University of Southern California Dissertations and Theses

Ellipsis constructions in Chinese

(USC Thesis Other)

Ellipsis constructions in Chinese

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ELLIPSIS CONSTRUCTIONS IN CHINESE Copyright 2002 by Hui-Ju Grace Li A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment o f the Requirements o f the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) August 2002 Hui-Ju Grace Li Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3094355 UMI UMI Microform 3094355 Copyright 2003 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA The Graduate School University Park LOS ANGELES, CALIFORNIA 90089-1695 This dissertation, w ritten b y / / c t / ___ ____________________ Under th e direction o f hex.... D issertation Com m ittee, and approved b y a ll its m em bers, has been p resen ted to an d accepted b y The G raduate School, in p a rtia l fulfillm ent o f requirem ents fo r th e degree o f DOCTOR OF PHILOSOPHY lean o f Graduate Studies D ate A ugust 6 , 2002 DISSER TA TION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Dedication To my beloved parents Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. iii Acknowledgements First and foremost, I am deeply grateful to the chair o f my dissertation committee, Yen-Hui Audrey Li, for her enlightenment and endless intellectual and personal support. Without her guidance and encouragement, I would not have entered the wonderland o f Linguistics. I am also deeply indebted to another committee member Hajime Hoji for his insightful suggestions and criticism on my work, which often led me to reconsider the data and arguments very carefully. His high standards have had a great influence on me. I consider it a great honor to have his valuable guidance. A warm and special thank you goes to Nam-Kil Kim, who generously agreed to be a member o f my committee. Despite his heavy schedule, he always spared the time to meet with me. I owe a great deal to my seniors Zoe Wu, Eric Guo and Luther Liu for useful discussions, comments and help throughout the writing o f this dissertation. I would like to reserve this last paragraph for three people who deserve the most mention o f all. I am thankful to Bob Hsin, who has witnessed and shared my happiness as well as sadness in all these years o f study. Without his love, understanding and unconditional help, my life would not have been easy and colorful. My parents, Pao-Pien Lee and Chin-Feng Lee, have always been there for me. Words cannot express my deep gratitude for their care and encouragement from childhood. I would like to dedicate this dissertation to them. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Dedication ii Acknowledgements iii Abstract viii Chapter 1: Overview 1 Chapter 2: Properties and Previous Analyses o f the Null Object Constructions 24 2.1 Introduction 24 2.2 Properties o f the Null Object Constructions 25 2.3 Previous Analyses o f the Null Object Constructions 27 2.3.1 Huang’s Analysis o f the Null Object Construction in Chinese 27 2.3.2 Otani and Whitman’s Analysis o f the Null Object Constructions in Chinese/Japanese/Korean 31 2.3.3 Hoji’s View o f the Null Object Construction in Japanese 33 2.3.4 Tomioka’s Analysis o f the Null Object Constructions in Chinese/Japanese/Korean 46 2.3.5 Kim’s Analysis o f the Null Object Construction in Korean 56 2.4 Properties o f the Chinese Null Object Construction Revisited 60 2.4.1 Differences between Stative/Resultative Verbs and Action Verbs 61 2.4.1.1 Interpretive Possibilities 63 2.4.1.2 Locality Effects 65 2.4.1.3 Requirement o f a Linguistic Antecedent 71 2.4.2 Similarities between Stative/Resultative Verbs and Action Verbs 74 2.4.2.1 Mix Readings 74 2.4.2.2 Confined Scope o f Adjuncts 78 2.4.3 Summary o f the Properties 80 2.5 Inadequacies o f the Previous Analyses o f the Null Object Construction in Chinese 81 2.5.1 Arguments against Huang’s and Otani and Whitman’s Analyses 81 2.5.1.1 Trichotomous Ambiguity 82 2.5.1.2 Absence o f Locality Effects 84 2.5.1.3 Blocking o f Auxiliaries 86 2.5.1.4 Confined Scope o f Adjuncts 88 2.5.1.5 Non-existence o f V-to-I Raising 92 2.5.2 Arguments against Hoji’s Suggestion 95 2.5.2.1 Availability o f Mix Readings 96 2.5.2.2 Unexpected Availability o f Sloppy Identity 97 2.5.2.3 Unexpected Unavailability o f Sloppy Identity 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. V 2.5.2.4 Unavailability o f Unspecified Readings 105 2.5.2.5 Interpretive Differences 106 2.5.2.6 Persistence o f Locality Effects 108 2.5.2.7 Relevance o f the Dependent Term 110 2.5.2.8 Relevance o f C-Command 113 2.5.3 Arguments against Tomioka’s Analysis 119 2.5.3.1 The Same Difficulties Faced by Hoji’s Suggestion 120 2.5.3.2 Unavailability o f Sloppy Identity for Carrot Sentences 122 2.5.4 Arguments against Kim’s Analysis 125 2.5.4.1 Interpretive Differences 125 2.5.4.2 Unexpected Unavailability o f Sloppy Identity 127 2.5.5 Arguments against the Null Topic Operator Analysis 130 2.5.5.1 Unexpected Unavailability o f Sloppy Identity 132 2.5.5.2 Unexpected Locality Effects 133 2.5.5.3 No Island Violations 135 2.5.6 Arguments against the Null Anaphor Analysis 137 2.5.6.1 Unexpected Availability o f Readings Other than the Anaphoric Reading 139 2.5.6.2 Unexpected Locality Effects and Absence o f Blocking Effects 141 2.5.7 Arguments against the Null Epithet Analysis 144 Chapter 3: The New Proposal o f the Null Object Construction in Chinese 148 3.1 V-to-v Movement 149 3.2 Revised VP Ellipsis Analysis o f the Chinese Null Object Construction 154 3.2.1 Deriving Strict/Sloppy Identity 155 3.2.2 Deriving Locality Effects 157 3.2.3 Surface Anaphora 159 3.2.3.1 Requirement o f a Linguistic Antecedent 159 3.2.3.2 Mix Readings 160 3.2.3.3 Missing Antecedent Phenomenon 162 3.2.4 Deriving Confined Adjunct Scope 165 3.2.5 No Island Violations 166 3.2.6 The Function o f Ye 167 3.3 Nominalization 168 3.3.1 Verbal Nouns 169 3.3.2 Deriving the Unspecified Reading 174 3.3.3 Deriving Absence o f Locality Effects 176 3.3.4 Stative Verbs Cannot Be Nominalized as Process Verbal Nouns 177 3.3.5 Resultative Verbs Cannot Be Nominalized 179 3.4 Predictions 186 3.4.1 Availability o f the Strict Reading When Reflexives Are Involved 187 3.4.2 Lexical Properties o f the Dependent Expressions 189 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. v i 3.4.3 The C-Command Condition 191 3.4.4 The Local Disjointness Effects 193 3.4.5 Mix Readings 195 3.4.5.1 Mix Readings and P -Occurrences 195 3.4.5.2 Mix Readings and the C-Command Condition 199 3.4.5.3 Mix Readings and Local Disjointness 201 3.4.6 Exchange Reading 202 3.5 Interaction between the Null Object Construction and Others 204 3.5.1 Topicalization 204 3.5.2 Double Object Structures 207 3.5.2.1 Goal DOS: V (gei) NP2 NP1 210 3.5.2.2 Goal DOS: V NP1 [v gei] NP2 216 3.5.2.3 Source DOS: V NP1 NP2 229 3.5.3 Control Constructions 234 3.5.4 Postverbal Elements 236 3.5.4.1 Quantity Adverbial Phrases 237 3.5.4.2 Predicative Complements 242 3.5.4.3 Purposive Clauses 245 3.5.4.4 Postverbal PPs 247 3.6 Differences between Chinese and Japanese 252 3.7 Differences between Chinese and English 254 3.8 Theoretical Implication 260 3.8.1 Syntactic Status o f the Empty Category in the Numeration 261 3.8.2 LF Copying vs. PF Deletion 266 3.9 Conclusion 275 Chapter 4: vP Ellipsis in Chinese 276 4.1 Dichotomous Ambiguity 278 4.2 Locality Effects 279 4.3 Surface Anaphora 280 4.3.1 Requirement o f a Linguistic Antecedent 281 4.3.2 Lexical Properties o f Dependent Expressions 282 4.3.3 The Local Disjointness Effects 284 4.3.4 Mix Readings 285 4.4 Scope o f Adjuncts 288 4.5 Postverbal Elements 289 4.5.1 Quantity Adverbial Phrases 290 4.5.2 Predicative Complements 292 4.5.3 Purposive Clauses 294 4.5.4 Postverbal PPs 295 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. v ii Chapter 5: The Null Topic Construction 296 5.1 Introduction 296 5.2 Structural Differences between the Null Object Construction and the Null Topic Construction 298 5.2.1 The VP Ellipsis Structure vs. the Null Topic Structure 298 5.2.2 Nominalization vs. the Null Topic Structure 304 5.2.3 Ambiguous Structures o f the Null Topic Construction 305 5.2.4 Confirmation 307 5.2.4.1 Distinct Predicates 307 5.2.4.2 Double Object Structures 309 5.2.4.3 Relative Clauses 311 5.3 Surface Anaphora vs. Deep Anaphora 313 5.4 Islands 318 5.5 Conclusion 322 Chapter 6: Disjunctive vs. A-Not-A Questions 323 6.1 Introduction 323 6.2 Disjunctive vs. A-Not-A questions 324 6.2.1 Lexical Integrity 326 6.2.2 Preposition Stranding 327 6.2.3 Island Constraints 328 6.3 Problems with Coordinate Deletion 332 6.3.1 The Directionality Constraint 332 6.3.2 Lexical Integrity 334 6.3.3 Preposition Stranding 335 6.3.4 Island Constraints 336 6.4 The Proposal 336 6.4.1 The Disjunctive Questions 337 6.4.2 The A-Not-A Questions 339 6.4.2.1 Interrogative Infl 339 6.4.2.2 Anaphoric Ellipsis or VP Ellipsis? 341 6.5 Conclusion 348 References 349 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract The major focus o f this study is to propose that the null object construction in Chinese involves VP ellipsis under the vP structure, which is obtained via V raising to v rather than to Infl. It is argued that distinction has to be made between the null object construction involving stative/resultative verbs and that involving action verbs, on the basis of their differences in interpretive possibilities and locality effects on sloppy identity. Although VP ellipsis is involved in both types o f the null object constructions, the observed differences arise due to an additional reanalysis process o f nominalization, which is only available to action verbs, but not stative or resultative verbs. It is also argued that anaphoric ellipsis involved in a type o f A-not-A questions is in fact an instance o f VP ellipsis. It is further suggested that the difference between Chinese and Japanese with respect to availability o f VP ellipsis is contingent upon that o f V-to-v raising. In Chinese, V is raised to v, making VP ellipsis possible; on the other hand, in Japanese, V stays in situ and thus VP ellipsis is not obtained. It is also suggested that the difference between Chinese and English with respect to availability o f the VP ellipsis representation might arise from the structural difference in subject positions between the two languages. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Overview l In this study, I examine the following constructions in Chinese that have been claimed to involve some type o f ellipsis: (i) null object constructions, (ii) null topic constructions, and (iii) A-not-A questions; and try to answer two questions: (i) do these constructions indeed involve ellipsis, and (ii) if yes, what type o f ellipsis is involved, assuming that ellipsis involves an elided part that is fully represented at LF as a copy o f its linguistic antecedent, much as in the case o f surface anaphora in the sense o f Hankamer and Sag (1976). Regarding the first question, I argue that the null object constructions and the A- not-A questions both involve ellipsis, whereas the null topic constructions, in line with earlier analyses, involve movement o f the null operator. Regarding the second question, I argue that both o f the constructions that are argued to be obtained via ellipsis involve VP ellipsis under the vP structure. Chapters 2, 3 and 4 examine the null object constructions in East Asian languages, with a focus on Chinese. There are two main views on the constructions. Since Huang (1988a), it is argued that the null object constructions as given in (1) and (2) should be treated on a par with VP ellipsis in English via an operation o f V- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 to-Infl raising, based on the observation that the former, like the latter, also exhibits strict/sloppy ambiguity and locality effects on sloppy identity. (1) Chinese (Huang 1988a, (21)) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mary ye [ln f l [v kanjian]-le] [w [v e][N P e]] also see Asp ‘Mary also saw (John’s mother).’ ‘Mary also saw (her mother).’ (2) Chinese (Huang 1988a, (24)) a. John kanjian-le tade mama see-Asp his mother ‘John saw his own mother.’ b. Mary zhidao Bill ye [I n f l [v kanjian]-le] [V P [v e][N P e]] know also see Asp ‘Mary knew that Bill saw John’s mother.’ ‘Mary knew that Bill saw his own mother.’ ‘*Mary knew that Bill saw her mother.’ In contrast, the other view, initiated by Hoji (1993, 1998a) for Japanese, argues against the VP ellipsis analysis o f the null object constructions, based on the following observations: (i) that the alleged sloppy identity reading is not always available, as in (3); (ii) that the null object constructions may yield a third reading in addition to strict/sloppy readings, as in (4); (iii) that the alleged locality effects on Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 sloppy identity can be nullified under appropriate contexts, as in (5); and (iv) that “Mix” readings that are typical o f VP ellipsis are unavailable, as in (6). (3) Japanese (Hoji 1998a, (12)) a. John-wa zibun(zisin)-o nagusameta Top self-Ace consoled ‘John consoled himself.’ b. Bill-mo [e] nagusameta also consoled ‘Bill consoled (John).’ ‘*Bill consoled (himself).’ (4) Japanese (Tomioka 1996, (19)) a. Ken-wa zibun-no kuruma-o arat-ta Top self-Gen car-Ace wash-Perf ‘Ken washed his car.’ b. Erika-mo [e] arat-ta also wash-Perf ‘Erica also washed (Ken’s car).’ ‘Erica also washed (her car).’ ‘Erica also washed (something).’ (5) Japanese (Hoji 1998a, (30)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 b. Mary-wa [C P Bill-ga [e] suisensita to] omotteita Top Nom recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘Mary thought that Bill also recommended (her student).’ (6) Japanese (Hoji 1997a, (5b)) Seihu-wa A sya-ni soko-no komonbengosi-ga soko-o uttaeta govemment-Top company A-Dat it-Gen attomey-Nom it-Acc sued to iw-ase-ta that say-make-Past ‘The government made Company A, say that its, attorney had sued it).’ B sya-ni mo iw-ase-ta company B-Dat also say-make-Past a. ‘(The government) made Company B2 say that its, attorney had sued it,.’ b. ‘(The government) made Company B2 say that its2 attorney had sued it2 .’ c. ‘*(The government) made Company B2 say that its2 attorney had sued it,.’ (Mix 1) d. ‘*(The government) made Company B2 say that itsi attorney had sued it2 .’ (Mix 2) Under this view, the alleged sloppy reading arises owing to the pragmatic (Hoji 1993, 1998a) or property anaphora (Tomioka 1999) or vehicle change (Kim 1999) use o f the null object. Some o f the arguments against the VP ellipsis analysis o f the null object constructions also hold true for Chinese. In particular, the null object construction in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chinese may also yield a third reading in addition to strict/sloppy readings, as in (7) and the locality effects on sloppy identity may be eliminated as well, as in (8). (7) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized John’s teacher, and Bill criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill criticized B ill’s teacher.’ ‘John criticized John’s teacher, and Bill criticized someone.’ (8) John piping-le tade laoshi, Bill zhidao Mark ye piping [e] le criticize-Asp his teacher know also criticize Asp ‘John criticized John’s teacher, and Bill knew that Mark criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill knew that Mark criticized Mark’s teacher.’ ‘John criticized John’s teacher, and Bill knew that Mark criticized someone (that could be Bill’s teacher).’ However, in contrast to Japanese, the sloppy identity reading is always available, as in (9), and so are “Mix” readings, as in (10). (9) a. John anwei-le ziji console-Asp self ‘John consoled himself.’ b. Bill ye anwei [e] le also console Asp ‘Bill consoled himself.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 (10) John shuo-guo tade laoshi piping-le ta, Bill ye shuo-guo [e] say-Asp his teacher criticize-Asp him also say-Asp ‘John, said his, teacher criticized him,; Bill2 said his, teacher criticized him,.’ ‘John, said his, teacher criticized him,; Bill2 said his2 teacher criticized him2 .’ ‘John, said his, teacher criticized himi; Bill2 said his2 teacher criticized him,.’ (Mix 1) ‘John, said his, teacher criticized him,; Bill2 said his, teacher criticized him2 .’ (Mix 2) The conflict thus arises with regard to the status o f the Chinese null object construction: is it an instance o f VP ellipsis or is it not? It is first noted in this study that the apparent conflict is readily resolved once a distinction is made between the types o f verbs involved in the relevant construction. To be specific, the Chinese null object construction that involves stative or resultative verbs exhibits the same range o f properties associated with VP ellipsis, namely, strict/sloppy ambiguity, locality effects on sloppy identity, and availability o f Mix readings, as shown in (11), (12), (13), respectively. On the other hand, the Chinese null object construction that involves action verbs yields strict/sloppy/unspecified ambiguity and does not exhibit locality effects on sloppy identity, as shown in (14) and (15), respectively. However, it also gives rise to Mix readings, as shown in (16). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 (11) a. John xihuan tade laoshi, Bill ye xihuan[e] like his teacher also like ‘John likes John’s teacher, and Bill likes John’s teacher.’ ‘John likes John’s teacher, and Bill likes Bill’s teacher.’ b. John kanjian-le tade laoshi, Bill ye kanjian [e] le see-Asp his teacher also see Asp ‘John saw John’s teacher, and Bill saw John’s teacher.’ ‘John saw John’s teacher, and Bill saw Bill’s teacher.’ (12) a. John xihuan tade laoshi, Bill zhidao Mark ye xihuan [e] like his teacher know also like ‘John likes John’s teacher, and Bill knows that Mark likes John’s teacher.’ ‘John likes John’s teacher, and Bill knows that Mark likes Mark’s teacher.’ ‘*John likes John’s teacher, and Bill knows that Mark likes B ill’s teacher.’ b. John kanjian-le tade laoshi, Bill zhidao Mark ye kanjian [e] le see-Asp his teacher know also see Asp ‘John saw John’s teacher, and Bill knows that Mark saw John’s teacher.’ ‘John saw John’s teacher, and Bill knows that Mark saw Mark’s teacher.’ ‘*John saw John’s teacher, and Bill knows that Mark saw B ill’s teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 (13) a. John shuo-guo tade laoshi xihuan ta, Bill ye shuo-guo [e] say-Asp his teacher like him also say-Asp ‘John! said hisi teacher liked him,; Bill2 said hisj teacher liked him,.’ ‘Johnj said hisi teacher liked him,; Bill2 said his2 teacher liked him2 .’ ‘Johni said his! teacher liked him,; Bill2 said his2 teacher liked him,.’ (Mix 1) ‘John, said his, teacher liked himi; Bill2 said his, teacher liked him2 .’ (Mix 2) b. John shuo-guo tade laoshi dashang ta, Bill ye shuo-guo [e] say-Asp his teacher hit-wound him also say-Asp ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him,.’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him2 .’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him,.’ (Mix 1) ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him2 .’ (Mix 2) (14) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized John’s teacher, and Bill criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill criticized Bill’s teacher.’ ‘John criticized John’s teacher, and Bill criticized someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (15) John piping-le tade laoshi, Bill zhidao Mark ye piping [e] le criticize-Asp his teacher know also criticize Asp ‘John criticized John’s teacher, and Bill knew that Mark criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill knew that Mark criticized Mark’s teacher.’ ‘John criticized John’s teacher, and Bill knew that Mark criticized someone (that could be Bill’s teacher).’ (16) John shuo-guo tade laoshi piping-le ta, Bill ye shuo-guo [e] say-Asp his teacher criticize-Asp him also say-Asp ‘Johni said his, teacher criticized him,; Bill2 said his, teacher criticized him,.’ ‘John, said his, teacher criticized him,; Bill2 said his2 teacher criticized him2.’ ‘John, said his, teacher criticized him,; Bill2 said his2 teacher criticized him,.’ (Mix 1) ‘John, said his, teacher criticized him,; Bill2 said his, teacher criticized him2 .’ (Mix 2) It is argued in chapter 3 that regardless o f the verb types, both o f the null object constructions in Chinese have access to the VP ellipsis representation, which is made possible via V-to-v raising, as illustrated in (17) and (18) for the null object constructions in (11) and (14), respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 10 (17) a. John [v.v xihuan] [V P tv [ Np tade laoshi]], Bill ye [V .v xihuan] [w tv [ n p e]] b. John [V - v kanjian] [w tv [ n p tade laoshi]], Bill ye [v.„ kanjian] [ w tv [ n p e ] ] (18) John [v.v piping] [V P tv [N P tade laoshi]], Bill ye [V .v piping] [ Vp tv [ n p e]] It is important to note that the notion o f “VP ellipsis” in this study is embedded under the vP structure. Hence, it is distinct from the “VP ellipsis” considered in the previous works, which is in fact “vP ellipsis” under our framework. In other words, unlike the previous analyses headed by Huang (1988a, 1988b, 1989, 1991a), this chapter argues that the Chinese null object constructions are the results o f VP ellipsis, but not vP ellipsis, based on the observation that V-to-v raising applies in the language, as independently argued in Huang (1994a, 1994b, 1997) and Tang (1998) whereas V-to-Infl raising does not, as argued in Tang (1998). It is further argued in this chapter that the availability o f the unspecified reading and the lack o f locality effects that are associated with the Chinese null object construction involving action verbs can be attributed to the fact that it also has access to the nominalization representation, in addition to the VP ellipsis representation, as shown in (19) for the null object construction in (14). This option, however, is not available for the one involving stative or resultative verbs, since stative or resultative verbs cannot be syntactically nominalized, as independently argued in Fu (1994). (19) Bill ye [„ DO] [N P [V P tv] [ v -n piping]] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 Given the proposed VP ellipsis and nominalization analysis o f the null object construction in Chinese, it is argued that the alleged similarities between the Chinese and Japanese null object constructions, as exemplified in (20) and (21) respectively, are more apparent than real. In particular, the former is the result o f VP ellipsis or nominalization, whereas the latter involves only the null object, but not VP ellipsis, as argued in Hoji (1993, 1998a), Tomioka (1996-1999), and Kim (1995, 1999). It is further suggested that the difference between Chinese and Japanese with respect to availability o f VP ellipsis is contingent upon that o f V-to-v raising. In Chinese, V is raised to v, making VP ellipsis possible; on the other hand, in Japanese, V stays in situ (as argued in Takano (1996)) and thus VP ellipsis is not obtained. (20) a. John tuijian-le tade xuesheng recommend-Asp his student ‘John recommended his student.’ b. Bill ye tuijian [e] le also recommend Asp ‘Bill also recommended (John’s student).’ ‘Bill also recommended (his student).’ ‘Bill also recommended (someone).’ (21) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Acc recommended ‘John recommended his student.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 b. Bill-mo [e] suisensita also recommended ‘Bill also recommended (John’s student).’ ‘Bill also recommended (Bill’s student).’ ‘Bill also recommended (someone).’ Given that the process o f V-to-v raising in Chinese makes the VP ellipsis representation possible, it is also tackled why English, which also has the process o f V-to-v raising, has no access to the VP ellipsis representation, namely, the null object construction, as indicated by the ungrammaticality o f (22), in contrast to the acceptable Chinese example in (23). (22) a. John likes his teacher, b. *Mary likes [e], too. (23) a. John xihuan tade laoshi like his teacher ‘John likes his teacher.’ b. Mary ye xihuan [e] also like ‘Mary also likes John’s teacher.’ ‘Mary also likes her teacher.’ It will be suggested that the difference between Chinese and English with respect to availability o f the VP ellipsis representation may arise from the structural difference in subject positions between the two languages. To be specific, in English, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject is always raised to [Spec, IP] while in Chinese, the subject occurs either in [Spec, IP] or in [Spec, vP] (as suggested in Aoun and Li (1989)). It is suggested in this chapter that the availability o f VP or vP ellipsis may correlate with the position o f a subject, as stated in (24). (24) Ellipsis o f a verbal projection is only possible with the subject occurring in the Spec position o f its governor. In other words, in order for VP ellipsis to be possible, the subject must occur in the Spec position o f v, which lexically governs the elided VP after V-to-v raising. On the other hand, in the case o f vP ellipsis, the subject must occur in the Spec position o f Infl, which lexically governs the elided vP due to the occurrence o f an auxiliary. (24) captures the fact that VP ellipsis such as (22) is unavailable in English, since an English subject is always raised to [Spec, IP]. In contrast, VP ellipsis like (23) is available in Chinese, since a Chinese subject may occur in [Spec, vP]. Although the statement in (24) is more like a preliminary generalization that calls for further articulation and deeper explanation, yet it is indirectly supported by the ellipsis facts concerning the double object structure in Chinese such as (25). (25) Zhangsan erzi kaoshang daxue le son pass college Asp ‘Zhangsan’s son passed the entrance exam for a college.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 In (25), Zhangsan is a major subject sitting in [Spec, IP] followed by a full clause whose subject, erzi ‘son,’ occurs in [Spec, vP] (as argued in Aoun and Li (1989)). Given that the major subject Zhangsan occurs in [Spec, IP] rather than [Spec, vP], (24) correctly predicts that only vP ellipsis, but not VP ellipsis, is available, as indicated by the contrast between (26b) and (26c). (26) a. Zhangsan erzi kaoshang daxue le son passed college Asp ‘Zhangsan’s son passed the entrance exam for a college.’ b. Lisi ye shi [v p e] also be ‘Lisi’s son also passed the entrance exam for a college.’ c. *Lisi ye erzi kaoshang [vp e] le also son pass Asp ‘Lisi’s son also passed the entrance exam for a college.’ Chapter 5 examines the similarities and differences between the null object construction in point and the null topic construction, as exemplified in (27). (27) A: John fasheng-le shemeshi? happen-Asp what ‘What happened to John?’ B: [T o p Johnji, Bill dashang-le [t]j hit-hurt-Asp ‘(John), Bill hurt.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 It is argued that the null object construction, or more accurately, both the VP ellipsis and nominalization representations associated with the null object construction, must be distinguished from the null topic representation despite their apparent similarity. First, although both the VP ellipsis and null topic representations may give rise to the strict reading, only the former, but not the latter, may give rise to the sloppy reading. (29) shows that both the VP ellipsis representation associated with the strict reading and the one associated with the sloppy reading are possible for (28b). On the other hand, only the null topic representation associated with the strict reading, but not the one that is intended to be associated with the sloppy reading is possible for (28b), as shown by the contrast between (30a) and (30b). (28) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Bill ye kanjian [e] le also see Asp ‘Bill also saw (John’s mother).’ ‘Bill also saw (his mother).’ (29) Strict Identity a. Billj ye [v [v kanjian]] [w tv [ n p Joluij de mama]] Sloppy Identity b. Billj ye [v [v kanjian]] [V P tv [N P tade, mama]] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 (30) Strict Identity a.. [John* de mama], Bill, ye kanjian-le [t] Sloppy Identity b. *[Tade, mama], Billj ye kanjian-le [t] Second, unlike the nominalization representation, the null topic representation does not yield the unspecified reading, as briefly argued below. Suppose that the unspecified reading in (3 IB) can result from the null topic representation in (3 IB’). We would expect the reading to be always available regardless o f the verb types involved. This is not the case, as indicated by the ungrammatically o f the following sentences involving a stative verb and a resultative verb. (31) A: Ni piping Johnle ma? you criticize Asp Q ‘Did you criticize John?’ B: Bu, danshi Bill piping [e] le; ta piping de shi Mark no but criticize Asp he criticize DE be ‘No, but Bill criticized (someone); the person whom he criticized is Mark.’ B ’: [ toP e], Bill piping [t] le; ta piping de shi Mark (32) A: Ni xihuan John ma? you like Q ‘Do you like John?’ B: *Bu, danshi [T o p e], Bill xihuan [t]; ta xihuan de shi Mark no but like he like DE be ‘No, but Bill likes (someone); the person whom he likes is Mark.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) A: Ni kandao John le ma? you see Asp Q ‘Did you see John?’ B: *Meiyou, danshi [T o p e], Bill kandao-le [t]; ta kanjian de shi Mark no but see-Asp he see DE be ‘No, but Bill saw (someone); the person whom Bill saw was Mark.’ Third, the null topic representation can be an instance o f deep anaphora, whereas the VP ellipsis representation can only be an instance o f surface anaphora, based on the observation that the latter requires a linguistic antecedent and allows a missing antecedent, but the former does not. Fourth, the null topic representation is subject to islands, whereas the VP ellipsis representation is not, as demonstrated below. In each o f the following examples, the first response, which additionally allows the VP ellipsis representation o f the object gap, is grammatical despite the fact that the object gap is inside an island; on the other hand, the second response is ungrammatical, since it only involves a null topic related to an object gap within an island. (34) CNPC - relative clause Q: Ni jiao-guo [vp ty nage xueshengi] ma? you teach-Asp that student Q ‘Have you taught that student?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. RlrMei, wo ye zhengzai zhao [tj jiao-guo [V P tv [e]( ]] de laoshij no I also Asp look for teach-Asp DE teacher ‘No, I am also looking for the teacher who has taught (that student).’ R2:*Mei, wo ye zhengzai zhao [t, jiao-guo [yp tv [e]; shuxue]] de no I also Asp look for teach-Asp math DE laoshij teacher ‘No, I am also looking for the teacher who has taught (that student) math.’ (35) CNPC - noun-complement clause S: Youren shuo John jiao-guo [w tv nage xueshengj; ta shi tade someone say teach-Asp that student he be his shuxue laoshi math teacher ‘Someone said that John has taught that student; he was his math teacher.’ R1:Shi ma? Buguo wo bu xiangxin [John jiao-guo [V P tv [e],]] zheju hua be Q but I not believe teach-Asp this word ‘Really? But I don’t believe the statement that John has taught (that student).’ R2:*Shima? Buguo, wo bu xiangxin [John jiao-guo [V P tv [e], shuxue]] be Q but I not believe teach-Asp math zheju hua this saying ‘Really? But I don’t believe the statement that John has taught (that student) math.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) Adjunct Condition Q: Bill zhidao John jiao-guo [V p tv nage xueshengj] ma? John shi tade( know teach-Asp that student Q be his shuxue laoshi math teacher ‘Does Bill know that John has taught that student? John was his math teacher.’ Rl:Zhidao. Bill hai [yinwei John jiao-guo [V P tv [e]j]] hen bu gaoxing. know because teach-Asp very not happy ‘Yes, Bill was unhappy because John has taught (that student).’ R2:*Zhidao. Bill hai [yinwei John jiao-guo [V P tv [e]; shuxue]] hen bu know because teach-Asp math very not gaoxing. happy ‘Yes, Bill was unhappy because John has taught (that student) math.’ Chapter 6 investigates the relevance o f the so-called “anaphoric ellipsis” to the null object construction. Anaphoric ellipsis is put forth in Huang (1991b) to account for the type o f A-not-A questions such as (37) and (38). (37) Takai chebu kai [chel? he drive car not drive (car) ‘Does he drive?’ (38) Tahui kai chebu hui fkai chel? he know-how-to drive car not know-how-to (drive car) ‘Does he know how to drive?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 20 According to Huang (1991b), the above A-not-A questions are derived from structures with coordinate vP’s o f the form [[AB] not [AB]], which may undergo a process o f anaphoric ellipsis that deletes the second occurrence o f B. However, it is argued in this chapter that Huang’s (1991b) anaphoric ellipsis, which is far from articulated, is nothing but a cover term for the already available mechanisms in the language, namely, VP and vP ellipses, based on the observation that the former simply duplicates precisely the effects o f the latter, as demonstrated in (39) for (37) and (40) for (38). (39) Ta kai che bu [V - „ kai] [V P tv [N P e]]? (40) Ta hui kai che bu [I n f l hui] [v P e]? That anaphoric ellipsis is a case o f VP/vP ellipsis, or in other words, the type o f A-not-A questions in (37) and (38) are the results o f VP and vP ellipses, respectively, is further supported by the following observations. First, the relevant type o f A-not-A questions patterns with a VP ellipsis sentence in that it is not subject to Ross’ (1967) Directionality Constraint. (41) Ta xihuan naben shu bu xihuan [e]? he like that book not like ‘Does he like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 21 (42) John xihuan tade shu, Bill ye xihuan [e] like his book also like ‘John likes his book, and Bill also likes (his book).’ In (41) and (42), the elided elements, which are traditionally derived by forward deletion, occur on a right branch. Second, both the relevant type o f A-not-A questions and a VP ellipsis sentence obey the Lexical Integrity Hypothesis. (43) *Ta xi-huan naben shu bu xi- [e]? he like that book not like ‘Does he like that book?’ (44) * John xi-huan tade shu, Bill ye xi- [e] like his book also like ‘John likes his book, and Bill also likes (his book).’ Third, both the relevant type o f A-not-A questions and a vP ellipsis sentence disallow preposition stranding. (45) *John hui dui tade nuuban shuo shihua bu hui dui [e]? will to his date say truth not will to ‘Will John tell his date the truth?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 22 (46) * John hui dui tade nuuban shuo shihua, Bill ye hui dui [e] will to his date say truth also will to ‘John will tell his date the truth, and Bill will also to (his date tell the truth).’ Lastly, postverbal elements such as duration/frequency phrases and predicative complements do not participate in a VP ellipsis sentence (see chapter 3) or the relevant type o f A-not-A questions, as illustrated in (47) and (48), respectively. (47) a. John nian shu nian-le san ci, Bill ye nian-le [e] read book read-Asp three times also read-Asp ‘John read three times, and Bill also read.’ b. John nian shu nian de hen lei, Bill ye nian-le [e] read book read DE very tired also read-Asp ‘John read so much that he was tired, and Bill also read.’ (48) a. * John nian shu nian-le san ci mei nian [e]? read book read-Asp three times not read ‘Did John read three times?’ b. *John nian shu niande hen lei mei nian [e]? read book read DE very tired not read ‘Did John read so much that he was tired?’ Given the proposed VP ellipsis analysis, the ungrammaticality o f the A-not-A questions in (48) is captured as follows. As shown in chapter 3, the interpretation o f the elided VP in the second conjunct does not include that o f the postverbal Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 frequency phrase or predicative complement in the first. Thus, the relevant LF representations of the ill-formed A-not-A questions in (48) are the ones given in (49). (49) a. *John [v P [A d v P san ci] [v P [v P [v.v nian][V P tv [N P shu]]][v . [v.v nian] three time read book read [v p ty tA d v p ]]]] mei [v p [v -v nian][V p tv [N P shu]]] not read book b. *John[v P [P r e C o m p hen lei] [v P [v P [V .v nian][w tv [N P shu]]][v . [v-v nian] very tired read book read [v p tv tp r eco m p ]]]] mei [v P [V .v nian]{yp. — v _[n p . shu]]] not read book The representations in (49) are ruled out for semantic reasons: the first conjunct questions the frequency or result with respect to an event o f reading, but at the same time, the second conjunct questions the existence of the event. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 2 Null Object Constructions 24 2.1 Introduction “What you see is what you get” is a common slogan in the business world. However, it does not always hold in the linguistic field. Under linguistic theories o f argument structures and full interpretation, what you see is oftentimes less than what you get. Issues o f the so-called “null arguments” have thus drawn tremendous discussion in the field. This chapter is dedicated to a special kind o f constructions involving null objects in VP ellipsis contexts, namely, the null object constructions. In particular, we will tackle the status and nature o f the null objects involved in East Asian languages, with a focus on Mandarin Chinese. This chapter is organized as follows. In section 2.2, general properties o f the null object constructions in Chinese, Japanese and Korean will be presented. In section 2.3, previous analyses o f the null object constructions in these languages will be reviewed. With this background knowledge in mind, we will revisit the Chinese null object construction in section 2.4, where a comprehensive discussion o f its properties will be given. In section 2.5, it will be demonstrated that none o f the previous Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 analyses presented can accommodate the whole range o f facts in Chinese. A new proposal will be put forth in chapter 3. 2.2 Properties of the Null Object Constructions in Chinese, Japanese and Korean The null object constructions refer to sentences where the object o f the verb in the second clause o f a pair o f sentences is apparently missing, as exemplified below. (1) Chinese (Huang 1988a, (21)) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mary ye kanjian [e] le also see Asp ‘Mary also saw (John’s mother).’ ‘Mary also saw (her mother).’ (2) Japanese (Otani & Whitman 1991, (4)) a. John-wa zibun-no tegami-o sute-ta. Nom self-Gen letter-Acc discard-Perf ‘John threw out se lf s letters.’ b. Mary-mo [e] sute-ta. also discard-Perf ‘Mary also threw out (John’s letters).’ ‘Mary also threw out (se lf s letters).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 (3) Korean (Otani & Whitman 1991, (3)) a. Chelswu-ka caki-uy phyenci-ul peli-ess-ta. Nom self-Gen letter-Acc discard-Past-Dec ‘Chelswu threw out se lf s letters.’ b. Yengmi-to [e] peli-ess-ta also discard-Past-Dec ‘Yengmi also threw out (Chelswu’s letters).’ ‘ Yengmi also threw out (se lf s letters).’ An important property o f the Chinese/Japanese/Korean null object constructions is that they all give rise to interpretive ambiguity between a referential or “strict” reading and a bound variable or “sloppy” reading, as indicated in their English translations. In (1), John and Mary saw the same person’s mother (the strict reading), or John and Mary saw their own respective mothers (the sloppy reading). In (2) and (3), the two people designated by the subjects threw out the same person’s letters (the strict reading) or their own respective letters (the sloppy reading). The distribution and interpretation o f the null objects in (l)-(3) is reminiscent o f the so-called VP ellipsis constructions in languages such as English. In the latter, one could find a VP gap instead o f an object gap, as shown below. (4) a. John threw out his letters. b. Mary did [V p e] too. ‘Mary threw out John’s letters.’ ‘Mary threw out her letters.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 As is well known, the VP ellipsis sentence in (4) also exhibits strict/sloppy ambiguity, as indicated in the interpretation. The first emerging question concerns the fundamental difference between Chinese, Japanese, and Korean on the one hand and English on the other with respect to possibility o f null objects, which we will take up in chapter 3. The similarity between the null object constructions in (l)-(3) and VP ellipsis in English has led Huang (1988a, 1988b, 1 9 8 9 ,1991a), followed by Otani and Whitman (1991), to treat the former on a par with the latter. In the following section, we will give a more detailed review o f their analyses as well as a different view initiated by Hoji (1993, 1998a), which is in turn followed by Kim (1995, 1999) and Tomioka (1996-1999). 2.3 Previous Analyses of the Null Object Constructions 2.3.1 Huang’ s Analysis o f the Null Object Construction in Chinese Based on the observation that Chinese null objects occurring in contexts such as (1) exhibit strict/sloppy ambiguity typically associated with VP ellipsis constructions such as (4), Huang (1988a, 1988b, 1989, 1991a) proposes to analyze such null objects, not as genuine null objects, but as null VPs in disguise. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It has been widely assumed that the English VP ellipsis sentence in (4) is made possible by a process o f cfo-support to license the empty VP in accordance with the ECP, which requires an empty category to be lexically governed. The Chinese sentence in (1), on the other hand, has the verb repeated in the second clause along with an empty object, instead o f having any auxiliary corresponding to do followed by an empty VP. Huang suggests that the repetition o f the verb in (1) serves the purpose o f “cfo-support” in English. To be precise, the second occurrence o f the verb is raised to Infl, which lexicalizes the abstract Infl node and enables it to properly govern the VP. The sentence then undergoes the process o f VP deletion. Under this hypothesis, what follows the verb in the second clause is then an empty VP, not a null object.1 In other words, VP ellipsis in Chinese may be obtained via VP deletion after a process o f V-to-Infl raising, as shown in the following representation. (5) Mary ye [I n f l [v kanjian]-le] [v p [v e][N P e]] According to Huang (1988a, 1988b, 1989, 1991a), the strict/sloppy ambiguity exhibited in the Chinese null object construction can now be straightforwardly accounted for by assigning the sentence the same kind o f LF representation assigned to English VP ellipsis in (4). In particular, Huang assumes that given the 1 Huang (1988a) suggests an alternative that the second occurrence of the verb is a “resumptive pro- VP,” which covers up the otherwise improperly goverened empty VP. However, he did not elaborate on this alternative. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 representation in (5), the LF interpretive rules o f Sag (1976) and Williams (1977) can be directly applied to derive the strict/sloppy readings.2 According to Sag-Williams’ account, the empty VP in (5) corresponds in LF to a lambda predicate, which is a variant o f the lambda predicate associated with the antecedent VP. If the antecedent VP is translated into Ax (x kanjian tade mama) ‘7.x (x saw his mother)’, the pronoun tade ‘his’ is taken to be referential and the strict reading is derived. On the other hand, if the antecedent VP is translated into Ax (x kanjian xde mama) ‘7.x (x saw x ’s mother)’, the pronoun is taken to be a variable bound by the subject the lambda predicate is a predicate o f and the sloppy reading is derived. Crucially, this line o f approach is not obtainable to a sentence involving merely a null argument rather than an empty VP, given the assumption that a null argument, which denotes individuals rather than properties, cannot be translated into a lambda predicate. Another piece o f argument is advanced in support o f the VP ellipsis treatment o f the relevant null object, which has to do with the locality effects with respect to sloppy identity, as in (6) and (7). 2 Note that Huang’s assuming Sag-Williams’ LF interpretive rules, which involve LF copying, seems in conflict with his another assumption that the alleged VP ellipsis is a result of the process of VP deletion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 (6) (= Huang 1988a, (23)) John saw his mother, and Mary knew that Bill did, too. ‘John saw his own mother, and Mary knew that Bill saw John’s mother.’ ‘John saw his own mother, and Mary knew that Bill saw his own mother.’ ‘*John saw his own mother, and Mary knew that Bill saw her mother.’ (7) (= Huang 1988a, (24)) John kanjian-le tade mama, Mary zhidao Bill ye kanjian [e] le see-Asp his mother know also see Asp ‘John saw his own mother, and Mary knew that Bill saw John’s mother.’ ‘John saw his own mother, and Mary knew that Bill saw his own mother.’ ‘*John saw his own mother, and Mary knew that Bill saw her mother.’ In both (6) and (7), the nonlocal sloppy interpretation ‘John saw his own mother, and Mary knew that Bill saw her mother’ is not available. Given Sag-Williams’ account, the availability o f the local sloppy reading on the one hand and the unavailability o f the nonlocal sloppy reading on the other hand are predicted, since the binder o f the sloppy pronoun is restricted to the subject o f the lambda predicate associated with the empty VP. In brief, the strict/sloppy ambiguity and locality effects exhibited in the Chinese null object construction under discussion has led Huang (1988a, 1988b, 1989, 1991a) to conclude that it may be better analyzed as analogous to VP ellipsis in English. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 2.3.2 Otani and Whitman’ s Analysis o f the Null Object Constructions in Chinese, Japanese and Korean Following Huang (1988a, 1988b, 1989, 1991a), Otani and Whitman (1991) also propose a similar analysis o f the null object construction in Japanese as involving VP ellipsis and claim that the results hold for Chinese and Korean as well. According to their view, the empty VP is derived by (i) the conditions that license a null object and (ii) an operation that raises the verb out o f VP into Infl. The LF interpretive rules o f Williams (1977) are then directly applied to derive sloppy identity. The procedure is exemplified below for the Japanese null object sentence in (2). (8) (= Otani and Whitman 1991, (11)) V-Raising Derivation a. John-wa [V P [ n p zibun-no tegami-o] [v sute]]-ta Mary-mo [V P [N P e] [v sute]]-ta V-Raising b. John-wa [V P [N P zibun-no tegami-o] tv][v sute]-ta Mary-mo [V P [N P e] tv][v sute]-ta D erived VP Rule c. John-wa [Lx [ v p x [N P zibun-no tegami-o] tv]][v sute]-ta Mary-mo [w [ n p e] tv][v sute]-ta Reflexive rule d. John-wa [Lx [ v p x [n p x-no tegami-o] tv]][v sute]-ta Mary-mo [V P [N P e] tv][v sute]-ta Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 32 VP Rule e. John-wa [Xx [V P x [N P x-no tegami-o] tv]][v sute]-ta Mary-mo [kx [V P x [N P x-no tegami-o] tv] ] [ v sute]-ta Otani and Whitman further show that the same restriction on locality o f the sloppy identity interpretation associated with VP ellipsis also applies to the null object construction in Japanese. (9) (= Otani & Whitman 1991, (26)) a. John-wa [C P NY Times-ga zibun-no kizi-o inyoosi-te i-ru to] John-Top NY Times-Nom self-Gen article-Acc quote-ing be-Imp C kik-ta hear-Perf ‘John heard that the NY Times is quoting se lf s article.’ b. Bill-mo [C P N Y Times-ga [e] inyoosi-te i-ru to] kik-ta Bill-also NY Times-Nom quote-ing be-Imp Comp hear-Perf ‘Bill also heard that the NY Times is quoting (John’s article).’ ‘*Bill also heard that the NY Times is quoting (se lf s article).’ As is clear from the English translations, the null object clause in (9b) only yields the local sloppy reading, but not the nonlocal sloppy reading. To sum up, Otani and Whitman (1991), following Huang (1988a, 1988b, 1989, 1991a), also claims that the null object construction in Japanese (as well as Chinese and Korean) can be treated as analogous to VP ellipsis in English via an operation o f V-to-Infl raising, based on the two observations as follows. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 (10) a. The null object construction in Japanese yields the sloppy identity reading just as VP ellipsis in English does, b. The null object construction in Japanese exhibits locality effects on the sloppy reading just as VP ellipsis does. 2.3.3 Hoji ’ s View o f the Null Object Construction in Japanese Hoji (1993, 1998a), followed by Kim (1995, 1999) and Tomioka (1996-1999), argues that the null object construction in Japanese cannot be analyzed on a par with VP ellipsis in English, in contrast to the view made in Otani and Whitman (1991). He also argues that the sloppy identity readings considered in the latter work are not genuine sloppy identity readings, and calls it sloppy-like readings, which, according to him, arise independently o f the alleged VP ellipsis status o f the construction. More specifically, he suggests that the relevant sloppy-like readings are obtained by the pragmatic contexts in which the null objects are used. In other words, the null object in question is an instance o f deep anaphora rather than surface anaphora in the sense o f Hankamer and Sag (1976).3 3 According to Hankamer and Sag (1976), “surface anaphora” such as VP ellipsis in English cannot be licensed pragmatically and needs a linguistic antecedent; on the other hand, “deep anaphora” such as do it can be licensed pragmatically and does not need a linguistic antecedent, as indicated by the contrast below. (i) [Hankamer attempts to stuff a 9-inch ball through a 6-inch hoop.] Sag: *It’s not clear that you’ll be able to. Sag: It’s not clear that you’ll be able to do it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 To argue that the null object construction in Japanese cannot be analyzed as analogous to VP ellipsis in English, Hoji (1998a) demonstrates that neither o f the two alleged empirical bases for the analysis as given in (10) is valid. Hoji first shows that the sloppy identity reading is not always available for the null object construction, unlike its VP ellipsis counterpart in English, as indicated by the contrast between (11) and (12). (11) (= Hoji 1998a, (12)) a. Johnrwa zibun(zisin)ro nagusameta Top self-Acc consoled ‘John consoled himself.’ b. *Billj-mo [e]j nagusameta also consoled ‘Bill consoled (e).’ (12) (= Hoji 1998a, (13)) a. John consoled himself. b. Bill did, too. Hoji argues that if a null object sentence such as (11) could be analyzed as an instance o f VP ellipsis, it should yield the sloppy reading, just as the English VP ellipsis sentence in (12) can. The fact is, however, contrary to the prediction made by the VP ellipsis analysis.4 4 We will return to Hoji’s account for the unacceptability of (1 lb) in footnote 11. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 Consideration o f the data involving otagai ‘each other’ as in (13)-(14) and paradigms with onazi ‘same’ and be tube tu ‘different’ as in (15)-(20) illustrates the same point.5 (13) (= Hoji 1998a, (14)) a. Subete-no nihonzin huuhu-ga otaeai-o nausameta, all-Gen Japanese couple-Nom each other-Acc consoled ‘Every Japanese couple consoled each other.’ b. Subete-no amerikazin huuhu-mo [e] nagusameta all-Gen American couple-also consoled ‘Every American couple also consoled (e).’ (14) (= Hoji 1998a, (16)) a. Every Japanese couple consoled each other. b. Every American couple did, too. The English VP ellipsis sentence in (14) can yield the sloppy reading that for each Japanese couple, the husband and the wife consoled each other, and for each American couple, too, the husband and the wife consoled each other. On the other hand, this reading does not seem to be available for the Japanese null object sentence in (13). 5 As Hoji notes, not all of the empirical paradigms in Hoji (1998a) are as compelling as they are wished to be, for there exist unwanted factors that may affect native speakers’ judgements. More sharpening experiments are conducted in Hoji (1998d). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 Similarly, the English VP ellipsis sentences involving same and different as in (15) and (16) can have the sloppy readings in (17) and (18), respectively. (15) (= Hoji 1998a, (17)) a. Every Japanese couple recommended the same student. b. Every American couple did, too. (16) (= Hoji 1998a, (18)) a. Every Japanese couple recommended different students. b. Every American couple did, too. (17) (= Hoji 1998a, (19)) For each Japanese couple, the husband recommended the same student as his wife; and for each American couple, too, the husband recommended the same student as his wife. (18) (= Hoji 1998a, (20)) For each Japanese couple, the husband recommended a student different from the one that his wife recommended; and for each American couple, too, the husband recommended a student different from the one that his wife recommended. However, the readings in (17) and (18) are not possible for the Japanese null object sentences in (19) and (20), according to Hoji (1998a). (19) (= Hoji 1998a, (25)) a. Subete-no nihonzin huuhu-ga onazi gakusei-o suisensita all-Gen Japanese couple-Nom same student-Acc recommended ‘Every Japanese couple recommended the same student.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 b. Subete-no amerikazin huuhu-mo [e] suisensita all-Gen American couple-also recommended ‘Every American couple also recommended (e).’ (20) (= Hoji 1998a, (26)) a. Subete-no nihonzin huuhu-ga betubetu-no gakusei-o all-Gen Japanese couple-Nom different-Gen student-Acc suisensita recommended ‘Every Japanese couple recommended different students.’ b. Subete-no amerikazin huuhu-mo [e] suisensita all-Gen American couple-also recommended ‘Every American couple also recommended (e).’ The contrast between Japanese comparative ellipsis and comparative deletion with respect to availability o f the sloppy reading provides compelling confirmation. (21) (= Hoji 1998a, (29)) a. Iintyoo-wa [subete-no nihonzin huuhu-ni yorij(-mo) sakini chairperson-Top all-Gen Japanese couple-Dat than early subete-no amerikazin huuhu-ni otagai-o nagusame-saseta all-Gen American couple-Dat each other-Acc console-caused ‘The chairperson made every American couple console each other earlier than every Japanese couple.’ (“sloppy reading” possible) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 b. Iintyooj-wa [[e; subete-no nihonzin huuhuk -ni chairperson-Top all-Gen Japanese couple-Dat [PROk [v p [e] nagusame]]-sase-ru] yori](-mo) sakini console-cause-Infl than early subete-no amerikazin huuhu-ni otagai-o nagusame-saseta all-Gen American couple-Dat each other-Acc console-caused ‘The chairperson! made every American couple console each other earlier than hei or she; made every Japanese couple console (e).’ (“sloppy reading” not possible) According to Hoji (1998a), the only difference between comparative ellipsis as in (21a) and comparative deletion as in (21b) is the absence in the former and the presence in the latter o f the predicate nagusame-sase-ru ‘console-cause-Infl’ inside the yo/7-clause/phrase. He thus assumes that the latter, where the object argument o f the predicate in the yorz'-clause/phrase is missing, contains a null object construction, but not the former. Now, it is argued in Hoji (1990, ch. 5) that Japanese comparative ellipsis is analogous to VP ellipsis in English in allowing the sloppy reading. He argues that if the null object construction could also be analyzed as analogous to VP ellipsis in English, then the comparative deletion example in (21b) should be able to yield the sloppy reading, just as the comparative ellipsis example in (21a) can. Contrary to this expectation, the example in (21b) does not seem to allow the sloppy reading, confirming Hoji’s claim that the null object construction in Japanese cannot be treated on a par with VP ellipsis in English. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 After arguing that the null object construction in Japanese cannot be analyzed on a par with VP ellipsis in English, Hoji (1998a) further demonstrates that the alleged “locality effects” on the “sloppy identity reading,” as illustrated in (22), are independent o f the properties o f VP ellipsis and can be eliminated under appropriate contexts. He argues in particular the null object sentence in (23), which results from substituting the Japanese nominative marker ga for mo ‘also’ in (22), allows the nonlocal sloppy reading, in contradistinction to (22).6 (22) (= Hoji 1998a, (8)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ b. Mary-wa [ Cp Bill-mo [e] suisensita to] omotteita Top also recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘*Mary thought that Bill also recommended (her student).’ (23) (= Hoji 1998a, (30)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ 6 Hoji (1998a) made a supplementary remark that the nonlocal sloppy reading is readily available under the following context. (i) John and Mary have been competing with each other in placing their students for good teaching positions. Ordinarily, whenever John recommends John’s student for a position, Mary also recommends Mary’s student for the same position. Now, Bill, Mary’s colleague, who used to be her student, does various things for Mary. He sometimes even recommends Mary’s students on behalf of Mary, so that Mary dos not have to do anything. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 b. Mary-wa [C P Bill-ga [e] suisensita to] omotteita Top Nom recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘Mary thought that Bill also recommended (her student).’ Hoji also notes that the same strategy is not available to English VP ellipsis. The elimination o f too in English does not make the nonlocal sloppy reading possible.7 (24) a. John recommended his student, b. Mary thought that Bill did. In addition to the above arguments against the VP ellipsis analysis, Hoji (1997a) gives another compelling and conclusive argument concerning what he calls “Mix readings,” as illustrated below. (25) Max said he saw his mother; Oscar did too. a. Maxi said he, saw hisx mother; Oscar2 said he! saw his, mother. b. Max, said hei saw hisi mother; Oscar2 said he2 saw his2 mother. c. Maxi said he] saw hist mother; Oscar2 said he2 saw his, mother. (Mix 1) d. *MaX] said he! saw hisi mother; Oscar2 said he, saw his2 mother. (Mix 2) 7 The locality effects persist for the English VP ellipsis sentence in (24), even under the favorable context as depicted in footnote 6. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 (26) Max said his mother saw him; Oscar did too. a. Maxi said hisi mother saw him,; Oscar2 said his, mother saw him,. b. Max, said hisi mother saw him,; Oscar2 said his2 mother saw him2 . c. Max! said hisx mother saw him,; Oscar2 said his2 mother saw him!. (Mix 1) d. Max! said hisi mother saw him,; Oscar2 said hisi mother saw him2. (Mix 2) (25) allows the across-the-board strict and sloppy readings in (25a, b) as well as what Hoji calls “Mix 1” reading in (25c), but not what Hoji calls “Mix 2” reading in (25d). On the other hand, (26) allows all o f the four readings in (26a, b, c, d). It has been argued in Hoji (1997a) that the so-called Mix readings are genuine sloppy identity readings in the sense that they are exclusively based on Formal Dependency8 and arise only in surface anaphora, such as VP ellipsis in English (cf. footnote 3).9 8 According to Hoji (1998d), Formal Dependency is an symmetrical relation of dependency and it can be established at LF only if the following conditions are met. (i) The three necessary conditions for an FD(A, B), where A and B are in argument positions: a. B is [+(}]. b. A c-commands B. c. A is not in the local domain of B. 9 According to Hoji (1997a), that an ellipsis site A , which is a linguistic object, is surface anaphora means that A , is a “reconstruction” of another linguistic object 8 in the sense of Fiengo and May (1994). Two linguistic objects are reconstructions of each other if they are identical except for (3- occurrences in them that appear in the identical structural configurations with respect to what they depend upon. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. N ow return to the following null object constructions in Japanese. 42 (27) Seihu-wa A sya-ni soko-ga soko-no komonbengosi-o govemment-Top company A-Dat it-Nom it-Gen attorney-Ace uttaeta to iw-ase-ta sued that say-make-Past ‘The government made Company A, say that if had sued its, attorney.’ B sya-ni mo iw-ase-ta company B-Dat also say-make-Past a. ‘(The government) made Company B2 say that iti had sued itSi attorney.’ b. ‘(The government) made Company B2 say that it2 had sued its2 attorney.’ c. ‘*(The government) made Company B2 say that it2 had sued its, attorney.’ (Mix 1) d. ‘*(The government) made Company B2 say that it! had sued its2 attorney.’ (Mix 2) (28) Seihu-wa A sya-ni soko-no komonbengosi-ga soko-o govemment-Top company A-Dat it-Gen attomey-Nom it-Acc uttaeta to iw-ase-ta sued that say-make-Past ‘The government made Company Ai say that itSi attorney had sued it!.’ B sya-ni mo iw-ase-ta company B-Dat also say-make-Past a. ‘(The government) made Company B2 say that itSi attorney had sued it,.’ b. ‘(The government) made Company B2 say that its2 attorney had sued it2 -’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 c. ‘*(The government) made Company B2 say that its2 attorney had sued it,.’ (Mix 1) d. ‘*(The government) made Company B2 say that its, attorney had sued it,.’ (Mix 2) The failure o f the null object constructions to yield Mix readings, as illustrated in (27) and (28), indicates that the relevant ellipsis sites cannot be instances o f surface anaphora and the sloppy readings observed in them cannot be based on Formal Dependency, unlike English VP ellipsis. In brief, based on the invalidity o f the two empirical bases for the VP ellipsis analysis and unavailability o f Mix readings for the Japanese null object construction, Hoji concludes that the null object construction in Japanese cannot be analyzed as analogous to VP ellipsis in English. Given the above conclusion, Hoji (1998a) claims that the “sloppy readings” o f the Japanese null object construction considered in Otani and Whitman (1991) are not genuine sloppy readings, but “sloppy-like readings,” which are obtained by the definite or “referential” and indefinite or “concept” uses o f the null object. To be more concrete, Hoji suggests that the content o f the N head o f the null object is supplied by the context o f discourse and the supplied N head may be a feature bundle, excluding phonological features, which can correspond to a definite Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 (including Names) or an indefinite (in English).10 Suppose the supplied N head is a Name. The null object can then enter into a coreference relation with another Name, as in (29), yielding the so-called “sloppy-like reading.”1 1 (29) (= Hoji 1998a, (33)) a. John-ga zibunzisin-o suisensita Nom self-Acc recommended ‘John recommended himself.’ b. Billj-mo [e]j suisensita also recommended ‘Bill also recommended (e).’ On the other hand, the supplied N head o f the null object can be a feature bundle that corresponds to a concept such as kuruma ‘car,’ as in (30). 10 Hoji’s suggestion is based on the assumptions that the structure of empty categories mirrors that of their overt counterparts and that a bare nominal in Japanese, such as kuruma ‘car,’ whose sole content is its N head, can be interpreted in various ways as a car, the car, cars, the cars, and more. I think you should mention that Hoji 1998a cites Ishii's work in this connection. 11 Hoji (1995) has argued that Principle B is not violated in (29b), since bound variable anphora is not involved, along the line of Reinhart (1983), wherein Principle B regulates only bound variable anaphora, not coreference. Therefore, the unacceptability of (1 lb) cannot be attributed to Principle B, either. Instead, Hoji (1995,268-269) suggests that the relevant factor regarding the ‘recommend’ and ‘console’ contrast, as in (29b) and (lib) respectively, is how easy it is to assign different guises to the coarguments that share the same denotation; apparently, it is easier to do so with suisens ‘recommend’ than with nagusame ‘console,’ presumably due to their semantico-functional properties. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 (30) (= Hoji 1998a, (35)) a. John-ga zibun-no kuruma-o aratta Nom self-Gen car-Ace washed ‘John washed s e lf s car.’ b. John igai-no subete-no hito-mo (minna) [e] aratta except-Gen all-Gen person-also (all) washed ‘Everyone other than John also washed (a car).’ As indicated in the English translation, the null object in (30b), where coreference relation cannot be established, may behave like an indefinite in English. In this case, indeterminacy may arise with regard to the possessor o f the car. However, it should be noted that the sentence can mean or imply that everyone other than John washed his or her own car. Note in passing that the supplied N head o f the null object in (29b) can also be a feature bundle that corresponds to a concept such as hito ‘man.’ In this case, indeterminacy may arise with regard to the person being recommended. However, it is possible to interpret the person being recommended to be Bill. Hoji’s view o f the null object as used conceptually (or as an indefinite) is supported by an observation made by Ueyama (personal communication with Hoji, 1994; see Hoji (1998a, 142-143)). She noted that the null object sentences in (29b) and (30b) can be followed by Dare(da)-ka siranai kedo ‘But I don’t know who’ and Dare-no kuruma(da)-ka siranai kedo ‘But I don’t know whose car,’ respectively. On Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 the other hand, VP ellipsis analogues o f (29b) and (30b) in English fail to yield the interpretive indeterminacy indicated in her Japanese examples. The contrast between the null object construction in Japanese and VP ellipsis in English with respect to interpretive determinacy thus constitutes another piece o f evidence in support of Hoji’s claim that the former cannot be analyzed on a par with the latter. To sum up, Hoji (1998a) argues that the null object construction in Japanese cannot be treated as analogous to VP ellipsis in English, in contrast to Otani and Whitman (1991). He further suggests that the sloppy identity readings considered in Otani and Whitman (1991) are not genuine sloppy identity readings, but sloppy-like readings, which are obtained by the referential or concept uses o f the null objects. 2.3.4 Tomioka’ s Analysis o f the Null Object Constructions in Chinese, Japanese, and Korean Tomioka (1996-1999), following Hoji (1993, 1998a), also argues against the VP ellipsis analysis o f the null object constructions. According to him, Otani and Whitman’s (1991) VP ellipsis analysis is too restrictive in that it predicts incorrectly the unavailability o f the sloppy identity readings in examples such as (31) and (32). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (31) (= Tomioka 1996, (10)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-kara-no self-Gen garden-Gen carrot-Nom McGregor Mr.-ffom-Gen okurimono desi-ta gift be-Perf ‘(Some) carrots from s e lf s garden were the gift (for us) from Mr. McGregor.’ b. Piitaa-mo [e] okutte-kuremasi-ta Peter-also send-give-Perf ‘Peter also sent us (some carrots from se lf s garden).’ (32) (= Tomioka 1996, (11)) a. John-ga yatto zibun-no apaato-made tadoritsui-ta sonokoro Nom finally self-Gen apartment-up to reach-Perf that time ‘By the time John finally arrived at his apartment,’ b. Bill-wa sudeni [e] soozisi-hazime-tei-ta Top already clean-began-be-Perf ‘Bill had already began cleaning (his apartment).’ In each o f (31) and (32), the internal structures o f the two VPs involved are distinct. Therefore, it is predicted that the sloppy identity readings are impossible for both (31) and (32) under the VP ellipsis analysis. This is, however, contrary to the fact, as indicated in the English translations. Despite agreeing with Hoji (1998a) in that the null object construction in Japanese cannot be analyzed on a par with VP ellipsis in English, Tomioka rejects Hoji’s suggestion on the indefinite or concept use o f the null object in question. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 According to Tomioka (1996), the sloppy-like readings in the sense o f Hoji (1998a) may not be attributed to pragmatic recoverability o f the null object as an indefinite, as suggested by the following contrast. (33) (= Tomioka 1996, (16)) a. Ken-wa zibun-ga sotugyoo-sita daigaku-ga kirai-da. Erika-mo Top self-Nom graduate-Perf college-Nom hateful-be also kirai-da hateful-be ‘Ken hates the college he graduated from. Erika also hates (the college she graduated from.’ b. Ken hates the college he graduated from. Erika also hates a college. In (33b), it is extremely hard to yield the sloppy-like reading for the indefinite pragmatically, in sharp contrast with the Japanese null object sentence in (33a).12 From a cross-linguistic perspective, Tomioka (1996) proposes to analyze the Japanese null object as a phonologically silent pronoun o f laziness or the laziest pronoun, which is a bare NP under the NP structure and is property-denoting. 12 According to Hoji (personal communication, 2002), Hoji (1998a) does not make a theoretical claim about the properties and the interpretations of indefinite NPs in English or those of the so-called null argument in Japanese. Therefore, it cannot be argued against, strictly speaking. It is clear, at least on the surface, that Japanese does not have an indefinite determiner and hence, the empty category in (33 a) does not quite corresponds to a college anyway if it is assumed that the property of the null argument mirrors that of an overt category in the language. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 To be specific, Tomioka (1996) observes that there is an otherwise near-complete parallelism between overt pronominals in English and p ro 's in Japanese: they both can be used as referential pronouns, bound variable pronouns, or E-type pronouns (see Tomioka (1997) for the relevant examples). This observation has motivated Tomioka (1996) to treat the Japanese null object, which he assumes to be a pro, as analogous to a pronoun o f laziness in English, which is used instead o f the repetition o f an antecedent, as exemplified by the well-known paycheck sentence. (34) A man who gives his paycheck to his wife is wiser than a man who gives it (= his paycheck) to his mistress Tomioka (1996) gives the following meaning to pronouns o f laziness. (35) (= Tomioka 1996, (18a)) A definite pronoun o f laziness: A .P e D<ejt>. 3x [Vy [II(y) <-> x = y] & P(x)] where II is a property denoting expression which is contextually salient. Given (35), the sloppy-like reading can be obtained in (36) by analyzing the object pro as in (37). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 (36) (= Tomioka 1996, (19)) a. Ken-wa zibun-no kuruma-o arat-ta Top self-Gen car-Acc wash-Perf ‘Ken washed his car.’ b. Erika-mo pro arat-ta also wash-Perf ‘Erica also washed.’ (37) X? € D < e,p.. 3x [Vy [[car(y) & belong-to(v)(y)] o x - y] & P(x)] where v is a variable corresponding to the reflexive zibun. According to Tomioka, (37) gives the interpretation, ‘Erica also washed her car.’ Tomioka (1996) further claims that there are in fact three properties available in the context o f (36): (i) XyeD. car(y) & belong-to(Ken)(y), (ii) AyeD. car(y) & belong-to(v)(y), and (iii) XyeD. car(y). (i) gives the strict reading, (ii) gives the sloppy-like reading, as in (33), and (iii) gives the unspecified reading, as observed by Ueyama (see the preceding section). Tomioka claims that his proposal is supported by the parallel contrast between English and Japanese with regard to the availability o f the sloppy-like reading. In particular, the “sloppy identity reading” considered in Otani and Whitman (1991) is unavailable for the null object sentence in (38), where the intended semantic antecedent, zibun-no hatake-no ninjin-ga ‘carrots from s e lf s garden’ is interpreted generically. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 (38) (= Otani and Whitman 1991, (7)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-no daikoobutu self-Gen garden-Gen carrot-Nom McGregor Mr.-Gen big favorite desi-ta be-Perf ‘The carrots from s e lf s garden were Mr. McGregor’s big favorite.’ b. Piitaa-mo [e] daisuki desi-ta Peter-also very fond o f be-Perf ‘Peter was also very fond o f (the carrots from Mr. McGregor’s garden).’ ‘*Peter was also very fond o f (the carrots from s e lf s garden).’ On the other hand, as discussed above, Tomioka (1996) claims that the reading becomes available for the null object sentence in (31), where the intended semantic antecedent, zibun-no hatake-no ninjin-ga ‘carrots from s e lf s garden’ is interpreted existentially. (31) (= Tomioka 1996, (10)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-kara-no self-Gen garden-Gen carrot-Nom McGregor Mr.-ffom-Gen okurimono desi-ta gift be-Perf ‘(Some) carrots from s e lf s garden were the gift (for us) from Mr. McGregor.’ b. Piitaa-mo [e] okutte-kuremasi-ta Peter-also send-give-Perf ‘Peter also sent us (some carrots from se lf s garden).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Pronominals in English exhibit a similar contrast, as shown below. 52 (39) (= Tomioka 1996, (25)) a. Carrots from his (own) garden were Mr. McGregor’s favorite. b. Peter liked them, too. ‘Peter liked the carrots from Mr. McGregor’s garden.’ Peter liked the carrots from his own garden.’ (40) (= Tomioka 1996, (26)) a. Some carrots from his (own) garden were the gift from Mr. McGregor. b. Peter sent us some, too. ‘Peter sent us some carrots from his own garden.’ However, Tomioka (1996, 1997) notes that there is a distributional mismatch between the Japanese pro o f laziness and pronouns o f laziness in English. Unlike the former, the distribution o f the latter is very restricted, as shown in (41), where the pronoun is put in conjunctions and the sloppy-like reading is not available. (41) (= Tomioka 1996, (27)) a. Gary likes his mother. b. Tim likes her, too. ‘Tim likes Gary’s mother.’ ‘*Tim likes Tim’s mother.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In addition to the distributional restriction, there is another property that English pronouns o f laziness do not share with Japanese pro o f laziness. That is, the pro o f laziness in Japanese can be indefinite, as in (42) and defined by Tomioka in (43). (42) (= Tomioka 1997, (10)) a. Ken-wa kuruma-o kat-ta Top car-Ace buy-Perf ‘Ken bought a car.’ b. Erika-mo pro kat-ta also buy-Perf ‘Erica also bought (a car).’ (43) (= Tomioka 1997, (11)) An indefinite pronoun o f laziness: XP e D<eit>. 3x [II(x) & P(x)] where II is a contextually salient property. The above facts have led Tomioka (1997) to look for another English candidate that comes the closest to the Japanese pro o f laziness. He suggests that the pro o f laziness in Japanese should be compared to N ’-deletion in English, as in (44). (44) (= Tomioka 1997, (12)) a. Ken bought a car, and Erica bought one (= a car), too. b. Professor Grim gave an A to many students and flunked only a few (= a few students). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 According to Tomioka, N ’-deletion in English, which is distinct from (definite) pronouns o f laziness, is a property anaphora. Likewise, the pro o f laziness in Japanese, which is a bare NP under the NP structure, is also property denoting. Moreover, the distribution o f N ’-deletion is much less restricted, as shown below. (45) (= Tomioka 1998, (27)) Conjunction Pablo sold many paintings o f his, but Juan sold only a few (a few paintings o f his). Tomioka (1998) further claims that the semantic tools used to derive the interpretation o f the relevant Japanese pro are 3-Closure (cf. Heim (1982) and Diesing (1992)) and type shifting o f a predicate to an individual, i.e., Iota (cf. Partee (1987)), both o f which are independently required to interpret a bare NP in the so- called “bare NP” or [+arg] languages in the sense o f Chierchia (1996). To be specific, it has been noted that languages such as Japanese allow a bare NP to occur in an argument position, with the definite/number specifications missing. The interpretation o f such an NP is largely contextually determined (cf. Kuroda (1992), Krifka (1995), and Chierchia (1996,1997)). For example, the bare NP object in (46) can be indefinite singular, definite singular, indefinite plural, or definite plural. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 (46) (= Tomioka (1998), (18)) Ken-wa ronbun-o yon-da Top paper-Acc read-Past ‘Ken read a paper / papers / the paper / the papers.’ Given these independently needed semantic tools for overt NPs in a bare NP language such as Japanese, Tomioka (1998) therefore proposes to use the same devices to interpret the pro, which is a property-denoting NP as well. In other words, the employment o f both semantic tools, namely, 3-Closure and Iota, makes the relevant pro o f laziness possible. As pointed out by Tomioka (1998) himself, the analysis he proposes has already been anticipated in Hoji (1998a). Note, however, that Tomioka’s analysis is embedded in a generalized proposal, where all instances o f Japanese pro, including the relevant null object, are reanalyzed as property anaphora analogous to N ’-deletion in English and their diverse interpretations are derived via the devices independently needed for ordinary bare NPs, i.e., 3-Closure and Iota. Readers are referred to Tomioka (1998) for the details. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 2.3.5 K im ’ s Analysis o f the Null Object Construction in Korean Kim (1995, 1999) argues against Otani and Whitman’s (1991) VP ellipsis analysis o f the null object constructions in Chinese, Japanese, and Korean and proposes a vehicle change account.13 Three arguments have been advanced in Kim (1995, 1999) to support the claim that the relevant null object constructions cannot be analyzed on a par with VP ellipsis in English. The first one concerns the double object constructions, as shown in (47), where the goal argument in the second VP, namely, amu ‘anybody,’ is different from that in the antecedent VP, namely, na ‘me.’ (47) (= Kim 1995, (7)) a. Peter-nun caki-uy sacin-ul na-eykey/lul poyecwu-ess-ta Top self-Gen picture-Acc I-Dat/Acc show-Past-Ind ‘Peter showed me his picture.’ b. Kulena Ruth-nun kyelko [e] amu-(eykey-)to an poyecwu-ess-ta But Top ever anybody-(Dat-)even not show-Past-Ind ‘But Ruth never showed anybody Peter’s picture.’ ‘But Ruth never showed anybody her picture.’ 13 Although Kim’s (1995) discussion is based on the Korean data, he claims that most of the points also appear to hold for Chinese and Japanese. Kim (1995) suggests that the following two alternatives might be adopted by Otani and Whitman to accommodate the double object construction in (47). One is to assume that what gets copied in (47) is some projection of V from the antecedent VP (V’ in this case) that does not contain the goal argument. The other is to move the goal arguments involved out of the VPs prior to LF. These two alternatives are, however, untenable when it comes to examples involving numeral quantifiers and part-whole relations. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 According to Kim, the availability o f the sloppy reading in (47b) poses a problem on Otani and Whitman’s (1991) VP ellipsis account for the following reason. Under the standard analysis o f the double object constructions, the second VP with the distinct goal argument still inside it cannot be made empty even after V-raising. As a result, the Sag-William’s VP copying adopted in Otani and Whitman (1991) could not apply and the sloppy reading would not arise.14 This, however, is contrary to the fact. The following sentences with a numeral quantifier as in (48) and part-whole relation as in (49) illustrate the same point. (48) (= Kim 1995, (8)) a. Alan-un caki-(uy) chinkwu-lul twumyeng-(ul) pul-less-ta Top self-(Gen) friend-Acc two.Cl-(Acc) invite-Past-Ind ‘Alan invited two o f his friends.’ b. Jane-un [e] seymyeng-(ul) pul-less-ta Top three.Cl-(Acc) invite-Past-Ind ‘Jane invited three o f Alan’s friends.’ ‘Jane invited three o f her friends.’ (49) (= Kim 1995,(13)) a. Jerry-nun caki-(uy) ai-lul phal-ul ttayli-ess-ta Top self-Gen child-Acc arm-Acc whip-Past-Ind ‘Jerry whipped/hit his child on the arm.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 b. Kulena Julia-nun [e] tali-lul ttayli-ess-ta but Top leg-Acc whip-Past-Ind ‘But Julia whipped/hit Jerry’s child on the leg.’ ‘But Julia whipped/hit her child on the leg.’ Under the VP ellipsis analysis o f Otani and Whitman, it is predicted that the sloppy identity readings would not be available for (48) and (49), since the second VPs in (48) and (49) with their respective numeral quantifier and part-NP inside them cannot feed VP copying.15 However, this prediction is disconfirmed. Both (48) and (49) allow a sloppy reading. 15 According to Kim (1995), the following representation for (48), whereby the relevant quantifier moves out of the VP, is impossible, due to a grammatical constraint on the distribution of floating quantifiers in Korean such that a floating quantifier must be c-commanded and preceded by its antecedent NP in syntax and the c-command requirement cannot be fulfilled by the trace of the floating quantifier left behind. (i) *... twumyeng-(ul)j [V p [n p caki-(uy) chinkwu-lul] [N P t,] tv]... Note that given the above constraint on the distribution of floating quantifiers, the other alternative of partial VP copying is also impossible, since there is no way to copy only the part of the antecedent VP that includes the desired materials but excludes the numeral quantifier. Similarly, both of the alternatives fail in the case of the part-whole construction such as (49), which has been argued to have the following structure (see Yoon (1989), Maling and Kim (1992) for the relevant discussion). (ii) [VP [whole-NP] [V’ [part-NP] [V]] Given the above representation, there is no way to copy only the V and the whole-NP, excluding the part-NP, since they do not form a constituent. Moreover, the other alternative of moving the part-NP out of the VP is not possible, since the part-NP, being nonreferential, is syntactically inert and must stay in place, as argued in Yoon (1989) and cited in Kim (1995, 1999). ^ The Dependency Theory proposed in Fiengo and May (1994) is a theory of indexical dependency. Under their view, an index is a complex object consisting of indexcial type, indicated by a superscripted a or p, and indexical value, indicated by a subscripted numeral. A nominal expression with a p-occurrence denotes an expression whose reference depends on another, whereas a nominal expression with an a-occurrence denotes an expression whose reference is independent of others. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 After arguing against the VP ellipsis analysis o f the null object constructions, Kim (1995, 1999) suggests a vehicle change account for the null objects. In particular, Kim assumes Dependency Theory proposed in Fiengo and May (1994) and extends its application to the null object.16 Under this analysis, the null object in question receives its interpretation via what Fiengo and May (1994) call “vehicle change,” which applies directly to the null object without VP reconstruction, as illustrated in (51) for the Korean null object sentence in (50). (50) a. John-un caki-uy emma-lul salangha-n-ta Topic self-Gen mother-Acc love-Pres-Ind ‘John loves his mother.’ b. Mary-ttohan [N P e] salangha-n-ta also love-Pres-Ind ‘Mary also loves his mother.’ ‘Mary also loves her mother.’ ‘Mary also loves John.’ (51) a. Strict: Mary“3 -ttohan [NPa! emma-lul]a2 salangha-n-ta b. Sloppy: MaryVttohan [NPP 3 emma-lul]p 4 salangha-n-ta According to Kim, the strict and sloppy readings o f the null object in (50) are obtained by reconstructing the indexical structure o f its antecedent NP, namely, caki- uy emma-lul ‘his mother,’ rather than the entire antecedent VP. The strict reading Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 obtains when the relevant pronoun bears an a-occurrence, as in (51a). On the other hand, the sloppy reading obtains when the pronoun bears a |3-occurrence, as in (51b). According to Kim (1995, 1999), his vehicle change account o f the null object, whereby only the antecedent NP instead o f the entire VP is reconstructed, is supported by the fact that the relevant null object can be co-referent to the subject o f the antecedent clause as well. In other words, the null object not only yields the strict/sloppy identity o f the familiar kind as English VP ellipsis does, but also allows the reading such that it refers back to the subject o f the antecedent clause. For example, the null object in (50b) can refer to the subject o f (50a), yielding the reading that Mary also loves John, as indicated in the translation. Note in passing that the three possible readings in (50) are also expected under the suggestion made in Hoji (1998a). The difference between Kim’s (1995, 1999) proposal and Hoji’s (1998a) suggestion seems to be that the former is too much more restrictive than the latter in the sense that it incorrectly predicts the unavailability o f the so-called “unspecified” reading as discussed in section 2.3.3, since there is no such an antecedent NP to be reconstructed. 2.4 Properties of the Chinese Null Object Construction Revisited It will be shown in the subsequent sections that the Chinese null object constructions involving different types o f verbs exhibit different properties regarding interpretive Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. possibilities and locality effects; in other words, the null object construction in Chinese is sensitive to verb types, which the previous researchers overlooked. The result o f this study suggests that distinction has to be made between stative and resultative verbs on the one hand and action verbs on the other. 2.4.1 Differences between Stative/Resultative Verbs and Action Verbs Stative verbs include verbs o f emotion such as xihuan ‘like’ and verbs o f cognition such as xinshang ‘appreciate.’ These verbs are compatible with a preceding hen ‘very,’ but incompatible with a following le ‘an aspect marker,’ as shown below. (52) a. John hen xihuan/xinshang Mary, very like/admire ‘John likes/admires Mary very much.’ b. * John xihuan/xinshang-le Mary like/admire-Asp ‘John likes/admires Mary.’ On the other hand, resultative verbs such as kanjian ‘look-see’ and dashang ‘hit- wound’ are incompatible with a preceding hen but compatible with a following le, in contrast to stative verbs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 (53) a. * John hen kanjian/dashang Mary. very see/hit-wound ‘John saw/wounded Mary very much.’ b. John kanjian/dashang-le Mary see/hit-wound-Asp ‘John saw/wounded Mary.’ Action verbs refer to verbs involving willful action on the part o f an initiator regardless o f results, such as piping ‘criticize,’ and da ‘hit.’ Like resultative compounds, these verbs are compatible with le but incompatible with hen, as shown below. (54) a. * John hen piping/da Mary. very criticize/hit ‘John criticizes/hits Mary very much.’ b. John piping/da-le Mary criticize/hit-Asp ‘John criticized/hit Mary.’ The distinction between stative and resultative verbs on the one hand and action verbs on the other that is crucial to the current discussion is that the null object construction involving stative/resultative verbs and that involving action verbs differ in two main aspects, namely, interpretive possibilities and locality effects, which I will now turn to. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 2.4.1.1 Interpretive Possibilities The null object construction involving stative/resultative verbs yields dichotomous strict/sloppy ambiguity as in (55), while that involving action verbs yields trichotomous ambiguity; i.e., it yields not only the familiar strict/sloppy identity readings but also what I will call “the unspecified reading,” as in (56). (55) a. John xihuan tade laoshi, Bill ye xihuan [e] like his teacher also like ‘John likes John’s teacher, and Bill likes John’s teacher.’ ‘John likes John’s teacher, and Bill likes Bill’s teacher.’ b. John kanjian-le tade laoshi, Bill ye kanjian [e] le see-Asp his teacher also see Asp ‘John saw John’s teacher, and Bill saw John’s teacher.’ ‘John saw John’s teacher, and Bill saw Bill’s teacher.’ (56) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized John’s teacher, and Bill criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill criticized Bill’s teacher.’ ‘John criticized John’s teacher, and Bill criticized someone.’ The above observation that the null object sentence involving a stative or resultative verb such as (55) does not give rise to the so-called unspecified reading, whereas that involving an action verb such as (56) does is supported by the following observation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 (57) a. John xihuan tade laoshi, Bill ye xihuan [e] like his teacher also like ‘John likes his teacher, and Bill likes (e), too.’ *Danshi wo bu zhidao Bill xihuan de shi shei but I not know like DE be whom ‘But I don’t know whom Bill likes.’ b. John kanjian-le tade laoshi, Bill ye kanjian [e] le see-Asp his teacher also see Asp ‘John saw his teacher, and Bill saw (e), too.’ *Danshi wo bu zhidao Bill kanjian de shi shei but I not know see D E be whom ‘But I don’t know whom Bill saw.’ (58) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized his teacher, and Bill criticized (e), too.’ Danshi wo bu zhidao Bill piping de shi shei but I not know criticize DE be whom ‘But I don’t know whom Bill criticized.’ As indicated, the null object sentences with stative/resultative verbs in (57) cannot be followed by sentences expressing referential indeterminacy o f the relevant null objects, whereas that with an action verb in (58) can. In other words, the former gives only the strict and sloppy readings, but not the unspecified reading. On the other hand, all o f the three readings are available for the latter. (Cf. Ueyama’s (1994) observation in section 2.3.3.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 2.4.1.2 Locality Effects The null object sentences with stative/resultative verbs exhibit locality effects, as shown in (59). (59) a. John xihuan tade laoshi, Bill zhidao Mark ye xihuan [e] like his teacher know also like ‘John likes John’s teacher, and Bill knows that Mark likes John’s teacher.’ ‘John likes John’s teacher, and Bill knows that Mark likes Mark’s teacher.’ ‘*John likes John’s teacher, and Bill knows that Mark likes Bill’s teacher.’ b. John kanjian-le tade laoshi, Bill zhidao Mark ye kanjian [e] le see-Asp his teacher know also see Asp ‘John saw John’s teacher, and Bill knows that Mark saw John’s teacher.’ ‘John saw John’s teacher, and Bill knows that Mark saw Mark’s teacher.’ ‘*John saw John’s teacher, and Bill knows that Mark saw B ill’s teacher.’ In both (59a) and (59b), the nonlocal sloppy reading such that the null object shares the same value with the matrix subject is not available. On the other hand, the null object construction involving action verbs seems to escape locality effects, as shown in (60). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (60) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized John’s teacher. ’ b. Bill zhidao Mark ye piping [e] le know also criticize Asp ‘Bill knew that Mark criticized John’s teacher.’ ‘Bill knew that Mark criticized Mark’s teacher.’ ‘Bill knew that Mark criticized someone (that could be B ill’s teacher).’ As indicated in the English translation, the null object sentence in (60b) allows the unspecified reading whereby the value o f the null object is indeterminate. However, what is crucial to our current discussion is that (60b) is compatible with the so-called nonlocal reading that Bill knew that Mark criticized Bill’s teacher, as supported by the fact that it can be followed by the sentence in (61), where the value o f the null object is explicitly specified. (61) Mark piping de jiushiB illde laoshi criticize DE be Gen teacher ‘The person whom Mark criticized was Bill’s teacher.’ Note that the null object sentences in (59), where stative/resultative verbs are involved, cannot be so followed. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 (62) *Mark xihuan/kanjian de jiushi Bill de laoshi like/see DE be Gen teacher ‘The person whom Mark likes/saw is/was Bill’s teacher.’ In section 2.3.3, we have discussed that one o f Hoji’s (1998a) arguments against Otani and Whitman’s (1991) VP ellipsis analysis o f the null object construction in Japanese concerns elimination o f the locality effects, as shown in (22)-(23). (22) (= Hoji 1998a, (8)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ b. Mary-wa [C P Bill-mo [e] suisensita to] omotteita Top also recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘*Mary thought that Bill also recommended (her student).’ (23) (= Hoji 1998a, (30)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ b. Mary-wa [C P Bill-ga [e] suisensita to] omotteita Top Nom recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘Mary thought that Bill also recommended (her student).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 The fact that the mere replacement o f mo ‘also’ in (22) by the nominative marker ga in (23) results in elimination o f the locality effects has led Hoji to conclude that the alleged “locality effects” on the “sloppy reading,” as in (22), must be independent o f the properties o f VP ellipsis. After arguing that the null object construction in Japanese cannot be treated on a par with VP ellipsis in English and Formal Dependency is not at issue for what he calls the sloppy-like readings, Hoji (1998a) further shows that the observed “locality effects” still persist even when zibun ‘s e lf is replaced by a Name, a category that clearly cannot be a P-occurrence. (63) (= Hoji 1998a, (53) a. John-ga John-no gakusei-o suisensita Nom Gen student-Ace recommended ‘John recommended John’s student.’ b. Mary-wa [C P Bill-mo [e] suisensita to] omotteita Top also recommended that thought ‘Mary thought that Bill also recommended John’s student.’ ‘Mary thought that Bill also recommended Bill’s student.’ ‘*Mary thought that Bill also recommended Mary’s student.’ By contrast, if mo ‘also’ is replaced with the nominative marker ga, the “nonlocal sloppy reading” becomes available, as shown in (64). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 (64) a. John-ga John-no gakusei-o suisensita Nom Gen student-Acc recommended ‘John recommended John’s student.’ b. Mary-wa [cp Bill-ga [e] suisensita to] omotteita Top Nom recommended that thought ‘Mary thought that Bill also recommended John’s student.’ ‘Mary thought that Bill also recommended B ill’s student.’ ‘Mary thought that Bill also recommended Mary’s student.’ The contrast between (63) and (64) therefore provides further support for Hoji’s claim that the “locality effects” reported for the Japanese null object construction are in fact due to the use o f mo and are independent o f VP ellipsis and Formal Dependency. N ow turn to the following null object sentences in Chinese, where the element corresponding to mo ‘also’ in Japanese, namely, ye ‘also,’ is eliminated. (65) a. John bu xihuan tade laoshi, danshi Bill zhidao Mark xihuan [e] not like his teacher but know like ‘John doesn’t like John’s teacher, but Bill knows that Mark likes John’s teacher.’ ‘John doesn’t like John’s teacher, but Bill knows that Mark likes Mark’s teacher.’ ‘*John doesn’t like John’s teacher, but Bill knows that Mark likes Bill’s teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 b. John mei kanjian tade laoshi, danshi Bill zhidao Mark kanjian [e] le not see his teacher but know see Asp ‘John didn’t see John’s teacher, but Bill knows that Mark saw John’s teacher.’ ‘John didn’t see John’s teacher, but Bill knows that Mark saw Mark’s teacher.’ ‘*John didn’t see John’s teacher, but Bill knows that Mark saw Bill’s teacher.’ It has just been discussed in the beginning o f this section that the Chinese null object sentences involving stative/resultative verbs exhibit locality effects on sloppy identity. Unlike Japanese, the locality effects are still detectable even with .ye ‘also’ eliminated, as indicated in the above interpretive possibilities. On the other hand, the absence o f locality effects in the null object construction with an action verb is not affected even when the relevant pronoun is replaced with a Name, as shown below. (66) A: Shei piping Johnde laoshi? who criticize DE teacher ‘Who criticized John’s teacher?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. B: John piping John de laoshi, Bill zhidao Mark ye piping [e] criticize DE teacher know also criticize ‘John criticized John’s teacher; Bill knew that Mark criticized John’s teacher.’ ‘John criticized John’s teacher; Bill knew that Mark criticized someone.’ Like (60) discussed earlier, the null object sentence in (66B) allows the unspecified reading, which is compatible with the “nonlocal sloppy reading.” In this connection, it is interesting to note that (66B), where a Name instead o f a pronoun is used, does not obligatorily gives rise to the local sloppy reading. However, as in the case o f the “nonlocal sloppy reading,” the “local sloppy reading” can be inferred. In brief, the null object construction involving stative/resultative verbs exhibits locality effects, which cannot be eliminated, while that involving action verbs escapes locality effects, 2.4.1.3 Requirement o f a Linguistic Antecedent It has been mentioned in footnote 3 that Hankamer and Sag (1976) classify anaphora into two types, namely, surface anaphora and deep anaphora, on the basis o f the following contrast. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 (67) (= Hankamer and Sag 1976, (3)-(4» [Hankamer attempts to stuff a 9-inch ball through a 6-inch hoop.] a. Sag: *It’s not clear that you’ll be able to. b. Sag: It’s not clear that you’ll be able to do it. They argue that “surface anaphora” such as English VP ellipsis in (67a) cannot be pragmatically licensed and requires a linguistic antecedent, whereas “deep anaphora” such as do it can be pragmatically licensed and does not need a linguistic antecedent. N ow turn to the Chinese null object construction. (68) a. [John looks at Mary with deep affection.] Mark: Bill ye xihuan [e] also like ‘Bill likes Mary, too.’ b. [John hit Mary on her face and Mary’s face swelled up.] Mark: Bill (shangge xiangqi) ye dashang-le [e] last week also hit-wound-Asp ‘Bill injured Mary (last week), too.’ (69) [John is criticizing Mary.] Mark: Bill ye piping [e] le also criticize Asp ‘Bill criticized Mary, too.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 Both (68), which involves a stative verb and a resultative verb, and (69), which involves an action verb, are acceptable. However, the result is more complicated when sloppy identity is considered, as show below. (70) a. [John looks at his date with deep affection.] Mark: Bill ye xihuan [e] also like ‘Bill likes John’s date, too.’ ‘*Bill likes Bill’s date, too.’ b. [John hit his date on the face, and her face swelled up.] Mark: Bill (shangge xingqi) ye dashang [e] le last week also hit-wound Asp ‘Bill wounded John’s date (last week), too.’ ‘*Bill wounded Bill’s date (last week), too.’ (71) [John is criticizing his date.] Mark: Bill ye piping [e] le also criticize Asp ‘Bill criticized John’s date, too.’ ‘Bill criticized someone, too.’ Like (68)-(69), the null object sentences in (70), where a stative verb and a resultative verb are used, and (71), where an action verb is used, allow the “strict” reading. However, the intended sloppy identity reading is not available for (70). On the other hand, the “sloppy” reading may be inferred from (71), as supported by the fact that it can be followed by the sentence Bill piping de jiushi taziji de nuuban ‘The Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 person whom Bill criticized was his own date.’ By contrast, (70) cannot be followed by the sentence Bill xihuan/dashang de jiushi taziji de nuuban ‘The person whom Bill likes/wounded is his own date.’ In brief, the Chinese null object constructions with either stative/resultative verbs or action verbs give rise to the strict reading in a context where no linguistic antecedent is present. However, only the ones with action verbs allow the reading that is compatible with the “sloppy reading.” Note in passing that it will be argued in chapter 5 that the “strict” readings observed above are in fact due to the force o f the null topic operator. 2.4.2 Similarities between Stative/Resultative Verbs and Action Verbs Although distinction has to be made between stative/resultative verbs and action verbs with regard to interpretation obtained in the null object construction, yet the two groups o f verbs in the constructions behave alike in terms o f Mix readings and adjunct scope. 2.4.2.1 Mix Readings As discussed in section 2.3.3, Hoji (1998d) has argued that Mix readings arise only in surface anaphora such as VP ellipsis in English, as demonstrated in (25)-(26), and they are based on Formal Dependency, which is subject to conditions concerning [5- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 occurrences o f the dependent expressions, c-command, and local disjointness. Therefore, the fact that the null object construction in Japanese does not give rise to Mix readings has been used in Hoji (1998d) as a confirming evidence for his claim that the Japanese null object construction cannot be treated on a par with VP ellipsis in English and the nature o f the “sloppy identity reading” observed in the former is distinct from that in the latter. (25) Max said he saw his mother; Oscar did too. a. Max, said he, saw his, mother; Oscar2 said he, saw his, mother. b. Max, said he, saw his, mother; Oscar2 said he2 saw his2 mother. c. Max, said he, saw his, mother; Oscar2 said he2 saw his, mother. (Mix 1) d. *Max, said he, saw his, mother; Oscar2 said he, saw his2 mother. (Mix 2) (26) Max said his mother saw him; Oscar did too. a. Max, said his, mother saw him,; Oscar2 said his, mother saw him,. b. Max, said his, mother saw him,; Oscar2 said his2 mother saw him2 . c. Max, said his, mother saw him,; Oscar2 said his2 mother saw him,. (Mix 1) d. Max, said his, mother saw him,; Oscar2 said his, mother saw him2 . (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 N ow turn to the null object construction in Chinese. (72) a. John shuo-guo ta xihuan tade laoshi, Bill ye shuo-guo [e] say-Asp he like his teacher also say-Asp ‘John! said hex liked his! teacher; Bill2 said he! liked hisi teacher.’ ‘John, said he, liked his, teacher; Bill2 said he2 liked his2 teacher.’ ‘John, said he, liked his, teacher; Bill2 said he2 liked his, teacher.’ (Mix 1) ‘*John, said he, liked his, teacher; Bill2 said he, liked his2 teacher.’ (Mix 2) b. John shuo-guo ta dashang tade laoshi, Bill ye shuo-guo [e] say-Asp he hit-wound his teacher also say-Asp ‘John, said he, wounded his, teacher; Bill2 said he, wounded his, teacher.’ ‘John, said he, wounded his, teacher; Bill2 said he2 wounded his2 teacher.’ ‘John, said he, wounded his, teacher; Bill2 said he2 wounded his, teacher.’ (Mix 1) ‘*John, said he, wounded his, teacher; Bill2 said he, wounded his2 teacher.’ (Mix 2) (73) a. John shuo-guo tade laoshi xihuan ta, Bill ye shuo-guo [e] say-Asp his teacher like him also say-Asp ‘John, said his, teacher liked him,; Bill2 said his, teacher liked him,.’ ‘John, said his, teacher liked him,; Bill2 said his2 teacher liked him2 .’ ‘John, said his, teacher liked him,; Bill2 said his2 teacher liked him,.’ (Mix 1) ‘John, said his, teacher liked him,; Bill2 said his, teacher liked him2 .’ (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 b. John shuo-guo tade laoshi dashang ta, Bill ye shuo-guo [e] say-Asp his teacher hit-wound him also say-Asp ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him,.’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him2 .’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him,.’ (Mix 1) ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him2 .’ (Mix 2) The null object sentences in (72), which involves a stative verb and a resultative verb, allows not only across-the-board strict and sloppy readings but also what Hoji calls “Mix 1” reading, but not “Mix 2” reading, whereas the sentences in (73), which involves the same stative/resultative verbs, allow all o f the four readings. The null object sentences involving action verbs behave alike, as shown below. (74) John shuo-guo ta piping-le tade laoshi, Bill ye shuo-guo [e] say-Asp he criticize-Asp his teacher also say-Asp ‘John, said he, criticized his, teacher; Bill2 said he, criticized his, teacher.’ ‘John, said he, criticized his, teacher; Bill2 said he2 criticized his2 teacher.’ ‘John, said he, criticized his, teacher; Bill2 said he2 criticized his, teacher.’ (Mix 1) ‘*John, said he, criticized his, teacher; Bill2 said he, criticized his2 teacher.’ (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 (75) John shuo-guo tade laoshi piping-le ta, Bill ye shuo-guo [e] say-Asp his teacher criticize-Asp him also say-Asp ‘John! said his, teacher criticized him,; Bill2 said his, teacher criticized him !.’ ‘Johni said hisi teacher criticized himi; Bill2 said his2 teacher criticized him2 .’ ‘Johni said hisi teacher criticized himi; Bill2 said his2 teacher criticized him,.’ (Mix 1) ‘John, said his, teacher criticized himi; Bill2 said his, teacher criticized him2 .’ (Mix 2) The fact that the null object constructions with stative/resultative verbs and action verbs do not show any difference with respect to availability o f the Mix readings is in fact expected, since what is at issue is the matrix verb rather than the embedded verb. Nevertheless, what is crucial is that the null object construction in Chinese gives rise to the Mix readings. 2.4.2.2 Confined Scope o f Adjuncts Under the VP ellipsis analysis o f the null object constructions, as proposed in Huang (1988a, 1988b, 1989, 1991a) and Otani and Whitman (1991), it is predicted that a VP-level adjunct in the antecedent clause should be interpreted as part o f the elided clause. This prediction, however, is not borne out. In particular, in the Chinese null object construction, no matter it involves a stative/resultative verb or an action verb, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 a VP-level adjunct in the first conjunct is not interpreted as part o f the gap in the second conjunct, as shown below. (76) a. John shenshende xihuan Mary, Bill ye xihuan [e] deeply like also like ‘John deeply likes Mary, and Bill likes Mary.’ b. John vinwei qian xihuan Mary, Bill ye xihuan [e] because o f money like also like ‘John likes Mary because o f money, and Bill likes Marv.’ (77) a. John qingchude kanjian-le Mary, Bill ye kanjian [e] le clearly see-Asp also see Asp ‘John clearly saw Mary, and Bill saw Mary.’ b. John vinwei yuanshi kanjian-le Mary, Bill ye kanjian [e] le because o f farsighted see-Asp also see Asp ‘John saw Mary because o f being farsighted, and Bill saw Mary.’ (78) a. John vanlide piping-le Mary, Bill ye piping [e] le harshly criticize-Asp also criticize Asp ‘John harshly criticized Mary, and Bill criticized Mary.’ ‘John harshly criticized Mary, and Bill criticized someone.’ b. John wei-le gongsi de hao piping Mary, Bill ye piping [e] for com pany‘s good criticize also criticize ‘John criticized Mary for the sake o f the company, and Bill criticized Mary.’ ‘John criticized Mary for the sake o f the company, and Bill criticized someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8° As indicated in (76), (77), and (78), which involve an emotion verb, a reception verb, and an action verb, respectively, the scope o f the VP-level adjuncts in the first conjuncts is confined to the clauses they occur, namely, the first adjuncts. 2.4.3 Summary o f the Properties The properties o f the null object constructions with stative and resultative verbs on the one hand and action verbs on the other are summarized in (79) and (80), respectively. A table showing the similarities and differences between the former and the latter is given in (81) as well. (79) Properties o f the null object construction with stative/resultative verbs a. It yields dichotomous strict/sloppy identity. b. It exhibits locality effects on sloppy identity, which cannot be eliminated. c. It gives rise to the Mix readings. d. The interpretation o f the gap in the second conjunct does not include the VP-level adjunct in the first conjunct. (80) Properties o f the null object construction with action verbs a. It yields trichotomous strict/sloppy/unspecified ambiguity. b. It does not exhibit locality effects. (The nonlocal “sloppy” reading is available.) c. It gives rise to the Mix readings. d. The interpretation o f the gap in the second conjunct does not include the VP-level adjunct in the first conjunct. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 (81) Similarities and Differences o f the Null Object Construction with Stative/Resultative Verbs and Action Verbs Strict Sloppy Nonlocal sloppy Unspecified Mix Stative/ Resultative + + + Action + + + + + 2.5 Inadequacies of the Previous Analyses of the Null Object Construction in Chinese Given the above more comprehensive discussion o f the properties associated with the null object construction in Chinese, we are now in a better position to evaluate the previous analyses that claim to be applicable to the Chinese null object construction. 2.5.1 Arguments against Huang’ s and Otani and Whitman’ s Analyses In this subsection, it will be shown that consideration o f the null object construction involving action verbs and other relevant data suggests that the aforementioned V-to- I raising/VP ellipsis analysis, as proposed in Huang (1988a, 1988b, 1989, 1991a) and followed by Otani and Whitman (1991), is inadequate. Moreover, the positioning o f negation, adverbs, and floating quantifiers seems to run afoul o f the postulation o f V- to-I raising. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5.1.1 Trichotomous Ambiguity 82 Recall that one o f Huang’s and Otani and Whitman’s arguments for the V-to-I raising/VP ellipsis analysis o f the null object construction in Chinese is based on the existence o f dichotomous strict/sloppy ambiguity, as shown in (1), repeated here. (1) (= Huang 1988a, (21)) John kanjian-le tade mama, Mary ye kanjian [e] le see-Asp his mother also see Asp ‘John saw his mother, and Mary also saw John’s mother.’ ‘John saw his mother, and Mary also saw her mother.’ The null object sentence in (1) involves a resultative verb, kanjian ‘see.’ However, consideration o f the null object construction involving action verbs as shown in (56), repeated here, obscures the argument. (56) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized John’s teacher, and Bill criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill criticized B ill’s teacher.’ ‘John criticized John’s teacher, and Bill criticized someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 The V-to-I raising/VP ellipsis analysis o f the null object construction expects that (56) should allow only the strict and sloppy readings, just as the English VP ellipsis sentence in (82). (82) John criticized his teacher, and Bill did, too. ‘John criticized John’s teacher, and Bill criticized John’s teacher.’ ‘John criticized John’s teacher, and Bill criticized Bill’s teacher.’ ‘*John criticized John’s teacher, and Bill criticized someone.’ However, as discussed in section 2.4.1.1, the Chinese null object sentence in (56) yields not only the strict and sloppy readings, but also a third unspecified reading that Bill criticized someone other than John’s and B ill’s teachers. The interpretive possibility o f the third reading in (56) is confirmed by the fact that the sentence can be followed by remarks on this “someone” as in (83), which is impossible for a VP ellipsis sentence in English, as shown in (84). (83) a. John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized his teacher, and Bill also criticized (e).’ b. Danshi Bill piping de shi tade tongxue but criticize DE be his classmate ‘But the person whom Bill criticized was his classmate.’ (84) a. John criticized his teacher, and Bill did, too. b. *But the person whom Bill criticized was his classmate. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 It is hard to see how the V-to-I raising/VP ellipsis analysis o f the relevant null object construction can accommodate this trichotomous interpretive ambiguity. 2.5.1.2 Absence o f Locality Effects As discussed in the previous section, the trichotomous strict/sloppy/unspecified ambiguity runs afoul o f one o f Huang’s (1988a, 1988b, 1989, 1991a) and Otani and Whitman’s (1991) arguments for the V-to-I raising/VP ellipsis analysis o f the null object construction in Chinese. In this subsection, it will be shown that the other argument based on the existence o f locality effects as in (7), repeated here, also runs into problems when more data are considered. (7) (= Huang 1988a, (24)) John kanjian-le tade mama, Mary zhidao Bill ye kanjian [e] le see-Asp his mother know also see Asp ‘John saw his own mother, and Mary knew that Bill saw John’s mother.’ ‘John saw his own mother, and Mary knew that Bill saw his own mother.’ ‘*John saw his own mother, and Mary knew that Bill saw her own mother.’ Like (1), the null object sentence in (7) involves a resultative verb, kanjian ‘see,’ and it exhibits locality effects, namely, the so-called nonlocal sloppy reading is impossible, as discussed in Huang (1988s, 1988b, 1989,1991a) and section 2.4.1.2. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. By contrast, the nonlocal sloppy reading becomes available, or more accurately, can be inferred when the relevant construction involves an action verb as in (60). (60) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized John’s teacher.’ b. Bill zhidao Mark ye piping [e] le know also criticize Asp ‘Bill knew that Mark criticized John’s teacher.’ ‘Bill knew that Mark criticized Mark’s teacher.’ ‘Bill knew that Mark criticized someone (that could be Bill’s teacher).’ The availability o f the nonlocal sloppy reading in (60) is confirmed by the fact that it can be followed by a remark that spells out the person criticized as B ill’s teacher. (85) Mark piping de jiushi Bill de laoshi criticize DE be Gen teacher ‘The person whom Mark criticized was Bill’s teacher.’ It has been shown that the V-to-I raising/VP ellipsis analysis o f the Chinese null object construction is not as adequate as it appeared to be, based on the observation that it only accounts for the ones involving stative/resultative verbs, but not the ones involving action verbs, which not only yield trichotomous ambiguity but also escape locality effects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 2.5.1.3 Blocking o f Auxiliaries As discussed in section 2.3.1, Huang attributes the possibility o f sloppy identity to the VP status o f the null object construction, which is obtained by raising V to Infl; in other words, V-to-I raising is the prerequisite to the existence o f sloppy identity. Huang (1988b, 174-178) argues that V-to-I raising does not take place if there is already an auxiliary in the Infl node.17 Given that the VP ellipsis configuration o f 1 7 This argument is advanced to account for the following contrasts. (i) (= Huang 1988b, (10)) Tapaode hen kuai he run DE very fast ‘He runs very fast.’ (ii) (= Huang 1988b, (13a)) Tapao de bu kuai he run DE not fast ‘He does not run fast.’ (iii)(= Huang 1988b, (13b)) *ta bu pao de kuai he not run DE fast ‘He does not run fast.’ (iv)(= Huang 1988b, (23)) ta bu hui pao de hen kuai he not will run DE very fast ‘He won’t run fast.’ The sentences (i)-(iv) each contain two verbal elements, the first of which, pao ‘run’ is the main verb and the second, kuai ‘fast’ is a depictive complement. When the sentence in (i) is negated, the complement verb but not the main verb is preceded by bu ‘not,’ as shown by the contrast between (ii) and (iii). Assuming that bu ‘not’ is a bound morpheme dominated by Infl and must be supported by an immediately following minimal verbal element, Huang accounted for the contrast between (ii) and (iii) as follows. Given the assumption, the bu in (iii) triggers the process of V-to-Infl raising by which the main verb pao ‘run’ is moved into Infl and combined with bu, forming a negated main verb. The resulting representation in which the negated main verb is followed by the depictive complement ‘fast’ is ruled out for semantic incompatibility: the sentence claims that he does not run but he is fast with respect to a non-existing action. On the other hand, the process of V-to-Infl raising does not take place if there is an auxiliary already in Infl, since the auxiliary, being a [+V] element, is already sufficient to support bu in Infl. Therefore, bu in (iv), which already has an auxiliary hui ‘will’ in Infl, does not trigger V-to-Infl raising and the sentence is grammatical, since there is no semantic incompatibility incurred. In other words, the occurrence of an auxiliary has the effect of preventing the main verb from being negated, thus no semantic incompatibility. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 the null object construction is created via V-to-I raising, and that the operation does not take place if there is an auxiliary, Huang would predict that what appears to be a null object in sentences with an occurrence o f an auxiliary cannot be a null VP in disguise, but a genuine null object, which cannot yield sloppy identity. The relevant cases, however, still give rise to sloppy identity, as shown in (86). (86) (At tomorrow’s school reunion....) John hui kandao tade laoshi, Bill ye hui kandao [e] will see his teacher also will see ‘John will see John’s teacher, and Bill will see John’s teacher.’ ‘John will see John’s teacher, and Bill will see B ill’s teacher.’ Given the interaction between Huang’s (1988b) hypothesis and his V-to-I raising analysis o f the null object construction, it is not clear how the availability o f the sloppy reading in (86) can be accommodated. In the following sections, it will be shown that the interpretation o f adjuncts and the positioning o f adverbs, negation, and floating quantifiers with respect to the verb shed serious doubt on the existence o f the raising o f V to Infl in Chinese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5.1.4 Confined Scope o f Adjuncts 88 It is noted that in an English VP ellipsis sentence, the adjunct in the first conjunct is interpreted as if it were part o f the elided VP in the second conjunct, as shown by the interpretation for the sentences in (87). (87) a. John clearly saw his mother, and Mary did, too. ‘John clearly saw John’s mother, and Mary clearly saw John’s mother.’ ‘John clearly saw John’s mother, and Mary clearly saw Mary’s mother.’ b. This couple resigned for different reasons, and that couple did, too. ‘This couple resigned for different reasons, and that couple resigned for different reasons.’ Such a fact is expected, given the widely held assumption that the auxiliary do occupies the Infl node, which is higher than VP, and a manner adverb such as clearly and an adjunct PP such as fo r different reasons are VP-level adjuncts. On the other hand, Hoji (personal communication, 1999) points out that the null object construction in Japanese fails to give rise to the interpretation o f the sort noted in (87), although the relevant examples he provides are more complicated. Hoji’s Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 observation is expected, given his claim that the null object in Japanese is not a VP in disguise, but a feature bundle whose content is supplied by the context o f discourse (cf. section 2.3.3). N ow consider the corresponding Chinese null object sentences in (88). (88) a. John qingchude kanjian-le tade mama, Mary ye kanjian [e] le clearly see-Asp his mother also see Asp ‘John clearly saw John’s mother, and Mary saw John’s mother.’ ‘John clearly saw John’s mother, and Mary saw Mary’s mother.’ b. Zhedui fufu wei-le butongde livou cizhi, nadui fufu ye cizhi this couple for different reason resign that couple also resign ‘This couple resigned for different reasons, and that couple resigned.’ Given Huang’s and Otani and Whitman’s V-to-I raising/VP ellipsis analysis, we would expect the manner adverb, qingchude ‘clearly’ in (88a) and the adjunct PP, wei-le butongde liyou ‘for different reasons’ in (88b) to be interpreted as part o f the elided conjuncts, as is the case for English, which is contrary to the fact. In the case o f (88b), where the verb cizhi ‘resign’ is intransitive, Huang and Otani and Whitman might argue that there is no ellipsis involved and the question just noted simply does not arise. However, it is not clear how they can account for the case in (88a), which does have a gap after the transitive verb kanjian ‘see.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 Although the elided part o f the second conjunct does not contain the meaning o f a manner adverb or an adjunct PP occurring in the first conjunct, it does contain that o f a temporal or locative adverb, as shown in (89). (89) John zuotian/zai xuexiao kanjian-le tade mama, Mary ye kanjian [e] le yesterday/at school see-Asp his mother also see Asp ‘John vesterdav/at school saw John’s mother, and Mary vesterdav/at school saw John’s mother.’ ‘John vesterdav/at school saw John’s mother, and Mary vesterdav/at school saw Mary’s mother.’ This fact, however, does not necessarily support the V-to-I raising/VP ellipsis analysis o f the null object construction in Chinese, given the observation that the second conjunct still contains the meaning o f a temporal or locative adverb occurring in the first conjunct even when there is no object gap in it, and hence no “VP ellipsis,” as shown in (90). (90) a. John zuotian/zai xuexiao kanjian-le tade mama yesterday/at school see-Asp his mother b. Mary ye kanjian-le tade mama also see-Asp her mother ‘John vesterdav/at school saw John’s mother, and Mary vesterdav/at school saw John’s mother.’ ‘John vesterdav/at school saw John’s mother, and Mary vesterdav/at school saw Mary’s mother.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Given that each event must have a temporal/locative anchor, it is possible that if the time/place is not specified in the very clause, it will be assumed to be the same as mentioned in the previous statement. In this connection, it is interesting to note a remark made by Hoji (personal communication, 1999). Consider the following sentence, where the temporal phrase o f the first conjunct contains a pronoun tade ‘his.’ (91) a. John zai tade shengri natian he-le pijiu on his birthday that day drink-Asp beer ‘John drank beer on his birthday.’ b. Bill ye he-le pijiu also drink-Asp beer ‘Bill also drank beer on John’s birthday.’ ‘*Bill also drank beer on B ill’s birthday.’ Hoji suggests that if (91), where clearly no ellipsis is involved, still gives rise to strict/sloppy ambiguity, then the mechanism that is responsible for the relevant ambiguity may account for the strict/sloppy ambiguity observed in the null object construction under discussion. However, as indicated, although the interpretation o f the second conjunct in (91b) may contain the temporal phrase occurring in the first conjunct, it does not give rise to strict/sloppy ambiguity, contrary to Hoji’s expectation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 2.5.1.5 Non-existence o f V-to-I Raising The most crucial claim o f Huang’s VP ellipsis analysis o f the Chinese null object construction is that the relevant verb is raised to Infl, creating an empty VP that feeds VP copying. However, consideration o f adverb placement and the positioning o f floating quantifiers and negation sheds serious doubt on the claim. First consider adverb placement. Pollock (1989) and Doron (1998) argue for the raising o f V to Infl in French and Hebrew, respectively, based on the fact that both French and Hebrew allow adverbs to intervene between the verb and its direct object as in (92) and (93), respectively. (92) Jean embrasse souvent Marie kisses often ‘John often kisses Mary.’ (93) dani menaSek li f amim et dina kisses sometimes Acc ‘Dani sometimes kisses Dina.’ In contrast with French and Hebrew, in Chinese adverbs do not intervene between a verb and its direct object; they occur only to the left o f a verb as in (94). (94) a. John changchang wen Mary often kiss ‘John often kisses Mary.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 b. * John wen changchang Mary kiss often It might still be argued that in Chinese, adverbs are originated higher than Infl. As a result, vacuous raising o f V to Infl gives rise to the word order in (94a) rather than (94b). However, this line o f argument is turned down by the fact that the relevant adverbs occur to the right o f an auxiliary, which is assumed to sit in the Infl node, as shown in (95). (95) John vinggai changchang wen Mary should often kiss ‘John should often kiss Mary.’ Another argument for the raising o f V to Infl in French and Hebrew is based on the positioning o f floating quantifiers to the right o f the verb, as in (96) and (97), respectively. (96) Mes amis aiment tous Marie my friends love all ‘My friends all love Mary.’ (97) ha- yeladim niSku Snevhem et dina the children kissed both Acc ‘The children both kissed Dina.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 In contrast to French and Hebrew, Chinese floating quantifiers occur to the left o f the verb as in (98). Yet, they can occur to the right o f an auxiliary as in (99). (98) a. Wode pengyou dou ai Mary my friends all love ‘My friends all love Mary.’ b. *Wode pengyou ai dou Mary my friends love all (99) Wode pengyou vinggai dou ai Mary my friends should all love ‘It should be the case that my friends all love Mary.’ Lastly consider the positioning o f negation. Pollock (1989) also gives the data involving the positioning o f negation to the right o f the verb as further argument for the raising o f V to Infl in French as in (100). In Chinese, the negative morpheme occurs to the right o f an auxiliary as in (101), but to the left o f the verb as in (102), unlike in French: (100) Jean (n’) aime pas Marie. like not ‘John does not like Mary.’ (101) John vinggai mei(you) chi should not eat ‘It should be the case that John didn’t eat.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 (102) a. John mei(you) kanjian Mary not see ‘John didn’t see Mary.’ b. * John kaniian mei(vou') Mary see not In brief, the above data concerning adverb placement, the positioning o f floating quantifiers and negation suggest that main verbs in Chinese do not raise across adverbs, floating quantifiers, and negation to Infl, thus undermining the V-to-I raising analysis o f the null object construction. It has been shown in this section that Huang’s (1988a, 1988b, 1989, 1991a) and Otani and Whitman’s (1991) V-to-I raising/VP ellipsis analyses o f the null object construction in Chinese are inadequate. In the subsequent sections, it will be shown that the suggestion made in Hoji (1998a), Kim (1995, 1999), and Tomioka (1996- 1999) cannot apply to the Chinese data, either. 2.5.2 Arguments against H oji’ s Suggestion As discussed in section 2.3.3, Hoji (1998a) argues against Otani and Whitman’s (1991) VP ellipsis analysis o f the null object construction on the basis o f the observation that unlike VP ellipsis in English, it does not always gives rise to the alleged sloppy identity and the locality effects exhibited in the construction can be eliminated. He further suggests that the sloppy identity considered is in fact sloppy- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 like identity, which is obtained by the pragmatic context in which the null object, a feature bundle, is used. In this section, it will be shown that Hoji’s view o f the null object construction in Japanese cannot be extended to that in Chinese. 2.5.2.1 Availability o f Mix Readings Recall that one o f Hoji’s arguments against treating the Japanese null object construction as analogous to English VP ellipsis concerns Mix readings, which are based on Formal Dependency and arise only in surface anaphora such as the latter. The unavailability o f Mix readings in the Japanese null object construction has led Hoji to conclude that the relevant construction cannot be an instance o f surface anaphora; instead, it is deep anaphora (see Hoji (1997a)). Suppose that Hoji’s claim applies to Chinese. We would expect that the null object construction in Chinese, being an instance o f deep anaphora rather than surface anaphora, would not give rise to Mix readings. This expectation, however, is disconfirmed, as indicated by the availability o f Mix readings in (103)-(104), repeated here (cf. section 2.4.2.1). (103) a. John shuo-guo ta xihuan/dashang/piping-le tade laoshi say-Asp he like/hit-wound/criticize-Asp his teacher ‘John, said he! liked/wounded/criticized hist teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 said he, liked/wounded/criticized his, teacher.’ ‘Bill2 said he2 liked/wounded/criticized his2 teacher.’ ‘Bill2 said he2 liked/wounded/criticized his, teacher.’ (Mix 1) ‘*Bill2 said he, liked/wounded/criticized his2 teacher.’ (Mix 2) (104) a. John shuo-guo tade laoshi xihuan/dashang/piping-le ta say-Asp his teacher like/hit-wound/criticize-Asp him ‘John, said his, teacher liked/wounded/criticized him,.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 said his, teacher liked/wounded/criticized him,.’ ‘Bill2 said his2 teacher liked/wounded/criticized him2 .’ ‘Bill2 said his2 teacher liked/wounded/criticized him,.’ (Mix 1) ‘Bill2 said his, teacher liked/wounded/criticized him2 .’ (Mix 2) The fact that the null object construction in Chinese, whether it involves a stative/resultative verb or an action verb, yields the Mix readings suggests that it cannot be treated on a par with the null object construction in Japanese. 2.5.2.2 Unexpected Availability o f Sloppy Identity It has been discussed in section 2.3.3 that another one o f Hoji’s (1998a) arguments against the VP ellipsis analysis o f the null object construction in Japanese concerns Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 unexpected unavailability o f certain sloppy identity, as exemplified in (11), which does not yield the intended sloppy reading, unlike its English YP ellipsis counterpart. (11) (= Hoji 1998a, (12)) a. John-wa zibun(zisin)-o nagusameta Top self-Acc consoled ‘John consoled himself.’ b. Bill-mo [e] nagusameta also consoled ‘Bill consoled (John).’ ‘*Bill consoled (himself).’ Under Hoji’s (1998a) suggestion that the available interpretations for the null object construction in Japanese depend upon the content o f the pragmatically supplied N head o f the null object, the unavailability o f the sloppy-like reading in (lib ) is attributed to the following two reasons: (i) the feature bundle that corresponds to ziburt ‘se lf cannot be supplied by the discourse context, and (ii) the coreferential reading between the subject Bill and the null object is impossible even if the supplied content o f the null object can correspond to that o f Bill; in other words, the null object sentence in (1 lb) is o f the same status as the sentence in (105). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 (105) *Billr ga Billi/karei/[e]i-o nagusameta Bill-Nom Bill/he/[e]-Acc consoled ‘Billj consoled Billi/himi/(e)i.’ N ow consider the relevant data in Chinese. Let us first look at the possibilities o f coreference. (106) a. * Billj anwei-le taj console-Asp him ‘B ilf consoled him;.’ b. *Bill; anwei-le [e]j console-Asp ‘B ilf consoled (e)j.’ c. ?Bill; anwei-le Bilf console-Asp ‘B ilf consoled Billj.’ Like its Japanese counterpart, the coreferential reading between the subject Bill and the object pronoun ta ‘him’ in (106a) is disallowed.18 Similarly, the coreference between the subject Bill and the null object in (106b) is impossible, whereas that 18 Note that unlike Japanese, coreference in Chinese seems to be subject to Principle B, for coreference within the local domain is still barred even if the verb anwei ‘console’ in (106a) is replaced by tuijian ‘recommend,’ as indicated by the unaccpetability of (i) (cf. footnote 11). (i) *Billj tuijian-le taj recommend-Asp him ‘Bilf recommended himj.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 between the two Names in (106c) is marginally acceptable. The relevant point here is that the intended coreference in (106b) is impossible in a pragmatically neutral context. However, it should be noted that the null object in (106b) may refer to someone whose reference is fixed outside o f the sentence. As a result, it may be interpreted as referring to Bill, given an appropriate context such as the following. (107) (Every colleague o f B ill’s knows that Bill gets upset easily, but he also calms down quickly by being able to console himself in the restroom. One day, Mark, who is new to the company, saw that Bill was extremely upset just a moment ago but already calmed down after returning from the restroom. He therefore asked....) M: Shei anwei-le Bill? who console-Asp ‘Who consoled Bill?’ (One o f the colleagues answered....) C: Bill anwei-le [e] console-Asp ‘Bill consoled (Bill).’ Likewise, the intended coreference in (106c) where Bill is used in the object position is readily acceptable in the context depicted above. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 Next look at the following null object sentence. (108) a. John anwei-le ziji console-Asp self ‘John consoled himself.’ b. Bill ye anwei [e] le also console Asp ‘Bill consoled John.’ ‘Bill consoled himself.’ ‘Bill consoled someone.’ As indicated, the sloppy reading is readily available for the Chinese null object construction in (108) without any favoring context.19 What is crucial here is that the relevant sloppy reading is o f the same status as that in (109) rather than that in (106). (109) Bill ye anwei-le ziji also console-Asp self ‘Bill also consoled himself.’ 19 Note that the availability of the sloppy reading in (108b) is irrelevant of the occurrence of ye ‘also,’ for the sloppy reading is still obtained even if ye ‘also’ is removed, as indicated below. (108) b’ Danshi Bill meiyou anwei [e] but not console ‘But Bill didn’t console John.’ ‘But Bill didn’t console himself.’ ‘But Bill didn’t console someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 In brief, if the null object construction in Chinese were o f the same nature as that in Japanese, we would expect at best the marginal availability o f the sloppy reading for (108) due to the marginal to impossible coreference in (106), which is contrary to the fact. Next recall that the fact that the null object construction in Japanese does not give rise to the sloppy identity reading when bound variable anaphora is at stake, as shown in (13) (repeated here), has been used in Hoji (1998a) as supporting evidence for the claim that the construction in question cannot be an instance o f surface anaphora; instead, it is deep anaphora. (13) (= Hoji 1998a, (14)) a. Subete-no nihonzin huuhu-ga otagai-o nausameta, all-Gen Japanese couple-Nom each other-Acc consoled ‘Every Japanese couple consoled each other; i.e. for each Japanese couple, the husband consoled his wife and the wife consoled her husband.’ b. Subete-no Amerikazin huuhu-mo [e] nagusameta all-Gen American couple-also consoled ‘♦Every American couple also consoled each other; i.e. for each American couple, the husband consoled his wife and the wife consoled her husband.’ Consider the following Chinese null object sentence, which involves bound variable anaphora. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 (110) a. Meidui Zhongguo fufu dou anwei-le bicide pengyou20 every Chinese couple all console-Asp each other’s friend ‘Every Chinese couple consoled each other’s friends; i.e. for each Chinese couple, the husband consoled his w ife’s friend and the wife consoled her husband’s friend.’ b. Meidui Meiguo fufu ye dou anwei [e] le every American couple also all console Asp ‘Every American couple consoled each other’s friends; i.e. for each American couple, the husband consoled his w ife’s friend and the wife consoled her husband’s friend.’ If the null object construction in Chinese were also an instance o f deep anaphora rather than surface anaphora, we would expect the sloppy reading to be impossible, just like the Japanese null object sentence in (13).21 The expectation, again, is not 20 The reason why a phrase containing pici ‘each other’ instead of pici itself is used in (110) is due to the fact that pici cannot occur in the object position; instead, it functions like an adverb, as indicated by the contrast in (i). (i) a. *Nadui fufu anwei-le pici that couple console-Asp each other ‘That couple consoled each other.’ b. Nadui fufu pici anwei-le that couple each other console-Asp ‘That couple consoled each other.’ 2 ' There was much discussion about bound variable anaphora involved in the null object construction in the course offered by Hoji (1997) at USC. The aspect of the discussion that is relevant here is that Japanese speakers also get the sloppy reading for the Japanese analogue of the null object sentence in (110). However, the sloppy reading gets tougher with the parallel reading of otagai ‘each other.’ Note that pici ‘each other’ in Chinese only allows the crossing reading, but not the parallel reading (i) Nadui fufu anwei-le picide pengyou that couple console-Asp each other’s friend ‘That couple consoled each other’s friends; i.e. the husband consoled his wife’s friend and the wife consoled her husband’s friend.’ ‘*The husband consoled his friend and the wife consoled her friend.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 bome out. The Chinese null object sentence in (110) gives rise to the intended sloppy reading. 2.5.2.3 Unexpected Unavailability o f Sloppy Identity In this section, it will be shown that consideration o f the data involving topicalization o f a category that contains a pronoun throws doubt on attributing the sloppy identity observed in the Chinese null object construction to the pragmatic use o f the relevant null argument, as proposed in Hoji (1998a) for the Japanese null object construction. Consider the following null object construction and its topicalized counterpart, as shown in (11 la) and (111b), respectively. (Ill) a. John hen xihuan wo gei ta de shu, Bill ye hen xihuan [e] very like I give him DE book also very like ‘John likes the book I gave John; Bill also likes the book I gave John.’ ‘John likes the book I gave John; Bill also likes the book I gave Bill.’ b. Wo gei ta de shu, John hen xihuan, Bill ye hen xihuan [e] I give him DE book very like also very like ‘John likes the book I gave John; Bill also likes the book I gave John.’ ‘*John likes the book I gave John; Bill also likes the book I gave Bill.’ As indicated, the null object sentence in (111a) gives rise to the sloppy reading as well as the strict reading. On the other hand, the sloppy reading is not available in (111b), where the phrase containing the relevant pronoun is topicalized. If the sloppy Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 105 reading observed in the Chinese null object construction arises from the pragmatic use o f the null argument itself, the contrast between (111a) and (111b) with respect to the availability o f the sloppy reading is puzzling (More discussion on the relevant issue will be given in chapter 3). 2.5.2.4 Unavailability o f Unspecified Readings In this section, I will present still another piece o f evidence in support o f my claim that Hoji’s suggestion regarding the Japanese null object construction cannot hold for Chinese. Recall that according to Hoji (1998a), the content o f the N head o f the null object is supplied by the context o f discourse and the supplied N head may be a feature bundle that corresponds to a concept such as hito ‘man,’ as in (29), repeated here (cf. section 2.3.3). (29) (= Hoji 1998a, (33)) a. John-ga zibunzisin-o suisensita Nom self-Acc recommended ‘John recommended himself.’ b. Bill-mo [e] suisensita also recommended ‘Bill also recommended (a man).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 In this case, indeterminacy may arise with regard to the person being recommended. In other words, in addition to the regular strict and sloppy-like readings, the Japanese null object sentence in (29) also yields the so-called “unspecified” reading. If the null object construction in Chinese were o f the same nature as that in Japanese, we would expect the three readings noted in the latter to be available for the former as well. The expectation, however, is not fully bome out. As discussed in section 2.4.1.1, null object sentences with stative/resultative verbs do not yield the third unspecified reading, as shown in (112). (112) a. John xihuan ziji, Bill ye xihuan [e] like self also like ‘John likes John, and Bill likes John.’ ‘John likes John, and Bill likes Bill.’ ‘*John likes John, and Bill likes someone.’ b. Zai jingzi li, John kanjian-le ziji, Bill ye kanjian [e] le in mirror inside see-Asp self also see Asp ‘In the mirror, John saw John, and Bill saw John.’ ‘In the mirror, John saw John, and Bill saw Bill.’ ‘*In the mirror, John saw John, and Bill saw someone.’ 2.5.2.5 Interpretive Differences According to Hoji (1998a), the sloppy-like reading may result from the referential or concept use o f the relevant null object. In other words, the sloppy-like reading Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. obtained in (29b), repeated here, is more o f the same status as that in (113a) rather than that in (1 13b). (29) (= Hoji 1998a, (33)) a. John-ga zibunzisin-o suisensita Nom self-Ace recommended ‘John recommended himself.’ b. Bill-mo [e] suisensita also recommended ‘Bill also recommended (e).’ (113) a. Bill-mo Bill-o suisensita also Acc recommended ‘Bill also recommended Bill.’ b. Bill-mo zibunzisin-o suisensita also self-Acc recommended ‘Bill also recommended himself.’ On the other hand, the “sloppy” reading noted in the Chinese null object construction such as (112), repeated here, is the opposite. It is more o f the same nature as that in (114b), rather than that in (114a) (cf. section 2.5.2.2). (112) a. John xihuan zij i like self ‘John likes himself.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 b. Bill ye xihuan [e] also like ‘Bill also likes (e).’ (114) a. Bill ye xihuan Bill also like ‘Bill also likes Bill.’ b. Bill ye xihuan ziji also like self ‘Bill also likes himself.’ 2.5.2.6 Persistence o f Locality Effects As reviewed in section 2.3.3, Hoji’s (1998a) another argument against the VP ellipsis analysis o f the null object construction is based on the observation that the alleged locality effects observed are independent o f the VP ellipsis status, for the locality effects can be nullified if mo ‘also’ is replaced by the nominative marker ga. (23) (= Hoji 1998a, (30)) a. John-wa zibun-no gakusei-o suisensita Top self-Gen student-Ass recommended ‘John recommended s e lf s student.’ b. Mary-wa [cp Bill-mo [e] suisensita to] omotteita Top also recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘*Mary thought that Bill also recommended (her student).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 b \ Mary-wa [cp Bill-ga [e] suisensita to] omotteita Top Nom recommended that thought ‘Mary thought that Bill also recommended (John’s student).’ ‘Mary thought that Bill also recommended (se lf s student).’ ‘Mary thought that Bill also recommended (her student).’ On the other hand, we have shown in section 2.4.1.2 that the locality effects are still detectable for the null object sentences involving stative/resultative verbs, even with ye ‘also,’ equivalent to the Japanese mo, eliminated, as shown below. (65) a. Suiran John bu xihuan tade laoshi, Bill zhidao Mark xihuan [e] although not like his teacher know like ‘John doesn’t like John’s teacher, but Bill knows that Mark likes John’s teacher.’ ‘John doesn’t like John’s teacher, but Bill knows that Mark likes Mark’s teacher.’ ‘*John doesn’t like John’s teacher, but Bill knows that Mark likes B ill’s teacher.’ b. Suiran John mei kanjian tade laoshi, Bill zhidao Mark kanjian [e] le although not see his teacher know see Asp ‘John didn’t see John’s teacher, but Bill knows that Mark saw John’s teacher.’ ‘John didn’t see John’s teacher, but Bill knows that Mark saw Mark’s teacher.’ ‘*John didn’t see John’s teacher, but Bill knows that Mark saw B ill’s teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 If the null object construction in Chinese patterns with that in Japanese, it is unclear how the contrast between the former and the latter with respect to locality effects can be captured. 2.5.2.7 Relevance o f the Dependent Term After proposing the analysis o f the “sloppy-like” readings observed in the Japanese null object construction as owing to the properties o f the null argument itself rather than the alleged VP ellipsis status, Hoji (1998a) further demonstrates that a number o f predictions made by the analysis are confirmed. One o f them involves the fact that the lexical properties o f the dependent term are irrelevant to the availability of the sloppy-like readings, as shown below. (115) (= Hoji 1998a, (48)) a. John-ga John-no kuruma-o aratta Nom Gen car-Acc washed ‘John washed John’s car.’ b. Bill-mo [e] aratta also washed ‘Bill washed (e).’ (116) (= Hoji 1998a, (35)) a. John-ga zibun-no kuruma-o aratta Nom self-Gen car-Acc washed ‘John washed s e lf s car.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I l l b. Bill-mo [e] aratta also washed ‘Bill washed (e).’ According to Hoji, despite the use o f the Name John in place o f zihun ‘s e lf in (115a), the null object sentence in (115b) seems to have a status comparable to that o f (116b): both give rise to the sloppy-like reading that Bill washed Bill’s car. Now turn to the data in Chinese. (117) a. Johnxi-le Johnde chezi wash-Asp Gen car ‘John washed John’s car.’ b. Bill ye xi [e] le also wash Asp ‘Bill washed John’s car.’ ‘Bill washed something.’ The Chinese null object sentence in (117), with the Name John in place o f ziji ‘self,’ can yield, or more accurately, imply the intended sloppy-like reading such that Bill washed B ill’s car, as supported by the fact that the sentence can be followed by the remark that spells out the value o f the null object in (1 17b) as that o f B ill’s car. (117) c. B illx i de shi zijide chezi wash DE be se lf s car ‘What Bill washed was s e lf s car.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 The null object sentence in (117), therefore, seems to suggest that like Japanese, the lexical properties o f the dependent term in the Chinese null object construction do not affect the availability o f the sloppy-like readings. However, this is just one side o f the story. Recall that there is distinction between the null object construction with action verbs and that with stative/resultative verbs with regard to their interpretive possibilities (see section 2.4.1.1). They again are distinct with respect to the sensitivity to the lexical properties o f the dependent term. In particular, the latter, unlike the former, is sensitive to the lexical properties o f the dependent term, as indicated below. (118) a. John xihuan John de chezi, Bill ye xihuan [e] like Gen car also like ‘John likes John’s car; Bill likes John’s car.’ ‘*John likes John’s car; Bill likes Bill’s car.’ ‘*John likes John’s car; Bill likes something.’ b. John kanjian-le John de chezi, Bill ye kanjian [e] le see-Asp Gen car also see Asp ‘John saw John’s car; Bill saw John’s car.’ ‘*John saw John’s car; Bill saw Bill’s car.’ ‘*John saw John’s car; Bill saw something.’ The use o f the Name John in place o f ziji ‘s e lf in (118), which involves a stative verb xihuan Tike’ and a resultative verb kanjian ‘see,’ renders the intended sloppy(- like) reading impossible. Note that the reading cannot be inferred either, since the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 so-called unspecified reading is not available, unlike the null object sentence in (117), which involves an action verb xi ‘wash.’ The unavailability o f the sloppy reading is supported by the fact that the sentences in (118) cannot be followed by a remark that specifies the value o f the null object as that o f B ill’s car, as indicated by the unacceptability o f the sentences in (119). (119) a. * Bill xihuan de shi zijide chezi like DE be s e lf s car ‘What Bill likes is se lf s car.’ b. *Bill kanjian de shi zijide chezi see DE be se lf s car ‘What Bill saw was se lf s car.’ In brief, the prediction made by Hoji’s (1998a) analysis o f the sloppy-like readings regarding the relevance o f the lexical properties o f the dependent term is partially bome out in Chinese. It is unclear how the Chinese data involving stative/resultative verbs can be accounted for under Hoji’s suggestion. 2.5.2.8 Relevance o f C-Command In addition to the prediction that the lexical properties o f the dependent term are irrelevant to the availability o f the sloppy-like reading, Hoji’s (1998a) analysis also predicts that the c-command requirement is irrelevant to the availability o f the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 relevant reading, since the reading is based on the pragmatic use o f the null argument itself rather than Formal Dependency that is subject to the c-command condition. The prediction is confirmed, as demonstrated below. (120) (= Hoji 1998a, (49)) a. [mukasi John-o osieta] sensei-ga kare (no kota)-o homete iru years ago Acc taught teacher-Nom him (about)-Acc praising is ‘The teacher [who taught John years ago] is praising him.’ b. [mukasi Bill-o osieta] sensei-mo [e] homete iru years ago Acc taught teacher-also praising is ‘The teacher [who taught Bill years ago] is praising Bill.’ According to Hoji, the sloppy-like reading obtained in (120) is due to coreference between the embedded object Bill and the matrix null object.22 Note in passing that 22 Hoji (1998a, (50)) has shown that John in (120) is not in a position to establish the relevant Formal Dependency with the matrix object kare ‘him,’ as indicated by the failure of bound variable anaphora in (ia), in contrast to (ib). (i) a. * [mukasi subete-no gakuseij-o osieta] sensei-ga soitUj (no koto)-o homete iru years ago all-Gen student-Acc taught teacher-Nom that guy (about)-Acc praising is ‘The teacher [who taught every studentj years ago] is praising him;.’ b. Subete-no gakuseirga [mukasi soitUi-o osieta] sensei-o homete iru all-Gen student-Nom years ao that guy-Acc taught teacher-Acc praising is ‘Every studentj is praising the teacher [who taught hinij years ago].’ Note that the intended dependent term in (120) is kare ‘him.’ Although kare generally does not allow a bound variable construal, it can be a P -occurrence and enter into Formal Dependency, as argued in Hoji (1997a, 1997b). However, since Formal Dependency is not at stake in (120), it is expected on the basis of the discussion in the previous section that the relevant sloppy-like reading continues to obtain even if kare is changed to the Name John, which is the case, as indicated below. (ii) a. [mukasi John-o osieta] sensei-ga John (no kota)-o homete iru years ago Acc taught teacher-Nom (about)-Acc praising is ‘The teacher [who taught John years ago] is praising John.’ b. [mukasi Bill-o osieta] sensei-mo [e] homete iru years ago Acc taught teacher-also praising is ‘The teacher [who taught Bill years ago] is praising John/Bill/someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 Hoji (1998a, 1998d) discusses the availability o f the sloppy identity reading in the following English VP ellipsis sentence, where neither him nor his is c-commanded by John. (The example is based on Fiengo and May (1994, (41a)), cited from Dalrymple et. al. (1991), originally due to Michael Wescoat.) (121) The policeman who arrested John read him his rights, and the one who arrested Bill did too. ‘The policeman who arrested John read John John’s rights, and the one who arrested Bill read Bill Bill’s rights.’ According to Hoji, the absence o f the necessary c-command condition for Formal Dependency in (121) indicates that the relevant reading is an instance o f the sloppy- like reading, which is confirmed by the fact that the relevant reading is still available even when the pronoun him is replaced by the Name John. (122) The policeman who arrested John read John his rights, and the one who arrested Bill did too. ‘The policeman who arrested John read John John’s rights, and the one who arrested Bill read Bill Bill’s rights.’ Now turn to the following null object sentence, where ta ‘him’ is not c- commanded by the embedded John. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 116 (123) a. [Jinianqianjiao-guo John de] laoshi zhengzai chengzan ta years ago teach-Asp DE teacher Prog praise him ‘The teacher [who taught John years ago] is praising him.’ b. [Jinianqianjiao-guo Bill de] laoshi ye zhengzai chengzan [e] years ag teach-Asp DE teacher also Prog praise ‘The teacher [who taught Bill years ago] is praising John.’ ‘The teacher [who taught Bill years ago] is praising someone.’ The fact that the null object sentence in (123) with an action verb chengzan ‘praise,’ can imply the relevant reading seems to suggest that the reading is a sloppy-like one whose availability is not subject to the c-command condition. That John in (123) is not in a position to establish Formal Dependency with ta ‘him’ can be seen from the impossibility o f the bound variable anaphora in (124a), in contrast to (124b). (124) a. * [Jinianqianjiao-guo meige xuesheng, de] laoshi zhengzai years ago teach-Asp every student DE teacher Prog chengzan ta; praise him ‘The teacher [who taught every student, years ago] is praising him;.’ b. Meige xueshengj dou zai chengzan [jinianqianjiao-guo ta; de] every student all Prog praise years ago teach-Asp him DE laoshi teacher ‘Every student; is praising the teacher [who taught him; years ago].’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 117 Note that the sloppy-like reading implied in (123) continues to obtain even when ta ‘him’ is replaced by the Name John, as indicated below. (125) a. [Jinianqianjiao-guo John de] laoshi zhengzai chengzan John years ago teach-Asp DE teacher Prog praise ‘The teacher [who taught John years ago] is praising John.’ b. [Jinianqianjiao-guo Bill de] laoshi ye zhengzai chengzan [e] years ago teach-Asp DE teacher also Prog praise ‘The teacher [who taught Bill years ago] is praising John.’ ‘The teacher [who taught Bill years ago] is praising someone.’ However, consideration o f the null object sentences involving stative/resultative verbs obscures the initial generalization that the sloppy identity reading observed in the Chinese null object construction is not sensitive to the c-command requirement. (126) a. [Jinianqianjiao-guo John de] laoshi hen xihuanta years ago teach-Asp DE teacher very like him ‘The teacher [who taught John years ago] likes him.’ b. [Jinianqianjiao-guo Bill de] laoshi ye hen xihuan [e] years ago teach-Asp DE teacher also very like ‘The teacher [who taught Bill years ago] likes John.’ ‘*The teacher [who taught Bill years ago] likes Bill.’ ‘*The teacher [who taught Bill years ago] likes someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 (127) a. [Jinianqianjiao-guo John de] laoshi zai wanhui kanjian-le ta years ago teach-Asp DE teacher at party see-Asp him ‘The teacher [who taught John years ago] saw him at the party.’ b. [Jinianqianjiao-guo Bill de] laoshi ye kanjian [e] le years ago teach-Asp DE teacher also see Asp ‘The teacher [who taught Bill years ago] saw John.’ ‘*The teacher [who taught Bill years ago] saw Bill.’ ‘*The teacher [who taught Bill years ago] saw someone.’ The structural relation between ta ‘him’ and John in (126) and (127) is identical to that in (123). However, unlike (123), the null object sentences in (126) and (127), which involve a stative verb and a resultative verb, respectively, cannot give rise to the intended sloppy-like reading. Note that the reading cannot be obtained via inference, since the unspecified reading that is available for (123) is unavailable here. Replacement o f ta ‘him’ with the Name John does not affect the result, as indicated by the failure o f the sloppy-like reading in (128) and (129). (128) a. [Jinianqianjiao-guo John de] laoshi hen xihuan John years ago teach-Asp DE teacher very like ‘The teacher [who taught John years ago] likes John.’ b. [Jinianqianjiao-guo Bill de] laoshi ye hen xihuan [e] years ago teach-Asp DE teacher also very like ‘The teacher [who taught Bill years ago] likes John.’ ‘*The teacher [who taught Bill years ago] likes Bill.’ ‘*The teacher [who taught Bill years ago] likes someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 (129) a. [Jinianqianjiao-guo John de] laoshi zai wanhui kanjian-le John years ago teach-Asp DE teacher at party see-Asp ‘The teacher [who taught John years ago] saw John at the party.’ b. [Jinianqianjiao-guo Bill de] laoshi ye kanjian [e] le years ago teach-Asp DE teacher also see Asp ‘The teacher [who taught Bill years ago] saw John.’ ‘*The teacher [who taught Bill years ago] saw Bill.’ ‘*The teacher [who taught Bill years ago] saw someone.’ To sum up, it has been shown that the null object constructions in Chinese and Japanese are distinct with respect to the interpretive possibilities, including Mix readings, sloppy readings and unspecified readings, as well as locality effects and relevance o f the dependent terms and the c-command condition. I therefore conclude that they cannot be treated on a par with each other. 2.5.3 Arguments against Tomioka ’ s Analysis In this section, it will be shown that Tomioka’s pro o f laziness analysis o f the Japanese null object construction as reviewed in section 2.3.4 cannot be extended to Chinese, either. As a result, the Chinese null object construction does not pose a problem on Huang’s (1984) claim that Chinese does not allow an object p ro ,23 23 According to Huang (1984), the ban on an object pro in Chinese is the result of a reductio from the Generalized Control Rule (GCR), which requires it to be coindexed with the closest nominal element, namely, the clause-mate subject, and Principle B, which requires it to be free from the clause-mate subject, thus the contradiction. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 120 2.5.3.1 The Same Difficulties Faced by H oji’ s Suggestion The application o f Tomioka’s pro o f laziness analysis to the null object construction in Chinese will encounter the same difficulties faced by Hoji’s NP suggestion discussed in the above section. That is, like Hoji’s NP analysis, it is unclear how Tomioka’s pro o f laziness analysis can account for (i) the availability o f Mix readings, which are assumed to be based on Formal Dependency rather than context- sensitive interpretation o f the relevant pro o f laziness; (ii) the availability o f sloppy identity readings when coreference between the relevant subject and pro is impossible; (iii) the unavailability o f sloppy identity readings when topicalization o f the element containing the relevant pronoun is involved; (iv) the unavailability o f unspecified readings in the null object constructions with stative or resultative verbs; (v) persistence o f locality effects in the null object constructions with stative or resultative verbs; and (vi) relevance o f the dependent terms and the c-command requirement to the availability o f sloppy identity readings. Note that the availability o f the sloppy readings in the Chinese null object construction involving bound variable anaphora as in (110), repeated here, also poses a problem on Tomioka’s pro o f laziness analysis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 (110) a. Meidui Zhongguo fufu douanwei-le bicide pengyou every Chinese couple all console-Asp each other’s friend ‘Every Chinese couple consoled each other’s friends; i.e. for each Chinese couple, the husband consoled his w ife’s friend and the wife consoled her husband’s friend.’ b. Meidui Meiguo fufu ye dou anwei [e] le every American couple also all console Asp ‘Every American couple consoled each other’s friends; i.e. for each American couple, the husband consoled his w ife’s friend and the wife consoled her husband’s friend.’ According to Tomioka (1997, 1998), the pro in Japanese can enter into bound variable anaphora. (130) (= Tomioka 1998, (2)) Dono gakuseij-mo Dan-ga pro, bu2yokushi-ta to it-ta Which student-even Nom insult-Perf Comp say-Perf ‘Every student said that Dan insulted him;.’ However, the Chinese counterpart o f (130) does not allow the intended bound variable anaphora. (131) * Meige xuesheng, dou shuo Jian qifu [e], le every student all say Jian insult Asp ‘Every student said that Jian insulted (him)j.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 Similarly, the intended bound variable anaphora is not available for the sentence in (132), which is obtained by eliminating the first conjunct and ye ‘also’ o f the second conjunct in (110). (132) *Meidui Meiguo fufu douanwei [e] le every American couple all console Asp ‘Every American couple consoled (the couple).’ ‘Every American couple consoled (each other).’ ‘Every American couple consoled (each other’s friends).’ Given the failure o f bound variable anaphora in cases involving an apparent object gap such as (131) and (132), it is not clear how Tomioka’s pro analysis can capture the sloppy identity readings observed in the Chinese null object construction which involves bound variable anaphora, such as (110). 2.5.3.2 Unavailability o f Sloppy Identity fo r Carrot Sentences Recall that one o f Tomioka’s (1996) arguments against Otani and Whitman’s (1991) VP ellipsis analysis and for his pro o f laziness analysis o f the Japanese null object construction concerns the availability o f the sloppy(-like) reading in the so-called Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 “carrot” sentences such as (31) (repeated here), in which the internal structures o f the two VPs involved are distinct.24 (31) (= Tomioka 1996, (10)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-kara-no self-Gen garden-Gen carrot-Nom McGregor Mr.-ffom-Gen okurimono desi-ta gift be-Perf ‘(Some) carrots from s e lf s garden were the gift (for us) from Mr. McGregor.’ b. Piitaa-mo [e] okutte-kuremasi-ta Peter-also send-give-Perf ‘Peter also sent us (some carrots from se lf s garden).’ N ow turn to the relevant null object sentences in Chinese. (133) a. Ziji huayuanli de huluobo shi Zhangsan song women de liwu self garden Gen carrot be send us DE gift ‘(Some) carrots from se lf s garden were the gift Zhangsan sent us.’ 24 Note that the sloppy(-like) reading in (31) becomes unavailable if the intended semantic antecedent zibun-no hatake-no ninjin-ga ‘carrots from self garden’ is interpreted generically, as in Otani and Whitman’s example (cf. section 2.3.4, (38)). (i) (= Otani and Whitman 1991, (7)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-no daikoobutu desi-ta self-Gen garden-Gen carrot-Nom McGregor Mr.-Gen big favorite be-Perf ‘The carrots from self s garden were Mr. McGregor’s big favorite.’ b. Piitaa-mo [e] daisuki desi-ta Peter-also very fond of be-Perf ‘Peter was also very fond of (the carrots from Mr. McGregor’s garden).’ ‘*Peter was also very fond of (the carrots from self s garden).’ The difference between the existential and generic uses of the relevant antecedent with respect to the availability of the sloppy(-like) reading still awaits explanation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 b. Lisi ye song-le [e] also send-Asp ‘Lisi also sent us (some carrots from Zhangsan’s garden).’ ‘Lisi also sent us (something).’ As shown in (133b), the corresponding carrot sentence in Chinese, which involves an action verb song ‘send’ can imply the intended sloppy(-like) reading. However, the relevant sentences that involve a stative verb such as xihuan ‘like’ and a resultative verb such as kanjian ‘see’ cannot, as in (134a) and (134b), respectively. (134) a. Lisi ye hen xihuan [e] also very like ‘Lisi also liked (some carrots from Zhangsan’s garden).’ ‘*Lisi also liked (some carrots from s e lf s garden).’ ‘*Lisi also liked (something).’ b. Lisi ye kanjian-le [e] also see-Asp ‘Lisi also saw (some carrots from Zhangsan’s garden).’ ‘*Lisi also saw (some carrots from s e lf s garden).’ ‘*Lisi also saw (something).’ In other words, Tomioka’s pro o f laziness analysis cannot accommodate the interpretive difference between the Chinese carrot sentence involving action verbs and that involving stative/resultative verbs regarding the availability o f the sloppy(- like) reading in question. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 2.5.4 Arguments against K im ’ s Analysis It will be shown that like Tomioka’s pro analysis, Kim’s (1995, 1999) vehicle change analysis o f the null object construction cannot be extended to that in Chinese. 2.5.4.1 Interpretive Differences Recall that according to Kim’s (1995, 1999) vehicle change account, which involves reconstructing the indexical structures o f the relevant antecedent NPs, the null object in (50b), repeated here, can give rise to the three readings in (135) (cf. section 2.3.5). (50) a. John-un caki-uy emma-lul salangha-n-ta Topic self-Gen mother-Acc love-Pres-Ind ‘John loves his mother.’ b. Mary-ttohan [e] salangha-n-ta also love-Pres-Ind (135) a. Mary loves John. b. Mary loves his (John’s) mother. c. Mary loves her mother. If Kim’s analysis were extendable to the Chinese null object construction, we would expect it to yield the same interpretive possibilities. However, the expectation is disconfirmed in two aspects. First, the reading in (135a) is not available for a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 Chinese null object sentence involving a stative verb or a resultative verb, as shown in (136) and (137), respectively. (136) a. John xihuan tade mama like his mother ‘John likes his mother.’ b. Mary ye xihuan [e] also like ‘*Mary likes John.’ ‘Mary likes John’s mother.’ ‘Mary likes her mother.’ (137) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mary ye kanjian [e] le also see Asp ‘*Mary saw John.’ ‘Mary saw John’s mother.’ ‘Mary saw her mother.’ Second, it is not clear how the vehicle change analysis can account for the unspecified reading obtained in a null object sentence involving an action verb such as (138), since the reading obviously does not involve reconstruction o f the indexical structure o f a certain antecedent NP. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 (138) a. John zanmei-le tade mama praise-Asp his mother ‘John praised his mother.’ b. Mary ye zanmei [e] le also praise Asp ‘Mary praised John’s mother.’ ‘Mary praised her mother.’ ‘Mary praised someone.’ 2.5.4.2 Unexpected Unavailability o f Sloppy Identity Recall that one o f Kim’s arguments against the VP ellipsis analysis and for his vehicle change analysis o f the null object construction in Korean concerns the availability o f the sloppy reading in the double object construction such as (47) (repeated here), where the two goal arguments involved are different and as a result VP copying is not applicable (cf. section 2.3.5). (47) (= Kim 1995, (7)) a. Peter-nun caki-uy sacin-ul na-eykey/lul poyecwu-ess-ta Top self-Gen picture-Acc I-Dat/Acc show-Past-Ind ‘Peter showed me his picture.’ b. Kulena Ruth-nun kyelko [e] amu-(eykey-)to an poyecwu-ess-ta But Top ever anybody-(Dat-)even not show-Past-Ind ‘But Ruth never showed anybody Peter’s picture.’ ‘But Ruth never showed anybody her picture.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 Similarly, the Chinese counterpart o f the Korean double object sentence in (47) yields the relevant sloppy reading as well, as indicated by the availability o f the sloppy reading in (139). (139) a. Johnsong-le Mary tade shu give-Asp his book ‘John gave Mary his book.’ b. Bill ye song-le Susan [e] also give-Asp ‘Bill also gave Susan John’s book.’ ‘Bill also gave Susan his (Bill’s) book.’ However, this is just one side o f the story. The sloppy reading in question is unexpectedly impossible for the double object sentences such as (140), which involves two different direct arguments, namely, kapian ‘card’ and hua ‘flower.’ (140) a. Johnsong-le tade laoshi kapian give-Asp his teacher card ‘John gave his teacher a card.’ b. Bill ye song-le [e] hua also give-Asp flower ‘Bill also gave John’s teacher flowers.’ ‘*Bill also gave his (Bill’s) teacher flowers.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 If Kim’s vehicle change analysis were tenable in Chinese, we would expect the intended sloppy reading to be available for the null object sentence in (140). This, however, is not the case. The contrast between (139) and (140) with respect to the availability o f the sloppy reading will be discussed in chapter 3. Kim’s another argument against the VP ellipsis analysis and for his vehicle change analysis o f the Korean null object construction concerns the availability o f the sloppy reading in (48) (repeated here), which, with the two distinct numeral quantifiers, cannot feed VP copying. (48) (= Kim 1995, (8)) a. Alan-un caki-(uy) chinkwu-lul twumyeng-(ul) pul-less-ta Top self-(Gen) friend-Acc two.Cl-(Acc) invite-Past-Ind ‘Alan invited two o f his friends.’ b. Jane-un [e] seymyeng-(ul) pul-less-ta Top three.Cl-(Acc) invite-Past-Ind ‘Jane invited three o f Alan’s friends.’ ‘Jane invited three o f her friends.’ Similarly, the Chinese counterpart o f the null object sentence in (48) also gives rise to the sloppy reading as well as the strict reading, as shown in (141). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 (141) a. Johnqing-le liangge tade pengyou invite-Asp two his friend ‘John invited two o f his friends.’ b. Bill ye qing-le sange [e] also invite-Asp three ‘Bill also invited three o f John’s friends.’ ‘Bill also invited three o f his friends.’ However, the sloppy reading obtained in (141) is not particularly surprising, if we consider (141) as an instance o f N ’-deletion, like its English equivalent in (142), which also yields the relevant sloppy reading. (142) a. Johnj invited two o f hiSj friends. b. Billj invited three [e] (= three o f hisj friends). Therefore, the availability o f the sloppy reading in the Chinese null object sentence involving numeral quantifiers may not be taken to support Kim’s suggestion that the null object construction in Chinese, like that in Korean, is amiable to the vehicle change analysis rather than the VP ellipsis analysis. 2.5.5 Arguments against the Null Topic Operator Analysis Huang (1982, 1984, 1987) claims that the null object in (143) is a variable bound by a null topic operator. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 (143) (= Huang 1984, (19d)) [T o p e; ], Zhangsan shuo Lisi bu renshi [ej say not know ‘ (Hinii), Zhangsan said that Lisi did not know (ej).’ The aspect o f Huang’s analysis that is relevant to the current discussion is whether the analysis can be applied to the null object construction. In other words, can a null object sentence such as (lb), repeated here, be treated on a par with the null topic sentence in (144), which in turn is comparable to (145) with an overt topic? (1) (= Huang 1988a, (21)) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mary ye kanjian [e] le also see Asp ‘Mary also saw (John’s mother).’ ‘Mary also saw (her mother).’ (144) [T o p e j, Mary ye kanjian [e,] (145) a. [John de mama]j 5 Mary; ye kanjian [e]j b. [tade; mamajj, Mary; ye kanjian [e]j Otani and Whitman (1991) argue that the null topic operator analysis cannot be applied to the null object construction in Japanese. It will be shown in the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 subsequent subsections that their arguments also hold for Chinese. Additional argument concerning island conditions will be presented as well. 2.5.5.1 Unexpected Unavailability o f Sloppy Identity One o f Otani and Whitman’s (1991) arguments against the null topic operator analysis o f the null object construction in Japanese concerns that it wrongly predicts the availability o f the sloppy reading in sentences such as (146), based on the acceptable overt topic sentence in (147). (146) (= Otani and Whitman 1991, (7)) a. Zibun-no hatake-no ninjin-ga Magguregaa ozisan-no daikoobutu self-Gen garden-Gen carrot-Nom McGregor Mr.-Gen big favorite desi-ta be-Perf ‘The carrots from s e lf s garden were Mr. McGregor’s big favorite.’ b. Piitaa-mo [e] daisuki desi-ta Peter-also very fond o f be-Perf ‘Peter was also very fond o f (the carrots from Mr. McGregor’s garden).’ ‘*Peter was also very fond o f (the carrots from s e lf s garden).’ (147) (= Otani and Whitman 1991, (8b)) [Zibunrno hatake-no ninjin-ga]j Piitaaj-mo [e]j daisuki desi-ta self-Gen garden-Gen carrot-Nom Peter-also very fond o f be-Perf ‘The carrots from s e lf s garden Peter was also very fond of.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 The corresponding null object sentence in Chinese shows the same pattern, as indicated below. (148) a. Ziji caiyuanli de huluobo shi John de zuiai self garden Gen carrot be Gen favorite ‘The carrots from se lf s garden were John’s favorite.’ b. Bill ye hen xihuan [e] also very like ‘Bill also liked (the carrots from John’s garden).’ ‘*Bill also liked (the carrots from s e lf s garden).’ (149) [T o p Zijk caiyuanli de huluoboj,, Bill, ye hen xihuan [e]j self garden Gen carrot also very like ‘The carrots from s e lf s garden, Bill also liked.’ If the null object construction in Chinese could be analyzed as involving a null topic operator, we would expect the availability o f the sloppy identity reading in examples such as (148), on the basis o f the acceptable overt topic sentence in (149), which is contrary to the fact. 2.5.5.2 Unexpected Locality Effects Another difficulty with the null topic operator analysis o f the null object construction is that it incorrectly predicts the availability o f the nonlocal sloppy identity reading in examples such as (150) and (151), which involve a stative verb and a resultative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 verb, respectively, based on the fact that the overt topic sentences parallel in the relevant aspects to (150b) and (151b) are possible, as shown in (152) and (153), respectively. (See Otani and Whitman (1991, 355) for the relevant Japanese examples.) (150) a. John xihuan zijide laoshi like se lf s teacher ‘John likes s e lf s teacher.’ b. Bill zhidao Mark ye xihuan [e] know also like ‘Bill knows that Mark also likes s e lf s (Mark’s) teacher.’ ‘*Bill knows that Mark also likes s e lf s (Bill’s) teacher.’ (151) a, John kanjian-le zijide laoshi see-Asp se lf s teacher ‘John saw se lf s teacher.’ b. Bill zhidao Mark ye kanjian [e] le know also see Asp ‘Bill knew that Mark also saw se lf s (Mark’s) teacher.’ ‘*Bill knew that Mark also saw se lf s (Bill’s) teacher.’ (152) [T o p Zijidej/j laoshi], Bill, zhidao Mark, ye xihuan [e] se lf s teacher know also like ‘S e lfs teacher, Bill knows that Mark also likes (e).’ (153) [T o p Zijidei/j laoshi], Billj zhidao Markj ye kanjian [e] le se lf s teacher know also see Asp ‘S e lfs teacher, Bill knew that Mark also saw (e).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 The acceptability o f (152) and (153) follows from the fact that ziji ‘s e lf is a long distance reflexive that is not clause-bound. The contrast between the null object sentences in (150b) and (151b) on the one hand and the overt topic sentences in (152) and (153) on the other with respect to the availability o f the relevant readings therefore suggests that the former cannot be treated as analogous to the null topic constructions. 2.5.5.3 No Island Violations Still another argument against the null topic operator analysis o f the null object construction concerns the absence o f island violations. To be specific, it is well known that movement is subject to island conditions, which also holds for Chinese, as illustrated in (154) and (155), where Mary cannot be topicalized out o f a complex NP island or an adjunct island. (154) Complex NP Island *Maryi? wo zhidao [piping tj de Bill haowu huiyi I know criticized DE have no regrets ‘As for Mary, I know that Bill, who criticized (her), has no regrets.’ (155) Adjunct Island *Mary; , wo [yinwei Bill piping-le t,] er shengqi I because criticize-Asp angry ‘As for Mary, I was angry because Bill criticized (her).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 If the null object under discussion were a variable bound by a null topic operator, we would expect it to obey the same island conditions. However, the expectation is not borne out. (156) Complex NP Island a. John xihuan/kanjian-le/piping-le Mary ma? like/see-Asp/criticize-Asp Q ‘Did John like/see/criticize Mary?’ b. Bu, Bill cai shi [xihuan/kanjian/piping [e]] de ren no, be like/see/criticize De person ‘No, Bill was the person who liked/saw/criticized (Mary).’ (157) Adj unct Island a. John xihuan Mary like ‘John likes Mary.’ John de mama [yinwei Bill ye xihuan [e]] er danyou Gen mother because also like worried ‘John’s mother is worried because Bill also likes (Mary).’ b. John kanjian-le xin xiaozhang see-Asp new principal ‘John saw the new principal.’ laoshi [zaiB illye kanjian-le [e] yihou] cai jiao tamen zuoxia teacher also see-Asp after ask them sit down ‘The teacher asked John and Bill to sit down after Bill also saw (the new principal).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 c. John piping-le Mary criticize-Asp ‘John criticized Mary.’ laoshi [zai Bill ye piping-le [e] yihou ] chufa-le tamen teacher also criticize-Asp after punish-Asp them ‘The teacher punished John and Bill after Bill also criticized (Mary).’ In (156), the null object construction involving either stative verbs or action verbs escapes the complex NP island; and in (157), it escapes the adjunct island. In other words, regardless o f the verb types, the null object construction exhibits no island effects. The absence o f island violations thus suggests that the null object construction cannot be analyzed on a par with the null topic construction that involves movement and obeys island conditions. In sum, it has been shown that the Chinese null object construction cannot be treated as involving a variable bound by a null topic operator. In the following subsections, it will be shown that the construction does not involve a null anaphor or a null epithet, either. 2.5.6 Arguments against the Null Anaphor Analysis Xu (1986) proposes a unified account o f all the null arguments found in Chinese, according to which, the null arguments are what he refers to as “free empty categories” that can be null anaphors as well as null pronominals. The aspect o f Xu’s Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 discussion that is relevant here is that he claims that the null object in sentences such as (158B) is a null anaphor, not a null pronominal, based on the contrast between (158C) and (158D). (158) (= Xu 1986, (50)) A: Shei piping-guo ziji? who criticize-Asp self ‘Who criticized himself?’ B: John piping-guo [e] criticize-Asp ‘John criticized (himself).’ C: John piping-guo ziji criticize-Asp self ‘John criticized (himself).’ D: *John piping-guo ta criticize-Asp him ‘John criticized him.’ As indicated, only (158B) and (158C) can serve as answers to the yes-no question in (158A), but not (158D), where coreference between the pronoun in the object position and the clause-mate subject is prohibited. The contrast between (158C) and (158D) has led Xu to conclude that the null object in (158B) is a null anaphor, but not a null pronominal. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 Xu also argues that the null object sentence in the above question-answer pair cannot be analyzed as an instance o f VP ellipsis, based on the observation that the verb in the answer can be distinct from that in the question, as shown below. (159) (= Xu 1986, (52)) A: Renren dou biaoyang ziji ma? everyone all praise self Q ‘Did everyone praised himself?’ B: Bu, John piping-le [e] no criticize-Asp ‘No, John criticized (himself).’ N ow the question arises: can all o f the relevant null object sentences involve a null anaphor? It will be shown in the following subsections that the answer is no. 2.5.6.1 Unexpected Availability o f Readings Other than the Anaphoric Reading First consider the following null object construction, which also involves an anaphor. (160) a. John piping-le ziji criticize-Asp self ‘John criticized himself.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 b. Mary ye piping [e] le also criticize Asp ‘Mary also criticized (John).’ ‘Mary also criticized (herself).’ ‘Mary also criticized (someone).’ As indicated, the null object sentence in (160b) yields not only the interpretation that Mary criticized herself but also the interpretations that Mary criticized John and that Mary criticized someone. If the null object sentence in question involved a null anaphor, we would expect the former interpretation to be the only one available, which is contrary to the fact. Next consider a null object sentence that involves a nominal phrase containing an anaphor. (161) (= Otani and Whitman 1991, (2)) a. Zhangsan bu xihuan guanyu zijide yaoyan not like about s e lf s rumor ‘Zhangsan does not like rumors about himself.’ b. Mali ye bu xihuan [e] also not like ‘Mali does not like (rumors about John), either.’ ‘Mali does not like (rumors about herself), either.’ The null object in (161b) clearly cannot be an anaphor, since it cannot refer to M ali; instead, it denotes the concept rumors. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5.6.2 Unexpected Locality Effects and Absence o f Blocking Effects 141 Still another piece o f evidence running afoul o f the null anaphor proposal concerns the peculiar characteristics associated with the anaphors in Chinese. It is well known that the anaphoric relation in Chinese that involves the bare reflexive ziji might be nonlocal as in (162) (see Battistella (1989), and Cole, Hermon, and Sung (1990)). (162) Mary; zhidao B ill xihuan zijii/j know like self ‘Mary knows that Bill likes himself/herj.’ In (162), the so-called long-distance reflexive, ziji ‘self,’ can be referentially dependent upon either the embedded subject Bill or the matrix subject Mary. However, this long-distance reflexive binding is limited in such a way that binding by a remote antecedent is possible only when the remote antecedent agrees in /?/i/-features (person and number) with the antecedent closer to the reflexive (i.e., the local subject); otherwise, nonlocal binding is blocked, as illustrated in (163) and (164) (see Wang and Stillings (1984), Huang (1985), Battistella and Xu (1986), and Huang and Tang (1991)). (163) Mary; zhidao wo xihuan ziji,i ( j know I like self ‘Mary, knows that I like myself/*her.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 (164) W oiwuyiwei Mary tuijian-le zij mistakenly believe recommend-Asp self ‘I mistakenly believed that Mary recommended herself/*me.’ In (163) and (164), where the matrix and embedded subjects do not agree in their />/»-features, the reflexive ziji can only be locally bound. N ow recall that a null object sentence involving a stative verb or a resultative verb cannot yield the nonlocal sloppy reading, as indicated in (165) and (166), respectively (cf. section 2.4.1.2). (165) a. John xihuan ziji like self ‘John likes himself.’ b. Mary zhidao Bill ye xihuan [e] know also like ‘Mary knows that Bill also likes (himself).’ ‘*Mary knows that Bill also likes her.’ (166) a. Zaijingzi li, John kanjian-le ziji in mirror inside saw-Asp self ‘John saw himself in the mirror.’ b. Mary zhidao Bill ye kanjian [e] le know also see Asp ‘Mary knew that Bill also saw (himself).’ ‘*Mary knew that Bill also saw her.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 If the relevant null object were a null anaphor, we would expect the null object sentences in (165b) and (166b) to be able to yield the nonlocal sloppy reading such that the null object, as a null anaphor, were referentially dependent upon the matrix subject Mary, just like the nonlocal anaphoric relation in (162). However, this is contrary to the fact. On the other hand, it has also been discussed that a null object sentence involving an action verb can give rise to the nonlocal sloppy-like reading, as indicated in (167) (cf. section 2.4.1.2). (167) a. John piping-le ziji criticize-Asp self ‘John criticized himself.’ b. Mary zhidao Bill ye piping [e] le know also criticize Asp ‘Mary knew that Bill also criticized (himself).’ ‘Mary knew that Bill also criticized (someone who could be Mary).’ This kind o f nonlocal sloppy-like reading is still possible in examples where the matrix subject and the embedded subject do not agree with respect to their phi- features, as illustrated below. (168) a. John piping-le ziji criticize-Asp self ‘John criticized himself.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 b. Mary zhidao wo ye piping [e] le know I also criticize Asp ‘Mary knew that I also criticized (myself).’ ‘Mary knew that I also criticized (someone who could be Mary).’ If the null object in (168b) were a null anaphor, the availability o f the nonlocal sloppy-like reading would be unexplained. In brief, it has been shown that Xu’s (1986) null anaphor proposal for sentences such as (158b) cannot be applied to accommodate all the instances o f the null object construction. 2.5.7 Arguments against the Null Epithet Analysis It has been argued in Lasnik (1982, 1986) that overt R-expressions should be further classified into two types: names and anaphoric epithets, based on the observation that anaphoric epithets behave like names in one way and like pronominals in another with respect to Binding Theory, which requires the latter to be free in their governing category (Principle B) and the former to be free throughout (Principle C). Assuming Lasnik’s typology o f NPs, Huang (1988a) suggests that the null object in Chinese can be a null epithet, on the basis o f the fact that the null object shares all the properties o f an epithet: (i) it cannot be A-bound, (ii) it can be A ’-bound, (iii) it need not be A ’- bound, and (iv) it may be coindexed with an argument that does not c-command it, as illustrated in (169), (170), (171), and (172), respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (169) a. (= Huang 1984, (19d)) *Zhangsanj shuo [Lisij bu renshi [e]^-] say not know ‘*Zhangsani said that Lisij did not know (him^).’ b. (= Huang 1988, (14a)) *Zhangsani yiwei [zhege bendanj hen youqian] think this idiot very rich Zhangsanj thought this idioti was very rich.’ (170) (Adapted from Huang 1988a, (14b)) Zhangsanj, wo bu xihuan [e]i/zhege wangbadanj I not like this bastard ‘Zhangsanj, I don’t like (himj)/this bastardj.’ (171) a. (= Huang 1984, (7)) Zhangsan kanjian Lisi le ma? see Asp Q ‘Did Zhangsan see Lisi?’ Ta kanjian [e] le he see Asp ‘He saw (him).’ b. (= Huang 1988a, (14c)) Wo bu xihuan zhege wangbadan I not like this bastard ‘I don’t like this bastard.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (172) (Adapted from Huang 1988a, (17)) Ruguoni xiang jian Zhangsanj, wo jiu pai ren quzhao if you want see I then send man go look for [e]j/zhe xiaozij this sissy ‘If you want to see Zhangsan,, I will send someone to look for (hinij)/this sissyj.’ The relevant question is: can this null epithet proposal be extended to the null object construction under discussion? Recall that one o f the properties associated with an overt epithet is that it may not be A-bound, as indicated in (173) (cf. (176b)). (173) a. *Billj piping-le zhege xiaozij criticize-Asp this sissy ‘*Billj criticized this sissy,.’ b. * Johnj zhidao Bill, piping-le zhege xiaozij know criticize-Asp this sissy ‘*Johrij knew that Billj criticized this sissy^.’ In (173), the epithet, zhege xiaozi ‘this sissy’ cannot corefer with the subject in the same or higher clauses. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. N ow look at the null object sentences in (174) and (175). 147 (174) a. John piping-le ziji criticize-Asp self ‘John criticized himself.’ b. Bill ye piping [e] le also criticize Asp ‘Bill also criticized (John).’ ‘Bill also criticized (himself).’ ‘Bill also criticized (someone).’ (175) a. John piping-le ziji criticize-Asp self ‘ John criticized himself. ’ b. Mary zhidao Bill ye piping [e] le know also criticize Asp ‘Mary knew that Bill also criticized (John).’ ‘Mary knew that Bill also criticized (himself).’ ‘Mary knew that Bill also criticized (someone who could be Mary).’ In (174b), the null object can corefer with its clause-mate subject. In (175b), the null object can corefer with the subject in the matrix clause as well as the one in the embedded clause. If the null object in question were a null epithet, the contrast between (173) on the one hand and (174) and (175) on the other would be unexpected. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 Chapter 3 The New Proposal of the Null Object Construction in Chinese It has been demonstrated in the previous chapter that the apparent object gap in the Chinese null object construction in question does not correspond to a phonetically unrealized feature bundle or a pro. Nor does it correspond to a variable bound by a null topic operator or a null anaphor or a null epithet. Moreover, the null object is also resistant to Kim’s (1995, 1999) vehicle change (or NP reconstruction) analysis. In other words, the relevant object gap resists being treated merely as a null argument. On the other hand, it has also been argued that the VP ellipsis analysis, as proposed in Huang (1988a, 1988b, 1989, 1991a) and Otani and Whitman (1991), raises its own questions with regard to the raising o f the verb to the Infl node. However, the arguments do not exclude the possibility that the verb raises to a lower position and still creates a null VP. In fact, it will be proposed that the Chinese null object construction involves a null VP under the vP structure. The relevant configuration is obtained via the raising o f V to v, thus giving rise to the partial range o f the properties typical o f VP ellipsis in English, such as strict/sloppy identity, locality effects, and Mix readings, but not the relevant adjunct scope. It will be further suggested that the action verb in the null object construction can be reanalyzed as a verbal noun, thus giving rise to the unspecified reading and overriding locality effects. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 3.1 V-to-v Movement Huang (1994a, 1994b, 1997), followed by Tang (1998), argues that the general process o f V-to-v movement exists in Chinese as it does in other languages, on the basis o f a number o f otherwise unexpected word order facts and apparent mismatches between syntax and semantics in Chinese. Huang’s first argument concerns sentences with expressions o f event quantification, as exemplified in (1). (1) (= Huang 1994a, (10b)) Takan-le san tian(de) shu he read-Asp three day Gen book ‘He read (books) for three days.’ In (1), the duration expression, san tian ‘three days,’ which quantifies over the action, does not occur in a syntactic position to modify the verb; instead, it occurs in construction with the object and in a position typically occupied by a nominal quantifier, as in (2). (2) Ta kan-le wu-ben (de) shu he read-Asp five-CL Gen book ‘He read five books.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 Huang argues that this mismatch is resolved by analyzing the sentence as involving a structure o f gerundive nominalization and a process that raises the verb out o f the VP into the position o f the abstract light verb DO, as illustrated in (3). NP I shu In (3), the VP kan shu ‘read books’ is part o f the gerundive phrase, i.e., the nominal IP, which is a complement o f the light verb and is quantified by the duration phrase san tian ‘three days.’ According to the structure, the sentence means “he did three days o f reading books.” Under this analysis, the duration phrase is appropriately analyzed as a nominal measure in the syntax quantifying over the nominal IP; and since this gerundive phrase it quantifies over denotes an action or event, the sentence is correctly interpreted as involving event quantification. Huang’s another argument in support o f the V-to-v raising concerns the genitive agent construction, as exemplified in (4). (3) n Subj v ’ V-v IP [+N] Spec I’ [+N] QP I[+N] VP Ta kan san tian Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 (4) (= Huang 1994a, (23c)) Ni haohaojiao ni-de yingwen ba you well teach your English Prt ‘You better teach your English well.’ In (4), the object contains a genitive modifier, nide ‘your.’ However, the genitive NP, which is syntactically a modifier o f a concrete noun, is semantically an argument o f a verb. As in the case o f (1), Huang argues that this mismatch disappears by assuming that the sentence in (4) also involves gerundive nominalization and that the genitive NP is an agent argument o f an event nominal whose verbal head moves to the light verb, as illustrated in (5). V-v IP [+N] Subj I’[+N] I [+N] VP tv NP I N i jiao ni-de yingwen The last piece o f Huang’s arguments I will present here concerns the possessive object construction, as exemplified in (6). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 (6) (= Huang 1994a, (37)) Tamen bang-le wo (de) piao they tie-Asp my Gen ticket ‘They kidnapped me.’ In (6), the expression bang piao is a V -0 construction with the idiomatic meaning ‘kidnap.’ The syntax-semantics mismatch arises from the fact that what is semantically the object o f this idiomatic expression, namely, wo ‘m e’ appears in the genitive form in construction with the object o f the V -0 construction. Huang argues that the possessive NP wo ‘m e’ is the outer object o f the idiomatic V -0 construction and occurs in the Spec o f the VP, whose verbal head, again, has raised to the light verb, as illustrated in (7). (7) vP Subj v ’ V-v ^ V P ^ Spec V ’ f r ^ ^ N P Tamen bang wo piao In this connection, it is interesting to note that according to Huang, not only action verbs undergo movement to the light verb; all o f the other types o f verbs also move to the light verb. However, they are different in that action verbs are complements to the light verb akin to DO, whereas the other types o f verbs are Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 embedded under some other light verbs. For example, stative verbs are embedded under BE or HOLD. Huang’s assumption that stative verbs undergo movement to the light verb akin to BE or HOLD is supported by the possibility o f the following sentence, as pointed out by Li (personal communication, 2002). (8) Wo xihuan yibeizi shu I like all one’s life book ‘I like books all my life.’ As in (1), the duration expression in (8), yibeizi ‘all one’s life,’ which quantifies over the state, does not occur in a syntactic position to modify the verb; instead, it occurs in construction with the object and in a position typically occupied by a nominal quantifier. This mismatch is again readily resolved if we analyze the sentence as involving a structure as in (9), where the verb is raised out o f the VP into the position o f the abstract light verb BE (cf. (3)). (9) [„ P Wo [v . [v- v xihuan] [I P yibeizi tv shu]]] Following Huang, I also assume that all verbs raise to the light verb, v, in Chinese. In the following section, I will show how this independently motivated assumption can account for the Chinese null object construction under discussion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.2 Revised VP Ellipsis Analysis of the Chinese Null Object Construction 154 Consider the examples in (10), which involve a stative verb, xihuan ‘like’ and a resultative verb kanjian ‘see,’ and the one in (11), which involves an action verb, piping ‘criticize.’ (10) a. John xihuan tade laoshi, Bill ye xihuan [e] like his teacher also like ‘John likes his teacher, and Bill also likes (John’s teacher).’ ‘John likes his teacher, and Bill also likes (Bill’s teacher).’ b. John kanjian-le tade laoshi, Bill ye kanjian [e] le see-Asp his teacher also see Asp ‘John saw his teacher, and Bill also saw (John’s teacher).’ ‘John saw his teacher, and Bill also saw (Bill’s teacher).’ (11) John piping-le tade laoshi, Bill ye piping [e] le criticize-Asp his teacher also criticize Asp ‘John criticized his teacher, and Bill also criticized (John’s teacher).’ ‘John criticized his teacher, and Bill also criticized (Bill’s teacher).’ ‘John criticized his teacher, and Bill also criticized (someone).’ Given the assumption that all verbs move to the light verb, v, in Chinese, the second conjuncts in (10) as well as (11) will have the following structure in (12). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 (12) vP Subj V-v VP NP As in (12), the null VP is obtained after the raising o f V out o f the VP into v. 3.2.1 Deriving Strict and Sloppy Identity Given the representation in (12), the LF interpretive rules o f Sag (1976) and Williams (1977) can be directly applied to derive the strict and sloppy readings associated with VP ellipsis, as illustrated in (13) for the null object sentence in (10a). (13) a. John [V P [v xihuan] [N P tade laoshi]] Bill ye [V P [v xihuan] [N P e]] V-to-v Raising (at Spell-Out) b. John [v-v xihuan] [V P tv [N p tade laoshi]] Bill ye [V .v xihuan] [V P tv [n p e]] D erived VP Rule (at LF) c. John [v-v xihuan] [V P A , x [x tv [N p tade laoshi]]] Bill ye [V _ v xihuan] [V P Ax [x tv [N p e]]] Pronoun Rule (at LF) d. John [v-v xihuan] [ v p A x [x tv [ n p x-de laoshi]]] Bill ye [V - v xihuan] [ v p A x [x tv [ n p e]]] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 VP Rule (at LF) e. John [ v .v xihuan] [ VPX x [x tv [ Np x-de laoshi]]] Bill ye [v.v xihuan] [V P A,x [x tv [ n p x-de laoshi]]] In (13), application o f the Pronoun Rule gives rise to the sloppy identity reading; otherwise, the strict identity reading is obtained. Note in passing that cases where the verbs involved are distinct still give rise to the strict and sloppy readings, as shown in (14). (14) a. John [V .v cuoguo][V P tv [ n p tade nuupengyou]] miss his girlfriend ‘John missed his girlfriend.’ b. Bill [v-„ jiandao][V p tv [ np e]] see ‘Bill saw (John’s girlfriend).’ ‘Bill saw (Bill’s girlfriend).’ Examples such as (14) can be readily accommodated under the VP ellipsis analysis, if we assume that the traces left behind by the verbs do not carry values and therefore VP copying does not result in conflict. The assumption is supported by the reduplicated fact in Hebrew, as shown in (15), which has been argued in Doron (1998) to involve VP ellipsis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 (15) (= Doron 1998, (20)) a. dani ohev et iSt-o Dani loves Acc wife-his ‘Dani loves his w ife.’ b. aval moSe sone [e] but Moshe hates ‘But Moshe hates (Dani’s w ife).’ ‘But Moshe hates (Moshe’s wife).’ 3.2.2 Deriving Locality Effects In addition to strict/sloppy ambiguity, the property o f locality effects on the sloppy identity reading that is associated with the null object construction involving stative/resultative verbs, as shown in (16), is straightforwardly captured under the current VP ellipsis analysis. (16) a. John xihuan tade laoshi, Bill zhidao Mark ye xihuan [e] like his teacher know also like ‘John likes John’s teacher, and Bill knows that Mark likes John’s teacher.’ ‘John likes John’s teacher, and Bill knows that Mark likes Mark’s teacher.’ ‘*John likes John’s teacher, and Bill knows that Mark likes B ill’s teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 b. John kanjian-le tade laoshi, Bill zhidao Mark ye kanjian [e] le see-Asp his teacher know also see Asp ‘John saw John’s teacher, and Bill knows that Mark saw John’s teacher.’ ‘John saw John’s teacher, and Bill knows that Mark saw Mark’s teacher.’ ‘*John saw John’s teacher, and Bill knows that Mark saw Bill’s teacher.’ Under the current analysis, the availability o f the local sloppy reading and the unavailability o f the nonlocal one can be explained as follows. The relevant LF representation o f (16a) after application o f the Pronoun Rule is as in (17). (17) a. John [V .v xihuan] [V P ^ x [x tv [N p x-de laoshi]]] b. Bill [V-v zhidao] [vp^y [y tv [ cp hdark ye [v-v xihuan] [vp tv [ n p e]]]]] Application o f the VP Rule then gives the representation in (18), which, however, only yields the local sloppy interpretation. (18) Bill [V-v zhidao] [ypA,y [y tv [C P Mark ye [v.v xihuan] [V P ^ x [x tv [N P x-de laoshi]]]]]] In order for the nonlocal sloppy reading to be obtained, the first conjunct o f (16a) should be able to be interpreted as in (19), which is contrary to the fact. (19) Xx [x ^ 0 [C P Mark ye [V -v xihuan] [V P tx ih u a n [N P x-de laoshi]]]] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 Note that even if the representation in (19) were available, application o f the VP Rule to (17b) would result in vacuous copying o f material other than the embedded VP in point, which is empirically and conceptually infelicitous. In brief, under the current analysis, the availability o f the local sloppy reading and the unavailability o f the nonlocal sloppy reading is expected, given that the binder o f the sloppy pronoun is restricted to the subject o f the lambda predicate associated with the null VP. 3.2.3 Surface Anaphora It has been discussed in chapter 2 that according to Hankamer and Sag (1976), VP ellipsis in English is an instance o f surface anaphora. The present proposal whereby the null object construction in Chinese involves VP ellipsis amounts to saying that it is also an instance o f surface anaphora. It is then expected that the Chinese null object construction shares the same properties o f surface anaphora, which we will show is indeed the case. 3.2.3.1 Requirement o f a Linguistic Antecedent As discussed in chapter 2, the null object conjunct in Chinese needs a linguistic antecedent when sloppy identity is at stake, as shown in (20). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 (20) a. [John looks at his date with deep affection.] Mark: Bill ye xihuan [e] also like ‘Bill likes John’s date, too.’ ‘*Bill likes Bill’s date, too.’ b. [John hit his date on the face, and her face swelled up.] Mark: Bill (shangge xingqi) ye dashang [e] le last week also hit-wound Asp ‘Bill wounded John’s date (last week), too.’ ‘*Bill wounded Bill’s date (last week), too.’ This is expected, given that the relevant construction involves VP ellipsis, which is an instance o f surface anaphora. 3.2.3.2 Mix Readings As discussed in chapter 2, Hoji (1997a) argues that Mix readings are available only for surface anaphora such as VP ellipsis in English, but not for deep anaphora such as the null object construction in Japanese. N ow recall that the null object construction in Chinese gives rise to Mix readings, as shown in (21) and (22). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 161 (21) a. John shuo-guo ta xihuan tade laoshi, Bill ye shuo-guo [e] say-Asp he like his teacher also say-Asp ‘John] said he] liked hisx teacher; Bill2 said he] liked hisi teacher.’ ‘Johni said hei liked hisj teacher; Bill2 said he2 liked his2 teacher.’ ‘John] said he, liked his, teacher; Bill2 said he2 liked hiS] teacher.’ (Mix 1) ‘*John] said he, liked hisi teacher; Bill2 said he, liked his2 teacher.’ (Mix 2) b. John shuo-guo ta dashang tade laoshi, Bill ye shuo-guo [e] say-Asp he hit-wound his teacher also say-Asp ‘John, said he, wounded hiS] teacher; Bill2 said he, wounded his, teacher.’ ‘John, said he, wounded his, teacher; Bill2 said he2 wounded his2 teacher.’ ‘John, said he, wounded his, teacher; Bill2 said he2 wounded his, teacher.’ (Mix 1) ‘*John, said he, wounded his, teacher; Bill2 said he, wounded his2 teacher.’ (Mix 2) (22) a. John shuo-guo tade laoshi xihuan ta, Bill ye shuo-guo [e] say-Asp his teacher like him also say-Asp ‘John, said his, teacher liked him,; Bill2 said his, teacher liked him,.’ ‘John, said his, teacher liked him,; Bill2 said his2 teacher liked him2 .’ ‘John, said his, teacher liked him,; Bill2 said his2 teacher liked him,.’ (Mix 1) ‘John, said his, teacher liked him,; Bill2 said his, teacher liked him2 .’ (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 162 b. John shuo-guo tade laoshi dashang ta, Bill ye shuo-guo [e] say-Asp his teacher hit-wound him also say-Asp ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him,.’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him2 .’ ‘John, said his, teacher wounded him,; Bill2 said his2 teacher wounded him,.’ (Mix 1) ‘John, said his, teacher wounded him,; Bill2 said his, teacher wounded him2 .’ (Mix 2) The possibility o f Mix readings for the Chinese null object sentences in (21) and (22) thus confirms the prediction that the relevant construction is an instance o f surface anaphora, which can give rise to Mix readings. 3.2.3.3 Missing Antecedent Phenomenon Still another piece o f data in support o f the claim that the Chinese null object construction in question is an instance o f surface anaphora comes from the so-called “missing antecedent” phenomenon, first observed in Grinder and Postal (1971) for VP ellipsis in English (namely, vP ellipsis under the current framework), as given in (23), taken from their example in (12). (23) a. Harry doesn’t have a wife, but Bill does have a wife, and she is a nag. b. Harry doesn’t have a wife, but Bill does, and she is a nag. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 In (23b), the sentence contains a pronoun, she, whose antecedent is understood as part o f the elided vP. The possibility o f the apparent missing antecedent has been taken by Grinder and Postal as an argument for the transformational nature o f vP ellipsis, which, in Hankamer and Sag’s (1976) term, is an instance o f surface anaphora. Bresnan (1971) further notes that this missing antecedent phenomenon is only possible with anaphora such as vP ellipsis, but not with anaphora such as do it, which is classified as deep anaphora in Hankamer and Sag (1976). (24) * Jack didn’t cut Betty with a knife; Bill did it, and it was rusty. (The underlined it refers to the knife Bill cut Betty with.) That the missing antecedent is possible only with surface anaphora, but not deep anaphora is expected, given the assumption that the former, but not the latter, is fully represented at LF; and as a result, the pronoun does have an antecedent at the level o f interpretation. N ow consider the relevant null object sentence in Chinese. (25) a. John meiyou qu-guo limian you nuuren de jiuba not go-Asp inside have woman DE bar ‘John has never been to a bar with women inside it.’ b. Bill qu-guo [e], hai shuo tamen dou hen piaoliang go-Asp and say they all very pretty ‘Bill has been to (a bar with women inside it), and said they all were pretty.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 164 As indicated above, it is possible for the sentence in (25b) to contain a pronoun, tamen ‘they,’ whose antecedent is understood as part o f the ellipsis. Note that the instance o f the NP nuuren ‘women’ in (25a) cannot serve as antecedent for a definite pronoun, as shown below. (26) *John meiyou qu-guo limian you nuuren de jiuba; tamen dou hen not go-Asp inside have women DE bar they all very piaoliang pretty ‘*.T ohn has never been to a bar with women inside it; and they all were pretty.’ The possibility o f the missing antecedent phenomenon in (25b) is in fact predicted under the present analysis, according to which, the relevant null object sentence involves VP ellipsis, which is an instance o f surface anaphora and is fully represented at LF. Note that if the antecedent utterance in (25a) is replaced by non-verbal pointing, the utterance in (25b) becomes impossible, as shown below. (27) [Pointing to a bar that cannot be seen through, Mark says:] *Bill qu-guo [e], hai shuo tamen dou hen piaoliang go-Asp and say they all very pretty ‘*Bill has been to (the bar), and said they all were pretty.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 The impossibility o f the missing antecedent phenomenon in (27), in contrast to (25), is again expected under the current proposal that the relevant null object sentence involves VP ellipsis, which, as surface anaphora, requires a linguistic antecedent. In brief, the claim that the null object construction in Chinese involves VP ellipsis and thus is an instance o f surface anaphora is confirmed on the basis o f the following observations: (i) it requires a linguistic antecedent; (ii) it gives rise to Mix readings; and (iii) it allows a missing antecedent for a pronoun. 3.2.4 Deriving Confined Adjunct Scope It has been shown in chapter 2 that in the Chinese null object construction, the interpretation o f the gap in the second conjunct does not contain that o f the adjunct in the first conjunct, in contrast to its vP ellipsis counterpart in English. The relevant example is repeated below. (28) a. John qingchude kanjian-le tade mama clearly see-Asp his mother ‘John clearly saw his mother.’ b. Mary ye kanjian [e] le also see Asp ‘Mary also saw (John’s mother).’ ‘Mary also saw (Mary’s mother).’ First consider the position o f the adjunct, qingchude ‘clearly,’ in (28a). Given the assumptions that V in Chinese moves to v rather than Infl and that adjuncts such as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 166 qingchude, which is a manner adverb, do not move, the word order in (28a) suggests that the position o f the manner adverb should be higher than v. Following Tang (1998), I will assume that the adverb is adjoined to vP, as shown below. (29) John [v P qingchude [v P [V .v kanjian] [V P tv [ n p tade mama]]]] Given the above representation, it follows under the current analysis that the manner adverb in the first conjunct is not interpreted as part of the ellipsis in the second, since the VP Rule will copy only the VP part, which does not contain the adverb. 3.2.5 No Island Violations The last property o f the Chinese null object construction we will review here concerns the fact that it does not exhibit island effects. Readers are referred to chapter 2 for relevant examples. This fact is expected under the current analysis that the null object construction involves VP ellipsis, which does not involve wh-movement and thus island conditions are irrelevant. To sum up, it has been shown that the V-to-v analysis o f the Chinese null object construction captures the properties regarding strict/sloppy ambiguity, locality effects, confined adjunct scope, and absence o f island effects. However, availability o f the unspecified reading and absence o f locality effects for the null object Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 construction involving action verbs still need an explanation, which we will turn to in section 3.3. 3.2.6 The Function o f Ye Before closing this section regarding VP ellipsis analysis o f the Chinese null object construction, I would like to address the function o f ye in the construction. (30) John xihuan tade laoshi, Bill *(ye) xihuan [e] like his teacher also like ‘John likes his teacher, and Bill likes (John’s teacher), too.’ ‘John likes his teacher, and Bill likes (Bill’s teacher), too.’ As indicated, the occurrence o f ye is obligatory in a null object sentence such as (30). I suggest that the status o f ye in (30) is equivalent to that o f too in English vP ellipsis such as (31). (31) John hit his classmate, and Bill did, *(too). Kaplan (1984) suggests that too in (31) is required to emphasize that the same thing is predicated about two contrasting items, i.e., John and Bill, by virtue o f its conventional implicature (that what is predicated about the contrasting item in one clause is also predicated about the contrasting item in the other) and its hypothesized discourse function (to emphasize the similarity between contrasting constituents). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 To be specific, the absence o f the second vP in (31) makes the contrasting items that remain relatively prominent. In other words, it increases the prominence o f the contrast between, and the focus on, the differing constituents. According to Kaplan, in the case o f (31), where there is a strong focus on contrasting items, too is needed to counterbalance the heightened effect o f the contrast by indicating that two items, despite their contrast, are significantly similar. 3.3 Nominalization Recall that a Chinese null object sentence involving an action verb yields not only strict/sloppy identity readings but also a third reading, namely, the so-called unspecified reading, as shown in (32). (32) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized John’s teacher.’ b. Bill ye piping [e] le also criticize Asp ‘Bill criticized John’s teacher.’ ‘Bill criticized Bill’s teacher.’ ‘Bill criticized someone.’ The VP ellipsis analysis proposed in the previous section only accounts for the strict/sloppy readings, but not the unspecified reading. How is this unspecified Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 reading obtained? I will propose in the following that an alternative process o f reanalysis that involves verbal nominalization makes it possible. 3.3.1 Verbal Nouns There is a class o f words in Chinese that can function both as verbs and as nouns. The word piping for instance serves as a verb in (33) and as a noun in (34). (33) John piping-le Taiwan de zhengzhi criticize-Asp Gen politics ‘John criticized the politics in Taiwan.’ (34) John (dui Taiwan zhengzhi) de piping jinxingj-le liangge toward politics Gen criticism proceed-Asp two xiaoshi hour ‘John’s criticism (of the politics in Taiwan) went on for two hours.’ The verbal status o f piping in (33) is clearly shown by the aspectual marker le, which attaches to verbs rather than nouns. In (34), the presence o f the genitive marker de between piping and its preceding PP, dui-phrase, indicates that it is a noun. According to Fu (1994), verbal nouns in Chinese, like English derived nominals, are ambiguous between a process and a result reading. Intuitively speaking, the verbal noun piping ‘criticism’ refers to a process in (34) and a result in (35).1 1 It is noted in Grimshaw (1990) that derived nominals in English are largely ambiguous between process and result readings. A derived nominal like examination has a process reading in (ia) and a rsult reading in (ib). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 (35) John (dui Taiwan zhengzhi) de piping chushu-le toward politics Gen criticism publish-Asp ‘John’s criticism (of the politics in Taiwan) got published.’ Further tests are given in Fu (1994) to distinguish process verbal nouns from result ones. One such test is the choice o f classifiers. Process verbal norms only co-occur with process classifiers such as ci ‘time,’ whereas result verbal norms only co-occur with non-process classifiers such as pian ‘article,’ as indicated in (36) and (37), respectively. (36) John (dui Taiwan zhengzhi) de na ci/*pian piping jinxing-le toward politics Gen that Cl criticism proceed-Asp liangge xiaoshi two hour ‘That criticism (of the politics in Taiwan) by John went on for two hours.’ (37) John (dui Taiwan zhengzhi) de na *ci/pian piping chuban-le toward politics Gen that Cl criticism publish-Asp ‘John’s criticism (of the politics in Taiwan) got published.’ Another test is the compatibility with the light verb construction. It is noted in Luu et al. (1981, 279) that only process verbal nouns, but not result ones, can serve as the subject and the object o f the light verb jinxing ‘proceed,’ as indicated in (34) and (38), respectively. (i) a. John’s examination of the patient took a long time, b. The examination was long. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 (38) John jinxing-le (dui Taiwan zhengzhi de) piping proceed-Asp toward politics Gen criticism ‘John did the criticism (of the politics in Taiwan).’ That the subject or the object o f the light verb are process nouns can be seen in (36) and (39), based on the grammaticality o f the process classifier ci ‘time’ and the ungrammatically o f the non-process classifier pian ‘article.’ (39) John (dui Taiwan de zhengzhi) jinxing-le san ci/*pian piping toward Gen politics proceed-Asp three Cl criticism ‘John did the criticism (of the politics in Taiwan) three times.’ Process and result verbal nouns can also be distinguished in terms o f their compatibility with the event expressions. While the process verbal noun in (40) admits the rationale clause weile chufengtou de ‘to show off,’ the result verbal noun in (41) does not. (40) John weile chufengtou de (dui Taiwan zhengzhi de) na ci piping to show off Gen toward politics Gen that Cl criticism jinxing-le liangge xiaoshi proceed-Asp two hour ‘John’s criticism (o f the politics in Taiwan) in order to show off lasted two hours.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 (41) * John weile chufengtou de (dui Taiwan zhengzhi de) na pian Given the above and other syntactic differences between process and result verbal nouns, Fu (1994), following Borer’s (1988, 1991, 1993) Parallel Morphology, suggests that the result verbal nouns are formed in the lexicon and are no different from any other noun as shown in (42), whereas the process verbal nouns are syntactically derived from their verbal counterparts by means o f zero-affixation, as shown in (43). to show off Gen toward politics Gen that Cl piping chuban-le criticism publish-Asp ‘John’s criticism (of the politics in Taiwan) in order to show o ff got published.’ (42) NP Spec N ’ PP N Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note that according to Fu (1994), the verbal head o f the nominalization as in (43) is deprived o f accusative Case assigning ability, given the assumption that accusative Case is licensed (or checked) in the extended projection o f V, namely, vP, which the nominalization does not have. It follows that the object NP, which needs accusative Case, is not licensed in the nominalization, as indicated by the contrast in (44). (44) a. John (dui Taiwan zhengzhi) de na ci piping jinxing-le toward politics Gen that Cl criticism proceed-Asp liangge xiaoshi two hour ‘That criticism (of the politics in Taiwan) by John went on for two hours.’ b. *Johnde na ci piping Taiwan zhengzhi jinxing-le liangge Gen that Cl criticism politic proceed-Asp two xiaoshi hour ‘That criticism (of the politics in Taiwan) by John went on for two hours.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 Following Fu (1994), I will also assume the structures in (42) and (43) for result and process verbal nouns, respectively. However, in contrast to Fu, I will contend that the PP headed by dui in the latter, as well as that in the former, is an adjunct rather than an argument, based on the fact that the <iw/-phrase is optional, not obligatory, as indicated in (34), (36), and (38)-(40). Note in passing that the nominalization under discussion is different from that discussed in Huang (1994a, 1994b, 1997) (cf. section 3.1). The former involves a nominalized verb, which is derived from V-to-N movement and is deprived o f Case- assigning ability, whereas the latter involves a nominalized IP, in which the verb still retains its Case-assigning ability. In other words, the former involves a verbal noun that is analogous to a derived nominal in English such as criticism, whereas the latter involves a gerund such as criticizing. 3.3.2 Deriving the Unspecified Reading The aspect o f the foregoing discussion that is relevant to the null object construction in question is that the action verb involved in a null object sentence such as (32) (repeated here) may be syntactically reanalyzed as a process verbal noun, serving as an argument to the abstract light verb, DO. The relevant structure after the reanalysis process o f nominalization is given in (45). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 (32) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized John’s teacher.’ b. Bill ye piping [e] le also criticize Asp ‘Bill criticized John’s teacher.’ ‘Bill criticized B ill’s teacher.’ ‘Bill criticized someone.’ v NP VP V-N V ’ I Bill DO tv piping That the action verb in (32) may undergo the process o f nominalization as in (45) is supported by the fact that the abstract light verb, DO, can also be lexically realized, as given in (46). (46) Bill ye [v zuo]-le [N P piping] also do-Asp criticism ‘Bill also did the criticism.’ The interpretation o f the nominalized structures in (214) and (215) is then akin to ‘Bill also did the process o f criticism’; and the interpretation o f the logical object o f Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 the verbal noun, which is not licensed in the nominalization, may be supplied by the context o f discourse in a similar fashion to the interpretation o f the logical subject in an English passive sentence such as (47). (47) Mary was criticized (by someone). In brief, the unspecified reading observed in the Chinese null object construction results from a reanalysis process o f nominalization, whereby the logical object is not specified. 3.3.3 Deriving Absence o f Locality Effects The absence o f locality effects in the Chinese null object construction involving action verbs like (48) can also be accounted for under the proposed nominalization reanalysis. (48) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized John’s teacher.’ b. Bill zhidao Mark ye piping [e] le know also criticize Asp ‘Bill knew that Mark criticized John’s teacher.’ ‘Bill knew that Mark criticized Mark’s teacher.’ ‘Bill knew that Mark criticized someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 Under the current proposal, the null object sentence in (48b) can be reanalyzed as involving the nominalization as in (49). (49) Bill zhidao Mark ye [v P [v DO][N P piping]] As discussed above, the nominalized structure in (49) gives rise to the unspecified reading, which could be compatible with the seeming nonlocal sloppy reading that Mark criticized someone who happened to be Bill’s teacher, thus nullifying the locality effects. It has been shown that the null object constructions involving action verbs may be reanalyzed as involving nominalization, thus availability o f the unspecified reading and absence o f locality effects. It will be shown in the following subsections that the null object constructions involving stative and resultative verbs, on the other hand, do not have access to the nominalization structure. 3.3.4 Stative Verbs Cannot Be Nominalized as Process Verbal Nouns It is intuitively clear that a verb that denotes a state is intrinsically incompatible with a process reading. As a result, a stative verb such as xihuan ‘like’ cannot be syntactically reanalyzed as a process verbal noun, in contrast to an action verb such as piping ‘criticize,’ as indicated below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 (50) a. John (dui Taiwan zhengzhi) de na ci piping jinxing-le toward politics Gen that Cl criticism proceed-Asp liangge xiaoshi two hour ‘That criticism (of the politics in Taiwan) by John went on for two hours.’ b. * John (dui M ali)de na ci xihuan jingxing-le liangnian toward Mary Gen that Cl like proceed-Asp two year ‘That liking o f Mary by John went for two years.’ That a stative verb cannot be syntactically reanalyzed as a nominal is further supported by the fact that unlike an action verb (or more accurately, a nominalized action verb), which can serve as the object o f an overtly realized light verb DO, a stative verb, cannot serve as the object o f an overtly realized light verb BE, as indicated by the contrast between (51a) and (51b), respectively. (51) a. Johnzuo-le piping do-Asp criticism ‘John did the criticism.’ b. *John shi xihuan be liking ‘John was the liking.’ Given that stative verbs cannot be syntactically reanalyzed as nominals, it follows that the null object construction involving a stative verb as in (52) has no Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 access to nominalization and as a result, the unspecified reading is not available, in contrast to the null object construction involving an action verb. (52) a. John xihuan tade laoshi like his teacher ‘John likes his teacher.’ b. Mary ye xihuan [e] also like ‘Mary also likes John’s teacher’ ‘Mary also likes Mary’s teacher.’ ‘*Mary also likes someone.’ 3.3.5 Resultative Verbs Cannot Be Nominalized We have seen that only action verbs, but not statives verbs, may be syntactically reanalyzed as nominals; and as a result, only the former, but not the latter, gives rise to the unspecified reading in the null object construction. Also excluded from the nominalization are resultative verbs, which are in the form o f verb-verb compounds. Each verb expresses an event or a state. The event or state expressed by the second verb is usually caused by the event expressed by the first verb. Unlike action verbs, resultative compound verbs can only be used as verbs, not nouns, as indicated by the contrast in (53) and (54). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 (53) a. John kanjian-le Mary see-Asp ‘John saw Mary.’ b. *John(dui Mary)de kanjian toward Gen sight ‘John’s sight o f Mary’ (54) a. John dashang-le Mary hit-wound-Asp ‘John wounded Mary.’ b. *John (dui Mary) de dashang toward Gen wounding ‘John’s wounding o f Mary’ The failure o f resultative verbs to undergo nominalization has been attributed to their intrinsic delimited aspect (see Fu (1994)). An event is said to be delimited if it comes to an end point. It follows that resultative verbs, which express events with results or end points, are delimited in nature. One the other hand, Fu (1994) argues that nominalizations have non-delimited aspect, based on the fact that they are compatible only with dwz-phrases, which are non-delimited, but not with foz-phrases, which are delimited, as indicated by the contrast between (55c) and (56c). In each o f (55) and (56), the verb in (a) takes a bare NP complement while the verb in (b) takes a corresponding PP (<iu/-phrase in (55) and da-phrase in (56)). In (c), we have the nominalized counterpart o f (a) and (b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 (55) (=Fu (1994, 152)) a. Tamen taolun-guo zhe-ge wenti they discuss-Asp this-Cl problem ‘They have discussed this problem.’ b. Tamen dui zhe-ge wenti taolun-guo they toward this-Cl problem discuss-Asp ‘ They have discussed this problem. ’ c. Tamen dui zhe-ge wenti de taolun they toward this-Cl problem Gen discussion ‘their discussion o f this problem’ (56) (=F u (1994, 151)) a. Budui jiefang-le chengshi army liberate-Asp city ‘The army liberated the city.’ b. Budui ba chengshi jiefang-le army BA city liberate-Asp ‘The army liberated the city.’ c. * Budui ba chengshi de jiefang army BA city Gen liberation ‘the army’s liberation o f the city’ That Jwz-phrases are non-delimited can be seen from the fact that they are compatible only with objects that are not affected by the events expressed by the verbs, as indicated by the contrast between (55) and (57) on the one hand and (58) on the other. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 (57) (= F u (1994, 145)) Ta dui zhe-jian shi kaoluu-guo he toward this-Cl matter think-Asp ‘He thought about this matter.’ (58) (=Fu (1994: 146)) *Ta dui xiaoshuo fanyi-guo he toward novel translate-Asp ‘He translated novels once.’ As indicated, dwi-phrases are allowed in (55) and (57), where the objects o f the verbs taolun ‘discuss’ and kaoluu ‘think about’ do not undergo change. When something is being discussed or thought about, it does not have to be different from a previous state. In contrast, t/wz'-phrase is not acceptable in (58), where the object o f the verb fanyi ‘translate’ must have undergone change. According to Tenny (1989), affectedness or undergoing change falls under the delimited aspect, given that the object o f a verb can realize the end point o f an event if it undergoes a change or is affected. That Jwz'-phrases are compatible only with objects that are not affected or do not undergo change, thus, suggests that dm-phrases are associated with non delimited aspect. On the other hand, fta-phrases occur with delimited predicates, unlike dui- phrases. It has been noted that objects o f 6a-phrases must be “disposed o f ’ (cf. Wang 1947) or affected by the events denoted by the verbs. For example, a ba- phrase may not occur with a predicate such as tui Mary ‘push Mary,’ which does not Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 denote the event where the object Mary is affected either by changing her location or by being injured. (59) a. Johntui-le Mary (danshi Mary meiyou dong) push-Asp but not-have move ‘John pushed Mary (but Mary did not move).’ b. *John ba Marytui-le BA push-Asp ‘John pushed Mary.’ In contrast to tui ‘push,’ the resultative compound verb tui-kai ‘push-away’ in (60) implies that the object undergoes a change o f location, and as a result, the ba- phrase can be used. (60) a. John tui-kai-le Mary push-away-Asp ‘John pushed Mary away.’ b. Johnba Mary tui-kai-le BA push-away-Asp ‘John pushed Mary away.’ Cheng (1990) notes that the occurrence o f 6a-phrases is also sensitive to the aspectual marking o f the verbs. In particular, for-phrases are compatible with aspectual markers that imply an end point to the event expressed by the verb, such as the perfective marker le; on the contrary, they are incompatible with aspectual Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. markers that do not provide an end point for the event expressed by the verb, such as the progressive marker zai, as illustrated in (61). (61) a. Johnba yan jie-le BA cigarette quit-Asp ‘John quit smoking.’ b. *Johnzai ba yan jie Prog BA cigarette quit ‘John is quitting smoking.’ Given that 6a-phrases occur only in contexts where the event is delimited, it is concluded that 6a-phrases are associated with delimited aspect, in contrast to dui- phrases. In view o f the fact that nominalizations only admit non-delimited <iw/-phrases, but not delimited 6a-phrases, Fu (1994) concludes that nominalizations are non delimited. In (62), we have the resultative verb da-shang ‘hit-wound.’ (62a) shows that it is compatible with a Z>a-phrase; (62b) shows that it is incompatible with a dui- phrase; and (62c) shows that it cannot be nominalized. (62) a. Johnba Maryda-shang-le BA hit-wound-Asp ‘John wounded Mary.’ b. *John dui Mary da-shang-le toward hit-wound-Asp ‘John wounded Mary.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 c. *John de da-shang Gen hit-wounding ‘John’s wounding’ Given Fu’s conclusion that nominalizations have non-delimited aspect, it follows that resultative verbs, which typically delimit an event, cannot be nominalized. Bear in mind that resultative verbs cannot be nominalized. Let us now turn to the null object construction involving resultative verbs, as shown in (63). (63) a. John kanjian-le tade laoshi, Bill ye kanjian [e] le see-Asp his teacher also see Asp ‘John saw John’s teacher, and Bill saw John’s teacher.’ ‘John saw John’s teacher, and Bill saw Bill’s teacher.’ ‘*John saw John’s teacher, and Bill saw someone.’ b. John dashang-le tade laoshi, Bill ye dashang [e] le wound-Asp his teacher also wound Asp ‘John wounded John’s teacher, and Bill wounded John’s teacher.’ ‘John wounded John’s teacher, and Bill wounded Bill’s teacher.’ ‘*John wounded John’s teacher, and Bill wounded someone.’ In (63), where resultative verbs are used, the unspecified readings are not available. Under the current analysis whereby the unspecified reading arises from reanalysis o f the verb as a nominalization, the lack o f the unspecified reading in (63) is expected because resultative verbs cannot undergo nominalization. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 That resultative verbs cannot be reanalyzed as nominalizations is further supported by the fact that unlike action verbs, they cannot serve as the objects o f the overtly realized light verb, as indicated in (64) (cf. (51)). (64) a. *Bill zuo-le kanjian do-Asp seeing ‘Bill did the seeing.’ b. *Bill zuo-le dashang do-Asp hit-wounding ‘Bill did the wounding.’ To sum up, the current reanalysis hypothesis correctly predicts that only the null object construction with action verbs may be reanalyzed as involving nominalization and gives rise to the observed unspecified reading; on the other hand, those with stative verbs or resultative compound verbs may not be so analyzed and thus the unspecified reading is not available. 3.4 Predictions I have argued in the previous sections that the sloppy reading observed in the Chinese null object construction is due to the properties o f VP ellipsis (under the vP structure) and the unspecified reading noted in the construction with action verbs (but not with stative verbs or resultative compound verbs) arises owing to a possible Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. syntactic reanalysis o f nominalization. The proposed analysis makes a number o f predictions, which I will turn to in the subsequent sections. 3.4.1 Availability o f the Strict Reading When Reflexives Are Involved Williams (1977), among others, notes that sloppy identity is obligatory for anaphors but is optional for pronouns, as illustrated by the contrast between (65) and (66). (65) John likes himself, and Mary does, too. ‘John likes John, and Mary likes Mary.’ ‘*John likes John, and Mary likes John.’ (66) John likes his colleague, and Mary does, too. ‘John likes John’s colleague, and Mary likes Mary’s colleague.’ ‘John likes John’s colleague, and Mary likes John’s colleague.’ Given Sag-Williams’ theory o f VP ellipsis, this contrast is not unexpected. An anaphor, unlike a pronoun, is not independently referential. Thus, the LF representation for a sentence involving an anaphor as in (65) is the one where the anaphor is translated into a variable bound by the subject o f the lambda predicate associated with the VP. N ow consider the following Chinese null object sentence. (67) a. John xihuan ziji like self ‘John likes himself.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 b. Mary ye xihuan [e] also like ‘Mary also likes John.’ ‘Mary also likes Mary.’ The proposed VP ellipsis analysis (under the vP structure) expects that in (67), the sloppy identity reading should be obligatory, just like its English counterpart in (65). Contrary to the expectation, however, the null object conjunct in (67b) yields not only the sloppy identity reading, but also the equally acceptable strict identity reading that Mary likes John. Note that the strict reading in question cannot be attributed to inference from the unspecified reading, since the latter is unavailable in (67), which involves a stative verb xihuan Tike’ and may not be reanalyzed as involving nominalization. Instead, it will be suggested in chapter 5 that the strict reading arises from the fact that the null object construction in (67) is ambiguous between the VP ellipsis structure and the null topic structure. The strict reading is expected to obtain in the latter, with the empty topic having its value set as John, as illustrated in (68). (68) [John];, Mary ye xihuan [t]; also like ‘(John), Mary also likes.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.4.2 Lexical Properties o f the Dependent Expressions 189 Given Sag-Williams’ (1976, 1977) analysis o f the sloppy identity reading in surface anaphora, it is suggested in Lasnik (1976) and Reinhart (1983) that the distribution o f a sloppy identity reading in surface anaphora is constrained in the same way as that o f bound variable anaphora. Hoji (1997a et esq.) argues that bound variable anaphora, as well as sloppy identity, is possible only if there is an asymmetrical relation o f dependency, which he calls Formal Dependency (FD), established between the relevant nominals expressions. He further illustrates that the establishment o f FD is subject to the following three conditions. (69) The three necessary conditions for an FD (A, B), where A and B are in argument positions: a. B is[-ffl]. b. A c-commands B. c. A is not in the local domain o f B. A nominal expression is [+P ], i.e., a ( 3 -occurrence in the terms o f Fiengo and May (1994) if the determination o f its value is dependent upon that o f another. On the other hand, a nominal expression is [-kx ] or ana -occurrence if it is not [-$ ]. Given that the establishment o f FD is necessary for sloppy identity, it is expected that the availability o f a sloppy reading in surface anaphora is also subject to the three conditions in (69). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 It will be shown in the subsequent sections that the proposed VP ellipsis correctly predicts that the availability o f sloppy identity in the Chinese null object construction is constrained by the three conditions in (69). The first prediction is that the sloppy identity reading in the Chinese null object construction is unavailable if the relevant NP is not a p -occurrence, as followed from (69a). This prediction is borne out, as indicated in (70) and (71). (70) a. John xihuan John de laoshi like Gen teacher ‘John likes John’s teacher.’ b. Bill ye xihuan [e] also like ‘Bill also likes John’s teacher.’ ‘*Bill also likes Bill’s teacher.’ (71) a. John kanjian-le John de laoshi see-Asp Gen teacher ‘John saw John’s teacher.’ b. Bill ye kanjian [e] le also saw Asp ‘Bill also saw John’s teacher.’ ‘*Bill also saw Bill’s teacher.’ As indicated, the sloppy identity reading becomes unavailable when the Name John, which is [+a], is in place o f the sloppy pronoun tade ‘his’, which is [+P]. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 Note, however, that despite the use o f the Name John in (72), which involves an action verb piping ‘criticize,’ the intended reading seems to be possible. (72) a. John piping-le John de laoshi criticize-Asp Gen teacher ‘John criticized John’s teacher.’ b. Bill ye piping [e] le also criticize Asp ‘Bill also criticized John’s teacher.’ ‘Bill also criticized someone (who may be Bill’s teacher).’ The possibility o f the intended sloppy-like reading in (72) is in fact predicted under the proposed analysis whereby an action verb such as piping ‘criticize’ in (72) may be reanalyzed as a nominalization and give rise to the unspecified reading. In this case, the sloppy-like reading may arise as an instance o f inference from the unspecified reading. 3.4.3 The C-Command Condition The proposed VP ellipsis analysis also predicts that the familiar c-command requirement is a necessary condition for the sloppy identity reading, as followed from (69b) above. This prediction is borne out as well. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 (73) a. [Shunianqian jiao-guo Johnde laoshi] hen xihuanta years ago teach-Asp Gen teacher very like him ‘The teacher who taught John years ago likes him. b. [Shunianqian jiao-guo B illde laoshi] ye hen xihuan [e] years ago teach-Asp Gen teacher also very like ‘The teacher who taught Bill years ago also likes John.’ ‘*The teacher who taught Bill years ago also likes Bill.’ As indicated, the sloppy reading is unavailable in (73), where the relevant NP John does not c-command the pronoun ta ‘him.’ That John in (73a) is not in a position to c-command the matrix object can be seen from the failure o f the bound variable anaphora in (74a), in contrast to (74b). (74) a. * [Shunianqian jiao-guo meige xueshengi de] laoshi hen xihuan taj years ago teach-Asp every student DE teacher very like him ‘The teacher [who taught every studentj years ago] likes him;.’ b. Meige xueshengi dou xihuan [shunianqian jiao-guo taj de] laoshi every student all like years ago teach-Asp him DE teacher ‘Every studentj likes the teacher [who taught himj years ago].’ Note again that the intended reading seems to be possible in (75), where an action verb piping ‘criticize’ is used in place o f the stative verb xihuan Tike.’ (75) a. [Shunianqianjiao-guo Johnde laoshi] piping-le ta years ago teach-Asp Gen teacher criticize-Asp him ‘The teacher who taught John years ago criticized him.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 b. [Shunianqian jiao-guo B illde laoshi] ye piping [e] le years ago teach-Asp Gen teacher also criticize Asp ‘The teacher who taught Bill years ago also criticized John.’ ‘The teacher who taught Bill years ago also criticized someone (who may be Bill).’ This is again expected under the proposed reanalysis hypothesis, according to which the action verb involved may be reanalyzed as a nominalization and yield the unspecified reading. As a result, the intended sloppy(-like) reading may be inferred even when the two relevant nominal expressions are not in c-command relation. 3.4.4 The Local Disjointness Effects The third prediction is that the sloppy identity reading in the Chinese null object construction is sensitive to the local disjointness condition in (69c). This prediction is also confirmed. Consider the following null object sentences. (76) a. Wo xihuan wode laoshi I like my teacher ‘I like my teacher.’ b. Wo yao ni ye xihuan [e] I want you also like ‘I want you to like (my teacher).’ ‘I want you to like (your teacher).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 (77) a. ?Wo xihuan wo I like me ‘I like me.’ b. Wo yao ni ye xihuan [e] I want you also like ‘I want you to like (me).’ ‘*1 want you to like (you).’ In (76a), wo ‘I’ is not in the local domain o f the possessive wode ‘m y.’ Hence, the relevant FD, i.e., FD (t, wode) with t being the trace o f wo ‘I,’ can be established and the sloppy identity reading is correctly predicted to be available. On the other hand, the first occurrence o f wo ‘I’ in (77a) is indeed in the local domain o f the second occurrence o f wo ‘me.’ Hence, the relevant FD (t, wo) with t being the trace o f wo ‘I,’ cannot be established and the sloppy identity reading is unavailable, as predicted. However, consider the example in (78), which involves an action verb. (78) a. ?Wo tuijian-le wo I recommend-Asp me ‘I recommended m e.’ b. Wo yao ni ye tuijian [e] I want you also recommend ‘I want you to recommend (me).’ ‘I want you to recommend (someone who may be you).’ Like (77a), the first occurrence o f wo ‘I’ in (78a) is in the local domain o f the second occurrence o f wo ‘m e.’ Consequently, the relevant FD (t, wo) with t being the trace Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 o f wo ‘I,’ cannot be established and the sloppy reading is predicted to be unavailable. Nevertheless, as indicated in the translation, the intended reading may be inferred from the unspecified reading that arises owing to the reanalysis o f the action verb as a nominalization. This, however, does not affect the confirmation presented above that the Chinese null object construction, as an instance o f surface anaphora, does not allow a sloppy identity reading in the local context, as indicated by the sharp contrast between (76) and (77) with respect to the availability o f the sloppy reading. 3.4.5 Mix Readings Recall that the fact that the null object construction in Chinese, like VP ellipsis in English, can yield Mix readings has led us to conclude that like the latter, the former is surface anaphora rather than deep anaphora, given that Mix readings are possible only in surface anaphora, but not in deep anaphora. Given Hoji’s (1997a et esq.) claim that the establishment o f FD is a necessary condition for a sloppy reading in surface anaphora, it is expected that the Mix readings observed in the Chinese null object construction are possible only if the three conditions on the establishment o f FD, as given in (69) are satisfied. We will discuss one by one in the following. 3.4.5.1 Mix Readings andp -Occurrences Recall that one o f the necessary conditions for the establishment o f FD and thus for the availability o f a sloppy reading is the use o f a P -occurrence. It is therefore Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 predicted under the proposed analysis that the Mix readings observed in the Chinese null object construction are not allowed if an a -occurrence is used in place o f a p - occurrence for the relevant nominal expressions. The prediction, however, cannot be tested due to the fact that the relevant first conjuncts with a -occurrences in every relevant position are already awkward to start with, as indicated in (79a) and (80a). (79) a. John shuo-guo John xihuan Johnde laoshi say-Asp like Gen teacher ‘*Johni said John! liked John’s! teacher.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said (e).’ (80) a. John shuo-guo John de laoshi xihuan John say-Asp Gen teacher like ‘*John, said John’Si teacher liked Johni.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said (e).’ The only way for (79a) and (80a) to be acceptable is to interpret John as referring to the speaker himself. In other words, the possible readings o f (79a) and (80a) are analogous to those o f (81a) and (81b), respectively, where the first person pronoun wo ‘I, me’ is used. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 (81) a. Wo shuo-guo wo xihuan wode laoshi I say-Asp I like my teacher ‘I said I liked my teacher.’ b. Wo shuo-guo wode laoshi xihuan wo I say-Asp my teacher like me ‘I said my teacher liked m e.’ This usage o f a Name as referring to the speaker him/herself is not uncommon, especially among kids and political figures. That a Name can have a status analogous to the first person pronoun suggests that a Name may be a p -occurrence. This is confirmed by the availability o f Mix readings for (79’) and (80’) when John is interpreted as akin to the first person pronoun wo. (82) and (83) illustrate that Mix readings are possible with the first person pronoun wo. (79’) a. John shuo-guo John xihuan Johnde laoshi say-Asp like Gen teacher ‘John/Ii said John/Ii liked John’s/my! teacher.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said John/Ij liked John’s/myi teacher.’ ‘Bill2 also said John/I2 liked John’s/my2 teacher.’ ‘Bill2 also said John/I2 liked John’s/my! teacher.’ (Mix 1) ‘*Bill2 also said John/h liked John’s/my2 teacher.’ (Mix 2) (80’) a. John shuo-guo John de laoshi xihuan John say-Asp Gen teacher like ‘John/Ii said John’s/myi teacher liked John/mei.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said John’s/my, teacher liked John/me!.’ ‘Bill2 also said John’s/my2 teacher liked John/me2 .’ ‘Bill2 also said John’s/my2 teacher liked John/mei.’ (Mix 1) ‘Bill2 also said John’s/my! teacher liked John/me2 .’ (Mix 2) (82) a. Wo shuo-guo wo xihuan wode laoshi I say-Asp like my teacher ‘Ii said I, liked myi teacher.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said I, liked my, teacher.’ ‘Bill2 also said I2 liked my2 teacher.’ ‘Bill2 also said I2 liked myi teacher.’ (Mix 1) ‘*Bill2 also said h liked my2 teacher.’ (Mix 2) (83) a. Wo shuo-guo wode laoshi xihuan wo I say-Asp my teacher like me ‘Ii said my! teacher liked mei.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said myi teacher liked mei.’ ‘Bill2 also said my2 teacher liked me2 .’ ‘Bill2 also said my2 teacher liked me!.’ (Mix 1) ‘Bill2 also said my, teacher liked me2 .’ (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 3.4.5.2 Mix Readings and the C-Command Condition Another necessary condition for the establishment o f FD, which in turn is necessary for the availability o f Mix readings, is the c-command condition. Given the proposed VP ellipsis analysis, it is predicted that the null object sentences in which the c-command condition is not satisfied will fail to yield Mix readings. This is the case, as indicated in (84) and (85). (84) a. Daizhu John de xiaojing shuo-guo ta dashang-le tade arrest DE campus-security say-Asp he hit-wound-Asp his tongxue classmate ‘The campus security who arrested John! said he! wounded hisi classmate.’ b. Daizhu Bill de xiaojing ye shuo-guo [e] arrest DE campus-security also say-Asp ‘The campus security who arrested Bill2 also said hei wounded hisi classmate.’ ‘The campus security who arrested Bill2 also said he2 wounded his2 classmate.’ ‘*The campus security who arrested Bill2 also said he2 wounded hisi classmate.’ (Mix 1) ‘*The campus security who arrested Bill2 also said hei wounded his2 classmate.’ (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 (85) a. Daizhu Johnde xiaojing shuo-guo tade tongxue da-shang arrest DE campus-security say-Asp his classmate hit-wound ta him ‘The campus security who arrested John) said his, classmate wounded him,.’ b. Daizhu Bill de xiaojing ye shuo-guo [e] arrest DE campus-security also say-Asp ‘The campus security who arrested Bill2 also said hisi classmate wounded himi.’ ‘The campus security who arrested Bill2 also said his2 classmate wounded him2 .’ ‘*The campus security who arrested Bill2 also said his2 classmate wounded him!.’ (Mix 1) ‘*The campus security who arrested Bill2 also said his! classmate wounded him2 .’ (Mix 2) In (84a), John does not c-command either o f the two pronouns ta ‘he’ and tade ‘his’; and (84) yields only the across-the-board strict and sloppy(-like) readings, but not the Mix readings. Likewise, in (85a), where John does not c-command either pronoun, the Mix readings are impossible. That John in (84a) and (85 a) is not in a position to c-command the following pronouns can be seen from the failure o f the bound variable anaphora in (86a), in contrast to (86b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 (86) a. *Daizhu meige xueshengi de xiaojing shuo-guo ta, dashang arrest every student DE campus security say-Asp he hit-wound xiaozhang principal ‘The campus security who arrested every student said that hei wounded the principal.’ b. Meige xueshengi dou shuo-guo taj dashang-le xiaozhang every student all say-Asp he hit-wound-Asp principal ‘Every student said that he, wounded the principal.’ 3.4.5.3 Mix Readings and Local Disjointness Given Hoji’s claim that Mix readings require the establishment o f FD, which in turn is subject to the local disjointness condition in (69c), it is predicted under the current VP ellipsis analysis that the Mix readings observed in the Chinese null object construction will fail to arise, if the antecedent is in the local domain o f its dependent. This prediction, however, cannot be tested, since the relevant first conjunct where the local disjointness condition is not satisfied, is illegitimate to begin with, as indicated in (87). (87) a. Johndaying-guo ta yao haohao zhaogu tadejiaren promise-Asp him to well take care o f his family ‘*Johni promised him! to take good care o f his! family.’ b. Bill ye daying-guo [e] also promise-Asp ‘Bill2 also promised (e).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As expected, if the relevant locality is removed, the Mix 1 reading becomes available, as illustrated in (88). (88) a. John shuo-guo ta hui haohao zhaogu tade jiaren say-Asp he will well take care of his family ‘Johni said that hei would take good care o f hisi family.’ b. Bill ye shuo-guo [e] also say-Asp ‘Bill2 also said that hei would take good care o f his, family.’ ‘Bill2 also said that he2 would take good care o f his2 family.’ ‘Bill2 also said that he2 would take good care o f hisi family.’ (Mix 1) ‘*Bill2 also said that he, would take good care o f his2 family.’ (Mix 2) 3.4.6 Exchange Reading It has been known since the works by Sag (1976) and Williams (1977) that VP ellipsis sentences in English such as (89) do not allow what Hoji (1997a) calls “exchange reading” in (90). (89) John recommended B ill’s student; but Bill didn’t. (90) John recommended B ill’s student; but Bill didn’t recommend John’s student. In contrast, a sentence with deep anaphora it such as (91) allows such a reading. (91) John recommended B ill’s student; but Bill didn’t do it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 Given the proposed VP ellipsis analysis, it is predicted that the Chinese null object construction, like VP ellipsis in English, does not allow the exchange reading. However, this prediction does not seem to be a correct one. (92) a. John tuijian-le B illde xuesheng recommend-Asp Gen student ‘John recommended Bill’s student.’ b. Danshi Bill meiyou tuijian [e] but not-have recommend ‘But Bill didn’t recommend (e).’ The null object conjunct in (92b), where an action verb tuijian ‘recommend’ is used, can imply the intended exchange reading that Bill didn’t recommend John’s student. Nonetheless, recall that an action verb may be reanalyzed as a nominalization and yield the unspecified reading. It is possible that the availability o f the exchange reading in (92) is due to inference from the unspecified reading that arises from nominalization o f the action verb. If this is the case, it is predicted that if we eliminate the possibility o f nominalization by replacing the action verb with a resultative compound verb, the possibility o f the exchange reading will also be eliminated. This is a correct prediction, as indicated by the failure o f the exchange reading in (93). (93) a. John dashang-le Bill de tongshi hit-wound-Asp Gen colleague ‘John wounded B ill’s colleague.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 b. Danshi Bill meiyou dashang [e] but not-have hit-wound ‘But Bill didn’t wound B ill’s colleague.’ ‘*But Bill didn’t wound John’s colleague.’ It has been shown in the previous sections that the proposed VP ellipsis analysis along with the reanalysis hypothesis correctly predicts that sloppy identity readings and Mix readings are subject to the three conditions on the establishment o f FD, namely, p -occurrences, the c-command condition, and local disjointness, and that the exchange reading is not allowed in the Chinese null object construction that does not involve action verbs. 3.5 Interaction between the Null Object Construction and Others In the following subsections, the interaction between the null object construction and other constructions in Chinese will be discussed. 3.5.1 Topicalization Recall that the sloppy reading is unavailable in cases where the phrase containing the relevant pronoun is topicalized, as in (94b) in contrast to its non-topicalized counterpart in (94a). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 (94) a. John hen xihuan wo gei ta de shu, Bill ye hen xihuan [e] very like I give him DE book also very like ‘John likes the book I gave John; Bill also likes the book I gave John.’ ‘John likes the book I gave John; Bill also likes the book I gave Bill.’ b. Wo gei ta de shu, John hen xihuan, Bill ye hen xihuan [e] I give him DE book very like also very like ‘John likes the book I gave John; Bill also likes the book I gave John.’ ‘*John likes the book I gave John; Bill also likes the book I gave Bill.’ Let us see how the current proposal can capture the above fact. Li (1996) proposes that reconstruction in Chinese is a last resort, based on the following observation. (95) a. Meige renf dou hen xihuan ziji; mai de shu every person all very like self buy DE book ‘Everyone likes the book he bought.’ b. Zijii mai de shu, meige renj dou hen xihuan self buy DE book every person all very like ‘The book he bought, everyone likes.’ (96) a. Meige renj dou hen xihuan wo gei taj de shu every person all very like I give him DE book ‘Everyone likes the book I gave him.’ b. *Wo gei taj de shu, meige renj dou hen xihuan I give him DE book every person all very like ‘The book I gave him, everyone likes.’ As in (95a) and (96a), both the reflexive ziji ‘se lf and the pronoun ta ‘him’ can be bound by the quantificational expression, meigeren ‘everyone.’ However, when they Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 are topicalized, only the reflexive, but not the pronoun, can be interpreted as being bound by the quantificational expression, as indicated by the contrast between (95b) and (96b). Li (1996) suggests that the contrast can be straightforwardly accounted for if reconstruction (in Chinese) is taken to be a process that takes place only when it is necessary. In the case o f pronouns, there is no requirement for the reconstruction to take place, since a pronoun does not need to be bound. Now return to the null object sentence in (94b). Given Li’s claim that pronouns do not reconstruct, the impossibility o f the sloppy identity reading in (94b) can be captured straightforwardly under the present VP ellipsis analysis. To be specific, the apparent null object, which in fact is a null VP in disguise, has no substance to copy, since the phrase that contains the relevant pronoun does not reconstruct back to its originated object position, thus unavailability o f the sloppy identity reading. Given that reflexives do reconstruct, unlike pronouns, it is predicted that the null object sentences involving topicalization o f the phrases containing reflexives behave on a par with their non-topicalized counterparts with respect to the availability o f the sloppy identity readings. This prediction is borne out, as indicated in (97). (97) a. John hen xihuan ziji mai de shu, Bill ye hen xihuan [e] very like self buy DE book also very like ‘John likes the book John bought; Bill also likes the book Bill bought.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 b. Ziji mai de shu, John hen xihuan, Bill ye hen xihuan [e] self buy DE book very like also very like ‘John likes the book John bought; Bill also likes the book Bill bought.’ 3.5.2 Double Object Structures Tang (1978) classifies double object structures (DOS) in Chinese into two types, namely, goal DOS and source DOS. Verbs that occur in goal DOS include j i ‘mail,’ song ‘send,’ and so on. For this type o f verbs, gei, which can be a verb ‘give’ and a preposition ‘to,’ is obligatory with verbs such as j i but optional with verbs such as song when the indirect object (i.e., the goal object) immediately follows the verb, as illustrated in (98) and (99), respectively. (98) Johnji * (gei) Mary yifeng qingshu mail to one-Cl love-letter ‘John mailed to Mary a love letter.’ (99) John song (gei) Mary yishu hua give to one-bouquet flower ‘John gave Mary a bouquet o f flowers.’ However, when the direct object immediately follows the verb, gei must occur before the indirect object, as illustrated in (100) and (101). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 208 (100) Johnji yifeng qingshu *(gei)M ary mail one-Cl love-letter to ‘John mailed a love letter to Mary.’ (101) John song yishu hua * (gei) Mary give one-bouquet flower to ‘John gave a bouquet o f flowers to Mary.’ On the other hand, verbs that occur in source DOS include qian ‘owe,’ qiang ‘rob,’ and so on. For this type o f verbs, there is only one ordering possible; that is, the indirect object (i.e., the source object) must immediately follow the verb and furthermore, gei or any other preposition is forbidden, as illustrated in (102). (102) John qian (*gei/cong) Mary yibai kuai qian owe to/from one-hundred Cl money ‘John owed Mary one hundred dollars.’ No other ordering such as the one in which the direct object immediately follows the verb is allowed, as illustrated in (103). (103) * John qian yibai kuai qian (gei/cong) Mary owe one-hundred Cl money to/from ‘John owed Mary one hundred dollars.’ Li and Thompson (1981) correlate the presence o f gei in (98)-(99) and its absence in (102) with its semantic function. As mentioned above, gei can be a verb denoting ‘give,’ as in (104). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 (104) John gei Mary yiben shu give one-Cl book ‘John gave Mary a book.’ Give signals a transaction. It requires a source from which a transaction originates, a goal at which the transaction terminates, and a theme that is transferred. In (104), the subject John is the source; the indirect object Mary is the goal; and yiben shu ‘one book’ is the theme being transferred. Given the meaning o f gei, Li and Thompson contends that it is natural that gei must occur with a verb denoting some sort o f transaction. In other words, according to Li and Thompson, the semantic function o f gei is to mark the goal o f the transaction named by the verb. It follows that the verb j i ‘mail’ in (98) and the verb song ‘give’ in (99) may co-occur with gei, since these verbs involve transactions in which the subject is the source and the indirect object the goal. On the other hand, the verb qian ‘ow e’ in (102) involves deprivation from rather than transacting to the indirect object. In other words, the subject instead o f the indirect object o f the verb qian ‘ow e’ is the goal, and the indirect object, from whom something is taken, is the source. Therefore, the indirect object o f the verb qian ‘ow e’ may not be preceded by gei. In the following subsections, we will discuss the interaction between the two types o f DOSs on the one hand and the null object structure on the other. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 3.5.2.1 Goal DOS: VNP2NP1 Consider the following null object sentence involving the goal DOS in which the indirect object immediately follows the verb and precedes the direct object. (105) a. John song tade jiaolian yishuang qiuxie give his coach one-pair sneakers ‘John gave his coach a pair o f sneakers.’ b. Bill ye song [e] le also give Asp ‘Bill also gave John’s coach a pair o f sneakers.’ ‘Bill also gave his (Bill’s) coach a pair o f sneakers.’ ‘Bill also gave someone something.’ As indicated, the null goal DOS in (105b), where both the indirect object and the direct object are missing, gives rise to the strict reading, the sloppy identity reading, and the unspecified reading. Before giving an account for the above interpretive possibilities under the proposed VP ellipsis analysis, let us first consider the relevant structure for the goal DOS in Chinese. Based on the Case consideration, Li (1990) proposes that the goal DOS such as (105a) has the following structure, in which the indirect object, namely, NP2 is adjoined to the verb (VI), creating a new node, namely, V2. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 (106) VI NP2 n : song tade yishuang jiaolian qiuxie Li’s proposal is supported by the extraction possibilities with respect to the direct object and the indirect object. It is noted in Li (1990) that the direct object (NP1) can be wh-moved, but the indirect object (NP2) cannot, as shown by the contrast between (107) and (108). (107) a. (= Li 1990, 69) Zhei jiushi [[wo song tamen] de shu] this is I give them DEbook ‘This is the book that I gave them.’ b. Naxie shu, wo song tamen le those book I give them Asp ‘Those books, I gave them.’ (108) a. (= Li 1990, 70) *Zhei jiushi [[wo song shu] de nage ren] this is I give book DE that man ‘This is the man whom I gave a book.’ b. *Tamen, wo song naxie shu le them I give those book Asp ‘Them, I gave those books.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 As indicated in (107), the direct object, namely, shu ‘book’ can be relativized and topicalized (instances o f wh-movement). On the other hand, none o f these structures are possible with the indirect object in (108). Given the proposed structure for the goal DOS as in (106), the above extraction possibilities are accounted for in Li (1990) in terms o f the Empty Category Principle (ECP) together with the notion o f exclusion in the definition o f government as in May (1985) and Chomsky (1986). (109) Chomsky’s (1986) definition o f government A excludes B if no segment o f A dominates B. A governs B iff A m-commands B and there is no C, C a barrier for B, such that C excludes A. The aspect o f the definition o f government in (109) that is directly relevant to Li’s account is that VI in (106) does not govern NP2 because VI does not exclude NP2, given that one o f the segments o f V I, namely, V2, dominates NP2. Given the ECP and Chomsky’s definition o f government that incorporates the notion o f exclusion, the extraction possibilities with respect to the goal DOS are explained as follows. Suppose NP1 is wh-moved. The ECP is satisfied because the trace o f NP1 is properly governed by the verb. On the other hand, wh-movement o f NP2 is ruled out by the failure to satisfy proper government. Under the current hypothesis that V in Chinese moves to v, the following structure for the relevant goal DOS is obtained after raising o f V to v. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 213 (110) vP VI V2 NP1 song tvi NP2 a . | tade yishuang jiaolian qiuxie Bear the above structure in mind and return to the null object sentence in (105), repeated here. (105) a. John song tade jiaolian yishuang qiuxie give his coach one-pair sneakers ‘John gave his coach a pair o f sneakers.’ b. Bill ye song [e] le also give Asp ‘Bill also gave John’s coach a pair o f sneakers.’ ‘Bill also gave his (Bill’s) coach a pair o f sneakers.’ ‘Bill also gave someone something.’ Given the proposed VP ellipsis analysis and the structure in (110), we correctly predict the availability o f the strict reading and the sloppy reading, as illustrated below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 214 (111) a. Strict reading for (105b) Bill [v [vi song]][vA x [x [v [tv i][N p 2 tade jiaolian]] give his (= John’s) coach [N P 1 yishuang qiuxie]]] one-pair sneakers ‘Bill gave John’s coach a pair o f sneakers.’ b. Sloppy reading for (105b) Bill [v [vx song]][v p A . x [x [v [tvi][N P 2 x de jiaolian]] give Gen coach [ N p i yishuang qiuxie]]] one-pair sneakers ‘Bill gave Bill’s coach a pair o f sneakers.’ Moreover, the unspecified reading is also available owing to the possible nominalization o f the action verb song ‘give.’ On the other hand, the sloppy reading and the unspecified reading are impossible with the following null object sentence involving the goal DOS, in which the second VP contains a direct object different from that in the antecedent VP. (112) a. John song tade jiaolian yishuang qiuxie give his coach one-pair sneakers ‘John gave his coach a pair o f sneakers.’ b. Bill ye song-le [e] yizhi qiupai also give-Asp one racket ‘Bill also gave John’s coach a racket.’ ‘*Bill also gave Bill’s coach a racket.’ ‘*Bill also gave someone a racket.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 215 The current analysis correctly predicts the unavailability o f the intended sloppy reading in (112), since the second VP with the distinct direct object still inside it cannot be made empty after V raising and as a result, the Sag-William’s VP copying cannot apply and the intended sloppy reading does not arise. Furthermore, that the verb song ‘give’ in the second VP still takes the direct object prevents it from the reanalysis process o f nominalization, given that a nominalized verb cannot take postverbal arguments. Therefore, the unspecified reading is barred. Similarly, the sloppy reading and the unspecified reading are not available for the following null object sentence with the goal DOS, which contains two distinct indirect objects. (113) a. John song Mary tade shu give his book ‘John gave Mary his book.’ b. Bill ye song Susan [e] also give ‘Bill also gave Susan John’s book.’ ‘*Bill also gave Susan Bill’s book.’ ‘*Bill also gave Susan something.’ Like (112), the second VP in (113) with the distinct indirect object still inside it cannot be made empty after V-raising and as a result, the Sag-William’s VP copying cannot apply and the intended sloppy reading does not arise. Furthermore, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 216 unspecified reading is impossible because the relevant verb song ‘give,’ which takes an indirect object, cannot undergo reanalysis o f nominalization. Note that the sloppy reading and the unspecified reading become possible once the indirect object as well as the direct object is null, as exemplified below. (114) a. John song Mary tade shu give his book ‘John gave Mary his book.’ b. Bill ye song [e] le also give Asp ‘Bill also gave Mary John’s book.’ ‘Bill also gave Mary Bill’s book.’ ‘Bill also gave someone something.’ This is not unexpected under the current analysis. The strict and sloppy identity readings are obtained via VP copying, while the unspecified reading arises due to the possible nominalization o f the action verb song ‘give,’ which does not take any arguments. 3.5.2.2 Goal DOS: VNP1 [ v gei]N P 2 Recall that the Chinese goal DOS such as (105a) has an alternative word order, as exemplified in (115). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 217 (105) John song tade jiaolian yishuang qiuxie give his coach one-pair sneakers ‘John gave his coach a pair o f sneakers.’ (115) Johnsong-le yishuang qiuxie gei tade jiaolian give-Asp one-pair sneakers give his coach ‘John gave one pair o f sneakers to his coach.’ Li (1990) demonstrates that gei in (115), which occurs postverbally, is a verb rather than a preposition by means o f the syntactic tests for verbs and prepositions such as the attachment o f aspect markers and function as a simple answer, as illustrated in (116) and (117), respectively. (116) a. Johnmai-le yishu hua buy-Asp one-bouquet flower ‘John bought a bouquet o f flowers.’ b. *Johncong-le huadian mai yishu hua from-Asp flower-shop buy one-bouquet flower ‘John bought a bouquet o f flowers from a flower shop.’ c. John song hua gei-le M aryyihoujiu hui jia give flower give-Asp after then return home ‘After John gave Mary the flowers, he went home.’ (117) a. John mai hua le ma? Mai le buy flower Asp Q buy Asp ‘Did John buy flowers?’ ‘Yes.’ b. John cong huadian mai hua le ma? *Cong (le) from flower-shop buy flower Asp Q from (Asp) ‘Is John buying flowers from the flower shop?’ ‘Yes.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 218 c. John song hua gei M arylem a? Gei le give flower give Asp Q give Asp ‘Who did John give flowers to?’ ‘Yes.’ Each o f (116) and (117) is a set o f three examples. The sentences in (a) indicate that verbs such as mai ‘buy’ can co-occur with aspect markers such as le and serve as a simple answer. On the other hand, the sentences in (b) indicate that prepositions such as cong ‘from’ cannot. Lastly, the sentences in (c) indicate that the postverbal gei patterns with verbs rather than prepositions with respect to the aspect marker and simple answer tests. That the postverbal gei functions like a verb rather than a preposition is further supported by the fact that it can appear independently without a co-occurring verb and be reanalyzed with another verb to form a new compound verb, whereas prepositions such as cong ‘from’ cannot, as indicated by the contrast between (118) and (119). (118) a. John gei Mary yishu hua give one-bouquet flower ‘John gave Mary a bouquet o f flowers.’ b. John mai gei-le Mary yishu hua buy give-Asp one-bouquet flower ‘John bought Mary a bouquet o f flowers. ’ (119) a. * John cong huadian from flower-shop ‘John was from a flower shop.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 219 b. * John mai cong(-le) huadian yishu hua buy from(-Asp) flower-shop one-bouquet flower ‘John bought a bouquet o f flowers from a flower shop.’ Given the verbal status o f the postverbal gei, the goal DOS in (115) is in fact an instance o f the serial verb construction. The so-called serial verb constructions refer to sentences that contain two or more verb phrases juxtaposed without any overt linguistic marker between them. The interpretive relationship between these verb phrases varies according to the verbs involved. For example, in serial verb sentences such as (120a) and (120b), the sequence o f verbs denotes two actions, with the second one depending on the first; in other words, the second action can only be carried out after the completion o f the first. In (120a), the flowers must be bought before the recipient can be given the flowers; in (120b), one has to take the chopsticks in order to get the peanuts. On the other hand, such a dependency relation does not exist in serial verb sentences such as (120c), in which the drinking does not depend on the completion o f the eating, and vice versa. (120) a. John [mai hua] [song Mary] buy flower give ‘John bought the flowers to give to Mary.’ b. John [na kuaizi] [jia huasheng] take chopsticks get peanut ‘John got the peanuts with chopsticks.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 220 c. John meitian [chi hanbao] [he kele] everyday eat hamburger drink kola ‘John eats hamburgers and drinks kola every day.’ Li (1990), in light o f the semantic contrast between (120a)-(120b) on the one hand and (120c) on the other, notes that the linear ordering between the two verb phrases in (120a)-(120b), but not in (120c), must reflect the actual sequence o f events. As the following examples indicate, permutation o f the two verb phrases is impossible with (120a)-(120b), but possible with (120c). (121) a. * John [song Mary] [mai hua] give buy flower ‘John bought the flowers to give to Mary.’ b. *John [jiahuasheng] [na kuaizi] get peanut take chopsticks ‘John got the peanuts with chopsticks.’ c. John meitian [he kele] [chi hanbao] every da drink kola eat hamburger ‘John drinks kola and eats hamburgers every day.’ Although the serial verb sentences in (120a) and (120b) shares the same dependency relation, yet they behave differently with respect to the extraction possibilities. Extraction o f the object out o f the first verb phrase is possible with (120a), but not with (120b), as indicated in (122a) and (122b), respectively. On the other hand, extraction o f the object out o f the second phrase is impossible with (120a), but Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. possible with (120b), as indicated in (123 a) and (123b), respectively. The examples in (122c) and (123c) indicate that none o f the extraction possibilities is possible with the serial verb sentence in (120c). (122) a. Hua, John mai [t] song Mary le flower buy give Asp ‘The flowers, John bought to give to Mary.’ b. * Kuaizi, Johnna [t] jiahuasheng le chopsticks take get peanut Asp ‘Chopsticks, John got the peanuts with.’ c. *Hanbao, John meitian chi [t] he kele hamburger every day eat drink kola ‘Hamburgers, John eats and drinks kola every day.’ (123) a. *Mary, John mai hua song [t] le buy flower give Asp ‘Mary, John bought the flowers to give to.’ b. Huasheng, John na kuaizi jia [t] le peanut take chopsticks get Asp ‘The peanuts, John got with chopsticks.’ c. *Kele, John meitian chi hanbao he [t] kola every day eat hamburger drink ‘Kola, John eats hamburgers and drinks every day.’ In brief, the serial verb construction in (120a) involves a dependency relation between the two actions denoted by the two verb phrases and the linear ordering is restricted. Furthermore, extraction o f the object out o f the first verb phrase is possible, while extraction o f the object out o f the second verb phrase is barred. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 222 Similarly, the serial verb construction in (120b) also involves a dependency relation between the two actions denoted by the two verb phrases and the linear ordering is restricted. However, extraction o f the object out o f the first verb phrase is barred, whereas extraction o f the object out o f the second verb phrase is allowed, in contrast to (120a). Lastly, in (120c), the two actions expressed by the two verb phrases are independent and the linear ordering is free; and unlike (120a) or (120b), none o f the extraction operations is possible. The distinct semantic and syntactic properties o f the serial verb constructions in (120a)-(120c) can be captured if they are assigned different structures, as indicated in (124), (125), and (126), respectively. (124) Spec v VP Spec V ’ V Adjunct John mai hua tv PRO song Mary Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 223 (125) (126) John chi tv hanbao he tv kele Huang (1994b) argues that the second verb phrase in (120a) (i.e., song Mary ‘give to Mary’), which functions as an adjunct purposive, belongs to the oblique category o f the Thematic Hierarchy {Agent > ... > Theme > ... > Oblique} (cf. Larson (1988)) and originates as the complement o f V, leaving the Spec o f the VP available for the object, hua ‘flower,’ in accordance with the principle o f argument projection that preserves the hierarchy, as illustrated in (124). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Spec Adjunct PROi na kuaizi huasheng ConjP 224 Tang (1998), following Travis (1988), further argues that adjuncts like the purposive in (124) are defective and have to be licensed at LF; however, they cannot be licensed by traces, given the intuition that traces do not have enough content to license adjuncts. It follows that the adjunct purposive in (124) must be adjoined to vP at LF without violating adjunct licensing. The relevant structure is given in (127). Note that given the LF structure in (127), PRO in the adjunct purposive is correctly controlled by the subject John, in accordance with the Minimal Distance Principle, which requires PRO to be controlled by the closest potential antecedent c- commanding it. Return to the sentences in (120). Given the structure in (127), extraction o f the object NP Mary out o f the second verb phrase, which is an adjunct purposive, results in violation o f the adjunct island. On the other hand, extraction o f the object NP hua ‘flower’ is allowed. In contrast, given the structure in (125), extraction o f the object NP kuaizi ‘chopsticks’ out o f the first verb phrase, which is also an adjunct clause, is (127) Adjunct vP V Adjunct PRO song Maty mai hua tv tAdjunct Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 225 barred for the same reason, whereas extraction o f the object NP huasheng ‘peanut’ out o f the second verb phrase is allowed. The structure in (126) is a coordinate structure with a null conjunction. Extraction o f the object NP hanbao ‘hamburger’ out o f the first conjunct or extraction o f the object NP kele ‘kola’ out o f the second conjunct violates the Coordinate Structure Constraint (CSC), which excludes wh- movement o f a constituent from one o f the two conjuncts. The aspect o f the above discussion that is relevant here is that the goal DOS in (115), repeated here, patters with the serial verb sentence in (120a) with respect to the dependency relation and the extraction possibilities, as indicated in (128). (115) Johnsong-le yishuang qiuxie gei tade jiaolian give-Asp one-pair sneakers give his coach ‘John gave one pair o f sneakers to his coach.’ (128) a. Nashuang qiuxie, Johnsong-le [t] gei tade jiaolian that-pair sneakers give-Asp give his coach ‘That pair o f sneakers, John gave to his coach.’ b. *Tade jiaolian, John song-le yishuang qiuxie gei [t] his coach give-Asp one-pair sneakers give ‘His coach, John gave one pair o f sneakers to.’ As indicated, extraction o f the object NP yishuang qiuxie ‘one pair o f sneakers’ out o f the first verb phrase is possible, while extraction o f the object NP tade jiaolian ‘his coach’ out o f the second verb phrase is not. The extraction possibilities therefore suggest that the goal DOS in point has the structure in (127), not the one in (125) or (126). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. N ow consider the null object examples with the relevant goal DOS in (129) and (130). The former involves two distinct goal arguments, namely, Mary and Susan. (129) a. John song tade shu gei Mary give his book give ‘John gave his book to Mary.’ b. Bill ye song [e] gei Susan also give give ‘Bill also gave John’s book to Susan.’ ‘Bill also gave Bill’s book to Susan.’ (130) a. John song tade shu gei Mary give his book give ‘John gave his book to Mary.’ b. Bill ye song [e] le also give Asp ‘Bill also gave John’s book (to Mary).’ ‘Bill also gave Bill’s book (to Mary).’ As indicated, the null object sentences in (129) and (130) yield the strict and sloppy readings. However, the sloppy reading is impossible with either o f the null goal DOSs in (131) and (132). The latter involves two distinct theme arguments, namely, yishuang qiuxie ‘one pair o f sneakers’ and yizhi quipai ‘one racket.’ (131) a. Johnsong-le yishuang qiuxie gei tade jiaolian give-Asp one-pair sneakers give his coach ‘John gave one pair o f sneakers to his coach.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 227 b. Bill ye song [e] le also give Asp ‘Bill also gave one pair o f sneakers to John’s coach.’ ‘*Bill also gave one pair o f sneakers to Bill’s coach.’ (132) a. Johnsong-le yishuang qiuxie gei tade jiaolian give-Asp one-pair sneakers give his coach ‘John gave one pair o f sneakers to his coach.’ b. Bill ye song-le yizhi qiupai [e] also give-Asp one racket ‘Bill also gave a racket to John’s coach.’ ‘*Bill also gave a racket to Bill’s coach.’ The contrast between (129)-(130) on the one hand and (131)-(132) on the other with respect to the availability o f the sloppy reading is in fact expected under the proposed VP ellipsis analysis. According to the above discussion, the goal DOS in (129a) and (130a) has the same status as the serial verb construction in (120a) and thus has an LF structure akin to (127), as shown in (133). (133) vP Adjunct vP V Adjunct PRO gei Mary song tade shu tv tAdjunct Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 228 Given the structure in (133), the sloppy reading in (129b) and (130b) is obtained via VP copying, regardless o f the presence or absence o f the goal argument that is contained in the adjunct clause. Similarly, the goal DOS in (131a) and (132a) has the structure in (134). (134) Adjunct PRO gei tade jiaolian Adjunct PRO gei tade jiaolian song yishuang qiuxie tv tAdjunct Given the structure in (134), it is clear that the intended sloppy reading is impossible for both (131b) and (132b), because the relevant sloppy pronoun tade ‘his’ does not occur within the VP, but inside o f the adjunct clause, which has been raised out o f the VP. In brief, it has been shown that the complicated interpretive possibilities o f the null object construction involving the goal DOS, which has the structure o f the serial verb construction, are fully captured under the proposed VP ellipsis analysis. In the next section, we will discuss the other type o f the DOS, i.e., the source DOS. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 229 3.5.2J Source DOS: VNP2 NP1 Unlike the goal DOS such as (135a), which involves a goal object (NP2) and a theme object (NP1), the source DOS such as (135b) involves a source object (NP2) immediately followed by a theme object (NP1). (135) a. John song [N P 2 Mary][N P 1 hua] give flower ‘John gave Mary flowers.’ b. John qiang [N P 2 Mary][N P 1 shubiao] rob school-bag ‘John robbed Mary o f her school bag.’ Li (1990) notes that the theme object (NP1) in the source DOS, in contrast to the goal DOS, cannot be wh-moved. (136) a. [N P 1 Hua], Johnsong-le [N P 2Mary] flower give-Asp ‘The flowers, John gave Mary.’ b. *[N P i Shubao], John qiang-le [N P 2 Mary] school-bag rob-Asp ‘The school bag, John robbed Mary of.’ Li (1990) further notes that the source DOS exhibits great similarity to the part- whole construction in semantic and syntactic behavior, as exemplified below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 230 (137) a. Tufei sha-le [N P 2 ta](de) [N P 1 fuqin] bandit kill-Asp he Gen father ‘The bandit killed his father.’ b. [N P 2 Ta] bei tufei sha-le [N P 1 fuqin] he BEI bandit kill-Asp father ‘He was affected by the bandit’s killing his father.’ c. Tufei ba [N P 2 ta] sha-le [N P i fuqin] bandit BA him kill-Asp father ‘The bandit disposed o f him by killing his father.’ (138) a. Johnqiang-le [N P 2 Mary] (de) [N P 1 shubao] rob-Asp Gen school-bag ‘John robbed Mary o f her school bag.’ b. [N P 2 Mary] bei John qiang-le [N P t shubao] BEI rob-Asp school-bag ‘Mary was affected by John’s robbing o f her school bag.’ c. John ba [N P 2 Mary] qiang-le [N P i shubao] BA rob-Asp school-bag ‘John disposed o f Mary by robbing o f her school bag.’ The examples in (137) are the so-called part-whole constructions, where NP2, ta ‘him,’ is the possessor o f N P I, fuqin ‘father.’ The possession relation between the two arguments is also indicated by the possibility o f de-insertion. Likewise, NP2, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 231 Mary, in the source DOS like (138) is the possessor of NPI, shubao ‘school bag’ before the action o f robbing; and de-insertion is also possible. In addition, NP2 in both constructions, namely, ta ‘him’ in (137) and Mary in (138) are affected by the respective actions. In (137), for example, ‘he’ is affected by the action o f killing, suffering from the loss o f his father. In (138), ‘Mary’ is affected by the action o f robbing, being the victim o f the robbery. The affectedness is also indicated by the possibility o f passivization and disposition o f NP2, as shown in the (b) and (c) examples, respectively. According to Huang (1994a), the part-whole constructions like (137) is derived from (139), where N P I, fuqin ‘father,’ is an inner object and NP2, ta ‘him,’ is an outer object. (139) vP Spec v v VP NP2 V ’ V NPI Tufei sha ta tv fuqin Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 232 As argued in Thompson (1973), sha ‘kill’ and fuqin ‘father’ together take an object ta ‘him.’ ‘Him’ is affected by the killing o f his father; in other words, NP2 is affected by the action expressed by V and N P I. Given the semantic and syntactic similarities between the source DOS and the part-whole construction, I suggest that the former be assigned the same structure as the latter. (140) vP Spec VP NP2 V ’ V NPI John qiang Mary tv shubao Bear the structure for the source DOS in mind, let us turn to the following null object sentence that involves the source DOS. (141) a. John qiang-le ta tongxue shubao rob-Asp his classmate school-bag ‘John robbed his classmate o f the school bag.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 233 b. Bill ye qiang-le [e] also rob-Asp ‘Bill also robbed John’s classmate o f the school bag.’ ‘Bill also robbed his (Bill’s) classmate o f the school bag.’ The proposed VP ellipsis analysis correctly predicts the availability o f the sloppy reading as well as the strict reading for the null source DOS in (141). However, the sloppy reading is impossible for the following null source DOS, which involves two distinct theme arguments, namely, shubao ‘school bag’ and biandang ‘lunch box.’ (142) a. John qiang-le ta tongxue shubao rob-Asp his classmate school-bag ‘John robbed his classmate o f the school bag.’ b. Bill ye qiang-le [e] biandang also rob-Asp lunch-box ‘Bill also robbed John’s classmate o f the lunch box.’ ‘*Bill also robbed his (Bill’s) classmate o f the lunch box.’ Under the proposed analysis, the unavailability o f the sloppy reading in (142) is expected because the second VP with the distinct theme argument still inside it cannot be made empty after V raising and as a result, the Sag-William’s VP copying cannot apply and the intended sloppy reading is not available. In sum, it has been shown in this section that the wide range o f interpretive possibilities exhibited in the null object constructions involving the goal and source DOS is accommodated under the proposed VP ellipsis analysis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 234 3.5.3 Control Constructions In this section, we will look at the null object construction involving the control structure. Let us first consider the control constructions in (143). The example in (143a) involves a verb o f object control, quan ‘persuade’ and the one in (143b) involves a verb o f subject control, daying ‘promise.’ (143) a. Johnj quan-guo Mark, [C P PRO.i/j tuixiu] persuade-Asp retire ‘John persuaded Mark to retire.’ b. Johni daying-guo Mark, [C P PRO^ tuixiu] promise-Asp retire ‘John promised his coach to retire.’ As the terms suggest, in an object control sentence such as (143a), the controller o f PRO must be the object. On the other hand, in a subject control sentence such as (143b), the controller must be the subject. Following the spirit o f Huang (1994b), I suggest that the relevant control constructions be assigned the following structure. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 235 (144) vP Spec v VP Spec V’ / \ V CP John quan Mark tv PRO tuixiu daying Bear the structure in (144) in mind and first consider the following null object sentence involving a verb o f object control, namely, quan ‘persuade.’ (145) a. John quan tade jiaolian tuixiu persuade his coach retire ‘John persuaded his coach to retire.’ Given the structure in (144), the proposed VP ellipsis analysis correctly predicts the availability o f the strict and sloppy readings. b. Bill ye quan-le [e] also persuade-Asp ‘Bill also persuaded John’s coach to retire.’ ‘Bill also persuaded his (Bill’s) coach to retire.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 236 Likewise, both the strict reading and the sloppy reading are possible for the null object sentence involving a verb o f subject control, namely, claying ‘promise,’ as shown in (146). (146) a. Johndaying tade jiaolian canjia bisai promise his coach participate in game ‘John promised his coach to participate in the game.’ b. Bill ye daying-le [e] also promise-Asp ‘Bill also promised John’s coach to participate in the game.’ ‘Bill also promised his (Bill’s) coach to participate in the game.’ 3.5.4 Postverbal Elements Given the proposed VP ellipsis analysis o f the null object construction in Chinese, we are concerned about the relevance o f the postverbal elements to the interpretive possibilities o f the construction. In the subsequent sections, we will discuss the null object constructions involving quantity adverbial phrases, predicative complements, purposive clauses and postverbal PPs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 237 3.5.4.1 Quantity Adverbial Phrases Quantity adverbial phrases, which are often referred to as duration/frequency phrases, occur after the verb, specifying the duration or frequency o f the activity denoted by the verb, as in (147). (147) a. Johnkan-le Harry Potter san tian read-Asp three day ‘John read Harry Potter for three days.’ b. John kan-le Harry Potter liang ci read-Asp two time ‘John read Harry Potter twice.’ Huang (1994b) argues that a duration/frequency phrase, like an adjunct purposive discussed in section 3.5.2.2, also belongs to the oblique category in the Thematic Hierarchy and originates as the complement o f V, leaving the Spec position o f the VP available for the object. N ow consider the sentences in (148), which are different from the ones in (147) only in the object. (148) a. *John kan-le shu san tian read-Asp book three day ‘John read books for three days.’ b. * John kan-le shu liang ci read-Asp book two time ‘John read books twice.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 238 The grammatical contrast between (147) and (148) has been attributed in Huang (1994b) to the referentiality o f the relevant objects. Huang (1994b) argues that an object like the one in (147) is referential and can occur in [Spec, VP]. On the other hand, an object like the one in (148), which is used as part o f a complex predicate denoting the activity o f book-reading rather than as a true argument, is non- referential and cannot occur in [Spec, VP]. Instead, it must be the first to enter into semantic composition with the verb. The ungrammaticality in (148) therefore results because a non-referential object and a duration/frequency phrase are competing to be the first to combine with the verb. One solution is verb reduplication so that one instance o f the verb is followed by a non-referential object and the other followed by a duration/frequency phrase, as shown in (149). (149) a. Johnkan shu kan-le san tian read book read-Asp three day ‘John read books for three days.’ b. Johnkan shu kan-le liang ci read book read-Asp two time ‘John read books twice.’ It is argued in Tai and Chou (1974) that in the verb-reduplicating sentences such as (149a) and (149b), the first verb phrase, consisting o f the verb and a non-referential object, serves as an adjunct to the second verb phrase, consisting o f the verb and a duration/frequency phrase, on the basis o f the observation that the second verb phrase constitutes the main assertion or the “center o f predication.” This observation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 239 is further strengthened syntactically by the fact that aspect markers such as the perfective marker le may only co-occur with the second verb, but not the first one, as indicated by the grammatical contrast between (149) and (150). (150) a. * John kan-le shu kan san tian read-Asp book read three day ‘John read books for three days.’ b. * John kan-le shu kan liang ci read-Asp book read two time ‘John read books twice.’ The sentences in (147) and (149) thus have the relevant LF structures in (151) and (152), respectively. (151) D/F Spec V ’ D/F san tian/liang ci kan Harry Potter tv tD/F Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 240 (152) D/F V D/F san tian/liang ci kan tv shu kan tv to/F Note that in (151) and (152), the duration/frequency phrase has been adjoined to v to be licensed by the raised verb, as argued in Travis (1988) and Tang (1998) (cf. section 3.5.2.2). Bear these LF structures in mind and let us turn to the null object constructions involving duration/frequency phrases. First consider the relevant sentence with a referential object. (153) a. John kan-le Harry Potter liang ci read-Asp two time ‘John read Harry Potter twice.’ b. Bill ye kan [e] le also read Asp ‘Bill also read (Harry Potter).’ As indicated in the English translation, the interpretation o f the elided conjunct does not contain that o f the frequency phrase that occurs in the first conjunct. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 241 Given the LF representation in (151), it follows under the VP ellipsis analysis that the frequency phrase in the first conjunct is not interpreted as part o f the ellipsis in the second conjunct, since the VP Rule copies only the VP part, which does not contain the frequency phrase. Next consider the following null object sentence with a non-referential object. (154) a. Johnkan shu kan-le liangci read book read-Asp two time ‘John read books twice.’ b. Bill ye kan [e] le also read Asp ‘Bill also read (books).’ As indicated in the English translation, the interpretation o f the elided conjunct in (154b) contains only that o f the non-referential object, but not that o f the frequency phrase. Given the LF representation in (152), the inclusion o f the non-referential object and the exclusion o f the frequency phrase in the ellipsis are explained as follows. Suppose that the VP Rule copies the second or main VP in (152) onto the elided conjunct in (154b). The result remains uninterpretable, since the VP copied contains no substantial content but traces. Yet, there is still another VP available for copying, namely, the first or adjunct VP. Indeed, LF copying o f the first/adjunct VP in (152) yields the interpretation, ‘Bill also read books.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 242 3.5.4.2 Predicative Complements In this section, we will discuss another type o f postverbal elements, namely, the so- called predicative complements, as exemplified in (155) and (156). (155) Johnpao-de hen man run-DE very slow ‘John runs slowly.’ (156) Johnzhou-de tui hen suan walk-DE leg very ache ‘John walked so much that his legs ached.’ The sentences in (155) and (156) each contain two verbal elements, the first o f which (pao ‘run’ and zhou ‘walk’) are the main verbs and the second are called predicative complements-- a depictive complement (kuai ‘fast’) in the case o f (155) and a resultative complement (tui hen suan Tegs ached’) in the case o f (156). As the terms suggest, depictive complements are interpreted as descriptions o f the manners in which the events denoted by the main verbs occur whereas resultative complements express the extents or states resulting from the events described by the main verbs. The status o f de has been analyzed as a complementizer introducing a depictive/resultative complement. Given clear phonological evidence, it has also been assumed that the complementizer obligatorily cliticizes to the preceding verb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 243 Recall that in sentences with duration/frequency phrases, distinction must be drawn between referential and non-referential objects. This kind o f distinction, however, is not observed in sentences with depictive/resultative complements. (157) a. *Johnkan Harry Potter de hen m an/lei read DE very slowly/ tired ‘John read Harry Potter slowly/ so much that he was tired.’ b. *Johnkan shu de hen m an/lei read book DE very slowly/ tired ‘John read books slowly/ so much that he was tired.’ As (157a) shows, a postverbal object is excluded even when it is used referentially. I suggest that these sentences be ruled out independently, due to the failure o f cliticization o f the complementizer de to V at PF because o f the intervening object. On the other hand, both sentences in (157) can be made acceptable by virtue o f verb reduplication, as shown in (158). (158) a. Johnkan Harry Potter kan-de hen m an/lei read read-DE very slowly/ tired ‘John read Harry Potter slowly/ so much that he was tired.’ b. John kan shu kan-de hen man/ lei read book read-DE very slowly/ tired ‘John read books slowly/ so much that he was tired.’ Both sentences in (158), regardless o f the referentiality o f the objects, thus have the relevant LF structure as given in (159). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 244 (159) vP D/R vP Spec vP V’ V VP V VP V NP V D/R de hen man/lei kan tv Harry Potter kan shu tD/R Note that like other adjuncts, the resultative complement is adjoined to v at LF without violating adverb licensing, as argued in Travis (1988) and Tang (1998) (cf. section 3.5.2.2). N ow let us consider the following null object sentence involving a depictive/resultative complement. (160) a. John kan shu kan-de hen m an/lei read book read-DE very slowly/ tired ‘John read books slowly/ so much that he was tired.’ b. Bill ye kan [e] le also read Asp ‘Bill also read (books).’ As indicated, the interpretation o f the ellipsis in the second conjunct does not contain that o f the resultative complement in the first conjunct. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 245 Given the LF structure in (159), it follows under the current VP ellipsis analysis that the resultative complement in the first conjunct is not interpreted as part o f the elided conjunct, since the VP Rule will copy only the VP part, which does not contain the resultative complement. 3.5.4.3 Purposive Clauses In this section, we will discuss still another type o f postverbal elements, namely, the purposive clauses, as exemplified in (161). (161) Johnkan-le yi ben shu [dafa shijian] read-Asp one Cl book while away time ‘John read a book to while away his time.’ As discussed in section 3.5.2.2, it has been assumed in Huang (1994b) that the postverbal adjunct purposive in (161) combines with V, to the exclusion o f the object NP, which, as a true argument, occurs in the Spec position o f the VP. The relevant LF structure after the raising o f the adjunct purposive clause is given in (162). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 246 Note that given the LF structure in (162), PRO in the adjunct purposive is correctly controlled by the subject John, in accordance with the Minimal Distance Principle, which requires PRO to be controlled by the closest potential antecedent c- commanding it. Now consider the following null object sentence involving a purposive clause. (163) a. Johnkan-le yi ben shu [dafa shijian] read-Asp one Cl book while away time ‘John read a book to while away his time.’ b. Bill ye kan [e] le also read Asp ‘Bill also read (a book).’ As indicated, the interpretation o f the ellipsis in the second conjunct does not contain that o f the purposive clause in the first conjunct. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 247 Given the LF structure in (162), it follows under the current VP ellipsis analysis that the purposive clause in the first conjunct is not interpreted as part o f the elided conjunct, since the VP Rule will copy only the VP part, which does not contain the purposive clause. 3.5.4.4 Postverbal PPs In this section, we will discuss postverbal locative phrases, which have the following structure. (164) zai NP(-locative particle) ‘at’ It is noted in Li & Thompson (1981) that most verbs in Chinese take a preverbal locative phrase, as in (165); on the other hand, only four classes o f verbs allow a postverbal locative phrase, as illustrated in (166). (165) John zai huayuan-li chucao at garden-in weed ‘John weeded in the garden.’ (166) a. Verbs o f Displacement John die zai chuang-shang fall at bed-on ‘John fell on the bed.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 248 b. Verbs o f Posture John shui zai di-shang sleep at ground-on ‘John slept on the ground.’ c. Verbs o f (Dis-)Appearing John si zai chuan-shang die at ship-on ‘John died on the ship.’ d. Verbs o f Placement John ba xiezi fang zai guizi-li BA shoes put at cabinet-in ‘John put the shoes in the cabinet.’ The verb o f displacement in (166a), die ‘fall,’ involves the movement or displacement o f the subject from one location to another and the postverbal locative phrase, zai chuang-shang ‘on the bed,’ expresses the new location o f the subject as a result o f the local displacement. In (166b), the verb shui ‘sleep’ depicts the posture o f the subject and the locative phrase, zai di-shang ‘on the ground,’ names the place where the subject assumes the posture. In (166c), the verb si ‘die’ signals the disappearing o f the subject and the locative phrase, zai chuan-shang ‘on the ship,’ names the location o f the disappearance o f the subject. In (166d), the verb fang ‘put’ involves the subject placing the direct object somewhere and the locative phrase, zai guizi-li ‘in the cabinet,’ names the place where the direct object ends up as a result o f the action. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 249 Li (1990) claims that the structure o f the sentences like (166) is not [V [pp P NP]]; instead, the structure is [[v V P] NP], where P (zai ‘at’) is reanalyzed with V and forms a complex verb. Li’s claim that the verbs in sentences like (166) form a unit with the following zai ‘at’ is based on the observation that no aspect markers can occur between the verbs and zai ‘at.’ (167) a. * John die le/guo/zhe zai chuang-shang fall Asp at bed-on ‘John fell on the bed.’ b. *Johnshui le/guo/zhe zai di-shang sleep Asp at ground-on ‘John slept on the ground.’ c. *John si le/guo/zhe zai chuan-shang die Asp at ship-on ‘John died on the ship.’ d. * John ba xiezi fang le/guo/zhe zai guizi-li BA shoes put Asp at cabinet-in ‘John put the shoes in the cabinet.’ Li also notes that no aspect markers can occur after what she considers as the newly formed verbs. (168) a. *John die zai le/guo/zhe chuang-shang fall at Asp bed-on ‘John fell on the bed.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 250 b. * John shui zai le/guo/zhe di-shang sleep at Asp ground-on ‘John slept on the ground.’ c. *John si zai le/guo/zhe chuan-shang die at Asp ship-on ‘John died on the ship.’ d. *John ba xiezi fang zai le/guo/zhe guizi-li BA shoes put at Asp cabinet-in ‘John put the shoes in the cabinet.’ Nevertheless, the unacceptability o f (168) need not be counterexamples to Li’s claim that the verbs and zai form a unit because, as Li notes, for independent reasons, zai cannot take aspect markers. (169) John zai (* 1 e/guo/zhe) chuang-shang at Asp bed-on ‘John was on the bed.’ However, the following fact that the object NP in sentences like (166) cannot be null sheds doubt on Li’s claim that zai ‘at’ is reanalyzed with V. (170) a. John shui zai di-shang sleep at ground-on ‘John slept on the ground.’ b. *Bill ye shui zai [e] also sleep at ‘Bill also slept (on the ground).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 251 The ungrammaticality o f (170) suggests that the preposition zai ‘at’ is not reanalyzed with V to form a complex verb and thus renders preposition stranding impossible. Note, however, that the above conclusion does not preclude the possibility that other prepositions may be reanalyzed with V to form complex verbs. It seems to be the case with the preposition dao ‘to,’ as indicated in (171). (171) a. John fei dao Meiguo le fly to America Asp ‘John has flown to America.’ b. Bill ye fei dao [e] le also fly to Asp ‘Bill has also flown to (America).’ The acceptability o f the null object sentence in (171) suggests that the preposition dao ‘to’ may be reanalyzed with V to form a complex verb, which is raised to v, making VP ellipsis possible. To conclude, it has been shown in this section that the interaction between the null object construction and other structures such as topicalization, double object structures, control structures, and postverbal elements is fully captured by the proposed VP ellipsis analysis, coupled with the reanalysis process o f nominalization. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 252 3.6 Differences between Chinese and Japanese Given the apparent similarities between the null object constructions in Chinese and Japanese as shown in (172) and (173) and our conclusion that the former in fact involves VP ellipsis rather than just an object gap as it appears, a question arises pertinent to the difference between Chinese and Japanese: why is the mechanism of VP ellipsis not available to the null object construction in Japanese? (172) Chinese (Huang 1988a, (21)) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mary ye kanjian [e] le also see Asp ‘Mary also saw (John’s mother).’ ‘Mary also saw (her mother).’ (173) Japanese (Otani & Whitman 1991, (4)) a. John-wa zibun-no tegami-o sute-ta. Nom self-Gen letter-Acc discard-Perf ‘John threw out s e lf s letters.’ b. Mary-mo [e] sute-ta. also discard-Perf ‘Mary also threw out (John’s letters).’ ‘Mary also threw out (se lf s letters).’ Recall that it has been argued in Hoji (1 9 9 3 ,1998a), Tomioka (1996-1999), and Kim (1995, 1999) that the null object construction in Japanese such as the one in (173) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 253 cannot be treated as VP ellipsis; instead, it indeed involves the object gap only, no matter it is a feature bundle as suggested in Hoji (1998a), a pro as assumed in Tomioka (1996-1999), or NP ellipsis as proposed in Kim (1995, 1999). I will suggest in the following that the difference between Chinese and Japanese with respect to the possibility o f VP ellipsis is contingent upon that o f V-to-v raising. In Chinese, the main verb is raised to the light verb, making VP ellipsis possible; on the other hand, in Japanese, the main verb stays in situ and VP ellipsis is not obtained. That the main verb in Chinese is raised to the light verb has been discussed in section 3.1. We will now turn to Japanese. That Japanese does not have V-to-v movement has been argued to reflect in its utilization o f what Kuroda (1988) calls “lower-case” case-marking, as in (174). (174) John-ga Mary-ni hon-o okutta Nom Dat Acc sent ‘John sent a book to Mary.’ Kuroda (1988) points out that syntactic or “upper-case” Case-marking is forced in languages such as English whereas it is not in Japanese. In this aspect, Japanese appeals to the mechanism o f lower-case case-marking to license NPs by assigning appropriate markers such as -ga and -o. To explain the difference between English and Japanese with respect to Case assignment, Takano (1996) further correlates accusative Case assignment with the possibility o f V movement. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 254 (175) (= Takano 1996,(68)) If a language lacks verb raising, it must invoke the particle system for accusative Case. The rationale for (175) is that the checking system for accusative Case is available for free in languages with V-to-v movement and the particle system is not available, assuming that vP is for checking accusative Case. According to Takano, V moves to v in English but it does not move in Japanese. Therefore, Japanese has to utilize the particle system for accusative Case. Assuming along with Takano (1996) that Japanese lacks V-to-v movement, I will suggest that the lack o f VP ellipsis in Japanese, in contrast to Chinese, may correlate with its lack o f what Huang calls a “lexicalized” v (after V-to-v movement) to license the empty VP in accordance with ECP. 3.7 Differences between Chinese and English We have discussed in the previous section that the difference between Chinese and Japanese with respect to the possibility o f VP ellipsis might result from the existence o f V-to-v raising in the former but not in the latter. In this section, we will tackle the difference between Chinese and English in this regard. Both Chinese and English have been assumed to involve V-to-v raising. It has been assumed that V moves to v in English for checking accusative Case. Chinese Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 255 has also been argued to involve V-to-v raising based on the observations o f word order facts and syntax-semantics mismatches (cf. section 3.1). Given that both languages have V-to-v raising, why is it the case that VP ellipsis is allowed in Chinese, but not in English? In other words, why does English lack an equivalent Chinese counterpart such as (176), as indicated by the ungrammaticality o f (177)? (176) a. John xihuan tade laoshi like his teacher ‘John likes his teacher.’ b. Mary ye xihuan [e] also like ‘Mary also likes John’s teacher.’ ‘Mary also likes her teacher.’ (177) a. John likes his teacher, b. *Mary likes [e], too. In the following I wish to suggest a possible syntactic basis for the above contrast. In particular, I will suggest that the difference between Chinese and English with respect to the availability o f VP ellipsis (namely, the so-called null object construction) results from structural differences between the two languages. It has been widely assumed that subjects in English originate in the Spec position o f VP (namely, vP under the current framework) and are then raised to the Spec position o f Infl. Under this view, an English sentence has the structure in (178a) instead o f the structure in (178b) or (178c). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 256 (178) a. [I P NPj [r I [v P tj vP]]] b. * [.p [.-I[v p N P vP]]] c. *[ipN P[rI vP]] On the other hand, Aoun and Li (1989) suggest that subject raising is not available in Chinese; instead, a Chinese subject occurs either in the Spec position o f VP or in the Spec position o f Infl, as supported by the existence o f double subject structures in the language. To be specific, Kuroda (1988) claims that both the Spec position o f Infl and the Spec position o f VP are Case positions, on a basis o f the existence o f double subject structures such as (179) in Japanese. Kitagawa (1986) further claims that both positions are Case and 0 positions. (179) Yamada-sanga go-tyoonanga daigakuni gookakusareta Nom eldest son Nom college Dat passed ‘Yamada-san’s eldest son passed the entrance exam for a college.’ In (179), Yamada-san is what Kitagawa calls “major subject,” followed by a full clause whose subject is go-tyoonan. The nominative marker ga occurs with both subjects. In line with Kuroda (1988) and Kitagawa (1986), Aoun and Li also suggest that both the Spec o f Infl and the Spec o f VP are Case and 0 positions in Chinese, as indicated by the existence o f the double subject structures such as (180) in Chinese. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 257 (180) (= Aoun and Li 1989, (26a)) Zhangsan erzi kaoshang daxue le son pass college Asp ‘Zhangsan’s son passed the entrance exam for a college.’ I will follow Aoun and Li (1989) in assuming that subjects in Chinese occur either in the Spec o f VP (namely, vP under the current framework) or in the Spec o f Infl; that is, the basic structure o f a Chinese sentence will be either (181a) or (181b). (181) a. [I P [r I [vpNP vP]]] b. [I P NP [r I vP]] Bear the structural differences between Chinese and English in mind. I would like to suggest that the availability o f VP ellipsis (namely, the null object construction) correlates with the position o f a subject, as stated in (182). (182) Ellipsis o f a verbal projection is only possible with the subject occurring in the Spec position o f its governor. In other words, in order for VP ellipsis to be possible, the subject must occur in the Spec position o f v, which lexically governs the elided VP after V-to-v raising. On the other hand, in the case o f vP ellipsis, the subject must occur in the Spec position o f Infl, which lexically governs the elided vP due to the occurrence o f an auxiliary, such as c/o-support in English. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Given (182), it follows that VP ellipsis such as (177) is not available in English, since an English subject has been raised out o f vP to the Spec o f Infl. However, vP ellipsis such as (183) is correctly predicted to be available for English. (183) a. John likes his teacher, b. Mary does [e], too. ‘Mary likes John’s teacher.’ ‘ Mary likes her teacher. ’ In contrast, both VP ellipsis such as (176) and vP ellipsis such as (184) (see the discussion in chapter 4) are available for Chinese, since a Chinese subject may occur either in the Spec o f vP or in the Spec o f Infl. (184) a. John xihuan tade laoshi like his teacher ‘John likes his teacher.’ b. Mary ye shi [e] also be ‘Mary also likes John’s teacher.’ ‘Mary also likes her teacher.’ Although the statement in (182) is more like a preliminary generalization that calls for further articulation and deeper explanation, yet it is indirectly supported by the ellipsis facts concerning the double object structure in Chinese such as (180), repeated here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 259 (180) (= Aoun and Li 1989, (26a)) Zhangsan erzi kaoshang daxue le son pass college Asp ‘Zhangsan’s son passed the entrance exam for a college.’ In (180), Zhangsan is a major subject in the Spec position o f Infl, followed by a full clause whose subject, erzi, occurs in the Spec position o f vP, as shown below. Given the structure in (185), it is predicted under (182) that the sentence allows only vP ellipsis, but not VP ellipsis, since the major subject Zhangsan occurs in the Spec position o f Infl, but not in the Spec position o f vP. This prediction is borne out, as indicated by the contrast between (186b) and (186c). (185) IP Spec I’ I vP Spec V-v VP V NP Zhangsan erzi kaoshang tv daxue Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 260 (186) a. Zhangsan erzi kaoshang daxue le son passed college Asp ‘Zhangsan’s son passed the entrance exam for a college.’ b. Lisi ye shi [v P e] also be ‘Lisi’s son also passed the entrance exam for a college.’ c. *Lisi ye erzi kaoshang [V P e] le also son pass Asp ‘Lisi’s son also passed the entrance exam for a college.’ In brief, it has been suggested in this section that the differences between Chinese and English with respect to the availability o f VP ellipsis, namely, the null object construction in the former but not in the latter, may arise from the structural differences in subject positions between the two languages. 3.8 Theoretical Implication In this section, we will address some theoretical implications raised by the current VP ellipsis analysis, coupled with the reanalysis process o f nominalization. First o f all, let us discuss the syntactic status o f the empty category in the null object construction when it is selected in the Numeration. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 261 3.8.1 Syntactic Status o f the Empty Category in the Numeration Recall that the apparent object gap in the Chinese null object construction has been argued to be an empty VP after the raising o f the verb to the light verb, v, as illustrated in (187). (187) a. John [v [v xihuan]] [w tv [N p fade laoshi]] like his teacher ‘John likes his teacher.’ b. Bill ye [v [v xihuan] [w tv [N p e]] also like ‘Bill also likes (John’s teacher).’ ‘Bill also likes (his teacher).’ Furthermore, the relevant null object construction that involves an action verb may undergo reanalysis whereby the verb is reanalyzed as a nominalization and serves as the object o f the empty light verb, as illustrated in (188). (188) a. John piping-le tade laoshi. criticize-Asp his teacher ‘John criticized his teacher.’ b. Bill ye [v P [v DO] [N P piping], buguo Bill piping de shi tade also criticizing but criticize DE be his tongxue. classmate ‘Bill also did the criticizing; but the person whom Bill criticized was his classmate.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 262 On this view, the second conjunct in (188) is comparable to the one with the light verb overtly realized, as shown in (189), and there is no object gap involved. (189) Bill ye [v P [v zuo] [N P piping] also do criticizing ‘Bill also did the criticizing.’ A question arises pertinent to the representation in (187b), namely, what is the syntactic status o f the empty category when it is “selected” in the Numeration? In the following, I will discuss the syntactic status o f various types o f empty categories within the Minimalist framework. The conclusion we will reach is that the relevant empty category in (187b) does not enter into the Numeration, which will lead to arguments in support o f the LF copying approach to VP ellipsis in the next subsection. Yasui (1997) tries to give a unified licensing condition o f elided constituents, such as VP ellipsis and other empty categories, such as pro and traces by proposing the following. Ellipses and empty categories are distinct instances o f the same empty category that lacks any phonetic, syntactic, or semantic features when selected for the Numeration. They are licensed by getting their syntactic categories and semantic contents specified in the course o f the syntactic derivation so that they can be visible for LF interpretive processes. Take pro, wh-traces, and VP ellipsis for illustration. Let us take up each o f them in turn. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 263 Consider the following pair o f examples involving null subjects, her examples in (1) from Italian and English. (190) a. Italian \pro\ ha parlato. spoke ‘(He) spoke.’ b. English *\pro] spoke. In both (190a) and (190b), the subject pro is unambiguously identified as a DP by the Infl, which has a strong D-feature. However, the semantic O-features o f pro can only be specified by the strong Infl in Italian but cannot be provided by the weak Infl in English, thus the contrast. Next consider the following sentence involving a wh-trace, taken from her example in (24a). (191) Who, did John kill [t,]? According to Yasui (1997), in a case involving wh-movement such as (191), the category and content o f the wh-trace is recovered through the chain link with its antecedent who, which is o f the category D and has some specific content such as [human]. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 264 Lastly consider the following sentence involving an empty vP, taken from her example in (2a). (192) John talked to Mary, but [iP B ilf [I n f l didn’t][v P [f] [e]]] Following Abney (1987), Yasui assumes that functional categories such as Infl, Comp(lementizer), and Det(erminer) have the property o f selecting a specific category as its complement. Therefore, the empty vP in (192) can be unambiguously identified as a vP by the Infl, which selects a vP as its complement. In addition to its categorial feature, the empty vP is also provided with some semantic content in the sense that it contains the trace bound by the subject, Bill, given the Internal Subject Hypothesis. The empty vP is therefore categorically and semantically licensed. On the other hand, unavailability o f an object gap, as given in (193), taken from her example in (10a), is captured as follows. (193) *Mary doesn’t expect [C P Bill to win], but she [v P wants [e]] Yasui claims that unlike functional categories, lexical categories cannot unambiguously identify the categorial feature o f the following elided constituent, because they may take more than one type o f categories as complements. For example, V may take a CP as well as a DP as its complement. Therefore, the categorial feature o f the missing object in (193) cannot be unambiguously identified, thus the ungrammaticality. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 265 Yasui’s (1997) analysis, however, runs into some difficulties. First, under the Minimalist framework, her claim that a single absolutely empty category is selected in the Numeration is hard to maintain, since a category without any syntactic or semantic feature is not allowed in the Numeration. Second, her analysis predicts that null objects other than traces are impossible, because their categorial features cannot be unambiguously identified. This is not the case, as indicated by the possibility o f an object gap in a Chinese null object construction such as (187), repeated here. (187) a. John xihuan tade laoshi like his teacher ‘John likes his teacher.’ b. Bill ye [v [v xihuan]] [V P tv [D p e]] also like ‘Bill also likes (John’s teacher).’ ‘Bill also likes (Bill’s teacher).’ Given the above discussion, I hereby render Yasui’s (1997) proposal as untenable. The initial question still remains with regard to the syntactic status o f the empty category when it is selected in the Numeration. I will follow Chomsky (1993, 1995) in assuming that empty categories such as PRO and pro are lexical items and can be selected for the Numeration, whereas empty categories such as traces are deleted copies created by Move and thus are not represented in the Numeration. Note that pro might only have its categorial feature when selected for the Numeration. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 266 Suppose that pro has (^-features when it is selected in the Numeration, just like an overt DP, and those features enter into a checking relation with the Infl in the same way as those o f an overt DP. It would call for explanation why an overt DP can appear in [Spec, IP] in English but pro cannot, whereas both can in Italian. Turn to the Chinese null object construction as given in (187). It has been argued that the null object in question is neither a PRO nor a pro. As a result, it is not represented in the Numeration, on a par with other empty categories. Nevertheless, for ease o f exposition, the null object is still represented as an empty category in the relevant examples. In the next section, I will discuss how the relevant construction, which contains a gap, is interpreted. 3.8.2 LF Copying vs. PF Deletion There has been a long debate between the PF deletion and LF copying approaches towards vP ellipsis in English (as in (194)) since Ross (1967). (194) a. John can play the piano, b. Bill can [v P e], too. The PF deletion approach proposes that an empty vP such as the one in (194b) is derived from a fully represented structure (Bill can play the piano) by a deletion process under identity with some other vP in the antecedent clause at the PF component. On the other hand, the LF copying (or interpretive) approach argues that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 267 the relevant empty vP starts out as a null element and is assigned a fully represented structure later at LF via copying. In the following I will review some o f the arguments which have been put forth to defend the two approaches and it will be suggested that the LF copying approach should be adopted. Williams (1977) defends the LF copying approach by giving the following set o f data involving auxiliary reduction. (195) (Adapted from Williams 1977, (26)) a. John can go tomorrow. b. John can go tomorrow. c. John can tomorrow. d. * John can tomorrow. As cited in Williams (1977), the above contrasts have been captured under Selkirk’s (1972) Monosyllable Rule, which applies to reduce auxiliaries when they are followed by a fully represented vP, but does not when the vP is missing. Williams (1977) then argues that under the deletion theory, in order to prevent (195d) from being generated, the Monosyllable Rule has to be ordered after vP deletion. The ordering, however, violates his principle o f strict utterance, which states that rules o f Sentence Grammar such as the Monosyllable rule must precede those o f Discourse Grammar such as vP deletion. Even in the current framework, where both vP deletion and the Monosyllable Rule presumably apply at the PF component, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 268 deletion approach is still less preferred because an extra ordering o f the rules needs to be stipulated. This problem, on the other hand, does not arise under the LF copying approach. Under this approach, the empty vP in (195) starts out as null and thus the Monosyllable rule cannot apply to derive (195d). Therefore, the principle o f strict utterance is maintained or under the current framework, the correct result is obtained without stipulation. Williams (1977) also gives the following data involving the interaction between vP deletion and gapping to support the LF copying approach. (196) (Adapted from Williams 1977, (30)) a: Did John hit Bill? b: No, Sam hit Bill, and Pete hit Mary, c: No, Sam did, and Pete hit Mary, d: *No, Sam did, and Pete Mary. He argues that under the deletion theory, gapping has to apply after vP deletion; otherwise (196d) would be generated, an undesirable result. The ordering, however, violates the principle o f strict utterance, since a rule o f Sentence Grammar such as gapping precedes a rule o f Discourse Grammar such as vP deletion. (Or, under the current framework, the ordering has to be stipulated.) Once again, the problem is avoided under the LF copying approach. Under this approach, (196d) has the structure in (197). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 269 (197) Sam did [v P [v-v e] [ w tv [ n p e]]], and Pete [V .v e] Mary In the LF copying theory, gapping, like vP ellipsis, is not the result o f deletion; instead, the gapped verb is interpreted via copying under certain conditions (see Stillings (1975) for the conditions on gapping). In (197), the gapped verb in the second conjunct cannot be properly interpreted, since there is no antecedent in the first conjunct, thus the ungrammaticality. The paradigm in (196) is again captured under the LF copying theory without any stipulation. On the other hand, Ross (1969), as cited in Hankamer and Sag (1976), gives several arguments for the deletion approach. Let us first look at the following two sets o f data in (198) and (199), taken from Hankamer and Sag (1976, 404). (198) a. We can’t prove that there are such rules, but there are. b. *We can’t prove that there are such rules, but there is. (199) a. He knows how to fly, but I don’t know how to fly. b. He knows how to fly, but I don’t know how to. c. *He knows how to fly, but I don’t know why to fly. d. *He knows how to fly, but I don’t know why to. According to Ross, the contrast in (198) with respect to verbal agreement and that in (199) with respect to collocation o f wh- words and to are straightforwardly accounted for under the deletion theory whereby the elided constituents are fully represented in the beginning. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 270 Ross’ argument for the deletion theory, however, does not go through because the same sets o f data can also be straightforwardly captured under the LF copying approach, whereby the elided constituents are assigned fully represented structures at LF. Next consider the contrast in (200). (200) a. Paul Anderson’s fat, and I am fat, too. b. Paul Anderson’s fat, and I’m fat, too. c. Paul Anderson’s fat, and I am, too. d. *Paul Anderson’s fat, and I’m, too. Following King’s (1970) claim that failure o f otherwise reducible items to reduce correlates with the presence o f an immediately following deletion site, Ross takes the above contrast in (200) as a piece o f supporting evidence for the deletion approach. Nevertheless, the contrast in (200), which is similar to the one in (195), is exactly the one which will argue for the LF copying approach. Under the deletion theory, ordering o f the rules o f auxiliary reduction and deletion has to be stipulated in order to bar the ungrammatical sentence in (200d); i.e., the former has to apply after the latter. On the other hand, the LF copying approach does not induce the stipulation, given that the rule o f auxiliary reduction is not applicable in the first place. In addition to Ross (1969), Grinder and Postal (1971) also argues for the deletion approach on the basis o f the “missing antecedent” phenomenon mentioned in section 3.2.3.3, repeated here as in (201). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 271 (201) a. Harry doesn’t have a wife, but Bill does have a wife, and she is a nag. b. Harry doesn’t have a wife, but Bill does, and she is a nag. The pronoun she in (201b) should have an antecedent; yet, on the surface, the sentence contains no NP that can serve as one. According to Grinder and Postal (1971), this phenomenon is straightforwardly explained under the deletion approach whereby the antecedent for the pronoun is present in the beginning. Again, the missing antecedent phenomenon as in (201) can also be readily accommodated under the LF copying theory whereby the elided vP is fully represented at LF and thus the antecedent for the pronoun is there for interpretation. In brief, we have seen that the LF copying approach to ellipses can account for the whole range o f data presented, and it does not invoke any stipulation, unlike the PF deletion approach. N ow turn to the Chinese null object construction, which has been analyzed as an instance o f VP ellipsis. (202) a. John [v [v xihuan]] [V P tv [N p fade laoshi]] like his teacher ‘John likes his teacher.’ b. Bill ye [v [v xihuan]] [w tv [N p e]] also like ‘Bill also likes (John’s teacher).’ ‘Bill also likes (his teacher).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 272 In the following, I will pursue the two logical possibilities to the ellipsis, namely, the PF deletion approach and the LF copying approach, and will conclude that the latter should be upheld. Let us start with the PF deletion approach. Under this approach, both VPs in (202a-b) are fully represented in the beginning and the VP in (202b) gets deleted at PF under identity with the VP in the previous clause, namely, (202a), as demonstrated below. (203) a. John [v [v xihuan]] [V P tv [ n p tade laoshi]] b. Bill ye [v [v xihuan]] [v p tv [N P tade laoshi] 1 As discussed in Hoji’s (1999) class, an immediate difficulty arises with regard to the requirement o f identity on the VPs involved. Given that VP deletion is operated at the PF component, whereas (non)identity o f the VPs, which obviously involves the interpretation o f the pronoun in question, is presumably established at the LF component, some mechanism has to be stipulated so that the information o f (non)identity is also accessible at the PF component. On the other hand, the difficulty does not arise under the LF copying approach. Under this approach, the object gap in (202b) is empty in the beginning. In other words, the empty category in the object position o f (202b) is not represented in the Numeration, given the previous conclusion that empty categories other than PRO and pro are not represented in the Numeration. However, at the LF component, the empty VP in (202b), which has been obtained via V-to-v raising, is assigned a fully Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 273 represented structure via an operation o f VP copying. Since the requirement o f identity is not at stake, the aforementioned difficulty with the PF deletion approach is avoided here. The VP ellipsis in Chinese therefore goes for the LF copying approach, like VP ellipsis in English. The conclusion we just reached is indirectly supported by the data involving tone sandhi in Taiwanese, a Chinese dialect spoken in southern China and Taiwan. It has been widely known that Taiwanese exhibits tone sandhi. Simpson and Wu (1999) reports that the tone o f a syllable undergoes tone sandhi when followed by another tone-bearing syllable in the same tone sandhi domain. On the other hand, tone sandhi does not occur if the syllable is followed by a syllable in the neutral tone or if it appears in the final position o f a tone sandhi domain. One o f the tone sandhi domains is defined as such that the head and its complement are in the same domain, as illustrated below, taken from their example in (14). (204) a. (Comp -- IP) na*si* [A*-sin m* la i...] if A-sin neg come b. (A u x -V ) e* [lai] will come c. (V - O ) be* [chhe] buy books Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 274 d. (P - O ) tui* [goan* lau*pe] to my father N ow consider the Taiwanese null object sentence in (205). (205) a. A*-hun kah*i* in* lau*su like her teacher ‘A-fen likes her teacher.’ b. A*-lan ma* kah*i [ec] also like ‘A-lan also likes (A-fen’s teacher).’ ‘A-lan also likes (A-lan’s teacher).’ The strict/sloppy ambiguity exhibited in (205) suggests that the sentence also involves VP ellipsis, on a par with a Chinese null object sentence. The aspect o f the example that is crucial to the present discussion is that if the VP ellipsis in point is obtained via PF deletion, we would expect tone sandhi o f the verb to be triggered in (205b) as well as in (205a), given that the conjunct in (205b) is also fully represented in the beginning. This expectation, however, is not borne out, as indicated by the citation tone o f the verb in (205b). Alternatively, a stipulation o f rule ordering has to be invoked to the effect that PF deletion applies before tone sandhi. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 275 Again, the problem or stipulation does not arise if the LF copying approach is adopted. The data concerning tone sandhi as in (205) is thus in support o f the claim that VP ellipsis in the Chinese language goes for the LF copying theory. 3.9 Conclusion It has been argued in this chapter that the Chinese null object construction is indeed a null VP in disguise under the vP structure, which is obtained via V raising to v rather than to Infl. Moreover, distinction has to be made between the null object construction involving stative/resultative verbs and that involving action verbs, on the basis o f their differences in interpretive possibilities and locality effects on sloppy identity. The observed differences have been attributed to an additional reanalysis process o f nominalization, which is available to action verbs, but not stative or resultative verbs. It is further suggested that the difference between Chinese and Japanese with respect to availability o f VP ellipsis is contingent upon that o f V-to-v raising. In Chinese, V is raised to v, making VP ellipsis possible; on the other hand, in Japanese, V stays in situ and thus VP ellipsis is not obtained. It is also suggested that the difference between Chinese and English with respect to availability o f the VP ellipsis representation might arise from the structural difference in subject positions between the two languages. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 276 Chapter 4 vP Ellipsis in Chinese Kim (1995) claims that Chinese, like Japanese and Korean, has no genuine VP ellipsis constructions (namely, vP ellipsis constructions within the current framework) exactly parallel to those in English. However, recall our discussion that the difference between Chinese and English with respect to the possibility o f the null object construction, namely, VP ellipsis (under the vP structure), is contingent upon the movement o f the internal subject. In English, only vP ellipsis, but not VP ellipsis, is possible because the subject has been raised to [Spec, IP]. Chinese, on the other hand, allows the subject either to stay in-situ or to move to [Spec, IP]; therefore, it is predicted that not only VP ellipsis but also vP ellipsis exists in the language. The prediction is in fact borne out. Chinese does have vP ellipsis parallel to those in English, as illustrated in (1). (1) a. John criticized his teacher, and Mary did [v P e], too. ‘John criticized John’s teacher, and Mary criticized John’s teacher, too.’ ‘John criticized John’s teacher, and Mary criticized Mary’s teacher, too.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. b. John piping-le tade laoshi, Mary ye sh i[vPe]. criticize-Asp his teacher also be ‘John criticized John’s teacher, and Mary criticized John’s teacher, too.’ ‘John criticized John’s teacher, and Mary criticized Mary’s teacher, too.’ As indicated, the vP ellipsis in Chinese is made possible by a supporting auxiliary shi ‘be’ sitting in the Infl node, which functions as do-support in English. However, Hoji (personal communication) questions the nature o f the above ye shi ‘also be’ construction in Chinese. Is it an instance o f surface anaphora like vP ellipsis in English or deep anaphora? Hoji’s suspicion is based on the fact that Japanese has a similar construction, namely, what Hoji (1998d) calls “Stripping without case-markers” or “Non-CM Stripping,” as in (2), which is an instance o f deep anaphora. (See Hoji (1998d) for detailed discussion.) (2) (= Hoji 1988d, (48)) John-ga John-no kuruma-o aratta; Bill-mo da. Nom Gen car-Acc washed also BE ‘John washed John’s car; Bill washed John’s car.’ ‘John washed John’s car; Bill washed B ill’s car.’ In the following sections, it will be demonstrated that the ye shi construction in Chinese exhibits the full range o f properties associated with vP ellipsis in English and is an instance o f surface anaphora, like vP ellipsis in English. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.1 Dichotomous Ambiguity 278 Let us return to the relevant Chinese sentence in (lb), repeated here. (1) John piping-le tade laoshi, Mary ye shi [e] criticize-Asp his teacher also be ‘John criticized John’s teacher, and Mary criticized John’s teacher, too.’ ‘John criticized John’s teacher, and Mary criticized Mary’s teacher, too.’ ‘♦John criticized John’s teacher, and Mary criticized someone, too.’ As indicated, the sentence in (1) is ambiguous between a strict reading and a sloppy reading o f the pronoun tade ‘his’, which is typical o f vP ellipsis. In (1), either John and Mary criticized the same teacher (the strict reading), or both o f them criticized their own respective teachers (the sloppy reading). Note that unlike the null object construction (namely, the VP ellipsis construction under the current analysis) discussed previously, there is no third reading for (1). If John and Mary criticized different people, they must have criticized their own respective teachers. The absence o f a third reading is in fact expected. Recall that a third reading is possible only when the retained action verb o f the construction is reanalyzed as a nominalization. Since there is no such a verb involved in ye shi sentences like (1), it follows that there is no third reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 279 Next consider the following sentence involving an anaphor. (3) John piping-le ziji, Mary ye shi [e] criticize-Asp self also be ‘John criticized John, and Mary criticized Mary/*John/*someone.’ The use o f an anaphor in (3) gives rise to the sloppy reading only, which again is typical o f vP ellipsis. Again note that the third reading is impossible for the reason that no retained action verb is involved here. 4.2 Locality Effects Next consider the following ye shi construction in (4), which involves an embedded clause. (4) John piping-le tade laoshi, Mary zhidao Bill ye shi [e] criticize-Asp his teacher know also be ‘John criticized John’s teacher, and Mary knew that Bill criticized Bill’s teacher.’ ““ John criticized John’s teacher, and Mary knew that Bill criticized Mary’s teacher.’ ‘*John criticized John’s teacher, and Mary knew that Bill criticized someone.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 280 As indicated, the ye shi sentence in (4) only yields the local sloppy reading, but not the nonlocal sloppy reading. Moreover, it does not give rise to the unspecified reading, as confirmed by the fact that it cannot be followed by the sentence in (5). (5) *Danshi wo bu zhidao Bill piping de shi shei but I not know criticize DE be whom ‘But I don’t know whom Bill criticized.’ The locality effects exhibited by the ye shi construction again are one o f the properties associated with vP ellipsis in English. The compelling similarities between the ye shi ‘also be’ construction and vP ellipsis in English with respect to strict/sloppy ambiguity and locality effects thus drive us to speculate that the ye shi construction is the genuine Chinese counterpart o f vP ellipsis in English. This amounts to saying that the ye shi construction in Chinese is an instance o f surface anaphora. It will be shown in the subsequent sections that our speculation is indeed on the right track. 4.3 Surface Anaphora Hoji (1998d) has summarized the properties associated with surface anaphora as listed below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 281 (6) (= Hoji 1998d, (122)) Properties o f surface anaphora a. It requires a linguistic antecedent. b. It cannot give rise to a sloppy identity reading with an a-occurrence. c. It cannot give rise to a sloppy identity reading in the local context. d. It can give rise to Mix readings. It will be demonstrated in the following subsections that the ye shi construction in Chinese exhibits all o f the above properties o f surface anaphora, thus confirming our claim that the construction is an instance o f surface anaphora, like vP ellipsis in English. 4.3.1 Requirement o f a Linguistic Antecedent It has long been pointed out in Hankamer and Sag (1976) that surface anaphora cannot be pragmatically licensed and requires a linguistic antecedent. This is also the case with the ye shi construction, as indicated by the contrast between (7) and (8) with respect to the interpretation. (7) a. John zhengzai zema tade xuesheng Prog scold his student ‘John is scolding his student.’ b. Bill ye shi [e] also be ‘Bill is also scolding John’s student.’ ‘Bill is also scolding his (Bill’s) student.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 282 (8) [John is scolding his student.] Mary: Bill ye shi [e] also be ‘*Bill is also scolding John’s student.’ ‘*Bill is also scolding his (Bill’s) student.’ As indicated, the ye shi sentence in (7), where a linguistic antecedent is present, gives rise to the strict and sloppy identity readings. On the other hand, the one in (8), where no linguistic antecedent is present, gives neither the strict reading nor the sloppy reading; in fact, it does not give any interpretation at all. 4.3.2 Lexical Properties o f Dependent Expressions It has been discussed in chapter 3 that the sloppy identity reading in surface anaphora requires establishment o f what Hoji (1997a et esq.) calls Formal Dependency (FD) between the relevant nominal expressions, which, in turn, is subject to the three conditions listed in (9). (9) The three necessary conditions for an FD (A, B), where A and B are in argument positions: a. B is [4^ ]. b. A c-commands B. c. A is not in the local domain o f B. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 283 Since the establishment o f FD is necessary for sloppy identity, it follows that the availability o f a sloppy identity reading in surface anaphora is also subject to the three conditions in (9). In this and the following sections, it will be shown that the availability o f the sloppy identity reading in the ye shi construction is constrained by the conditions in (9), again confirming that the construction is an instance o f surface anaphora, like vP ellipsis in English. The first piece o f confirmation comes from the fact that the sloppy identity reading in the ye shi construction is unavailable when the relevant NP is not a P - occurrence, as shown by the contrast between (10) and (11). (10) a. John piping-le tade laoshi criticize-Asp his teacher ‘John criticized his teacher.’ b. Bill ye shi [e] also be ‘Bill also criticized John’s teacher.’ ‘Bill also criticized B ill’s teacher.’ (11) a. John piping-le Johnde laoshi criticize-Asp Gen teacher ‘John criticized John’s teacher.’ b. Bill ye shi [e] also be ‘Bill also criticized John’s teacher.’ ‘*Bill also criticized B ill’s teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 284 As indicated, the sloppy identity reading becomes unavailable when the Name John, which is [+a], is in place o f the sloppy pronoun tade ‘his’, which is [+P]. 4.3.3 The Local Disjointness Effects Another piece o f confirmation that the ye shi construction is surface anaphora comes from the fact that the sloppy identity reading in the construction is sensitive to the local disjointness condition in (9c). Consider the following ye shi sentences. (12) a. Wo xihuan wode laoshi I like my teacher ‘I like my teacher.’ b. Wo xiang ni ye shi [e] I think you also be ‘I think you like (my teacher).’ ‘I think you like (your teacher).’ (13) a. ?Wo xihuan wo 1 like me ‘I like m e.’ b. Wo xiang ni ye shi [e] I think you also be ‘I think you like (me).’ ‘*1 think you like (you).’ In (12a), wo ‘I’ is not in the local domain o f the possessive wode ‘m y.’ Consequently, the relevant FD, i.e., FD (t, wode) with t being the trace o f wo, can be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 285 established and the sloppy identity reading is available. On the other hand, the first occurrence o f wo ‘I’ in (13a) is in the local domain o f the second occurrence o f wo ‘me.’ Therefore, the relevant FD (t, wo) with t being the trace o f wo, cannot be established and the sloppy identity reading is unavailable. However, as pointed out by Li (personal communication, 2002), the sloppy identity reading becomes available for (13) in the following linguistic context. (14) a. Ren dou xihuan ziji man all like self ‘Everyone likes him/herself.’ Wo xihuan wo I like me ‘I like me.’ b. Wo xiang ni ye shi [e] I think you also be ‘?I think you like (me).’ ‘I think you like (yourself).’ The availability o f the sloppy identity reading in (14b) is not surprising, since the FD required can be established between (t, ziji) with t being the trace o f ren ‘man.’ 4.3.4 Mix Readings Hoji (1998d) argues that Mix readings arise only in surface anaphora such as vP ellipsis in English, as in (15)-(16), and they are based on Formal Dependency, which Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 286 is subject to conditions concerning p-occurrences o f the dependent expressions, c- command, and local disjointness. Therefore, the fact that the null object construction in Chinese, but not that in Japanese (cf. Hoji (1998d)), gives rise to Mix readings has led us to conclude that the Chinese null object construction is surface anaphora. (15) Max said he saw his mother; Oscar did too. a. Maxi said he, saw hisx mother; Oscar2 said he, saw hisi mother. b. Maxi said he! saw his! mother; Oscar2 said he2 saw his2 mother. c. Max! said he! saw hisi mother; Oscar2 said he2 saw his, mother. (Mix 1) d. *Max, said he, saw hisi mother; Oscar2 said hei saw his2 mother. (Mix 2) (16) Max said his mother saw him; Oscar did too. a. Maxi said hisi mother saw him,; Oscar2 said his, mother saw him,. b. Maxi said his, mother saw him,; Oscar2 said his2 mother saw him2 . c. Max, said his, mother saw him,; Oscar2 said his2 mother saw him,. (Mix 1) d. Max, said his, mother saw him,; Oscar2 said his, mother saw him2 . (Mix 2) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Now turn to the ye shi construction in Chinese. 287 (17) John shuota piping-le tade laoshi, Bill ye shi [e] say he criticize-Asp his teacher also be ‘John! said hei criticized hisi teacher; Bill2 said he, criticized hisi teacher.’ ‘John! said he, criticized his, teacher; Bill2 said he2 criticized his2 teacher.’ ‘John, said he, criticized his, teacher; Bill2 said he2 criticized his, teacher.’ (Mix 1) ‘*Johni said he, criticized his, teacher; Bill2 said he, criticized his2 teacher.’ (Mix 2) (18) John shuo tade laoshi piping-le ta, Bill ye shi [e] say his teacher criticize-Asp him also be ‘John, said his, teacher criticized him,; Bill2 said his, teacher criticized him,.’ ‘John, said his, teacher criticized him,; Bill2 said his2 teacher criticized him2 .’ ‘John, said his, teacher criticized him,; Bill2 said his2 teacher criticized him,.’ (Mix 1) ‘John, said his, teacher criticized him,; Bill2 said his, teacher criticized him2 .’ (Mix 2) The ye shi sentence in (17) allows not only across-the-board strict and sloppy identity readings but also Mix 1 reading, but not Mix 2 reading, whereas the sentence in (18) allows all o f the four readings. In brief, it has been shown that the ye shi construction in Chinese requires a linguistic antecedent and can give rise to Mix readings; however, it cannot give rise to a sloppy identity reading with an a-occurrence, nor can it give rise to a sloppy Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 288 identity reading in the local context. These facts confirm our initial speculation that the construction is an instance o f surface anaphora, on a par with vP ellipsis in English. 4.4 Scope of Adjuncts It has been discussed in chapter 2 that in an English vP ellipsis sentence, the adjunct in the first conjunct is interpreted as if it were part o f the elided vP in the second conjunct; on the other hand, in a Chinese null object/VP ellipsis sentence, the interpretation o f the elided VP in the second conjunct does not contain that o f the adjunct in the first conjunct, as indicated by the interpretive contrast between (19) and (20). (19) John clearly saw his mother, and Mary did, too. ‘John clearly saw John’s mother, and Mary clearly saw John’s mother.’ ‘John clearly saw John’s mother, and Mary clearly saw Mary’s mother.’ (20) John qingchude kanjian-le tade mama, Mary ye kanjian [e] le clearly see-Asp his mother also see Asp ‘John clearly saw his mother, and Mary saw John’s mother.’ ‘John clearly saw his mother, and Mary saw Mary’s mother.’ Such a contrast has been accounted for on the basis that in vP ellipsis, the auxiliary do, which occupies the Infl node, is higher than vP and a vP-level adjunct such as clearly whereas in VP ellipsis, a vP-level adjunct still sits in a position higher than the elided VP. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 289 Return to the ye shi construction discussed in the previous section. Given the conclusion that the construction should be analyzed on a par with vP ellipsis in English, it is predicted that the adjunct in the first conjunct be interpreted as part o f the elided vP in the second conjunct. This is a correct prediction. (21) a. John qingchude kaniian-le tade mama clearly see-Asp ‘John clearly saw John’s mother.’ b. Mary ye shi [e] also be ‘Mary clearly saw John’s mother.’ ‘Mary clearly saw Mary’s mother.’ As shown in (21), the interpretation o f the ellipsis in the second conjunct contains that o f the manner adverb qingchude ‘clearly’ in the first conjunct. 4.5 Postverbal Elements We have seen in chapter 3 that the interpretation o f an elided VP in the null object construction does not contain that o f a postverbal element such as a quantity adverbial phrase, a predicative complement, a purposive clause, or a postverbal PP in the antecedent conjunct. In the following sections, we will see the relevance o f these postverbal elements to the interpretation o f the ye shi construction. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 290 4.5.1 Quantity Adverbial Phrases Recall that quantity adverbial phrases such as the ones in (22) and (23) are adjoined to vP at LF; i.e., they are within vP and lower than the Infl node, as in (24) and (25). (22) a. John kan-le Harry Potter san tian read-Asp three day ‘John read Harry Potter for three days.’ b. John kan-le Harry Potter liang ci read-Asp two time ‘John read Harry Potter twice.’ (23) a. Johnkan shu kan-le san tian read book read-Asp three day ‘John read books for three days.’ b. John kan shu kan-le liang ci read book read-Asp two time Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 291 D/F V D/F san tian/liang ci kan Harry Potter tv to/F (25) vP V D/F san tian/liang ci kan tv shu kan tv to/F Given the structures in (24) and (25) and the conclusion that the ye shi construction involves vP ellipsis, which is made possible by the supporting auxiliary shi ‘be’ sitting in the Infl node, it is expected that in the ye shi construction, the interpretation o f the elided vP in the second conjunct contain that o f the postverbal quantity adverbial phrase in the first conjunct, in contrast to the null object construction. The expectation is borne out, as indicated in (26) and (27). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 292 (26) a. John kan-le Harry Potter san tian/ liang ci read-Asp three day/ two time ‘John read Harry Potter for three days/ twice.’ b. Bill ye shi [e] also be ‘Bill also (read Harry Potter for three days/ twice).’ (27) a. John kan shu kan-le san tian/ liang ci read book read-Asp three day/ two time ‘John read books for three days/ twice.’ b. Bill ye shi [e] also be ‘Bill also (read books for three days/ twice).’ 4.5.2 Predicative Complements Again recall that the predicative complements such as the one in (28) are also adjoined to vP at LF, as shown in (29). (28) John kan shu kan-de hen m an/lei read book read-DE very slowly/ tired ‘John read books slowly/ so much that he was tired.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. de hen man/lei kan tv Harry Potter kan tv to/R shu Given the structure in (29), it is again predicted that in the ye shi construction, which involves vP ellipsis, the postverbal predicative complement in the first conjunct will be interpreted as part o f the elided vP in the second. The prediction is correct, as indicated in (30). (30) a. John kan shu kan-de hen m an/lei read book read-DE very slowly/ tired ‘John read books so much that he was tired.’ b. Bill ye shi [e] also be ‘Bill also (read books slowly/ so much that he was tired).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 294 4.5.3 Purposive Clauses Like the quantity adverbial phrases and predicative complements, the purposive clauses such as the one in (31) are adjoined to vP at LF; in other words, they are still within vP and lower than the Infl node, as shown in (32). (31) John kan-le yi ben shu [dafa shijian] read-Asp one Cl book while away time ‘John read a book to while away his time.’ (32) vP PC vP PRO da la shijian Spec yi ben shu tv V’ PC I tPC Given the structure in (32), it is expected that in the ye shi construction, the interpretation o f the elided vP in the second conjunct contain that o f the postverbal purposive clause in the first conjunct. The expectation is again borne out, as indicated in (33). (33) a. John kan-le yi ben shu [dafa shijian] read-Asp one Cl book while away time ‘John read a book to while away his time.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 295 b. Bill ye shi [e] also be ‘Bill also (read a book to while away his time).’ 4.5.4 Postverbal PPs It has been discussed in chapter 3 that the structure o f a postverbal locative phrase such as the one in (34) is [Y [pp P NP]]. (34) John shui zai di-shang sleep at ground-on ‘John slept on the ground.’ It is clear then that the relevant PP is still within vP. Therefore, it is predicted that in the ye shi construction, the PP in the first conjunct will be interpreted as part o f the elided vP in the second. This is a correct prediction, as indicated in (35). (35) a. John shui zai di-shang sleep at ground-on ‘John slept on the ground.’ b. Bill ye shi [e] also be ‘Bill also (slept on the ground).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 5 The Null Topic Construction 296 5.1 Introduction In chapter 3, we have examined the syntactic properties and structures o f the Chinese null object construction such as (1) and concluded that the apparent object gap is in fact an empty VP, which is made possible via V-to-v raising, as illustrated in (2). (1) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Bill ye kanjian [e] le also see Asp ‘Bill also saw (John’s mother).’ ‘Bill also saw (his mother).’ (2) Bill ye [v [v kanjian]] [V P tv [ n p e]] It is also proposed that the null object construction involving an action verb that has a corresponding verbal norm, as exemplified in (3), can be additionally reanalyzed as a structure where the relevant verb is in fact an object verbal noun o f the empty light verb, as shown in (4). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 297 (3) a. John piping-le tade laoshi. criticize-Asp his teacher ‘John criticized his teacher.’ b. Bill ye piping le, danshi Bill piping de shi tade tongxue. also criticize Asp but criticize DE be his classmate ‘Bill also did the criticizing; but the person whom Bill criticized was his classmate.’ (4) Bill ye [v P [v DO] [N P piping]]1 A much related question arises in regard to the status o f the null object that occurs in an answer to a previous question such as (5). (5) A: John fasheng-le shemeshi? happen-Asp what ‘What happened to John?’ B: Bill dashang-le [e] hit-hurt-Asp ‘Bill hurt (John).’ (5B) is often called the null topic construction with an empty topic that presumably has its value set as John in (5 A). Given the superficial similarity between the null object construction and the null topic construction, is it possible to analyze the former as an instance o f the latter or vice versa? Or do they exhibit different syntactic properties and thus have to be distinguished from each other? To anticipate, the conclusion I will reach is that the 1 [v DO] stands for the empty light verb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 298 null object construction, or more accurately, both the VP ellipsis and nominalization representations associated with the null object construction, must be distinguished from the null topic construction despite the apparent similarity. 5.2 Structural Differences between the Null Object Construction and the Null Topic Construction It will be shown in this section that both the VP ellipsis and nominalization structures involved in the null object construction must be distinguished from the null topic structure. 5.2.1 The VP Ellipsis Structure vs. the Null Topic Structure Consider the example in (1) again, a null object sentence involving a resultative verb. (1) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Bill ye kanjian [e] le also see Asp ‘Bill also saw (John’s mother).’ ‘Bill also saw (his mother).’ As indicated in the English translation, the sentence in (1) is ambiguous between a strict identity reading (whereby John and Bill saw the same person’s mother) and a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 299 sloppy identity reading (whereby John and Bill saw their own respective mothers). Under the VP ellipsis analysis proposed in chapter 3, the strict/sloppy ambiguity is straightforwardly accounted for, given the LF interpretive rules o f Sag (1976) and Williams (1977). Nevertheless, can the null topic analysis accommodate the interpretive possibilities as well? In other words, can the object gap be treated as a variable bound by a null topic? The answer is partially yes. Theoretically speaking, nothing prevents the second conjunct in (1) from involving a null topic that has its value set as John’ s mother, thus giving rise to the strict reading. On this view, the clause will be comparable to the one with an overt topic, as shown below.2 (6) [John de mama];, Bill ye kanjian-le [t]; Gen mother also see-Asp ‘(John’s mother), Bill also saw.’ The next question is: can the sloppy identity reading be derived in a similar fashion? Conceptually speaking, a topic usually has a determined value instead o f being indeterminate. In other words, the second conjunct in (1) presumably may not involve a null topic containing a pronoun whose value is not fixed in the previous discourse. Nevertheless, suppose we suppress this conceptual infelicity and allow his mother to serve as a null topic for the second conjunct, with the value o f the pronoun his being supplied later by that o f the subject Bill. 2 It is implicitly assumed here that a null topic patterns with an overt one. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 300 (7) *[Tadej mama]k , Bill, ye kanjian-le [t]k his mother also see-Asp ‘(His mother), Bill also saw.’ As indicated, the sentence is unacceptable to begin with, let alone yielding the intended sloppy reading. We may still try to defend the null topic analysis o f the sloppy identity reading by stating the following. In (1), in addition to John’ s mother, what can serve as a null topic for the second conjunct is the notion o f self’ s mother. Therefore, the representation in (8), instead o f (7), is the correct one, which gives rise to the intended sloppy identity reading. (8) [Zijij de mama]k , Billj ye kanjian-le [t]k self Gen mother also see-Asp ‘(His own mother), Bill also saw.’ It thus seems that the null topic analysis o f the object gap also gives rise to the strict/sloppy ambiguity. However, consideration o f more data shows that this line o f reasoning cannot be upheld. It has been shown in chapter 3 that the null object construction exhibits the so- called locality effects on sloppy identity, as demonstrated below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 301 (9) a. John kanjian-le tade mama see-Asp his mother ‘John saw his mother.’ b. Mark zhidao Bill ye kanjian-le [e] know also see-Asp ‘Mark knew that Bill also saw (John’s mother).’ ‘Mark knew that Bill also saw (Bill’s mother).’ ‘*Mark knew that Bill also saw (Mark’s mother).’ In (9), the local sloppy reading that Bill saw his own mother is available, while the non-local sloppy reading that Mark knew that Bill saw his (Mark’s) mother is not. The unavailability o f the non-local sloppy reading is, again, straightforwardly captured under Sag-Williams’ account that the binder o f the sloppy pronoun is restricted to the subject o f the lambda predicate associated with the embedded empty VP, namely, Bill. Let us turn to the null topic analysis. As in (8), the notion o f self’ s mother should be available as a null topic for the second conjunct in (9). On this view, the second conjunct in (9) is comparable to (10). (10) [Zijij/kde mama),, Markj zhidao Billk ye kanjian-le [t], self Gen mother know also see-Asp ‘(His mother), Mark knew that Bill also saw.’ The sentence in (10) gives rise to not only the local sloppy reading whereby the value o f self is set as Bill but also the non-local sloppy reading whereby the value o f Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 302 self is set as Mark. In other words, the null topic analysis incorrectly predicts the availability o f the non-local sloppy reading, thus undermining the idea o f treating the notion o f self’ s mother as a potential null topic for the null object sentences in (1) and (9). In brief, the null object construction, as exemplified in (1) and (7), is structurally ambiguous between the VP ellipsis and the null topic structures. However, the sloppy identity readings exhibited in (1) and (7) can only be correctly derived under the VP ellipsis analysis, but not the null topic analysis. The first different property between the null topic construction on the one hand and the VP ellipsis structure on the other thus emerges. Whereas both can yield the strict interpretation, only the latter gives rise to the sloppy identity reading. That the null object construction such as (1) is structurally ambiguous is supported by the observation that the null object sentence involving an anaphor gives rise to not only the sloppy identity reading but also the strict identity reading. It has been noted that in an English VP ellipsis sentence such as (11), which involves an anaphor, the sloppy identity reading is either obligatory or at least much preferred over the strict identity reading. (11) John shot himself and Bill did too. ‘John shot John and Bill shot John too.’ ‘??John shot John and Bill shot Bill too.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 303 On the other hand, a Chinese null object construction that also involves an anaphor gives rise to the sloppy reading and equally acceptable strict reading. (12) a. John sheshang-le taziji shoot-hurt-Asp himself ‘John shot himself.’ b. Bill ye sheshang-le [e] also shoot-hurt-Asp ‘Bill also shot himself.’ ‘Bill also shot John.’ It is unexpected under the proposed VP ellipsis analysis that the strict reading is readily available. However, if the null object construction in (12) is ambiguous between the VP ellipsis structure and the null topic structure, the strict reading is expected to obtain in the latter, with the null topic having its value set as John. (13) [John];, Bill ye sheshang-le [tf also shoot-hurt-Asp ‘(John), Bill also shot.’ Note, however, that the structural ambiguity o f the null object construction should not obscure the differences between the VP ellipsis structure and the null topic structure: only the former can give rise to the sloppy identity reading, but not the latter. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2.2 Nominalization vs. the Null Topic Structure 304 Now let us turn to the null object construction involving an action verb that has a corresponding verbal noun, as given in (14). (14) a.. John piping-le tadelaoshi criticize-Asp his teacher ‘John criticized his teacher.’ b. Bill ye piping-le [e] also criticize-Asp ‘Bill also criticized (John’s teacher).’ ‘Bill also criticized (his teacher).’ ‘Bill also criticized (someone).’ As indicated in the English translations, in addition to the strict and sloppy identity readings generated by VP ellipsis, a third unspecified reading is also possible, which has been attributed to the nominalization o f the verb into a verbal noun. However, can the null topic analysis account for this unspecified reading? The answer is no. If the unspecified reading could be obtained in a null topic representation, we would expect it to be available for null topic sentences involving verbs other than the relevant action verb. This is not the case, as indicated by the ungrammaticality o f the following sentences involving a stative verb and a resultative verb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 305 (15) A: Ni xihuan John ma? you like Q ‘Do you like John?’ B: *Bu, danshi [top e], Bill xihuan [t]; ta xihuan de shi Mark no but like he like DE be ‘No, but Bill likes (someone); the person whom he likes is Mark.’ (16) A: Ni kandaoJohnle ma? you see Asp Q ‘Did you see John?’ B: *Meiyou, danshi [xop e], Bill kandao-le [t]; ta kanjian de shi Mark no but see-Asp he see DE be ‘No, but Bill saw (someone); the person whom Bill saw was Mark.’ The above fact thus suggests that the null topic structure must be separated from the nominalization structure as well. In sum, it has been shown that the so-called Chinese null object construction as discussed in chapter 3 may be structurally three-way ambiguous among the null topic representation, the VP ellipsis representation, and the nominalization representation. However, the three representations must be separated from one another. 5.2.3 Ambiguous Structures o f the Null Topic Construction We have just seen that the null object construction may be ambiguous between the VP ellipsis structure and the null topic structure. Can the null topic construction such as the question-answer pair in (17) be structurally ambiguous as well? The answer is yes. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 306 (17) A: Ni kandaoJohnle ma? you see Asp Q ‘Did you see John?’ B: Meiyou, buguo Bill kandao-le [e] no but see-Asp ‘No, but Bill saw (John).’ The readers should be clear from the earlier discussion that the answer in (17B) allows two structures. In one structure, what is missing is the object, which is treated as a variable bound by the null topic, John, as in (18a). In the other structure, the whole VP is missing, with the verb kandao ‘see’ stranded in v, as in (18b). The empty VP is then recovered from the VP in the antecedent clause. (18) a. [Johnjj, Bill kandao-le [t]i see-Asp ‘(John), Bill saw.’ b. Bill [v [ v kandao]] [v p W [ dp c ] j Since there is no anaphor or pronoun involved here, both structures give rise to the same strict interpretation. N ow consider the following null topic construction involving a pronoun. (19) A: John kanjian-le tade laoshi ma? see-Asp his teacher Q ‘Did John see his teacher?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 307 B: Meiyou, buguo Bill kanjian-le [e] no but see-Asp ‘No, but Bill saw (John’s teacher).’ ‘No, but Bill saw (his teacher).’ As discussed earlier, the null topic analysis o f the object gap in (19B) can only yield the strict interpretation, with the null topic having its value set as John’ s teacher. The sloppy interpretation, however, can be obtained by treating the apparent object gap as an instance o f VP ellipsis. 5.2.4 Confirmation In the following, three sets o f data will be presented in support o f the conclusion we reached in the earlier discussion that the VP ellipsis structure gives rise to strict/sloppy ambiguity, whereas the null topic structure only yields the strict interpretation. 5.2.4.1 Distinct Predicates The first set o f data we will consider concerns the question-answer pair where the internal structures o f the predicates involved vary arbitrarily. (20) A: John cong xuexiao huilai yihou, yizhi hen xingfen. from school return after constantly very excited ‘John has been very excited after returning from school.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 308 Tade laoshi hen piaoliang ma? his teacher very pretty Q ‘Is his teacher pretty?’ B: Dui, Bill kandao-le [e] yes see-Asp ‘Yes, Bill saw John’s teacher.’ ‘*Yes, Bill saw B ill’s teacher.’ The relevant predicates in (20A-B) have the following representations after the verbs raise to v. (21) a. (=(19A )) Tade laoshi [A P hen piaoliang] ma? b. (= (1 9 B » Bill [v [v kandao]] [ V P tv [ n p e]] As is clear from the representations given in (21), the LF interpretive rules o f VP ellipsis cannot apply to reconstruct the empty VP in (21b), because in (21a), there is no node corresponding to it. In other words, the VP ellipsis representation is not available for a sentence like (20). The only possibility we are left with is then the null topic representation. The fact that the sentence in (20) is unambiguously represented as a null topic structure, coupled with the fact that it only has the strict reading therefore indicates that the null topic structure does not gives rise to the sloppy reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 309 5.2.4.2 Double Object Structures Next consider the following double object sentences. (22) A: John song tade shiyou yiding maozi ma? give his roommate one hat Q ‘Did John give his roommate a hat?’ B: Dui, Bill ye song-le [e] [e] yes also give-Asp ‘Yes, Bill also gave John’s roommate a hat.’ ‘Yes, Bill also gave B ill’s roommate a hat.’ The VPs in (22) have the following representations after the verbs raise to v. (23) a. John [,, [v song]] [V P [ v tv [ n p tade shiyou]] [N P yiding maozi]] b. Bill ye [v [v song]] [V P [v tv [N P e]][N P e]] Unlike the answer in (20B) discussed above, the answer in (22B) allows the VP ellipsis structure as well as the null topic structure, because there is a corresponding VP node in (22A) to serve as the antecedent for the empty VP in (22B). Note that the sentence exhibits strict/sloppy ambiguity. Now look at the double object sentence in (24B), with the representation after the V-to-v raising given in (25). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 310 (24) A: John song tade shiyou yiding maozi ma? give his roommate one hat Q ‘Did John give his roommate a hat?’ B: Dui, Bill ye song-le [e] yishuang xiezi yes also give-Asp one pair shoe ‘Yes, Bill also gave John’s roommate a pair o f shoes.’ ‘*Yes, Bill also gave Bill’s roommate a pair o f shoes.’ (25) Bill [v [v song]] [w [v tv [N P e]] [N P yishuang xiezi]] As indicated by the translation, the sloppy reading becomes unavailable for the sentence in (24B), which forms a minimal pair with that in (22B). The major difference between the two sentences is as follows. The sentence in (22B), with two objects missing, can be made into an empty VP after V-to-v raising, and thus is amiable to the LF interpretive rules o f VP ellipsis (in particular, the rule o f VP copying). On the other hand, the sentence in (24B), with one object stranded, cannot be made empty even after the raising o f the verb to v, as shown in (25); and as a result, the rule o f VP copying would fail. Given the unavailability o f the VP ellipsis structure, the sentence in (24B), like that in (20B), is unambiguously treated as an instance o f the null topic representation. The fact that the sentence only has the strict interpretation, again, supports the claim that only the VP ellipsis structure, but not the null topic structure, can yield the sloppy reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 311 5.2.4.3 Relative Clauses The last set o f data we w ill consider involves relative clauses such as (26). (26) A: John zuo-le [ta shangke yong t,] de zika, ma? make-Asp he teach use DE flash card Q ‘Did John make the flash cards John used in class?’ B: Meiyou, buguo, Bill zuo-le [e] no but make-Asp ‘No, but Bill make the flash cards John used in class.’ ‘No, but Bill make the flash cards Bill used in class.’ Once again, the answer in (26), which exhibits strict/sloppy ambiguity, allows both the VP ellipsis and null topic structures. Given the claim that the sloppy reading can only be derived from the former, but not the latter, absence o f the sloppy reading is predicted if the possibility o f the VP ellipsis structure is eliminated. This prediction is borne out, as illustrated in (27). (27) A: John zuo-le [ta shangke yong t,] de zika, ma? make-Asp he teach use DE flash card Q ‘Did John make the flash cards John used in class?’ B: Meiyou, buguo, Bill zuo [e] gei ta le no but make to him Asp ‘No, but Bill make the flash cards John used in class.’ ‘*No, but Bill make the flash cards Bill used in class.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 312 The possibility o f the VP ellipsis representation is eliminated from (27B) by adding to the sentence another argument o f VP, the goal PP. As a result, (27B) only has the null topic structure, which only allows a strict interpretation, unlike (26). The sloppy reading therefore distinguishes the VP ellipsis structure from the null topic structure. The same conclusion is reached in Doron (1998), where it is presented that the null object sentence allowing the VP ellipsis representation yields strict/sloppy ambiguity, whereas the one not allowing it only gives the strict reading, as shown in (28) and (29), respectively (his examples (16) and (17) from Hebrew). (28) A: dina soreget et ha- svederim Se- hi loveSet Dina knits ACC the sweaters that she wears ‘Does Dina knit the sweaters that she wears?’ B: lo, aval ima Sela soreget no but mother hers knits ‘No, but Dina’s mother knits the sweaters that Dina wears.’ ‘No, but Dina’s mother knits the sweaters that Dina’s mother wears.’ (29) A: dina soreget et ha- svederim Se- hi loveSet Dina knits ACC the sweaters that she wears ‘Does Dina knit the sweaters that she wears?’ B: lo, aval ima Sela soreget 1-a no but mother hers knits to-her ‘No, but Dina’s mother knits the sweaters that Dina wears to Dina.’ ‘*No, but Dina’s mother knits the sweaters that Dina’s mother wears to Dina.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 313 Like (26), the answer in (28B) allows both the null topic and VP ellipsis representations and gives the strict reading as well as the sloppy reading. On the other hand, the answer in (29B), which allows only the null topic representation because o f the added goal PP, gives the strict reading only. To sum up, we have seen that both the null object and null topic constructions may be structurally ambiguous between the VP ellipsis and null topic representations. However, only the VP ellipsis structure, not the null topic structure, gives rise to the sloppy identity reading. 5.3 Surface Anaphora vs. Deep Anaphora In addition to the interpretive difference, we will look at another difference between the null topic structure and the VP ellipsis structure in this section. It will be shown that the former can be an instance o f deep anaphora, whereas the latter can only be an instance o f surface anaphora, in the sense o f Hankamer and Sag (1976). Hankamer and Sag (1976) note that anaphoric forms such as VP ellipsis in English require linguistic antecedents, whereas anaphoric forms such as do it do not, as shown by the contrast in (30), taken from their examples in (3) and (4). (30) [Hankamer attempts to stuff a 9-inch ball through a 6-inch hoop.] Sag: *It’s not clear that you’ll be able to. Sag: It’s not clear that you’ll be able to do it. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 314 Based on the above contrast, Hankamer and Sag (1976) claim that anaphora can be classified into two types. “Surface anaphora” such as VP ellipsis in English cannot be pragmatically licensed and requires a linguistic antecedent; on the other hand, “deep anaphora” such as do it can be pragmatically licensed and does not need a linguistic antecedent. Bear this distinction in mind and return to sentences such as (20), repeated here. (20) A: John cong xuexiao huilai yihou, yizhi hen xingfen. from school return after constantly very excited ‘John has been very excited after returning from school.’ Tade laoshi hen piaoliang ma? his teacher very pretty Q ‘Is his teacher pretty?’ B: Dui, Bill kandao-le [e] yes see-Asp ‘Yes, Bill saw John’s teacher.’ ‘*Yes, Bill saw B ill’s teacher.’ We have seen that when there is no appropriate linguistic antecedent, the VP ellipsis representation o f the object gap is impossible, as indicated by the unavailability o f the sloppy identity reading in (20B). This fact thus suggests that the proposed VP ellipsis representation is an instance o f surface anaphora. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 315 The same conclusion is suggested by the observation that the otherwise possible sloppy interpretation is again unavailable in the following context, where no linguistic antecedent is present. (31) [Observing John get a crush on his date in the prom, Mark asks Bill,] Ni ye xihuan ma [e]? you also like Q ‘Do you like John’s date?’ ‘*Do you like your date?’ Note that whereas the sloppy interpretation is impossible, the strict interpretation is possible, which suggests that the null topic representation is available for the sentence in (31). Therefore, in addition to showing that the VP ellipsis structure under discussion is an instance o f surface anaphora, the utterance in (31) also indicates that the null topic structure can be an instance o f deep anaphora.3 The deictic use o f the null object/topic as in (32) illustrates the same point that the null object/topic can be an instance o f deep anaphora, which does not require a linguistic antecedent. 3 As noted by Hoji (personal communication, 1999), the test of the sloppy-like interpretation with deep anaphora in English without a linguistic antecedent as discussed in Dalrymple (1991) is restricted to action verbs. However, in the case of Chinese, the use of a stative Verb such as xihuan ‘like’ in (31) is forced, because an action verb in Chinese can be reanalyzed as a nominalization, thus giving rise to a reading that may be compatible with the sloppy-like interpretation, as indicated below, (i) [Observing John fail to kiss his date in the prom, Mark asks Bill,] Ni wen [e] le ma? you kiss Asp Q ‘Did you kiss John’s date?’ ‘Did you kiss someone (whom could be your date)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 316 (32) [In the study where the father points to the computer, the son says,] Duibuqi, wo yonghuai le [e] sorry I break Asp ‘Sorry, I broke (it).’ Still another piece o f evidence in support o f the claim that VP ellipsis in question is an instance o f surface anaphora comes from the so-called “missing antecedent” phenomenon, first observed in Grinder and Postal (1971) for VP ellipsis in English, as given in (33), taken from their example in (12). (33) a. Harry doesn’t have a wife, but Bill does have a wife, and she is a nag. b. Harry doesn’t have a wife, but Bill does, and she is a nag. The fact that the sentence in (33b) contains a pronoun o f which the antecedent is part o f the elided VP has been taken by Grinder and Postal as an argument for the transformational nature o f VP ellipsis, which, in Hankamer and Sag’s (1976) term, is an instance o f surface anaphora. Bresnan (1971) further notes that this missing antecedent phenomenon is only possible with anaphora such as VP ellipsis, but not with anaphora such as do it, which is classified as deep anaphora in Hankamer and Sag (1976). (34) *Jack didn’t cut Betty with a knife; Bill did it, and it was rusty. (The underlined it refers to the knife Bill cut Betty with.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 317 Now consider the relevant VP ellipsis in Chinese. (35) a. John meiyou qu-guo limian you nuuren de jiuba not go-Asp inside have woman DE bar ‘John has never been to a bar with women inside it.’ b. Bill qu-guo, hai shuo tamen dou hen piaoliang go-Asp and say they all very pretty ‘Bill has been to (a bar with women inside it), and said they all were pretty.’ Like in (33b), the sentence in (35b) can contain a pronoun o f which the antecedent is part o f the elided VP. Note that the instance o f the NP nuuren ‘women’ in (35a) cannot serve as an antecedent for the pronoun tamen ‘they,’ as indicated by the fact that the sentence in (35a) cannot be followed by a sentence such as (36). (36) *Tamendouhen piaoliang they all very pretty ‘They all were pretty.’ The possibility o f the missing antecedent phenomenon in VP ellipsis under discussion thus supports the claim that it is an instance o f surface anaphora, which is fully represented at some point. Also note that if the antecedent utterance in (35a) is replaced by non-verbal pointing, the utterance in (35b) becomes ungrammatical, as shown below. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 318 (37) [Pointing to a bar that cannot be seen through, Mark says:] *Bill qu-guo, hai shuo tamen dou hen piaoliang go-Asp and say they all very pretty ‘Bill has been to (the bar), and said they all were pretty.’ The impossibility o f the missing antecedent phenomenon in (35) is in fact expected, given the above claim that the null object/topic structure, which is the only available representation for the object gap in (37), is an instance o f deep anaphora. In brief, we have seen that the null topic representation can be distinguished from the VP ellipsis representation in two aspects. The former is an instance o f deep anaphora and only yields the strict interpretation. On the other hand, the latter is an instance o f surface anaphora and can yield the sloppy identity reading as well as the strict identity reading. 5.4 Islands In this section, I will present another way o f distinguishing the VP ellipsis representation from the null topic representation based on locality effects. It is well known since Ross (1967) that VP ellipsis in English is not sensitive to constraints on movement. On the other hand, the null topic representation is predicted to be constrained by conditions on movement, given Huang’s (1984) movement analysis. Indeed, a null topic cannot be related to an object gap within an island, as shown by the discourse in (38a-c). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 319 (38) [Pointing to a particular student in the office] CNPC - relative clause a. *[Nage xuesheng],, wo zhengzai zhao [tjjiao-guo [e],] de laoshij that student I Asp look for teach-Asp DE teacher ‘(That student), I am looking for the teacher who has taught (him).’ (From Qu (1994,14)) CNPC - noun-complement clause b. *[Nage xuesheng]i, wo bu xiangxin [Johnjiao-guo [e],] that student I not believe teach-Asp zheju hua this saying ‘(That student), I don’t believe the statement that John has taught (him).’ Adjunct Condition c. *[Nage xuesheng]j, Bill [yinwei Johnjiao-guo [e]i] hen that student because teach-Asp very bu gaoxing not happy ‘(That student), Bill is unhappy because John has taught (him).’ Note that the null topics pattern with the overt topics with respect to locality effects. On the other hand, the VP ellipsis representation under discussion is not sensitive to locality effects. The following examples illustrate this difference between the two representations with respect to islands such as relative clauses (in (39)), noun- complement clauses (in (40)), and adjuncts (in (41)). In each o f these examples, the first response, which additionally allows the VP ellipsis representation o f the object gap, is grammatical despite the fact that the object gap is inside an island, whereas Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 320 the second response is ungrammatical, since it only involves a null topic related to an object gap within an island. (39) CNPC - relative clause Q: Ni jiao-guo [V P tv nage xueshengi] ma? you teach-Asp that student Q ‘Have you taught that student?’ RlrMeiyou, wo ye zhengzai zhao [tjjiao-guo [V P tv [e],]] de laoshij no I also Asp find teach-Asp DE teacher ‘No, I am also looking for the teacher who has taught (that student).’ R2:*Meiyou, wo ye zhengzai zhao [tjjiao-guo [V P tv [e]i shuxue]] de no I also Asp find teach-Asp math DE laoshij teacher ‘No, I am also looking for the teacher who has taught (that student) math.’ (40) CNPC - noun-complement clause Q: Youren shuo Johnjiao-guo [vp tv nage xueshengi];tashitade someone say teach-Asp that student he be his shuxue laoshi. math teacher ‘Someone said that John has taught that student; he was his math teacher.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 321 R1 :Shi ma? Buguo wo bu xiangxin [Johnjiao-guo [w tv [e],]] be Q but I not believe teach-Asp zheju hua this saying ‘Really? But I don’t believe the statement that John has taught (that student).’ R2:*Shima? Buguo, wo bu xiangxin [John jiao-guo [V P tv [e]( shuxue]] be Q but I not believe teach-Asp math zheju hua this saying ‘Really? But I don’t believe the statement that John has taught (that student) math.’ (41) Adjunct Condition Q: Bill zhidao Johnjiao-guo [ Vp tv nage xueshengi] ma? John shi tade, know teach-Asp that student Q be his shuxue laoshi. math teacher ‘Does Bill know that John has taught that student? John was his math teacher.’ R1:Zhidao. Bill hai [yinwei Johnjiao-guo [w tv [e]i]] hen know because teach-Asp very bu gaoxing. not happy ‘Yes, Bill was unhappy because John has taught (that student).’ R2:*Zhidao. Bill hai [yinwei Johnjiao-guo [V P tv [e], shuxue]] hen know because teach-Asp math very bu gaoxing. not happy ‘Yes, Bill was unhappy because John has taught (that student) math.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 322 5.5 Conclusion To conclude the discussion, despite their superficial similarity, the VP ellipsis representation o f the object gap can be distinguished from the null topic representation, based on the observed differences with respect to availability o f the sloppy identity reading, surface/deep anaphora distinction, and locality effects, as summarized in (42). (42) Differences between the VP ellipsis and the null topic representations a. VP ellipsis gives rise to strict/sloppy ambiguity, whereas the null topic representation only yields the strict interpretation. b. VP ellipsis is an instance o f surface anaphora, whereas the null topic representation can be an instance o f deep anaphora. c. VP ellipsis is not sensitive to locality effects, whereas the null topic representation is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 6 Disjunctive vs. A-Not-A Questions 323 6.1 Introduction In contrast to Li and Thompson (1981) and Luu (1981), Huang (1991b) argues that the so-called “A-not-A” questions like (1) are not the results o f applying a single rule o f coordinate deletion to a full-fledged disjunctive question in (2); instead, the A-not- A question in (la) is derived from a simplex sentence with an interrogative Infl constituent that is phonetically realized by a rule o f reduplication whereas the A-not- A question in (lb ) is derived from a structure o f juxtaposed vPs which may be subject to anaphoric ellipsis. (1) a. Ta xihuan bu xihuan ni? he like not like you ‘Does he like (you) or does (he) not like you?’ b. Ta xihuan ni bu xihuan? he like you not like ‘Does he like you or does (he) not like (you)?’ (2) Ta xihuan ni haishi ta bu xihuan ni? he like you or he not like you ‘Does he like you or does he not like you?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 324 In the face o f the foregoing discussion on ellipsis in chapter 3, it is interesting to investigate the process o f anaphoric ellipsis as proposed in Huang (1991b) for the A- not-A question like (lb). In particular, what is the relevance o f anaphoric ellipsis to other ellipses like VP and vP ellipsis? Given that the process o f anaphoric ellipsis is especially put forth to generate A-not-A questions, which in turn have been assumed to be related to ordinary disjunctive questions, we will start by examining the full paradigms o f both question forms. 6.2 Disjunctive vs. A-Not-A Questions The ordinary disjunctive questions may include a paradigm like (3). (3) a. Ta xihuan naben shu haishi ta bu xihuan naben shu? he like that book or he not like that book ‘Does he like that book or does he not like that book?’ b. Ta xihuan naben shu haishi bu xihuan naben shu? he like that book or not like that book ‘Does he like that book or does (he) not like that book?’ c. Ta xihuan naben shu haishi bu xihuan? he like that book or not like ‘Does he like that book or does (he) not like (that book)?’ d. Ta xihuan haishi bu xihuan naben shu? he like or not like that book ‘Does he like (that book) or does (he) not like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 325 e. *Ta xi- haishi bu xi-huan naben shu? he like or not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ Like the ordinary disjunctive questions in (3), the A-not-A questions as shown in (4) are also a kind o f yes/no questions, taking a disjunctive form requesting the addressee to identify either the affirmative or the negative from its two disjuncts. However, unlike an ordinary disjunctive question, an A-not-A question contains no conjunction like haishi ‘or’ between the two disjuncts. (4) a. *Ta xihuan naben shu ta bu xihuan naben shu? he like that book not like that book ‘Does he like that book or does he not like that book?’ b. Ta xihuan naben shu bu xihuan naben shu? he like that book not like that book ‘Does he like that book or does (he) not like that book?’ c. Ta xihuan naben shu bu xihuan? he like that book not like ‘Does he like that book or does (he) not like (that book)?’ d. Ta xihuan bu xihuan naben shu? he like not like that book ‘Does he like (that book) or does (he) not like that book?’ e. Ta xi- bu xi-huan naben shu? he li- not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 326 Early linguists have taken the A-not-A questions as a special type o f disjunctive questions and accounted for the formation o f the questions in (3b)-(3d) as well as those in (4b)-(4e) as the result o f application o f a single rule o f coordinate deletion to a full coordinate structure in (3a). Thus, forward deletion o f the second subject in (3a) gives (3b) and (4b). Further forward deletion o f the second object in (3b) and (4b) gives (3c) and (4c). On the other hand, backward deletion o f the first object in (3b) and (4b) gives (3d) and (4d). In the case o f backward deletion, it is also possible with the A-not-A question form to delete an identical subpart o f the first verb, giving (4e). This, however, is impossible with the ordinary disjunctive question form, as indicated by the ungrammaticality o f (3e). However, Huang (1991b) argues that the A-not-A questions cannot be treated on a par with the disjunctive questions, based on the observation that they behave differently with respect to lexical integrity, preposition stranding, and island constraints. Furthermore, problems arise with the coordination deletion analysis o f the disjunctive and A-not-A questions, which we will turn to in section 6.3. 6.2.1 Lexical Integrity The first difference between disjunctive and A-not-A questions concerns preservation o f lexical integrity. The grammatical contrast between the disjunctive question in (3e) and the A-not-A question in (4e) shows that violation o f the Lexical Integrity Hypothesis, as stated in (5), is possible with the latter, but not with the former. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 327 (3) e. *Ta xi- haishi bu xi-huan naben shu? he like or not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ (4) e. Ta xi- bu xi-huan naben shu? he li- not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ (5) Lexical Integrity Hypothesis Phrase-level rules belonging to the syntactic component cannot affect a proper subpart o f a lexical category (word). 6.2.2 Preposition Stranding The second difference between disjunctive and A-not-A questions concerns the general prohibition against preposition stranding, which may be expressed as the filter in (6). (6) *[P P P [e]] Thus, unlike an A-not-A question such as (7), a disjunctive question such as (8) is unacceptable. ( 7 ) Ni gen bu gen ta qu? you with not with him go ‘Will you go with him?’ (8) *Ni gen haishi bu gen ta qu? you with or not with him go ‘Will you go with him?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 328 6.2.3 Island Constraints The last difference between disjunctive and A-not-A questions we will discuss concerns island constraints. In particular, the latter exhibits island effects whereas the former does not. Thus, a disjunctive question with haishi ‘or’ may be embedded in a sentential subject, but an A-not-A form cannot, as shown by the contrast between (9) and (10). (9) a. [Ta huiqu Taiwan haishi ta bu huiqu Taiwan] bijiao hao? he return or he not return more good ‘Is it better that he goes back to Taiwan or that he does not go back to Taiwan?’ b. [Ta huiqu Taiwan haishi bu huiqu Taiwan] bijiao hao? he return or not return more good ‘Is it better that he goes back to Taiwan or that (he) does not go back to Taiwan?’ c. [Ta huiqu Taiwan haishi bu huiqu] bijiao hao? he return or not return more good ‘Is it better that he goes back to Taiwan or that (he) does not go back to (Taiwan)?’ d. [Ta huiqu haishi bu huiqu Taiwan] bijiao hao? he retumor not return more good ‘Is it better that he goes back to (Taiwan) or that (he) does not go back to Taiwan?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (10) a. * [Ta huiqu Taiwan bu huiqu Taiwan] bij iao hao? he return not return more good ‘Is it better that he goes back to Taiwan (or) that (he) does not go back to Taiwan?’ b. * [Ta huiqu Taiwan bu huiqu] bij iao hao? he return not return more good ‘Is it better that he goes back to Taiwan (or) that (he) does not go back to (Taiwan)?’ c. *[Ta huiqu bu huiqu Taiwan] bijiao hao? he return not return more good ‘Is it better that he goes back to (Taiwan) (or) that (he) does not go back to Taiwan?’ d. *[Tahui- bu hui-qu Taiwan] bijiao hao? he return not return more good ‘Is it better that he goes back to (Taiwan) (or) that (he) does not go back to Taiwan?’ Likewise, a disjunctive question may occur within a relative clause, but an A- not-A form cannot. (11) a. Ni xihuan [piping ni haishi bu piping ni] de ren? you like criticize you or not criticize you DE person ‘Do you like people who criticize you or people who do not criticize you?’ b. Ni xihuan [piping ni haishi bu piping] de ren? you like criticize you or not criticize DE person ‘Do you like people who criticize you or people who do not criticize (you)?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 330 c. Ni xihuan [piping haishi bu piping ni] de ren? you like criticize or not criticize you DE person ‘Do you like people who criticize (you) or people who do not criticize you?’ (12) a. *Ni xihuan [piping ni bu piping ni] de ren? you like criticize you not criticize you DE person ‘Do you like people who criticize you (or) people who do not criticize you?’ b. *Ni xihuan [piping ni bu piping] de ren? you like criticize you not criticize DE person ‘Do you like people who criticize you (or) people who do not criticize (you)?’ c. *Ni xihuan [piping bu piping ni] de ren? you like criticize not criticize you DE person ‘Do you like people who criticize (you) (or) people who do not criticize you?’ d. *Ni xihuan [pi- bu pi-ping ni] de ren? you like criticize not criticize you DE person ‘Do you like people who criticize (you) (or) people who do not criticize you?’ Similarly, a disjunctive question may occur within an adjunct clause, but an A- not-A form cannot. (13) a. Ni yinwei [tapiping ni haishi tabu piping ni] er shengqi you because he criticize you or he not criticize you then angry ‘Are you angry because he criticized you or because he did not criticize you?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 331 b. Ni yinwei [ta piping ni haishi bu piping ni] er shengqi you because he criticize you or not criticize you then angry ‘Are you angry because he criticized you or because (he) did not criticize you?’ c. Ni yinwei [ta piping ni haishi bu piping] er shengqi you because he criticize you or not criticize then angry ‘Are you angry because he criticized you or because (he) did not criticize (you)?’ d. Ni yinwei [ta piping haishi bu piping ni] er shengqi you because he criticize or not criticize you then angry ‘Are you angry because he criticized (you) or because (he) did not criticize you?’ (14) a. *Ni yinwei [ta piping ni bu piping ni] er shengqi you because he criticize you not criticize you then angry ‘Are you angry because he criticized you (or) because (he) did not criticize you?’ b. *Ni yinwei [ta piping ni bu piping] er shengqi you because he criticize you not criticize then angry ‘Are you angry because he criticized you (or) because (he) did not criticize (you)?’ c. *Ni yinwei [ta piping bu piping ni] er shengqi you because he criticize not criticize you then angry ‘Are you angry because he criticized (you) (or) because (he) did not criticize you?’ d. *Ni yinwei [tapi- bu pi-ping ni] er shengqi you because he criticize not criticize you then angry ‘Are you angry because he criticized (you) (or) because (he) did not criticize you?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 332 In brief, as argued in Huang (1991b), there is a systematic difference between disjunctive questions with haishi ‘or’ and A-not-A questions without haishi with respect to lexical integrity, preposition stranding, and island constraints. Thus, A- not-A questions cannot be treated as analogous to disjunctive questions. 6.3 Problems with Coordinate Deletion In the following subsections, we will present the problems associated with the coordinate deletion analysis o f the constructions under discussion, namely, the disjunctive and A-not-A questions. 6.3.1 The Directionality Constraint It has been known since Ross (1967) that coordinate deletion is subject to Directionality Constraint to the effect that deletion must go forward if the identical elements in a coordinate structure occur on a left branch o f a tree, but backward if they each occur on a right branch. Huang (1991b) notes that the constraint also applies in Chinese, accounting for the fact that each (a) sentence below can be turned into (b) but not into (c). (15) a. John changge, John tiaowu sing dance ‘John sang and John danced.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 333 b. John changge, tiaowu sing dance ‘John sang and danced.’ c. *changge, John tiaowu sing dance (16) a. John changge, Bill changge sing sing ‘John sang and Bill sang.’ b. John gen Bill (dou) changge and both sing ‘John and Bill sang.’ c. *John changge, Bill (ye/dou) sing also/both Huang (1991b) further notes that the well-formed A-not-A question in (4c) (repeated here), however, clearly does not obey the constraint. (4) c. Ta xihuan naben shu bu xihuan? he like that book not like ‘Does he like that book or does (he) not like (that book)?’ The object naben shu ‘that book’ occurs on a right branch o f the structure, but (4c) is the result o f forward deletion, in violation o f the constraint. This is also the case with the well-formed disjunctive question in (3c), repeated here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 334 (3) c. Ta xihuan naben shu haishi bu xihuan? he like that book or not like ‘Does he like that book or does (he) not like (that book)?’ The above facts show that at least some o f the disjunctive and A-not-A questions cannot be derived by coordinate deletion. 6.3.2 Lexical Integrity The Lexical Integrity Hypothesis, as stated in (5), correctly predicts that the rule o f coordinate deletion cannot delete a subpart o f a word, as illustrated in (17). (17) a. Johndai-le qian-bi genmao-bi bring-Asp lead-pen and hair-pen ‘John brought pencils and brushes.’ b. *John dai-le qian- gen mao-bi bring lead- and hair-pen However, as discussed in section 6.2.1, the A-not-A question in (4e), repeated here, disobeys this principle. (4) e. Ta xi- bu xi-huan naben shu? he li- not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 335 The previous analysis that derives the entire paradigm o f the A-not-A questions in (4) via a single rule o f coordinate deletion thus has to explain why the derivation o f the relevant question may violate the Lexical Integrity Hypothesis. 6.3.3 Preposition Stranding As the example in (18) shows, Chinese does not permit preposition stranding as a result o f coordinate deletion. (18) a. John gen Mary tiaowu, ye gen Mary shuohua with dance also with talk ‘John danced with Mary and talked with Mary.’ b. *John gen [e] tiaowu, ye gen Mary shuohua with dance also with talk ‘John danced with and talked with Mary.’ However, as shown in section 6.2.2, an A-not-A question like (7), repeated here, allows preposition stranding. (7) Ni gen bu gen ta qu? you with not with him go ‘Will you go with him?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 336 The possibility o f preposition stranding with A-not-A questions thus poses problems on the previous analysis that derives all the A-not-A questions via the rule o f coordinate deletion. 6.3.4 Island Constraints We have seen in section 6.2.3 that the distribution and interpretation o f an A-not-A question exhibits island effects. Therefore, the previous analysis that derives A-not- A questions via coordinate deletion has the burden o f accounting for the island facts, since it is generally assumed that island constraints do not apply to deletion processes at all. 6.4 The Proposal It has been shown in the previous sections that A-not-A questions must be separated from disjunctive questions and moreover the former cannot be derived via the process o f coordinate deletion, as argued in Huang (1991b). Huang proposes that while all the disjunctive questions are derived by coordinate deletion, two kinds o f A-not-A questions should be distinguished, one o f which is derived from a simplex sentence with an interrogative Infl constituent that is phonetically realized by a rule o f reduplication, and the other from a structure o f juxtaposed vPs which may be subject to anaphoric ellipsis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 337 In contrast to Huang and the previous linguists, I will argue that not all o f the disjunctive questions are derived from coordinate deletion; instead, some are instances o f VP ellipsis (under the vP structure). Furthermore, I will also argue that what Huang (1991b) terms as “anaphoric ellipsis” is in fact an instance o f VP ellipsis. 6.4.1 The Disjunctive Questions Following traditional analyses, I assume that the disjunctive questions in (3b) and (3d), repeated here, are derived from the full structure in (3a), repeated here, by coordinate deletion. (3) a. Ta xihuan naben shu haishi ta bu xihuan naben shu? he like that book or he not like that book ‘Does he like that book or does he not like that book?’ b. Ta xihuan naben shu haishi bu xihuan naben shu? he like that book or not like that book ‘Does he like that book or does (he) not like that book?’ d. Ta xihuan haishi bu xihuan naben shu? he like or not like that book ‘Does he like (that book) or does (he) not like that book?’ As the readers can verify, the process obeys Ross’ (1967) Directionality Constraint. Note that the disjunctive question in (3e), repeated here, is ruled out because o f the violation o f the Lexical Integrity Hypothesis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 338 (3) e. *Taxi- haishi bu xi-huan naben shu? he like or not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ Next consider the disjunctive question in (3c). (3) c. Ta xihuan naben shu haishi bu xihuan? he like that book or not like ‘Does he like that book or does (he) not like (that book)?’ As discussed in section 6.3.1.1, application o f coordinate deletion to derive a disjunctive question like (3c) is in violation o f the Directionality Constraint. Thus, unlike the other disjunctive questions in (3), the one in (3c) should be derived in some other way. With a closer look, the disjunctive question in (3c) is reminiscent o f the following VP ellipsis sentence. (19) John xihuan naben shu, Bill bu xihuan [e] like that book not like ‘John likes that book; Bill does not like (that book).’ Given the ready availability o f VP ellipsis in the language, I will thus suggest that the disjunctive question in (3c) is in fact an instance o f VP ellipsis, as illustrated by the relevant LF structure in (20). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 339 (20) Ta [v-v xihuan] [y p tv [ n p naben shu]] haishi bu [V .v xihuan] [V P tv [N P e]]? In brief, there are two types o f disjunctive questions. One is derived from a full structure via coordinate deletion, and the other is an instance o f VP ellipsis. 6.4.2 The A-Not-A Questions Following Huang (1991b), I assume that two kinds o f A-not-A questions should be separated, one o f which, what Huang calls [A not AB] type, is derived from a simplex sentence with an interrogative Infl constituent. However, in contrary to Huang (1991b), I will argue that the other, namely, what Huang calls [AB not A] type, is obtained as a result o f VP ellipsis rather than anaphoric ellipsis. 6.4.2.1 Interrogative Infl Huang (1991b) proposes that the [A not AB] type o f A-not-A questions such as (4e), (4d), and (4b) (where A is taken to include the entire vP and B is null) is derived from a simplex structure with an interrogative Infl constituent, as depicted in (21). (4) e. Ta xi- bu xi-huan naben shu? he li- not li-ke that book ‘Does he li-(ke that book) or does (he) not like that book?’ d. Ta xihuan bu xihuan naben shu? he like not like that book ‘Does he like (that book) or does (he) not like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 340 b. Ta xihuan naben shu bu xihuan naben shu? he like that book not like that book ‘Does he like that book or does (he) not like that book?’ (21) Subj I’ I vP V-v VP tv NP Ta [+Q] xihuan naben shu According to Huang, the interrogative Infl is realized by a reduplication rule, which copies a sequence immediately following Infl and inserts the morpheme bu ‘not’ between the original and its copy. Depending on the length o f the reduplicated sequence, the result can be xi-bu-xi, xihuan-bu-xihuan, or xihuan naben shu bu xihuan naben shu. These results correspond to (4e), (4d), and (4b), respectively. Note that the Lexical Integrity Hypothesis is irrelevant in the case o f (4e), since phonological reduplication processes typically do not obey the requirements o f the Lexical Integrity Hypothesis, which obtains only as a principle o f syntax. Also note that the question in (4a), cannot be generated by reduplication on a par with the other [A not AB] type o f A-not-A questions, since the second occurrence o f ta ‘he’ intervenes between A and not-A, thus the ungrammaticality. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 341 (4) a. *Ta xihuan naben shu ta bu xihuan naben shu? he like that book not like that book ‘Does he like that book or does he not like that book?’ 6.4.2.2 Anaphoric Ellipsis or VP Ellipsis? Huang (1991b) further proposes that the [AB not A] type o f A-not-A questions, as exemplified by (4c), is derived from a structure with a coordinate vP o f the form [[AB] not [AB]], which may undergo a process o f anaphoric ellipsis that deletes the second occurrence o f B. (4) c. Ta xihuan naben shu bu xihuan? he like that book not like ‘Does he like that book or does (he) not like (that book)?’ Thus, the A-not-A question in (4c) is obtained as a result o f anaphoric ellipsis that deletes the second occurrence o f naben shu ‘that book.’ However, the anaphoric ellipsis analysis proposed in Huang (1991b) is not articulated enough. It does not specify what elements under what conditions can undergo the process. It seems that apparent NPs as well as vPs can undergo the process, as indicated in (22) and (23), respectively. (22) Ta kai che bu kai [n p e]? he drive car not drive ‘Does he drive?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 342 (23) Ta hui kai che bu hui [v p e]? he know-how-to drive car not know-how-to ‘Does he know how to drive?’ Nevertheless, with a closer look, the sentences in (22) and (23) are the results that can be obtained via the mechanisms that already exist in the language, namely, VP and vP ellipses (cf. chapter 3). In the face o f the argument that anaphoric ellipsis is not articulated and the fact that it simply duplicates precisely the effects o f the already available mechanisms o f ellipsis, I therefore suggest that the process o f anaphoric ellipsis is in fact an instance o f VP ellipsis (in the case o f (22)) or vP ellipsis (in the case o f (23)). If this is the case, the A-not-A question in (4c) is obtained via VP ellipsis. (4) c. Ta [v.v xihuan] [N P naben shu] bu [V - „ xihuan] [N P [e]]? Our line o f speculation is empirically supported by the similarities between the A-not-A question o f the type in (4c) and VP ellipsis sentences, as demonstrated in the following. First, the A-not-A question in (4c) patterns with a VP ellipsis sentence in that it is not subject to Ross’ (1967) Directionality Constraint. (4) c. Ta xihuan naben shu bu xihuan [e]? he like that book not like ‘Does he like that book?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 343 (24) John xihuan tade shu, Bill ye xihuan [e] like his book also like ‘John likes his book, and Bill also likes (his book).’ In each o f (4c) and (24), the elided element, which is traditionally derived by forward deletion, occurs on a right branch. Second, both the A-not-A question o f the type in (4c) and VP ellipsis sentences obey the Lexical Integrity Hypothesis. (25) *Ta xi-huan naben shu bu xi- [e]? he like that book not like ‘Does he like that book?’ (26) * John xi-huan tade shu, Bill ye xi- [e] like his book also like ‘John likes his book, and Bill also likes (his book).’ Third, both the A-not-A question o f the type in (4c) and vP ellipsis sentences disallow preposition stranding.1 (27) *John hui dui tade nuuban shuo shihua bu hui dui [e]? will to his date say truth not will to ‘Will John tell his date the truth?’ 1 The reason why a vP ellipsis sentence is used here is because it is impossible to have a minimal pair of an A-not-A question and a VP ellipsis sentence with regard to preposition stranding, since a relevant VP ellipsis sentence like (i) still retains the main verb shuo ‘say,’ which has raised to v. (i) *John dui tade nuuban shuo shihua, Bill ye dui [e] shuo [e] to his date say truth also to say ‘John told his date the truth, and Bill also to (his date) told (the truth).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 344 (28) *John hui dui tade nuuban shuo shihua, Bill ye hui dui [e] will to his date say truth also will to ‘John will tell his date the truth, and Bill will also to (his date tell the truth).’ Lastly, it has been shown in chapter 3 that postverbal elements such as duration/frequency phrases and predicative complements do not participate in VP ellipsis, as exemplified in (29). It is the same with the A-not-A question o f the type in (4c), as shown in (30). (29) a. John nian shu nian-le san ci, Bill ye nian-le [e] read book read-Asp three times also read-Asp ‘John read three times, and Bill also read.’ b. John nian shu niande hen lei, Bill ye nian-le [e] read book read DE very tired also read-Asp ‘John read so much that he was tired, and Bill also read.’ (30) a. *John nian shu nian-le san ci mei nian [e]? read book read-Asp three times not read ‘Did John read three times?’ b. * John nian shu nian de hen lei mei nian [e]? read book read DE very tired not read ‘Did John read so much that he was tired?’ The ungrammaticality o f the A-not-A questions in (30) is just what is expected under the proposed VP ellipsis analysis. As mentioned above, the interpretation o f the elided VP in the second conjunct does not include that o f the postverbal frequency Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 345 phrase or predicative complement in the first. Thus, the relevant LF representations o f the ill-formed A-not-A questions in (30) are like those given in (31) at a certain point. (31) a. *John [v P [A d v P san ci] [,,P [v P [v.v nian][V P tv [ n p shu]]] three time read book [v [v-v nian][V p t v tA d V p]]]] mei [v p [v-v nian][V P ty [N p shu]]] read not read book b. *John [v P [ p r e c o m p hen lei] [v P (v P [v.v nian][V P t v [N P shu]]] very tired read book [,• [v-v nian][V P tv t P r e C o m p ] ] ] ] mei [v P [V - v nianliyp ty [n p shu]]] read not read book The representations in (31) are ruled out for semantic reasons: the first conjunct questions the frequency or result with respect to an event o f reading, but at the same time, the second conjunct questions the existence o f the event. On the other hand, it is predicted that the representations will become well- formed once the semantic anomaly is eliminated by making the sentences question only the frequency or result with respect to an event. This is a correct prediction, as given in (32), where the frequency phrase or predicative complement is added to the second conjunct. (32) a. John nian shu nian-le san ci mei nian [e] san ci? read book read-Asp three times not read three times ‘Did John read three times?’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 346 b. John nian shu nian de hen lei mei nian [e] de hen lei? read book read DE very tired not read DE very tired ‘Did John read so much that he was tired?’ Similarly, in VP ellipsis sentences like (33), the interpretation o f the second conjunct also includes that o f the frequency phrase or predicative complement that repeats in the conjunct. (33) a. John nian shu nian san ci, Bill ye nian [e] san ci read book read three times also read three times ‘John read three times, and Bill also read three times.’ b. John nian shu nian de hen lei, Bill ye nian-de [e] hen lei read book read DE very tired also read-DE very tired ‘John read so much that he was tired, and Bill also read so much that he was tired.’ In contrast, a frequency phrase or predicative complement participates in vP ellipsis (namely, the ye shi construction), as illustrated in (34). (34) a. John nian shu nian-le san ci, Bill ye shi [e] read book read-Asp three times also be ‘John read three times, and Bill also (read three times).’ b. John nian shu nian de hen lei, Bill ye shi [e] read book read DE very tired also be ‘John read so much that he was tired, and Bill also (read so much that he was tired).’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 347 Given our claim that anaphoric ellipsis involved in the relevant A-not-A questions is an instance o f VP or vP ellipsis, we expect that a frequency phrase or predicative complement also participates in the relevant A-not-A questions that are obtained via vP ellipsis. This is the case, as indicated in (35), where you ‘have’ serves as an auxiliary licensing the elided vP. (35) a. John nian shu nian-le san ci meiyou [e]? read book read-Asp three times not-have ‘Did John read three times?’ b. John nian shu nian de hen lei meiyou [e]? read book read DE very tired not-have ‘Did John read so much that he was tired?’ The relevant LF representations o f the A-not-A questions in (35) are given in (36). (36) a. John [„ P [A d v P san ci] [v P [A d j u n c t nian shu] [,, [V .v nian][V P tv tA d v P ]]]] three time read book read [tea meiyou][v P [A d v P san ci] [v P [A d j u n c t nian shu] [v . [V .v nian][V P tv tA d v P ]]]] not-have three time read book read b. John [v P [P r e C o m p hen lei] [v P [A d j u n c t nian shu] [v . [V .v nian] very tired read book read [vp tv t p r e C o m p ] ] ] ] ^ meiyou] [v P [ p r e C o r a p hen lei] [v P [A d j u n c t nian shu] not-have very tired read book [„ • [v-v nian][V P tv t P r e C o m p ] ] ] ] read Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 348 In brief, it has been argued in this section that the so-called anaphoric ellipsis as proposed in Huang (1991) for the [AB not A] type o f A-not-A questions is in fact an instance o f VP or vP ellipsis. 6.5 Conclusion To conclude, it has been argued that the paradigm o f the disjunctive questions in (3) is derived from two sources. The questions in (3b) and (3d) are the results o f coordinate deletion whereas the question in (3 c) is an instance o f VP ellipsis. Similarly, the paradigm o f the A-not-A questions in (4) is also derived from two sources. As proposed in Huang (1991), the questions in (4b), (4d), and (4e), namely, the [A not AB] type o f A-not-A questions, are derived via the interrogative Infl that is phonetically realized by reduplication. However, in contrast to Huang (1991), it is argued that the question in (4c), namely, the [AB not A] type o f A-not-A questions, is obtained via VP ellipsis rather than anaphoric ellipsis. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 349 References Abney, S. (1987) The English Noun Phrase in Its Sentential Aspect, Doctoral dissertation, MIT. Aoun, J. (1986) Generalized Binding, Foris, Dordrecht. Aoun, J. and Y.-H. A. Li (1989) “Scope and Constituency,” Linguistic Inquiry 20, 141-172. Battistella, E. (1989) “Chinese Reflexivization: A Movement to Infl Approach,” Linguistics 27, 987-1012. Battistella, E. and Y. Xu (1986) “Remarks on the Reflexive in Chinese,” Ms., University o f Alabama, Birmingham. Borer, H. (1988) “On the Morphological Parallelism between compounds and Constructs,” Yearbook o f Morphology 1, 45-65. Borer, H. (1991) “The Causative-Inchoative Alternation: A Case Study in Parallel Morphology,” The Linguistic Review 8, 119-158. Borer, H. (1993) Parallel Morphology, Book Manuscript, University o f Massachusetts at Amherst. Bresnan, J. (1971) “Note on the Notion ‘Identity o f Sense Anaphora’,” Linguistic Inquiry 2, 589-597. Chao, W. (1987) On Ellipsis, Doctoral dissertation, University o f Massachusetts, Amherst. Cheng, L. (1990) “Aspects o f Ba-Construction,” Ms., MIT. Chierchia, G. (1996) “Reference to Kinds across Languages,” Ms., University o f Milan. Chierchia, G. (1997) “Plurality o f Mass Nouns and the Notion o f ‘Semantic Parameter’,” Ms., University o f Milan. Chomsky, N. (1981) Lectures on government and Binding, Foris, Dordrecht. Chomsky, N. (1986) Barriers, MIT Press, Cambridge, Massachusetts. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 350 Chomsky, N. (1993) “A Minimalist Program for Linguistic Theory,” in The View from Building 20, Hale K. and S. J. Kayser (eds.), Cambridge, Massachusetts: The MIT Press, 1-52. Chomsky, N. (1995) The Minimalist Program, Cambridge, Massachusetts: The MIT Press. Cole, P., G. Hermon, and L.-M. Sung (1990) “Principles and Parameters o f Long- Distance Reflexives,” Linguistic Inquiry 21,1-22. Dahl, O. (1974) “How to Open a Sentence: Abstraction in Natural Language,” Logical Grammar Reports 12, University o f Goteberg. Dalrymple, M., S. M. Shieber, and F. C. N. Pereira (1991) “Ellipsis and Higher- Order Unification,” Linguistics and Philosophy 14, 399-452. Doron, E. (1998) “V-Movement and VP-Ellipsis,” in Lappin, S. and E. Benmamoun (eds.), Fragments: Studies in Ellipsis and Gapping, London. Fiengo, R. and R. May (1994) Indices and Identity, Cambridge: MIT Press. Fu, J.-Q. (1994) On Deriving Chinese Derived Nominals: Evidence for V-to-N Raising, Doctoral dissertation, University o f Massachusetts, Amherst. Grimshaw, J. (1990) Argument Structure, Cambridge, Mass: MIT Press. Grinder, J., and P. Postal (1971) “Missing Antecedents,” Linguistic Inquiry 2, 269- 312. Hankamer, J. and I. Sag (1976) “Deep and Surface Anaphora,” Linguistic Inquiry 1, 391-428. Hoji, H. (1990) “Theories o f Anaphora and Aspects o f Japanese Syntax,” Ms., University o f Southern California, Los Angeles. Hoji, H. (1993) “Null Object and Sloppy Identity in Japanese,” Ms., University o f Southern California, Los Angeles. Hoji, H. (1995) “Demonstrative Binding and Principle B,” NELS 25, 255-271. Hoji, H. (1997a) “Sloppy Identity and Formal Dependency,” WCCFL 15,209-223. Hoji, H. (1997b) “Sloppy Identity and Principle B,” in Bennis H., P. Pica, and J. Rooryck (eds.), Atomism and Binding, Foris Publications. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 351 Hoji, H. (1998a) “Null Object and Sloppy Identity in Japanese,” Linguistic Inquiry 29, 127-152. Hoji, H. (1998b) “Formal Dependency, Organization o f Grammar, and Japanese Demonstratives,” Ms., University o f Southern California. Hoji, H. (1998c) “Surface and Deep Anaphora, Sloppy Identity, and Experiments in Syntax,” Ms., University o f Southern California. Hoji, H. (1998d) “Surface and Deep Anaphora, Sloppy Identity, and Experiments in Syntax,” to appear in Barss A. and T. Langendoen (eds.), Explaining Linguistics, Blackwell. Huang, C.-T. J. (1982) Logical Relations in Chinese and the Theory o f Grammar, Doctoral dissertation, MIT. Huang, C.-T. J. (1984) “On the Distribution and Reference o f Empty Pronouns,” Linguistic Inquiry 15, 531-574. Huang, C.-T. J. (1987) “Remarks on Empty Categories in Chinese,” Linguistic Inquiry 18, 321-337. Huang, C.-T. J. (1988a) “Comments on Hasegawa’s Paper,” in Proceedings o f Japanese Syntax Workshop: Issues on Empty Categories, Connecticut College, N ew London. Huang, C.-T. J. (1988b) “Verb-Second in German and Some AUX Phenomena in Chinese,” in A. Bramkamp et al. (eds.), Chinese-Western Encounter: Festschrift fo r Franz Giet, SVD on his 85th Birthday, Fu Jen University, Taipei. Huang, C.-T. J. (1989) “Pro-Drop in Chinese: A Generalized Control Theory,” in Jaeggli O. and K. Safir (eds.), The Null Subject Parameter, Dordrecht: Kluwer Academic Publishers, 185-214. Huang, C.-T. J. (1991a) “Remarks on the Status o f the Null Object,” in R. Freidin (ed.), Principles and Parameters in Comparative Grammar, Cambridge: MIT Press, 56-76. Huang, C.-T. J. (1991b) “Modularity and Chinese A-Not-A Questions,” in Georgopolous C. and R. Ishihara (eds.), Interdisciplinary Approaches to Language: Essays in Honor o f S.-Y. Kuroda, Kluwer Academic Publishers, Dordrecht, 305-332. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 352 Huang, C.-T. J. (1993) “Reconstruction and the Structure o f VP: Some Theoretical Consequences,” Linguistic Inquiry 24, 103-138. Huang, C.-T. J. (1994a) “Verb Movement and Some Syntax-Semantics Mismatches in Chinese,” in Li J.-K. P., C.-R. Huang and C.-C. J. Tang (eds.), Chinese Languages and Linguistics 2, 587-613, Taipei: Academia Sinica. Huang, C.-T. J. (1994b) “More on Chinese Word Order and Parametric Theory,” in Lust B., M. Suner, and J. Whitman (eds.), Syntactic Theory and First Language Acquisition: Cross-Linguistic Perspectives 1, Heads, Projections, and Learnability, 15-35, Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. Huang, C.-T. J. (1997) “On Lexical Structure and Syntactic Projection,” in Tsao F.- F., and H. S. Wang (eds.), Chinese Languages and Linguistics 3, 45-89, Taipei: Academia Sinica. Huang, C.-T. J. and C.-C. J. Tang (1991) “The Local Nature o f the Long Distance Reflexive in Chinese,” in Long Distance Anaphora, Koster J. and E. Reuland, (eds.), Cambridge: Cambridge University Press, 263-282. Huang, Y.-H. (1985) “Chinese Reflexives,” in Studies in English Literature and Linguistics 10, National Taiwan Normal University, Taipei. Johnson, K. (1994) “Bridging the Gap,” Ms., University o f Massachusetts, Amherst. Kim, S. (1995) “Sloppy/Strict Identity and V-Raising: A Reply to Otani & Whitman,” Ms., University o f Washington. Kim, S. (1999) “Sloppy/Strict Identity, Empty Objects, and NP Ellipsis,” Journal o f East Asian Linguistics 8.4, 255-284. King, H. (1970) “On Blocking the Rules for Contraction in English,” Linguistic Inquiry 1, 134-136. Kitagawa, Y. (1986) Subjects in Japanese and English, Doctoral dissertation, University o f Massachusetts, Amherst. Kitagawa, Y. (1989) “Deriving and Copying Predication” in Proceedings o f Cornell’ s NELS 19, 279-300. Kortobi, I. (1997) “Gapping and VP Deletion in Moroccan Arabic,” Ms., University o f Southern California. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 353 Krifka, M. (1995) “Common Nouns: A Contrastive Analysis o f Chinese and English,” in The Generic Book, Carlson G. and F.-J. Pelletier (eds.), Chicago: University o f Chicago Press. Kuo, S.-Z. (1996) “Null Objects in Chinese: Covert Pronominal, But Not Variables Nor Silent Epithets,” NACCL 1, 161-170. Kuroda, S.-Y. (1988) “Whether We Agree Or Not: A Comparative Syntax o f English and Japanese,” in Japanese Syntax and Semantics: Collected Papers, Dordrecht: Kluwer Academic Publishers, 315-357. Kuroda, S.-Y. (1992) Japanese Syntax and Semantics, Dordrecht: Kluwer Academic Publishers. Larson, R. (1988) “On the Double Object Construction,” Linguistic Inquiry 19, 335- 392. Lasnik, H. (1976) “Remarks on Coreference,” Linguistic Analysis 2, 1-22. Lasnik, H. (1982) “Is Condition C Necessary?” Talk given at the Cornell Conference on Government and Binding Theory, Cornell University. Lasnik, H. (1986) “On the Necessity o f Binding Conditions,” Paper presented at the Princeton Workshop on Comparative Syntax, Princeton University. Lasnik, H. (1994) “Verbal Morphology: Syntactic Structures Meets the Minimalist Program,” Ms., University o f Connecticut. Li, H.-Y. A. (1990) Order and Constituency in Mandarin Chinese, Dordrecht: Kluwer Academic Publishers. Li, C. N. and S. Thompson (1981) Mandarin Chinese: A Functional Reference Grammar, University o f California Press. Li, H.-Y. A. (1996) “Minimalism and Optionality,” Ms., University o f Southern California. Luu, S.-X. et al. (1981) Xiandai Hanyu Babai Ci (Eight Hundred Words in Modem Chinese), Shangwu Yinshuguan, Beijing. Maling, J. and S. Kim (1992) “Case Assignment in the Inalienable Possession Constmction in Korean,” Journal o f East Asian Linguistics 1, 37-68. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 354 May, R. (1985) Logical Form: Its Structure and Derivation, MIT Press, Cambridge, Massachusetts. McCawley, J. D. (1994) “Remarks on the Syntax o f Mandarin Yes-No Questions,” Journal o f East Asian Linguistics 3,179-194. Mei, K. (1972) Studies in the Transformational Grammar o f Modern Standard Chinese, Doctoral dissertation, Harvard University. Mei, K. (1978) “On Verb-Phrase Complements in Chinese,” in A Festschrift fo r Professor Chu Wan Li, Taipei, Taiwan. Otani, K. and J. Whitman (1991) “V-Raising and VP-Ellipsis,” Linguistic Inquiry 22, 345-358. Partee, B. H. (1978) “Bound Variables and Other Anaphors,” in D. Waltz (ed.), Proceedings ofTINLAP 2, University o f Illinois, 79-85. Paul, W. (1996) “Verb Raising in Chinese,” NACCL 1, 260-276. Pollock, J.-Y. (1989) “Verb Movement, Universal Grammar, and the Structure o f IP,” Linguistic Inquiry 20, 365-424. Reinhart, T. (1983) “Coreference and Bound Anaphora: a Restatement o f the Anaphora Questions,” Linguistics and Philosophy 6, 47-88. Rizzi, L. (1986) “Null Objects in Italian and the Theory o f P r o f Linguistic Inquiry 17, 467-558. Ross, J. R. (1967) Constraints on Variables in Syntax, Doctoral dissertation, MIT. Ross, J. R. (1969) “On the Cyclic Nature o f Pronominalization,” in Reibel D. and S. Schane (eds.), Modern Studies in English, Prentice-Hall, Englewood Cliffs, New Jersey. Sag, I. (1976) Deletion and Logical Form, Doctoral dissertation, MIT. Selkirk, L. (1972) The Phrase Phonology o f English and French, Doctoral dissertation, MIT. Simpson, A. and X.-Z. Z. Wu (1999) “On PF Clausal Raising- Evidence from Tone Sandhi Phenomena,” Paper presented in LSA, Los Angeles. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 355 Stillings, J. (1975) “The Formulation o f Gapping in English as Evidence for Variable Types in Syntactic Transformation,” Linguistic Analysis 1.3. Tai, J. (1984) “Verbs and Times in Chinese: Vendler’s Four Categories,” CLS Parasession on Lexical Semantics, Chicago, Illinois, 289-296. Tai, J. and J. Chou (1974) “On the Center o f Predication in Chinese Resultative Verb Compounds,” Mimeographed, Southern Illinois University. Takano, Y. (1996) Movement and Parametric Variation in Syntax, Doctoral dissertation, University o f California, Irvine. Tang, T.-C. C. (1978) “Double Object Constructions in Chinese,” in Cheng, R. C., Y. C. Li, and T. C. Tang (eds.), Proceedings o f Symposium o f Chinese Linguistics, LSA, Student Books, Taipei, 197-210. Tang, S.-W. (1998) Parametrization o f Features in Syntax, Doctoral dissertation, University o f California, Irvine. Tenny, C. (1989) “The Aspectual Interface Hypothesis,” Lexicon Project Working Paper 31, Center for Cognitive Science, MIT. Thompson, S. A. (1973) “Transitivity and the Ba Construction in Mandarin Chinese,” Journal o f Chinese Linguistics 1.2. Tomioka, S. (1996) “The Laziest Pronouns,” Paper presented in Japanese and Korean Linguistics 7, UCLA. Tomioka, S. (1997) “Property Anaphora and the Typology o f P r o f Paper presented in Theoretical East Asian Linguistics Workshop, UC-Irvine. Tomioka, S. (1998) “The Semantics o f Japanese Null Pronouns in a Crosslinguistic Context,” Paper presented in University o f Delaware. Tomioka, S. (1999) “A Sloppy Identity Puzzle,” Natural Language Semantics 1, 217-241. Travis, L. (1988) “The Syntax o f Adverbs, in McGill Working Papers in Linguistics: Special Issue on Comparative Germanic Syntax. Wang, J. and J. Stillings (1984) “Chinese Reflexives,” in Proceedings o f the First Harbin Conference on Generative Grammar, X.-Y. Li et al. (eds.), Harbin: Heilongjiang University Press, 100-109. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 356 Wang, L. (1947) Zhongguo Xiandai Yufa (Modem Chinese Grammar), Shanghai: Zhonghua Shuju. Whitman, J. (1987) “Discourse Ellipsis and the Identity o f Zero Pronouns,” in Selected Papers from SICOL-1986, Linguistic Society o f Korea, Seoul. Williams, E. (1977) “Discourse and Logical Form,” Linguistic Inquiry 8, 101-139. Wu, X.-Z. Z. (1998) “The Syntax and Interpretation o f Sentence-Final DE,” Ms., University o f Southern California. Xu, L.-J. (1986) “Free Empty Category,” Linguistic Inquiry 17, 75-93. Yasui, M. (1997) “Identification o f Ellipses and Other Empty Categories,” in Liu, C.-S. L. and K. Takeda (eds.), UCI Working Papers in Linguistics 3, 227-255. Yoon, J. H.-S. (1989) “The Grammar o f Inalienable Possession Constructions in Korean, Mandarin and French,” in Harvard Studies in Korean Linguistics III, S. Kuno et al. (eds.), 357-368, Seoul: Hanshin. You, Y.-L. (1996) “Interpreting Chinese Zero-Anaphora: Examining the Recovery Rules in Text,” NACCL 1, 395-412. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Grammaticalization and the development of functional categories in Chinese

