UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Object noun phrase dislocation in Mandarin Chinese Qu, Yanfeng 1994-06-11

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


831-ubc_1995-983418.pdf [ 4.85MB ]
JSON: 831-1.0088905.json
JSON-LD: 831-1.0088905-ld.json
RDF/XML (Pretty): 831-1.0088905-rdf.xml
RDF/JSON: 831-1.0088905-rdf.json
Turtle: 831-1.0088905-turtle.txt
N-Triples: 831-1.0088905-rdf-ntriples.txt
Original Record: 831-1.0088905-source.json
Full Text

Full Text

OBJECT NOUN PHRASE DISLOCATION IN MANDARIN CHINESEbyYanfeng QuB.A., Shandong Teachers’ University, 1982M.A., Beijing Foreign Studies University, 1987A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Department of Linguistics)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIADecember 1994©Yanfeng Qu 1994In presenting this thesis in partial fulfilmentof the requirements for an advanceddegree at the University of British Columbia, Iagree that the Library shall make itfreely available for reference arid study. Ifurther agree that permission for extensivecopying of this thesis for scholarlypurposes may be granted by the head of mydepartment or by his or her representatives.It is understood that copying orpublication of this thesis for financial gain shallnot be allowed without my writtenpermission.(Signature)______________Department of_________________The University of BritishColumbiaVancouver, CanadaDate &L2t 30,/99DE-6 (2/88)11ABSTRACTThis dissertation studies leftward dislocation of object Noun Phrases in Mandarin Chinesewithin the framework of Government and Binding theory. Although the canonical word orderin Chinese is S(ubject)-V(erb)-O(bject), it also exhibits OSV and SOV word orders. After anintroduction in Ch. 1, I discuss OSV constructions in Ch.2. I argue that the S-initial object NPis moved there, since its association with a gap in the canonical object position obeys thesubjacency condition. Based on several diagnostic tests, I propose that Chinese has two kindsof short-distance NP fronting: one is A’-movement and the other is A-movement. Adopting theSplit Infi Hypothesis, I postulate a fully articulated clause structure for Chinese. In particular,I propose that the fronted NP in A-movement lands in [Spec AgrOP] as a kind of overt raising,while the one in A’-movement further leaves that spec position and is CP-adjoined. I alsoexamine long-distance NP fronting, showing that it is invariably A’-movement.In Ch. 3, I investigate object shift, which yields SOV constructions. I argue that thissyntactic process represents a type of A-movement, not A’-movement as concluded in previousstudies. Specifically, I propose that the subject NP and the object NP in this construction overtlyraise to [Spec AgrSP] and [Spec AgrOP] respectively.In Ch.4, I examine the interactions between an object wh-NP and dou, the adverb ofuniversal quantification. I propose that wh-phrases, like indefinites, can be eitherpresuppositional or existential. If they are within VP (i.e. remain postverbal), they are subjectto existential closure and get an existential/interrogative reading. If, however, they are outsideVP (i.e. shifted to the left of dou), they define the range of the quantifier dou and obtain thepresuppositional/universal reading. The conclusion is that there is a strict correlation betweenthe S-structure positions of the wh-phrase and its interpretations. The exhaustive list reading of111the in-situ wh-object associated with the interrogative reading is derived from the fact that it isin the scope of dou at S-structure.In Ch. 5, I summarize the major findings of this dissertation and raise several issuesforfuture research.ivTABLE OF CONTENTSAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivAcoledgeent . . . . . . . . . . . . . . . . . . . . . . . . •....... . . . . . . . . . . .vuChapterl. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .’reliniinaries . . . . . . . . . . . . . . •Theoretical Background . . . . . . . .A Brief Introduction to Mandarin ChineseAn Outline of the Dissertation . . . .. .1• • . . . . . • . . . . . . . . . . . .3• . • . • . •.•..... . . . . .5• . . . . . . . . . . . . . . . • • . .9Chapter 2. Object NP Fronting . . . . • . . . . • • . . • . . • • . . . . . . . • . . . 1229313435353639434450525657586063• . . . . • . • • • . 1313232.1. The S-initial NP: Base-generation vs. Movement .2.1.1. Relating to a Gap in a Relative Clause2.1.2. Relating to a Gap in a Sentential Subject2.1.3. Relating to a Gap in a Sentential Complement to an NP2.1.4. Relating to a Gap in the Fronted Object2.1.5. Summary2.2. Types of Movement and Their Landing Sites2.2.1. Two Types of Movement2.2.1.1. A’-movement2.2.1.2. A-movement2.2.2. Landing Sites2.2.2.1. The Landing Site of A-movement:[Spec AgrOP] The Landing Site of A’-movement: CP-adjunction . . Long Distance Fronting2.3. Residual Issues • . • • . . . • . . • • • • .2.3.1. Subjacency Effect and Empty Operator Movement2.3.2. Subjacency Suspension2.3.2.1. Review of Previous Analyses2.3.2.2. A’-position Voids Barrierhood2.4. Summary . . . . . . . . . . . • • • • • • • . . . 66VChapter3. ObjectlPShift . . . . . . . . . . . . . . . . . . .673.1. Iroperties of Object Shift . . . . . . . . . . . . . . . . . . . . . . . . . 673.1.1. Semantic Restrictions 673.1.2. Object Shift and BA Construction 733.2. Interaction of the Shifted Object with Other Constituents. . . . . . . . . 763.2.1. Interaction with Adverbs 763.2.2. Interaction with Modals 813.2.3. Interaction with Negators 833.3. Object NP Shiftand its Landing Site...... . . . . . . . . . . . . . . . . . . 863.3.1. Object Shift as A-movement 873.3.2. The Shifted NP in [Spec AgrOP] 933.3.3. The Definiteness of the Shifted NP and the Postulation of Agr in Chinese973.3.4. Evaluation of the Proposal 1003.4. Suniinary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Chapter 4. Wh-object NP Shift and Dou Interaction 1114.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1114.2.AnalysisoftheWh-object ............. 1174.2.1. Diesing’s Mapping Hypothesis 1174.2.2. The S-structure Position of the Wh-object and its Interpretations 1204.2.3. The Landing Site of the Shifted Wh-object 1284.2.3.1. Reviewof Cheng (1991) 1294.2.3.2. Wh-object NP Shift as A-movement and its Landing Site 1314.3. 14ore Case Studies . . . . . . . . . . . . . . . . . . . . . 1424.3.1. Wh-phrase as the Subject 1424.3.2. The Wh-NP Downstairs and its Interrogative/List Reading 1444.3.3. The Wh-NP Upstairs and its Universal Reading 1474.4. Summary . . . . . . . . . . . . . . . 155Chapter 5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1575.1. Research Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . 1575.2. Residual Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1615.2.1. Optionality of Definite NP Dislocation 161vi5.2.2. Feature Checking and the Tree-splitting Model.1635.2.3. The Licensing Puzzle of Object Shift 1645.2.4. The Interpretative Puzzle of Object Shift 167R.FF1?RF1”CFS . . . . . . . . . . . . . . . . 171viiACKNOWLEDGEMENTI would like to thank the members of my thesis committee, Henry Davis (Chair andSupervisor), Hamida Demirdache, Mike Rochemont and Leslie Saxon for their encouragement,professional advice and various kinds of assistance for this research. Henry Davis has beensupportive and helpful throughout the writing of this thesis. His volunteered reading course andsustained interest in Chinese syntax helped me to decide the path of my dissertation research.Working under his supervision is defmitely an enriching intellectual experience. I am gratefulto Hamida Demirdache for her critical acumen and her constant and constructive input for thisresearch. Her influence on my syntactic thinking is palpably felt throughout this thesis. Myinterest in GB syntax was initiated by Mike Rochemont, and I have continued to benefit fromhis enthusiasm and expertise in this field over the years. Especially at the preliminary stage ofthis research, he helped me to shape many of the ideas presented here. I gratefully acknowledgethe financial support he provided me (January 1993 - April 1993 and September 1993 - March1994) as his Research Assistant (SSHRC #410-92-1379 awarded to him). I am indebted to LeslieSaxon, who, from the other side of the waters, is always ready to come to my rescue wheneverI call for help. Her generous, disciplined and detailed attention to the thesis has improved itimmeasurably. It is a great honour to have James Huang as my External Examiner and I amvery thankful to him for his extremely insightful comments on this thesis.I have profited greatly from the linguistics community at UBC. I thank David Ingram forhis regular meetings with me in the Spring of 1994 and for his valuable advice on theadministrative aspect of finishing a degree. M. Dale Kinkade is always ready to stop his ownwork to listen to my ‘troubles’ and to offer me generous help and advice. Pat Shaw has beenencouraging and supportive from the very beginning of my studies at UBC, and I regard it asa privilege to be her student. I also thank Guy Carden, Rose-Marie Dechaine and DougPulleyblank for their teaching and professional advice. I am grateful to Ken Reeder for providingme with opportunities to apply my knowledge in linguistics to practice. Thanks are also due tomembers of the student population for their moral support and comradery, especially to NilocaBessell, Eleanor Blain, Susan Blake, Helmi Braiches, Ping Jiang-King, Lisa Matthewson,Monica Sanchez, William Thompson, and Akihiko Uechi. I am grateful to Carmen de Silva foradministration in the department and for her friendship.My family have been of invaluable support over the years. I am immensely indebted tomy parents, Zuo Chuan-shi and Qu Xiu-ying, for bringing me up and for all the things that theyhave done for me. I thank my sister and her family for being so understanding and supportive.Last, but by no means the least, I would like to thank my wife, Shaopeng Lu, for beingincredibly encouraging, patient and helpful in every possible way. This dissertation is dedicated,with love, to all of them.1Chapter 1 Introduction1.1. PreliminariesThis dissertation studies leftward dislocation of an object Noun Phrase in MandarinChinese within the theoretical framework of Government and Binding (GB) theory. Althoughthe canonical word order in Chinese is SVO (Li and Thompson 1981, Huang 1982, Travis1984), it is by no means the only possible S-structure word order. In particular, the object NPcan appear S-initially as in (2), resulting in OSV word order; or it can appear in a position afterthe subject but before the verb as in the SOy order shown in (3):(1) wo kan-guo zheben shu. (SVO: canonical word order)I read-asp this book‘I have read this book.’(2) zheben shu, wo kan-guo. (OSV)this book I read-asp‘This book, I have read.’(3) wo zheben shu kan-guo. (SOV)I this book read-asp‘This book, I have read.’For ease of exposition, I call cases like (2) ‘object NP fronting’ and those like (3) ‘object NPshift’. In both cases, the interpretation of the moved object NP is the same in terms of truthconditions as when the object is in-situ. This contrasts with cases which contain a wh-object andan adverb dou ‘all’. In these cases, different positions of the wh-object with respect to the adverbresult in different interpretations. As is shown in (4a) below, when the wh-object is in situ tothe right of the adverb, it has the interrogative reading; however, when it is shifted to the leftof the adverb as in (4b), it obtains the universal reading:2(4) a. Wh-object to the right of dou:John dou chi-guo shenme?John all eat-asp what‘What has John eaten?’ (interrogative reading)b. Wh-object to the left of dou:ta shenme1dou chi-guo.he what all eat-asp.‘He has eaten everything.’ (universal reading)NOT: ‘What has he eaten?’Sentences like (2), (3) and (4) constitute the empirical cases for investigation in the followingthree chapters. Specifically, I will address the following questions:(A) In the literature, the S-initial NP in (2) has been analyzed either as being moved there(cf. Huang 1982, Li 1990 among others), or as being base-generated (cf. Xu and Langendoen1985, Cheng 1991). The first question I address is which of these two conflicting analyses ispreferable. If there is movement involved, what kind of movement does this syntactic processrepresent? Is it uniformly A’-movement as traditionally held (cf. Huang 1982), or can it also beA-movement as in Mahajan’s account of Hindi (1990)? Furthermore, given the current Split InfiHypothesis (Pollock 1989, Chomsky 1992), what is the syntactic position of the fronted NP?(B) Object NP shift, as shown in (3), has not received as much attention in Chinesesyntax as NP fronting. I first examine what kinds of NPs are eligible to undergo this syntacticprocess. I then apply several diagnostic tests commonly used in the GB framework to determinethe nature of this movement and its landing site.(C) For cases like (4), I first establish a descriptive generalization for the S-structurepositions of wh-objects under the interrogative and the universal readings. I then address thequestion of the role the syntactic representation plays in the derivation of the semanticrepresentation of wh-NPs. Specifically, I consider the relation between the S-structure3representation and the LF representation for the interrogative and quantificational sentencesrespectively.In the rest of this chapter, I will briefly discuss the theoretical assumptions I adoptthroughout this dissertation and introduce some syntactic characteristics of Mandarin Chinese.1.2. Theoretical BackgroundThe analysis in this dissertation is presented within the Government and Binding syntacticmodel, developed in Chomsky (1981, 1986, 1991, 1993). This theory aims to account forcomplex distributional problems through the interaction of relatively autonomous modules ofgrammar. Each module is characterized by a small number of simple, universal, but in somecases parameterized principles. The particular model of grammar I adopt is represented in(5):1(5) D-structureS-structureI \Phonetic Form (PF) Logical Form (LF)D-structure is related to S-structure by the general transformational rule ‘Move a’. S-structureis related to the interpretive components, PF and LF, which provide the interface betweenformal grammar and something in the ‘real world’: acoustic/articulatory properties on one hand,and semantic properties on the other.‘Differing from Chomsky (1993) who eliminates D-structure and S-structure, I assume the conventional T-modelof GB syntax which postulates the existence of both structures. Nothing in this thesis crucially hinges on thisdifference.41.2.1. LexiconThe lexicon is the component of the grammar where lexical items are stored, along withidiosyncratic information about them, such as their category. There exists a distinction betweentwo major types of syntactic category: lexical categories suchas nouns, verbs and adjectives,and functional categories. The latter group constitute closed class items, which are generallyphonologically and morphologically dependent and sometimes phonologically empty. These twocategory types are crucially exploited by the Split Infi Hypothesis to be discussed in detail inCh. Phrase StructurePhrase structure is projected from elements which are selected from the lexicon.Projections are assumed to conform to the following schema (Chomsky 1986):(6) XP/ \YPx,Ix zPIn this schema, all phrases have a tri-level structure: the head X and its complement ZP forman X’-level constituent and the specifier YP is attached at the higher XP level. Chomsky (1993)proposes that there are only a small number of syntactically relevant relations within thisstructure. One is the head-specifier relation, which is exemplified in the tree by the relationshipbetween X and YP. Another holds between different heads in a tree, such as X and Z, the headof its complement ZP. In addition, YP and ZP are relevant for binding purposes. One furtherkind of structural relationship to which I will refer in this dissertation is ‘c-command’, whichcan be defined as follows: a c-commands B iff a does not dominate B and every ‘r that dominates5a dominates B (Chomsky 1986b:8). In the tree in (6), X c-commands ZP and everythingcontained in ZP.1.2.3. Checking Theory and MovementChomsky (1993) proposes that lexical items are inserted into syntactic structures withinflectional features already attached. During the course of the derivation, each feature mustbechecked, or licensed under identity with a feature on another element within the structure. Thechecking requirements of a lexical item can potentially motivate syntactic movement (see alsoChomsky and Lasnik 1991). Although Chinese has no overt morphology, I adopt the feature-checking mechanism as a null hypothesis. In particular, I assume that the in-situ object inChinese raises to [Spec AgrOP] at LF for feature checking.This section has presented a brief introduction to some of the notions which are centralto GB theory and relevant to the discussion in later chapters. I delay presenting a more detaileddiscussion of certain concepts and assumptions until later.1.3. A Brief Introduction to Mandarin ChineseIn this section, I review some major properties of Mandarin Chinese and discussassumptions which are relevant to the dislocation of object NPs.1.3.1. Word OrderHuang (1982) correctly points out that Chinese is head-final except for subeategorizedcomplements. Within NPs, however, either Ns do not have subcategorized complements andhence are always head-final, or N is simply an exception to the generalization that6subcategorized complements follow their heads. In brief, Huang’s observation captures the factthat Chinese is VO, prepositional and N-final.Although SVO is the canonical word order in Chinese, OSV and SOy are also allowedunder some specific conditions, for instance, when the object NP is definite or generic.Indefinite NPs are excluded from this process:(7)*yibenshu wo kan-guo.one book I read-asp*‘Onebook, I have read.’In later chapters, I will provide an account for why such a constraint holds in Chinese.1.3.2. ASPPExcept for two controversial future tense markers (see Y. A. Li 1990:22), Chinese hasno overt markers for tense, only postverbal aspect markers: the (durative), le ((im)perfective)and guo (perfective). Given these facts, I assume that TENSE in Chinese is syntacticallynon-existent (at least in non-future sentences). Instead, it has ASP as the head of the clause (cf.Cheng 1991). Adopting the Split Infi Hypothesis, I further assume that Chinese has AgrSP andAgrOP despite the absence of morphological evidence. I delay a detailed discussion of Chineseclausal structure until Ch. LF Movement of “V+Aspect Marker”Adapting Chomsky (1993), I assume that the aspect marker is generated as an affix onthe verb at D-structure, and that the verb-aspect amalgamation remains under the V node at S7structure2.Since Chinese is a morphologically poor language, the empty 4’ features (i.e. person,number) of the ASPECT head are invisible at PF. In accordance with the Economy andProcrastination principles (Chomsky 1993), the V+aspect amalgamation does not have to raiseto the ASPECT head for 4’ feature checking until LF.1.3.4. Bounding nodes in Chinese: NP and SAs in English, the bounding nodes for subjacency in Chinese are NP and S (Huang1982). As illustrated in example (8), if an NP moves across more than one bounding node, asubjacency violation will occur, making the sentence ungrammatical:(8)*zhebenshu1,wo renshi[,[,t1 xie t1] de rena].this book I know write DE person*‘Thjsbook, I know [the man [who wrote1].In this dissertation, I use the subjacency effect as the major test to argue for syntactic movementof the S-initial NP in the OSV word order.1.3.5. No S-structure wh-movement, but Weak Crossover (WCO) effect obtainsOne noted feature of Chinese is that it has no S-structure wh-movement. In particular,the word order of a wh-question does not differ from that of its declarative counterpart:2Evidence for such an assumption comes from word order facts. For instance, insertion of material between theverb and the aspect marker is forbidden as is shown in (ii), and a preverbal PP argument or adjunct must precedethe V + aspect amalgamation, as is shown in (iii):(i) wo von unzi da-guo Bill.I with stick hit-asp Bill‘I have hit Bill with a stick.’(ii)*woda yong gunzi guo Bill.I hit with stick asp Bill(iii)*woda-guo yon unzi Bill.I hit-asp with stick Bill8(9) a. ta mai-le yiben shu.he buy-asp one book‘He bought a book.’b. ta mai-le shenme?he buy-asp what‘What did he buy?’In his seminal work on Chinese syntax, Huang (1982) proposes that, despite its absence at Sstructure, wh-movement occurs at LF in wh-in-situ languages like Chinese. This analysis issupported by the WCO effect shown in (lob), on a par with the English example in (lOa):(10) a.*Whodid his, brother see ti?’b.*tade1didi kanjian-le shei1?his brother see-asp who*Whodid his1 brother see ti?’In contrast, Aoun and Li (1994) propose that there is no LF wh-movement. Their account forthe WCO effect is that a [+OJ morpheme in COMP binds both the wh-phrase and a co-indexedpronoun at LF. In this thesis, I adopt Aoun and Li (1994) as will be shown in Ch. BA constructionsIn Chinese, a definite NP can not only be preposed to a preverbal position as in (1 lb),but also appear in BA constructions as is shown in (llc):(11) a. wo kanwan-le zheben shu. (canonical word order)I finish-read-asp this book‘I have finished reading this book.’b. wo zheben shu kanwan-le. (object NP shift)I this book finish-read-asp‘I have finished reading this book.’9c. wo BA zheben shu kanwan-le. (BA construction)I BA this book finish-read-asp‘I have finished reading this book.’BA has been analyzed as a coverb, a preposition ora dummy Case-assigner (cf. Li andThompson 1981, Huang 1982, and Cheng 1991:14). Though both BA-NP and the shifted NPare in preverbal but post-subject positions, they are not derived by the same process, and arelocated in different positions. I will discuss these differences in detail in Ch.3.1.4. An Outline of the DissertationThis dissertation is organized as follows. In Ch.2, I first argue that when an S-initialobject NP in OSV constructions is construed with a gap in the canonical object position, thisempty category should be analyzed as the trace left by the movement of the NP. The major pieceof evidence for this claim is that if the empty category is embedded in an island, the sentencedisplays a subjacency effect. In the course of the discussion, I also examine eases Cheng (1991)and others use to argue for the non-movement analysis. Having established that movement isinvolved in OSV constructions, I then apply several diagnostic tests to this syntactic process. Ipropose that Chinese has two kinds of short-distance NP fronting. Specifically, as is the case inHindi (Mahajan 1990) and Japanese (Saito 1992), one type of fronting is A’-movement and theother is A-movement. In accordance with the Split Infi Hypothesis, I postulate a fully articulatedclause structure for Chinese. In particular, I propose that the fronted NP in A-movement landsin [Spec AgrOP] via a kind of overt raising for feature-checking, while the one in A’-movementfurther moves from that spec position and gets CP-adjoined. This chapter ends with a discussionof an alternative analysis based on empty operator movement, and on the suspension ofsubjacency effects observed in extraction out of the S-initial domain.In Ch.3, I discuss object NP shift, which yields SOV constructions. This process is10similar to NP fronting in that (a) the object NP is also ‘dislocated’, and(b) it must be defmiteor generic. It differs from the latter in that it lands in a position between the subject and theverb. Based on diagnostic tests including the absence of reconstruction effects and localityrestrictions, I conclude that object shift represents a type of A-movement, just like the short-distance shift found in Hindi or Japanese (Hindi 1990, Saito 1992). I further propose that intheobject shift cases, both the subject NP and the object NP overtly raise to [Spec AgrSP] and[Spec AgrOP] respectively, resulting in the SOy word order. This chapter concludes byaddressing some problems inherent in the A’-type analysis proposed in Z. Lu (1991) and F. Lu(1993).Ch.4 investigates the position of a wh-object with respect to dou and the various readingsassociated with it. I first discuss the existence of the interrogative/exhaustive list readingassociated with a wh-object to the right of dou at S-structure, an empirical fact which has notbeen discussed in the GB literature. I also provide data showing that a universal reading obtainsif the wh-object is shifted to the left of the adverb. Based on Diesing (1992), I propose that whphrases, like indefinites, can be either presuppositional or non-presuppositional. If they arewithin VP, they are subject to existential closure and get the existential/interrogative reading.If, however, they are outside VP, they define the range of the adverb of quantification dou andobtain the presuppositional/universal reading. The conclusion is that there is a strict correlationbetween the S-structure positions of a wh-NP as an indefinite in the context of dou and its LFinterpretations (cf. Diesing 1992 on German and other languages). The exhaustive list readingof the wh-phrase associated with the interrogative reading is derived from the fact that it is inthe scope of dou at S-structure. In the last section of this chapter, I examine different syntacticstructures where a wh-phrase and dou occur, and the resultant readings associated with the11former.In Ch. 5, I summarize the major fmdings of this research and discuss their relevancetothe study of Universal Grammar. I also bring up some questions as potential research topicsforthe future.12Chapter 2 Object NP FrontingChinese has constructions where a sentence-initial (S-initial hereafter) NP is related eitherto a full NP in the clause following it as in (1), or to a pronoun as in (2), or to an empty NPposition as in (3). There is usually a pause between this S-initial NP and the rest of the sentence,which is traditionally known as the ‘comment clause’ (cf. Li and Thompson 1981).(1) zheke shu1,yjhen da.this tree leaves very large‘This tree, its leaves are very large.’(2) zhege ren1,wo jian-guo t.this man, I meet-asp him‘This man, I have met him.’(3) zhege ren1,wo jian-guo .this man, I meet-asp‘This man, I have met.’In the literature on Chinese syntax, the S-initial NPs in cases like (1) and (2) have been analyzedas base-generated while the one in (3) has received different treatments (cf. Huang 1982, Xu andLangendoen 1985, Li 1990, Cheng 1991, Xue 1991, Z. Lu 1991). In this chapter, I focus oninvestigating sentences like (3). I argue that (a) the S-initial NP is not base-generated, but ismoved there; and (b) in contrast to the traditional analysis, this syntactic operation can be eitheran A-movement or an A’-movement. I use the theory-neutral term ‘object NP fronting’ to referto the movement exemplified in (3), since topicalization as used in the literature implies A’movement.132.1. The S-initial NP: Base-generation vs. MovementIn sentences like (3) repeated below,(3) zhege ren1,wo jian-guo .this man, I meet-asp‘This man, I have met.’the origin of the S-initial NP has been analyzed in various ways in the literature: Huang (1982)and Li (1990) argue that it is moved to the S-initial position while Xuand Langendoen (1985)and Cheng (1991) claim that it is base-generated, co-referentto a pro in the sentence by a kindof Left Dislocation. The purpose of this section is to examine the relevantdata and to maintainthe movement analysis. I will show that a subjacency effect is observed if the S-initial NP isrelated to a gap embedded in an island. I use this as a test to argue for the movement analysisfor the NP. In the process of the discussion, I will also show that subjacency is suspended insome structural configurations where it is expected to obtain. A structural generalization isproposed governing these cases.2.1.1. Relating to a Gap in a Relative ClauseAs discussed in Huang (1982), a relative clause is an island for movement in Chinese.In this chapter, I use subjacency as a test to support the movement analysis for object NPfronting: if a subjacency effect obtains in relating an element and a gap embedded inan island,it indicates that there must be syntactic movement involved.1’2Forease of presentation, I use‘One may argue that sensitivity to subjacency does not necessarily indicate that the NP is moved. What is movedmay be an empty operator predicated of the S-initial NP. See Section 2.3.1. for a discussion of this alternativeanalysis.14the subscript[fl,...]to mark clause boundaries, although I assume a fully articulated clausalstructure for Chinese (see Pollock 1989 and others) which will be discussed later. Relevant datais given below, organized into five sets.SET A. The gap is embedded in a relative clause in object position. As is shown below, relatingto a gap in such a configuration is ungrammatical, no matter whether the gap itself is in thesubject or the object position of the relative clause:(4) The gap is the subject of the relative clause:*nageren1,[,wo du-guo[,,, [,t, xie i] de shu]].that man I read-asp write DE books‘*Thatman1,I have read books which t1 wrote.’(5) The gap is the object of the relative clause:*nagexuesheng1,[,wo zhengzai zhao t jiao-guo t1] de laoshi]].that student I now look-for teach-asp DE teacher*‘Thatstudent1,I am looking for the teacher who has taught t1.’In contrast to the two cases above, if the gap is replaced by a resumptive pronoun, both (4) and(5) become grammatical, exactly like their English counterparts:(6) nage ren1,[,wo du-guo[ta1 xie tIde shu]].that man I read-asp he write DE books‘That man, I have read books which he wrote.’2Xue (1991) explores similar extraction facts in Chinese. As discussed below, the conclusion he reaches isdifferent from mine, as is his analysis, which is conducted within the theoretical framework of Generalized PhraseStructure Grammar.15(7) nage xuesheng, wo zhengzai zhaot jiao-guo ta,] de laoshi]].that student I now look-for teach-asp him DE teacher‘That student, I am looking for the teacher who has taught him.’If we assume that there is movement involved in (4) and(5), their ungrammaticality is accountedfor. Subjacency is violated, since the fronted NP has to cross three boundingnodes (which areIP and NP in Chinese) to get to the S-initial position.SET B: The gap is in a relative clause contained in a prepositional object NP.As is shown inexamples (8-9), such structures are also ungrammatical:(8)*thebenshu1,[,jingli yijing[gel[,t xie t1Ide ren3I]fa-lejiangjin].this book manager already to write DE man give-asp reward‘*Thisboolq, the manager has already given a reward [to [the man [who wroteii]]].(9)*thegeren1,[,jingli yijing[gel[, [,t xie t] de shu1]chu-le jiaqian].this man manager already to write DE book give-asp price‘*Thisman1,the manager has already offered a price for [the book [which wrote]].Again, the fact that subjacency obtains indicates that the S-initial NP is moved out of an islandto its S-structure position.SET C. The gap is in a relative clause in subject position. In contrast to the data in Sets A andB, relating the S-initial NP with a gap in such a position does not induce ungrammaticality:(10) The gap is the subject of the relative clause:nage ren1,[i,2 [ [t1 chuan t] de yifu] hen zang].that man wear DE clothes very dirtyLit.: ‘The man1,[the clothes [which t wears]] are very dirty.’16(11) The gap is the object of the relative clause:naben shu,[1 [,t kan-guo t1] de rena] bu duo].that book read-asp DE people not manyLit.: ‘That book, [people [who have readt]] are not many.’In these two examples, the S-initial NP can relate to thegap across three bounding nodes withoutshowing subjacency effects, a fact which may cast doubt on themovement analysis I have arguedfor so far. However, I propose that the absence ofa subjacency effect in these cases does notnecessarily mean that the S-initial NP is NOT moved there. Based onthe empirical facts in SetsA and B, I maintain the movement analysis even forcases in Set C and attribute the lack ofsubjacency to the particular structural position of the gap. Notice that the gap is in an islandcontained in the S-initial subject position. This suggests that this particular position maypossesssome special property that somehow voids the subjacency effect (see for anaccount).In fact, in addition to crosslinguistic evidence,3this hypothesis is supported by data in latersections where the S-initial NP relates to a gap in a sentential subject ora sentential complementto a subject NP.SET D. An S-initial PP or adverb is related to a gap in a relative clause. I am not focusing onadverbial movement in this dissertation, but for completeness, I will briefly look at casesinvolving an S-initial PP and adverb. Notice that if the gap is not in an island, an S-initialPPor adverb can be construed with either the matrix clause or (marginally) withan embeddedclause:3Hoji (1985:158-62) shows that in Japanese the contrastive topic NP-wa can be associated with a gap inarelative clause in subject position, but not in object position. Specifically, the gap as the subject/object ofa relativeclause in subject position is ok while the gap as the object of a relative clause inobject position is bad. Myprediction that the gap as the subject of a relative clause in object position is also badhas been confirmed by myJapanese consultants.17(12) zai lou-shang, wo t1 tingshuo [ni1 fang-le yiben shu1.at upstairs I hear you put-aspa book‘Upstairs, I hear that you put a book’(13)ZUotianI,?,wo zhidao t [ni I duwan-le zheben shu].yesterday I know you finish-read this book‘Yesterday, I know that you finished reading this book.’If the gap is embedded in a relativeclause, the construal of the S-initial PP or adverbiscomplicated. It can be easily construed with the matrixclause, but its construal with the relativeclause depends on the position of the NP containingthe island.(I) The gap is in a relative clause in object position:In the following set of data, an S-initialPP or adverb can only be construed with the matrix clause,not with the relative clause containedin an NP in object position (see also Li 1990:198).(14) zai najian wu-lij,.j, wo t1 zhengzai zhao[ L1John t xietkIde nabenshuk].at that room-in I now look-for John write DE that book‘In that I am looking for [the book [that John wrotet]] ti.’(15)zuotian,wo t1 kanjian-le yixietktj shou-le shang] de xuesheng].yesterday I see-asp some receive-asp wound DE students‘Yesterday,., I saw [some students [who were wounded t]]ti.’(II) The gap is in a relative clause in subject position: In contrastto the data in (I), an S-initialPP or adverb can be (marginally) construed witha relative clause in the subject NP. Admittedly,the construal is not as natural as the one with the matrix clause, butit is better than the construalwith the relative clause contained in an NP in object position, as shownin (I) above:18(16) zai zhuozi-shang.,, wot xie kJ de naben shuj t bei yahuai-le.at desk-on I write DE that bookpassive-by crushed-asp‘On the desk11,[the book [that I wrote t]Iwas crushed ti.’(17) zuotian4J?, wo 1 biaoyang-guo tJ de nagerenj t tuixiu-le.yesterday, I praise-asp DE that manretire-asp‘Yesterday1/,J, [the man [that I praised tI]retired t4.’In brief, the data in (I) is parallel to those wherean S-initial NP is related to a gap ina relativeclause contained in an object NP or in a prepositionalobject NP; the data in (II) is similar tothose where an S-initial NP relates to the gap ina relative clause in subject position. Theconclusion that emerges is that (a) the S-initialPP/adverb is moved there, and (b) lack ofsubjacency in (II) should be explained along the same lineas for the data in Set C.SET B: Summary and Apparent Counterexamples. To sumup, except when relating the S-initialNP to a gap in a relative clause in subject position, relatingto a gap in a relative clause displaysa subjacency effect. This constitutes the major piece of evidence for the claimthat the S-initialNP does involve movement.In recent studies on Chinese syntax, examples are provided to show thatrelating an 5-initial NP to a gap in a relative clause in object positiondoes not produce subjacency effects.Here are the two sentences culled from the relevant works:(18) zhexie hua1,wo dou mei kanjian t xihuane1] de rena].these painting I all not see like DE men‘These paintings1,I have not met a person who likes [themj.’(From Cheng 1991:147)19(19) zhege wenti1,wo conglai mei yudao-guo[‘t neng huida e1] de ren].this question I ever not meet-asp can answerDE man‘This questions, I have never meta person who can answer [itj.’(From Xu and Langendoen 1985: 15)According to these authors, lack of a subjacencyeffect indicates that the S-initial NP isbase-generated, and that the gap e, is a base-generatedpro. Apparently, the movement approachasdefended here has a problem: it predicts that examples(18-19) should be ungrammatical sincethe trace is embedded in an island contained in an objectNP.Notice that the fronted NPs in (18-19) are [-human] NPs whichnormally can not bereferred to by overt pronouns in Chinese:(20) zhexie xuesheng wo dou jiao-guo, wo hen xihuan tamen.these students I all teach-asp I very like them‘I have taught these students, and I like them very much.’(21) zhexie shu wo dou kan-guo, wo hen xihuan (*tamen).these books I all read-asp I very like them‘I have read all these books, and I like them very much.’Based on the prohibition against using pronouns for [-human]NPs, one may hypothesize that(18-19) are generated as Left Dislocation constructions with anovert resumptive pronoun in theposition of the gap. Later at PF, the language-specific prohibition deletesthe overt pronouns,leaving a gap behind. However, there are reasons to believe that thishypothesis is not a validone.First, if e1 in (18-19) is the trace of PF deletion ofan overt resumptive pronoun or is a4They acknowledge that the acceptability of (19) is questioned by some speakers.20real pro as claimed by Xu and Langendoen (1985)and Cheng (1991), it should not exhibitasubjacency effect if it is further embedded in another island:(22)*zhexiehua1,wo hal mei kanjian[, [,.,t sanbu[[John xihuan e1] zheyi yaoyan]these painting I yet not see spread John likethis rumourde rena11.DEman*‘Thesepaintings1,I have not met a person who spread the rumourthat John likes’In (22), e is separated from the S-initial NPby two islands. Specifically, it is embedded in thesentential complement to the NP ‘rumour’ which is in turn embeddedin a relative clause inobject position. If this e, were a pro or a trace of PF deletion of a resumptive pronoun,it wouldbe judged on a par with the following sentence which contains an overt pronoun:(23) zhexie yanyuan1,wo han mei kanjiant sanbu[John xihuan tamen1]zheyithese actors I yet not see spread John like themthisyaoyan] de rena]Erumour DE man‘These actors1,I have not met a person who spreads the rumour that John likes them1.’The reason why (23) is grammatical is that both the S-initial NP and theresumptive pronoun arebase-generated. Since they are not related by movement, they are insensitiveto the interveningislands. Returning to (22), the fact that it exhibits a subjacency effectsuggests that the S-initialNP must be moved there from the object position, and that the object empty categoryis indeedthe trace of movement, not a pro or the trace left by PFdeletion of a resumptive pronoun.Second, the head nouns in both (18-19) are non-specific.If they are changed to[+ specific] nouns by being modified by ‘that’, the sentences becomevery bad:21(24)*zhexiehua1,wo dou mei kanjian txihuan ej de nage renJ.these painting I all not seelike DE that person*Thesepaintings1,I have not met that personwho likes ti.’(25)*thegewenti1,wo conglai mci yudao-guot neng huida ej de nage rena].this question I ever not meet-aspcan answer DE that man‘*Thisquestion1,I have never met that man whocan answer tj.’Contrast these two examples with the followinggrammatical example where an overt resumptivepronoun is embedded in a relative clause ina specific NP:(26) John1,wo hal mci jian-guot xihuan ta1] de nage ren].John I yet not meet-asp like him DEthat person‘John1,I have not found that person who likes him1.’The significance of the contrast is that if the empty categoryin (24-25) is a pro or the trace ofPF deletion of an overt pronoun, they should receive the samegrammaticality judgement as (26).This prediction is not fulfilled.I propose that the contrast between (24-25) and (26) is accountedfor if the emptycategory in the former two examples is analyzed as a trace left by themovement of the S-initialNP. Since the trace is in an island contained ina specific NP, movement out of this NPconstitutes a violation of subjacency and of the Specificity Conditionwhich prevents extractionout of a [+ specific] NP (Huang 1982:4 19 quoting May 1977 and others;see also Nishigauchi1990). In contrast, in example (26) with the resumptivepronoun, there is no movementinvolved, and the Specificity Condition has no effect onthe sentence.Also notice that if (18-19) are changed fromnegative sentences into affirmative ones,they become ungrammatical, presumably because the nonspecific readingof the containing NPbecomes unavailable, or difficult to get, in the absence of negation:22(27)*thexjetwa1,wo dou kanjian[[j’.,t xihuan eJ de ren].these painting I all see like DE men*Thpaintings1,I have met the people who likeIi.’(28)*thegewenti1,wo yudao-guo t neng huidaej de renj].this question I meet-asp can answer DE man*Thjsquestion1,I have met the people who can answer t1.’The proposal that specificity plays a role in (18-19) is supported by thefollowingobservation: if the object NP is modified by renhe‘any’ under negation and is thusunambiguously nonspecific, extraction out of it is more acceptable. Lookat the followingsentences:(29) zhexiehua1,wo mei kanjian renhe t xihuan ej de rena].these painting I not see any like DE men‘These painting, I have not met any person who likes [them1].’ (cf.(18))(30) zhege wenti1,wo conglai mei yudao-guo renheL[,t neng huida eJ de rena].this question I ever not meet-asp any can answer DE man‘This question1,I have never met any person who can answer [it1].’ (ef.(19))What the last two sentences suggest is the mirror image of the Specificity Condition.5Specifically, the generalization is that the more non-specific an NP is, the easier extraction outof it becomes. Assuming this generalization is correct, the next question is howto implementit syntactically, an issue which I leave it open for future research.65jowe my discussion of this part to Henry Davis and Hamida Demirdache.6James Huang (p.c.) suggests that in cases like (18-19), there is an ‘invisible but overt’ inanimate pronoun. Butthe question is if there is such an invisible pronoun, why does the sentencedisplay a subjacency effect if the gapis further embedded in another island, in contrast to an overt pronoun? He also correctly points out thatthese twoexamples show that non-referentiality may be a necessary condition for fronting, but it still remains that Subjacencywould block any such extraction. One possible solution to this problem is to hypothesize that L-marking targetingthe head NP penetrates to its relative clause CP if the NP is non-specific (Henry Davis p.c.), thus voidingthe CPof its barrierhood. This technical speculation entails that L-marking or subjacency is inherentlytied to the SpecificityCondition, an implication which needs to be further studied in the long run.Also see Ch.4. for the proposal of (wh-)feature percolation out of non-specific NPs.23In brief, the exceptional cases (18-19), cited from Xuand Langendoen (1985) and Cheng(1991), do involve movement. This is becausethe embedded empty object displays islandeffectsif it is further embedded in another island,as is shown in (22). Moreover, these two examplesshow Specificity Condition effects as illustrated in (24-25).Their grammaticality should perhapsbe attributed to some property associated withthe non-specific interpretation of the containingobject NP.To sum up section 2.1.1, there is a subjacency effect observed inconstruing the S-initialNP with the gap in a relative clause. This indicatesthat the NP is moved there from the relativeclause, not base-generated S-initially. Thedata also exhibit two types of exceptions to thesubjacency effect. One concerns construal between an S-initialNP and a gap contained in arelative clause in subject position (i.e. in[ [a,..e..]]. . .1);an account for it will be offeredin section 2.3. The other cases are discussed by Cheng (1991)and Xu and Langendoen (1985);I suggest that an account for these cases lies in the non-specificnature of the containing NPunder negation.2.1.2. Relating to a Gap in a Sentential SubjectIn this section, I investigate cases where an NP is relatedto a gap contained in asentential subject. Surprisingly, as in relativization out ofa relative clause in subject position,there is no subjacency effect, even thoughtwo bounding nodes are crossed. I conclude,therefore, that the Sentential Subject Condition (SSC)does not hold in Chinese. Later in thissection, I will discuss cases Huang (1982) cites for theopposite claim. Relevant data illustratingthe point are provided below:24(31) The gap is in the subject positionof a sentential subject:zhexie ren1,wo bingbu renwei[p2[p[p1t gan zhezhong shi]] bu heshi/hen qiguai].these people I not thinkdo this thing not proper/very strange‘These people, I do not think that (for them)to do this kind of thing is not proper/verystrange.’(32) The gap is in the object positionof a sentential subject:zhexie shi1, wo bingbu juedeL[t’ta shuo tJ]bu heshi/hen qiguai1.these matter I not feel he saynot proper/very strange‘These things, I do not think that for himto talk about (them) is improper/verystrange.’In these sentences, the S-initial NPcan be related either to the subject orobject NP of thesentential subject, even though three boundingnodes are crossed. Notice again that the domainof the gap is the S-initial subject, i.e.[, [, [..gap..]..]. This recalls the case involvingthe gapin a relative clause in subject position,discussed in the section 2.1.1. Despitethe lack ofsubjacency, I still maintain that there is movementinvolved, and that what suspends thesubjacency effect has to do with a particular structuralconfiguration.7Using lack of subjacency effects as evidence,Xu and Langendoen (1985:16) claimthata gap embedded in a sentential subject is in realitya pro. However, ungrammaticality will resultif the gap is further embedded in an islandlike a relative clause, a surprise under the proapproach:(33)*zhexieren1,wo jian-guo t shuoe1 gan zhezhong shi]] bu heshi] de rena1].these people I meet-aspsay do this thing not proper DE man*‘Thesepeople1,I met a man who said that [fort1] to do this kind of thing is notproper.’Koster (1978:57) proposes that ‘sentential subjects [in English]do not exist. Instead, there are satellitesentences binding the (phonologically zero) NP subject in themain sentence. In this conception, the main sentenceis taken as an open sentence, satisfied by the satellite’.If this ‘satellite sentence’ is A’-adjoined, extraction out of itcan be accounted for by using the proposaladvanced in Lasnik and Saito (1990). (See section2.3.2. for details). I thank Leslie Saxon fordirecting me toKoster’s work.25Compare (33) with the following sentencewhere the gap is replacedby an overt resumptivepronoun:(34) zhexie ren1,wo jian-guot shuo tamen1gan zhezhong shi]]bu heshi]these people I meet-asp saythey do this thing not properde rena]].DEman‘These people1,I met a man who said that forthern to do this kind of thing isnotproper.’To conclude, the fact that thegap displays a subjacency effect if it is further embeddedin anisland rules out the possibility that it isa pro. It must be a trace left by movement.As part of his account of subjacency, Huang (1982)claims that Chinese displays theSSC.Note that in one of his later works (1984:562), hesuggests that Chinese appears to lack the SSCeffect as argued here. Below, I will review thedata used in Huang (1982) to illustrate theSSC.I am interested in finding out how theexamples formerly attributed to the SSC canbe explainedindependently of this constraint.8Notice that Huang’sjudgement differs for each case in thefollowing set of data.(35) mai-le[ [, [[,Lisi mei kan ti]] zhen giuai] de nabenshu.J.I bought-asp Lisi not read really strangeDE that book‘I bought the book which [[the fact that Lisi didnot read tIis very strange].(From Huang 1982:459 his (18)):8Huang (1984) does not explain why (35) below (from Huang1982) is ungrammatical if it is not attributed tothe SSC. See section for more discussion of Huang (1984).26(36) ??(wo kanjian-le)[ [Lisi da-let1]] shi wo hen bugaoxing]de nage renj.I saw-asp Lisi hit-asp made me very unhappyDE that man‘I saw the man1 who [[the fact that Lisi hit [him1]]made me unhappy1.(From Huang 1982:496 his (99b))(37) ??naben shu1,[Lisimei kant1]] zhen qiguai1.that book Lisi not read really strange‘That book, [the fact that Lisi did not read[itJ] is really strange.’(From Huang 1982:459 (his (19))Notice that (35) and (36) are cases of NP relativizationout of a sentential subject, but theydiffer in the predicate used in the relative clause.(37) uses the same predicate(then qigual‘really strange’) for the sentential subjectas (35), but it is a case of NP fronting out of asentential subject. Also notice that of the three sentences,only (35) is really bad. Though Huang(1982:496) claims that the other two are ‘ill-formed’, hedoes NOT give them a*,only ??. Myjudgement is that (36-37) are much better than (35),and they should be treated as marginal withat most one?.Let us analyze the contrast by looking at the predicates involvedin the three sentences.First, if we change the embedded predicatethen qiguai ‘really strange’ in (35) to that as usedin (36), (35) becomes acceptable.(35’) ?wo mai-le[, [,,,[Lisimei kan t1]] shi wo hen bugaoxinIde naben shu,I.I buy-asp Lisi not read made me very unhappyDE that book‘I bought the bookj which [[the fact that Lisi did not read[it1]] made me veryunhappy].‘Below is another example like (35),but with a different embedded predicate, and henceadifferent judgement. This again shows that thetype of predicate involved is relevant in27determining grammaticality judgement.(35”) ?wo mai-le[ [ [,[,Lisi mei kant1]] hen bu heshi] de naben shu1].I buy-asp Lisi not read very not proper DE that book‘I bought the book1 which [[that Lisi did not read [it]] is not proper].‘Conversely, if we replace the embedded predicate in (36)by the predicate used in (35), (36)becomes ill-formed:(36’)*(wokanjian-le)[[S[I1P[SLisi da-let1]] zhengiguai] de nage ren1].I see-asp Lisi hit-asp really strange DE that man‘I met the man, who [[the fact that Lisi hit tj was really strange].’Now the crucial question is why then qiguai ‘really strange’ is different from such predicatesas shi wo bugaoxing ‘make me unhappy’ and buheshi ‘not proper’ in causing the difference ingrammaticality judgement. Huang (p.c.) recently suggests that (35) is bad because an elementis extracted out of a factive sentential subject predicated by then qiguai ‘really strange’. But thequestion is why such a prohibition is not observed in (37).Right now, I have no syntactic analysis to offer as to why the use of the predicatethenqiguai ‘really strange’ makes a difference in grammaticality judgement. Nonetheless, I will makea descriptive generalization about its distribution, following a suggestion made by Xue. Xue(1991:63-64 fn 36) notes that then qiguai ‘really strange’ has an exclamatory effect. Anexamination of exclamatory phrases reveals that such phrases cannot be used as the predicate ofa relative clause. If then qiguai is analyzed as an exclamatory phrase, we may find anexplanation for the ungrammaticality of (35).28In Chinese, exclamatory phrases are AdjPs or VPs usually followed by anexclamationparticle a. This particle has various allomorphs depending onthe preceding syllable.(38) a. zhepian wenzhang1, Lisi pang wo xiugai tj] jiao wo hao gaoxinga.this paper Lisi help me revise make me how happy particleLit.: ‘This paper1,[q,that Lisi helps me to revise ti]] makes me how happy.’Intended.: ‘This paper1,how happyftmakes me that [Lisi helps me to reviseb. zhejian chenyi1, ta chuant1]] hao heshi ya.this shirt he wears how fit particleLit.: ‘This shirt1,L[1that hewears till is how fit/suitable.’Intended: ‘This shirt1,how fitftis [for him to wear [itjl.’If an NP is relativized out of a sentential subject with an exclamatory phrase as its predicate,the sentence becomes bad:9(39) a. dawan-leL[11,[1Lisibang wo xiugai tj] jiao wo hao gaoxing a] de napianI fmish-type-asp Lisi help me revise make me how happy particle DE thatwenzhangl.paperIntended:K’Ifinished typing the paper that how happyftmakes me [that Lisihelped me revise ritli.’b.*womai-le chuan t1]] hao heshi ya] de najian chenshan].I buy-asp Lisi wear how fit particle DE that shirtIntended:*’Ibought the shirt that how fitftis [that Lisi wears [itil.’Now let us return to example (35) repeated below,91n English, an exclamatory phrase is introduced by ‘how’ for an AdjP or by ‘what’ for a NP. As in Chinese,it cannot be used in a relative clause:(i) a. This is the movie star [whom I am very happy to meet].b.*Thisis the movie star [how happy I am to meet].(ii) a. This is the movie star [whom it is my pleasure to introduce to you].b.c?Thisis the movie star [what a pleasure it is to introduce to you].29(35)*mai-leL[[Lisi mei kant1]] zhen giuai1 de naben shuj.I bought-asp Lisi not read really strangeDE that book*‘Ibought the book which [[the fact that Lisi did notread t1] is very strange].If we treat zhen qigual as an exclamatory phrasefollowed by a covert exclamation particle, wecan then rule it out on a par with those in (39): anexclamatory phrase cannot occur as thepredicate in a relative clause.To sum up, Huang (1982) argues for the SSC in Chinese based onsentences like (35-37).However, I have shown that there is a grammaticality contrastbetween (35) and (36-37): only(35) is really bad. Since (35) and (36) involve the samekind of movement and (36) isgrammatical, the ungrammaticality of (35) should notbe attributed to a subjacency violation.Instead, I propose that it should be independently ruledout, since it involves embedding anexclamatory phrase in a relative clause. In other words, movementout of a sentential subjectdoes not display the SSC, and this constraint does NOT really hold inChinese, just as it doesnot in Japanese (Kuno 1973 :240). In the following section, I will providemore data to show thatextraction from an NP in subject position does not cause ungrammaticality.2.1.3. Relating to a Gap in a Sentential Complement to an NPIn this section, I show that relating an S-initial NPto a gap contained in the sententialcomplement to an NP displays a subjacency effect, except whenthe containing NP is in subjectposition. Once again, this indicates that (a) syntactic movement is involved forthe S-initial NP,and (b) lack of subjacency in the exceptional cases isdue to the special properties of theextraction domain in subject position.30SET A. The gap is in the sentential complement to anNP in object position. The gap in (40a)is in the subject position of the sentential complementwhile the one in (40b) is in the objectposition.(40)a.*nageren1,wo xiangxin[,t1 jian-guo John] zhege yanchuan].that man I believe meet-asp Johnthis rumour‘*Thatman1,I believe the rumour that t1 has met John.’b.*nagerenj, wo bu xiangxin Lisi kanjian tJzheju hua.that man I not believe Lisi see this saying*‘Thatman1,I don’t believe the statement that Lisisaw t1.’(From Huang 1982:459)In both examples, there is a subjacency effect in relating the S-initial NPto the gap in the island.This suggests that the relation between the two is one ofsyntactic movement. But as repeatedlyshown in this section, this subjacency effect is suspended if the extractiondomain is located insubject position. This is clearly shown in SET B.SET B. The gap is in the sentential complement to an NP in subject position.In the followingpair of sentences, although the fronted NP crosses more than onebounding node, there is nosubjacency effect and the sentences are acceptable. Notice again that the gapis located in theS-initial subject:[, [, [,. .gap..]]...].(41) a. ?nage ren1,wo bu zhidao t1 jian-guo John]zhege xiaoxi] shi dajia namethat man I not know meet-asp John this news make everybodysochijing].surprised‘That man, I do not know that the news that [he] has met Johnmakes everybodyso surprised.’b. ?nage ren1,wo bu zhidao1 [ [,,John jian-guo tj zhege xiaoxi] shi dajiathat man I not know John meet-asp this newsmake everybodyname chijing].so surprised31‘That man, I don’t know that the news that John has met [him] makes everybodyso surprised.’In (41a), the subject of the sentential complement is fronted; in (41b), it is the object that isfronted. Both sentences are grammatical.’°In brief, the recurrent pattern emerging from the data so far is that the S-initial NP mustbe moved there at S-structure, since there is a subjacency violation if the gap is in an island. Butif the extraction domain is in the subject position, subjacency is somehow suspended. In the nextsection, I will show subject position is not the only extraction domain that can void islandeffects.2.1.4. Relating to a Gap in the ‘Fronted’ Object:I have claimed in section 2.1.1. above that it is grammatical to extract out of a relativeclause contained in a subject NP, but not out of a relative clause contained in an object NP.‘°Xue (1991:75) provides example (i) to claim that extraction of the object of the sentential complement to thesubject NP is not grammatical, in contrast to the acceptable extraction of the subject as in (ii):(1)*hepjp.g1,Lisi xiangxin [[xuesheng xunqiu t1] de fang’anj hui yingde guangfande zhichi.peace Lisi believe students seek DE plan will win wide support‘Peace, Lisi believes the plan [that students seek [it]]will win wide support.’(ii) ?xuesheng, Lisi renwei [[ti xunqiu heping] de fang’an] hui yingde guangfande zhichi.students Lisi think seek peace DE plan will win wide support‘Students, Lisi thinks that the plan that [they] seek peace will win wide support.’But heping ‘peace’, an abstract noun, cannot normally be fronted as shown below:(iii)*?hepjpg1,xuesheng xunqiu t1.peace students seek‘*peace,students seek.’Thus it is not adequate to rely on (i) only to claim that extraction of the object of the sentential complementto asubject NP is ungrammatical, particularly in view of the grammaticality of examples like (41b).32Actually, the second half of the generalization is not entirely true: if the object NP containingthe relative clause is fronted, extraction out of it becomes acceptable.11 Consider the followingpair of sentences:(42) a. Object NP containing the relative clause is in-situ:*zhaflgg,Lisi bu zhidao ni yijing mai-le[, [,e, xie 1 de] shu1].Zhangsan Lisi not know you already buy-asp write DE bookLit.:*‘ZhaI1g.Lisi does not know that you have already bought [the books that twrote].’b. Object NP containing the relative clause is fronted to the beginning of the embeddedclause:Zhangsan, Lisi bu zhidao[i,, [,e1 xie 1 de] shu] ni yijing mai-le tj]].Zhangsan Lisi not know write DE book you already buy-aspLit.: ‘Zhangsan1,Lisi does not know that [the books that[hei]wrote], you have alreadybought ti.’Moreover, if the embedded object is fronted further to the front of the matrix clause, extractionout of this embedded object also becomes acceptable:“This was first noted in Huang (1984:562). The following example is his:(i) Zhangsan,, e1 changge de shengyin wo hen xihuan.Zhangsan sing DE voice I very like‘Zhangsan,, the voice with which [he,] sings, I like.Contrast (i) with (ii) below from which (i) is derived:(ii)*Thangsanj,wo hen xihuan [e, changge de shengyin].Zhangsan I very like sing DE voice‘*Zhangsaii,I like [the voice with which [e.J sings].Huang’s solution for (i) is to adopt the Generalized Control Rule (GCR) for the data. Specifically, e, is controlledby Zhangsan since it is the ‘closest’ nominal element for it. See section for more discussion of the GCR.33(43) ?Zhangsan1,[[,e, xie 1 deIshuJ[,Lisi bu zhidao ni yijing mai-le t,]]J.Zhangsan write DE book Lisi not know you alreadybuy-aspLit.: ‘Zhangsan1,[the books that [hej wrote], Lisi does not know you havebought tjalready.’In short, examples which involve a gap in a relative clause but which are grammaticalshare the following property: the gap is embedded in the S-initial subjectNP or in the S-initialfronted object as in (42b) and (43). One may stop here and list these two NPdomains asexceptions to the subjacency effect. But there are reasons to believe that thegeneralization cango even deeper. Notice that once the object is fronted before the subject NP, extractionout ofthe relative clause in the subject position is no longer acceptable. First, consider example (44)which illustrates fronting out of the relative clause in subject position with the matrix object insitu:(44) zhege laoshi1, t1 jiao-guo t] de xuesheng] dou jian-guo zhezhonghuak.this teacher teach-asp students all see-asp this-kind plant‘This teacher, the students [[he] taughtIhave all seen this kind of plant.’However, when the matrix object NP is fronted and the NP ‘this teacher’ is extractedout of thesubject NP across this fronted object as in (45), grammaticality deteriorates in contrast to (44):(45)?*zhegelaoshi1,zhezhonghuakt1 jiao-guo t] de xuesheng] dou jian-guotk.this teacher this-kind plant teach-asp DE students all see-asp*Thisteacher1,this kind of plant, the students [t1 taught] have all seentk.’Furthermore, with the matrix object fronted, fronting the NP ‘this teacher’ out of thesubject NPto a position after the fronted matrix object induces ungrammaticality:34(46)?*zhezhonghuak, zhege laoshi1, t jiao-guo t] de xuesheng] dou jian-guotk.this-kind plant this teacher teach-asp DE students allsee-asp*Thjskind of plant, this teacher1,the students [t1 taughtt] have all seentk.Given the data in (44-46), the correct generalizationis that the extraction domain that voidssubjacency must be in an S-initial position at S-structure, whether it is thesubject or a frontedobject. This applies to the S-initial NP of either the matrix clause(cf. 43) or an embedded clause(cf. 42).2.1.5. Summary:The central claim of this section is that movement is involved when the S-initialNP isrelated to a gap in the following ‘comment clause’. This is supportedby the subjacency effectwhich obtains if the gap is embedded in an island.13 However, there is a. complicationto theeffect that subjacency does not obtain in the following environments:(47) A. Movement out of the relative clause contained in an NP insubject position (section2.1.1).B. Movement out of a sentential subject (section 2.1.2).C. Movement out of the sentential complement to the subject NP (section 2.1.3).‘2To ensure that ‘this teacher’ is really extracted out of the relative clause, (46) should beread with a pause afterthis NP.‘3My conclusion is different from Xue’s (1991), who, as mentioned early, also discusses extractionfacts inChinese. His conclusion is that the Complex NP constraint in general is NOT relevant to Chinese(Xue 199 1:76).In particular, he claims that the impossibility of extraction from a complex NP in object positionshould not beattributed to the effects of the Complex NP constraint. His solution for the antecedent-gaprelations is couched inthe theoretical framework of Generalized Phrase Structure Grammar. The gist of hisanalysis for the extractionasymmetry is represented by his Linear Precedence rule: NP[SLASH) <[+ V.-Nj. This means that an NP whichproperly contains an empty category necessarily precedes the V. Since an object follows the verbin Chinese, onlysubject NPs can contain traces. In that work, Xue does not discuss extraction out of a frontedobject NP, and thestructural property shared by a complex NP in subject position and the one in the frontedobject NP.I thank Leslie Saxon for directing me to Xue (1991).35D. Movement out of a fronted object (section 2.1.4).Section 2.3.2. attempts to account for why subjacency is suspended in movementout of the Sinitial NP domain.2.2. Types of Movement and their Landing SitesHaving established that the derivation of S-initial NPs involves movement, the nextlegitimate questions concern what type of movement it is and what is its landing site.In thissection, I argue that, in contrast to the traditional view that NP fronting ortopicalization isuniformly A’-movement in Chinese, A-movement takes place in some eases of short-distancefronting, as observed for Hindi by Mahajan (1990) and for Japanese by Saito (1992). Thisconclusion is based on the results of applying to the Chinese data some diagnostic testscommonly used in the literature. Assuming the Split Infi Hypothesis for Chinese clause structure,I also discuss the landing sites for the fronted object in both A and A’-type movements.2.2.1. Two Types of MovementBased on Hindi, Mahajan (1990:46) argues that argument shift and adjunctionto XP existas separate operations. Argument shift yields a structure that enters binding theory, overridesWeak Crossover (WCO) effects and is not reconstructible. It has the properties of A-movement(his ‘L-movement’). Adjunction to XP, on the other hands, displays A’-movement properties,such as reconstruction effect. In this section, I will apply the same tests to Chinese, arguing thatthe distinction between argument shift and XP adjunction also holds in Chinese short-distanceNP fronting where the object is fronted to the S-initial position of its own clause. Iwill discusslong-distance fronting in section A’-movementThe claim that some cases of NP fronting in Chinese are A’-movement is supported bythe following tests.Test A: Reconstruction Effect In Chinese, as in Hindi, the anaphor contained in thefronted NP can still be bound by the matrix subject. This indicates that reconstruction appliesto the moved NP and that the NP fronting in question is an A’-movement. In the following pairsof sentences, (48a) shows the canonical word order while (49b) has a fronted NP containing ananaphor.(48) a. Bill1 zui xihuan[,ziji1 de haizi].Bill most like self DE kidsLit: ‘Bill1 most likes self1’s kids.’Intended: ‘Bill likes his own kids best.’b.[,ziji, de haizi],Bill1zui xihuan t.self DE kids Bill most likeLit: ‘SeIf1’s kids, Bill1 likes most.’Intended: ‘Bill likes his own kids best.’In (48a), the anaphor contained in the object NP is bound by the subject. In (48b), the fact thatanaphor binding by the subject still obtains indicates that t left by the fronted object is avariable, and that the movement is an A’-movement: reconstruction applies to make bindingpossible.1414As in Japanese (cf. Hoji 1985:126), variable binding under reconstruction in Chinese varies from speaker tospeaker. In (i), the pronoun ta is bound by the universal quantifier:(i) meigeren1dou yijing qisu-le [cia ta1 de ren].everybody all already sue-asp hit him DE man‘Everybody, has sued the mans that hit him.’In (ii), after the object is fronted, the pronoun, according to some speakers, can still be marginally bound by thematrix subject quantifier:(ii) ?[da ta1 de ren], meigeren1dou yijing qisu-guole t.hit him DE man everybody all already sue-asp.37Test B. Condition (C) Type Reconstruction Effect As discussed in Saito (1992:90),there is a Condition (C) type reconstruction effect which is observed only with A’-movement.An R-expression contained in a moved constituent cannot serve as the antecedent of a (matrixsubject) pronoun unless it is more deeply embedded in that constituent. Here I use this test toargue that the landing site in some cases of NP fronting in Chinese is an A’-position. Look atthe following pair of sentences. The NP ‘John’ contained in the direct object cannot act as theantecedent for the subject pronoun, whether the direct object is in-situ or fronted.(49) a. Canonical word order:ta yijing jigei-le Mary [John1deshu]k.he already send-asp Mary John DE book‘He.1., already sent Mary John1’s book.’b. Direct object fronted:[John1deshu]k,yijing jigei-le Marytk.John DE book he already send-asp Mary‘John1’s book, he. already sent to Mary.’However, as is shown in (50b) below, the NP ‘John’ can become the antecedent if it is moredeeply embedded in the fronted direct object.(50) a. Canonical word order:*tyijing jigei wo [[wo tuo John1zai zhongguo mal] deshulk.he already sent me I ask John at China bought DE book*He1has sent me the book that I asked John1 to buy in China.’b. Direct object fronted:[[wo tuo John1 zai zhongguo maiIdeshulk, ta1yijing jigei-le wotk.I ask John at China bought DE book he already sent-asp me‘The nian that hit him, everybody has sued.’If the judgement for (ii) is reliable, it should be taken to indicate that reconstruction occurs, and that the frontingof the object NP is A’-movement.38‘The book that I asked John, to buy in China, he1 has sent to me.’‘John’ in (50b), but not in (49b), is deeply embedded in the relative clause contained in thefronted object NP, and it can function as the antecedent for the matrix subject pronoun. The factthat Condition (C) type reconstruction effect obtains indicates that the fronting in (50b) must bean A’-movement.One question concerning this test is why it should be a test for A’-movement (DavidIngram p.c.). This question can be answered by looking at a raising structure, a prototypical Amovement:(51) [Johns de didi1]zai ta kanlai haoxiang[ t hen congming].John DE brother to him look seem very bright‘Johns’ s brother1seems to him [t1 to be very bright].‘In (51), ‘John’ wealdy c-commands ta ‘him’ as in (49b) above; yet it can be co-indexed with thelatter. As (51) is an A-movement, the grammaticality contrast between (51) and (49b) under theco-indexed reading can only be explained if we assume that (49b) involves an A’-movement. Ifit is also an A-movement, it should not be different from (51) in grammaticality judgement.’5James Huang (p.c.) points out that if object-fronting can be freely A’- or A-movement, then all object-frontedsentences should be fully grammatical with respect to crossover effects, etc. In particular, examples like (49b)should be ok under the co-indexed reading as a kind of A-movement. Huang suggests an alternative account for(49b) based on Reinhart (1981). According to Reinhart (1981:614), the domain of a node A consists of all and onlythe nodes c-commanded by A. She also proposes (1981:617) that ‘a given NP cannot be interpreted as coreferentialwith a distinct nonpronoun in its c-command domain.’ This is how she explains the contrast between (i) and (ii):(i)*NearDan1,he saw a snake.(ii) Near him1,Dan1 saw a snake.In (i), ‘Dan’ is in the domain of the subject ‘he’ and is not a pronoun; therefore, coreference between the two isruled out. In (ii), however, ‘Dan’ is not in the domain of ‘him’ (which only extends to the PP) and coreferencebetween the two NPs is allowed. To return to the Chinese cases, (49b) may be treated as a legitimate A-movementas Huang suggests, but it is ruled out under the intended coreferential reading exactly for the same reason as for(i): the possessor NP ‘John’ is in the c-command domain of the subject ‘he’ and it is not a pronoun.Note that in (50b), ‘John’ is still in the c-command domain of the subject, but its coreference with the39Summing up, based on the tests that NP fronting displaysa Condition (A) reconstructioneffect and a Condition (C) type reconstruction effect, I conclude thatthis process is an A’-movement, at least in some cases. A-movementIn this section, I show that NP fronting can also display A-movementproperties basedon three tests.Test A: Remedy WCO A pronoun can be boundby a wh- or quantifier phrase in aconstruction involving A movement. Take the following raising construction for instance,(52) a. [shei1 zai tade1 mama kanlai haoxiang [t zui xingfu]]?who at his mother view seem most happy‘Who1 seems to his1 mother [t to be most happy]?b. [meige haizi1 zai tade1 mama kanlai haoxiang [t1 dou shi tiancai]].every child at his mother view seem all are genius‘Every child, seems to his1 mother [t, to be a genius].’In contrast, a quantifier undergoing LF A’ movement cannot bind a pronoun it crosses. Thissubject pronoun is acceptable. Reinbart (198 1:632) discusses the same kind of contrast found in English as acounterexample to her proposed anaphora restriction:(iii)*Ben’s box, he1 put his cigars. (her (81a))(iv) In the box that Ben1 brought from China, he1 put cigars. (her (82c))Reinhart’s solution is that the anaphora rule obeys subjacency: in (iv), the antecedent is not subjacent tothepronoun, being separated by two bounding (NP) nodes, so the anaphora rule cannot apply to mark the two NPsasnoncoreferential. Similarly, the contrast in Chinese between (49b) and (50b) can be explained along the samelineas an alternative account. Also note that in (51), the c-command domain of the prepositional object also only extendsto PP. thus allowing its coreference with ‘John’.There are two noteworthy points regarding Reinhart’s solution for (1) and (ii). One is that it is no differentfrom a Condition C type solution: (i) is ruled out since the name is bound. The other point is that the Split InfiHypothesis is assumed in this dissertation. To adopt Reinhart’s notion of c-command domain, we have to furtherassume that a clause basically has two c-command domains: VP-internal and VP-external. For the latter one,allfunctional projections constitute one c-command domain. Under the A-movement analysisfor (49b), the subjectpronoun ta ‘he’ would be in [Spec AspP], c-commanding the possessor name ‘John’ in the fronted object NP landingin the higher [Spec AgrOP]. See section for the proposed landing sites forA-type fronting.40WCO effect is illustrated by LF Quantifier Raising (QR), an A’-movement,in the followingEnglish sentence:(53)*Hjs1mother loves everybody1.Likewise, in Chinese, a quantified object NP in-situ cannot refer to a precedingpronoun as aresult of WCO induced by LF QR. However, if this NP is fronted at S-structure,it can binda pronoun without inducing the WCO effect, a fact which supports the claim thatthe landing siteof the moved NP must be an A-position on a par with the examples in (52), notwith the one in(53). This is clearly illustrated in comparing (54a) with (54b):(54) a. Quantificational object is in-situ:*woxiang tamende1jiazhang biaoyang-guo suoyoude xuesheng.I to their parents praise-asp all student*‘Infront of their parents, I praised all the students.’b. quantificational object is fronted:suoyoude xuesheng1,wo dou xiang tamende1jiazhang biaoyang-guo t.all student I all to their parents praise-asp‘All the students,, I have praised [them] in front of their, parents.’As the S-structure fronting of the quantifier object NP overrides the WCO effect,the landingsite in (54b) must be an A-position.’616According to Lasnik and Stowell (199 1:697), wh-phrases and other QPs do not occur as topics.This impliesthat topicalization and QR are separate operations. Since the topicalizable NPs (in English) are not realoperators(while QPs and wh-phrases are), WCO does not obtain when these NPs cross a co-indexedpronoun to be fronted:(i) This book1,I will never ask its author to read.In Chinese, however, at least QPs can be fronted/topicalized, as is shown in (54b). Note that QPsare realoperators, and that there is nothing to prevent this overt fronting from being analyzed as S-structureQR. Yet,different from LF QR, there is no WCO effect in (54b). Therefore, the only way we cancapture these facts is toclaim that S-structure QR is different from LF QR in that it is an A-movement, withoutinducing WCO. Also notethat as in English, non-operator NPs in Chinese can also be fronted:41Test B: No Reconstruction When both the direct object and indirectobject are in situ,the preceding indirect object cannot be a pronoun co-indexed withan element of the NP in thedirect object position. But when the direct object is fronted leftward, coindexingbecomesacceptable.(55) a. Canonical word order:*iyijing gei-le ta1 [John1de shu].I already give-asp him John DE book‘I already gave him John1’s book.’b. Direct object fronted:[John1de shu], wo yijing gei-le ta1 t.John DE book I already gave-asp him‘John1’s book, I already gave him1.’Since reconstruction applies to the site of a variable, and there is no reconstruction effect,t in(55b) must be an NP trace, not a variable. This indicates that the movement involved isan Amovement, not an A’-movement.’7(ii) zhexie xuesheng1,wo yijing xiang tamende1jiazhang biaoyang guole.these students I already to their parents praise-asp‘These students, I have already praised (them) to their parents.’Lack of WCO in (ii) is subject to two possible analyses: (a) it is an A-movement as in(54b); (b) it is an A’-movement as in English, but since the fronted NP is not an operator, it does not induce WCO effect.I thank James Huang for reminding me of Lasnik and Stowell (1991). Also see Choi (1991) for the viewthat the remedy of WCO does not necessarily differentiate between A- and A’-movements.‘71n section, I use the following sentences to show that the S-initial NP is moved to anA’-position,based on the Condition (C) type reconstruction effect:(49) b.*[J0hn1deshu]k, ta1yijing jigei-le Marytk.John DE book he already send-asp Mary‘John’s book, he3 already sent to Mary.’(55) b. [wo tuo John1 zai zhongguo mai deshujk, ta1yijing jigei-le wo t.I ask John at China bought DE book he already sent-asp me‘The book that I asked John to buy in China, he has sent to me.’One may argue that (55b) and (49b) seem to have the opposite judgement, andmay thus doubt the analysisproposed. However, there is one difference between them which deserves our attention.The coindexed pronoun42Test C: The fronted NP as an A-binder: Another piece of evidence for theA-statusof some fronted NPs is that they can function as an A-binder for a reflexivein the clause asshown in (56b) below. (56a) is a sentence with the canonical word order whereanaphor bindingdoes not obtain:(56) a.*zjjjjde laoshi biaoyang-guo [zuozai qianpai de xuesheng]1.self DE teacher praise-asp seated front DE studentsLit.:*‘Self’steacher has already praised students1sitting at the front.’b. ?[zuozai qianpai de xuesheng]1,ziji1 de laoshi yijing biaoyang-guole t1.seated front DE students self DE teacher already praise-aspLit.: ‘Students, sitting at the front, self’s teacher has already praised[t].’The fact that anaphor binding obtains in (56b) demonstrates that thefronted object NP must havelanded in an A-position.To sum up, we have applied three tests in this section. They are (a) remedy of WCO;(b) no reconstruction; (c) the fronted NP as the A-binder. The conclusion that emerges fromthese tests is that NP fronting in Chinese can be A-movement.Since Chinese has both A-type and A’-type NP fronting,a question arises as to when aparticular case of NP fronting belongs to one type, not the other one. Compare the followingpair of sentences:(57) a. taziji1,John1 zui xihuan t.himself John most like‘Himself, John, likes t best.’in (55b) is in object position while the one in (49b) is in subject position. Since all the Condition(C) typereconstruction effect examples given in (Saito 1992:90-91) use subject pronoun, we may concludethat the differencein judgement between (55b) and (49b) is due precisely to the difference in the position of thepronoun, and thereis nothing contradictory about the judgement. Under Reinhart’s analysis(1981), the possibility of coreferencebetween ‘John’ and ta ‘him’ in (55b) (in contrast to that in (49b)) isattributed to the fact that ‘John’ in (55b) islocated outside the c-command domain of the indirect pronoun, whichonly extends to VP.43b. John1,[zijide, haizi] dou bu xihuan t.John self’s kids even not like‘John1,even his own kids do not like [him]1.’Examples like (57a) are analyzed as A’-movement since they displayreconstruction effect; incontrast, the one in (5Th) is analyzed as A-movementsince the fronted object can act as an A-binder for the anaphor contained in the subject. Nowthe question is why (57a) is not analyzedas A-movement while (5Th) is not analyzed as A’-movement? I followDeprez (1989:183) inassuming that both A and A’-possibilities exist in each frontingconstruction. But if someviolation would result, the syntax dictates that movementto the position which leads to the lesserviolation is the only one possible. Deprez’s solution isapplicable to the analysis of (57a-b). In(57a), under the A-movement analysis, the sentence should be bad since the anaphoris unbound;under the A’-movement analysis, it should be good sincethe fronted anaphor can be boundunder reconstruction. Given the outcomes of the twoanalyses, the A’-type analysis is preferredto the A-type analysis. In (5Th), the two options also exist.If it is A-movement, the anaphor canbe bound and the sentence is good; if it is A’-movement, it is predictedto be bad since theanaphor would be left unbound. In this case, the A-type movementanalysis wins over the A’-type one.2.2.2. The Landing Sites of NP FrontingIn the previous section, I have established that there are two kindsof movement involvedin cases of NP fronting. The next question is what arethe landing sites for the A and A’-movements. Below I first make explicit my assumptionsabout the clausal structure in Chinesebased on some recent theoretical proposals. I then advancemy proposal for the landing sites forthe two kinds of movement. Specifically, I claim thatA-movement in Chinese lands in the44specifier position of AgrOP while A’-movement is CPadj unction. The landing site of A-movement: [SpecAgrOP]Mahajan (1990) proposes an articulated clausal structurefor Hindi based on the Split InfiHypothesis (cf. Chomsky (1989) and Pollock (1989)).Under this hypothesis, a clause is not asimple projection of inflection syntactically representedas IP. Instead, inflection is split intovarious functional projections like AgrSP, AgrOPand TP etc. Such an expansion of IP makesadditional Spec positions available, internal to theclause. The following tree is proposed byMahajan (1990:14) for Hindi:(58)AG1PSPECAGESSPECISPECATJXSPECAGRO45In addition, Mahajan recognizes an L/non-Ldistinction among positions (1990:10). He proposesthat L-related positions are specifiers andcomplement positions of a lexical item and functionalheads projected from it. Within the clausalsystem, this includes Spec and Complement positionsof V, Agr and T. In contrast, non L-relatedpositions refer to all the other positions, includingSpec CP and adjunction positions.Adopting the same hypothesis, Chomsky and Lasnik (1991:33,82) further propose thattypological variations among languages should reduceto the ordering parameters and propertiesof functional elements. They try to derive variousword orders by postulating overt or covertNP raising. Take a right-headed language forinstance. If object raising is overt to [Spec AgrOP]at S-structure and subject raising to [Spec AgrSP] covert atLF, it results in what they call“scrambling”, i.e. the OSV order at S-structure. If bothsubject and object raise overtly, we willget the SOV order. They also state that other options are possible.The case of Chinese is more complicated than the casesdiscussed in Chomsky andLasnik. Chinese first of all exhibits disharmony in its headpositions across categories: it isstrictly head-final for NP, head-initial for PP and head-medial forVP (Huang 1982, Li 1990,Goodall 1991, Travis 1989, Mulder and Sybesma 1992).Second, although its basic S-structureorder is SVO, it can also have OSV (‘NP fronting’) orSOV (‘object shift’) orders. Given thesefacts and the theoretical assumptions outlined above,I propose that Chinese has the followingD-structure representation:46(59) CPI’AgrSP/ \AgrOP/ \MPI\AspPI\VP/ \subj. V’I\V obj.This structure postulated for Chinese differs from the one proposed for English(as in Chomskyand Lasnik 1991) in the following respects: (a) it contains anAspP instead of TP, as in Cheng(1991); (b) AgrOP is higher than AspP. These possibilitiesare countenanced, though notexplicitly discussed, in Chomsky and Lasnik (1991:23), as legitimate parametricvariations.18I further postulate an MP projected by a modal verb(cf. Cheng 1991:131), given data like (60c):(60) a. ta kan-guo zheben shu.he read-asp this book‘He has read this book.’b. ta yiding kan zheben shu.he must read this book‘He must read this book.’c. ta yiding kan-guo zheben shu.he must read-asp this book‘He must have read this book.’‘8My proposal coincides with the one proposed by Deprez (1989:223) for Mainland Scandinavianand Icelandic:AgrOP is also postulated higher than TP in these languages in order to accommodate wordorder facts.47(60c) shows that the modal verb and the aspect marker can co-occur in onesentence. When theV+asp amalgamation raises for feature checkingat LF, it should move to a head distinct fromthe one headed by the modal verb, since it is impossiblefor a head to accommodate both themodal and the V+asp at LF. Hence, the postulation ofboth MP, headed by the modal, andAspP, headed by the raised V+asp at LF.Regarding the SVO and OSV word orders found in Chinese, I proposethat they arederived as follows:SVO: the canonical word order: The subject NP, base-generated in[Spec VP], moves overtlyto [Spec AspP] or [Spec MP] as an intermediate step for its final raisingto [Spec AgrSP] forfeature checking. (Alternatively, the subject can overtly land in [SpecAgrSP] for featurechecking, though the S-structure effect would be vacuous in the SVOorder.) The object NPremains as sister to the V and later undergoes covert movementat LF to [Spec AgrOP] forfeature checking.’919This contrasts with Aoun and Li’s (1993:23) claim that subject NP in Chinese remains in [Spec VP]at Sstructure. One major problem with their proposal is that in the presence ofa modal, the subject is obligatorily toits left:(i) wo gan qu.I dare go‘I dare to go.’(ii)*ganwo qu.dare I go‘I dare to go.’If the ModalP is generated higher than VP, as commonly assumed, their no-subject-raising analysispredicts theopposite judgement for examples (i-u).48(61) The S-structure representation of the canonical SVO order in Chinese:AgrSPI \subj.1 AgrOP/ \MPI \t”, AspPI’t’ VP/ \t V’V obj.OSV: the word order for A-type fronting: The object NP undergoes overt raising to [SpecAgrOP]. The subject NP moves to and remains in [Spec AspP] or [Spec MP] at S-structure,butit later moves to [Spec AgrSP] at LF for feature-checking. Notice that what is novel about itsraising is that it can stop either in [Spec AspP] (or [Spec MP]) as in OSV order, or it canovertly raise to [Spec AgrSP] as in SVO order. There is nothing in principle to stop this kindof semi-overt raising in [Spec AspP], given the peculiar facts of Chinese. This kind of partialraising is also discussed in Pollock (1989) and Sportiche(1988).20In VSO languages likeArabic, if a sentence has an auxiliary, the subject can occur between the auxiliary or the verb,or before the auxiliary. Schematically, the order will be as follows: Aux-V-S, Aux-S-V, S-AuxV. This shows that the subject, though not the object, can move through and stop in theintermediate spec positions of the functional projections.201thank Hamida Demirdache (p.c.) for pointing out to me the VSO facts.49The S-structure representation for the A-type fronting derivingOSV order is as follows:(62)AgrSPI \AgrOPI \Obj., MPI \subj., AspP/ \t,i VP/ \ti V,/ \VGiven Mahajan’s assumption that the spec position ofa functional projection is an A-position,the A-movement properties of certain fronted NPsfollow naturally from our proposal here.The question is then how plausible it is to treat [Spec AgrOP] asthe landing site for theA-type NP fronting, given that Chinese has no morphological agreement.Recall that one of theconstraints for the fronted object NP is that it mustbe definite or generic. Notice that thedefiniteness requirement for the frontedobject NP is also found in morphologically richlanguages such as Turkish, Hindi and some European languages (Mahajan1991, Runner 1994).Furthermore, these fronted NP are analyzedas landing in [Spec AgrOP] in these languages.This indicates that there may be a correlation between the definitenessof the fronted NP and theavailability of [Spec AgrP], as attested by the morphologyof these languages. Returning toChinese, given the crosslinguistic correlation and thedefiniteness constraint for the fronted NPin Chinese, I propose that postulating [Spec AgrOP] as the landingsite for the fronted object inthis language is not as arbitrary as it at first seemsto be. In Ch.3, I will discuss NP object shift(i.e. SOV) which is also subject to the same constraint. I delay a detaileddiscussion on thejustification of Agr in Chinese until section 3.3.3 of that chapter.502.2.2.2. The landing site of A’-movement: CP adjunctionRochemont (1989) argues that topicalization in English is usually IPadjunction. But toaccount for two sets of his data, he also proposes adjunction to the matrix CP. Following him,I propose that A’-type NP fronting in Chineseis also CP-adjunction. This CP-adjunctionproposal may run counter to one of the assumptions inGB syntax. According to Chomsky(1986b:6), adjunction is possible only to a maximal projectionthat is not an argument.Rochemont (1989) allows adjunction to the matrix CP precisely becauseit is not an argument,unlike embedded CPs. Notice that in Chinese, embedded A’-movementalso occurs:(63) wo xiang[naben guanyu ziji de shu]1,John yijing kanwan-let..I think that about self DE book John already read-aspLit.: ‘I think that [the book about self], John, has already finished readingti.’As the anaphor in the fronted NP can be bound by the embeddedsubject under reconstruction,the landing site of the containing NP must be an A’-position. If weassume that A’-fronting isCP adjunction and this embedded CP is an argument, this violatesthe ban on adj unction to anargument. One possibility to accommodate this embedded CP-adjunctionis to propose thatChomsky’s constraint against argument CP adjunction is parameterized.21The anaphor binding facts illustrated in (63) notonly force an A’-movement analysis, butalso oblige us to assume that the ‘variable’ left by the CP-adjunction ofthe object must be lowerthan the subject. In particular, it must be in [Spec AgrOP] withthe subject in the higher [SpecAgrSP]. Recall that the object NP has to move to [SpecAgrOP] for feature checking. Therefore,21Rochemont (1989:155) also provides data from Italian, showingthat topicalization must be CP-adjoined incertain cases. This conclusion is based on examples like(i) below, where the embedded PP is topicalized to the leftof the complementizer:(i) Mi domando a Giovanni, se gli hanno presentatoPiero.me I-ask to John that they have presented Peter‘I ask if they have presented Peter to John.’51it cannot front from its postverbal D-structure position directlyto the CP-adjoined position, forthis would deprive it of the opportunity to be in [Spec AgrOP] for featurechecking. In order tocapture the A’-movement properties such as reconstruction and to accordwith feature checkingtheory, we have to assume not only that the S-structureposition of the subject must be in [SpecAgrSP], but also that the object first moves to[Spec AgrOP] for feature checking before itadjoins to CP, leaving a ‘variable’ in that spec position. Here is the representation forthe A’-type fronting deriving the OSV order:(64) CPIobj.3 CPI \AgrSP/ \subj. AgrOPI \t’ MP/ \t”, AspPI \t’ VP/ \t V,/ \ViIn (64), while t, t’1, t”, and t are all traces of A-type movementlanding in a Spec position, t’in [Spec AgrOP] is the variable left by the A’-type frontingof the object which is adjoined toCP. Note that since t’3 is the intermediate landing site forthe object, it predicts that SOV is apossible S-structure word order with the subject and the object in[Spec AgrSP] and [SpecAgrOP] respectively. This prediction is indeed borneout by the data in Chinese which will beexplored in detail in Ch.3. The correctness of theprediction, in turn, lends support to theproposed analysis for the various positions of relevant tracesin A’-type fronting. Lastly, the fact52the fronted object in A’-movement has the same definiteness requirement as the fronted objectin A-movement follows from our account: at S-structure, the fronted object NP in either casemoves through or lands in [Spec AgrOP] which is associated with a definite reading (cf. section2.2.2.1. above and section 3.3.3. below).To sum up, in Chinese, A-type object NP fronting involves raising to [Spec AgrOP] asa kind of overt raising for feature checking while A’-type NP fronting involves CP-adjunctionwith an intermediate trace in [Spec AgrOP]. Long-distance FrontingAll the examples provided above are short-distance fronting, i.e. the object is fronted tothe S-initial position of its own clause. What happens if the object from an embedded clause isfronted to the S-initial position of a matrix clause? Mahajan (1990) shows that in Hindi, whileshort-distance scrambling is either A or A’-movement, long-distance scrambling can only be A’-movement. I will show that the same conclusion applies to Chinese, based on the results of thefollowing tests.Test A: Reconstruction Effect In Chinese as in Hindi, an anaphor contained in anembedded object which is fronted before the matrix clause can still be bound by the embeddedsubject. This indicates that reconstruction applies to the moved NP, and that the NP frontingin question is an A’-movement. In the following pairs of sentences, (65a) shows the canonicalword order, while (65b) has a fronted NP containing an anaphor.(65) a. ni zhidao Bills zui xihuan[taziji de haizi].you know Bill most like himself DE kidsLit: ‘You know that Bill1 most likes sell’s kids.’Intended: ‘You know that Bill likes his own kids best.’53b. [,,taziji1de haizi], ni zhidao [t” Bill1 t’ zui xihuan ti].himself DE kids you know Bill most likeLit: ‘Se1f’s kids, you know that Bill1 likes best.In (65a), the anaphor contained in the embedded object NP isbound by the embedded subject.In (65b), anaphor binding by the embedded subject still obtains for the frontedembeddedobject. This shows that (a) t’ left by the fronted object is a variable, and (b) the position oft’‘is an A’-position, and so is the final landing site of the frontedembedded object NP. This NPis “reconstructed” from its S-initial position back to t” andeventually to t’ to make bindingpossible.Test B: Condition (C) Type Reconstruction Effect First of all, letus review the datain (66) below which illustrate Condition (C) Type Reconstruction Effect in cases of shortdistance fronting. Specifically, the name in the embedded object frontedto the initial positionof its own clause cannot refer to the embedded subject (as in(66-b)) unless it is more deeplyembedded in the object as shown in (66-d).(66) Short-distance fronting:a. Canonical word order:wo zhidao [ta yijing jigei-le Mary [John1de shu]J.I know he already send-asp Mary John DE book‘I know that he already sent Mary John1’s book.’b. object NP fronting:wo zhidao [[John1deshulk, ta yijing jigei-le Mary tk].I know John DE book he already send-asp Mary‘I know that John1’s book, he already sent to Mary.’c. Canonical word order with the name more deeply embedded:*wozhidao [ta1 yijing jigei wo [[wo tuoJohn1zai zhongguo mai] de shu]J].I know he already sent me I ask John at Chinabuy DE book*1know that h; has sent me the book that I askedJohn, to buy in China.’54d. Direct object fronted:wo zhidao[[[wo tuo John1zai zhongguo mai] deshu]k, ta1I know I ask John at China bought DEbook heyijing jigei-le wo tJ.already sent-asp me‘I know that the book that I asked John to buy inChina, he has sent to me.’In parallel, the same effects obtain inlong-distance fronting as illustrated in (67): the name inthe embedded object fronted to the initial positionof the matrix clause cannot referto thematrix subject as shown in (67-b) unlessit is more deeply embedded in the fronted NP as shownin (67-d):(67) Long-distance fronting:a. Canonical word order:ta. yiwei [ta. yijing jigei-le Mary [John1deShu]k].he think he already send-asp Mary JohnDE book‘He. thinks that he. has already sent Mary John1’s book.’b. Long-distance fronting:[John1deshu]k,ta. yiwei [ta. yijing jigei-le Marytk].John DE book he think he already send-aspMary‘John1’s book, he thinks that he already sentto Mary.’c. Canonical word order with the name more deeply embedded:Ta3 yiwei [ta. yijing jigei-le Mary [wotao John1 zai zhongguo mai de shu]J.he think he already send-asp Mary I ask John inChina buy DE book‘He. thinks [that he. already sent to Mary[the book that I asked John1 to buy inChina]J.’d. Long-distance fronting:[wo tao John1zai zhongguo mai deshu]k,ta yiwei [ta, yijing jigei-le Marytk].I ask John in China buy DE book he thinkhe already send-asp Mary‘[The book that I askedJohn1 to buyinChina]k, he11thinks that he11 already sent toMary.’55Examples (66b-d) and (67b-d) show that Condition(C) Type Reconstruction Effect obtains inboth short and long distance fronting. If the formertype of fronting is analyzed as A’-movementbased on the test, the long-distance fronting casesshould be analyzed the same way.Test C: WCO Effects In Chinese, a quantifier objectNP in-situ cannot refer to apreceding pronoun as a result of WCO inducedby LF movement. While short-distance frontingof such an NP suppresses WCO as discussed above,long-distance fronting does induceWCO:(68) a. Canonical word order:gaosu-guo tamende1jiazhang [ni xihuan suoyoude xuesheng].I tell-asp their parents you likeall student‘I told their1 parents that you like all the students1.’b. Wh-object fronted:?*suoyoudexuesheng1,wo gaosu-guo tamende1jiazhang[ni xihuan t1].all student you tell-asp theirparents you like*‘Allthe students1,I told their parents thatyou like ti.’In both cases, the embedded quantifier cannot referto the pronoun in the preceding matrixclause. This is because the quantifier phrase in (68a) hasto cross the co-indexed pronoun whenundergoing Quantifier Raising (QR) at LF for the matrixscope reading; likewise, the quantifierphrase in (68b) is fronted to the matrix clause,though at S-structure. Notice that (68a) is ruledout because LF QR is an A’-movement resulting in WCOeffects. If we assume that the Sstructure long-distance fronting of the embedded quantifierin (68b) is also an A’-movement, itsungrammaticality can be accounted for in parallelto that in (68a): the embedded quantifier landsin an A’-position, A’-binding both the pronounand its variable.Test D: The fronted NP cannot be an A-binder Anotherpiece of evidence for the A’-status of the long-distance fronted NPs is that they cannotfunction as an A-binder for a reflexivein the subject in the matrix clauseas shown in (69b) below. (69a) is asentence with the56canonical word order where anaphor binding does not obtain either.(69) a.*zjji1de laoshi zhidao[q,wo biaoyang-guo zhexie xuesheng].self DE teacher know I praise-asp these studentsLit: ‘Sel1’s teachers know that I already praised these students1.’b.*?[zhexiexuesheng]1,ziji1 de laoshi zhidao [wo yijing biaoyang-guot1].these students self DE teacher know I already praise-aspLit.: ‘These students1,self’s teachers know that I have already praised ti.’The fact that anaphor binding does not obtain in (69b) indicates the fronted object NP must landin an A’-position.To sum up, the long-distance fronted object shows reconstruction effect, induces WCO,and has other A’-movement properties. I therefore conclude that long-distance fronting inChinese can only be A’-type movement. Recall that A-type fronting is analysed as overt raisingof the object to [Spec AgrOP] for feature checking. Our finding that long-distance frontingcannot land in an A-position follows naturally from this feature checking theory. For the Specposition of the matrix AgrOP is reserved for the raising of the matrix object, and it is thereforeunavailable for accommodating an embedded object as this would result in a feature clash. Thisline of argument, based on feature clash, also accounts for why the embedded object cannot landin a position between the matrix subject and the matrix verb (i.e.*S0V-[S,,--V-tj])to bediscussed in Ch. 3.2.3. Residual IssuesIn this section, I address two residual issues: (a) the fact that a subjacency effect obtainsin relating the S-initial NP to a gap in an island is also open to an alternative analysis, that ofempty operator movement; and (b) the subjacency effect is suspended in extraction out of the57S-initial subject or fronted object.2.3.1. Empty Operator Movement: An Alternative AnalysisIn Section 2.1, I use sensitivity to subjacency as a diagnostic test to argue that there issyntactic movement for the S-initial NP. In this section, I will briefly considerone alternativeanalysis: the fact that subjacency obtains may indicate that what is moved is not theS-initial NP,but an empty operator predicated of the S-initial NP. Under this analysis, the example in (70a)should have the structural representation indicated in (70b), where an empty operator movesacross an island, resulting in a subjacency violation:(70) a.*John,wo renshi [[ti jiao-guo e] de laoshi].John I know teach-asp DE teacher*‘John,I know the teacher who has taught t.’b.*John1,[Op,[,wo renshi jiao-guo t] de laoshi]]].John I know teach-asp DE teacher*‘John,I know the teacher who has taught t.’As discussed above, there are two types of NP fronting in Chinese: A-type fronting andA’-type fronting. But in the literature (cf. Chomksy 1986 among others), empty operatormovement has only been defined as A’-movement. Now an interesting question arisesas to howto apply this alternative analysis to the A-type movement cases.22 Look at the followingsentence, which has been treated as A-type fronting of the S-initial NP:(71) John1, [ziji1 de haizi] dou bu zunjing t.John self DE kid even not respectLit.: ‘John1,even self’s kids do not respect.’am indebted to Hamida Demirdache for her insight on this section.58Under the empty operator movement analysis, example (71)will be represented as follows:(72) John1,[Op1[[ziji de haizi] dou bu zunjing t]].John self DE kid even not respectLit.: ‘John1,even self’s kids do not respect.’On a par with the structure proposed for the A-type NP fronting, one mightpropose that in (71),the empty operator first moves in [Spec AgrOP], A-bindingthe anaphor in the subject NPlocated in the lower [Spec AspP]. Then it is further CP-adjoined.In recent studies on NP fronting in other languages, the analysiswhich varioussyntacticians have adopted is the movement of NP itself, not ofan empty operator (see Deprez(1989) for Scandinavian languages, Webethuth (1989) for German, Mahajan(1990) for Hindiand Saito (1992) for Japanese). The data in Chinese are compatible with eithertype of analysis,and I have no compelling evidence to rule out one for theother. However, following theseprevious works and by Occam’s Razor, I opt for the conventional NP frontinganalysis as thedefault analysis for Chinese; however, I still entertain the possibility of an empty operatoranalysis. Further research on both Chinese and other languagesmay determine the final choicebetween the two analyses.2.3.2. Subjacency SuspensionThe central claim of Section 2.1. is that there is a subjacencyeffect if the S-initial NPis moved out of an island. However, this effect does not show up if the extractiondomain is theS-initial subject or fronted object. The data are summarized below:59(73) A. Movement out of the relative clause in asubject NP (section,wo bu zhidao[,2 [[t1 xie t] de shu] name changxiao].Zhangsan I not know write DE book so popular‘Zhangsan, I do not know that the books which [he] wrote are so popular.’23B. Movement out of a sentential subject (section renj, wo bingbu renweiLP2LPLPIt1 gan zhezhong shill hen heshi/qiguai].these people I not think do this thing very proper/strange‘These people, I do not think [(for them) to do this kind of thing] is proper/strange.’C. Movement out of the sentential complement to a subject NP (section ren1,wo bu zhidao jian-guo John] zhege xiaoxi] shi dajianame chijing.that man I not know meet-asp John this news make everybody so surprised‘That man, I do not know that the news that [he] has met John makes everybodysosurprised.’D. Movement out of a fronted object (section Zhangsan, wo bu zhidao t1 xie t deIshuIk [,,ni yijing mai-letk]].Zhangsan I not know write DE book you already buy-aspLit.: ‘Zhangsan1,I do not know that [the books that [he]wrotelk, you have alreadyboughttk.In this section, I try to answer the question as to why subjacency is suspended in movement outof the S-initial subject or the S-initial object. I will first review two analyses from Huang (1982,1984) and then propose my own account, based on Lasnik and Saito (1992).use the negative form of the matrix verb in order to exclude the potential parenthetical reading of the matrixclause. Using an affirmative sentence does not affect the grammaticality of the sentence:(i) Zbaflgsafl, wo zhidao[,2 [e xie] de shu] hen changxiao].Zhangsan I know write DE book very popular‘Zhangsan, I know the books that [he] wrote are very popular.’602.3.2.1. Review of Previous AnalysesThe first analysis to review is the ‘alternative construal’ analysis, proposed for sentenceslike the following (Huang 1982:459):(74) zheben shu1,[,[[,Lisi mei kan t1]] zhen qiguai].this book Lisi not read really strange‘*Thisbook1,that Lisi didn’t read t is really strange.’In (74), object extraction is allowed out of a sentential subject. According to Huang, the reasonthat (74) is not ungrammatical is that the S-initial NP ‘that book’ may be alternatively construedas occurring within the sentential subject, thus violating no known constraints of grammar:24(75)[,[zheben shu1, Lisi mci kan t1] zhen qiguai].this book Lisi not read real strange‘[The fact] that thisbook1,Lisi didn’t read t, is really strange.’Likewise, we may extend this ‘alternative construal’ analysis to subject extraction out of a, relative clause in a subject NP where no subjacency effect shows up. Thus, the example in (76)may be reanalysed as having the structure in (76’):(76) naben shu1,[, [,.,t kan-guo t.]de ren3Ibu duo].that book read-asp DE people not many‘The book1,the people who have read [it]1 are not many.’(76’)[,, [, [,naben shu1,t kan-guo tIde renJ bu duo].that book read-asp DE people not manySame as (76)In the above cases, the “alternative analysis” is possible since the S-initial NP and the matrixsubject containing the relative clause are immediately adjacent to each other. However, this4Though Huang (1982:459) indicates?? for sentence (75), Ijudge it as grammatical. Myjudgement is consistentwith the ‘alternative construal’ analysis he proposes.61“alternative construal” is not available for cases like the following:(77) Zhangsan1,wo bu zhidaoL2 [,[11)1t1 xie tjJ de shu] name changxiao].Zhangsan I not know write DE bookso popular‘Zhangsan, I do not know that the books [he] wrote are so popular.’In (77), the S-initial NP is related to an empty category ina relative clause contained in thesubject NP in an object complement clause. Notice that alternative construal is impossible forlack of immediate adjacency between the S-initial NP and the embedded subject NP containingthe co-indexed gap. Similarly, alternative construal should notbe available in (78) below, sincethe matrix clause ‘he thinks’ intervenes between the S-initial NP and the embedded sententialsubject containing the co-indexed gap; yet the sentence is still grammatical:(78) zheben shu1,ta juede[., [[8Lisi mei kan ti]] zhen qiguai].this book he think Lisi not read real strange‘*Thatbook, he thinks that Lisi didn’t read t is really strange.’Given data like (77) and (78), I do not think that “alternative construal” is the right analysis forthe lack of subjacency effect in fronting out of a relative clause in subject position, or out of asentential subject.25Huang (1982:457) also gives the following sentence and marks it ungrammatical:(i)*Nageren1,[[t1 kanjian l]de xuesheng] lai-le.that man see de student come-asp‘The man1,the students whom [he.] saw t have come.’If ‘alternative construal’ applies in cases like (75), the question arises as to why it does not apply in (i). Huang doesnot discuss this difference in grammaticality judgement.One solution to the contrast between (i) and (75) is that it has to do with the matrix predicate used. Noticethat the matrix predicate used in (i) denotes an action and is a stage-level predicate, while the one in (75) denotesa state and is an individual-level predicate (cf. Diesing 1992). This observation is supported by the following pairof sentences:(ii) Nage ren1,[[t, jiao-guo t,] de xuesheng3]hen congming.that man teach-asp de student very smart‘That man1,the students whom [he] has taught are very bright.’62In a footnote in one of his later papers, Huang (1984:562 fn 30) proposesan operatormovement account for sentences like (77). This solution is based on the functionaldeterminationof empty categories and the Generalized Control Rule (GCR). Accordingto the GCR, an emptypronominal should be coindexed with the closest nominal element (Huang 1984:552).Takeexample (77) for instance. Under Huang’s analysis (1984), ‘Zhangsan’ originates from theoperator position of the embedded complement clause under ‘I do not know’,and the subjectempty category ‘e’ is base-generated as a pronominal. Because ‘Zhangsan’ in theembeddedoperator position is the closest nominal element for the empty category in subjectposition, theGeneralized Control Rule applies to co-index these two elements. Then ‘Zhangsan’ undergoesmovement into the matrix topic (or Comp) position, and the pronominal becomes a variableafter co-indexing.There are two problems with this proposal. One is theoretical, regarding the claim thatthe empty category starts out as a pronominal and becomes a variable after it is co-indexed withan NP. Current GB theory no longer endorses such “chameleon” empty categories (Brody 1985among others). Second, such a proposal will force us to lose an account for the grammaticalityof (79):(with an individual-level predicate. cf. (i))(iii)*zhang,[[ej xie] de shuImai-le.Zhangsan write DE book sell-asp‘Thangsan1,the books that [he.] wrote were sold.’(with a stage-level predicate denoting action.)I propose that in general, subjacency is suspended in movement out of a relative clause in subject position inChinese. The fact that (i) is ungrammatical should be attributed to other factors. For instance, one speculation isthat the subject of a stage-level predicate must be within the VP domain at LF or the post-LF level (cf.Diesing1992). But the subject in (i) cannot move back to the VP at LF as it is semantically incomplete with ‘that man’moved out of it at S-structure.63(79) ?zhexie xuesheng1,wo bu zhidao tj jiao-guo e1]de laoshiIyijing tuixiu-le.these students I not know teach-asp DE teacher already retire-asp‘These students1,I do not know that the teacher who has taught[themi has alreadyretired.’Suppose, as in Huang (1984), that ‘these students’ is base-generated in the operator position ofthe embedded clause under ‘I do not know’. In contrast to (77), the empty category is in theobject position, and therefore the NP ‘these students’ in this embedded operator position is notthe closest nominal for e1. Instead, the closest nominal ist, and, according to the GCR, ‘thesestudents’ should not be co-indexed with e1. Yet, co-referencebetween these two elements isacceptable, and ‘these students’ can be construed as the object of the verb ‘teach’ in the relativeclause. Given these two objections to Huang’s account, I seek an alternative analysis, to bepresented in the next section. A’-position voids barrierhoodLasnik and Saito (1992:102) propose that “A’-binders are not barriers for furtherextraction”. Compare the following two English sentences. Extraction out of the embeddedsubject in-situ is bad since the embedded subject is a barrier. In contrast, as is shown in the(80b), extraction is grammatical if the subject NP itself is first topicalized. This, according toLasnik and Saito, is due to the fact that the topicalized subject is in an A’-position and ceasesto be a barrier. They further note that (80b) is marginal not because of subjacency, but becauseof the ‘internal constituent effect’ (Kuno 1973), which obtains in movement out of a nonrightmost branch constituent:(80) a.*Who1do you think that Mary believes [pictures oft1]2 are on sale?b. ??Who1do you think that [pictures oft1]2, Mary believes t2 are on sale?64Following Lasnik and Saito (1992), I propose that both the subjects and fronted objects can beCP-adjoined as A’-binders to void their barrierhood. Notice that the S-structure word ordereffect for such an adjunction is vacuous for some cases. Take adjunction of the S-initial subjectfor instance. We have proposed that S in the SVO order is in [Spec AgrSP], an A-position. Inorder to void its barrierhood, the S can be CP-adjoined (with vacuous S-structure effect) to allowextraction out of it. One may frown upon the idea that subject is vacuously moved. This isdisallowed in English (Fiengo et al 1988:89):(81)*John,came yesterday.In Chinese, subjects have certain properties that may justify vacuous subject movement. One isthat Chinese has been treated as a topic-prominent language, and that its subject is oftenidentified as the topic (Li and Thompson 1981). One piece of evidence for this claim is that topictends to be definite crosslinguistically. In Chinese, subject must be a definite NP.26 The secondproperty of the Chinese subject is that if it is a long one, it can be separated from the rest of thesentence by a comma in writing. In English, a topic is typically set off from the rest of thesentence by a comma. Given these properties, Chinese subjects should be identified as topics andvacuous subject movement should be allowed.It is proposed above that A’-type NP fronting is CP-adjunction and hence the fronted NPin this position should not be a barrier for extraction, as proposed in Lasnik and Saito. In thecase of an A-type fronted object in [Spec AgrOP], I propose that, just like the subject in [SpecAgrSP], the object NP is further adjoined to CP, as Mahajan (1991) does for Hindi. Specifically,the fronted object NP is first moved to the A-position, thus showing all the properties of AIndefinite subjects in Chinese are obligatorily preceded by you ‘exist’.65movement. It then moves vacuously to the CP-adjoined position to void its barrierhood. Hereis the structural representation for extraction out of a fronted object NP:(82) CP/ \NP, CPI \NP0 AgrSPA /\ti.. AgrOPI\t0 MPNP0indicates that the object is CP-adjoined from [Spec, AgrOP], leaving a trace in that position.NP, is the element extracted out of the extraction domain NP0 in the A’-position.27To sum up, CP-adjunction provides a unified account for the data presented in thissection. It correctly rules in all the grammatical sentences where the topic NP is extracted outof a relative clause in an S-initial subject NP or in a fronted object NP. It also correctly rulesout ungrammatical sentences such as (83) below: the NP containing the gap in the object positioncannot be a CP-adjoined extraction domain:(83)*zhangsan1,wo bu zhidao[,ni yijing mai-le[, [,e1 xie t] de shuJ].Zhangsan I not know you already buy-asp write DE book*‘Zhangsan,I do not know that you have already bought books that t wrote.’27AS James Huang (p.c.) pertinently points out, Lasnik and Saito’s proposal only exempts the S-initialphrase from barrierhood, but not any barrier it contains. Thus a relative clause contained in an NP is not debarrier-ized if the containing NP occurs in an A’-position. Mike Rochemont (p.c.) suggests that the extraction inquestion can be accounted for if we assume that (i) the relative clause itself first moves out of the containing NP,and lands in an A’-position, (ii) the containing NP then moves, again to an A’-position, (iii) extraction occurs outof the relative clause. But the question is that there is no independent evidence to show that both the head and therelative clause can move either simultaneously as proposed here or separately.Another hypothesis is that since the relative clause is in an A’-position, extraction out of it is allowed solong as the containing NP is A’-adjoined. In other words, a relative clause in Chinese is not a barrier for extractionout of it: whether extraction is eventually allowed or not depends on the position the containing NP occupies. Thishypothesis is in line with Lasnilc and Saito’s proposal adopted in the text, but it also shares the stipulative natureof their proposal.Fiengo et al (1988) also propose that QR at LF voids barrierhood.66Extractions out of a sentential subject and out of a sentential complement to a subject are subjectto the same analysis.2.4. SummaryIn this chapter, I have investigated various kinds of object NP fronting. I propose thatwhen an S-initial NP is construed with a gap in the following ‘comment clause’, this gap shouldbe analyzed as the trace left by the movement of the NP. The major piece of evidence for thisclaim is that if the empty category is embedded in an island, a subjacency effect obtains.Furthermore, I show that there is a complication regarding the subjacency effect: it is suspendedwhen the NP containing the gap is an S-initial subject or an object fronted to an S-initialposition. I propose, based on Lasnik and Saito (1992), that it is because they are located in anA’-position that they cease to be barriers for extraction.Having established that there is movement in cases of NP fronting, I then apply severaldiagnostic tests to this syntactic process. I propose that Chinese has two kinds of short-distanceNP fronting: one is A’-movement and the other is A-movement, as is the case in Hindi andJapanese. Assuming the Split Infi Hypothesis, I further propose that A-movement involvesraising to [Spec AgrOPJ as a kind of overt raising for feature-checking, with the subject in thelower [Spec AspP]; on the other hand, the fronted NP displaying A’-movement properties is CPadjoined after it first lands in [Spec AgrOP] (with the subject in the higher [Spec AgrSP]).67Chapter 3 Object NP ShiftThis chapter focuses on investigating the syntactic properties of objectshift, as illustratedin (1):(1) wo zheben shu1 yijing zai huayuan-li kanwan-let.I this book already at garden-in finish-read-aspLit: ‘I, this book, have already finished reading t in the garden.’Intended: ‘I have already finished reading this book in the garden.’In example (1), the object NP is shiftedto a position after the subject. Phonologically, there isa slight pause after the subject. If the subject is a long one, a comma is used in writing.For easeof presentation, I refer to cases of this kind as object shift, reservingthe term “NP fronting”for cases where the object is moved to the S-initial positionas discussed in Ch. 2.3.1. Properties of Object ShiftObject NP shift in Chinese has not received as much attention asNP fronting (but seeZ. Lu 1991, F. Lu 1993). In this section, I examine the properties ofthis process in an attemptto establish a descriptive generalization about it.3.1.1. Semantic RestrictionsThere are semantic restrictions regarding what kind of objectscan be shifted. As withthe fronted NP, the shifted object must be a definite or genericNP denoting theme, and thesubject is preferably an agent (Liu et al 1983:270). In the following sentences, theserequirements are not met, making it impossible for the object to be shiftedor fronted:68(2) a. ta daozai-le di-shang. (object isa locative NP.)he fall-asp floor-on‘He fell onto the floor.’b.*tadi-shang daozai-le t. (No object shift)he floor-on fall-asp‘He fell onto the floor.’c.*dishang1ta daozai-le t1. (No object fronting)floor-on he fall-asp‘He fell onto the floor.’(3) a. bianlun chixu-le sange xiaoshi. (object is a temporal NP)debate last-asp three hours‘The debate lasted three hours.’b.*biajflunsange xiaoshi1chixu-le t1. (No object shift)debate three hours last-asp‘The debate lasted three hours.’c.*sangexiaoshi, bianlun chixu-le t1. (No object fronting)three hours debate last-asp‘The debate lasted three hours.’The theme NP in locative inversion (cf. Qu 1993), which mustbe indefinite, cannot be shiftedor fronted either:(4) a. tai-shang zuo-zhe yibang xuesheng.rostrum-on sit-asp one-group students‘On the rostrum sit a group of students.’b.*j..shgylbang xuesheng1zuo-zhe t1.rostrum-on one-group students sit-asp‘On the rostrum sit a group of students.’c.*yibangxuesheng tai-shang zuo-zhe t.one-group students rostrum-on sit-asp‘On the rostrum sit a group of students.’Although object shift and NP fronting are subject to the same semantic restrictions, theformer is more restricted than the latter. F. Lu (1993:4) proposes the followingconstraint69regarding object shift: when both the subject and object are animate NPs,the object cannot beshifted, only fronted. Her constraint is formulated as follows:(5) Animacy generalization:*Subject NP + Object NP + V[+animate] [+animate]Her constraint correctly rules out reading (b) of the followingsentence:(6) John Bifi jian-guo t.John Bill met-asp(a) ‘Johns, Bills has met ti.’(‘John’ as the fronted object)(b) Lit. ‘John1,Bills, has metIntended:*‘Johnhas met Bill.’(‘Bill’ as the shifted object)Reading (6a) is allowed since there is no constraint against an animate subjectNP preceded bya fronted animate object NP. Reading (6b) is out since an animate objectNP is shifted andfollows an animate subject NP. However, her constraint fails to rule out reading(b) in thefollowing sentence:(7) niuyue sbibao, tade wenzhang yijing piping-guolet.New York times his article already criticize-asp(a) New YorkTimes,his articles has already criticized ti.’(‘New York Times’ as the fronted object.)(b) Lit. ‘New York Times1,his articles, has already criticized t.’Intended:*NewYork Times1has already criticized his articlej.’(‘His article’ as the shifted object.)In this sentence, both NPs are inanimate, but the second NP cannot be interpretedas the shiftedobject as is shown in reading (7b). This fact cannot be accounted for by Lu’s animacy constraint.70Z. Lu (1991:5 1) also proposes a constraint, and he intends itto apply to both NP frontingand object shift. His idea is that the NP inobject position (NP0)cannot undergo either frontingor shift if the NP in subject position (NP,) and NP0 can switch thematicroles without violatingselectional restrictions of the verb. Look at the following sentence withsuch a symmetricpredicate as ‘meet’:(8) John jian-guo Bill.John meet-asp Bill‘John has met Bill.’If we switch the positions of the two NPs in (8), we get (8’): the selectionalrestrictions of theverb jian ‘meet’ are not violated though the meaning of the sentenceis different:(8’) Bifi jian-guo John.Bill meet-asp John‘Bill has met John.’According to Z. Lu, since the two NPs in (8) can switch thematic roles without violatingtheselectional restrictions of the verb, NP0 in (8) should not be fronted or shifted,or to quote him(1991:5 1), it should not “undergo either pre-O movement or topicalization”.Suppose we frontand shift the object in (8):(9) a.*Bffl,John1jian-guo t. (NP fronting)Bill John meet-asp‘Bill, John has met.’ (The judgement is Lu’s.)b.*JohnBIHJ jian-guo t. (Object shift)John Bill meet-asp‘John has met Bill.’71Crucially, Lu judges both examples in (9) as ungrammatical,which conforms to his constraint.But evidence abounds in the literature showing that sentenceslike (9a) are grammatical (Xu andLangendoen 1985, F. Lu 1993). I also judge (9a)as grammatical, and thus consider hisconstraint too strong, although his basic insight is correct.Given the empirical facts in (9), I will restate Lu’s (1991) descriptivegeneralization asin (10):(10) (a) If NP, and NP0 cannot switch theta roles, NP0 canbe either fronted to the S-initialposition or shifted to a position after the subject, provided the definitenessrequirement of theNP is met.(b) If NP, and NP0 can switch theta roles, then NP0 can only be frontedto the S-initialposition, not shifted.The constraint in (10) can be formalized as follows:(11) Formalization of the constraint:*NP,+ NP01+ V + t(where NP, and NP0 can switch 0 roles)One way to interpret this constraint is that if there is an ambiguity due to the fact that either oneof the two S-initial NPs can be the thematic agent, it shouldbe disambiguated by treating thefirst NP as the fronted object. The reason that OSV is chosen over SOV isthat OSV is thedefault option. Support for this claim comes from crosslinguistic evidence:fronting (or‘topicalization’) is more common than object shift. For instance, English only hastopicalization,72but no object shift.’Here is an example where the subject NP and the object NP CANNOT switche roles:(12) John jian-guo zhezhong hua.John see-asp this kind plant‘John has seen this kind of plant.’(12’) After switching e role:*zhezhonghua jian-guo John.this plant see-asp John*‘Thiskind of plant has seen John.’As the two NPs cannot switch e roles without violating selectional restrictions of the verb (see(12’)), it falls outside of the constraint and the object NP can be both fronted and shifted:(13) a. Object NP fronting of (12) is OK:zhezhong hua, John jian-guo t.this plant John see-asp‘This kind of plant, John has seen.’b. Object shift of (12) is also OK:John, zhezhong hua jian-guo t.John this plant see-asp‘John has seen this kind of plant.’‘James Huang (p.c.) also notes that object shift in Chinese is a fairly marked construction as opposed to objectfronting, usually requiring a contrast. He further observes that although there is a general preference for interpretingNP NP V as a case of object fronting (OSV), but not as a case of object shift (SOV), this preference is completelynon-existent in the following sentence, with the focus marker han. The following sentence allows both SOV andOSV reading equally:(i) Zhangsan han Lisi dou bu pa.Zhangsan even Lisi all not feara. ‘Zhangsan is not even afraid of Lisi.’ (SOV)b. ‘Lisi is not even afraid of Zhangsan.’ (OSV)I have no explanation for why the presence of han undermines the reading preference noted in the text.73To sum up, NP fronting and object shift are very similar processes. The formeris likethe default case of object dislocation while the latter is an additional optionfor an object NPwhich cannot switch theta roles with the subject.2’33.1.2. Object Shift and BA ConstructionThe object-shift construction looks similar to the BA-construction in having theobjectpreverbally (see Li and Thompson 1981, Cheng 1991, Z. Lu 1991 on BA constructions):(14) wo yijing zai huayuan-li BA zheben shu kanwan-le.I already at garden-in BA this book finish-read-asp‘I have already finished reading this book in the garden.’2For some speakers, if the subject is a pronoun and can switch theta roles with the object, theobject can be notonly fronted as in (ib), but also shifted as in (ic), contraiy to our constraint:(1) a. tamen jian-guo John.they meet-asp John‘They have met John.’b. John1 tamen jian-guo t. (NP fronting)John they meet-asp‘They have met John.’c. ?tamen Johns jian-guo t. (Object shift)they John meet-asp‘They have met John.’ (or: ‘John has met them.’)There are two possible solutions to this exception: one is that a pronoun is usually not topicalized(Henry Davisp.c.); the other is that a pronoun is deeper in the animacy hierarchy than anNP and is hence more agentive (JamesHuang p.c.). These two ideas may explain why the pronoun in (c) can be treatedas the agent.3One may replace the idea of theta role switching by the notion of ‘symmetric predicate’. For instance, one mayclaim that (8) involves a symmetric predicate, and that it is the use of this particular predicate that makes itimpossible for the object NP to be shifted. But as Mike Rochemont (p.c.)points out, the use of a ‘symmetricpredicate’ is not isomorphic to switching theta roles. For instance, (12) uses the same ‘symmetric’ predicatejian‘see, meet’ as in (8), but its subject is [+human] and its object is [-human]. Therefore the two NPsin (12) cannotswitch theta roles as is shown in (12’). In other words, it is not true that the two NPs ofa symmetric predicate canalways switch theta roles. Because the NPs in (12) cannot switch thetaroles, the object NP can be shifted asillustrated in (13b) despite the use of a ‘symmetric’ predicate. Given the set of data in (12-13),I follow Z. Lu(1991) in using the non-theoretical term ‘theta-role switching’, rather than ‘symmetric predicate’,as the constraintfor object shift.74However, there are differences between object shift and BA construction. One ofthem is thatthe shifted NP can only precede an adjunct PP, as shown in (15a) below,not follow it, as isshown in (15b):(15) a. wo zheben shu1 yijing zai huayuan-li kanwan-let.I this book already at garden-in finish-read-asp‘I have already finished reading this book in the garden.’b.*wo yijing zai huayuan-li zheben shut kanwan-le t1.I already at garden-in this book finish-read-asp‘I have already finished reading this book in the garden.’In contrast, the BA-NP in (14) above can follow the PP, but not precede itas in (l6):(16)*?woBA zheben shu vijing zai huayuan-li kanwan-le.I BA this book already at garden-in fmish-read-asp‘I have already finished reading this book in the garden.’Another difference between these two constructions is the position of the negator: the BA-NP must follow the negator (see Li and Thompson 1981:478-480) while the shifted NP mustprecede the negator in the object shift constructions.4Liu et al (1983:476) distinguish two kinds of positions for adjunct PPs with respect to the BA-NP. Takethelocative PP for instance. If the PP also denotes the location of the subject, then it should precede the BA-NP.Thisis exemplified in example (14), which means that the subject ‘I’ is ‘in the garden’. If, however, the locative PP onlydenotes the location of the BA-NP, it must follow the latter as is illustrated by the contrast in (i-u):(i) wo BA daozi zai moshi-shang mohao-le.I BA knife on grindstone sharpen-asp‘I have sharpened the knife on the grindstone.’(The agent is not on the grindstone, only the knife.)(ii) wo zai moshi-shang BA daozi mohao-le.i on grindstone BA knife sharpen-asp‘I sharpened the knife (while I was) on the grindstone.’(semantically anomalous)75(17) a. wo mel ba Zhangsan qinglai.I not BA Zhangsan invite.‘I failed to get Zhangsan to come over.’a’. ba Zhangsan mel qinglai.I BA Zhangsan not invite‘I failed to get Zhangsan to come over.’b. mel Zhangsan1qinglai t.I not Zhangsan invite‘I failed to get Zhangsan to come over.’b’. wo Zhangsan1mel qinglai t1.I Zhangsan not invite‘I failed to get Zhangsan to come over.’The third difference is that the BA-NP must be definite, not generic, whilethe shiftedNP can be generic or definite.5(18) a. ba pijiu he-guo.I BA beer drink-asp‘I have had the experience of drinking beer before.’b. wo pljiu he-guo.I beer drink-asp‘I have had the experience of drinking beer before.’c. wo ba pijiu he-le.I BA beer drink-aspI drank (a specific bottle of) beer.d. ?wo pijiu he-le.I beer drink-aspI drank (a specific bottle of) beer.As shown by examples (18b) and (18d), it is theaspect marker that seems to determine thegeneric or definite reading of the shifted NP. Even if weuse the “right” aspect marker guo, the5Mahajan (1991) observes that the scrambled object in Hindi is specific. Diesing(1992) also notes that the NPoutside the VP domain is definite or generic. See below for more discussion.76BA-NP in (18a) does not get the generic reading, unlike object shift in (18b).6This establishesthat the BA-construction is not the same operation as object shift.In this chapter, we will focus on object shift constructions and discuss the syntacticproperties of this process. Below I will first discuss the structural position of the shifted objectwith respect to other constituents such as adverbs and modal verbs. Then I will apply somediagnostic tests to the syntactic operation. I argue, based on these tests, that object shift is anA-type movement, as in some Scandinavian languages (Deprez 1989). Adapting Chomsky andLasnik (1991), I further propose that the shifted object lands in the specifier position of AgrOP.3.2. Interaction of the Shifted Object with Other Constituents3.2.1. Interaction with Adverbs: the Shifted NP is to the Left of VPChinese has three kinds of adverbs, which have different distributions (Li and Thompson1981:320, Chen 1987). They are (1) sentential adverbs, (2) non-manner adverbs and (3) manneradverbs. I assume that sentential adverbs are either VP-adjoined, which allows us to mark theVP boundary (cf. Diesing 1992:32 and Deprez 1989), or they are adjoined to any of thefunctional projections above VP; manner adverbs are V-adjoined (or V’-adjoined butimmediately adjacent to the verb); non-manner adverbs are V’-adjoined.76The function of these different aspect markers constitutes a very interesting topic for future research. See Z.Lu (1991:38) for some discussion of aspect and object shift. He proposes that perfective aspect licenses thissyntactic process. However, this is falsified by data like (i):(i) wo zhexie dianying hen xiang kan.I these movies very like see‘I would like to see these movies very much.’7Sportiche (1988:432) assumes that sentential adverbs in French are adjoined to IP or I. Rizzi (1990:50) assumesthat sentential adverbs like reason are adjoined to TP or AgrP and manner adverbs are adjoined to VP.77(A). Sentential adverbs occur either before or after the subject, ora fronted NP if any.Examples of sentential adverbs are jintian ‘today’, dagai ‘probably’ and xingkui‘fortunately’.They do not occur after the PP adjunct. Like other adverbs,they never occur between the verband the object or sentence-finally.(19) S—initially:xingkui wo zai jia-li kan-guo zheben shu.fortunately I at home-in read-asp this book‘Fortunately, I read this book at home.’(20) Between the subject and the PP adjunct:wo xingkui zai jia-li kan-guo zheben shu.I fortunately at home-in read-asp this book‘Fortunately, I read this book at home.’(21) Either before or after the fronted NP:a. xingkui zheben shut, wo kan-guo t1.fortunately this book I read-asp‘Fortunately, this book I have read.’b. zheben shu1 xingkui wo kan-guo t1.this book fortunately I read-asp‘Fortunately, this book I have read.’(22) Not after the PP adjunct, between the verb and the object or sentence-finally:a.?*wozai jia-li xingkui kan-guo zheben shu.I at home-in fortunately read-asp this book(*if using other sentential adverbs)b. zai jia-li kan-guo xingkui zheben shu.I at home-in read-asp fortunately this book‘Fortunately, I read this book at home.’c. wo zai jia-li kan-guo zheben shu xingkui.I at home-in read-asp this book fortunately‘Fortunately, I read this book at home.’78(B). The distribution of manner adverbs is limited tothe position immediately before the verb.Such adverbs include manmande ‘slowly’renzhende ‘conscientiously’, xiaoxinde ‘carefully.’(23) Immediately before the verb:wo zai jia-li manmande kanwan-le zheben shu.I at home-in slowly read-asp this book‘I finished reading this book at home slowly.’(24) Not S-initially:*manjnandewo zai jia-li kanwan-le zheben shu.slowly I at home-in read-asp this book‘Slowly, I finished reading this book at home.’(25) Not between the subject and the PP adjunct:*womanmande zai jia-li kanwan-le zheben shu.I slowly at home-in read-asp this book‘Slowly, I finished reading this book at home.’(C). Non-manner adverbs, like manner adverbs, cannot occur sentence-initially;like sententialadverbs, they can occur between the subjectand the PP adjunct. Such adverbs include henshao‘rarely’, guyide ‘purposely’ (Chen 1987).(26) Before PP (and after the subject) or after PP:a. wo jingchang zai jia-li kan shu.I frequently at home-in read book‘Frequently, I read books at home.’b. wo zai jia-li jingchang kan shu.I at home-in frequently read book‘I frequently read books at home.’(27) Not immediately before the verb (after the manner adverb):zai jia-li manmande jingchang kan shu.I at home-in slowly frequently read book‘Frequently, I read books slowly at home.’79(28) Not S-initially:*jjngchangwo zai jia-li kan shu.frequently I at home-in read book‘Frequently, I read books at home.’Given the distributions of various types of adverbs,I will provide object shift data below toestablish where the shifted object is with respect to theseadverbs.The shifted object can occur either before or aftera sentential adverb such as xingkui‘fortunately’ whose position is marked by #:(29) #1 wo ?#2 zheben shu, #3 zai huayuan-li kanwan-let1.I this book at garden-in finish-read-asp‘Fortunately, this book, I have finished reading in the garden.’The shifted object occurs before (the PP and) the manner adverbas in (30a), not after them asin (30b-c):(30) a. wo zheben shu1 zai huayuan-li manmande kanwan-let,.I this book at garden-in slowly finish-read-asp‘I have finished reading this book slowly in the garden.’b.*wzai huayuan-li zheben shut mamnande kanwan-let1.I at garden-in this book slowly finish-read-asp‘I have finished reading this book slowly in the garden.’c.*wozai huayuan-li manmande zheben shut kanwan-let1.I at garden-in slowly this book finish-read-asp‘I have finished reading this book slowly in the garden.’Further as shown in (31a) below, the shifted object must precede thenon-manner adverb, butnot follow it, as is shown in (31b):80(31) a. ta zhezhong shi1 henshao zai xuexiao-lishuo t1.he this thing rarely at school-in say‘He rarely mentions this kind of thingsat school.’b.*henshao zhezhong shi1 zai xuexiao-li shuo t.he rarely this thing at school-insay‘He rarely mentions this kind of things at school.’Notice that even if the shifted object precedes thenon-manner adverb, the sentence is still badif it follows the PP:(32)?*tazai xuexiao-li zhezhong shi1 henshao shuo t.he at school-in this thing rarely say‘He rarely mentions this kind of things at school.’(*if using such non-manner adverbs asguyide ‘deliberately’.)This indicates that the shifted object must precede not onlythe non-manner adverb, but also theadjunct PP.To summarize the interactions between the shifted object and adverbs:(1) it can be preceded or followed by the sentential adverb;(2) it must be followed by the non-manner adverb and the manneradverb as well as PP,The conclusion based on these interactions is that the shiftedobject must be to the leftof VP. Schematically, the positions of the constituents involvedare represented below:(33) Subject + S-adv. + Shifted Object, + S-adv. + non-manneradv. + PP + non-manneradv. + manner adv. + V + t.88Au exemplification of (33) with all the adverbs present is givenin (i):(i) ?wo xinkui zheben shu jintian zai jia-li guyide manmande kanwan-le.I fortunately this book today at home-in deliberately slowly finish-read.813.2.2. Interaction with modals: the Shifted Object isto the Left of Modals:Chinese modals include the following: yinggai ‘should’, gan ‘dare’, bixu ‘must’,neng‘can’.9 As in English, they always precede the main verb.But they must follow the shiftedobject, if any. Interaction with modals and PP:Take the modal verb neng ‘can’ for instance. The canonical word order ofa sentencecontaining this modal is as follows:(34) ta neng zai tianhei-qian ganwan zhejian shi.he can at dark-before complete this matter‘He can finish this thing before dark.’Consider cases where the object is shifted. As shown in(35a), it is acceptable for the object toland in a position before the modal verb, but not after the modal verband before the PP as in(35b), or after both the modal verb and the PP as in (35c):(35) a. ta zhejian shi, neng zai tianhei-qian ganwan t.he this matter can at dark-before complete‘He can finish this thing before dark.’b.*taneng zhejian shi1 zai tianhei-qian ganwant.he can this matter at dark-before complete‘He can fmish this thing before dark.’‘Fortunately, I read this book deliberately slowly at home.’The unnaturalness of (i) is due to the stacking up of adverbs, just as in its English translation.9Modals are not raising verbs, since they never take aspect markers as verbs do. Also noticethat yinggai‘should’ has some idiosyncratic properties which are not shared by othermodal verbs. For instance, it can co-occurwith another modal verb. Moreover, it can occur S-initially. Cheng (1989) treatsit as a kind of sentential adverb.82c.*taneng zai tianhei-qian zhejian shi1 ganwan t1.he can at dark-before this matter complete‘He can finish this thing before dark.’The descriptive generalization from the data above isthat the shifted NP must precede the modalverb. In other words, it looks as if itoccupies a position higher than the functional projectionheaded by the modal verb. Interaction with a modal verb and anadverb:In this section, we briefly examine the interaction of the shiftedobject with respect to themodal verb and adverbs.A. Sentential adverb: As is shown in (36),the shifted object can precede the modal verb, butfollow the sentential adverb:(36) ta jintian zheben shu, neng kanwant.he today this book can finish-read‘He can finish reading this book today.’In (37a-b) below, the shifted object can precede both the modalverb and the sentential adverb:(37) a. ta zheben shut neng jintian kanwan t1.he this book can today finish-read‘He can finish reading this book today.’b. ta zheben shut jintian neng kanwan t1.he this book today can finish-read‘He can finish reading this book today.’(38) below is ruled out since the modal precedes the shifted NP.83(38)*taneng zheben shu1jintian kanwan t1.he can this book today finish-read‘He can finish reading this book today.’B. Manner adverb: The following set of data indicate thatthe shifted object must precede themodal verb which must in turn precede the manner adverb:(39) a. ta zheben shu1 neng renzhende kanwan t4.he this book can conscientiously finish-read‘He can finish reading this book conscientiously.’b.*zheben shu, renzhende neng kanwan t1.he this book conscientiously can finish-read‘He can finish reading this book conscientiously.’c.*neng zheben shu1 renzhende lcanwan t1.he can this book conscientiously finish-read‘He can finished reading this book conscientiously.’To conclude, the interaction of the shifted object with respect to the modal and adverbsconfirms the descriptive generalization reached earlier. It must precedethe modal and themanner adverb, but it can precede or follow the sentential adverb.3.2.3. Interaction with Negators: the Shifted Object isto the Left of the NegatorChinese has different kinds of negators; the most common ones are meiyou‘not’(perfective) and bu “not” (future or volitional) (cf. Li and Thompson Ch.8). Both of themprecede the main verb. Now consider the following negative sentences:(40) a. Canonical word order:John meiyou kan-guo zheben shu.John not read-asp this book‘John has not read this book.’84When the object is shifted, it must precede the negator, notfollow it:(40) b. Object shift:John zheben shu, meiyou kan-guo t1.John this book not read-asp‘John has not read this book.c.*Johnmeiyou zheben shu1 kan-guo t1.John not this book read-asp‘John has not read this book.Examples (40b-c) show that the object must be shifted to the left of the negator,not to the rightof it. Let us assume that a negator in Chinese is an adverb adjoinedto V’ or to the single barlevel of one of the functional projections’°. The data above are consistentwith our conclusionreached earlier: the shifted NP is to the left of VP.‘°The claim that a negator is just an adverb is supported by the following data:(i) ta bu keyi laihe not may come‘He may not come.’(or: ‘It is not all right for him to come.’)(ii) ta keyi bu lai.he may not come‘He may not come.’(or: ‘It is all right for him not to come.’)The fact that a negator can occur either before or after the modal indicates that thereis no NEGP, and that thenegator is simply adjoined as an adverb to V’ (like a non-manner adverb) or to thesingle-bar level of a functionalprojection. Deprez (1989:223) proposes that Negation is a sentential adverb, and doesnot head its projection inIcelandic and Mainland Scandinavian languages.Chen (1987) claims that a sentential adverb cannot occur in a position aftera negator. But the followingexample shows that this is a not valid claim:(iii) wo bu neng jinatian lai.I not can today come‘I cannot come today.’Under my analysis, bu as an adverb can be adjoined to a functional projection whilejintian can be VP-adjoined.85What if there is a negator modifying a modal? Look at the following sentence:(41) ta bu keyi kan zheben shu. (canonical construction)he not may read this book‘He may not read this book.’If the object is shifted, it can only land between the subject and the negator as in (42a), notbetween the negator and the modal verb as in (42b), or between the modal verb and the verb asin (42c):(42) a. ta zheben shu1 bu keyi kan t.he this book not may read‘This book, he may not read.’b.*tabu zheben shu1 keyi kan t.he not this book may read‘This book, he may not read.’c.*tabu keyi zheben shu1 kan t..he not may this book read‘This book, he can not read.’This group of data indicate that the correct linear order of the shifted object with respect to thenegator and the modal verb is that it should be the first element after the subject. Now comparewith the following sentence, where the negator is between the modal verb and the main verbwhich is modified by the negator.(43) ta yinggai bu kan zheben shu. (canonical construction)he should not read this book‘He should not read this book.’Notice that even in this case, the shifted object must be the first element after the subject, as isshown in (44a). It cannot land between the modal and the negator (44b) or between the negatorand the verb (44c):86(44) a. ta zheben shu1 yinggai bu kan t1.he this book should not read‘This book, he should not read.’b.?*tayinggai zheben shut bu kan t1.he should this book not read‘This book, he should not read.’c.*tayinggai bu zheben shu1 kan t.he should not this book read‘This book, he should not read.’The significance of this set of data is that, whether the negator modifies the modal verb or themain verb, the shifted object must precede the modal and the negator.To conclude, interactions of the shifted NP with adverbs, modals and negators indicatethat except for the sentential adverb, the shifted NP should be the first constituent after thesubject, preceding the modal and its negator as well as other types of adverbs and PPs.3.3. Object NP Shift and its Landing Site:To summarize the descriptive generalizations discussed above, here is a schematicrepresentation of the positions of the elements involved:(45) subject - S-adv. - Shifted Obj. - S-adv. - modal - S-adv. non-mnr adv. - PP -non-mnr adv. - manner adv. - V - t]In this section, we will examine the syntactic position of the shifted object NP. Z. Lu(1991:53, 69) assumes, without arguments, that the shifted object is VP-adjoined, while F. Lu(1993:7-9) claims that it is adjoined to ModaiP since it is higher than the modal verb. What87these two proposals imply is that the shifted object lands in an A’-position. However, therearereasons to believe that their analyses are not correct.3.3.1. Object NP Shift as an A-movement:In this section, I apply some of the diagnostic tests used in Section of Ch. 2. Iwill argue that the shifted object displays A-movement properties, not A’-movement properties.The three tests used in section of Ch.2 for A-movement are (a) the ability to remedyWCO violations, (b) the absence of reconstruction effects, and (c) the ability of the displacedobject NP to serve as an A-binder for an anaphor in the subject position. The last test does NOTapply here since the object does not cross the subject.”Test A: The ability to remedy WCO violations: In (46a) below, the pronoun cannotbe co-indexed with the universal quantifier object to its right. But if the quantified NP is shifted,co-indexing becomes acceptable as is shown in (46b):“I used two tests for A’-movement in section of Ch.2. One was ‘Reconstruction Effect’ test where theanaphor contained in the fronted object can still be bound by the subject. The other test was ‘Condition C typeReconstruction Effect’ test where a name contained in the fronted object NP cannot be co-indexed with thepronominal subject unless the name is further embedded in the fronted NP. As both tests rely on the subject acrosswhich the object NP is fronted to the S-initial position, they are not applicable here since thesubject is not crossed,at least in the S-structure representations. For instance,(i) John, bu xihuan [ziji, de wenzhang].John not like self DE papersLit.: ‘John, does not like self’s, papers.’(ii) John, [ziii; de wenthang dou bu xihuan .John self DE papers even not likeLit.: ‘John,, self’s1 papers, does not like.’Although ‘self’ in the shifted object is still bound by the subject ‘John’ as is shown in (ii), it does not tell usanything about the nature of the position of the shifted object NP. Also note that the adverb dou in (ii) means ‘even’(see Liu et al 1983).88(46) a.*wo[zai tamende, jiaoshi-li] jiandao-le suoyoude xuesheng,.I in their classroom-in meet-asp all students‘I have met all the students in their classroom.’b. wo suoyoude xuesheng, dou[zai tamende1jiaoshi-li] jiandao-le t1.I all students all in their classroom meet-asp‘I have met all the students in their classroom.’12The ungrammaticality of (46a) is due to a WCO configuration:the universal quantifier is subjectto LF Quantifier Raising which is an A’-movement, A’-binding both the crossed pronoun andits variable. In (46b), although the universal quantifier object moves across the co-indexedpronoun at S-structure, it does not induce a WCO effect. This indicates that the S-structureshiftof the object is an A-movement, not an A’-movement.Test B: The absence of reconstruction effects: As demonstrated in Section ofCh.2, when both the direct object and indirect object are in situ, the preceding indirect objectcannot be a pronoun co-indexed with the NP in direct object position. But when the direct objectis fronted, coindexing becomes acceptable. This co-indexing possibility also appliesto objectshift.‘2Notice that in (46b) there is the adverb of universal quantification dou though it is absent in (46a). Thedifference in grammaticality judgement between these two sentences has nothing to do with its presence or absence.Dou is independently required when a universal quantifier object is shifted or fronted, as observed in Ch.2:(i) wo(*doudu-guo suoyoude jufashu.I all read-asp every books-on-syntax‘I have read all the books on syntax.’(ii) wo suoyoude jufashu1*(dou)du-guo t1.I every books-on-syntax all read-asp‘I have read all the books on syntax.’89(47) a. Canonical word order:gei-le ta1 John1 de shu.I gave-asp him John DE bookgave him1 John1’s book.’b. Direct object fronted (from Ch.2):[John1de shu] wo yijing gei-le ta1 t.John DE book I already gave-asp him‘John1’s book, I already gave him1.’c. Direct object shifted:?wo [John1de shu] yijing gei-le ta1 t.I John DE book already gave-asp him‘John1’s book, I already gave him1.’Since reconstruction applies to the site of a variable, and there is no reconstruction effect,i in(47b-c) must be an NP trace, not a variable, and the movement is an A movement.Another example can be constructed using the Chinese equivalent of the English verb‘show’, as in (48) below:(48) a. wo geiimfl1kan-guo[zhezhang zhaopian].I to John see-asp this photo‘I showed John this photo.’b. wo geii2fl1kan [taziji1de zhaopian].I to John see himself DE photos‘I showed John his own photos.’The complex anaphor in (48b), unlike a simple anaphor, is not subject-oriented(cf. Tang 1989)and can be bound by the prepositional object. If we shift the object containing the anaphor asin (49b), the sentence becomes ungrammatical, even though theobject in this kind ofconstructions can be shifted independently as shown in (49a):90(49) a. wo [zhezhang zhaopian]gei John, kan-guo t1.I this photo to John see-asp‘I showed John this photo.’b.*i[taziji de zhaopian] gei John kan-guo t.I himself DE photos to John see-asp‘I showed John his own photos.’If the shifted object lands in an A’-position, the anaphor shouldbe able to be bound underreconstruction at LF by the prepositional object and the sentence shouldbe good. Lack ofreconstruction effect in (49) indicates that the landing site of the shifted directobject is not anA’-position, but an A-position.Test C: Locality Constraint: It has been shown in GB literature that A-type movementis bounded (cf. Chomsky 1986ab among others). For instance, “superraising”,an unbounded A-movement, is ruled out as in the following English sentence:(50)*John1seems [that it is likely [t to win the game]].As is shown below, Chinese object shift is also bounded. While the embedded object can beshifted to a position after the subject of its own clause as in (51a), it cannot be shiftedto aposition after the subject of the matrix clause, as is shown in (51b):(51) a. wo renwei ta zheben shu1 yijing kanwan-le tj.I think he this book already read-asp‘I think that he has finished reading the book.’b.*wozheben shu, renwei ta yijing kan-guo ti].I this book think he already read-aspLit.:*‘I,this book, think that he has read ti.’91If object shift is an A’-movement, we should expect that long-distance shiftas in (51b) shouldbe grammatical as in other long-distance A’-movement. Compare with the following sentence,which is a topicalization or fronting process across a CP boundary:(52) zheben shu1,wo renwei [ta yijing kanwan-le t1.this book I think he already finish-read-asp‘This book, I think he has finished reading [it] already.’As argued in Ch. 2, long-distance fronting in Chinese is an A’-movement, like long-distancescrambling in Hindi. The contrast between (51b-52) indicates that while NP fronting involvesA’-movement and is therefore unbounded, object shift can only be A-movement and is thereforelocal. See section 3.3.4. for why object shift can only be local.To conclude, the shifted object displays three syntactic properties: (a) it remedies WCO;(b) it shows no reconstruction effect; (c) it is clause-bounded. These properties indicate thatobject shift is an A-movement, not an A’-movement. The fact that the shifted NP exhibits A-properties thus invalidates the A’-adjunction analysis as proposed in Z. Lu (1991) and in F. Lu(1993).A-type object shift is also found in other languages. For example, Tada (1990) as quotedin Saito (1992:111 fn. 7 and p.114 fn. 37) notes that Japanese has a so-called VP-internalscrambling where the scrambled object does NOT cross the subject. Further, this VP-internalscrambling has all the properties of A-movement. It remains to be seen whether this VP-internalscrambling is really VP-internal. The fact that scrambling/shift in both Japanese and Chineseexhibits A-properties suggests they may be the same process.Mahajan (1990) claims that short distance scrambling can be either L- or non-Lmovement. In the data he provides (1990:34), the direct object does not display reconstruction92effect when it is shifted to a position after the subject,but it does when it is topicalized beforethe subject:(53) raam-ne1mohan-ko apnli kitaab lotaaii.Ram (SUB) Mohan (JO) self’ s book-f(D0) return-perf-f‘Ram, returned self book to Mohan.’ (his (45))(54) raam-ne, [apnlikitaab]kmohan-ko3tklotaaii.Ram (SUB) self’s book-f(D0) Mohan (10) return-perf-f‘Ram, returned self’s book to Mohan1.’(55) [apnlikitaab]kraam-ne1tkmohan-kotklotaaii.self’s book-f(D0) Ram (SUB) Mohan (JO) return-perf-f‘Ram1 returned se1f’s book to Mohan.’(54) resembles our object shift in that the direct object landsin a position after the subject.Notice that the anaphor contained in the shifted direct object can no longerbe bound by theindirect object across which the direct object has shifted. This, according toMahajan, indicatesthattkis not a variable and the shift is an L-movement. In contrast, the anaphor in thetopicalized direct object in (55) can still be bound by the subject, thus indicatingthat t’ is avariable and that the second step of movement is a non-L movement. What is relevantto ourdiscussion here is that the object shift demonstrated in (54) displays A-movement,exactly as inChinese and Japanese.13‘3According to James Huang (p.c.), in contrast to the unambiguous (55)in Hindi, the Chinese counterpart isambiguous, admitting both i and j reading:(i) taziji de fenshu, Zhangsan, yijing gaosu Lisi3 le.himself DE grades Zhangsan already tell Lisi particle‘His own grades, Zhangsan already told Lisi.’As Huang suggests, one way to account for this ambiguous reading is to explore the possibility to directlyfront thedirect object into the pre-subject position, rather than to first shift the direct object beforefronting it. One potentialproblem with this direct A’-fronting is that the direct object would be deprived of an opportunityof moving through[Spec AgrOP] for feature/Case checking. I have no solution to this problem.93Deprez (1989) discusses object shift in such European languages as MainlandScandinavian, Icelandic, German and Dutch. Her conclusion, based on similar teststo those usedhere, is that the object shift in these languages is also A-movement.3.3.2. The Shifted Object in [Spec AgrOP]As discussed in Ch.2, a fronted object NP can land in an A position since it can serveas an A-binder for an anaphor in the subject NP. For example:(56) John1,ziji, de haizi dou bu zunjing t1. (OSV)John self DE kids even not respectLit.: ‘John1,even self’ s kids do not respect ti.’Intended: ‘Even John’s kids do not respect John.’In (56), the fronted object NP acts as an A-binder for the anaphor in thesubject NP. Hencethe conclusion that there must be an A-position before/above theposition of the subject. I havealso shown in this chapter based on several diagnostic tests that a shifted NP lands inan A-position. This entails that there should be an A-position after the subject, but before the modalverb and VP. As subject is canonically in an A-position, there should be at least two A-positionsS-initially in a sentence with the OSV or SOV order:(57)a. A-positions in the OSV word order (in A-type fronting):A-position + A-position + modal + VP...(for fronted obj.) (for subject)b. A-positions in the SOV word order:A-position + A-position + modal + VP...(for subject) (for shifted obj.)94(57a) is illustrated by example (56) provided above, and (5Th) by example (58)given below:(58) John zheben shu1 bu xihuan t1. (SOV)John this book not like‘John does not like this book.’In Ch.2, adapting Pollock (1989), Mahajan (1990) and Chomsky and Lasnik(1991:33,82), I proposed the following clausal structure for Chinese:(59) CPI \AgrSP/ \AgrOP/ \(MP)/AspP/ \VP/ \subj. V’/ \V obj.I further proposed that the SVO and OSV word orders found in Chinese are derivedas follows:SVO: the canonical word order: The subject NP, base-generated in [Spec VP], movesovertly to [Spec AspP] as an intermediate step for Case assignment. (Alternatively,the subjectcan further move to [Spec AgrSP] for feature checking, though the S-structure effect could bevacuous in the SVO order.) The object NP remains as sister to the V and later undergoes covertmovement at LF to [Spec AgrOP] for feature/Case checking.95OSV: the word order for A-type fronting: The object NP undergoes overt raising to[Spec AgrOP]. The subject NP moves to and remains in [Spec AspP] at S-structure, but it latermoves to [Spec AgrSP] at LF for feature-checking.Now for the NP shift word order SOV, I propose that both the subject NP and the objectNP overtly raise to [Spec AgrSP] and [Spec AgrOP] respectively, and the result is the followingS-structure word order:’4(60) CPI \AgrSP/ \subj.1 AgrOP/ \obj. AspPI \t,i VP/ \ti V’I \VtjRecall that in discussing short distance A’-fronting in Ch. 2, I proposed that both the subject andthe object raise overtly to [Spec AgrSPJ and [Spec AgrOP] respectively before the object isfurther CP-adjoined, thereby displaying A’-movement properties. The fact that Chinese overtlyexhibits A-type object shift yielding SOV constructions lends strong empirical support for myt4Travis (1991) cites four analyses for the derived or scrambled object in the literature:(a) Mahajan’s analysis (1990): it is in [Spec AgrOP], hence outside VP (or [Spec AGR2 for ‘VP-internalshift’ 1990:28-29);(b) Johnson (1990): it is in [Spec VP], hence inside VP;(c) Sportiche (1990): it is inside VP in the Spec of the lower VP in the Larsonian sense of a layered VP.(d) Travis (1991): it is in Spec of a second ASPP located between a higher VP and a lower VP.The adopted structure for Chinese object shift resembles Mahajan’s proposal for Hindi.96proposal. Specifically, it justifies the ‘overt raising’ of the subject NP and the object NPas alegitimate intermediate step for deriving short-distance A’-fronting.Take the following sentence for instance:(61) John zheben shu1 kan-guo t.John this book read-asp‘John has read this book.’Its D-structure and S-structure are represented in (62a-b) respectively:(62) a. D-structure for (61):cPI \AgrSP/ \AgrOP/ \AspP/ \VPI \John, V’/ \VNPJkan-guo [zheben shu]97(62) b. S-structure for (61):CPI \AgrSPI \John1 AgrOP/[zheben shu] AspP/ \t’1 VP/ \1V,/ \kan-guoTo sum up, as NP shift is uniformly A-movement, I propose that the landing site of thisshifted NP is [Spec AgrOP]. In the next section, I will justify the postulation of AgrP in Chineseeven in the absence of morphological agreement.153.3.3. The Definiteness of the Shifted NP and the Postulation of Agr in ChineseThe three types of word order in Chinese are derived from the movement of the subjectand/or the object to the Spec positions of Agr projections. Given that Chinese has nomorphological agreement, an immediate question arises: how plausible is it to postulate an AgrPin the absence of morphological agreement in Chinese? It has been proposed in the literaturethatthere is a correlation between the definiteness of the moved NP and agreement. For instance,Mahajan (1991) observes that the shifted object in Hindi has the specific readingand it isanalyzed as moving into [Spec AgrOP]. Runner (1994:153) recently shows that there are five‘51t has been proposed in the literature that overt object shift is predicated on overt verb movement (Deprez1989, Mahajan 1990). Although Chinese has object shift as shown in this chapter, it does not exhibit overt verb (toInfi) movement. This suggests either that the proposed correlation between object shift or verb movement is wrong,or that object shift is not a unitary process crosslinguistically. See section 5.2. for more discussion.98ways in which the specificity reading of an NP correlates with Agr:(63) Five Manifestations of the Correlations between Specificity and Agr (Runner 1994:153):(a) DP itself bears certain Case, as in Turkish and Finnish;(b) DP triggers object agreement on the verb, as in Hindi;(c) both DP and the verb show Agr, as in Greenland Eskimo;(d) DP overtly surfaces in [Spec AgrP], as in Spanish, German and Catalan;(e) verbs overtly moves to Agr, as in Hungarian.What is relevant for our discussion here is (63d). For instance, in Catalan, when an indefinitesubject is postverbal as in (Ma), it has the existential/non-specific reading; in contrast, whenit is preverbal, i.e. in [Spec, AgrP] as shown in (64b), it gets the specific reading:(64) (a) Ha caigut un roc (b) Un roc ha caigut.has fallen a rock a rock has fallen‘A stone fell.’ ‘One (of the) stones fell.’The same paradigm obtains for subject NPs of Chinese unaccusative verbs as shown in (65):(65) (a) Zhao huo le. (a’) Huo zhao le.burn fire particle, fire burn particle‘Fire! Fire!’ ‘TIlE fire has started.= there exists A fire.(b) Chen chuan le. (b’) Chuan chen le.sink boat particle boat sink particle‘There sank A boat’ ‘THE boat has sunk.’= ‘A boat has sunk.’Since subject NPs of unaccusative verbs in Chinese behave like their counterparts in Catalan,I propose that the correlation between specificity and Agr in Chinese should belong to group (d)as discussed in Runner, i.e. DP overtly surfaces in [Spec Agr]. In fact, this conclusion basedon the unaccusative data provides independent support for my proposal for treating [Spec99AgrOP] as the landing site of the shifted object, which must be definite/specific.Without assuming the Split Infi Hypothesis, Diesing (1992) also shows that an (indefinite)NP in German must be outside the VP domain to obtain a specific reading. Recall that inChinese, the shifted and fronted NP must be definite or generic. The relevant data are repeatedbelow.(66) A. definite NP:a. zheben shu1,wo yijing kanwan-le t1. (fronting)this book I already finish-read-asp‘This book, I have already finished reading.’b. wo zheben shu yijing kanwan-le t. (shift)I this book already finish-read-asp‘This book, I have already finished reading.’B. generic NP:’6a. pijiu1,wo he-guo t1. (fronting)beer I drink-asp‘I have had beer before.’b. wo pijiu1 he-guo t1. (shift)I beer drink-asp‘I have had beer before.’C. indefinite NP:a.*yibenshut, wo yijing kanwan-le t. (fronting)one book I already finish-read-asp‘One book, I have already finished reading.’b.*iyiben shu1 yijing kanwan-le t. (shift)I one book already finish-read-asp‘One book, I have already finished reading.’‘6Following a proposal by Kratzer, Diesing (1992:95-96) argues that generics are associated with apresuppositional reading, and are hence specific. I assume that in Chinese generics also have a presuppositionalreading and therefore can be fronted and/or shifted on a par with definite NPs.100Since there is a correlation between the definiteness of the shifted NP and agreement inmorphologically rich languages, the obligatory definiteness of the shifted and fronted NP inChinese suggests that there are Agr projections in this language despite the absence ofmorphological evidence. This is on a par with the postulation of abstract Case in Chinese despitethe absence of its morphological manifestation. Furthermore, this correlation further motivatesour analysis that the landing site of the shifted NP and fronted NP (in A-type movement) inChinese should be the Spec of an Agr projection.’73.3.4. Evaluation of the Proposal:The [Spec AgrP] substitution analysis allows us to capture a number of facts. First, itaccounts for why only NPs, not other maximal projections like AP, can be shifted, for Spectypically accommodates NPs. In addition, the clausal structure postulated here gives us someintermediate A-positions before the VP: the spec positions of AspP, AgrOP and AgrSP. It thusaccounts for why the shifted NP displays only A-movement properties, given that the specposition of AgrOP is an L-related position, as proposed in Mahajan (1990).Second, it captures the same semantic restrictions for the fronted NP (to the S-initialposition) and for the shifted NP. They must be a definite or generic NP, as is shown at thebeginning of this chapter. For instance:(67) a. ta daozai-le di-shang. (Object is a locative NP.)he fall-asp floor-on‘He fell onto the floor.’‘7Some explanations have been attempted in the literature regarding why Agr is relevant for the specificityreading. Mahajan (1991) propose that Agr has pronominal features, thus making the NP in its Spec specific viaSpec-head agreement. Runner (1994) suggests that Agr can be anaphoric to discourse. I leave the fmal decision openfor future research.101b.*di-shang daozai-le.he floor-on fall-asp‘He fell onto the floor.’ (Object shift disallowed)c.*di.shangta daozai-le.floor-on he fall-asp‘He fell onto the floor.’ (Object fronting disallowed)Our analysis entails that object fronting and shift are similar syntactically in that both are overtraising of the object NP (either generic or definite). Therefore, it is only natural that these twoprocesses should have the same semantic requirements.Third, it accounts for the position of the sentential adverb with respect to the shiftedobject. Assuming that the sentential adverb is adjoined to VP or to any functional projectionsabove VP, it predicts that a sentential adverb like jintian ‘today’ can occur in any of thepositions marked by #:(68) # ta # zhejian shi1 # neng # ganwan t1.he this thing can accomplish‘He can finish doing this thing today.’Under our analysis, the position of the sentential adverb between the shifted NP and the modalverb is expected. This is because the adverb can be analyzed as adjoined to MP, which isdominated by AgrOP:102(69) CP/ \AgrSPI \ta, AgrOPhe1 I \zizejian shi MPthis matter I \jintian MPtoday / \t’’ N’l’/ \M ASPPneng / \can t’, VPI \t V’Aganwan taccomplishFinally, my analysis solves two related problems noted in F. Lu (1993). The first oneis that while the embedded object can be fronted to the S-initial position of either the embeddedclause or to the matrix clause, it can only be shifted to the position after the embedded subject,not after the matrix subject:(70) Wangwu shuo [Lisi duwan-le naben xiaoshuo].Wangwu say Lisi read-asp that novel‘Wangwu said that Lisi finished reading that novel.’(71) Wangwu shuo [naben xiaoshuo1Lisi duwan-le tj. (short-distance fronting)Wangwu say that novel Lisi read-asp‘Wangwu said that that novel, Lisi finished reading (it).’(72) naben xiaoshuo1,Wangwu shuo [Lisi duwan-le tJ. (long-distance fronting)that novel Wangwu say Lisi read-asp‘That novel Wangwu said that Lisi finished reading (it).’103(73) Wangwu shuo [Lisi naben xiaoshuo1duwan-le tj. (short-distance shift)Wangwu say Lisi that novel read-asp‘Wangwu said that Lisi, that novel, finished reading (it).’(74)*wangwunaben xiaoshuo, shuo [Lisi duwan-le tj. (long-distance shift)Wangwu that novel say Lisi read-asp‘Wangwu that novel said that Lisi finished reading (it).’Lu (1993:14) claims without argument that the object is shifted to obtain scope. Her solution forthe contrast between (73) and (74) is that since only the finite clause can bear scope, themovement to the immediate sentential position as in (73) will satisfy the scope requirement. ThePrinciple of Economy excludes further movement to the secondary preverbal position in thematrix clause as is shown in (74).Lu’s scope account can be challenged for the following reasons. First, she does notprovide evidence to substantiate her claim that an object NP is shifted to obtain scope. Second,as we have shown above, NP fronting and NP shift are subject to the same requirements andrepresent very similar processes. If an NP is shifted to obtain scope, it is only expected that afronted NP moves to obtain scope too. But if this is the case, Lu’s account then predicts thatthe matrix fronting/topicalization case of (72) should be ruled out by Economy on a par withmatrix shift case of (74): the NP can obtain scope in the S-initial position of the embeddedclause as in (71); therefore, there should be no need for it to be further fronted to the S-initialposition of the matrix clause. As (72) shows, however, fronting an embedded object to thematrix S-initial position is acceptable.The contrast between (72) and (74) follows from our account. We have assumed withMahajan (1991) that NP shift is an A-movement which is a local process while long-distancefronting is obligatorily an A’-movement. In (74), the object NP is shifted and it should be anA-movement, but it is a long-distance process, thus violating the locality restriction of A-104movement. This is in contrast to (73) which involves ‘local’ NP shift. (72) is grammaticalbecause (a) it involves fronting, and (b) long-distance fronting as an A’-movement is allowed.Hence the contrast between (72) and (74). Alternatively, the ungrammaticality of (74) can beaccounted for as follows. The embedded object is in the matrix [Spec AgrOP]. But this matrix[Spec AgrOP] should accommodate the object complement clause, not the embedded object.Due to this conflict, feature checking crashes. Notice that this is consistent with the localityconstraint for A-movement if this process is viewed as a kind of feature checking as envisagedin the minimalist program.’8Another, and related, problem noted by Lu is that in a pivot construction, an embeddedNP has to cross over the matrix verb to be shifted. A canonical pivot construction in Chineseis given in (75):(75) Wangwu BI Lisi du-guo naben xiaoshuo.Wangwu force Lisi read-asp that novel‘Wangwu FORCED Lisi to read that novel.’In contrast to English, although the embedded verb can take an aspect marker, this aspectmarker can only be construed with the matrix verb, not the embedded verb. (75) means that ‘theforcing by Wangwu took place in the past, and it could be very well true that Lisi did not readthat novel despite Wangwu’s forcing.’ (see Cheng 1989 and Li 1990 for more discussion). Lucorrectly observes that whether it is fronting or shift, the NP involved cannot land within thelower clause, which, she claims, is an infmitive clause:18Another way to rule out (74) is that the embedded object undergoes improper movement (Mike Rochemontp.c.): it is first adjoined to the embedded CP in order to get out of the clause, then it raises to the matrix [SpecAgrOP], an A-position.105Fronting cases:(76)*wangwuBI naben xiaoshuo1Lisi du-guo t.Wangwu force that novel Lisi read-aspLit.:*‘wajgwuforces, that novel1,Lisi to read ti’.(77) Naben xiaoshuo1Wangwu BI Lisi du-guo t1.that novel Wangwu force Lisi read-aspLit.: ‘That novel1,Wangwu forces Lisi to read ti.’Shift cases:(78)*wangwuBI Lisi naben xiaoshuo, du-guo t1.Wangwu force Lisi that novel read-aspLit.:*‘wangwuforces Lisi that novel, to read ti.’(79) Wangwu naben xiaoshuo, BI Lisi du-guo t1.Wangwu that novel force Lisi read-aspLit.:*‘Wangwu,that novel1,forces Lisi to read t1.’Lu’s account (1993:14) is that an NP cannot be shifted within an infinitive clause since infinitiveclauses are ‘tense dependent’ and do not bear scope. However, this explanation is not adequatefor the following reasons. First, it is not clear how she accounts for the fronting case in (76)which is also bad. If she assumes that a fronted NP moves in order to obtain scope as well, (76)can then be accounted for on a par with (78) since an infinitive clause is not a projection inwhich it can obtain scope. But under such an assumption, she loses her account for why (72) isgood as we discussed earlier: the fact the embedded object can ‘obtain scope’ at the S-initialposition of the embedded clause, as shown in (71), should preclude its further fronting to thematrix S-initial position by Economy. In short, Lu’s scope account cannot capture the facts in(71-79) consistently.The second problem with Lu’s proposal is that it is questionable whether an infinitiveclause cannot delimit a scope domain. Look at the following sentences:106(80) a. wo bu quan ta qu.I not persuade him go‘I did not persuade him to go.’b. wo quan ta bu qu.I persuade him not go‘I persuaded him not to go.’In (80a) the negation has scope over the entire clause, meaning ‘It is not true that I did anypersuasion.’ In contrast, the negation in (80b) only has scope over the embedded clause,meaning that ‘I persuaded him and the persuasion is that he should NOT go’. The interpretationof (80b) indicates that the infinitive clause can be used as a projection to delimit scope.Under my analysis of NP-shift, the fact that an NP cannot be shifted to the embeddedclause in a pivot construction has to do with the structural configuration of the latter. AdaptingCheng (cf. 1989:29), I assume that the infinitival at least in Chinese lacks AgrP and theconstruction in question is as shown below.19As in Cheng (1989:29), the verb bi ‘force’ andthe proposition it selects jointly select the NP Lisi. Later, the verb bi further moves up to thehigher (empty) verb position (cf. Larson 1988).‘9llamida Demirdache (p.c.) informs me that Portuguese has infinitives with agreement (cf. Raposo 1987).107(81)VP/ \I \VVPe /\NPV’Lisi I \BIVPforce / \PRO V’/ \VNPdu naben xiaoshuoread that novelThe proposed structure in (81) amounts to the claim that the whole sentence is monoclausal.Evidence for this claim comes from the following facts:2°(82) Aspect construal: The aspect marker attached to the lower verb is construed with thematrix verb (cf. Li 1990).a. wo bi-guo John du zheben shu.I force-asp John read this book‘I have forced John to read this book’b. wo bi John du-guo zheben shu.I force John read-asp this book‘I have forced John to read this book’(83) VP conjunction and the identical aspect constraint: Not only two lower VPs can beconjoined, they must also have the same aspect marker.a. wo bi ta kan-guo shu, xie-guo zi.I force him read-asp books write-asp words‘I have forced him to read books and to write characters.’b.*wobi ta kan-guo shu, xie-le zi.I force him read-asp books write-asp words201thank Rose-Marie Dechaine for discussing this part with me.108(84) Anaphor binding: The anaphor in the embedded object position can be bound by the matrixsubject.a. ?wo1 bi Mary xihuan woziji,.I force Mary like myselfLit.: ‘I force Mary to like myself.’b.*renwei [Mary xihuan woziji.JI think Mary like myselfLit.: ‘I think that Mary likes myself.’(85) Negative polarity: The embedded object can be modified by ‘any’ if the matrix verb isnegated.a. wo mci bi ta du renhe shu,I not force him read any books‘I did not force him to read any books.’b.?*Ibu zhidao [ta du-guo renhe shu].I not know he read-asp any books‘I did not know that he had read any books.’To return to our discussion of the contrast between (78) and (79), lack of AgrP in the infinitivalentails lack of the embedded landing site for the shifted NP. This accounts for theungrammaticality of (78), and the grammaticality of (79). Specifically, only the “matrix” clausehas [Spec AgrOP] which under my analysis accommodates the shifted object.2’How is the fronting case of (76) to be captured in my account? I have shown that afronted NP either raises to [Spec, AgrOP] as in A-movement or is further CP adjoined as in A’-movement. Since there is no AgrOP and CP in the embedded clause in the pivot construction,21A question arises as to how the “upper” object Lisi in (79) gets Case. Hamida Demirdache (p.c.) suggests thatit may get the inherent Case. Alternatively, one may posit three Agr projections as in the dative verb constructions.Another interesting point regarding the “upper” object and “lower” object in the pivot construction is thatthey are complementary in terms of undergoing syntactic processes. For instance, the “upper” object can only bepassivized, but not fronted, shifted, relativized or pseudoclefted. In contrast, the “lower” object cannot bepassivized, but can participate in all the other processes. The passive facts also obtain in English in the ‘persuade-type’ constructions.109embedded topicalization, along with embedded shift, is prohibited for lack of an embeddedlanding site.While the proposed analysis has quite a few advantages, there remains at least one issueopen to further research. In my analysis, even though Chinese has no overt agreement (cf.Chomsky 1993), overt raising for the subject in Chinese is obligatory, whether it is to [SpecAspP] or further to [Spec AgrSP]. There is a parallel between overt NP raising and proidentification. ‘Pro-drop’ has been found in languages which either have rich morphology likeItalian or have no morphology such as Chinese (Huang 1984, Speas 1994). Overt raising isproposed for languages with rich morphology (Pollock 1989, Chomsky 1992). However, asdiscussed here, it is also found available in languages with no morphology at all as in Chinese.Now the question is why pro-drop and overt raising involve the two extreme cases ofmorphology.One speculation for the obligatory raising of subject in Chinese is that since subject inChinese is definite, it has to get out of the VP domain at S-structure (cf. Diesing (1992). SinceChinese has no morphology, the subject does not have to land in [Spec AgrSP] at S-structure,though it has to at LF ultimately. It is therefore free to raise half way to the Spec of AspP orMP as in the SVO or OSV word orders.3.4. SummaryIn this chapter, I have presented and analysed the ‘object shift’ phenomenon in Chinese.It is similar to fronting in that a definite or generic NP is literally ‘dislocated’; but it differsfrom the latter in that the object NP lands in a position which follows the subject. Based on threediagnostic tests such as ‘No reconstruction’ and ‘Locality Restriction’, I conclude that the shifted110object NP lands in an A-position, just like the short-distance shift found in Hindi or Japanese.As for the exact landing site of the shifted NP, I follow the fully articulated clausal structureproposed in Ch.2, and propose the following analysis: both the subject NP and the object NPovertly raise to [Spec AgrSP] and [Spec AgrOP] respectively.111Chapter 4 WIT-Object NP Shift and DOU LicensingIn Ch.3, I show that a definite object NP can be shifted to a position after the subject.In this chapter, I will investigate cases where a wh-object is shifted to a position after thesubject, but before the adverb dou, and the interpretation resulting from the wh-object shift. Insection 4.1, I provide the data and a descriptive generalization. In section 4.2, I undertake adetailed analysis of this syntactic process, based on the tree-splitting model advanced in Diesing(1992). Finally in section 4.3, I extend the analysis proposed for wh-object shift to a broaderrange of data in Chinese, followed by a summary in section IntroductionIn a sentence containing a wh-phrase and the adverb dou ‘all’, the wh-phrase has aninterrogative reading when it is to the right of the adverb. However, when it is to the left ofdou, it obtains a universal reading. Here are some basic facts:Wh-nhrase as the direct object:A. Without dou:(1) ta chi-guo shennie?he eat-asp what?‘What has he eaten?’B. In interaction with dou:(2) Wh-phrase to the right of dou:John dou chi-guo shenme?John all eat-asp what‘What has John eaten?’(interrogative/exhaustive list reading)112In (1) and (2), the wh-phrase has the interpretation of a question word. For (2), the speakerimplies that he wants as an answer a complete list of things which ‘John’ has eaten. If the hearerreplies with an incomplete list of things, he then fails to answer this question. Hence, theexpected answer for (2) is ‘John has eaten apples, oranges and bananas’ or ‘John has only eatenapples.’ (cf. Kuno 1973 calls this kind of reading in Japanese “exhaustive list reading”).(3) Wh-phrase to the left of dou:ta shenme1dou chi-guo t..he what all eat-asp.‘He has eaten everything.’ (universal reading)NOT: ‘What has he eaten?’Here, the wh-object is between the subject and the verb, like the object NP shift discussed inCh.3. For some speakers, the wh-object can also be fronted S-initially:(4) ?Shemne1ta dou chi-guo t1.what he all eat-asp‘He has eaten everything.’Notice that (4) is not as good as (3). Both S. Lu (1981) and Liu et al (1983) only give data like(3) where the wh-phrase is immediately adjacent to dou.Dou, like other adverbs, cannot occur postverbally as is shown in examples (5-6):(5) chi dou shenme?he eat all what‘What does he eat?’(6)*Tachi shenme dou.he eat what all.‘He eats everything’113The data in (5-6) show that dou cannot occur immediately after the verb, nor sentence-finally.Wh-phrase as the indirect object:A. Without dou:(7) ta gei-le shei F?he give-asp who F‘Who did he give an F?’B. In interaction with dou:(8) Wh-phrase to the right of dou:ta dou gei-le shei F?he all give-asp who F‘To whom did he give an F?’ (exhaustive list reading)(9) Wh-phrase to the left of dou:ta shei1 dou gei-le t F.he who all give-asp F‘He gave everybody an F.’ (universal reading)Notice again that, like the direct object, the indirect wh-object in (9) is located in a position afterthe subject. Furthermore, it can also be sentence-initial:(10) ?nage xuesheng1 ta dou gei-le t1 F.which students he all give-asp F‘He gave every student an F.’Wh-ohrase as the subject:A. Without dou:(11) shei qu-guo?who go-asp‘Who has been there?’114B. In interaction with dou:(12) Wh-phrase to the right of dou:dou you shei qu-guo?all exist who go-asp‘Who has been there?’ (exhaustive list reading)Notice that when the subject NP in Chinese is indefinite, it is preceded by you ‘exist’ as in (12).(13) Wh-phrase to the left of dou:shei dou qu-guo.who all go-asp‘Everybody has been there.’ (universal reading)Generally speaking, the wh-phrase and dou should be within the same clause in order forthe former to obtain a list or universal reading.’ Otherwise, neither reading will obtain:A. No list reading:(14) ni(*dou)renwei[,,ta mai-le shenme1?you all think he buy-asp what‘What(*llst)do you think he has bought?’In the above example, although shenme occurs to the right of dou, it is impossible for shenmeto obtain the list reading. The reason is that these two elements are in different clauses. Lackof the universal reading in the following example follows for the same reason.‘Milce Rochemont (p.c.) points out that dou used in collocation with han ‘even’ can be related to an NP locatedin a preceding clause:(i) [han zhepian wenzhang gaidong-le najige zi][ta dou jide ti].even this article change-asp which words he even remember‘He even remembers which words are changed in this article.’ (S. Lu 1981:326)He also suggests that such a usage of dou should be considered as having the universal quantification force as well.115B. No universal reading:(15) a.*shejrenwei [ta dou hui lai].who think he all will come.‘Everybody thinks he will come.’b.?*nagexuesheng1wo renwei [ta dou gei-le t F].which student I think he all give-asp F‘I think he gave every student an F.’In (15b), even though the wh-phrase is the object of the embedded clause containing dou, it isfronted to the S-initial position of the matrix clause at S-structure. As it is separated from douwhich remains in the embedded clause, it fails to obtain the universal reading. What isinteresting about this example is that it shows that the locality constraint between dou and a whphrase cannot be satisfied by a trace left by the moved wh-NP.What emerges from the data above is the following descriptive generalization(preliminary):A wh-phrase must be to the right of a clausemate dou for the interrogative and listreading; it must be to thelfof the adverb for the universal reading;This generalization is stated in linear terms. However, the contrastive judgement in thetwo pairs of sentences in (16) below indicates that translating the generalization intoconfigurational terms is not only possible, but also preferable. In examples (16a-a’), neither whobject gets the exhaustive list reading even though it is to the right of the adverb. In (16a), theadverb dou is embedded in the relative clause in subject position. It is not only in a differentclause from the matrix object, but also fails to c-command the latter. The judgement for thissentence is that it is entirely gibberish. This judgement contrasts with that for (16a’) where douc-command obtains, but not the clause-mate requirement.116(16) a.**[flidou renshi de nage ren] aishang-le shei?you all know DE that person love-asp who*‘Who(list) did the person that you know fall in love with?a’.*nidou renwei ta aishang-le shei]?you all think he love-asp who*‘who(list) do you think that he fell in love with?’For the universal reading, c-command is also required apart from the locality constraint. In(16b), the wh-object and dou are in different clauses; moreover, the wh-phrase does not ccommand the adverb, and the sentence is entirely ungrammatical. This judgement contrasts withthat for (16b’) where the fronted wh-object c-commands the adverb which is embedded in thesubordinate clause.(16) b.**[[tamai-le shenme] zhege xiaoxiIwo dou zhidao.he buy-asp what this news I all knowIntended: ‘I know the news that he bought everything.’b’.*nagexuesheng wo renwei [ta dou gei-le t F1.which student I think he all give-asp F‘I think he gave every student an F.’Given the judgement contrast in the two pairs of sentences in (16), I propose that the lineardescriptive generalization should be reformulated in configurational terms as below:A wh-phrase must be c-conunanded by a clausemate dou at S-structure for theinterrogative and exhaustive list reading. For the universal reading, a wh-phrase mustcommand dou in the same clause at S-structure.2The central issue I address in this chapter is the following: what role does the syntacticrepresentation play in the derivation of the semantic representation of wh-NPs? Specifically, for2Given the data so far, the wh-phrase must c-command dou at S-structure for the universal reading. In section4.3, I will present and discuss some exceptional cases where c-command apparently does not obtain.117the interrogative sentences and quantificational sentences, what is the relation between their Sstructure representation and their LF representation? I assume with Cheng (1993) that whphrases in Chinese are indefinites. But I further assume as in Diesing (1992) that indefinites canbe either specific or non-specific. If they are specific, they are then presuppositional (to bedefined in section 4.2.2.) and, following Diesing, must escape from the VP domain at 5-structure. By this syntactic movement, they form a restrictive clause defining the range of thequantifier dou. This explains how a wh-NP to the left of the adverb obtains the universalreading, a reading which is presuppositional in nature. Wh-phrases can also be non-specific ornon-presuppositional. In this case, they remain within the VP domain at S-structure, obtainingthe interrogative reading as a result of existential closure and binding by a [+wh] trigger. Inaddition, I apply some diagnostic tests to the syntactic process of wh-object shift resulting in theuniversal reading. I argue that (a) the wh-object shift exhibits A-movement properties, as in thenon-wh-object NP shift discussed in Ch. 3; and (b) its landing site is also [Spec AgrOP]. At theend of this chapter, I investigate the interaction of a wh-phrase and dou in other structuralconfigurations.4.2. Analysis of the Wh-objectBefore presenting my analysis for the Chinese data, I will first review Diesing’s (1992)model, which I adopt below.34.2.1. Diesing’s Mapping Hypothesis (1992)Within the GB-framework, the LF level is the intermediary between the syntax and theI thank Haniida Demirdache and Henry Davis for the general line of research for this chapter.118logical representation. It is from this abstract level of syntactic representation that logicalrepresentations are derived. Diesing proposes that VP and the area above VP (at the IP level)are distinct domains for different kinds of quantification. Adopting a two-subject model of phrasestructures ([Spec IP] and [Spec VP]) (see Kuroda 1988, Koopman and Sportiche 1991 amongothers), she proposes that the two subject positions are distinguished in mapping from the 5-structure to LF representations. Her tree-splitting model is given in (17). Adopting terms fromHelm’s (1982), she calls the area above VP the ‘restrictive clause’ and the area below VP the‘nuclear scope’:(17) IP Above VPI \ “Restrictive clause”Spec I’I \I VPI Below VPSpec V’ “Nuclear scope”/ \V NPDiesing’s Mapping Hypothesis is as follows:Presuppositional NPs, including specific indefinites, form restrictive clause structures(cf. Berman’s 1991 Presupposition Accommodation) and therefore must undergo QuantiferRaising;Non-presuppositional phrases remain in the nuclear scope and are bound by existentialclosure, licensing an existential reading.For instance, the sentence “Every llama ate a banana” has the following Logical Form:(18)Every [x is llama] (Er) y is a banana /\ x ateyt tquantifier restrictive clause nuclear scope119In this example, the quantificational NP ‘every llama’ is presuppositional: it presupposes a setof llamas over which the Quantifier ‘every’ ranges. In Diesing’s tree-splitting model in (17), thisNP gets mapped from [Spec VP] into [Spec IP] to form the restrictive clause, specifying the setof things the quantifier ‘every’ quantifies over. On the other hand, ‘a banana’ is a non-specificindefinite: it remains within VP (the domain of Nuclear Scope), and receives the existentialreading via existential closure by a default E operator introduced at the VP level. Notice that thequantifier itself constitutes the third part of the logical representation, in addition to therestrictive clause and nuclear scope. Specifically, Diesing proposes that in the presence of QR,there must be a means of excluding the actual quantifier from both the restrictive clause and thenuclear scope. Following Heim’s term (1982:133), she assumes a rule “Quantifier Construal”which adjoins every quantifier to IP, leading to a tripartite tree. Below is the LF representationof example (18) using Diesing’s model:(19) IP/ \every IP/ \QC llama VPt / \QR e V’/ \ate a bananaIn English both QR and QC occur at LF. As a parametric variation, QR can also occurat S-structure. In German, for instance, whether an indefinite NP receives a presuppositional ora non-presuppositional reading can be ‘read’ off its S-structure position. The followingexamples are from Diesing (1992:78):120(20)a. . . .weil ja doch zwei Cellisten in diesem Hotel abgestiegen sind.since ‘indeed’ two cellists in this hotel have-taken-roomsb. . . . weil zwei Cellisten ja doch in diesem Hotel abgestiegen sind.since two cellists ‘indeed’ in this hotel have-taken-roomsShe assumes that when the subject appears to the right of the particle ‘indeed’, it is in [SpecVP]; and when it appears to the left of the particle it is in [Spec IP]. Thus the particles are usedto diagnose the position of the subject. Specifically, she proposes (1992:78) that varying theposition of the subject in this way produces an alternation in the reading of the subject NP zweiCellisten ‘two cellists’. In (20a), the subject NP is analyzed as in [Spec VP], and the cardinalreading is most salient. The sentence asserts the existence of two cellists who have taken roomsin the hotel. This existential or cardinal reading follows from her tree-splitting algorithm whichmaps the subject occupying the [Spec VP] position into the nuclear scope of the logicalrepresentation. In (20b), however, the subject NP is in [Spec IP], and the presuppositionalreading becomes available. In this case, the two cellists are two of some larger set of cellists.The indefinite subject in [Spec IP] is mapped into a restrictive clause by tree splitting. Thisrestrictive clause represents the existential presupposition of the subject NP, leading to theinterpretative contrast between (20a) and (20b). As will be shown in the next section, thepresuppositionality of indefinites in Chinese is positionally marked at S-structure, as in German.4.2.2. The S-structure Positions of the Wh-object and its InterpretationsBefore I propose my analysis, I will briefly review Cheng’s (1993) analysis of wh-phrasesand dou since, to my knowledge, it represents the latest analysis within the GB paradigm (seealso Huang 1982, Lee 1986). Cheng (1993:26) proposes that the universal reading of a wh121phrase obtains if the adverb rn-commands the wh-phrase. This is how she accounts for thefollowing sentence:(21) ta shemne, dou chi-guo t,.he what all eat-asp‘He has eaten everything.’One problem with this rn-command account is that in the following wh-in-situ cases which areabsent from her database,(22) ta dou chi-guo shenme?he all eat-asp what‘What has he eaten?’ (interrogative and list)dou rn-commands the wh-object; yet it does not trigger the universal quantifier reading in thelatter.4 To put it differently, if dou-m-cornmand is the real requirement, there is then nomotivation for the wh-object in (22) to be shifted as in (21) to obtain the universal reading.Given this objection, I seek an alternative account for the Chinese data.5Following Cheng (1991, 1992, 1993), I assume that a wh-phrase in Chinese is indefinite:it can have an interrogative, universal or indefinite reading depending on the environment (cf.4Data of this kind are also provided in the traditional studies of Chinese grammar, e.g. S. Lu (1981:153) andLiu et al (1983:137).5Another potential problem with Cheng’s rn-command account is that a wh-object can precede the subject as in(i):(i) ?shennie ta dou chi.what he all eat‘He eats everything.’If rn-command refers to the first maximal projection (cf. Chomsky 1986), clearly dou does not rn-command thefronted wh-object and the universal reading should not obtain, contrary to fact. But as Leslie Saxon (p.c.) pointsout, if rn-command is understood in terms of the entire category rather than just the lower segment, this criticismdoes not go through. It is not clear which notion of rn-command Cheng adopts (1993:8).122Heim on indefinites 1982:127). In contrast to Cheng, I further assume that an indefinite NP inChinese is either quantificational (i.e. presuppositional), or non-quantificational (i.e. nonpresuppositional), depending on its S-structure positions. Note that my use of the term‘presuppositional’ is syntactically defined and is thus more restricted than its common usage. Itis commonly assumed that wh-phrases as question words are also presuppositional. For instance,a non-D-linked wh-phrase ‘what’ or ‘who’ carries a presupposition associated with the wholesentence, whereas a D-linked wh-phrase like ‘which’ carries a presupposition associated with acommon noun. In this dissertation, NPs associated with the particular notion of presuppositionare operationally defined as those which cannot occur in the existential you context in Chinese,such as the underlined NPs in (23a-b). If an NP can occur after you, it is then not associatedwith such an ‘existential presupposition’ as in (23c-d):(23) a.*yjmeige ren dou zai kanshu.exist every person all presently study.‘Everybody is studying.’b.*yjshei/nage xuesheng DOU lai-guo.exist who which student all come-asp‘Everyone/every student has come.’c. you ren zai qiao men.exist person presently knock door‘Somebody is knocking on the door.’d. (you) shei/nage xuesheng lai-guo?exist who/which student come-asp‘Who/which student has come?’As will be discussed below, such existential presuppositions induced by quantified NPs must beincorporated into a restrictive clause, i.e. above VP (cf. Diesing 1992:62).In addition to these assumptions about Chinese wh-phrases and presupposition, I furtherassume that the adverb of quantification dou is adjoined to the bar-level of any of the functional123projections above VP (see section for evidence and discussion). For ease ofpresentation, I focus on wh-objects in this section. The data are repeated below:(24) a. ta dou chi-guo shenme? (interrogative and exhaustive list reading)he all eat-asp what‘What has he eaten?’b. ta shenme1dou chi-guo t1. (universal reading)he what all eat-asp‘He has eaten everything.’Adopting Diesing’s framework, I propose that the different interpretations of the wh-phraseresult from its different S-structure positions. In (24a), the wh-object shenme remains within VPat S-structure. It receives the default existential reading via existential closure. Furthermore, itobtains the interrogative reading because it is bound by default by the covert [+wh] morphemein COMP (cf. Cheng1991).6Crucially, the wh-object does not provide the restriction for theadverb of quantification dou because of its S-structure position (i.e. to the right of the adverb).Notice that there is a correlation between the interrogative reading and its S-structure position.As suggested by Partee (1991) and Diesing (1992), VP is the focus domain. As the wh-phrasein (24a) remains within the VP domain at S-structure, it automatically gets the focus reading thatis typical of interrogative wh-phrases.6Note that the post-verbal wh-object is interrogative because it follows from the tree-splitting hypothesis (whichgives it an existential interpretation), and from the existence of a (covert) Q-operator (which makes it aninterrogative). However, in an appropriate ‘affective’ context (Neg. etc.), the wh-phrase becomes ambiguous. Lookthe following sentence (Li 1992, James Huang p.c.):(i) ta mei chi shenme?he not eat whata. What did he not eat? (interrogative existential)b. He did not eat anything? (non-interrogative existential under negation)The conclusion is that binding by the covert Q-operator is by default, and that the presence of an affective contextwill induce the existential but non-interrogative reading.124(25) S-structure for (24a) (The clause structure is simplified)cP/ \QIP E represents the existential/ \ operator in VP.he I’I \dou VP/ \E VP/ \V’/ \V NPeat what (x)There is an added fact in example (24a): the wh-phrase also has the ‘exhaustive list’ reading.To my knowledge, there is no discussion on this topic in the GB literature on Chinese syntax.Its presence is clearly related to the presence of dou. I propose that this exhaustive list readingis derived from the fact that the wh-phrase is within the scope of dou at S-structure. A parallelcan be drawn between this list reading under the scope of dou and the licensing of polarity itemsunder the scope of negators.7In Chinese, the use of negative polarity items like renhe ‘any’ islicensed by a c-commanding negator as is shown in the following group of sentences:(26) a. wo mel kanjian renhe ren.I not meet any people‘I haven’t met anybody.’b. kanjian renhe ren.I meet any people‘*1have met anybody.’7Here it is important to differentiate the polarity reading of English ‘any’ and its free choice reading. The latterreading obtains in the presence of a modal (including an invisible modality operator as proposed in Cheng1991:134).Henry Davis and Leslie Saxon (p.c.) inform me that in English it is acceptable to say ‘Who all did yousee?’ with the wh-phrase having an exhaustive list reading.125c.*renheren meiyou lai.any people not come‘No one came.’In short, just as the use of ‘any’ is licensed by a c-commanding negator at S-structure, theexhaustive list reading of a wh-object should be attributed to the c-commanding trigger dou atS-structure as well.8Let us look at example (24) repeated as (27) below:(27) ta shenme1dou chi-guo t1. (universal reading)he what all eat-asp‘He has eaten everything.’Here, the indefinite wh-object provides the set over which the quantifier dou quantifies over. Toserve as the range of a quantifier, it has to be outside the domain of VP at S-structure to formthe restrictive clause. Being outside the VP domain triggers the indefinite wh-phrase for thepresuppositional reading. At LF, dou further moves out to become the leftmost immediateconstituent of S by the rule ‘Quantifier Construal’ (cf. Heim 1982). Since I am using CP, I8James Huang (p.c.) suggests an alternative analysis for the postverbal wh-phrases in the context of dou. Forexample, in a sentence like (i),(i) John dou mai-le shemne?John all buy-asp what‘What (interrogative and exhaustive list) has John bought?’dou quantifies over several events, i.e. the event of ‘buying’ is dissolved into several distinct buying events, andthe speaker wants to find out what John has bought during each distinct event of buying. The advantage of thisalternative analysis is that it successfully maintains the uniform leftward quantification of dou. Yet it remains to bedetermined how to represent quantification over these ‘subevents’ syntactically, and how to explain sentences likethe following which does not contain an action verb:(ii) m jia-li dou shenme ren? (Liu et al 1983:137)your family all has what people‘Who are the members of your family?’126assume that the dou is adjoined to CP.(28) (Irrelevant details about the structures are ignored here.)a. S-structure for (27) b. LF representation of (27):CP CPI\ IIP doug CP/ \ / \NP ... IPhe I ... /\shenme, I VP NPdou / \ he /shenme/\ I VPVt1 t I \eatI \Vt1eat(28a) is the S-structure representation for (27): the wh-object moves outside the VP domain andit forms a restrictive clause for the quantifier dou. At LF as is shown in (28b), dou furtheradjoins to CP for Quantifier Construal. Notice that the contrast between (23) and (27) resemblesthe German examples in (20) in that it is the S-structure position of the indefinite NP (withrespect to dou in Chinese) that determines whether it is interpreted as quantificational or nonquantificational. Also notice that as in English, the quantifier undergoes ‘Quantifier Construal’.However there is a morphological difference between English and Chinese. In the Englishsentence ‘Everyone has come’, the subject NP consists of both the quantifier ‘every’ and itsrange ‘one’: the whole NP ‘everyone’ first undergoes QR as a whole at LF; then the quantifieris separated from its range and undergoes Quantifier Construal. In Chinese, the quantifier douand its range shenme are separated morphologically to begin with. The indefinite NPdenoting127the range is shifted (QRs) on its own at S-structure and lands to the left of dou so as to escapefrom the Nuclear Scope and form the restrictive clause of the quantifier.Note that the tree-splitting model only gives the wh-phrase to the left of dou the‘presuppositional’ or ‘quantificational’ interpretation (as opposed to the ‘existential’ or ‘cardinal’reading), but does not make it non-interrogative or universal. In other words, why does a shiftedD-linked wh-object such as the one in (29) lose its interrogative reading even though its Dlinking carries with it a presuppositional reading (James Huang p.c.)?(29) ta naben shu kan-guo.he which book all read-asp‘He has read every book.’NOT: ‘Which books has he all read?’Note in Chinese, a wh-phrase with the interrogative reading cannot be shifted or fronted (cf.Lasnik and Stowell 1991).(30)*tanaben shu kan-guo?he which book read-asp‘Which book has he read?’In the literature (see Heim 1987, Diesing 1992 among others), the term ‘presuppositional orquantificational reading’ has been used to refer to either the non-interrogative universal reading(‘everything’) or the interrogative universal reading (‘which X all’). In this thesis, I adopt asyntactically defmed notion of presuppositional reading as discussed above, i.e. it refersto thenon-interrogative universal reading. Such a restrictive use of the term is consistent with what thedata in (29-30) reveal. Specifically, since the interrogative reading for the shifted D-linked whobject is independently ruled out by the syntactic position it occupies as shown in(30), what weget for the shifted wh-phrase in the context of dou in (29) is, by default, the non-interrogative128universal reading. What emerges from this empirical fact is that in the presence of dou,thedistinction between the D-linked wh-phrase (‘which X’) and non-D-linked wh-phrase (‘what’)is lost: if they are (descriptively) to the right of the adverb, theymaintain their interrogativereading with an added exhaustive list reading; if, however, they are shiftedto its left, they losetheir interrogative reading (independently) and obtain thepresuppositional, and morespecifically, universal reading by c-commanding the adverb.To summarize our analysis, if the wh-phrase remains downstairsat S-structure (i.e.within VP), it receives the non-presuppositional reading of interrogatives. If, however, the whphrase moves upstairs at S-structure, i.e. if it is above the VP domain, it defines the rangeofdou, triggering the presuppositional reading. This analysis is consistent with Huang’s IsomorphicPrinciple (1982) in the sense that S-structure configuration determines LF interpretation inChinese. In Section 4.3, I will show that the proposed analysis is applicable to morecomplicated cases as well.4.2.3. The Landing Site of the Shifted Wh-objectIt is proposed above that the presuppositional wh-object moves outside the VP domain.The next legitimate question is what is the nature of this movement and whatis its landing site.In this section, I first review Cheng’s analysis for the position of the wh-object. Then Iarguethat the wh-object shifting is an A-movement and that it is moved to[Spec AgrOP] position ona par with non-wh-object shift discussed in Ch.3.1294.2.3.1. Review of Cheng (1991):Cheng (1993:6, 29) observes that the wh-object is “preposed”, but she does not specifywhere it lands. In her earlier dissertation (1991), Cheng is more explicitabout the position, butshe does not assume a movement analysis. She gives the following example:(31) Qiaofeng[shenme,[,pro, dou chi pro11]].Qiaofeng what all eat‘Qiaofeng eats everything.’In (31), Cheng (1991:166, 167) proposes that for an object to be modified by dou, it has to bein the topic position, and if ‘what’ is a topic, then ‘Qiaofeng’ has to be a topic as well.Specifically, ‘Qiaofeng’ is in the position for aboutness topics, which she laterexplains are“topics without a gap” (1993:5). Structurally, she assumes that ‘Qiaofeng’ is adjoined to CP, coindexed with pro, in [Spec IP], and that ‘what’ is base-generated in [Spec CP], co-indexed withpro1 in the canonical object position.This analysis is questionable for the following reasons. First, Cheng does not justify whythe wh-object has to be in [Spec CP]. If nothing prevents it from being moved into the spec ofone of the functional projections, then there is no motivation for the subject ‘Qiaofeng’ to bebase-generated adjoined to CP. Second, Cheng posits two pros in bothsubject and objectpositions in (31). What this entails is that the two lexical NPs are base-generated S-initially andare not related to the subject and object positions by movement. This predicts that there shouldbe no subjacency effect if these two empty categories are embedded in an island. However, thisprediction is not fulfilled. For example, in (32), the empty categories are in the sententialcomplement of the object NP and the sentence is ungrammatical:130(32) *[ Qiaofeng[,shenme,[,wo zhidao pro, dou chi proj] zhejian shi]]]].Qiaofeng what I know all eat this fact‘I know the fact the Qiaofeng eats everything.’The ungrainmaticality (due to the subjacency effect) of this sentence indicates that the postulationof pro is untenable, and that the S-initial NPs and the empty categories in both the subject andthe object positions are related by movement, as indicated by the island effect.9Third, it is not clear how pros are identified in her account. There is a problem if sheassumes Huang’s (1989) generalized control theory. The subject pro does not have a controldomain in the sense of Huang (1989), and therefore should not be controlled. But the factis thatthe pro subject, if present, obligatorily refers to, or is “controlled” by, ‘Qiaofeng’, and thesentence unambiguously means ‘Qiaofeng is the person who eats everything’, not ‘As forQiaofeng, somebody else eats everything’. Incidentally, this may be a general problem for thegeneralized control theory in accounting for left dislocation in Chinese.In short, given these objections, I reject the non-movement hypothesis of Cheng (1991).Instead, I argue for a movement analysis and discuss its syntactic properties in the next section.9One may argue that (32) is ruled out independently since dou and the wh-phrase are in separate clauses. Recallthat example (15a) repeated as (i) below is ruled out precisely because of the violation of the locality constraint:(i)*shejrenwei [ta dou hui lai].who think he all will come.‘Everybody thinks he will come.’However, (32) is worse than (i). I propose that the deterioration in (32) is caused by the subjacency violation inaddition to the locality constraint illustrated in (i).Also notice that (32) is worse than ordinary subjacency violations presumably becausethe paths are crossedbetween NP1 and the gap coindexed with it on one hand, and NP and its co-indexedgap on the other.1314.2.3.2. Wh-NP Shift as A-movement and its Landing SiteIn this section, I will first discuss the position of dou with respect toother elements. Iwill then argue that the movement of the wh-object is an A-movementand that its landing siteis [Spec AgrOP]. The Position of DouAssuming that a clause is projected as AspP instead of IP and thatadverbs are licensedby a head feature, Cheng (1993:14) proposes that dou can be adjoinedto ASP’, ASP as well asV’ and V since these elements are extended projections of a verb. In contrast,I propose that theadjunction site of this adverb should be restricted to the bar-level of anyof the functionalprojections. One advantage of such a restriction is that it can be usedto demarcate the dividingline between the “restrictive clause” and the “nuclearscope”.[1] Sentential adverbs: I assume that sentential adverbs are adjoinedsentence-initially, to afunctional projection or to VP, serving to mark the VP boundary(cf. Ch.3). The adjunction siteof dou can be ascertained by its position with respectto these adverbs. The conclusion is thatits distribution is consistent with adjunction to the bar-level of any functionalprojection.Set One: dou quantifies over a subject: dou can follow or precede thesentential adverb, thoughthis results in different scope readings.(33) a. tamen zuotian dou shi-guo.they yesterday all try-asp‘They all tried yesterday.’132b. ?tamen dou zuotian shi-guo.they all yesterday try-asp‘All of them tried yesterday.’Supposing that the sentential adverb is adjoined to one of the functional categories in both cases,the fact that dou can follow or precede it is consistent with the proposal that dou is analyzed asadjoined to (the bar-level of) any functional projection.Set Two: dou quantifies over the shifted non-wh-object: dou must follow the shifted object,which is in [Spec AgrOP] (cf. Ch.3). Further, this shifted object can follow (as in (34a)) orprecede (as in (34c)) the sentential adverb.(34) a. ta xingkui zhexie shu1 don kan-guo t1.he fortunately these books all read-asp‘He has fortunately read all these books.’b.*xingkui don zhexie shu1 kan-guo t1.he fortunately all these book read-asp‘Fortunately, he has read all the books.’c. ? ta zhexie shu1 dou xingkui kan-guo t.he these book all fortunately read-asp‘Fortunately he has read all these books.’d. ta zhexie shut xingkui dou kan-guo t1.he these books fortunately all read-asp‘He has fortunately read all these books.’As a sentential adverb can be adjoined to any of the functional projections, the fact that theshifted object can precede or follow it, but invariably precedes dou is consistent with theproposal that this adverb of quantification should be adjoined to the bar level of any functionalprojection. If the shifted object is a wh-phrase, the same distributional facts obtain as shown in(34’) below:133(34’) a. ta xingkui shenme shu1 dou kan-guo t1.he fortunately what books all read-asp‘He has fortunately read every book.’b.*xingkui dou shenme shut kan-guo t1.he fortunately all what book read-asp‘Fortunately, he has read every book.’c. ?ta shenme shut dou xingkui kan-guo t1.he what book all fortunately read-asp‘Fortunately he has read every book.’d. ta shenme shu1 xingkui dou kan-guo t..he what books fortunately all read-asp‘He has fortunately read every book.’[2] Manner adverbs: As discussed in Ch.3, I assume that manner adverbs are adjoined to V orto V’ but must occur immediately adjacent to V. This predicts that dou cancmiyoccur beforea manner adverb, not after it. This prediction is borne out by the following data.Set One: dou quantifies over a subject: dou can only precede the manner adverb.(35) a. tamen dou manmande gan.they all slowly do‘They all did it slowly.’b.*tamenmanmande dou gan.they slowly all do‘They all did it slowly.’Set Two: dou quantifies over a shifted non-wh-object: dou which follows the shifted object mustprecede the manner adverb.(36) a. ta zhexie shu1 dou manmande kanwan-le t.he these books all slowly read-asp‘He finished reading all these books slowly.’134b. zhexie shu1 manmande dou kanwan-le t1.he these books slowly all read-asp‘He finished reading all these books slowly.’c.*tamanmande zhexie shu1 dou kanwan-le t.he slowly these book all read-asp‘He finished reading all these books slowly.’d.*dou zhexie shu manmande kanwan-le t1.he all these book slowly read-asp‘He finished reading all these books slowly.’‘dou’ must precede the V-adjoined manner adverb as in (36a) and (36b), but followthe shiftedobject as shown in (36a) and (36d). Its interaction with manner adverbs indicatesthat its positionis consistent with its adj unction to the bar-level of a functional projection.If dou can be Vadjoined as assumed in Cheng (1993), it should be able to be ordered freely withthe V-adjoinedmanner adverb, contrary to fact:(37) a.*tii1enmanmande dou gan.they slowly all do‘They all did it slowly.’b.*?tazhexie shu manmande dou kanwan-le.he these book slowly all read-asp‘He finished reading all these books slowly.’Cheng herself also notes (1993:3) that ‘for some unknown reason dou cannotappear aftermanner adverbs.’ Our assumption that dou is adjoined to the bar-levelof a functional projectionwhile a manner adverb is V’/V-adjoined providesan answer to this mystery.If the shifted object in (36) is replaced by a wh-phrase,we get the same distributionalpatterns with respect to dou:135(38) a. ta shenme shut dou manmande kanwan-le t.he what books all slowly read-asp‘He finished reading every book slowly.’b. shenme shu1 manmande dou kanwan-le t1.he what books slowly all read-asp‘He finished reading all these books slowly.’c.*tamanmande shenme shut dou kanwan-le t.he slowly what book all read-asp‘He finished reading every book slowly.’d. dou shenme shu manmande kanwan-le t.he all what book slowly read-asp‘He finished reading every book slowly.’[3] Modal verbs: In Ch. 3, I assume that a modal verb heads MP in Chinese. In the followingexample, I use the modal verb bixu ‘must’ for illustration.’0Set One: dou quantifies over a subject: dou can precede or follow the modal verb bixu.(39) a. tamen dou lai.they all must come‘They all must come.’b. tamenjjdoulai.they must all come‘They all must come.’Examples (39a-b) are consistent with the proposal that dou is adjoined to the bar level of a1°Note that not all the modals behave the same. Take gan ‘dare’ for instance, dou somehow cannot appear afterit, in contrast to (39b) in the text below:(i)?*tamengan dou lai.they dare all come‘They all dare to come.’Another modal verb which cannot precede dou is ken ‘will/be willing to’. Leslie Saxon (p.c.) suggests that thisdistributive difference may have to do with the fact that some modals are more verb-like than others. I have nothinginteresting to offer on this point.136functional projection.Set Two: dou quantifies over a shifted wh-object: The shifted NP must precede the modal verb,which, in turn, can precede (see 40c) or follow (see 40a) dou.(40) a. ta shenme shu1 douijduwan t.he what books all must finish-read‘He must finish reading every book.’b.*?tabjicjjshenme shu1 dou duwan t.he must what books all finish-readc. ta shemne shuijdou duwan t1.he what books must all finish-readIn (40a), the shifted object precedes the modal verb, and the sentence is grammatical (sinceAgrOP is higher than MP which accommodates the modal). (40b) is bad due to the wrong wordorder with respect to the modal and the fronted wh-NP. In (40c), the shifted object and dou canbe separated because the adverb has the option of adjoining to the bar-level of a functionalprojection lower than MP.To conclude, the interaction of dou with adverbs (in particular with manner adverbs) andmodals indicate that dou should be adjoined to the bar-level of any functional projection. Thereis no compelling evidence to show that it can be V-adjoined as claimed in Cheng (1993).4.232.2. Wh-object Shifting as an A-movementAs shown in Ch.3, non-wh object shift displays A-movement properties and its landingsite is [Spec AgrOP]. In this section, I propose that wh-object shift is also A-movement, basedon several diagnostic tests. Furthermore, I will show that it also lands in [Spec AgrOP], as innon-wh object shift.137Test A: No WCO effect: Wh-object shift does not induce WCO effect. Look at thefollowing sentence containing a universal quantifier:(41) ta biaoyang-le suoyoude xuesheng.he praise-asp all student‘He praised all the students.’When we add a preverbal PP containing a pronoun co-indexed with the postverbal universalquantifier, the sentence becomes ungrammatical:(42)*taxiang tamende1jiazhang biaoyang-guo suoyoude xuesheng.he to their parents praise-asp all student‘He praised all the students to their parents.’The ungrammaticality of (42) is due to its WCO configuration. At LF, the universal quantifierundergoes Quantifier Raising across the co-indexed pronoun, A’-binding both the pronoun andits variable. However, as discussed in Ch.3, when the universal quantifier is shifted at Sstructure, the co-indexing becomes acceptable:(43) ta suoyoude xuesheng1dou xiang tamende1jiazhang biaoyang-guo t.he all students all to their parents praise-asp‘He praised all students to their parents.’“Notice that (43) contains the adverb of universal quantification dou though it is absent in (42). The differencein grammaticality between these two sentences has nothing to do with its presence or absence. Dou is independentlyrequired when a universal quantified object is fronted even in sentences without the PP:(i) wo(*dou)du-guo suoyoude jufashu.I all read-asp all books-on-syntax‘I have read all the books on syntax.’(ii) wo suoyoude jufashu,*(dou)du-guo tj.I all books-on-syntax all read-asp‘I have read all the books on syntax.’138Consider the cases involving a wh-phrase. There is a contrast in grammaticality between (44a)where the wh-object is in-situ and (44b) where the wh-object is shifted:(44) a.*pjdou xiang tamende, jiazhang biaoyang-guo nage xuesheng1?you all to their parents praise-asp which student‘Which students have you praised to their parents?’b. wo nage xuesheng dou xiang tamende1jiazhang biaoyang-guo t1.I which student all to their parents praise-asp‘I have praised all the students to their parents.’In (44a), the postverbal wh-object cannot be co-indexed with the pronominal object of thepreverbal proposition due to WC012.In (44b), the wh-object is shifted leftward obtaining theuniversal reading. Although it moves across the co-indexed pronoun, it does not induce theWCO effect as in (43). This indicates that the shifting of the wh-object is an A-movement, notan A’-movement.Test B: No Reconstruction Effect: Look at the following pair of examples first. In (45),the wh-direct object is in situ; in (46) it is shifted and obtains the universal reading:(45) Wh-in-situ with the interrogative reading:ni dou gei-le [ta] [shenme shu1?you all give-asp him what book‘What books did you give him?’(46) Wh-direct object is shifted for the universal reading:wo [shenme shuI,don gei-le [taJt1.I what book all give-asp him‘I have given him every book.’There are two ways to account for the WCO in (44a). One is to assume with Aoun and Li (1993) that whphrases do not move at LF. Under this assumption, (44a) is ruled out because a [+Q]morpheme in COMP bindsboth the wh-phrase and a pronoun co-indexed with the wh-phrase. The other way to nile out (44a) is to adopt LFmovement as standardly assumed (Huang 1982). Under this assumption, (44a) is ruled out as the wh-phrase movesto [Spec CP] across a co-indexed pronoun. To be consistent with Diesing’s approach, I follow Aoun and Li inassuming that the wh-phrase with the interrogative reading remains in the Nuclear Scope, i.e. VP, at LF.139Suppose that we add to the direct object a possessor co-indexed with the indirect object. Whenboth the direct object and indirect object are in situ, the preceding indirect object pronoun cannotbe co-indexed with the possessor NP in direct object position. But when the direct object isshifted, coindexing becomes acceptable.(47) Canonical word order: no co-indexingdon gei-le [ta1} [John1de shenme shu]?you all give-asp him John DE what book*‘Whatbooks by John did you give him1?’(48) Direct object shifted:?wo [John1de shenme shu] dou yijing gei-le ta1 t.I John DE what book all already gave-asp him‘John1’s every book, I have already given him1.’Notice that (48) is derived by movement from (47); yet (47) is bad while (48) is acceptable.Based on this paradigm, I argue that since reconstruction applies to the site of a variable, andthere is no reconstruction effect in (48), 1 in (48) must be an NP trace, not a variable, and themovement is an A movement. Also notice that if there were reconstruction, the wh-phrase wouldget an interrogative reading. But since the wh-phrase does not have this reading, it further showsthat there is no reconstruction, and the movement in question is an A-movement.Test C: Locality Constraint: As is shown in Ch.3, non-wh object shift is bounded inChinese. Here, I will show that the shifting of a wh-phrase is also clause-bounded. For instance,while the embedded object can be shifted to a position after the subject of its own clause as in(49), it cannot be shifted to a position after the subject of the matrix clause, as is shown in (50):(49) wo renwei[ta naben shu1 dou kan-guo t,].I think he which book all read-asp‘I think that he has read every book.’140(50)Kwonaben shu1 renwei[4,ta dou kan-guo t1].I which book think he all read-aspLit.:*‘J,every book, think that he has read [it].’If wh-object shift is A’-movement, we should expect that long-distance shifting as in (50) shouldbe grammatical as in other long-distance A’-movement. Compare with the following sentencewhich is a fronting process across a CP boundary:(51) zheben shu1,wo renwei [ta yijing lcanwan-le ti].this book I think he already finish-read-asp‘This book, I think he has finished reading [it] already.’As argued in Ch. 2, long-distance fronting in Chinese is an A’- movement as long-distancescrambling in Hindi. The contrast between (50) and (51) indicates that while fronting can be A’-movement and therefore unbounded, object shift can only be A-movement and is thereforelocal. [Spec AgrOP]: the Landing Site of the Shifted wh-ObjectBased on the results of the diagnostic tests, I conclude that wh-object NP shift in Chineseis A-movement. The next question is what is its landing site? Adopting the Split Infi Hypothesis(cf. Pollock 1989, Mahajan 1990), Chomsky and Lasnik (1991) analyse [Spec AgrOP] as thelanding site of ‘scrambling’ for the SOy word order. In Ch. 3, I also assume a fully articulatedclausal structure for Chinese, and propose that [Spec AgrOP] is the landing site for the non-whNP shift. Given that wh-NP shift is an A-movement as in non-wh-NP shift, and that [SpecAgrOP] is an A-position, I propose that [Spec AgrOP] is also the landing site of wh-NP shift.‘3Alternatively, the clause-boundedness of wh-NP shift can be accounted for by the independent requirementthat dou quantification over a wh-phrase, as in other cases of quantification, is clause-bounded. No matter whichaccount we adopt, the clause-boundedness of wh-NP shift is consistent with its A-movement properties.141Notice that Diesing does not use a fully articulated clausal structure in her work. In her model,the area above VP (i.e. the traditional IP level) is the ‘restrictive clause’. To transfer thearticulated clause structure to Diesing’s tree splitting model, I assume that the ‘restrictive clause’is the area above VP, including the lowest functional projection. For instance, the sentence tashenme dou chi ‘He eats everything’ has the following S-structural representation:(52)AgrSPI \ta4 AgrOPI \shenme MP/dou MPI \t”, AspP/ \t,i VP/ \t V’I \VeatHere, both the subject NP and the object NP raise overtly to [Spec AgrSP] and [Spec AgrOP]respectively.The proposal that the wh-object lands in [Spec AgrOP] predicts that it must not onlyprecede the argument PP which is canonically contained within VP, but also precede the modalverb which heads the MP. As is shown below, this prediction is indeed borne out:(53) a. Ni dou NENG gei ta zuo shenme fan?you all can for him cook what meal‘What kind of meals can you cook for him?’142b. wo shenme fan dou NENG gei ta zuo t.I what meal all can for him cook‘I can cook all kinds of meals for him.’c.*1dou NENG shenme fan gei ta zuo t.you all can what meal for him cook‘I can cook all kinds of meals for him.’The ungrammaticality of (53c) indicates that a wh-object cannot be shifted only to a positionbefore the VP-internal PP but after the modal verb. It must precede the modal verb as well, andthis lends support to our analysis that it must land in [Spec AgrOP].4.3. More Case studiesIn this section, we will extend our analysis for the direct wh-object to examples wherethe wh-NP occurs in different environments with respect to dou. We first look at cases involvingwh-subjects and then go over other sets of data.4.3.1. Wh-phrase as the subject: The data are repeated below:A. Without dou:(54) shei qu-guo?who go-asp‘Who has been there?’B. In interaction with dou:(55) Wh-phrase to the right of dou:don you shei qu-guo?all exist who go-asp‘Who has been there?’ (Exhaustive list reading)Recall that when the subject in Chinese is indefinite, it is preceded by you ‘exist’ as in (55).143(56) Wh-phrase to the left of dou:shei dou qu-guo.who all go-asp‘Everybody has been there.’As in the case of wh-objects, the explanation for the two readings shown in(55) and (56)follows from Diesing’s theory. Adopting the VP-internal subject for Chinese, I propose that thepresence of you in (55) is to ensure that the wh-phrase should remain in VP, or specifically in[Spec VP] at S-structure, as it follows you. The word you literally means ‘exist’, and may beanalyzed as the overt ‘existential operator’. Given this structural representation, the wh-subjectis subject to existential closure and gets the existential reading.’4In example (56), the universalreading for the subject NP is expected since it is outside the VP domain (i.e. to the left of douwhich is adjoined to the bar-level of a functional projection). As it is in the domainof therestrictive clause, it defines the range for the quantifier and obtains the presuppositional/universalreading. At LF, dou adjoins to CP by ‘Quantifier Construal’. In the next two sections, we willexamine different structural configurations where the wh-phrase interacts with the adverb doufor the interrogative or universal reading.‘4Altematively, we may assume the wh-subject has moved from [Spec VP] to [Spec AgrSP]. You is insertedto induce the existential reading as the marked case. For the sake of consistency with Diesing (1992), however, Iprefer the solution proposed in the text, i.e. the wh-subject is in VP.Liii (1993) proposes that the object expletive ta in examples like (i) functions as an existential element:(i) wo hen xiang clii ía liangci Beijing kaoya.I very like eat TA twice Beijing roast-duck‘I would like very much to eat Beijing roast duck twice.’He further proposes that ta is located in [Spec AgrOP]. Under this proposal, it is not clear how theobject NP getsfeature checking since this spec position is already plugged up.1444.3.2. The Wh-NP downstairs and its Interrogative/List Reading:SET A: A wh-phrase contained in a Possessor/Possessedobject NP in situ: Like the simplewh-object NP, the possessor/possessed object NP which remains within VP issubject toexistential closure, and thus obtains the existential reading.(57) Possessor wh-NP:Ni dou jian-guo[,,shei de jiazhang]?you all meet-asp who DE parents‘Whose parents have you met?(58) Possessed wh-NP:Ni dou du-guo[tade shenme wenzhang]?you all read-asp his what papers‘What papers of his have you read?’As in the case of the simple direct object, the expected answer to the above twoexamples is that‘I have met John’s parents, Mary’s parents, etc.’ or ‘I have read his papers on syntax,onphonology, etc.’SET B: A wh-phrase contained in a preverbal PP: I assume that PPis canonically V’adjoined; thus it is within the VP-domain and the prepositional object receives theexistentialreading:(59) a. Ni dou[zai nage danwei]shixi-guo?you all at which organization practise-asp‘At which organizations did you have your practicum?’b. Ni dou[zai natian] shangban?you all at which-day go-to-work‘On which days did you go to work?’145SET C: The violation of locality constraint: Our account not only captures all the grammaticalsentences as shown so far, it also successfully excludes the interrogative (and list) reading of awh-phrase which fail to meet the S-structure requirement. For instance, in the followingsentence, dou which is in the matrix clause cannot quantify over a wh-phrase embedded in anobject complement clause:(60) dou renwei[ta chi-guo shenme]?you all think he eat-asp what‘What do you think he has eaten?’The reason for the ungrammaticality in (60) is that the range and the quantifier are in differentclauses, thus violating the locality constraint. Another example is that the wh-object NP is in thematrix clause while the adverb is in the adverbial clause:(61)*[[Ni dou jinlai de shihou],[,,ta zai gan shenme]]?you all enter DE time, he currently do what‘What was he doing when you entered?’This example is ruled out for the same reason as (60).SET D: The wh-phrase in a relative clause: Complication arises in cases involving a relativeclause. Look at the following sentence where interaction is possible between the wh-phrasecontained in a relative clause in object NP and dou in the matrix clause:(62) ni dou renshi gan shenme] de ren]?you all know do what DE manLit. ‘People that do what (list) do you know?’Although the wh-phrase and dou are in different clauses, the former can have the interrogative146and list reading. Contrast this case with the NP containing a complement clause. In the followingsentence, dou is in the matrix clause while the wh-phrase is in the embedded complement clause:(63)*ijdou xiangxin[,Bill da-guo shei] de zhengci]?you all believe Bill hit-asp who DE testimonyLit. ‘You believe the testimony that Bill hit who (list)?’As expected, the wh-phrase in the complement clause cannot get the list reading since dou andthe wh-NP are not in the same clause. This contrast between (62) and (63) indicates that therelative clause with a gap is not an island for construal, in contrast to an NP complement clause.Look at the following English NP:’5(64) the rumour that [three men were killed]In this example, the NP contains a complement clause and is unambiguous. It means that therumour is about the fact that there are three men who were killed. The NP ‘three men’ alwayshas the narrower scope than the head noun ‘rumour’. Compare with the following NP whichcontains a relative clause:(65) the rumour, [that three men heard t]In (65), the relative clause has a gap, and the head noun is construed as the object of the relativeclause. This phrase is ambiguous: it means either that the three men heard the same rumour orthat each of the three men heard a different rumour.’6The scope interactions between the head‘I thank Hamida Demirdache for drawing my attention to Haik’s work (1984) on this topic.The ambiguity of the example in the text becomes more salient in sentences like the following (Leslie Saxonp.c.):(i) The rumour1[that1three men heard tImade (each of) them write letters to the editor.147noun ‘rumour’ and the NP ‘three men’ support the claim that a relative clause is not an islandfor construal/quantification. To return to our discussion of (63), the fact that the relative clauseis not an island for scope interaction explains why dou can quantify over a wh-NP contained ina relative clause, but not in an NP complement clause.17To conclude, we have shown that dou cannot license a non-clausemate wh-NP for thelist/interrogative reading unless the latter is contained in a relative clause.4.3.3. The Wh-NP upstairs and the Universal Reading:In this section, we will investigate cases where the wh-phrase is configurationally higher,but is located in more complicated structural environments.SET A: A wh-phrase contained in a Possessor/Possessed NP, a relative clause and asentential subject: In the following examples, the containing NP is always to the left of dou.This indicates that the wh-phrase contained in it is outside the VP and defines the range for thequantifier.The ambiguity noted in (65) is attributed by Haik to the fact that the head noun ‘rumour’ c-commands the indefiniteNP which in turn c-commands the trace of the relative pronoun construed with the head noun. Look at the followingNP:(ii) the mmour [that1t hurt three men]Example (ii) is not ambiguous: the head noun unambiguously has the wide scope. The reason is that the head nounand the trace construed with it both c-command the indefinite NP.‘7One way to implement the idea that a relative clause is not an island for construal between dou and a wh-NPtherein contained is to adopt feature percolation as proposed by Nishigauchi (1990). Specifically, we may proposethat the dominating NP node acquires a [+ indefmitej feature which is percolated from within the relative clause.Nishigauchi does not discuss cases involving sentential complements to NPs.See also Cinque (1990:136, 195), Tsai (1993) and Cole et al (1993) for their use of feature percolation.148Let us first examine cases with a possessor/possessed wh-NP contained in an NP:(66) The containing NP is the subject:a. The Wh-NP is the possessor of the subject NP:[nage xuesheng de jiazhang] don zhichi zhege jueding.which student DE parents all support this decision‘Every student’s parents support this decision.’b. The Wh-NP is the possessed NP:[tade napian wenzhang] dou keyi fabiao.his which paper all can publish‘Every paper of his can be published.’(67) The containing NP is a fronted or shifted object:a.[tade shenme wenzhang], wo dou du-guo t.his what papers I all read-asp‘I have read every paper of his.’b. Wo[,shei de jiazhang]1dou jian-guo t1.I who DE parents all meet-asp‘I have met everyone’s parents.’In (67a-b), the object NP is dislocated to the left of dou, thus escaping the VP domain and losingthe existential reading as well. The same explanation applies to examples (68) below:(68) wo[ [,gan shenme] de renI,dou renshi t1.I do what DE people all know‘I know people who do all kinds of things.’Example (68) illustrates cases where the shifted object NP contains a relative clause. Here, theentire fronted NP is syntactically moved out of VP, and is thus construed as quantificational orpresuppositional. This again shows that relative clause is not an island for construal, presumablyvia feature percolation. Another possible account for (68) is to follow Fiengo et al (1988:93) insemantically defining the entire NP as quantificational. Semantically, the sentence means ‘I know149[people who repair cars, people who write computer programs and people who wash dishes,etc.]’. The shifting of the NP in (68) represents a kind of S-structure QR, just like the simplewh-object shift.Finally, let us look at a case of sentential subject containing a wh-phrase:(69) shei mai cal]] dou xing.who buy grocery all ok[For anyone to buy groceries] is ok.’Cases like (69) can be accounted for by a proposal made by Cheng and Huang (1994) for Douconditionals in Chinese. An example of Dou-conditionals is given in (69’):(69’) (buguan) shei1 lai, wo dou bu zaihu.no matter who come I all not care‘No matter who comes, I do not care.’They propose (1994:33) that the antecedent clause which can be headed by buguan is anelliptical phrase containing an embedded question. Given the standard semantics of questions,it is interpreted as having the force of existential quantification. Dou in the followingconsequent clause triggers universal quantification, and it quantifies over the set ofpropositions that jointly define the ‘truth conditions’ of questions. This analysis correctlycaptures the distributive meaning of this sentence: ‘if John or Mary or Bill comes, I don’t care.’Cheng and Huang’s analysis is readily extendable to the sentential subject cases like (69).Specifically, the sentential subject can be analysed as an embedded question and has the forceof existential quantification; dou in the main clause universally quantifies over the set ofpropositions that jointly define the ‘truth conditions’ of questions. Again such an analysis issupported by the semantics of the sentence: ‘it is ok for John or Mary or Bill to buy groceries’.150One may suspect that Dou-conditionals and sentential subject cases like (69), in reality, belongto one syntactic construction. If (69) is analyzed as a Dou-conditional, we then have to postulatea covert subject.’8I leave this issue open.SET B. A wh-phrase contained in a preverbal PP: In the following example, the wh-NPasthe object of a preposition has the universal reading:(70) ta[jshei] dou hen keqi.he to who all very polite‘He is polite to everybody.’This is in contrast to example (59) where the PP is to the right of the adverb:(59) Ni dou[,,,,zai nage danwei]shixi-guo?you all at which organization practise-asp‘At which organizations did you have your practicum?’Example (70) is noteworthy in that in order for the PP to escape the Nuclear Scope, it mustmove outside the VP domain, i.e. to the left of dou. In (59), however, the PP has to be withinVP (canonically V’-adjoined) so that it can get the existential/interrogative reading.One may question the analysis that PP moves (out of VP) in cases like (70). But thereis evidence to show that PP can move leftward. Consider the following sentences:‘8Below, I use example (77) to show that an adverb in the matrix clause cannot be related to a wh-phrasein anadverbial clause.(77)*[[,ta zai za shenme de shihou],[,wo dou jinlai-le1].he currently trash what DE time, I all enter-asp‘I entered when he was trashing everything.’If the sentential subject in (69) is analysed as an adverbial clause, one needs to capture the contrast between (69)and (77).151(71) ?[j shei] ta dou hen keqi.to who he all very polite‘He is polite to everybody.’Liu et al (1983:476) point out that most of the PPs in Chinese can appear either before or afterthe subject. Since PP in (71) is fronted overtly to the left of the subject, it is not unreasonableto assume that the PP in (70) also moves leftward, though NOT to the S-initial position, but toa position between the subject and the adverb dou.’9Recall that I proposed above that dou is adjoined to a functional projection. This proposalis further supported by the data involving PPs. Consider examples (72) which has a PP, but nodou. Notice that the PP can precede or follow the modal verb:(72) a. wo neng xiang ta]jiedao qian.I can from him borrow money‘I can borrow money from him.’thank Leslie Saxon for reminding me of the PP data.Two of the few non-frontable PPs are headed by ba and bei (passive ‘by’) (Liu et al 1983:476). Since thesetypes of PP can also appear to the left of dou, a question arises as to how to justify the movement analysis for themin the absence of their overt fronting. Take ba-NP for instance:(i) ta dou ba shenme chi-le?he all BA what eat-asp‘What did he eat?’ (Exhaustive list reading)(ii) ta ba shenme dou chi-le.he BA what all eat-asp‘He ate everything.’ (Universal reading)(iii)*bashenme ta dou chi-le.BA what he all eat-asp‘He ate everything.’I assume that the ba-NP in (ii) is also moved. Specifically, I propose that it is adjoined to a functional projection.This analysis can be justified in two ways. First, as shown in Section 3.1.2. of Ch. 3, the ba-NP can appear indifferent position with respect to other PPs (though with different scope readings). This indicates that the positionof the ba-NP is not fixed. Further, it parallels the behaviours of non-manner adverbs: it cannot be fronted, but itsposition is not fixed either. Second, as Henry Davis points out (p.c), ba-NP is usually the object of the verb, andthe whole phrase is therefore closely connected to the argument structure of the verb. As functional projections areextended projections of the verb, there is nothing incompatible for the ba-NP to adjoin to a functional projection.152b. wo[,xiang ta] neng jiedao qian.I from him can borrow money‘I can borrow money from him.’However, when the object of the preposition is a wh-phrase and is quantified over by the adverbfor the universal reading, the PP and dou MUST be to the left of the modal verb.(73) a. wo[xiang sheiIdon neng jiedao qian.I from who all can borrow money‘I can borrow money from anybody.’b.*yneng[xiang sheiIdon jiedao qian.I can from who all borrow money‘I can borrow money from anybody.’c.*wo[xiang shei] neng dou jiedao qian.I from who can all borrow money‘I can borrow money from anybody.’In contrast, in the case of interrogative/exhaustive list reading, dou is preferably to the left ofthe modal verb with the PP after it:(74) a. ni dou neng xiang shei]jiedao qian?you all can from who borrow money‘From whom can you borrow money?’b.?*nineng don[xiang shei]jiedao qian?you can all from who borrow money‘From whom can you borrow money?’c.*ijdon[xiang sheiIneng jiedao qian?you all from who can borrow money‘From whom can you borrow money?’The contrast between (73) and (74) provides strong support for the proposal that for dou toquantify over the object of a prepositional phrase for the universal reading, both the PP and theadverb must be adjoined to a functional projection higher than MP since they must be to the left153of the modal verb. In Diesing’s terms, this means that the wh-object of a preposition associatedwith the presuppositional reading must be in the restrictive clause, the domain above VP.Apart from ruling in grammatical sentences, our account also successfully excludes theuniversal reading of a wh-phrase in the following set of data:SET C: Violation of the locality constraint and the absence of the universal reading: Thereare three cases to consider:(a) A wh-phrase in the main clause with dou in an object complement clause:(75)*nagexuesheng bu renwei ta dou lai].which students not think he all come‘All the students do not think that he will come.’Example (75) is ruled out since the locality constraint governing the range and the quantifier isviolated. As shown in (76), the sentence is equally bad if an embedded wh-object is fronted tothe S-initial position:(76)*Nabudianying1,ta bu xiangxin[,wo dou kan-guo tj.which films he not believe I all see-asp‘He does not believe that I have seen every film.’(b) In the following example, the adverb dou is in the matrix clause while the wh-phraseis in an adverbial clause:(77)*[[ta zai za shenme de shihou], wo dou jinlai-le]].he currently trash what DE time, I all enter-asp‘I entered when he was trashing everything.’Again the locality constraint is violated in (77).(c) The adverb of quantification is in the matrix clause, but the wh-phrase as its range154is in the complement clause to an NP:(78) ta da-guo sheiIde zhengci] dou zai zhe’er.he hit-asp who DE testimony all at here‘The testimony that he hit everybody is here.Example (78) is ruled out by the locality constraint. This follows from our discussion above thatan NP complement clause is an island for quantification/construal (while a relative clause is not).To summarize, if a wh-phrase as an indefinite NP is outside the VP domain, it then candefine the range of the adverb of quantification dou within the same clause and obtain thepresuppositional/universal reading.To briefly evaluate our analysis, our proposal is based on the current theoreticalframework proposed by Diesing. Her model, which interfaces syntax and semantics, has beenproved successful in analyzing indefinites in English and other languages. Its application toChinese, which differs from English both typologically and genetically, further demonstrates thevalidity of her approach. In general, our analysis conforms to the convention that interpretationis an LF operation. In Chinese, the reading of a wh-phrase in the context of dou is marked byits S-structure position. Specifically, we claim that the obligatory shifting of indefinite NPs forthe universal reading is only an S-structure reflex of some LF requirement: thepresuppositional NP is shifted to escape the Nuclear Scope for the correct LF mapping.Furthermore, the S-structure requirement that dou can only quantify over a clausemate wh-NPagain follows from an LF requirement since the locality constraint is one of the characteristicsof LF QR. If the S-structure wh-shift is analyzed as a kind of S-structure QR, its clauseboundedness only reflects its inherent connection to its LF counterpart. Finally, S-structure QR(in the form of the wh-NP shift in Chinese) is not an isolated case. It has been proposed in the155literature that Hungarian (Kiss 1990) and German (Fanselow 1990) have S-structure QR. Alsorecall that in languages like Dutch and German, the S-structure position of an indefinite NPdetermines its LF interpretation (Diesing 1992). One significant generalization emerging fromour current study is that Chinese, together with Hungarian and German, represents one type oflanguages where the LF interpretations are syntactically signalled at S-structure.4.4. SummaryThis chapter investigates the position of a wh-phrase with respect to dou and the variousreadings associated with it. More specifically,(I) I provide a broader database. In particular, I discuss the existence of theinterrogative/exhaustive list reading associated with a wh-phrase to the right of dou at 5-structure, an empirical fact which has not been discussed in the GB literature.(II) Based on Diesing (1992), I propose that a wh-phrase as indefinites can be eitherpresuppositional or existential. If it is outside VP, it defines the range of the adverb ofquantification dou and obtains the presuppositional/universal reading. If, however, it is withinVP (i.e. the nuclear scope), it is subject to existential closure and gets theexistential/interrogative reading. The exhaustive list reading of the wh-phrase with theinterrogative reading is derived from the fact it is in the scope of dou.(III) In contrast to the latest work on dou quantification (Cheng 1993), I propose that douis adjoined to the bar level of any functional projection. Further, I argue that the wh-phrase withthe universal reading is not in some ‘aboutness’ topic position, but is in [Spec AgrOP] as a resultof A-movement.(IV) I examine different syntactic structures where a wh-phrase and dou occur, and the156resultant readings associated with the former.As an epilogue to this section, let us explore the ambiguous readings of the followingsentence:(82) Shei dou xiang chi shemne?who all want eat what(a) ‘What does everybody want to eat?’(b) ?‘Who wants to eat what (list)?’(c) NOT: ‘What (list) does everybody want to eat?This sentence has two wh-NPs. One is the subject NP shei and the other is the object NPshenme. The wh-subject shei ‘who’ is outside the VP domain as it is to the left of dou, andhence reading (a). In contrast, the object NP is in situ since it is to the right of dou and theverb, and it receives, among other things, the existential reading (reading b) as a result ofexistential closure. Reading (c) does not obtain. The reason is that dou can only quantify overone item at one time (Huang 1982), but it quantifies over both the subject and the object underthis reading.157Chapter 5 ConclusionIn this chapter, I first summarize what has been accomplished in the previous chapters.I then consider certain issues that are unresolved in the current work but represent interestingtopics for future research.5.1. Research FindingsThis dissertation studied NP leftward dislocation in Mandarin Chinese. I first of alldiscussed NP fronting cases which are commonly known as ‘topicalization’. As illustrated inexample (1) below, the object NP is fronted to the S-initial position:(1) zhege ren1,wo jian-guo t.this person I meet-asp‘This person, I have met.’Based on subjacency tests, I argued that sentences like (1) involve syntactic movement of theobject NP. In other words, the S-initial NP is not base-generated S-initially, with the emptycategory being a pro, as proposed in Xu and Langendoen (1985). After establishing themovement thesis, I applied several diagnostic tests to determine the nature of the syntacticprocess. My conclusion is that short-distance fronting can be either A or A’-movement whilelong-distance fronting is invariably A’-movement. I further proposed that A-type frontinginvolves raising to [Spec AgrOP], following current assumptions about clause structures, whileA’-type fronting involves adjunction to CP. I also discussed subjacency suspension in cases ofextraction out of the S-initial domain. In Ch.3, I investigated ‘object shift’ where the object isdislocated to a position between the subject and the verb. An example is provided below:158(2) wo zheben shu1 yijing kanwan-le t.I this book already read-asp‘I already finished reading this book.’I proposed that the movement exemplified in (2) is an A-movement and that its landing site is[Spec AgrOP] as a manifestation of overt raising for feature-checking. Based on data from otherlanguages, I further proposed that the correlation of the definiteness of the shifted NP andthepossibility of object shift justifies our postulation of [Spec AgrOP] as the landing site (cf. Runner1994), even in the absence of morphological agreement in Chinese.’In Ch.4, I explored the interactions between dou, an adverb of quantification, and whphrases. Descriptively, an in-situ wh-object to the right of the adverb keeps its interrogativereading; in contrast, a wh-object shifted to the left of the adverb is universally quantified.Consider the following pair of sentences:(3) a. John dou chi-guo shenme?John all eat-asp what‘What has John eaten?’ (exhaustive list reading)b. John shenme1dou chi-guo t1.John what all eat-asp‘John has eaten everything.’ (universal reading)Adopting Diesing (1992), I proposed that in the context of dou, the reading of a wh-phrase is‘Another possibility for deriving OSV order is that both the object and the subject leave their [Spec AgrP]positions and get CP-adjoined, resulting in O-S-ç-t0-V. However, this derivation is prohibited by relativizedminimality (which prevents t0 from being antecedent governed by ‘0’ due to the presence of ‘S’), or by relativizmgthe Economy principle (Henry Davis p.c.). Specifically, we may propose that each A’-movement must lead to aseparate interpretation, and multiple A’-movements which lead to the same interpretations are barred. Note that thiskind of A’ -multiple movements is different from further extraction out of a fronted object NP as discussed in section2.1.4 in Ch. 2. That is because in the latter case, the reading for the fronted object NP itself is different from thatfor the NP that is further extracted out of it. In other words, while the fronted object NP is a ‘topic’ in relation tothe rest of the sentence, the further extracted NP is a ‘topic’ in relation to this fronted NP.The consecutive A’-adjunction possibility applies to deriving SOV as well, i.e. both NP are CP-adjoinedleaving traces in the [Spec AgrPJ positions, resulting S-O-t,-t0-V. Again this possibility is ruled out by the Economyprinciple (which bars longer derivation), the “path containment condition” (which prohibits A’-paths from crossingeach other) or by the relativized minimality.159positionally determined at S-structure (cf. German and Dutch). Specifically, if the wh-objectremains in situ as in (3a), it obtains the existential reading due to existential closure applying tothe VP domain at LF. It further keeps its interrogative reading due to the default binding by[+wh] in COMP. If, however, the wh-phrase is shifted out of the VP domain as in (3b), itforms the restrictive clause and defines the range of quantifier dou, and thus obtains theuniversal/presuppositional reading.As shown in previous chapters, not any NPs can be dislocated: only definite and genericNPs and universal quantifiers are eligible for participating in this syntactic process. For instance,an indefinite object can neither be fronted nor shifted:(4) a.*yibenshu1 wo kan-guo t.one book I read-asp‘*Onebook, I have read.’b.*wo yiben shut kan-guo t1.I one book read-asp‘*1one book have read.’A closer look at the dislocated object NPs reveals that what group them together is that they areall presuppositional in the sense of Diesing (1992). Specifically, the existence of entities whichthey denote has already been presupposed in the previous discourse. Given this shared property,I propose that what ties the three chapters together is leftward dislocation of NPs associated withpresuppositional readings.In addition to addressing some empirical issues internal to Chinese syntax, the currentwork also attempts to contribute to a more general understanding of the theory of grammar. Forinstance, Diesing’s model is developed on the basis of Indo-European languages. To myknowledge, this model has not been applied to the analysis of S-structure dislocation of object160NPs in Chinese. Its successful application, as shown in previous chapters, not only captures theempirical facts in Chinese, but also demonstrates its explanatory value for a broader range ofdata. In particular, employing Diesing’s distinction between presuppositional and nonpresuppositional NPs and her tree splitting model, we can provide a plausible analysis for theobligatoriness of dislocation for a wh-phrase when used as a universal quantifier and for theprohibition of dislocation for indefinite NPs.Another way in which the current work contributes to our understanding of the theoryof grammar is that it provides empirical justification for the postulation of Agrin Chineseconsistent with current assumptions about clause structure. Accordingto the Split InfiHypothesis, a finite clause contains various Agreement Phrases. Though by null hypothesis wecan posit the existence of AgrP in Chinese even in the absence of morphological evidence, westill preferably need some evidence in support of this postulation. This missing evidence isprovided indirectly by the crosslinguistic correlation between dislocated NPs and their obligatorydefiniteness in morphologically rich languages. According to Runner (1994), there are fivemanifested correlations between specific NPs and Agr. One of them is that specific NPs canovertly surface in [Spec AgrP], as in Spanish, German and Catalan. Given this correlation, theobligatory definiteness of the shifted and fronted NP in Chinese should by deduction suggest thatthere are Agr projections in this language whose specs accommodate thedislocated NPs. Thisis on a par with the postulation of abstract Case in Chinese (Li 1990). Despite the absence ofmorphological case, abstract Case plays a regulatory role in the distribution of NPs in Chinese.Similarly, the morphologically invisible Agr exerts its impact in Chinese syntax by hosting thedislocated NP in its Spec position.1615.2. Residual Issues:There are certain issues that the current work is unable to fully resolve. In this section,I will briefly discuss them and offer some speculations to be pursued in future research.5.2.1. Optionality of Definite NP DislocationRecall that a definite NP can be dislocated, but it does not have to be. This contrasts withthe case of a wh-phrase: it must be dislocated to the left of dou for the universal reading. Takeobject NP shift for instance:(5) a. wo yijing kan-le zhexie shu.I already read-asp these books‘I have already read these books.’b. wo zhexie shu1 yijing kan-le t.I these books already read-asp‘I have already read these books.’(6) a. ni ijj kan-guo shenme shu?you all read-asp what books‘What books have you read?’b. ni shemne shu, kan-guo t1.you what books all read-asp‘You have read all kinds of books.’In (5), the definite object NP can either stay in-situ or be shifted. In (6), a universal readingobtains for the wh-phrase only when it is shifted to the left of dou. The question is why objectshift for the ordinary definite NP is optional while the shift for the wh-object for the universalreading is obligatory.I propose that this obligatory/optional contrast is attributed to the presence of dou.162Specifically, an NP which is to be quantified by dou is strongly presuppositional.2In orderto achieve strong presupposition, it must be shifted to the left of the adverb as the latter onlyquantifies to the left. In contrast, NPs which are weakly presuppositional can remain in-situ orbe shifted at S-structure. In (5a-b), since the defmite NP is not quantified by dou, it is wealdypresuppositional and can thus remain in-situ or shifted. Contrast (5a-b) with (7a-b) which containdou:(7) a.*rdou yijing kan-le zhexie shu.I all already read-asp these books‘I have already read all of these books.’b. wo zhexie shu1 dou yijing kan-le t1.I these books all already read-asp‘I have already read all of these books.’In (7a), the definite NP can not be quantified by the adverb since it is to its right. As the adverbhas nothing to quantify over, the sentence is ruled out. To remedy the situation, the defmite NPwhich is quantifiable must be shifted to the left of this adverb to be quantified over by it. Hencethe obligatory shift of the definite NP as is shown in (7b).A related question arises as to why dou must quantify to the left. A possible answer isthat given the assumptions adopted in Ch.4, the area to the left of dou (i.e. above VP) formsthe ‘restrictive clause’ which accommodates the range of a quantifier. This determines that thedefmite NP must move leftward in order to be in the restrictive clause andto function as therange of the quantifier. Hence the observation that this adverb must quantify to the left. Again21thank Mike Rochemont (p.c.) for pointing out the possibility of distinguishing NPs which are stronglypresuppositional and those that are weakly presuppositional. Note that this contrast parallels the case of determinerswhich are either strong or weak (Milsark 1974). Only the weak ones can appear in ‘there-constructions’.To further pursue the possible distinction between strong/weak presuppositionality, I need to findindependent evidence where some NPs are obligatorily shifted, but are not quantified over by dou. I alsoneed sometests to make a semantic distinction between strongly presuppositional NPs and wealdy presuppositionalNPs.163the directionality of dou quantification should be perceived as an S-structure reflex of theabstract LF operations (quantification, restrictive clause formation, etc.).35.2.2. Feature Checking and the Tree-splitting ModelIn her tree-splitting model, Diesing (1992) does not adopt the Split Infi Hypothesis. Shetreats VP as the dividing line between Restrictive Clause and Nuclear Scope. In Ch.4, I adoptboth Diesing’s model and AgrPs for clause structures. My working hypothesis is to simplyfollow Diesing in treating VP as the dividing line between those two domains. Although thisworking assumption is valid for the analysis for the data, it remains to be determined whetherthe dividing line should really be drawn at VP or somewhere above VP. For instance, could itbe that MP demarcates the two domains since the shifted wh-object must be to the left of themodal verb? In other words, could it be that the area above the MP constitutes the ‘restrictiveclause’ while the area below it represents the ‘nuclear scope’? More work is needed before anyconclusive decisions are made.Related to the demarcation issue is the implementation of feature checking. For Diesing,LF is the domain for interpretation while for Chomsky (1993) it is also the domain for featurechecking. If we assume the minimalist approach, the question is how to reconcile featurechecking with Diesing’s claim that the S-structure positions of, say, bare plural NPs determinehow they should be mapped to logical representations. In the German sentence in (8) below, theindefinite NP is to the right of the adverb:3The leftward quantification of dou is further motivated if (i) it further adjoins to the definite NP at LF as Chengproposes (1993), and (ii) adjunction in general is only leftward in Chinese.164(8) weil ja doch zwei Cellisten in diesem Hotel abgestiegen sind.since ‘indeed’ two cellists in this hotel have-taken-roomsAccording to Diesing, the position of the indefinite NP in (8) is within VP at S-structure, andthe tree-splitting algorithm will map the subject occupying the [Spec VP] position into thenuclear scope of the logical representation, giving rise to the existential, or cardinal, reading.4Under Chomsky’s minimalist program, object NP should raise to [Spec AgrOP] for featurechecking either at S-structure or at LF. Following this proposal, the NP in (8) has to raise at LFto check its features. But this entails that the object is above the VP domain at LF, and that itcannot be correctly mapped to the nuclear scope of the logical representation. This predicts thatthe indefinite object cannot have the cardinal reading, contrary to fact. One way to reconcileraising for feature checking and correct mapping is to assume that at LF’ (cf. Cheng 1991), theraised object moves back to its S-structure position before it is finally mapped to the nuclearscope for the existential reading.5.2.3. The Licensing Puzzle of Object ShiftObject shift has been found in languages where verb movement is attested, languagessuch as Mainland Scandinavian, Icelandic, German, Dutch, Hindi and Japanese (Deprez 1989,Mahajan 1990). According to Deprez (1989:227-236), S-structure verb movement to a functionalhead will nullify the barrierhood of VP, thus making it possible for the object to move out.Leaving the technical implementation aside, a question arises as to why Chinese licenses object4Contrast (8) with (1) where the indefinite NP is to the left of the adverb and has a presuppositional reading:(i) weil zwei Cellisten Ia doch in diesem Hotel abgestiegen sind.since two cellists ‘indeed’ in this hotel have-taken-rooms165shift, even if it does not exhibit verb movement.5The claim that Chinese has no verb movementcan be established by looking at the position of the verb with respect to adverbs (cf. Pollock1989:367). In Chinese, adverbs must be preverbal at S-structure as shown in (9a), assuming asbefore that manner adverbs are base-generated immediately to the left of the verb. S-structureverb movement (to Asp or Agr) predicts that the adverb should end up to the right of the verb.But as shown in (9b), this word order leads to ungrammaticality.(9) a. wo renzhende kan-le zheben shu.I conscientiously read-asp this book‘I read this book conscientiously.’b.*kan-le renzhende zheben shu.I read-asp conscientiously this book‘I read this book conscientiously.’The fact that (9b) is ungrammatical supports the claim that there is no S-structure movement ofthe verb (to I) in Chinese.Lack of correlation between object shift and verb movement in Chinese leads to twopossible conclusions: either the claim that object shift is dependent on verb movement is wrong,or object shift is not a unitary process crosslinguistically. This reminds one of the case of pro-drop, which is found in languages which are either morphologically rich or poor (cf. Italian,Spanish on one hand and Chinese on the other). As has been attempted for the pro-drop data,one might posit an object shift parameter. However, this solution is the least interesting one asit would only shift the burden of explanation for the data per se to the explanation for theparameter.5Huang (1992) proposes a kind of verb movement to account for the distribution of frequency/durationexpressions in Chinese. In his analysis, the verb only moves to a higher empty verb position; it is not the type ofV-to-I movement as discussed in the object-shift languages. When I use the term ‘verb movement’, I am referringto the V-to-I movement.166One promising line to pursue is to examine the nature of the relevant nodes whose specsaccommodate the shifted object. According to Chomsky (1993:15), ‘the N-feature of Tense isstrong in English. . .hence NP must raise to SPEC-[AGR, T] in English prior to SPELL-OUT orthe derivation will not converge.. . Note that the N-feature of AGR is weak in English, or wewould have overt object shift.’ Following Chomsky, we may suggest that the N-feature of Agrcan be strong in Chinese, and this makes object shift possible. But once again, this solution facesthe same problem of circularity as the parameter approach. Under this hypothesis, object shiftis allowed because the N-feature of Agr in Chinese can be strong; the evidence that N-Featureof Agr can be strong is that object shift obtains. In other words, there is no independentevidence to back the claim that the N-feature of Agr is strong, a claim which is intended tomotivate object shift.6Note that we cannot claim that the morphological features of Chinese nouns are strong,so that they must be overtly raised for feature checking before SPELL-OUT (like the overtraising of French verbs as proposed in Pollock 1989). This is because like Chinese verbs,Chinese nouns do not have overt morphological marking. Further, even if we were to claim thatthey were strong, we would have no explanation for why they can remain in-situ. Chomskyproposes (1993:15) that ‘a language might allow either weak or strong inflection, hence weakor strong N-features: Arabic is a suggestive case, with SVO versus VSO correlation with therichness of visible Verb-inflection.’ This may suggest an answer to the question of why definiteNPs can be shifted or remain in-situ. Specifically, one may claim that Chinesealso has weakand strong N-features to allow the alternation of being in-situ or shifted for the definite object.6No that we cannot say that it is definiteness of object NPs that licenses object shift. This is because definiteobject NPs can remain in-situ in Chinese. Moreover, although English has definite object NPs, it cannot have objectshift.167But one recalcitrant problem remains. In Arabic, the strength of N-feature correlates with therichness of verb-inflection. Since Chinese has no morphological inflections whatsoever, we donot have this overt correlation. To posit an abstract variation of strong and weak N-feature toaccount for the optionality of object shift amounts to nothing but a veiled stipulation. I leave thefinal solution for the licensing puzzle of object shift open for future research.5.2.4. The Interpretative Puzzle of Object ShiftThe process of fronting the object NP yielding the OSV order has been traditionallycalled ‘topicalization’, and the fronted object has been treated as the ‘topic’ of the sentence(Huang 1982, Li 1990). As object shift also obtains in Chinese, yielding SOV order, a questionarises as to what is the interpretative function of the shifted object. In particular, does it functionas the topic like its S-initial counterpart in OSV sentences?I argued in Chs. 2 and 3 that the dislocated object in both OSV and SOV orders landsin [Spec AgrOP] as overt raising for feature checking.7Suppose we assume that topicalizationis a kind of overt movement motivated for phi-feature agreement between the object and theAgrO head. Since both OSV and SOy involve raising to [Spec AgrOP], object shift should beconsidered as a kind of ‘topicalization’ as well, like the prototypical OSV constructions.The claim that SOy also represents ‘topicalization’ must address two questions: (a) doesthe object in SOV syntactically qualify as the topic? (b) why does Chinese license both SOyand OSV as two manifestations of one topicalization operation? Let us consider the firstquestion. There are at least two tests which we can use to determine the eligibility of the shiftedNP as a topic. One is that topic NPs must be definite or generic. Evidence abounds in showing7hi the A’-movement yielding OSV, the object first lands in [Spec AgrOP] before it is further A’-adjoined,leaving an intermediate trace in that spec position.168that only these two types of NPs can be shifted as in the fronting cases (cf. Ch.3):(10) a. wo pijiu he-guo. (Generic NP is shifted)I beer drink-asp‘I have had beer before.’b. wo zheben shu kan-guo. (Definite NP is shifted)I this book read-asp‘I have read this book.’b’ yiben shu kan-guo (Indefinite NP is shifted)I one book read-asp‘I have read a book.’Second, topic and focus are conflicting discourse functions (Bresnan and Mchombo 1987: 757-759). For instance, a questioned constituent or the clefted NP in cleft constructions universallybears the FOCUS function as shown in (1 la); in contrast, the relative pronoun and relativizedconstituent universally bears the TOPIC function as shown in (1 ib). When we relativize theclefted NP in a cleft construction, the sentence deteriorates in grammaticality as is shown in(1 ic):(11) a. It was John [that Marilyn suspected_____Ifocusb. The car which you don’t want is a BMW.topicc.*theperson who it was _[that Marilyn suspected].topic focusConsider the following wh-question/answer pair in Chinese (cf. Rochemont 1986):(12) Q: ni kan-guo naben shu?you read-asp which book‘Which book have you read?’169Al: wo kan-guo zheben shu. (canonical word order)I read-asp this book‘I have read this book.’A2:*thebenshu wo kan-guo. (NP fronting)this book I read-aspA3:*izheben shu kan-guo. (NP shift)I this book read-aspIn (12Q), the object NP is questioned. Of the three replies, only (12A1) where the object is in-situ is acceptable. (12A2) is bad because the fronted (“topicalized”) object is used as a reply tothe questioned element in (12Q). According to Rochemont (1986), the questioned constituentcorresponds to the focus of a relevant construction. But the fronted NP in (12A2) is a topic aswe argue above. Therefore (12A2) is ruled out as a reply due to the incompatibility betweentopic and focus. (12A3) is as bad in reply to (12Q) as (12A2). If the shifted object NP isanalyzed as a (secondary) topic as we do here, its ungrammaticality in this question/answercontext follows for the same reason as that in (12A2). Specifically, my proposal that the shiftedobject is a topic predicts that it cannot be a focus at the same time, a prediction which is borneout by the data in (12). In short, based on the definiteness and topic/focus tests, I conclude thatthe shifted object qualifies as a topic, which, according to Diesing (1992), is presuppositional.Let us now address the second question: why Chinese allows both NP fronting andshifting as topicalization processes. One possible answer is that Chinese is a topic-orientedlanguage (Huang 1984:549). The discourse structure of Chinese is such that speakers have thefreedom to highlight the topic of their utterances by placing it at the beginning of a sentence.In contrast, English, according to Huang (1984:549), is a sentence-oriented language, and this170may explain lack of the freedom to shift objects as in Chinese.8 In addition, in the previouschapters, we propose the following:(13) a. In QSV constructions, the subject is in [Spec AspP/MP] and the object is in [SpecAgrOP] (cf. Ch 2.);b. In SQV constructions, the subject is in [Spec AgrSP] and the object is in [SpecAgrOP] (cf. Ch 3.);Given this proposal, we may further speculate that NPs raised to functional projections (whether[Spec AgrP] or [Spec AspP]) are topic positions. Specifically, in both OSV and SOV, there aretwo topic NPs since both the subject and object are in the spec of a functional projectionrespectively.9”0Right now I see little distinction in terms of topic functions between thefronted object NP and the shifted object NP, and I leave this issue open for future research.8Henry Davis (p.c.) informs me that German and Dutch, which have object shift, are V2-languages, and thatthey are generally treated as ‘topic-prominent’ languages (cf. Koster 1978, Thiersch 1978).91n the canonical SVO word order, there is only one topic NP since only the subject, not the object, is locatedin the spec position of a functional projection.‘°The idea that there are two topics in a sentence is not a novel one. Aissen (1992) proposes that Mayanlanguages have two topic positions. Under her analysis, the external topic is base-generated in an adjoined positionto E(vent)P and the internal topic is in [Spec CP]. See also Kuno (1973), Kuroda (1988), Guilfoyle, Hung andTravis (1992) and Rice and Saxon (1994) on this subject.171REFERENCES:Abney, Steven Paul1987 The English Noun Phrase in Its Sentential Aspect. PhD thesis, MIT.Aissen, Judith1992 Topic and Focus in Mayan. Language 68:43-80.Aoun, Joseph, and Norbert Hornstein1990 Bound and Referential Pronouns. Pp. 1-24 in Logical Structure and LinguisticStructure: Cross-linguistic Perspective. C. -T. James Huang and Robert May eds.Kiuwer Academic Publishers.Aoun, Joseph, and Yen-hui Audrey Li1989 Scope and Constituency. Linguistic Inquiry 20:141-172.1991 Minimal Disjointness. Linguistics 28:189-203.1993 Syntax of Scope. Cambridge, Mass.: MIT Press.1994 Wh-elements in situ: syntax or LF? Linguistic Inquiry 25: 199-238.1994 On Some Differences Between Chinese and Japanese WH-elements. LinguisticInquiry 25:365-371.Balcer, Mark1988 Incorporation: A Theory of Grammatical Function Changing. University ofChicago Press.1991 On Some Subject/Object Non-Asymmetries in Mohawk. Natural Language andLinguistic Theory 9: 537-576.Barss, Andrew1986 Chains and Anaphoric Dependence: On Reconstruction and its Implications. PhDthesis, MIT.Belletti, Adriana1988 The Case of Unaccusatives. Linguistic Inquiry 19:1-34.Berman, Stephen Robert1991 On the Semantics and Logical Form of WH-Clauses. PhD thesis, UMass.Borer, Hagit1994 On the Projection of Arguments. Pp. 19-48 in Functional Projections (UMassOccasional Papers 17). Elena E. Benedicto and Jeffrey Runner eds. GLSA,UMass, Amherst.172Bresnan, Joan1990 Levels of Representations in Locative Inversion: a Comparison of English andChichewa. Ms.Bresnan, Joan, and Jonni M. Kanerva1989 Locative Inversion in Chichewa: A Case Study of Factorization in Grammar.Linguistic Inquiry 20:1-50.Bresnan, Joan, and Sam A. Mchombo1987 Topic, Pronoun, and Agreement in Chichewa. Language 63: 740-782.Bresnan, Joan, and Lioba Moshi1990 Object Asymmetries in Comparative Bantu Syntax. Linguistic Inquiry 21:147-185.Brody, Michael1985 On the Distribution of Empty Categories. Linguistics Inquiry 16:505-546.Browning, Marguerite1987 Null Operator Constructions, PhD thesis, MIT.Carsten, Vicki, and Kasangati Kinyalolo1990 On IP Structure: Tense, Aspect and Agreement. Ms.Chao, Yuan-ren1968 A Grammar of Spoken Chinese. Berkeley: University of California Press.Chen, Chung-yu1978 Aspectual Features of the Verb and the Relative Position of the Locatives.Journal of Chinese Linguistics 6:76-103.Chen, Matthew Y.1987 The Syntax of Xiamen Tone Sandhi. Pp. 109-156 in Phonology Year Book: 4.Cohn Ewen and John Anderson eds. Cambridge: Cambridge University Press.Cheng, Lisa L.-S.1988 Aspects of the BA-construction. Pp. 73-84 in Studies in Generative Approach toAspect. Carol Tenny, ed. Cambridge, Mass.: MIT Press.1988 Raising Structures and Passives in Mandarin Chinese. Ms.1989 Aspectual Licensing of pro in Mandarin Chinese. Ms.1991 On the Typology of wh-Questions. PhD thesis, MIT.1992 Deriving the Locality Restrictions in DOU in Mandarin Chinese. Paper givenatthe Canadian Linguistics Association meeting, Charlottetwon, PEI.1731993 On Dou-quantification. Ms.Cheng, Lisa.L.-S. and C.-T. James Huang1994 Two Types of Donkey Sentences. Ms.Cheng, Robert L.1984 Chinese Question Formation and their Meaning. Journal of Chinese Linguistics12: 86-147.Choe, Jae W.1987 LF Movement and Pied Piping. Linguistic Inquiry 18: 348-353.Choi, Jai-Hyoung1991 Scrambling as Non-operator A’-movement: Variables vs. Null Epithet. Ms.Chomsky, Noam1986a Knowledge of Language: Its Nature. Origin and Use New York: Praeger.1986b Barriers. Cambridge, Mass.: MIT Press.1991 Some Notes on Economy of Derivation and Representation. Pp.417-454 inPrinciples and Parameters in Comparative Grammar. Robert Freidin ed.Cambridge, Mass.: MIT Press.1993 A Minimalist Program for Linguistic Theory. Pp. 1-52 in The View from BuildingQ.Ken Hale and Samuel J. Keyser eds. Cambridge, Mass.: MIT Press.Chomsky, Noam, and Howard Lasnilc1991 Principles and Parameters. Ms.Chung, Sandra and James McCloskey1983 On the Interpretations of Certain Islands Facts in GPSG. Linguistics Inquiry14:704-713.Cinque, Guglielmo1990 Types of A’-dependencies. MIT Press.Clark, Robin1992 Scope Assignment and Modification. Linguistic Inquiry 23: 1-28.Cole, Peter, Gabriella Hermon and Li-May Sung1990 Principles and Parameters of Long-Distance Reflexives. Linguistic Inquiry 21:1-22.1993 Feature Percolation. Journal of East Asian Linguistics 2:91-118.Culicover, Peter, and Michael Rochemont1741992 Adjunct Extraction from NP and the ECP. Linguistic Inquiry 23:496-504.Demirdache, Hamida1991 Resumptive Chains in Restrictive Relatives, Appositives and DislocationStructures. PhD thesis, MIT.Deprez, Viviane M.1989 On the Typology of Syntactic Positions and the Nature of Chains: Move a to thespecifier of functional projections. PhD thesis, MIT.Diesing, Molly1992 Indefinites. Cambridge, Mass.: MIT Press.Epstein, Samuel D.1992 Derivational Constraints on A’-chain Formation. Linguistic Inquiry 23: 235-259.Fiengo, Robert, C.-T. James Huang, Howard Lasnik and Tanya Reinhart1988 The Syntax of WH-in-situ. Pp. 8 1-98. Proceedings of the Seventh WCCFL. ed.H. Borer, Stanford University, CSLI.Frampton, John1990 Parasitic Gaps and the Theory of Wh-chains. Linguistic Inquiry 21: 49-77.Franslow, Gisbert1990 Scrambling as NP-movement. Pp. 113-140 in Scrambling and Barrier. GuntherGrewendorf and Wolfgang Stemefeld, eds. Amsterdam/Philadephia: JohnBenjamins Publishing CompanyFreidin, Robert.1991 Principles and Parameters in Comparative Grammar. Cambridge, Mass.: MITPress.Fukui, Naoki1988 LF Extraction of Naze: some Theoretical Implications. Natural Language andLinguistic Theory 6:503-526.Goodall, Grant1990 X’-internal Word Order in Mandarin Chinese and Universal Grammar.Linguistics 28:241-261.Goodluck, Helen, and Michael Rochemont1992 Island Constraints: Theory. Acquisition and Processing. Kluwer AcademicPublishers.Greenberg, Gerald R.1984 Left Dislocation, Topicalization and Interjection. Natural Language andLinguistic Theory 2: 283-287.175Grewendorf, Gunther, and Wolfgang Sternefeld1990 Scrambling and Barriers. Amsterdam/Philadephia: John Benjamins PublishingCompany.Grimshaw, Jane1990 Argument Structure. Cambridge, Mass.: MIT Press.Guilfoyle, Eithne, Henrietta Hung and Lisa Travis1992 Spec of IP and Spec of VP: Two Subjects in Austronesian Languages. NaturalLanguage and Linguistic Theory 10: 375-414.Haik, Isabelle1984 Indirect Binding. Linguistics Inquiry 15:185-223.Han, Hak-sung1992 Notes on Reflexive Movement. Journal of East Asian Linguistics 1:215-218.Heim, Irene1982 The Semantics of Definite and Indefinite NPs. PhD thesis, UMass.1987 Where Does the Definiteness Restriction Apply? Pp.21-42 in The Representationsof (Indefiniteness. Eric. J. Reuland and Alice G.B. ter Meulen, eds. Cambridge,Mass.: MIT Press.Hoji, Hajime1985 Logical Form Constraints and Configurational Structures in Japanese. PhD thesis,Uw.1986 Weak Crossover and Japanese PS Structure. Pp. 163-202 in Issues in JapaneseLinguistics. Takashhi Imai and Mamoru Saito, eds. Foris Publications.Hu, Yu-shu1983 Modern Chinese. Shanghai: Shanghai Education Press.Huang, C.-T. James1982 Logical Relations in Chinese and the Theory of Grammar. PhD thesis, MIT.1983 On the Representation of Scope in Chinese. Journal of Chinese Linguistics 11:36-91.1984 On the Distribution and Reference of Empty Pronouns. Linguistic Inquiry15:531-574.1987 Existential Sentences in Chinese and (In)definiteness. Pp. 227-253 inThRepresentations of (In)definiteness. Eric. J. Reuland and Alice G.B. ter Meulen,eds. Cambridge, Mass.: MIT Press.1761989 Pro-Drop in Chinese: A Generalized Control Theory. Pp. 185-2 14 in The NullSubject Parameter. Osvaldo Jaeggli and Kenneth 3. Safir, eds. Dordrecht:Kiuwer.1991 Remarks on the Status of the Null Object. Pp. 56-76 in Principles and Parametersin Comparative Grammar. Robert Freidin ed. MIT Press.1992 Verb movement, (In)definiteness and the Thematic Hierarchy. Ms.1993 Reconstruction and the Structure of VP: Some Theoretical Consequences.Linguistic Inquiry 24, 103-138.Huang, C.-T. James, and Robert May1990 Logical Structure and Linguistic Structure: Cross-linguistic Perspective. KluwerAcademic Publishers.Huang, Lillian, and Philip Davis1989 An Aspectual System in Mandarin Chinese. Journal of Chinese Linguistics17:128-166.Iatridou, Sabine1990 About Agr(P). Linguistic Inquiry 21:551-576.Imai, Takashi and Mamoru Saito1986 Issues in Japanese Linguistics. Foris Publications.Ishihara, Roberta1984 Clausal Pied Piping: A Problem for GB. Natural Language and Linguistic Theory2: 397-418.Jiang, Zixin1990 A Constraint on Topic in Chinese. Journal of Chinese Linguistics 18:23 1-260.Johnson, Kyle1991 Object Positions. Natural Language and Linguistic Theory 9:577-636.Kayne. Richard S.1983 Connecteclness. Linguistic Inquiry 14: 223-249.Kim, Soo Woo1989 Wh-Phrases in Korean and Japanese as Quantifiers. Ms.Kitagawa, Yoshihisa1986 Subjects in Japanese and English. PhD thesis, UMass.Koopman, Hilda1984 The Syntax of Verbs: From V Movement in the Kru Languages to Universal177Grammar. Dordrecht: Foris.Koopman, Hilda, and Dominique Sportiche1991 The Position of Subjects. Pp. 211-258 in The Syntax of Verb-initial Languages(Special Issue of Lingua 85). James McCloskey ed. Amsterdam: Elsevier.Koster, Jan1978 Why Subject Sentences Don’t Exist. Pp.53-64 in Recent Transformational Studiesin European Languages. Jay Keyser ed. Cambridge, Mass.: MIT Press.1978 Locality Principles in Syntax. Cambridge, Mass.: MIT Press.Kuno, Susumu1973 The Structure of the Japanese Language Cambridge Mass.: MIT Press.Kuroda, S-Y1988 Whether We Agree or Not: A Comparative Syntax of English and Japanese.Lingvisticae Investigationes 12:1-47.1987 Movement of Noun Phrases in Japanese. Pp. 229-272 in Issues in JapaneseLinguistics. Imai and Mamoru Saito eds. Fons Publications.Larson, Richard K.1988 On the Double Object Construction. Linguistic Inquiry 19:335-39 1.Lasnik, Howard, and Mamoru Saito1984 On the Nature of Proper Government. Linguistic Inquiry 15: 235-290.1992 Move a: Conditions on its Applications and Output. Cambridge, Mass.: MITPress.Lasnik, Howard and Tim Stowell1991 Weakest Crossover. Linguistic Inquiry 22: 687-720.Lee, Thomas Hun-talc1986 Studies on Quantification in Chinese. PhD thesis, UCLA.Li, Charles, and Sandra Thompson1981 Mandarin Chinese: A Functional Reference Grammar. Berkeley: University ofCalifornia Press.Li, Lin-ding1989 Xiandai Hanyu Yufa de Tedian [Properties of Chinese Syntax]. Beijing: People’sEducation Press.Li, Yafei1993 What Makes Long Distance Reflexives Possible? Journal of East Asian178Linguistics. 2:135-166.Li, Y.-H. Audrey1990 Order and Constituency in Mandarin Chinese. Dordrecht: Kiuwer.1992 Indefinite Wh in Mandarin Chinese. Journal of East Asian Linguistics 1:125-155.Lin, Jo-wang1993 The Syntax of Zenmeyang ‘How’ and Weishenme ‘Why’ in Mandarin Chinese.Journal of East Asian Linguistics 1:293-33 1.Liu, Feng-hei1987 A Note on Word Order in Chinese. Chicago Linguistics Society 23.Liu, Yuehua, Wenyu Pan, and Hua Gu1983 Practical Chinese Grammar. Beijing: Foreign Language Teaching and ResearchPress.Lu, Felisa Hui-chuan1993 Preverbal Objects in Chinese. Ms.Lu, Shuxiang1981 The Usage of 800 Words in Modern Chinese. Beijing: Commercial Press.Lu, Zhong-ying1991 The Preverbal NPs in Chinese. MA thesis, Memorial University of NewFoundland, St. John’s.Mahajan, Anoop K.1990 The A/A-bar Distinction and Movement Theory. PhD thesis, MIT.1991 Operator Movement, Agreement and Referentiality. Pp. 77-96 in More Papers onWH Movement.1992 The Specificity Condition and the CED. Linguistic Inquiry 23: 510-5 16.Mallen, Enrique1992 Subject Topicalization and Inflection in Spanish. Theoretical Linguistics 18:179-208.May, Robert1985 Logical Form: Its Structure and Derivation. Cambridge, Mass.: MIT Press.Milsark, Gary Lee1974 Existential Sentences in English. PhD thesis, MIT.Mulder, Rene and Rint Sybesma1791992 Chinese is a VO Language. Natural Language and Linguistic Theory 10: 439-476.Murasugi, Kumiko1992 Crossing and Nested Paths: NP Movement in Accusative and Ergative Languages.PhD thesis. MIT.Nishigauchi, Taisuke1990 Quantification in the Theory of Grammar. Kiuwer Academic Publishers.Partee, Barbara1972 Opacity, Coreference and Pronouns. Pp. 415-441 in Semantics of NaturalLanguage. Donald Davidson and Gilbert Harman, eds. Dordrecht-Holland: D.Reidel Publishing Company.1991 Topic, Focus and Quantification. Proceedings of SALT 1.Pollock, Jean-Yves1989 Verb Movement, Universal Grammar, and the Structure of IP. Linguistic Inquiry20:365-424.Qu, Yanfeng1993 Locative Inversion in Mandarin Chinese. Canadian Journal of Linguistics38(3): 305-330.1994 The Dual Status of the Null Object in Chinese. Pp.149-164 in Kansas WorkingPapers in Linguistics Vol. 19, No. 1. Linda Roby ed. Linguistics GraduateStudent Association, University of Kansas.Rado, Janina1994 Object Positions, Interpretation and Agreement in Hungarian. Pp. 13 1-152 inFunctional Projections (UMass Occasional Papers 17). Elena E. Benedicto andJeffrey Runner, eds. GLSA, UMass, Amherst.Raposo, Eduardo1987 Case Theory and INFL to COMP: the Inflected Infinitive in European Portuguese.Linguistic Inquiry 18:85-109.Reinhart, Tanya1981 Definite NP Anaphora and C-Command Domains. Linguistic Inquiry 12:605-636.Rice, Keren, and Leslie Saxon1994 The Subject Positions in Athapaskan Languages. In The Morphology-SyntaxConnection MITWPL 22. Heidi Harley and Cohn Philips, eds. Cambridge:Department of Linguistics, MIT.Rizzi, Luigi1801990 Relativized Minimality. Cambridge, Mass.: MIT Press.Rochemont, Michael1986 FOCUS in Generative Grammar. Amsterdam: J. Benjamins.1989 Topic islands and the Subjacency Parameters. Canadian Journal of Linguistics34:145-170.Rochemont, Michael, and Peter Culicover1990 English Focus Constructions and the Theory of Grammar. Cambridge: CambridgeUniversity Press.Runner, Jeffrey1994 A Specific Role for Agr. Pp. 153-178 in Functional Projections (UMassOccasional Papers 17). Elena E. Benedicto and Jeffrey Runner, eds. GLSA,UMass, Amherst.Saito, Mamoru1985 Some Asymmetries in Japanese and Their Theoretical Implication. PhD thesis,MIT.1992 Long-distance Scrambling in Japanese. Journal of East Asian Linguistics 1: 69-118.Sells, Peter1984 Pied Piping and the Feature [+wh]. Ms.1985 Lectures on Contemporary Syntactic Theories. CSLI, Stanford.Shi, Dingxu1989 Topic Chains as a Syntactic Category in Chinese. Journal of Chinese Linguistics17:223-262.Speas, Margaret1990 Phrase Structure in Natural Language. Dordrecht: Kluwer.1994 Null Arguments in a Theory of Economy of Projections. Pp.179-208 inFunctional Projections. Elena E. Benedicto and Jeffrey Runner eds. GLSA,UMass, Amherst.Sportiche, Dominique1988 A Theory of Floating Quantifiers and its Corollaries for Constituent Structure.Linguistic Inquiry 19: 425-449.1990 Movement, Agreement and Case. Ms.Sternefeld, Wolfgang1811990 Scrambling and Minimality. Pp. 239-257 in Scrambling and Barrier. GuntherGrewendorf and Wolfgang Sternefeld, eds. Amsterdam/Philadephia: JohnBenjamins Publishing CompanyStowell, Tim1989 Subjects, Specifiers and X-bar Theory. Pp. 232-262 in Alternative Conceptionsof Phrase Structure. Mark Baltin and Anthony S. Kroch eds. Chicago, Illinois:University of Chicago Press.Tang, Jane1989 Chinese Reflexives. Natural Language and Linguistic Theory 7:93-121.1990 A Note on the DP Analysis of the Chinese Noun Phrases. Linguistics 28: 337-354.Teng, Shou-hsin1974 Double Nominative in Chinese. Language 50:455-473.Tenny, Carol1987 Grammaticalizing Aspect and Affectedness. PhD thesis, MIT.Thiersch, Craig Lee1978 Topics in German Syntax. PhD thesis, MIT.Travis, Lisa1984 Parameters and Effects of Word Order Variations. PhD thesis, MIT.1989 Parameters of Phrase Structures. Pp. 263-279 in Alternative Conceptions ofPhrase Structure. Mark R. Baltin and Anthony S. Kroch, eds. Chicago, illinois:University of Chicago Press.1991 Derived Objects, Inner Aspect, and the Structure of VP. paper given at NELS,University of Delaware.1992 Inner Tense with NPs: The Position of Number. Pp. 329-345 in Cahiers delinguistigue.Tsai, Ching-yuan1990 Question Words as Quantifiers in Chinese. Journal of Chinese Linguistics 12,125-158.Tsai, Wei-Tien D.1993 On Nominal Islands and LF Extraction in Chinese. Ms.Uechi, Aid1994 Focus and Quantification: A Case Study of Japanese Subjects. Ms.182Ura, Hiroyuki1992 Inverted Minimality and Topic-island Effects. Lingua 88: 21-65.Vinet, Marie-Therese1993 On Shi and You in Mandarin Chinese. Ms.Watanabe, Akira1992 Subjacency and S-structure Movement of Wh-in-situ. Journal of East AsianLinguistics 1:255-291.1992 Larsonian CP Recursion, Factive Complements and Selection. NELS 23proceedings.1993 Agr-Based Case Theory and Its interaction with the A-bar System. PhD thesis,MIT.Webeihuth, Gert1990 Diagnostics for Structures. Pp.41-76 in Scrambling and Barrier. GuntherGrewendorf and Wolfgang Sternefeld, eds. Amsterdam/Philadephia: JohnBenjamins Publishing Company1992 Principles and Parameters of Syntactic Saturation. OUP.Williams, Edwin1980 Predication. Linguistic Inquiry 11: 203-238.Xu, Liejiong1986 Free Empty Categories. Linguistic Inquiry 17:75-93.1990 Remarks on LF Movement in Chinese Questions. Linguistics 28:355-382.Xu, Liejiong, and Terrence Langendoen1985 Topic Structures in Chinese. Language 61:1-27.Xue, Ping1991 Syntactic Dependencies in Chinese and Their Theoretical Implications. PhDthesis, University of Victoria.Zanuttini, Raffella1990 Two Types of Negative Markers. Ms.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items