, 11, 26, 35, 44, 76, 219, 220, 221, 222, 223, 231, 244, 251, 262 identity condition 136, 157 Inclusiveness 163, 168 indefinite 129, 134, 137, 150, 155, 156, 158, 171, 172, 173, 217, 219, 224, 233, 236, 241 individual reading 10, 11, 157, 217, 219, 220, 221, 222, 223, 224, 229, 230, 231, 242 interface 1, 126, 245 Interface Legibility 86 internal structures 16, 21, 23, 26, 25, 26, 77, 108, 133, 129, 130, 131, 132, 150, 159, 160, 163, 165, 166, 167, 168, 169, 178, 182, 194, 196, 197, 198, 199, 201, 202, 205, 206, 207, 217, 218, 219, 220, 221, 222, 228, 240, 241, 243, 244, 248, 249, 251, 253, 254, 256 interpretability 95, 101 interpretable feature 94, 95, 99, 100, 115, 118, 119, 123, 128 interpretative effects 21, 22, 27, 132, 137, 156, 163, 196, 198, 205, 206, 216, 217, 220, 245, 252 intervention effects 26, 85, 120, 121 intrusive pronoun 21, 25, 44, 45, 51, 108, 110, 132, 136, 130, 131, 138, 168, 175, 178, 181, 186, 197, 202, 205, 211, 217, 220, 221, 219, 231, 234, 240, 243, 244, 251, 257, 262 intrusive use 3, 4, 6, 9, 21, 22, 23, 24, 27, 43, 45, 145, 210, 253, 254, 256, 263 Irish 13, 14, 22, 42, 44, 55, 57, 59, 76, 89, 90, 91, 92, 98, 109, 110, 133, 256, 259, 260, 275 island 3, 4, 5, 6, 12, 13, 14, 15, 16, 18, 19, 21, 24, 25, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 37, 38, 40, 41, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 74, 75, 77, 78, 79, 84, 85, 92, 98, 101, 106, 107, 108, 109, 110, 113, 116, 120, 121, 123, 125, 130, 132, 134, 135, 136, 129, 131,

  Subject index 137, 138, 139, 140, 141, 144, 145, 146, 147, 148, 149, 152, 153, 154, 157, 175, 176, 181, 187, 194, 199, 201, 202, 203, 204, 208, 220, 222, 217, 245, 246, 247, 254, 256, 257, 261, 262, 263, 276, 294 island effects 4, 12, 16, 22, 37, 38, 50, 51, 75, 76, 80, 92, 98, 102, 108, 110, 120, 121, 130, 132, 133, 131, 208, 217, 254, 263 island-free 5, 74, 181, 187 Jordanian Arabic 17, 20, 57, 58, 60, 104, 105, 129, 130, 142, 144, 145, 146, 149, 150, 153, 154, 155, 156, 158, 166, 167, 217, 219, 224, 225, 226, 228, 231, 244, 251 last resort 4, 9, 35, 42, 43, 51, 76, 136, 140, 231, 256 LD-structures 23, 27, 28, 30, 33, 44, 47, 51, 64, 65, 66, 70, 72, 73, 74, 78, 107, 132, 199, 245, 246, 247, 249, 253, 254, 255, 257, 289, 293 Lebanese Arabic 6, 17, 129, 137, 139, 145, 150, 151, 153, 167, 217 left-periphery 30, 107, 245, 249, 259, 273, 276, 277, 280 linking 30, 108, 140, 165 locality constraints 3, 4, 13, 15, 16, 19, 21, 24, 26, 30, 35, 37, 40, 42, 43, 44, 45, 51, 77, 79, 89, 98, 106, 121, 130, 134, 136, 131, 138, 145, 156, 163, 181, 187, 246, 247, 254, 256, 257, 262 Logical Form 13, 15, 27, 30 Long distance 84 macro-variation 4, 6, 9, 21, 136, 256 marked form 151, 166, 197, 198 Match 18, 23, 24, 26, 27, 30, 80, 81, 82, 83, 108, 109, 110, 111, 112, 113, 115, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 135, 137, 131, 181, 187, 200, 201, 203, 204, 205, 219, 221, 243, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 256, 257

Matching 18, 81, 82, 83, 109, 125, 129, 137, 222, 247 matrix clause 19, 271 Merge 25, 80, 81, 84, 85, 131, 136, 202, 203, 281 Minimalist Program 12, 13, 17, 18, 27, 29, 30, 31, 80, 83, 84, 85, 86, 98, 102, 107, 120, 134, 135, 168, 200, 202, 245, 246 Move 17, 18, 27, 30, 80, 81, 84, 85, 86, 92, 96, 97, 98, 106, 107, 108, 109, 110, 113, 130, 132, 133, 137, 132, 138, 139, 140, 141, 149, 153, 154, 200, 201, 204, 245, 248, 249 movement 6, 7, 12, 13, 14, 15, 16, 17, 18, 20, 25, 28, 30, 31, 52, 53, 54, 55, 56, 57, 58, 74, 76, 78, 79, 80, 84, 85, 86, 88, 90, 93, 97, 98, 102, 103, 104, 105, 106, 107, 109, 121, 134, 135, 137, 129, 132, 133, 134, 135, 137, 140, 141, 144, 145, 150, 152, 155, 156, 157, 160, 163, 165, 200, 201, 202, 203, 204, 217, 245, 259, 260, 263, 277, 281, 282, 283, 284, 292 Multiple Transfers 121, 122, 126, 130 Narrow Syntax 18, 19, 30, 86, 98, 125, 132, 133, 202, 245, 247 nominal restriction 157, 172, 175, 178 null objects 294 operator-variable 15, 104, 112, 113, 114, 116, 117, 120, 124, 128, 135, 247 pair-list reading 10, 20, 129, 130, 150, 155, 156, 158, 159, 160, 166, 167, 217, 218, 219, 220, 221, 222, 224, 225, 226, 228, 230, 231, 234, 235, 238, 239, 240, 241, 242, 243, 244, 245, 250, 251, 252, 255 Phasal Agree 86 phase 18, 24, 30, 31, 84, 86, 93, 98, 104, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 131, 132, 134, 247 Phase Impenetrability 120, 121 pied-piping 37, 85, 287, 293 possessor 7, 38, 288, 296

Subject index   preposition stranding 287, 288, 289 prepositional object 6, 8, 257, 287, 288, 289, 291 Probe 17, 18, 27, 30, 81, 82, 83, 84, 85, 86, 89, 92, 93, 94, 97, 99, 100, 103, 104, 109, 111, 112, 113, 115, 118, 119, 121, 124, 125, 126, 127, 128, 129, 133, 134, 135, 132, 246, 255 pseudo-cleft 266, 267 real resumption 136 reconstructed site 150, 178 reconstruction effects 17, 20, 21, 26, 27, 25, 28, 92, 129, 130, 131, 132, 133, 134, 135, 137, 138, 139, 140, 141, 142, 144, 145, 147, 149, 150, 151, 153, 157, 159, 160, 163, 164, 165, 166, 167, 168, 170, 196, 198, 200, 206, 207, 215, 220, 221, 222, 228, 245, 248, 249, 252, 254, 255 reduced form 17, 21, 23, 26, 25, 108, 131, 132, 151, 163, 165, 166, 168, 169, 171, 175, 177, 178, 180, 181, 186, 187, 192, 194, 196, 197, 198, 201, 203, 205, 207, 217, 219, 220, 221, 218, 229, 240, 241, 242, 243, 244, 245, 248, 249, 250, 251, 253, 254, 255, 256 referential pronoun 95, 101 relative clauses 4, 10, 31, 35, 37, 39, 87, 95, 99, 101, 107, 115, 124, 131, 135, 137, 142, 172, 177, 179, 201, 261, 262, 263, 285, 297 relativization 4, 9, 35, 37, 88, 257, 261, 288, 292 restrictive relative clause 15 resumptive dependencies 1, 2, 20, 27, 28, 55, 57, 76, 77, 104, 137, 149, 157, 159, 168, 246, 248, 251 resumptive relatives 45, 59, 61, 62, 68, 69, 80, 101, 106, 109, 112, 137, 160 scope 20, 25, 26, 105, 129, 130, 134, 139, 140, 145, 146, 148, 150, 155, 156, 158, 159, 161, 164, 166, 167, 168, 169, 170, 171, 172, 173, 174, 177, 178, 182, 188, 195, 196, 207, 208, 216, 219, 225, 229, 234, 238, 239, 241, 242,

244, 245, 250, 251, 255, 259, 260, 273, 275 Scottish Gaelic 87, 88, 90, 91, 95, 96, 109, 133 sentential subject 32, 41, 48, 52 shortest link condition 120, 121 specific reading 92, 217, 218, 220, 232, 233, 240, 241, 252 specificity 26, 217, 229, 242, 250 spelled out traces 2, 6, 11, 22, 136, 217 SPELL-OUT 122, 123, 127, 130 strong crossover 14, 54, 55, 57, 66, 68, 69, 70, 71, 72, 75, 78, 79, 102, 111 strong pronouns 17, 150, 151, 153, 154, 165, 167 strong resumptives 42, 217, 218, 244 Subjacency 3, 4, 15, 16, 31, 37, 261 subordinate clauses 34, 261 surface-structure 15, 245 Swedish 6, 12, 22, 136, 219, 230, 231, 251, 256 syntactic mechanism 22, 30, 108, 113, 130, 166, 246, 252 topic 13, 25, 26, 30, 44, 51, 78, 79, 107, 112, 114, 132, 148, 150, 202, 213, 219, 233, 244, 263, 269, 270, 273, 274, 276, 280, 282 topic markers 270 topicalization 25, 48, 49, 50, 54, 55, 67, 114, 117, 133, 134, 135, 201, 257, 263, 265, 266, 276, 277, 278, 279, 280, 281, 284, 285, 286 Transfer 18, 86, 120, 122, 123, 124, 126, 127, 128, 130 true resumption 137, 138, 149 uninterpretable features 17, 81, 82, 83, 84, 85, 86, 94, 97, 101, 104, 111, 113, 115, 119, 120, 126, 128, 129, 130, 134, 135, 246, 247 universal quantified phrase 10, 130, 133, 134, 140, 146, 147, 148, 152, 158, 159, 171, 172, 173, 180, 181, 207 unvalued features 17, 83, 84, 113, 115, 118, 128

  Subject index valuation 81, 82, 83, 88, 104, 122 value 17, 45, 81, 82, 84, 86, 87, 88, 89, 94, 97, 99, 100, 103, 104, 111, 115, 118, 119, 120, 121, 124, 128, 133, 134, 133, 282 variable sites 23, 130, 157, 205, 219, 218, 284 Vata 6, 12, 22, 76, 135, 256 weak crossover 29, 55, 56, 57, 58, 59, 60, 61, 62, 64, 65, 66, 67, 69, 105 weak pronouns 17, 150, 151, 152, 153, 166, 167

weak resumptives 14, 16, 41, 217, 218, 244 Welsh 6, 7, 8, 14, 15, 16, 19, 20, 22, 26, 41, 42, 76, 92, 93, 94, 96, 98, 99, 100, 101, 106, 133, 130, 131, 141, 149, 150, 159, 160, 161, 163, 166, 167, 217, 218, 219, 229, 230, 231, 241, 242, 251, 256, 287, 288, 289 wh-dependency 12, 15, 257, 285 wh-fronting 257, 277, 278 wh-in-situ 15, 260, 276, 285 wh-question 135, 259, 260, 276, 278, 285, 286, 287 wh-traces 1, 12, 13, 136, 256