PDF

Form and meaning: Negation and question in Chinese

PDF

Beyond words and phrases: A unified theory of predicate composition

PDF

A comparative study of focus constructions

PDF

Aspects of questions in Japanese and their theoretical implications.

PDF

Children's scope of indefinite objects

PDF

Basolateral transport of nucleosides in rabbit tracheal epithelial cells: An equilibrative nucleoside transporter outcome

PDF

Asymmetry of scope taking in wh -questions

PDF

Comparing social studies planning processes of experienced teachers and student teachers

PDF

Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic

PDF

Discourse functional units: A re-examination of discourse markers with particular reference to Spanish

PDF

Connectionist phonology

PDF

Counter -matching in nested case -control studies: Design and analytic issues

PDF

Abstract case in Chinese

PDF

Children in transition: Popular children's magazines in late imperial and early Soviet Russia

PDF

Locatives in Chinese and Japanese: distribution, case assignment and c-command

PDF

"Master of many tongues": The Russian Academy Dictionary (1789--1794) as a socio -historical document

PDF

Antibody-cytokine/chemokine fusion proteins in the immunotherapy of solid tumors

PDF

The Japanese demonstratives ko, so and a

PDF

Gossip, letters, phones: The scandal of female networks in film and literature

Asset Metadata

Creator Li, Hui-Ju Grace (author)

Core Title Ellipsis constructions in Chinese

School Graduate School

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Li, Audrey (committee chair), Hoji, Hajime (committee member), Kim, Nam-Kil (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-548547

Unique identifier UC11339838

Identifier 3094355.pdf (filename),usctheses-c16-548547 (legacy record id)

Legacy Identifier 3094355.pdf

Dmrecord 548547

Document Type Dissertation

Rights Li, Hui-Ju Grace

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA