Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Grammaticalization and the development of functional categories in Chinese
(USC Thesis Other)
Grammaticalization and the development of functional categories in Chinese
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. ProQuest Information and Learning 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. G r a m m a t ic a l iz a t io n AND t h e D e v e l o p m e n t o f F u n c t io n a l C a t e g o r ie s in C h in e s e ©2000 by Xiu-Zhi Zoe Wu A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA in Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (LINGUISTICS) December 2000 Xiu-Zhi Zoe Wu Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3041544 Copyright 2000 by Wu, Xiu-Zhi Zoe All rights reserved. ___ ® UMI UMI Microform 3041544 Copyright 2002 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA T he G raduate School U n iversity Park LOS ANGH.ES, CALIFORNIA 90089-1695 T h is d isserta tio n , w ritte n b y YIU-2HI ^ 0 6 tJU_______ U n d er th e d ire ctio n o f h. D issertation C om m ittee, a n d a p p ro v ed b y a il its m em bers, h a s been p re se n te d to a n d a ccep ted b y The G radu ate School, in p a r tia l fu lfillm en t o f req u irem en ts fo r th e d eg ree o f D O C TO R O F PH ILOSOPH Y a v » £ Dean o f Graduate S tu dies D a te December 18, 2000 DISSER TA T IO N CO M M ITTEE Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. To Lord Jesus Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A c k n o w l e d g m e n t s Many people have played a crucial role in the conception, development and refinement of this dissertation. To my professors and friends in USC I am deeply indebted for what they have taught me and for what they have shared with me. There is much more that is harder to identify: my intellectual, emotional and social debts through the years in Los Angeles. Thanks are due to many. In the following, however, I can only try to express my gratitude to some of the many. First of all, to my advisor, Dr. Audrey Li, who is one of the brightest and most hard-working women I know. I am deeply indebted to her for countless discussions about Chinese and linguistics, for her continual encouragement through the years, and for all the hospitable invitations to her home every holiday season (for which I also thank Audrey's family: her husband and her son, who have always treated the "homeless" students as part o f the family). To Dr. Jean-Roger Vergnaud, who always places encouragement above criticism and who has always been there to help whenever problems (both in life and in linguistics) occur. To Dr. Joseph Aoun, who has taught me passion and hope in the research of linguistics. To Dr. Hajime Hoji, who has been dedicated in his efforts to train a "scientific mind" out of students and who has read my drafts so painstakingly that one of his comments was in fact "one dot is missing here." To Dr. Nam-kil Kim, who is the outside member in my committee and has been extremely encouraging about the final dissertation. To the other faculty in the department: Dr. Hagit Borer, Dr. Barry Shein and Dr. Maria- iii permission of the copyright owner. Further reproduction prohibited without permission. Luisa Zubizarreta. They all taught me a great deal and their research serves to remind me of what linguistics can be. To my friends and colleagues in the department of East Asian Languages and Cultures. During those years I was teaching Chinese there to support my study, they supported me morally (sometimes even financially) and shared with me their precious intuitions on Chinese sentences. Among them I especially thank Chunya Chen, Sanhui Chuang, Sumei Ho, Yi-chia Hsu, Fang Li, Grace Li and Linlin Wang. To my friends and fellow students in the department of linguistics. Lina Choueiri has been the best of friends, always by my side through thick and thin. Among other fellow students I would like to thank Miao-ling Hsieh, Uffe Larsen, Karine Megerdoomian and Marcello Modesto. I am also indebted to the departmental co-ordinators Laura Reiter and Linda Williams-Culver. They have been very supportive throughout the years. To my brothers and sisters in Lord, especially those who meet with me in fellowship in the Church of Los Angeles. To Grandpa and Grandma Chen, who looked after me just as if I was their real grandchild. To Linlin's mother, who often provided me with free meals and acupuncture treatment during the writing of this dissertation. To the other members o f the church, who fed me, talked to me and helped me every way they could. Without their prayers, this dissertation would never have come to be. To Dr. Andrew Simpson, my dear friend and colleague in SOAS, London. He read my drafts and commented on all the different versions. One of the chapters iv permission of the copyright owner. Further reproduction prohibited without permission. of the dissertation sprang from our co-authored work and a number of other ideas came about during our correspondence. Without his tremendous help and encouragement, I would not have completed this work. Being the only native speaker o f English among the linguists who read my drafts, Andrew is nevertheless responsible for any mistakes you can find in the dissertation Oust kidding, Andrew). Finally, to my family, for the love and support I have been taking for granted. To my parents and grandmother, who don’t understand what linguistics is but believe it is a great and worthy subject to study because of their beloved child/grandchild. To my two brothers and their beautiful wives, who support me morally and financially whenever I need help. To their kids, Yu-hsin, Jyun-wei and Yu-heng, who have been perfect children and who remind me of what kind of role model I should be. I thank all of you from the bottom of my heart. Without the love, security, optimism and support that you have constantly given me over the years, this dissertation would never have come into existence. permission of the copyright owner. Further reproduction prohibited without permission. T a b l e o f C o n t e n t s Acknowledgments....................................................................................... iii Abstract..........................................................................................................ix 1. Overview................................................................................................... 1 1.0 Introduction..............................................................................................................1 1.1 Chapter 2 and the Classifier Ge: Movement and Re-analysis.......................2 1.2 Chapter 3: Relative Clause De, Tone Sandhi and S-final Particles.............6 1.3 Chapter 4: De in the Shi-de Construction and 'Horizontal Grammaticalization' from D to T .................................................................... 11 1.4 Chapter 5: Resultative Constructions, Re-analysis and Directionality.... 16 1.5 Chapter 6: Verbal -Le, Aspect and Tense...................................................... 22 2. The Classifier G ei Movement and Re-analysis...............................28 2.0 Introduction...........................................................................................................28 2.1 Ge as a General Classifier for Nouns...............................................................28 2.2 Ge as a Classifier for Events Denoted by the Predicates..............................30 2.2.1 Occurrence Between a Verb and Its Cognate Object.................................. 30 2.2.2 Occurrence Between a Verb and an Indefinite Object................................32 2.2.3 Ge Introducing a Simple Resultative/Descriptive Adjective.......................36 2.2.4 Ge Introducing an Idiom or Fixed Expression.............................................41 2.2.5 Ge with Perfective but not Imperfective Aspect...........................................42 2.2.6 Apparent Constraints on Verb Selection.......................................................43 2.3 Syntactic Properties.............................................................................................45 2.3.1 Event Ge Introducing a Non-predicational Adjective................................. 45 2.3.2 Event Ge Licensing an Extra Argument........................................................52 2.3.3 Ge Licensing the Expletive 'It'....................................................................... 57 2.4 The Proposal - Ge as a Weak Unselective Determiner in D°.......................59 2.4.1 As a Nominalizer.............................................................................................60 2.4.2 In D° Position.................................................................................................. 64 2.4.3 Further Evidence.............................................................................................65 2.5 Grammaticalization, Re-analysis and G e........................................................74 2.6 Concluding Remarks.......................................................................................... 85 3. Relative Clause D e i Directionality, Clausal Raising and Sentence- final Particles....................................................................................... 87 3.0 Introduction...........................................................................................................87 3.1 Relative Clauses in Government and Binding Theory.................................88 vi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.1.1 Chiu (1993/1995) and SuoP: Ning (1993)....................................................91 3.1.2 DE as a Complementizer.................................................................................96 3.2 Directionality and C-selection— a Theoretical Problem for Standard Analyses.................................................................................................................. 98 3.3 Kayne (1994) and a Uniform Theory of Relativization.............................. 106 3.3.1 Determiners, Demonstratives and Definiteness Agreement......................I l l 3.3.2 Further Evidence fo r the NP-raising Analysis: (1) Language Acquisition: Chiu (1998); (2) Connectivity and Idiom-chunks.........................................119 3.3.3 Relativization in Japanese, Murasugi (1991/1998)...................................124 3.3.4 Process Nominals and the Structure ofNoun-Complement Clause CNPs ............................................................................................................................. 134 3.3.5 Other Noun-Complement Clause CNPs...................................................... 138 3.4 Taiwanese Tone Sandhi and the Clausal Raising Hypothesis...................145 3.4.1 Tone Sandhi Patterns in Taiwanese.............................................................145 3.4.2 Tone Sandhi in Taiwanese Relative Clauses.............................................. 151 3.5 Extensions: Grammaticalized Complementizers in Taiwanese............... 158 3.5.1 Tone Sandhi Patterns with Kons..................................................................164 3.5.2 Grammaticalization o f K om and Motivations fo r IP-movement............. 172 3.5.3 Evidence for P F Movement?........................................................................ 178 3.5.4 An Alternative— Cyclic Spell-Out..................................................................182 3.6 Concluding Remarks......................................................................................... 191 4. D e in Focus Sentences: from D° to T°............................................. 193 4.0 Introduction..........................................................................................................193 4.1 Object-.De Repositioning in the Shi-de Construction................................. 196 4.2 Post-object De...................................................................................................... 208 4.3 The Syntax of Re-analysis in Shi-de Sentences............................................225 4.4 Summary: D-to-T and Paths of Grammaticalization................................. 247 4.5 Appendix: on the Focus Interpretation of Shi-de Forms...........................253 5. Resultative Constructions: Directionality and Re-analysis 260 5.0 Introduction......................................................................................................... 260 5.1 Previous Analyses and Current Objections................................................. 264 5.1.1 Resultative Verb Constructions....................................................................264 5.1.2 The Lexical Approach—Yafei Li (1990).......................................................268 5.1.3 The Syntactic Approach— Zou (1994).......................................................... 272 5.1.4 Sybesma (1999), Hoekstra (1992)............................................................... 275 5.1.5 The Objections to the Syntactic Approach................................................. 281 5.2 An Alternative Analysis— V 2 as Aspect........................................................... 294 5.2.1 Telicity and Non-predication........................................................................294 5.2.2 Re-analysis and Aspect..................................................................................298 5.2.3 Literal Vi and Re-analysis............................................................................ 310 vii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2.4 Responses to Apparent Objections.............................................................. 317 5.3 V2 Re-positioning and Directionality.............................................................. 331 3.3.1 Arguments Against a Phonological Explanation........................................332 5.3.2 V 2 Elements as Verbal Suffixes......................................................................338 5.3.3 V 2 Elements Do Not Undergo Raising to Vi................................................ 341 5.4 Related Consequences to Other Structures................................................... 358 5.5 English Resultatives— a Speculation................................................................ 364 5.6 Summary................................................................................................................375 6. Verbal — L ei Aspect and Tense.......................................................... 377 6.0 Introduction.......................................................................................................... 377 6.1 The Re-positioning of Le/Liao...........................................................................378 6.1.1 The Current Status o f Verbal Le and Completive Aspect...........................381 6.1.2 Smith (1997): Two Different Types o f Aspect.............................................386 6.1.3 Verbal -Le and Perfectivity...........................................................................395 6.1.4 Grammaticalization and the Dual Status o f Verbal -Le.............................405 6.1.4.1 Parallel grammaticalization of independent X°-heads and affixes ... 406 6.1.4.2 Current distinction between completive and perfective -le\ evidence for diachronic development.................................................................... 419 6.2 Verbal -Le and Tense..........................................................................................432 6.2.1 Tense, Aspect and Perfectivity.......................................................................437 6.2.2 Possible Objections to an Analysis o f -Le as Tense................................... 446 6.2.3 Objections to the Objections.........................................................................449 6.2.3.1 Optionality o f- le ..................................................................................... 450 6.2.3.2 Lack of generality: with certain simple verbs...................................... 454 6.2.3.3 Lack of generality: with statives............................................................456 6.2.3.4 Subordinate clauses.................................................................................468 6.2.3.5 Imperatives............................................................................................... 478 6.2.3.6 Matrix jiu sentences................................................................................479 6.3 Summary................................................................................................................487 Bibliography............................................................................................... 490 viii Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A b s t r a c t Gramaticalization and the Development of Functional Categories in Chinese This dissertation examines how functional categories have developed in Chinese and argues for the existence of a variety of processes of grammaticalization and reanalysis on the basis o f changes found in Chinese. "Vertical grammaticalization" is suggested to be a common movement-dependent reanalysis process in which elements undergo raising into the extended functional structure dominating a lexical category and then become reanalysed in the higher target head position, this process occurring in the creation o f a new indefinite determiner from the general classifier ge (chapter 2) and in the creation of aspect and tense morphemes (chapter 6, verbal -le). "Horizontal grammaticalization" is argued to be a second type of cross-domain reanalysis where a functional head in one type of domain is re-interpreted as an equivalent corresponding functional head in a second parallel type of domain, this occurring in modem Mandarin with the ongoing creation of a new perfect tense/T° morpheme from an original D° element (chapter 4, D-to-T reanalysis and the shi-de cleft construction). The dissertation also considers how restructuring operations may result in the creation o f new functional categories. In chapter 3 it is argued that a new instantiation of evidential Mood/C° is created when an original bi-clausal structure is simplified as a mono-clausal structure and the predicate of the original matrix clause becomes re-analyzed as a high functional head in the new single clause ix Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. structure (chapter 3, Taiwanese hong as an S-final particle derived from kong ‘to say’). Restructuring is also suggested to be responsible for the creation of a new completive aspect projection in Chinese from an original resultative construction and significantly to be triggered by reasons of directionality and selection (chapter 5, resultative verb constructions and aspect). Generally the dissertation attempts to provide answers to a number of puzzles relating to the present identity and syntax of functional elements in Chinese and to argue for synchronic analyses of such elements in a way which also accounts for how such grammatical elements could have most plausibly come into existence. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1. Overview 1.0 Introduction The central focus o f this thesis is an investigation of the phenomenon of grammaticalization and its manifestation in the development and the synchronic status of a number o f functional categories in Chinese. A primary aim of the thesis is therefore to see how a study of various functional paradigms in Chinese may perhaps shed further light on the mechanisms o f language change and grammaticalization that are generally available cross-linguistically, and also to examine how the use of diachronic information may usefully allow for insightful analyses o f certain otherwise problematic synchronic phenomena in Chinese. The thesis is divided into six chapters, each o f which attempts to address the above-stated dual aims of: (a) providing a synchronic analysis of a certain functional paradigm in Chinese assisted by considerations of how functional categories may develop over time, and (b) investigating the nature of general processes of grammaticalization. Concerning the latter goal, each chapter argues for and attempts to illustrate a different process of grammaticalization at work in Chinese, and shows how grammaticalization may have important effects on aspects of surface linear word-order, frequently causing significant distortions of the underlying syntactic structure. General conclusions of the thesis are provided chapter by chapter and outlined in brief below where an overview of the various chapters is set out. The I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. basic background framework and the set of assumptions adopted for the thesis is simply a broad Minimalist/Chomskyean approach with a transformational component (i.e., movement) and two interface levels, PF and LF. As in Chomsky (1993/1995a/1995b), the combination of lexical items via Merge into initial syntactic structures is assumed to lead first to a point of Spell-Out feeding PF and phonetic interpretation, and then derivationally continue on to a level o f LF and semantic interpretation. Various other assumptions relating more specifically to language change and the phenomenon of grammaticalization will be introduced as the chapters proceed and as such notions become relevant. 1.1 Chapter 2 and the Classifier G e: Movement and Re-analysis The thesis begins in chapter 2 with an investigation of the general classifier ge in its full present-day distribution. Like other classifiers, ge frequently occurs with a numeral and an NP as in (1). However, ge is now also found in additional environments such as in (2) and (3) where other classifiers may not occur. The aim of the chapter is therefore to reach an understanding of the underlying synchronic syntax of this functional element and to see how it may have grammaticalized into a functional type different from ordinary classifiers: (1) ta chi-le liang ge pingguo. he eat-Perf two ge apple 'He ate three apples.' 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) wo yao chi ge bao I want eat ge full 'I want to do a satisfying eating' (3) ta yi-kouqi he-le ge san-ping jiu he one-breath drink-Perf ge three-CL wine ’ He drank three bottles of beer in one breath.' Considering a wide range of data, it is seen that ge appears to co-occur with a variety of complement types, and not just NPs/nominal complements as in (1) and (3), but also adjectival complements as seen in (2) and sentential complements in other cases. In the sense that ge therefore tolerates a range of comeplement types, it is taken to be (largely) 'unselective' unlike other classifers. Because the output of applying ge to adjectives, sentential complements and nominal complements is always a nominal category, ge is then suggested to be functioning as a nominalizer. Such a hypothesis is subsequently supported via a comparison of ge with nominalizers in Korean and English and the observation that certain restrictions on the use o f ge with adjectival and sentential complements seem to mirror common restrictions found with nominalizers in other languages. Ge analyzed as a nominalizer is then suggested to have an important aspectual function, often converting non-nominal categories into DP/NP objects which serve to provide either an end point for an activity predicate or an object which can be 'measured-out' in the course of the event. Ge-XPs thus introduce a natural aspectual bound to predicates and function to make such predicates telic. 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Concerning this aspectual function it is significantly found that ge-XPs can only occur with perfective and not imperfective aspect-markers. The chapter goes on to consider what syntactic category ge instantiates in its wider distribution. It is first pointed out that ge does not appear to be simply replacing other more regular classifiers in Cl° as seen in (4): (4) tiao ge/*tiao sheng jump ge CL-strip rope 'do a rope jumping/*jump one rope' Note that ge also occurs with nominals which importantly already have a numeral+classifier pair as in (3). It is argued that ge must therefore occupy a higher functional head position in the DP than Num° or Cl° and suggested that ge in fact now occurs base-generated in the D° position as a determiner. This is further supported by the observation that ge-sequences o f this type co-occur with the object expletive ta which Lin (1993) observes to only tolerate DP associates headed by weak determiners if without ge, as in (5): (5) a. kan ta san-ben xiaoshuo read it three-CL novel 'do a reading of three novels' b. chi ta ge tongkuai eat it ge satisfied 'have a satisfying eating' c. *chi ta tongkuai eat it satisfied 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ge is consequently concluded to be an unselective weak determiner in D° which results in non-specific indefinite interpretations o f the DP it heads. This further explains the observation that such ge-XPs may not occur in either subject or topic positions (which are restricted to specific/definite DPs). Diachronically, it is suggested that ge was originally base-generated in Cl° as other regular classifiers but later underwent raising to the higher D° position and then ultimately became re-analyzed as an element o f type D° directly base-generated in the D° head. (6) below notes the suggested stages of development and (7) the intermediate stage 2: (6) Stage 1: ge was base-generated and remained in Cl° as a classifier. Stage 2: ge was base-generated in Cl° and also raised to the D° position. Stage 3: ge was fully re-analyzed as a D° and consequently base generated directly in the D° position. (7) Stage 2 DP D' Do / / ^ ^ ^ N u m P urn C1P urn cr NP 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The development of ge is thus suggested to be an example o f grammaticalization which involves a simple three-stage sequence of movement and re-analysis in the functional super-structure dominating a lexical category (the noun). In such a process of 'vertical' grammaticalization, an element is originally base-generated in a lower position A and then raised higher to a second functional position B and eventually re-analyzed as being base-generated directly in B. During the intermediate stage 2 of the process, the element will instantiate both categories A and B, and in stage 3 only B, allowing for A to be occupied by a new overt instantiation of this category. This is the stage assumed to characterize ge in modem Mandarin examples such as (3) where ge base-generated in D° co-occurs with a distinct numeral and classifier respectively in Num° and Cl°. Chapter 2 thus introduces and illustrates a first type o f grammaticalization process which results in categorial re-analysis critically via movement upwards in a functional structure and subsequent re-analysis ('vertical grammaticalization'). 1.2 Chapter 3: Relative Clause D e, Tone Sandhi and S-final Particles Chapter 3 essentially has two major parts. The first focuses on the syntactic status of the functional morpheme de which occurs in relative clauses such as in (8): (8) [wo zuotian mai] de shu I yesterday buy DE book 'the book(s) I bought yesterday' 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Standard GB-style analyses such as Chiu (1993/1995) and Ning (1993) assign relative clauses structures such as (9) with de analyzed as a complementizer selecting a leftward IP complement: (9) NP CP NP S|3ec Oi IP C’ C shu 'book' wo zuotian mai tj 'I yesterday bought^' de Such analyses however go against the observation that the general direction of selection in Chinese is rightward as in other regular V-O languages. (9) also does not recognize in any way the fact that relative clauses in Chinese are typologicaly extremely marked and that Chinese is a rare case of V-O language having pre- nominal relative clauses. Following discussion in Simpson (1997), the chapter then argues for and defends a Vergnaud (1985)/Kayne (1994) analysis of relative clauses in Chinese and suggests that de is actually a bleached determiner in D° grammaticalized from the element zhi which occurred as an early demonstrative in examples such as (10): (10) zhi er chong you he zhi these two worm again what know 'And what do these two worms know?' (Zhuangzi 1.10) 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It is suggested that in relative clauses such as (8) the head noun shu 'book' first undergoes raising to SpecCP, and then that this is followed by raising of the IP remnant to SpecDP, D° being headed by de as in (11): ( 11) IPk D ’ wo zuotian mai tj D CP 'I yesterday bought tj' | Spec The chapter both draws on existing observations in Simpson (1997) and also provides a whole range of new arguments in favor of an analysis such as (11). The new arguments includes child language data as in (12), comparisons with prenominal relative clauses in Japanese, and tone sandhi phenomena in relative clauses in Taiwanese: (12) luotuo chi caomei de caomei camel eat strawberry DE strawberry 'the strawberry which the camel ate' (Chiu 1998) In Taiwanese it is shown that tone sandhi patterns interestingly mirror syntactic structure, clearly indicating Specifier, head and complement relations. The tone sandhi patterning found in relative clauses is then seen to provide strong support for 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the Kaynean structure in (11) and significantly cannot be captured by the more traditional analysis in (9). Tone sandhi also provides the link into the second basic part of chapter 3 which focuses on the grammaticalization of sentence-final particles in head-initial languages. If such particles are taken to instantiate functional heads such as Mood, Aspect and +Q/C due to their functional roles and interpretations, their occurrence in sentence-final position is problematic, as in head-initial languages they are instead expected to be located in clause-initial positions. In Chinese in particular, the occurrence o f sentence-final question particles such as ma has led to claims that CP is a head-final projection in Chinese and that Chinese therefore has to be accepted as being a mixed head-initial and head-final language. Taking such a view to be rather unsatisfactory, the second major part of chapter 3 examines the ongoing creation of a new particle in Taiwanese, the element kong which is grammaticalizing from the verb 'to say' as a complementizer-type particle in an unexpected sentence-final position as illustrated in example (13): (13) A*-hui m* lai kong* A-hui NEG come KONG 'A-hui is not coming.' In contrast with older S-final particles elsewhere where the origin and creation of such elements is often unknown, the source of Taiwanese kong is still very clear and the syntax underlying its development is also largely transparent due to patterns of tone change which occur when kong is used. The combination o f information 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. available with kong is shown to strongly suggest that its sentence-final position in fact results from an operation of IP clausal raising schematized in (14) and that the apparently head-final surface form found with this S-final particle actually hides a fully regular head-initial structure: (14) CP Spec C' kong It is argued that the S-final particle kong began to develop when an original bi- clausal structure collapsed into a mono-clausal form with the verb in the higher clause grammaticalizing as a functional head in the new mono-clausal structure. Increased dependency following grammaticalization combined with the stress-initial property of syntactic phrases in Chinese and also considerations of topic-comment structure are suggested to have subsequently led to an inversion of the original linear order and caused raising of kong's new IP complement to its Specifier position (Spec of CP). The active grammaticalization patterning found with kong provides new insights into the debate about whether Chinese should be considered to be exceptionally head-final in CP. It indicates that one should not automatically assume that the surface linear ordering of a phonologically dependent particle-head and its complement necessarily mirrors the underlying order of these constituents. It is also 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. suggested that the kong paradigm reveals a significant route o f grammaticalization by means of which head-initial languages may quite generally develop S-final particles. Finally, under the assumption that tone sandhi alternation is a phonological process, it is shown that the patterning with kong provides strong evidence either for the occurrence of large-scale clausal movement at PF, or the alternative idea of 'cyclic Spell-Out,' and that a derivational rather than a purely representational model of grammar is required to accommodate the patterns found. 1.3 Chapter 4: D e in the S h i-d e Construction and 'Horizontal Grammaticalization' from D to T Chapter 4 continues to investigate the categorial status of the element de as it occurs in clause-final position in cleft-like sentences such as (15), and attempts to account for the alternation where the object o f the verb optionally appears positioned after de as in (16): (15) ta shi zuotian mai piao de he BE yesterday buy ticket DE 'It was yesterday that he bought the ticket.' (16) ta shi zuotian mai de piao he BE yesterday buy DE ticket 'It was yesterday that he bought the ticket.' Most research on the shi-de construction has concentrated on offering accounts of how the focus interpretation o f such forms may be syntactically encoded li Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and this has led to interest being firmly centered on the element shi. Comparatively little attention has however been given to the role and status o f the element de and there is almost no discussion o f the alternation in (15) and (16) in the current literature. The chapter suggests that a study of the role played by de and the alternation in (15) and (16) leads to a better understanding of the shi-de construction and the interesting conclusion that de in such structures is currently undergoing a significant re-analysis and re-grammaticalization. The chapter begins by showing that there is considerable cross-linguistic and cross-dialectal evidence to support the suggestion that shi-de forms incorporate a clausal nominalization selected by shi. Assuming the analysis of relative clause de as a determiner outlined in chapter 3, it is suggested that the nominalization hypothesis would essentially result in (15) translating into a structure and derivation such as (17) with de in D°: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. pro zuotian mat piao 'pro yesterday buy A tickets’ I Following this however, it is pointed out that there is a variety of other evidence which might seem to argue against assuming a nominalization structure of the type illustrated in (17), notably problems relating to the interpretation of adverbs and the positioning of w/z-adjuncts and object DPs. The chapter then investigates what possible accounts might be given for the object alternation in (15) and (16) and argues that the only plausible explanation of the wider patterning observed is that de is actually undergoing movement to the verb as a new verbal clitic. Considering further aspects of the interpretation of shi-de forms in order to understand what kind of verbal clitic de might be coming to instantiate, it is noted that the strong 1 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. background presupposition present in shi-de forms frequently results in a conversational implicature that the event described in shi-de sequences occurred in the past. It is therefore suggested that as the past time conversational implicature has analysis is argued to explain the fact that where de hypothetically raises over the object to encliticize to the verb as in (16), such forms can only have a past time interpretation, and the past time implicature can no longer be over-ridden by future- oriented adverbials. It is therefore suggested that an original nominalization structure such as (17) is currently undergoing re-analysis as (18) with de formerly in D° being re-interpreted as a past tense morpheme in T° (and raising to cliticize to the verb from its underlying T° position): strengthened over time, this has now reached the point where the D° determiner de has now become re-analyzed as an instantiation o f (past) Tense/T°. Such a re- 08) ta Top 'he' ShiP Shi' [pro zuotian mai piaoji 'pro yesterday bnv tickets' ti 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The re-analysis with de in the shi-de construction is suggested to be an example of 'horizontal grammaticalization,’ where a functional head in one domain (here the nominal domain) is simply re-analyzed as a corresponding functional head in another domain (here the clausal domain). In contrast to the 'vertical grammaticalization' process illustrated in chapter 2 where an element undergoes movement within the functional structure of a single domain and is eventually re analyzed as instantiating the higher functional head, in horizontal grammaticalization there is no movement o f the element which undergoes re-analysis, and the relevant element is simply re-analyzed as instantiating a related functional head in a different but parallel domain. With de, it is suggested that although new past tense elements may otherwise commonly arise from the grammaticalization o f perfective aspectual verbs, the re-analysis o f a D° element as a T° morpheme is in fact a rather natural developmental shift. Noting that D° and T° are both functional heads which serve to provide reference to their respective (NP and VP) complements, in the D-to-T re analysis it is argued that the definite reference-fixing property of the D°-element simply becomes re-interpreted in the locus of temporal reference, the T° position. Finally, the chapter re-considers non-past interpretations in shi-de sentences and argues that speakers actually maintain a dual analysis of de as both a T° and a D° element in different structures such as (17) and (18), this having direct effects on a number of syntactic phenomena. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.4 Chapter 5: Resultative Constructions, Re-analysis and Directionality Chapter 4 discussed the ongoing re-positioning o f the element de from sentence-final position to a position right-adjacent to the verb and preceding the verb's object. Such a change in basic linear ordering schematized in (19) has historically occurred elsewhere in Chinese in resultative verb constructions and with the development of the aspectual morpheme -le from the verb liao. (19) V Object de — > V-de Object Resultative verb constructions (RVCs) such as (20) and (21) synchronically consist in two adjacent verbal elements V t and V2, where the V2 encodes the result of an action represented by Vj and the object of the action follows the V 1-V2 pair: (20) ta xi-ganjing yifu le he wash clean clothes LE 'He washed the clothes clean.' (21) wo kan-wan shu le I look-finish book LE 'I have finished reading the book.' Historically such sequences derived from strings in which the object in fact occurred between Vj and V2. Consequently there has occurred over time a change which seems to resemble that in (19) with the V2 in RVCs apparently repositioning itself over the object and right-adjacent to Vi as in (22): 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (22) Vi Object V2 -> V i - V 2 Object A similar change also occurred historically when the verb liao re-positioned itself right-adjacent to the verb as an aspect-marker and reduced into -le, this being schematized in (23): (23) V Object liao — > V-liao/le Object Chapters 5 and 6 focus on the developments in these constructions and attempt to answer the important questions in (24): (24) a. What was the reason for the re-positioning of the sentence- final element to a position right-adjacent to the main lexical verb? b. What is the current syntactic status of such constructions? Chapter 5 first considers the case of RVCs and endeavors to see how a better understanding of the ongoing change in (19) established in chapter 4 now offers useful insights into the earlier change which took place in the similar patterns (22) (and (23) in chapter 6). Broadly the chapter argues that the second verb V2 in RVCs synchronically has the status of an aspectual suffix combined with the main verb V i in the lexicon and licensed/checked by an Aspect projection dominating the VP, against recent syntactic accounts of resultative V-V sequences in Zou (1993) and Sybesma (1999). Historically it is suggested that the movement/re-positioning of V2 to its present position right-adjacent to V i occurred as part o f a significant re-analysis 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. process in which the V2 underwent grammaticalization as a functional element and became re-analyzed as instantiating an aspectual head. An important part of the argumentation concerns claims that the V2 in RVCs is predicated of the object, hence that ganiing 'clean' is predicated o f vifu 'clothes' in (20) as in structure (25): (25) 'wash' clothes ganjing 'clean' Chapter 5 argues that V2 elements originally brought two discrete properties to RVCs. First of all the V2 is suggested to have predicated directly onto the object. Secondly the V2 is also taken to have regularly encoded a telic bound on the activity represented by Vi just as in present day RVCs, the result-state represented by V2 indicating the end of the activity depicted by Vj. In the earliest RVCs it is assumed that V2 elements were syntactically licensed in such structures primarily in virtue of their predication relation with the object NP. Later on however, it is suggested that a significant increase in focus on the telic aspectual contribution of V 2 led to the re analysis o f V 2 as primarily an aspect marker (with certain additional inherent lexical content). This re-orientation of the primary contribution of V2 in RVCs from being 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. one of direct predication of a lexical property onto the object of Vj to being one of aspectual modification of the predicate is suggested to have resulted in V2 elements being re-analyzed as predicating o f the Vj and its object together. Formally this is taken to have resulted in V 1 and its object being re-analyzed as a VP occurring in the Spec o f an Aspect projection headed by V2 as in (26): (26) AspP Vi Object I (V 2) It is argued that V 2(asp) elements therefore came to be licensed in RVCs not in virtue of any predication relation holding between V2 and the object NP, but solely due to V2(asp> providing a telic aspectual bound to the predicate [Vi + NP], A direct consequence of this change in focus of the licensing o f V2 elements is suggested to be that certain V2 S began to occur in RVCs which did not predicate of the 'object' NP— the 'phase'-type V2 elements which are found in RVCs only from a historically later time. (27a) is an example o f such a phase V 2 zhu 'live' in modem Mandarin which does not predicate of the object NP ta 'he/him' as shown by the fact that (27a) does not entail the result in (27b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) a. wo liu-zhu-le ta I retain-live-LE he 'I stopped him.' b [ta zhu-le = he lived] + [this was a result of a retaining event] Following this, V2 taken to be re-analyzed as Asp0 underwent its critical re positioning to Vi. Here it is suggested that the re-ordering of V2 elements right- adjacent to V 1 was not triggered by any phonological reasons or enclitic properties of the V2 and in fact did not result from any movement of V 2 at all, unlike the attraction of de to the verb in the shi-de construction. Instead it is proposed that the linear re positioning occurred as an instance of simple re-structuring, in the following way. It is first suggested that objects in RVCs were frequently either topicalized as old information or occurred in null pronominal form and that this would have significantly given rise to regular natural instances of Vi V2 adjacency as schematized in (28): (28) a. NPj Subject Vi tj V2 topic structure b. Subject V 1 pro V2 object as null small pro With Vi and V2(asp) naturally adjacent in this way, it is suggested that the V2 aspectual head became re-analyzed as an aspectual SUFFIX taken to be combined with V 1 in the lexicon. The motivation for such a re-analysis is suggested to be the directionality of selection. It is argued that because the structure in (26) has a non branching Asp' node, such a structure is effectively interpreted not as an AspP with a leftward Specifier containing the VP argument o f Asp0, but instead as the structure in 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (29) which has no vacuously non-branching nodes and Asp0 with a leftward complement VP rather than a leftward Specifier: AspP (29) Given that Chinese is essentially head-initial in its basic direction of selection, structures such as (29) would seem to go against this canonical direction of selection. However, if V2A S P is re-analyzed as a suffix base-generated on Vj and only checked/licensed against Asp0 at LF, this automatically allows for the AspP to be mentally re-analyzed as preceding the VP as in (30), with the V 1-V2A S P unit raising to Asp0 at LF for licensing of the aspectual specification (i.e. Asp-feature-checking): subjecti P AspP Asp' VP Asp NP Vi-Asp(V2) Object The re-analysis of V2A S P as a suffix therefore critically allows for a non-head-initial structure to be naturally re-analyzed as being head-initial and in harmony with the canonical direction of selection in Chinese. The chapter thus adds formal 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. understanding to typological claims that there is language-internal pressure for a single direction of selection and that examples of non-canonical directionality in selection relations may undergo 'correction' over time wherever this is possible. Finally, the chapter notes that the occurrence of suffixation in head-initial languages is typologically much less common than in head-final languages for simple reasons relating to the linear ordering o f elements in SOV-type languages (Greenberg 1978 and Dryer 1992). A further important conclusion of the chapter is consequently the principled explanation it offers for how and why less frequent instances of suffixation in head-initial languages might in fact develop when a combination of grammaticalization and re-structuring co-occur. 1.5 Chapter 6: Verbal -L e, Aspect and Tense Chapter 6 turns to verbal -le and consider how the re-positioning of -le/liao relative to the main verb and its object noted in (23) relates to parallel changes observed in resultative constructions. Largely because o f the completive meaning o f liao/le and its re-positioning adjacent to the verb at a time when other V2 elements also became right-adjacent to the verb, it is concluded that liao/le historically underwent structural re-analysis as COMPLETIVE ASPECT similar to other V2 elements in RVCs. The chapter then concentrates on the synchronic status of verbal -le and notes that -le may occur together with other overt V2-completive elements as in (31). This suggests that le has a function different from simple completive aspect in these cases, since an explicit completive element is already present. 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (31) wo yijing kan-wan-le nei-ben-shu le I already look-finish-LE that-CL-book LE 'I already finished reading that book.' In order to account for the patterns found, I adopt Smith's (1997) two-tiered approach to aspect which suggests that there are two fundamental types o f aspect— VIEWPOINT ASPECT (including perfective and imperfective aspect), and SITUATION ASPECT (relating to the telicity and aspectual bounding of a predicate). I suggest that since its initial grammaticalization, verbal -le has developed from being a completive SITUATIONAL aspect marker to instantiate structurally higher PERFECTIVE viewpoint aspect, a view which is in line with common synchronic perceptions of -le. The chapter then attempts to provide a formal modeling of this hypothetical change of -le from completive aspect to perfective aspect and argues for a development o f the approach to grammaticalization introduced in Simpson (1998b) and assumed in chapter 2 for ge. Whereas this approach was initially intended to account for the re-analysis of free-standing morphemes, chapter 6 suggests that SUFFIXES are also subject to further grammaticalization and re-analysis in a highly similar way, and that the model of grammaticalization initiated for free morphemes can be naturally developed to account for this. It is argued that grammaticalization and re-analysis in both free and bound morpheme cases is critically movement- dependent and results from raising of an element upwards in the functional structure dom inating a lexical projection to successively highly positions. This hypothetical 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. re-analysis of -le is shown to have the further significant consequence that it strongly supports the Minimalist view that functional affixes are base-generated together with their lexical host and then licensed via movement to a functional head, such movement furthermore often being covert and taking place at LF (Chomsky 1993/1995a/1995b). (32) represents the structure argued to underlie (31) with wan being a completive aspect suffix checked against the lower situation Asp2° and -le being a perfective viewpoint aspect suffix licensed against the higher Aspi° when raised at LF: ( 3 2 ) Asp | P(im/perfective) The chapter also argues that with certain verbs -le may in fact still instantiate lower completive aspect as well as higher perfective aspect and that -le in such cases carries two sets of aspectual feature specificications, being sequentially checked Aspi' Aspi (le ) Asp2P(completive) Asp2 ' Asp2 °(lv a „) VP ▲ V' V k a n -w a n ^ ^ rle^ )) 'read-finish-LE' against both Asp2° and Aspi0 as in (33): 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) Asp i P (im/perfective) Aspi' Aspufc) A A S P 2P (completive) Asp2' ▲ V' V mai-le (A sp 2 . A sp i) 'sdll-LE' Synchronically then, the suffix -le is potentially ambiguous and (with certain verbs) allows for use as both as completive aspect (here reflecting its original source as a completive V2) and as perfective aspect. Following this, the chapter attempts to account for other aspects of the interpretation o f -le and makes the contentious speculation that -le has now grammaticalized and developed further to instantiate past tense as well as perfective aspect. This accounts for the necessary past time interpretation which accompanies its use in sentences such as (34). Note that the use of -le as just perfective aspect should be compatible with future time reference. However as (34) shows, this is not possible. 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) wo (*mingtian) (*hui) zai Beijing canguan-le Gugong I tomorrow will) in Beijing visit-LE Imperial Palace 'I visited the Imperial Palace in Beijing.' NOT POSSIBLE: 'Tomorrow I will visit/will have visited the Imperial Palace.' While the possibility that -le encodes past tense has been explicitly rejected by some linguists (e.g. in Li & Thompson 1981), it is argued that this is largely due to the narrow view that overt morphemes must necessarily stand in a fully rigid one- to-one correspondence relation with a single meaning/function (such as past). Chapter 6 suggests instead that functional meanings/interpretations are associated with morphemes as the result of an active combination process linking functional features and lexical hosts. In such a process it is suggested to be possible for a single morpheme-host to be combined with more than a single functional interpretation/feature (and therefore instantiate multiple functions). It is also argued that in certain circumstances (discussed in the chapter) the association of a particular function with a particular morpheme may be either optional or even blocked. (35) represents -le as being combined with and carrying both perfective aspect and past tense features and being licensed/checked against Aspi0 and T° during the course of LF raising: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. "T'O 1 de) AspP l=im/perfcctive A s p P 2=compIetive A s p i (|e) Asp2 (wan) kQn-WCM(Asp2)~te(Aspl, Tns) 'read-finish-LE' Such assumptions and a careful re-consideration of the objections to an analysis of -le as tense raised in Li & Thompson (1981) and elsewhere then ultimately allows for a full account of the distribution and interpretation of verbal -le and a principled defense of the hypothesis that -le instantiates tense. The chapter concludes that this interpretation is significantly not random but actually determined by clear syntactic factors, and that the further grammaticalization of perfective -le as tense is one which is fully expected given the model of grammaticalization developed and defended in the thesis. 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2. The Classifier Ge: Movement and Re-analysis 2.0 Introduction This chapter is an examination of the status of the classifier ge in nominal and verbal environments. There are two types of (related) elements pronounced as ge in Mandarin Chinese. One is a general classifier which is commonly used to "type" or individuate a noun which follows it for purposes of counting. The other ge will be shown to be used to count the event denoted by the predicate (in which ge occurs) as a single event. While the first use of ge has long been known as the most general classifier in Mandarin, the second type of ge has not been investigated thoroughly in the literature. The discussion in this chapter will be focused on the second type of ge. Section 2.1 presents ge as the most general classifier for count nouns. Section 2.2 discusses the distribution and the semantic contribution of the event classifier ge. Section 2.3 presents the syntactic properties of the event classifier ge. Section 2.4 makes a proposal to structurally encode ge within the DP in a way which mirrors its central syntactic properties. Section 2.5 then argues for the grammaticalization of ge from a classifier to a determiner. Section 2.6 is a brief conclusion. 2.1 Ge as a General Classifier for Nouns The most frequent use of ge is to serve as a classifier for its following noun. Mandarin nouns all behave like mass nouns in English in the respect that a classifier (or a measure word) is needed in counting. For example, to count "glasses" or 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. "water," one could use "a pair o f glasses" and "two drops o f water" in English. Ge is found to be the most general classifier for nouns. When ge occurs as a general noun classifier, it is preceded by a numeral and directly followed by a noun, as in (1): (1) a. wo you liang ge pingguo. I have two ge apple 'I have two apples.1 b. Zuotian si-le san ge ren. yesterday die-LE three ge person 'Three people died yesterday.1 When the numeral preceding a classifier is "one," it is well-noted that this numeral can in fact be deleted, as shown in example (2): (2) wo chi-le (yi) ge juzi. I eat-LE (one) ge tangerine 'I ate a tangerine.' Ge as a noun classifier is defined as "individual classifier" by Chao (1968). The noun following ge is a count noun denoting an object which usually does not have big volume and has to be a separable entity in the physical world. That is, the entity is physically "bounded" in Allen's (1966) sense. The verbs that can occur with the noun classifier ge are diverse. Stative verbs (as in la), action verbs (as in 2) and achievement verbs (as in lb) are all possible. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.2 Ge as a Classifier for Events Denoted by the Predicates The second use of ge, which constitutes the main focus of this chapter, is where it occurs preceded by a verb and followed by a bare noun, a classifier-measure string, an adjective, or a clause-like idiom. In what follows, this second ge will be shown to contribute to the aspectuality of the predicate and makes the denoted activity counted as a single event. 2.2.1 Occurrence Between a Verb and Its Cognate Object Although in its most common use ge follows a numeral and introduces regular concrete count nouns, it may also occur between certain V-O compounds. In (3a), the object part of the compound is only licensed to occur when together with the verbal part of the compound. In (3b), the verbal part 'jump' can mean 'dance' only when it occurs with the object part. When ge occurs separating the verb and its object, the ge-O sequence functions like an "individuated" cognate object o f the verb: (3) a. women qu you ge yong ba! we go swim ge swim Prt 'Let's go for a swim (do a swimming)!' b. women zuotian qu na-jia wuting tiao-le ge wu. we yesterday go that-CL club jump-LE ge dance 'We went to that club and did an activity of dancing.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Some speakers even accept examples such as (4). in which ge intervenes between the verbal members of a V-V compound: (4) xue ge xi study ge study 'do an activity of studying' In these cases, it can be argued that ge is used to help count out a single event and cause the action to be interpreted as being telically bounded. For example, if ge is omitted in (3a), the interpretation of the verb youyong is as a simple unbounded action of "swimming." However, when ge occurs between the two syllables of "swim," (3a) in fact means to "do a swimming" (and there is a single swimming event). Similarly in (3b), ge allows in the interpretation of one dancing event. This event may consist of one dance or more than one dance; the main point is that ge helps count out the event as a whole. Above in example (2) with nominal classifier ge it has been observed that the classifier is optionally able to appear without a preceding numeral, and that when this happens, the numeral "one" is implied. Consequently here where these V-O and V- V sequences occur with a 'bare' ge intervening, it can similarly be suggested that the numeral one is implied and that the occurrence of ge makes the action verb countable~as one event. Indeed, it is possible to insert the element yi 'one' in examples (3) and (4): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (5) you yi ge yong swim one ge swim 'do a swimming' (6) xue yi ge xi study one ge study 'do an activity of studying' 2.2.2 Occurrence Between a Verb and an Indefinite Object Ge can also directly occur between a verb and a regular non-cognate object nominal. This post-verbal nominal is usually indefinite and is often a bare noun, as in (7). The appropriate classifiers are tiao 'stip' for "fish" and gen 'stick' for "rope," as in (8). As in the examples in (3), it can be suggested that when ge comes in between the V-N sequence it functions to quantify over the predicate/event and results in a single buying event. Note that in (7a) there could be a single buying event of several pieces of rope, whereas in (8a) with the regular classifier gen for "rope," the interpretation is that only a single piece of rope is purchased. Similarly, in (7b) there could be a delivery of several fish, whereas in (8b) with the regular classifer tiao for "fish," the interpretation is that only a single fish is delivered. Consequently the use of ge might seem to relate more to the event/predicate than to the nominal: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) a. wo qu mai ge sheng. (from Liu 1993) I go buy ge rope 'Let me go do a rope-buying.1 b. wo qu song ge yu. I go deliver ge fish 'Let me go do a fish-delivering.' (8) a. wo qu mai gen sheng. I go buy CL rope 'Let me go buy a rope.' b. wo qu song tiao yu. I go deliver CL fish 'Let me go deliver a fish.' Examples in (9) shows the occurrence of a locative/instrumental object introduced by ge: (9) a. women lai tiao ge sheng. (see also Liu 1993) we come jump ge rope 'Let's do a rope-jumping.' b. women qu zou ge lu. we go walk ge road 'Let's go take a walk (do a walking on the road).' Here (9) is in contrast with (7) and (8). In (7) and (8) it is possible to have either the specific classifiers for specific nouns or the 'event' classifier ge, but examples with locative objects such as in (9) do not allow similar substitution of the specific classifiers for nouns, as shown in (10): 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (10) a. * women lai tiao gen sheng we come jump CL rope '*Let's jump one rope.' b. *womenqu zou tiao lu. we go walk CL road '* Let's go walk on one road.1 From this it can be concluded that ge is not a simple replacement for more specific nominal classifiers and does not simply function to introduce a regular NP object. For example, in (9a) the sequence tiao-sheng 'jump-rope' does allow ge to introduce sheng 'rope,1 but in (10a) it does not allow a regular classifier such as gen (which would be an appropriate classifier for sheng in other situations such as in 8a). There is consequently an important difference between ge and other classifiers whose function is to introduce and individuate simple NP objects. Next, the post-verbal nominal following ge is often a numeral-measure phrase. This numeral-measure phrase can be an object of an action verb as in (11) or a duration/frequency phrase as in (12). (11) a. ta chi-le ge san wan fan he eat-LE ge three bowl rice 'He did an eating of three bowls of rice.' b. ta xiang shang-tai chang ge san duan qur he want get-on-stage sing ge three session song 'He wants to get on the stage and do a singing of three operas.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (12) a. women qu Bali wan-le ge liang tian. we go Paris play-LE ge two days 'We went to Paris and did a two-day playing (a two-day trip).' b. qing ba zhe-ben shu zixinde kan ge wu bian. please BA this-Cl book carefully read ge five time 'Please do a five-time reading o f this book carefully.' c. yundongyuan dao cao-chang pao-le ge_ san quan. athletes go exercise-filed run-LE ge three rounds 'The athletes went to the track and did a three-lap running' Comparing (11) and (12) with (13) and (14) it is found that when ge is used, the whole verb phrase is interpreted as "one" single event or "one group" of events. For example, in (1 la), the inclusion of ge results in the interpretative effect that the eating of the three bowls of rice has to be understood to take place at one specific time, while in (13a), the three bowls of rice could be eaten at different meals. (13) a. chi-le san wan fan. eat-LE three bowl rice 'ate three bowls of rice' b. chang san duan qur. sing three session song 'sing three operas' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14) a. wan-le liangtian play-LE two days 'played for two days' b. kan wu bian read five time 'read five times' c. pao-le san quan run-LE three lap 'ran three laps' This single-event interpretation associated with the predicate can be highlighted by the interaction o f quantified or plural NP subjects and ge-marked objects. Sentence (15) means Zhangsan ate three bowls of rice at one time and Lisi ate three bowls of rice at one time. Strictly speaking, although there may be two events involved, ge functions to set the boundary of the activity denoted by the predicate: for each person the three bowls of rice is necessarily consumed during one eating event. (15) Zhangsan he Lisi ge chi-lege san wan fan. Zhangsan and Lisi each eat-LE ge three bowl rice 'Zhangsan and Lisi each at one time ate three bowls of rice.' 2.2.3 Ge Introducing a Simple Resultative/Descriptive Adjective Ge can also be vised to follow an action verb and precede certain adjectives in an adverbial usage. As shown in (16), ge can be used to introduce a descriptive adjective, which otherwise can not follow the verb. This intervention of ge again results in the interpretation that a single event (of 'fast-running') occurred: 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (16) Pingchang kan-bu-chu-lai, zuotian de yundonghui ta dao pao-le in daily life can-not-see yesterday's sports event he however run-LE ge kuai. GE fast 'I can't see that in him in daily life, but he demonstrated a fast running in yesterday's sports meet.' A more frequently seen case is when the following adjective is a resultative adjective. The interpretation here may be that the activity depicted by the verb is carried out until the point is reached where the subject can be said to have the quality described by the adjective, as illustrated in (17) and (18):1 (17) a. ta xiang zai na-jia canting chi ge bao he want at that-CL restaurant eat ge stuffed ' He wants to do a satisfying eating (keeping eating until stuffed) in that restaurant.' b. tazai na-jia canting chi-le ge bao he at that-CL restaurant eat-LE ge stuffed 'He did a satisfying eating (kept eating until stuffed) in that restaurant.' 1 Here the ge construction exhibits interesting differences from resultative constructions in English where a resultative secondary predicate may only refer to an object (or to the subject o f an unaccusative verb, i.e., an underlying object) but not to regular transitive/unergative subjects: (i) *John swam tired. (ii) John ran his horse tired. (iii) John arrived tired. (iv) John laughed "(himself) hoarse. 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (18) wo mama hui ba wo ma ge gou I mother will BA me scold ge enough 'My mother will do a satisfying scolding (keep cursing until enough) to me.' (19) wo zuotian qu zhao ta, jieguo pu-le ge kong I yesterday go look-for him, in-the-end catch-LE ge empty 'I went to visit him yesterday, but he wasn't home (didn't catch him)' (20a) and (20b) are further examples of adjectives combining with ge in similar fashion: (20) a. sha-le ge ganjing kill-LE ge clean 'did a clean killing (killed everything)' b. shuo ge qingchu say ge clear 'do a clear explanation' In these cases, it might be suggested that ge intervenes in the middle of some kind of action-result compound. However, the examples in (21) indicate otherwise - without ge these sequences of verb-adjective would require some other elements such as a sentence-final -le (as in 21b) or other continuing elements (as in 21c) to make the sentences acceptable. The contrast in (22) shows that when the action and the result are separated, and where there consequently cannot be any V-Adj compounding, ge shows that it is able to license the following adjective as a legitimate post-verbal element. 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (21) a. *ta zai na-jia canting chi-bao. he at that-CL restaurant eat-full b. ta zai na-jia canting chi-bao le. he at that-CL restaurant eat-full le 'He is already full from eating at that restaurant.' c. ta zai na-jia canting chi-bao cai lai de. he at that-CL restaurant eat-full then come Prt 'He came here after he was full from eating at that restaurant.' (22) a. *ta chi-le bao. he eat-LE stuffed b. ta chi-le ge bao. he eat-LE ge stuffed 'He had a completely satisfying eating.' Concerning the use o f ge with adjectives here, it can be suggested that ge in fact has an aspecf-related function. Following ideas in Tenny (1987) and Borer (1994), it can be assumed that the occurrence of certain nominal objects may be necessary to make atelic activity predicates into telic bounded events, and that ge here critically functions to convert a non-nominal category into a nominal type which will result in such an interpretation (nominal elements 'measuring-out' an event and resulting in an event being bounded when the object is fully 'consumed'— see Tenny (1987)/Borer (1994). The use of ge as a simple nominal classifier has 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. therefore become extended to a wider use and may now be used to convert elements of adjectival type into nominal type elements.2 A further peculiarity of the ^e-construction is that when the adjective following ge can be made more complex by being combined with some kind of degree-type modifier, the expression with ge generally sounds better, as seen in (23) and (24). The modifier-head adjectives in these examples denote an extreme result or an extreme description. (23) a. he-le ge da-zui drink-LE ge big-drunk 'did a completely drunk drinking' b. ?he-le ge zui drink-LE ge drunk 'did a drunk drinking' 2 In this way ge is somewhat similar in function to other numeral-classifier strings, as illustrated in the examples in (i). It seems that the presence of a numerically-quantified object is necessary in (i) to provide a natural measure and endpoint for the event (compare here (i) with (22) above): (i) a. *ta chi-le fan. he eat-LE rice 'He ate.' (bad unless "rice" is in contrast to "noodle," for example) b. ta chi-le yi-wan fan. he eat-Prf one-bowl rice 'He ate a bowl of rice.' 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (24) a. zou-le ge fei-kuai walk-LE ge fly-fast 'did a as-fast-as-flying walking' b. ?zou-le ge kuai walk-LE ge fast 'did a fast walk' Concerning this effect o f adding a 'modifier' to an adjective in (23a) vs. (23b) and (24a) vs. (24b), it can be argued that if the object is modified and in a certain way made more specific/identifiable then this enables it to function as a delimiter more easily, either providing an end-point or otherwise measuring out the event to make it telic. If the end-point is some kind of extreme (a 'big drunk') then the telic reading is all the more easy to license. 2.2.4 Ge Introducing an Idiom or Fixed Expression In section 2.2.3 it was shown that the element introduced by ge can be an adjective, and the suggestion was made that ge may be functioning to provide an essentially nominal complement to the verb which precedes it. Here it can be noted that the element which is introduced into the VP via ge can actually also be a clause-like idiom, expressing the extreme result of some action.3 Again in such instances the preceding verb is an action verb and the ge-XP contributes to convert the predicate into an accomplishment: 3 These sequences are 'clause-like' in the sense that a full (and overt) argument structure is present. 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) a. pin ge ni-si-wo-huo fight ge you-die-I-live 'make a terrible fight (aim to kill the other party)' b. nao-le ge ji-quan-bu-ning fuss-LE ge chicken-dog-not-peaceful 'caused a big turmoil' c. wen ge shui-luo-shi-chu ask ge water-fall-rock-out 'do a thorough interrogation until everything is clear' 2.2.5 Ge with Perfective but not Imperfective Aspect There are also certain aspectual restrictions on ge-forms, and when the predicate has an overt aspect marker, as in (26) and (27), it is found that event classifier ge can co occur with the perfective marker -le in (28), but not with the imperfective markers zai in (26) and zhe in (27). This may perhaps be because imperfective aspect markers function to focus on the internal stages o f an event and event classifier ge objects are interpreted as constituting an unanalyzable single event with no visible internal stages, in this sense being more like instantaneous achievement-type predicates (see here Smith 1997 and chapter 6 for more discussion o f the role of perfective/imperfective aspect). Such an unanalyzable single event property will also account for the fact that in ge objects which have numerals, such as (11/12/15) above, there can only be a single event interpretation and no possibility of a multiple event reading— the internal parts o f a ge object are simply unavailable for individual interpretation/imperfective aspectual modification: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (26) *Wo zheng zai die ge tou-po-xie-liu. I right at fall ge head-break-blood-flow '*1 am right at falling so hard that I got a broken bleeding head.' (27) *Wo die-zhe ge tou-po-xie-liu. I fall-Dur ge head-break-blood-flow '*1 am at the state o f falling so hard that I got a broken bleeding head.' (28) Wo die-le ge tou-po-xie-liu. I fall-LE ge head-break-blood-flow 'I had a fall so bad that I got a broken bleeding head.' 2.2.6 Apparent Constraints on Verb Selection While it was noted that the noun classifier ge can occur with a stative verb, an action verb, or an achievement verb, most of the occurrences of the event classifier ge actually follow an action verb, and often it may not occur with a stative verb: (29) *Ta you-le ge san zhi bi. he have-LE ge three CL pen 'He had three pens.' This constraint on verb selection can be shown to be aspectual in nature. If the ge- XP is able to express an end point and make the whole VP as an identifiable event, even stative verbs are in fact permissible. For example, the stative verb "love" in (30), with the help of ge-XP, similarly expresses an identifiable event: the subject wants to have had "one" exciting romance. Once again it seems therefore that the ge-XP functions to change an atelic predicate into a telic predicate. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (30) a. Ta xiangyao ai ge si-qu-huo-lai. ta want love ge die-go-live-come 'He wants to experience an extreme romance.' b. Ta ai-le ge si-qu-huo-lai. ta love-LE ge die-go-live-come 'He experienced an extreme romance.' In addition to action verbs, certain achievement verbs can also occur with the event ge: (31) a. Ta xiang bazhejian shi w angge ganganjingjing. he want ba this CL matter forget ge clean 'He wants to forget this completely.' b. Zhe ci wenyi, cun-li de ren si-le ge jingguang. this CL plague village-in's person die-LE ge bright-clean 'The people in the village died out because of the plague.' As achievement verbs by definition inherently encode end points, it might perhaps be incorrect to suggest that the ge-XP functions to make the predicate telic by specifying the end point of the action. The ge-XP in (31) might instead be said to reinforce the boundary of the event and make the event be perceived as a whole. However, without the ge-XP, the event boundary would in fact have to be expressed by some other means, for example by the use of a perfective aspect-marker as in (32a) and (32b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (32) a. Ta xiang bazhejian shi wang *(le). he want ba this CL matter forget le 'He wants to forget this.' b. Zhe ci wenyi, cun-li de ren *(dou) si-guang *(le). this CL plague village-in's person all die-clean le 'The people in the village all died because of the plague.' Consequently it may in fact be the case that the ge-XP also in these cases does function to delimit the predicate and make it telic. Elsewhere in Chinese it is found that verbs which would be classed as telic in English and other languages need to be overtly specified as bounded by the addition of an aspect-marker or resultative complement. For example, (33) is possible in Chinese, though not in English. With English a predicate like 'kill' plus past tense implies that the event depicted by the predicate did have a successful completion, whereas in Chinese it is not necessarily true. It is suggested here that ge-XP functions as one of the means that make a predicate telic. (33) Zuotian Zhangsan sha Lisi, keshi mei sha-si. yesterday Zhangsan kill Lisi, but not kill-dead '[Yesterday Zhangsan killed Lisi, but Zhangsan didn't kill Lisi.' 2.3 Syntactic Properties 2.3.1 Event Ge Introducing a Non-predicational Adjective Here I would like to note a further restriction on the event ge which can be observed when ge combines with adjectives. Basically it is found that the adjective cannot be 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in 'predicational' form but rather must be 'bare.'4 When adjectives occur as predicates in declarative sentences in Chinese, they are accompanied by a modifier such as hen 'very' or tai ’ too’ as in (34), whereas when they occur attributively they appear bare, as in (35):5 (34) a. Ta hen piaoliang. ta very pretty ’ she is very pretty.’ b. Zhe-ge fangzi tai jiu. this-CL house too old 'This house is too old.' (35) xin shu new book 'new books'6 Turning to ge-forms it is found that adjectives which occur following ge cannot be modified by hen 'very' or tai 'too,' as shown in (36).7 4 Note that earlier it was observed that sometimes an adjective is more natural with ge if it is however combined with another element indicating high/extreme degree of the adjective, as in examples (23) and (24). This kind o f modification does not result in predicational forms, unlike the use of hen and tai attempted in (36) and so still leaves the adjectives 'bare' in a certain sense. 6 Note that the use of hen does not necessarily result in an interpretation of'very' but appears to be necessary to fulfill some rather unclear structural requirement that the pre-adjectival position be filled by some overt element in clauses. 6 If hen 'very' is added to the adjective, the element de must also occur, indicating that this is a relative clause structure and a predication, e.g.: (i) hen xin *(de) shu very new DE book *books that are very new’ 7 Compare (36) to the sentences with the resultative de in (i). 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) a. *shuo-le ge hen qingchu say-LE ge very clear b. *chi-le ge tai bao eat-LE ge too full Secondly, there is a further restriction found in adjectival reduplication forms. Chinese allows for bisyllabic adjectival reduplication in two basic forms commonly referred to as the AABB Form and the ABAB form. The AABB reduplication form is referred to as the 'vivid' adjectival form which expresses an extreme description, while an ABAB reduplication is said to be predicational reduplication and is shown to occur in standard predicate type positions in (37) (see e.g. Chao 1968): (37) a. Mingtian women chuqu tongkuaitongkuai. tomorrow we go out enjoy 'Let's go out and enjoy ourselves tomorrow.' b. Jintian ni hua-yixia-zhuang, rang ziji piaoliangpiaoliang. today you put one makeup let self beautify 'Why don't you put on makeup and make yourself look beautiful today.' (i) a. shuo de hen qingchu say de very clear 'said it very clearly.’ b. chi de tai bao eat de too hill 'eat too full' Resultative de selects for a complement which may be instantiated by a predicational adjective here, this contrasting with ge which cannot introduce predicational form adjectives. 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the ^e-construction it is found that adjectives may be reduplicated according to the AABB pattern but not the ABAB predicational pattern as shown in examples in (38) and (39): (38) a. chi ge tongtongkuaikuai eat ge over-enjoyed 'do a completely satisfying eating1 b. daban ge piaopiaoliangliang dress-up ge beautiful 'dressed up to be completely beautiful' (39) a. *chi ge tongkuaitongkuai eat ge over-enjoy 'ate until satisfied' b. * daban ge piaoliangpiaoliang dress-up ge beautify 'dressed up beautifully' A third restriction on the combination of ge with adjectives is that the adjective following ge can not be syntactically negated. As shown in (40a), ge can not introduce an adjective which is syntactically negated by bu 'not.'8 (40) *shuo-le ge bu qingchu say-LE ge not clearly 'say (it) not clearly.' 8 Compare (40) the sentence with the resultative de. (i) Ta shuo de bu qingchu. he say de not clearly 'He said it unclearly (He did not say it clearly).' 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, if the negation is part of a fixed negative adjective, as in (41), it is a legitimate item to follow ge. Being able to be syntactically negated is another characteristic of a syntactic predicate, so again there is evidence that a predicational adjective cannot follow ge. (41) shuo-le ge bu-qing-bu-chu say-LE ge not-clearly 'make a vague statement' Lastly, the adjective following ge can not be A-not-A questioned. As shown in (42), ge-sentences cannot be made into an A-not-A question by questioning the adjective.9 This is another predicational characteristic that the adjective selected by ge is lacking. (42) *Ta shuo ge qing-bu-qing-chu? he say ge clear-not-clear 'Did he make it clear?' I would like to suggest that all of these restrictions are actually rather typical of nominalization forms in certain languages and may be explained if it is assumed that (a) ge is functioning as a nominalizer (a categorization in fact suggested in S. Lu 1984), and (b) ge selects for a complement which is syntactically 'less' than a full clausal predicate. 9 Compare ge in (42) to the resultative de\ (i) Ta shuo de qing-bu-qing-chu? he say de clear-not-clear 'Did he make it clear?' 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Specifically concerning point (b), it has been noted that in languages such as Korean there is variation in the type o f predicate constituent selected by different nominalizers. Following initial work in Yoon (1990), Lapointe & Nielsen (1996) show that Korean Type II' and 'Type III' nomimalizations have certain critically different clusters of properties. The nominalizer in the former selects for a full clausal complement which has an overt tense morpheme, a nominative subject and allows for sentential negation, sentential adverbs and clause-internal scrambling. The nominalizer in the latter however seems to select for a more restricted constituent which does not allow scrambling, negation, sentential adverbs, nominative subjects or tense/aux elements (subjects instead occur in genitive case). Such a contrastive patterning can be captured quite simply if it is assumed that the Type II nominalization involves selection by the nominalizer of a TP-level constituent, whereas in Type III nominalizations the nominalizer selects a VP-type level constituent. Returning to Chinese, with ge one finds that although a predicate may be selected and even project its arguments (as in the idiom cases in (25)), there are certain syntactic properties associated with full regular clauses which may not occur in the predicate sequence following ge, e.g., clausal negation and A-not-A question forms.1 0 Such syntactic properties, and possibly also the type of reduplication found 10 Note that the negative element bu in (41) occurs in a frozen expression and therefore is not part o f the process of a syntactic negation, while (40) is a case o f syntactic negation. I assume that syntactic negation in examples such as (40) involves a clausal structure. According to Cinque (1999), clausal negation can occur in different positions cross-linguistically. It is therefore not unreasonable to find two or more clausal negative elements co-occurring in one sentence: 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with predicational adjectives noted, may all be assumed to require a certain minimal amount of clausal structure in order to be licensed. For example, clausal negation would occur located in the functional part of a clause and not low down within VP/AdjP, and A-not-A forms arguably require the presence of a Q-morpheme in I, hence the functional projection IP, etc. This absence of clausal negation, A-not-A questions and predicational adjectival reduplication can consequently be suggested to indicate that the functional structure necessary to support such properties is also absent, in a way which is rather similar to Korean Type III nominalizations. Similar to Korean Type III nominalizations, I would therefore now like to suggest that ge is in fact a nominalizer which takes as a potential complement simply a bare predicate (maybe an AdjP) without allowing any of the higher clausal functional projections which would be necessary to support negation, A-not-A and predicational reduplication, etc.1 1 (i) Ta meivou bu xiang qu. (p.c., Audrey Li) he didn’ t not want go 'He didn’ t not want to go.' The argument here is that the XP following ge can not be a clausal structure. Different negative items can occur as long as it does not involve a process o f syntactic negation (and hence a clausal structure). For example, the negative item mei can also occur in the frozen expression following ge in (ii): (ii) Ta shuo-le ge mei-wan-mei-liao. he speak-LE GE not-end-not-finish 'He made an endless talk.’ 1 1 Given the well-formedness o f (41) one might have to assume that some instances of negation may in fact be combined with other elements in the lexicon. This is a position which has actually frequently been adopted in the past, for example by certain authors working on negative auxiliary elements such as English can't, won't, etc. (in Zwicky 1977), or Cantonese mo = mei + vau in (Matthews & Yip 1994). 51 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.3.2 Event Ge Licensing an Extra Argument Thus far we have seen sequences o f a verb followed immediately by ge and some other element. However, elsewhere it is also possible to find sequences of [ verb + (in-)direct object + ge-object ], i.e., the ge-XP can also follow an object of the verb.1 2 These sentences are more acceptable to speakers from Mainland China; the examples in (43) are cited from novels written by writers from Mainland China, recorded in Fang Li (1993). 12 Objects occurring between the verb and ge are however rather uncommon, it being usually preferred to position them pre-verbally licensed by ba as in (i): (i) wo ba ta da-le ge ban-si I BA he hit-Asp GE half-dead 'I hit him half-dead.' 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (43) a. ta da-le wo ge ban-si. he hit-LE I ge half-dead 'He hit me to the extent that I almost died.' (Wang Meng's <Qingchunmeng> [Youth dreams]) b. da ni ge tou-po-xue-liu hit you ge head-break-blood-flow ■ hit you until you have a broken bleeding head' (Wang Shuo's <Wo Shi Ni Baba> [I am your father]) c. Bie yiwei rumen shi nusheng, fan-le gui zhaoyang da nimen ge don't assume you are girls, violate rule same hit you ge si-j iao-chao-tian. four-foot-face-sky 'Don't just assume you are girls. I will still hit you to the extent that you all fall down on you back, if you break the rules.' (Wang Meng's <Qingchunmeng> [Youth dreams]) d. Er guniang cui-le baba yi ge man-lian-hua. two miss spit-LE father one ge full-face-messy 'The second daughter spit at her father to the extent that his face got all messy.' (Lao She's <Liutunde> [Women from the Willow Village]) As is well-known, in Chinese it is generally not possible to have two discrete elements following the verb, unless one of these is a frequency complement as in (44): (44) wo qu-guo Beijing hen-duo-ci I go-Exp Beijing many times 'I have been to Beijing many times.' 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Normally if an element such as a descriptive or resultative complement occurs and the object o f the verb is overt, the verb must be repeated to license the object: (45) *Zhangsan da erzi de ban-si Zhangsan hit son DE half-dead (46) Zhangsan da erzi da de ban-si Zhangsan hit son hit DE half-dead 'Zhangsan beat his son until he was half-dead.' However, it is found that a ge-XP may in fact co-occur with an overt object DP as in (47), in direct contrast to the non-ge resultative structure (45): (47) Zhangsan da-le erzi ge bansi. Zhangsan hit-LE. son ge half-dead 'Zhangsan hit his son to the extent that his son almost died.' Such a structure with two elements following the verb in fact appears very similar to double object constructions (DOCs), roughly shown as (48). (48) a. V + NP Object + ge-XP ge-construction b. V + Object 1 + Object 2 DOC According to Li and Thompson (1981), an important characteristic of the double object construction in Chinese is that Objectl which directly follows the verb is always either an animate noun or an institution name, this being illustrated in (49). 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (49) a. Wo song-le ta yi ben shu. I give-LE he one CL book 'I gave him a book.' b. Tamen huan-le yinhangyi bi daikuan. They retum-LE bank one CL loan 'They returned a loan to the bank.' c. *Ta song-le xinjia yixie jiaju. he give-LE new-house some furniture Interestingly, ge-sentences seem to exhibit the same constraint on the object, as shown in (50), the (c) example is unacceptable because the direct object vifu is inanim ate:1 3 13 An explanation o f the animacy patterns in DOCs put forward in Kayne (1994) suggests that the fact that the indirect object in DOCs must always be +animate (or an institution) is because the indirect object comes to be in possession of the direct object in DOCs. The evidence from the Re construction here might seem to indicate that such a generalization may not be entirely correct— basically a parallel [+animate] constraint characterizes both DOCs and the ge-construction, but there is no possession relation arising in the latter syntactic forms. However, note that in 'di-transitive' ge constructions such as (50) the result depicted by the ge-XP always predicates o f the NP following the verb rather than the subject o f the sentence. In order to explain this, one might assume that the post verbal NP and the ge-XP form a single small clause-type constituent with the ge-XP naturally predicating o f the post-verbal NP as the small clause subject. Such an analysis would then actually come to resemble the small clause-type analysis of indirect object and direct object proposed by Kayne. 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (50) a. ta xiangyao xia wo ge hun-fei-po-san he want frighten I ge soul-fly-spirit-scatter 'frighten me until my soul and spirit leave my body' b. Naxieren wei-le yiyuan yi ge Those people surround-LE hospital one ge shui-xie-bu-tong. water-leak-not-through 'Those people surrounded the hospital to the extent that nothing can get through in the hospital.' c. *Ta xi-le yifu ge ganganjingjing. he wash-LE clothes ge clean One might suggest that the occurrence of two overt DPs post-verbally in DOCs is permitted because the indirect object can be licensed via incorporation into the verb (cf. chapter 4 and Audrey Li 1990), leaving the direct object to receive the verb's structural accusative case. The same incorporation process might then be taken to be responsible for the well-formedness of sentences with a post-verbal direct object and age-DP.1 4 !5 14 A second possible explanation might be that the ge-XP is always non-specific indefinite and hence can be licensed via some kind of inherent partitive case. 15 Elsewhere however the parallelism in behavior between DOCs and di-transitive ge-forms diverges somewhat. Specifically the object directly following the verb exhibits differences in the two constructions in its ability to undergo movement. Whereas the object following the verb can be topicalized in both constructions and also ba-raised or bei-raised in ge-sentences, object 1 in the double object construction cannot occur raised with either ba or bei: (i) a. Nei-ge-ren, wo da-le ge bansi. he I hit-Prf ge half-dead 'Him, I hit him until he almost died.' b. ?Nei-ge-ren, wo song-le yi-ben-shu. he I send-Prf one-CL-book 'Him, I gave a book.' 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.3.3 Ge Licensing the Expletive ' It' Here we note a last interesting aspect of the ge paradigm: its occurrence with ta in object expletive contexts. Lin (1993) shows that expletive ta can often co-occur with object DPs and suggests that an expletive-associative chain is formed between ta and the real argument DP, the argument raising to replace ta in SpecAgrOP at LF. (51) illustrates this use of ta with an object associate: (51) wo kan-le ta san-ben-xiaoshuo I read it one-Cl-novel ’ I read three novels (at one time).' Here it can be shown that ge-XPs may also often co-occur with the expletive ta 'it' (called by Lin 1993 as an expletive object). The element introduced by ge can be either an adjective or an indefinite nominal which is a measure phrase (or more marginally a bare noun), as shown in (52): (ii) a. Wo ba ta da-le ge ban-si. I ba him hit-Prf ge half-dead 'I hit him until he almost died.' b. *Wo ba ta song-le yi-ben-shu. I ba him send-Prf one-CL-book 'I gave him a book.' (iii) a. Ta bei wo da-le ge ban-si. he bei I hit-Prf ge half-dead 'He was hit by me and he almost died because o f it.' b. *Ta bei wo song-le yi-ben-shu. he bei I send-Prf one-CL-book 'He was given a book by me.' Both ba-constructions and bei-constructions involve affectedness o f the theme or the patient. In the double object construction, what is affected is the direct object (i.e., the book in the examples), not the indirect object. I assume that it is the affectedness effect which in fact prevents the indirect object of a di-transitive verb from being ba- or bei-preposed. In the di-transitive ge-construction by way of contrast, the NP following the verb is affected and so can be ba-/bei-preposed. Consequently the syntactic parallelsim between DOCs and di-transitive ge-forms still holds. 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (52) a. chi ta ge tongkuai eat it ge satisfied 'have a satisfying eating' b. pao ta ge san quan run it ge three CL-rounds 'do a three-lap running' c. ??tiao tage wu dance it ge dance 'do an activity of dancing' What is important to note is that ta in these object expletive structures always requires a DP associate. This therefore implies that ge is crucially functioning to provide a nominal DP expression in the event ge construction, sometimes converting non-nominal categories (such as adjectives) into DP elements. Secondly ta has been noted by Lin to tolerate only DP associates which are headed by weak determiners such as ji-CL ('several N1 ) and yi-xie ('some NP'), but not nei-CL ('that NP') or mei- CL ('every NP'): (53) wo yao kan ta ji-ben/yi-xie lishi-xiaoshuo I want look it few/some history novel 'I want to read a few/some history novels.' (54) *wo yao kan ta nei-ben/mei-ben lishi-xiao-shuo I want look it few/some history novel 'I want to read that/every history novel.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This would seem to indicate that the ge-headed nominals must also be DPs with some weak determiner element in the D-head position.1 6 1 7 2.4 The Proposal - G e as a Weak Unselective Determiner in D° Having provided a characterization of a number of restrictions and patterns found with what has been referred to loosely as 'event ge,' I would now like to argue that the evidence presented would all seem to indicate a single solution: the element ge is currently used not only as a regular numeral classifier, but also now as a weak unselective determiner with a non-specific indefinite specification. It will be discussed that this element is similar in certain respects to the unselective determiners reported to exist in Salish languages by Davis and Matthewson (1997), and it also bears similarities with a certain use of Cantonese classifiers reported by Cheng & Sybesma (1998) and Audrey Li (1998). 16 Lin (1993) and Borer (1994) essentially both assume that the DP-object-associate raises to SpecAgrO/AspP at LF. For Borer this has the effect of inducing a telic reading, activating the Asp- head which is associated with telicity. 17 Note that when ge is not present, only the (b) example in (52) survives, as shown in (i). (i) a. *chi ta tongkuai eat it satisfied 'ate until fully satisfied' b. pao ta san quan run it three CL-rounds 'do a three rounds of running' c. *tiao ta wu dance it dance 'do a dancing' This is because the expletive ta requires a weakly-quantified DP associate, as noted. Where ge is present, this provides the necessary type o f DP object associate for ta, but when it is removed only (b) remains a weakly-quantified DP; (a) and (c) can be suggested to unacceptable because tongkuai and wu are interpreted as verbal/non-nominal and hence of the wrong categorial type to constitute associates for ta. 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.4.1 As a Nominalizer In order to reach the conclusion above, a brief review of the patterning observed in section 2.2 will be helpful. It was observed in Section 2.2 that the element ge may occur preceding nouns which are already quantified by numeral-classifier expressions, as for example in (55): (55) he ge san-ping jiu (see also 11 and 12) drink ge three-CL wine 'drink three bottles of beer at one time' It was also significantly shown that ge may occur introducing not only elements which are clearly nominal (as in 55), but also other XP-types which may not be nominal at all, a common case being adjectives as in (56a), but also clause-like idiom expressions as in (56b) and sub-parts of V-V compounds as in (56c): (56) a. chi ge bao (see also 17) eat ge full 'do a satisfying eating' b. wen ge shui-luo-shi-chu (=25c) ask ge water-fall-rock-out 'do a thorough interrogation until everything is clear' c. xue ge xi (= 4) study ge study 'do an activity of studying' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In this sense it seems that ge is rather unselective in terms o f the syntactic category type it selects and introduces into the object position.1 8 A comparison was also made between ge and other regular classifier types and it was argued that ge is not simply replacing a more specialized regular classifier when it occurs in certain objects, as the attempt to replace ge with a regular classifier is often found to be unacceptable: (57) tiao ge/*gen sheng (see also the discussion in 7-10) jump ge/CL-strip rope 'do a rope jumping/*jump one rope1 This would seem to indicate that ge is not simply occurring in the standard classifier position and that 'event ge' forms do not result from simple deletion of the numeral yj 'one.' Rather, ge would appear to have a rather different status from standard classifiers and arguably is occurring in a different syntactic position. It was also shown that the ge-XP has an important aspectual function, often serving to provide either an end point for an activity predicate or an object which can be measured-out in the course of the event (and so also naturally bounding the predicate in this function too). Data reviewed indicated that ge-XPs could only occur with perfective and not imperfective aspect-markers (and marginally with stative predicates if an inchoative interpretation is forced, example 30), this being fully consistent with the assumption that ge-XPs function to convert a non-telic 18 The output o f applying ge to any o f its possible complement XP types is however assumed to be consistent, and ge as a nominalizer applied to any o f these XP will result in a constituent YP [Y p ge XP] which is outwardly specified as being a nominal-type projection. 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. predicate into one which is aspectually bounded. Noting that work carried out by Tenny (1987) and Borer (1994) has argued that underlying nominal objects (or unaccusative subjects) may be necessary for the interpretation of a predicate as telic, it was suggested that ge might seem to be functioning as a nominalizer in part, converting certain non-nominal categories into object DPs which could then serve as the necessary input for the interpretation of a predicate as telic. It was further noted that there are various restrictions on ge-XPs - it was shown that adjectives introduced by ge may not undergo predicate-type ABAB reduplication, that regular negation may not occur between ge and its complement, and that adjectives in ge-XPs may not undergo A-not-A question formation. It was suggested that restrictions of this type might be explained by likening such patterns to other similar restrictions found with nominalization phenomena in languages such as Korean. Nominalizers in Korean would seem to differ crucially with regard to the size of the syntactic constituent they may embed, Type II nominalizations embedding a full clausal constituent and allowing for all regular clausal elements and syntax to occur internal to the nominalization, while Type III nominalizations only embed VPs or smaller clausal/lexical projections and disallow elements and syntactic phenomena which are dependent on higher functional heads in the clause. Such differences have also been noted to occur in English (e.g., in Abney 1987), with the -ing nominalizer arguably being associated with either an IP-like constituent as in (58), this licensing nominative case and accusative case, or selecting a VP-size complement and forcing 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. genitive case and o/^insertion for objects (due to the lack of T and AgrO heads) as in (59): (58) [John reading his book] was encouraging. (59) [John's reading o f his book] was encouraging.1 9 Consequently there might seem to be good evidence for a nominalization hypothesis, the absence of ABAB reduplication, A-not-A question forms and regular negation being attributed to the possibility that ge as a . nominalizer might be selecting a bare Adj(P) which would not be able to support such syntactic processes. A final set of evidence relating to patterns found with object expletive ta was then argued to confirm this suggestion. Observing that object expletive ta always seems to require a weakly-quantified DP associate, a natural conclusion to make is that ge-XPs also have such a status, and that the element ge is critical in performing a nominalization category-conversion function. Putting all this information together, it seems that ge can be convincingly argued to function as some kind of nominalizer. A natural question to ask now is what the descriptive term nominalizer actually corresponds to in formal syntactic terms, and what position in the Chinese DP/NP/C1P ge actually occurs in? 19 Note also that negation can occur in the (58)-type form and not in (59), supporting the suggestion that (58) contains an IP clausal projection whereas (59) is a simple nominalized VP: (i) a. [John not reading his book] upset us. b. ^[John's not reading of his book] upset us. 63 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.4.2 In D° Position Here I would like to suggest that the most natural interpretation of the evidence and patterning found may be to conclude that the element ge is actually a determiner in the D position selecting for a variety of complement types and converting them into a 'nominal' output which can then function to provide an aspectual bounding for an activity predicate. The clues which lead to such a conclusion also come from various parts of the evidence presented. First the nominalization function of ge has been seen to be suggested by a number of phenomena. The patterns with object expletive ta then indicate that the overall category of the ge-XP can reasonably be argued to be that of DP, headed by some overt weak determiner (noting that object expletive ta will not combine with an NP which does not have any overt weak determiner, cf: *kan ta shu 'read it book(s)'). This consequently might seem to indicate that ge itself must be interpreted as the weak determiner head of the DP. As ge is elsewhere assumed to have the category of classifier however, one might perhaps question whether this suggested (re-)assessment of (event) ge as a determiner type is actually correct. Other evidence indicates that such a conclusion is in fact rather well-supported. First of all it was seen that (event) ge may often not be simply replaced by a more specialized classifier (example (57)), which may be taken to indicate that ge is not just instantiating the regular classifier position but is inserted as some other syntactic category type. Secondly we saw that (event) ge may also precede a full numeral-classifier expression as in (55), which indicates rather clearly that ge cannot in fact be occurring in the regular classifier position but must 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in fact be located in some higher functional head.20 A natural conclusion then is that this higher functional head is in fact D, and that the occurrence o f ge in D results in the nominalization of whatever complement it is that ge introduces. By drawing a parallelism between a V-O-ge-XP string and the double object construction, I have also shown how the event ge makes it possible for the ge-XP to act like an (extra) argument of the verb, which can furthermore be said to be a typical function of a determiner. Szabolsci (1994) has proposed that the primary function of determiners and complementizers (which occur in similar structural positions in nominal and clausal constituents) is to act as "embedders" and enable a proposition or a DP to refer as an argument. 2.4.3 Further Evidence Event ge is therefore now suggested to be a weak determiner instantiating D and higher in the nominal functional structure than simple numeral-classifier units. There are two further consequences o f such an analysis which can be argued to support it further. First of all if ge is in D and not the regular Classifier position Cl, we can explain why ge may not be preceded by numerals such as and san 'three' and si 'four,' etc.— numerals will occur lower in the functional structure and not raised in D or SpecDP: 20 It might perhaps be suggested that ge in these cases corresponds to something like a 'group classifier1 and therefore does still occur in a higher classifier position. However, event ge can also embed adjectives with no group meaning. If one intends to aim for a uniform analysis of event ge it would therefore seem that ge is best analyzed as something other than a straightforward classifier in Cl. 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (60) a. Wo xiang chi ge bao. I want eat ge full 'I want to keep eating until I am stuffed.' b. Wo xiang chi yi ge bao. I want eat one ge full 'I want to keep eating until I am stuffed.' c. *Wo xiang chi san/si ge bao. I want eat three/four ge full Elsewhere (e.g., in S. Lii 1984) it has been suggested that event ge is actually the abbreviated form of yi-ge 'one-ge' and hence that ge should in fact be in the regular Cl position, simply with a deleted number one. Here I would like to suggest that there may in fact be a deleted or covert yi occurring with ge, but that this element is not in fact the regular number one but rather an equivalent to the indefinite article a/an in English. While certain languages do have indefinite articles which are phonologically distinct from the number one (as e.g., English), there are many languages which make use of the number one also as an indefinite article (e.g., German ein. French un, etc.). Chinese can therefore be suggested to belong to this latter (far larger) group and allow for yi to be syntactically employed either as the number one or as an indefinite article.2 1 The distributional facts reported here would also seem to support the treatment of yi as an indefinite article syntactically distinct 21 Concretely I would actually like to assume \hzt yi-ge is a single grammaticized X° head and does not consist in two discrete parts/heads (noting that neither y i nor ge can be substituted for by any other numbers or classifiers when this unit occurs in the D° position preceding other numerals and classifiers or adjectives, etc.). It is yi-ge therefore which is the D° indefinite determiner equivalent to English 'a/an.' I furthermore suggest that simple bare ge is fully equivalent to yi-ge and results from simple reduction o f yi-ge to ge. 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from the number one yi too— only the article yi can occur (with ge) preceding a numeral-classifier-N sequence:22 (61) a. ta e jile, henhende chi-le (yi) ge san-wan-fan. he hungry extremely fiercely eat-LE one ge three-Cl-rice 'He was very hungry, so he fiercely ate three bowls of rice (at one time).' b. *ta e jile, henhende chi-le si ge san-wan-fan. he hungry extremely fiercely eat-LE four ge three-Cl-rice Such a syntactic difference between an indefinite article and a homophonous number one can be found elsewhere in Thai and supports the distinction suggested here. In standard Thai the number 'one' nung occurs preceding the classifier as in 22 Note that although (yij-ge as an indefinite determiner can freely precede other numerals and classifiers as in (61) and other examples in the chapter, it cannot occur before the numeral one and a classifier: (i) *wo chi-le (yi)-ge yi wan fan I eat-ASP a one CL/bowl rice A restriction on the legitimate use o f the indefinite determiner (yi-)ge seems to be that if the complement introduced by (yi-)ge is already nominal, it necessarily has to be a plural-collective XP. This may well be a primarily pragmatic restriction on the use of the determiner (yi-)ge. ( Yi-)ge is commonly used to introduce a complement which is unusual or sometimes excessive in some way as e.g. in (61a). Singular NP complements (such as for example yi wan fan) however normally do not represent unusual/excessive amounts and so the use of the (yi-)ge determiner is pragmatically unjustified. If a situation can however be created in which the consumption of a single unit of the N may in fact be interpreted as being excessive/unusual, the use o f (yi-)ge actually is permittable with a following y i + CL: (ii) Ta pingchang bu he jiu. Jintian tai gaoxing le. he regularly Neg drink alcohol tomorrow too happy LE Jingran zai wu-fenzhong nei he-le ge yi da-ping jiu. surprisingly at 5-minute within drink-LE GE one big-bottle wine 'He does not drink regularly. He was too happy today. Surprisingly he drank a big bottle o f wine in 5 minutes.' 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (62).23 However, when the same element has the interpretation of an indefinite article 'a/an' significantly it occurs after the classifier, as in (63a). The post-classifier position is also the position occupied by definite articles, as in (63b), confirming that post-classifier nung is in fact an article rather than a numeral: (62) a. phuu-ying nung-khon woman one CL 'one woman' b. phuu-ying soong-khon woman two CL 'two women' (63) a. phuu-ying khon-nung woman CL a 'a woman' b. phuu-ying khon-nii woman CL this 'this woman' The second set o f evidence supporting treatment of ge as a weak determiner giving rise to an interpretation of the DP it heads as a non-specific indefinite comes from the unacceptability o f ge-XPs in environments other than object position. The examples in (64) and (65) show that ge-XPs cannot occur in subject position and cannot be topicalized. This can be simply ascribed to the suggestion that ge, as a weak determiner with a non-specific indefinite interpretation, is barred from 23 See Simpson (2000) for further analysis o f the Thai patterns here. Contemporary Thai still in fact shows many interesting features of early Chinese due to Chinese influence on the Thai language when the Thai people still inhabited South-east China (up until the 13th Century). 68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. occurring in positions which only allow for either definite or specific indefinite DPs, hence neither subject nor topic position. Note that bare NPs/DPs may in fact occur in subject/topic position (with definite/generic interpretations); the fact that ge-XPs are fully unacceptable in subject/topic position can therefore only be related to the presence of ge and the proposal that ge as a weak determiner forces a particular non specific indefinite interpretation. (64) a. *ge wu tiao-wan-le. ge dance dance-finish-LE b. *ge san quan pao-wan-le. ge three lap run-finish-LE (65) a. *ge wu, ta tiao-le. ge dance he dance-LE b. *ge san quan, ta pao-le. ge three lap he run-LE c. *ge bao, ta chi-le. ge full he eat-LE d. *ge j i-quan-bu-ning, ta nao-Ie. ge chicken-dog-not-peaceful he fuss-LE Similar considerations account for the non-occurrence of ge-XPs in ba- and bei- constructions - in both such environments the ba-/bei-raised DP must be specific, but ge-XPs are necessarily non-specific. 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (66) a. *Ta bage wu tiao-le. he ba ge dance dance-LE b. *Ta ba ge san quan pao-le. he ba ge three lap run-LE c. *Ta ba ge bao chi-le. he ba ge full eat-LE d. *Ta bage ji-quan-bu-ning nao-le. he ba ge chicken-dog-not-peaceful fuss-LE (67) a. *ge wu bei ta tiao-le. ge dance bei he dance-LE b. *ge san quan bei ta pao-le. ge three lap bei he run-LE c. *ge bao bei ta chi-le. ge full bei he eat-LE d. *ge ji-quan-bu-ning bei ta nao-le. ge chicken-dog-not-peaceful bei he fuss-LE Finally, the weak determiner analysis of ge can be suggested to account for the non occurrence o f ge with demonstratives as in (68): (68) *Ta chi-le ge na san wan fan. (cf. (11a)) he eat-LE ge that three CL rice 'He ate those three bowls of rice at one time.' Examples such as (68) can be ruled out in either of two possible ways. Possibly it could be suggested that the demonstrative must be located in D and therefore if ge is also an instantiation of D they simply cannot occur. A second way to account for the unacceptability of (68) is to suggest that demonstratives are base-generated lower in 70 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the DP and not in D (as per Simpson 1997), but that there must be necessary definiteness concord/agreement obtaining between a lower demonstrative/quantifier and any instantiation of D, and as ge is specified as being indefinite, this will conflict with the [+definite] specification of the demonstrative, so resulting in the unacceptability of (68). Concluding this section, one might ask whether there is possibly any supporting precedent for the analysis of ge as an element of the type proposed. Essentially I have argued that ge is an unselective determiner which is able to embed a complement which may be o f a variety of syntactic types, from clearly nominal types through to adjectives and apparent clausal idioms. Whereas determiners (and functional elements in general) cross-linguistically are commonly assumed to be lexically specified for a single complement type, the proposal that ge as a nominalizing determiner should tolerate a variety of complement types might seem to be somewhat unusual. However, looking further across languages one finds that a proposal o f a highly similar type has in fact been made on the grounds of quite different evidence for patterning found in the Salish languages of North America. Davis and Matthewson (1997) argue at length and present convincing evidence that there are determiners in Salish which are unselective in terms of their complement category, optionally selecting either for a nominal or a verbal complement, an NP or a VP: 'In Lilloet Salish the lexical projections of N and V are morphologically and syntactically distinct. However, this distinction is neutralized at the DP/IP level. In (69) [numbering mine] the determiner /kwu/ indiscriminately takes either an NP or 71 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. an (infinitival small clause) VP as its complement:' (Davis & Mathewson 1997, handout p.3, numbering here is mine): (69) wa ka ama-s-as [kwu calis]/[kwu qwul-xal] Prog.appar. good-caus-3erg [Det cherry]/[Det cook-intr] 'He likes cherries/likes to cook.' Likewise the determiner [ti..a] introduces either an NP or a (finite small clause) VP: (70) ama [ti citx sw a] /[ti s xiq sw a] good [Det house 2sg.poss Det]/[Det finite arrive 2sg.poxss Det] 'Your house is good/Your arriving is good.' The existence of unselective determiners of different types might then actually be a phenomenon which exists across a number of languages and would not be linguistically odd in any way. Secondly, if one turns to another Chinese language - Cantonese - one can find further supporting evidence for the classifier-determiner connection. In Cantonese it has long been noted that bare classifiers occurring with nouns automatically give rise to readings of definiteness (see Cheng & Sybesma 1998 and Audrey Li 1998). (71) bo-sue hai bin do a? Cantonese Cl-book be where Q 'Where is the book?' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (72) goh-ging-chaat w aangohjii Iak Cantonese Cl-policeman say I know Asp 'The policeman told me (that).' Given that definiteness is standardly assumed to be a property of the D° position in DPs, if definiteness readings result from the use of a bare classifier in Cantonese, it may naturally be suggested that this is a function of the classifier occurring in D. Then a clear parallel is found with the ge-XP construction - in both instances there is the occurrence of a bare classifier in the D position and a particular type of definite/indefinite interpretation results. The difference between Mandarin and Cantonese here being that indefiniteness is a lexically-associated property of Mandarin event ge, whereas all classifiers in Cantonese trigger definiteness readings when hypothetically occurring in D°. Cross-linguistically then it would seem that there is in fact support for the analysis proposed here, and that the functional element ge arguably can be taken to occur high in the nominal functional structure as an unselective weak determiner, necessarily interpreted as being non-specific indefinite and functioning to provide a telic nominal bound to aspectually unbounded predicates.24 24 Earlier I have noted that there is a special restriction that indefinite determiner ge seems to occur only as part o f an aspectually-bounded predicate: ge-XP either follows an activity verb and provides the predicate with an end point or follows an achievement verb to reinforce the end point (the restriction addressed in examples 29-31). If one follows Borer (1994) and assumes that the object DPs which provide measures for telic predicates are licensed in the Spec o f an AspP dominating VP, it can now be suggested that the event ge is an instantiation of D° which will formally only be licensed when such an AspP is projected in the structure. Consequently, where stative predicates occur no such AspP will be projected and the indefinite determiner ge will not be licensed as an instantiation of D°: 73 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5 Grammaticalization, Re-analysis and G e In this last main section o f the chapter, I would finally like to approach the question o f how it is that ge might come to be an element in the Deposition. I would like to suggest that a plausible explanation can be offered in terms of an approach to grammaticalization initially suggested in Simpson (1998b). Such an approach can be shown to not only allow for a hypothetical account of ge, but will also be further developed in chapter 6 to account for changes which have arguably occurred with verbal -le. In Simpson (1998b) it is suggested that one possible way of understanding certain occurrences of grammaticalization is to see grammaticalization as frequently resulting from a combination of movement and re-analysis within the functional structure projected above a lexical element. It is proposed that various instances of 'category change' (i.e., re-analysis of an element of one category-type as instantiating a second category-type) may occur when a lexical element first raises up to some functional head dominating the lexical head and then later on becomes re-interpreted as actually being base-generated in the higher functional head position. For example, if a functional structure consisting of three functional heads F1-F3 is projected dominating a lexical head L as in (73), it is possible that F1-F3 may be filled by discrete overt functional elements which are actually base-generated in Fl- (i) *ta you ge san-ge haizi he have GE three-CL child (ii) *ta xihuan ge san-ge nu-ren he like GE three-CL women 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. F3, or it is possible that the lexical head L raises to these heads. In the latter case the lexical element may be taken to be base-generated with morphological features which are checked against the higher F-heads either overtly (as in N-to-D raising in DPs in Romance languages, see e.g., Longobardi 1994, Grosu 1988) or later on at LF. F3 F2 L Considering still the second case where L raises up to one or more of the functional heads, supposing the morphological features corresponding to these heads are covert/phonologically zero, it is possible that the lexical element itself may become interpreted as actually instantiating the features of the higher functional head(s). If the lexical element also undergoes a loss of descriptive content ('bleaching') while simultaneously becoming more associated with the functional properties of the F- head/s it raises to, it is argued that the lexical element may then ultimately undergo categorial re-analysis and be re-interpreted as base-generated in the functional super structure dominating L rather than originating in L. This will consequently result in a syntactic category change and grammaticalization of the lexical element L as a functional element o f type F i (or F2/3). 75 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As a concrete illustration of this hypothetical process, Simpson discusses the creation of the French negative morpheme pas. Originally negation in French was expressed just by a pre-verbal negative element ne. Later on this changed somewhat and the pre-posing o f certain verbal objects to a non-canonical object position preceding the verb came to signal emphatic negation. Among these emphatic re inforcers of negation was the element pas literally meaning 'a pace' or 'a step' and occurring naturally as the object of a variety o f verbs of motion such as 'walk,' 'run,' etc. The effect o f emphatic negation with pas was very similar to that in the English example (74): (74) I didn't walk a STEP! With other types o f verbs different appropriate objects would be used, amongst which those in (75) below (see Gamillscheg 1957): (75) mie 'crumb' - with verbs of eating gote 'drop' - with verbs of drinking point 'dot, point' - with verbs o f writing English again has certain similar uses, e.g., (76) I didn't drink/touch a DROP! From a wide range of objects used as emphatic re-inforcers of negation in early French, by the sixteenth century only pas and the three elements in (75) actually remained, these elements having undergone sufficient generalization that they could be used with a wide variety of verbs. Later on still in modem French only pas and 76 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. point are found, (without their original literal meaning) with point being more restricted in its use and only occurring to emphasize the contradiction o f a previous statement (see here Hopper and Traugott 1993). Essentially then pas over time lost its original purely literal meaning of 'step' and has come to be used as a functional emphatic re-inforcer of negation with verbs which have no connection at all with walking or running or actions involving 'steps' as potentially genuine objects. In modem day French the original emphasis present with pas and other emphatic re inforcers of negation has also significantly been lost and pas is now interpreted as signalling simple (non-emphatic) negation. Finally another highly important aspect of the process of historical change is that pas may now be used to express negation with transitive verbs which have overt objects. Whereas pas was originally understood to be the object of the verb, it has now been re-analyzed as a purely functional morpheme occurring in some other position. Because of this re-analysis, transitive verbs projecting other genuine object DPs may now co-occur negated with pas as in (77) :2 5 (77) Je n'ai pas vu Jean. I NEG-have PAS seen John 'I haven't seen John.' (Simpson 1998b) In view of this collection of synchronic and diachronic properties, Simpson suggests that pas was originally base-generated as a simple object NP which optionally 25 It can also be noted that in colloquial French the original negative marker ne is fast disappearing so that pas is taking over as the sole marker of negation. 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. underwent raising to a pre-verbal emphatic-focus-type projection selected by the negation head ne when focus-features were added on to pas, as illustrated in (78): (78) NegP Neg' N eg^ ^ ^ ^ ^ F ocP ne S p e c ^ ^ ^ '^ T o c VP pas F o < Spec N P (Object) With the continued association of pas with focus and negation and the specialization of pas over other emphatic negative objects Simpson suggests that pas was mentally re-analyzed as occurring base-generated in the Spec of the focus-projection. Such re-analysis then allowed for the object position to be occupied by a genuine non- emphatic NP and gave rise to forms with overt objects in addition to pas, as seen in (77) and the tree in (79): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (79) NegP Neg' Ne ne S^ec V' je V T I NP (Object) 'seen' vu Jean 'John' The possibility of explaining the clustering of properties and developmental changes with pas in this way thus arguably supports the suggestion that grammaticalization may indeed consist in a sequence o f movement and subsequent re-analysis in a functional projection targeted by a lexical element. Here pas is originally a lexical object which begins to undergo raising to the pre-verbal SpecFocP when it signals emphatic negation. Significantly later on after continued association of pas with negation and focus it is re-analyzed as being only a functional element and as therefore being base-generated in the functional projection itself. Such re-analysis frees up the object position in VP and a genuine thematic object can then be base- generated in this lexical position, leading to the co-occurrence of pas with overt lexical objects. Simpson suggests that another instantiation of this 'upward' process of grammaticalization is the frequent change of lexical verbs into auxiliary-verb elements, a development in which a lexical V° first raises up to a higher Infl-type 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from V° and then is later simply base-generated in the higher functional head. 2 6 In support o f the movement-based approach to grammaticalization, Simpson also notes that changes going in a 'downward' direction do not seem to take place. For example, whereas root modals which are canonically situated low down in the functional structure dominating VP commonly turn into structurally higher epistemic modals with sentential scope, the opposite type of change of epistemic modal into root modal is critically not found (see e.g., Bybee, Perkins & Pagliuca 1994). Significantly it would seem to be the case that change and re-analysis consistently occurs in an upwards direction, mirroring the direction o f syntactic movement. Turning back to Chinese and ge, I would like to suggest that a rather simple natural hypothesis can now be made about ge's occurrence in the D° position. It is fairly clear that ge must have originated as a classifier and would therefore have been originally base-generated in the classifier head position Cl°. Synchronically it has been suggested that ge has however a rather different function and occurs as an unselective determiner in D° embedding a variety of complement types as a nominalizer, functionally allowing these elements to occur as the nominal objects of a verb. Historically I would like to suggest that the current status of ge is a direct result o f ge undergoing a sequence of movement and re-analysis in the functional super-structure of the DP essentially very similar to that outlined for pas above. It can be suggested that ge was indeed originally base-generated in Cl° but then later underwent raising to the higher D° position and then ultimately became re-analyzed 26 Such a change and re-analyis is indeed assumed to have occurred in the 'creation' o f the 80 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. as an element of type D° directly base-generated in the D° head. Concerning the motivation for the hypothetical movement to D°, here a useful clear parallel can be drawn with the development o f English a/an. The English indefinite determiner a/an is well-documented as having developed historically from the numeral one (see e.g., Osawa 1998); if one assumes that numerals are base-generated in a distinct functional head Num° (as e.g., in Ritter 1991) and that indefinite determiners occur in D°, then a/an can be taken to have developed from one via raising from Num° to D° and eventual frill re-analysis in D°. It may be assumed that a/an/one underwent a change in its interpretation and functional role and that from a time when it was licensed as a numeral signaling the number one in contrast to other numbers it later came to be interpreted primarily as a simple indefinite element which 'embeds' an NP as its argument. As an indefinite counterpart to the it would therefore have raised to D° and possibly then have undergone re-analysis as a D° element. Now, Chinese ge is in ways quite similar to English a/an, functioning as an embedding element with a clear indefinite value like English a/an. Recall here also the fact noted earlier that when no numeral occurs with the simple classifier ge it has a natural default interpretation of 'one'; in such cases there is critically no feeling of contrast between the numerical value one and other possible numbers and ge is much more naturally translated as English a/an. Consequently it can be suggested that when ge was not accompanied by a numeral in the past, it came to be potentially licensed not just as a classifier associated with a covert numeral 'one' but also as an modal verb series in English (Lightfoot 1979). 81 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. indefinite embedding element higher in D°. Essentially then in both English and Chinese an element such as ge/a/an/one would be base-generated and remain in Cl°/Num° if its role was primarily numerical and contrastive with other possible number values, but raise to D° or be directly base-generated in D° if its primary function were to serve as an indefinite embedding element. The hypothetical raising of ge to the D° position can therefore be given a reasonable explanation which furthermore has cross-linguistic support. 2 7 Diachronically, I would therefore like to suggest the following sequence o f developments: (80) Stage 1: ge was base-generated and remained in Cl° as a classifier. Stage 2: ge was base-generated in Cl° and also raised to the D° position Stage 3: ge was fully re-analyzed as a D° and consequently base generated directly in the D° position Supposing one were to assume a simple DP structure with D° selecting for a NumP which in turn selects a C1P headed by a classifier, (81-83) would represent the hypothetical sequence of change: 2 8 27 See here Szabolsci (1994) and chapter 3 on the suggestion that elements in D° may be licensed in virtue o f being embedding elements similar to complementizers in the clausal domain. See also Longobardi (1994) and Vergnaud & Zubizarreta (1992) on the idea that the D° position sometimes must be filled, even if this is with an expletive determiner element which has no obvious definiteness value. Here I do not attempt to approach the question o f whether the Deposition in Chinese might always need to be filled, and if so how this is achieved. For interesting relevant discussion here see Audrey Li (1998). 28 There are many suggestions as to how D°, Num° and Cl° might be structurally related. I have adopted the structures in (81M83) for ease of explanation, but other structures might also be accommodated with the basic proposals put forward here. See Audrey Li (1999a) for further analysis o f the internal structure o f Chinese DPs. 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (81) Stage 1 0 NumP I Num' Num (yi)/er/san... '(one)/two/three' (82) Stage 2 DP I D' D° umP I m n' ▲ Num (yi 'one') C1P Cl' NP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (83) Stage 3 DP I D' Num 0 C1P I Cl' Q X ^ ^ N P I N° In (82) where no numeral (other than 'one') occurs in Num 0 ge raises here and is essentially interpreted as 'one,' and then moves higher to D° where it is licensed as an embedding indefinite determiner-type element. 2 9 In stage 3, ge is re-analyzed as being base-generated in D° and functions just as an embedding indefinite determiner. Significantly at this point when ge is not interpreted as having undergone raising from Cl° and Num0 it then becomes possible for these positions to become filled with other overt elements and sequences such as (84) are fully possible: (84) ta he-le ge san-ping-jiu he drink-LE GE 3-CL-wine 'He did a drinking of three bottles o f wine.' 29 Earlier it was noted that ge alternates with yi-ge as an indefinite determiner in D 0 and that ge and yi-ge are essentially equivalent (see footnote 21). There are two possible ways in which ge could have come to combine w ithy/ in D°. One possibility is that y i was located in Num0 in Stage 2 and allowed for ge first to raise up to it and then y i and ge together raised higher to D° as yi-ge. A second possibility is that yi was actually located in D° as an indefinite determiner in Stage 2 and ge raised up to adjoin to it in D°. 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The sequence in (81-83) then really has a close resemblance to the changes found with the French negation element pas. Like pas, after ge has raised from its base generated position to a higher functional projection and eventually undergone re analysis and full grammaticalization as instantiating only the higher functional position, its original base-generated site then once more becomes available for the lexical insertion of other elements. Somewhat different to the development with pas though, because ge has become an unselective determiner, once stage 3 is reached ge's complement is actually not forced to be a NumP but as noted in earlier sections can also be a variety o f other syntactic types. Ultimately then it has been possible to arrive at a plausible and reasoned explanation of how ge has grammaticalized and developed from a classifier into a functional element o f some higher type within the DP, thereby adding further good support and justification for its synchronic treatment as an element of type D° as indeed proposed here. 3 0 2.6 Concluding Remarks In this chapter I have shown that the general classifier ge can be employed to count out an event. Semantically this ge functions to make an atelic predicate telic by bringing out (or reinforcing) the end point o f the event depicted by the predicate. Syntactically event ge has been suggested to be a weak determiner in the D position 30 When the element yi occurs with event ge and has an interpretation as an indefinite determiner rather than a numeral, I would like to suggest that yi and ge are a single lexical head unit in D°, combined/fused in the lexicon as a single grammaticalized word/X°. 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. selecting for a variety o f complement types and converting them into a 'nominal' output which can then function to provide an aspectual bounding for an activity predicate. As a D°, ge is essentially equivalent to a non-specific indefinite article, this property also accounting for the fact that the DP headed by ge can only occur in object position in Chinese. Finally, I attempted to show how ge may have diachronically undergone change from being an instantiation of Cl° to become a D°, basically following a hypothetical path o f change upwards in the DP equivalent to the development of the English indefinite article 'a/an' from the numeral 'one.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3. Relative Clause Dei Directionality, Clausal Raising and Sentence-final Particles 3.0 Introduction This chapter examines the structure of relative clauses in Chinese and aims to develop and defend an analysis of relativization which will be made critical use of in the examination of shi-de cleft structures in Chinese in chapter 4. Section 1 of the chapter considers certain very general properties of relative clauses cross- linguistically and how these have been modeled in Chinese by standard Government and Binding style analyses. Section two then goes on to examine the problem of directionality and headedness with regard to relative clause structures in Chinese, highlighting the fact that Chinese is an exceptional language which combines prenominal relative clauses with a basic V-O word order. Concerning the ‘relativizing’ element de, it is argued that GB analyses of de as a complementizer are at odds with the general directionality of Category-selection in Chinese and that there is evidence that elements in C in Chinese do indeed select for rightward complements. In section 3.3 I then attempt to show how adopting a Kaynean approach to relativization may resolve these problems, following in part a discussion of relative clauses in Simpson (1997). This Kaynean analysis is subsequently compared with an alternative approach to prenominal relative clauses suggested in Murasugi (1998) for Japanese. I argue that the Kaynean account is preferable for Chinese for three essential reasons. First it allows for a more uniform account of 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. other noun-complement clause structures which make use of the element de. Secondly it is able to resolve the headedness/directionality issue (which is not a problem o f relative clauses in Japanese). Thirdly some of the critical features of relativization in Japanese which lead Murasugi to propose a fully base-generated analysis o f Japanese relative clauses are absent from relative clauses in Chinese. In section 3.4 I discuss the relevance of tone sandhi phenomena in Taiwanese for the account of relativization developed, indicating how it can be shown to provide interesting support for an analysis in which there are leftward clausal movement as part of the construction of relative clauses. Section 3.5 is an extension of the Taiwanese tone sandhi patterns discussed in section 3.4. The grammaticalization process of Taiwanese sentence-final "particle" -kong adds further support for the leftward clausal movement proposed for relative clauses. Section 3.6 is a brief conclusion. 3.1 Relative Clauses in Government and Binding Theory Relative clauses are standardly taken to be descriptions which semantically represent properties or sets, e.g., the relative clause: ‘who helped John on June 12th 1998’ linguistically represents the set of all (human) entities which have the property that they helped an individual named John on a certain day. Such elements are optionally combined with nouns which also represent sets (e.g., the bare noun ‘policeman’ being a property which characterizes all those human individuals with a particular profession) and in doing so simply function to pick out the intersection of the sets 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. represented by both noun and relative clause. Such a subset may then be combined with a determiner or other quantifier to yield a nominal phrase with particular reference, as in: ‘the policeman who helped John.' This assumption that relative clauses first combine with nominals and then with quantifier/determiner naturally leads to the view that relative clause structures have the form in (1) below. 1 Due to their optionality, relative clauses are syntactically taken to be adjuncts and so adjoin to NP before application of any element in D°: ( 1) N ’ who helped John N man From a syntactic point of view, within Government and Binding Theory it has been widely argued that relative clauses themselves are CPs, commonly with an operator element o f some type raised to the Specifier of C from a position within the IP 1 With non-restrictive relative clauses such as (i) below it is commonly assumed that the head-noun first combines with the determiner/demonstrative to form a DP and this DP is then joined up with the relative clause. This encodes the intuition that in non-restrictive relative clause structures the reference o f the DP is established independently of the relative clause, and the relative clause simply adds in a further property attributed to the DP (see Safir 1986 for further discussion of the interpretation o f non-restrictive relative clauses). Syntactically then, the relative clause is adjoined to a full DP rather than an NP: (i) [dp [dp that man] [R C who you m et last night]] 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. complement o f C, starting in Chomsky (1977). While this may often be an overt relative pronoun in many languages (e.g., in English), in the absence o f such an overt element Subjacency tests still provide good evidence for assuming the existence of null operators undergoing the same type of movement. A simple ‘that’-relative in English may therefore be represented as in (2): (2) DP I D’ D NP the NP CP N ’ Spec C’ that man John saw t; Turning to Chinese, relative clauses are found to precede the nominal element with which they are combined, the ‘linking’ element de occurs between the relative clause and the nominal and no relative pronoun is present. Quantifiers and demonstratives are optional: (3) [wo zuotian mai] de shu I yesterday buy DE book ‘the book(s) I bought yesterday’ The purpose of this chapter is to arrive at an analysis of how the various parts of relative clause structures in Chinese are syntactically related, i.e., how the noun head, 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the actual relative clause and the functional element de are formally combined with one another. If such structures do indeed have a similar underlying syntax to relative clauses in English and other languages (as e.g. represented in (1)), a first question which arises is whether there is any evidence of a null operator and null operator movement in Chinese. Although an answer to this question is not immediately obvious, Chiu (1993), (1995) presents interesting arguments that one can in fact conclude that at least a subset of relative clauses in Chinese must be formed by some kind of movement strategy and suggests a structure for Chinese relative clauses which in many ways is very close to that in ( 1 ), this also being largely endorsed by further argumentation in Ning (1993). 3.1.1 Chiu (1993/1995) and SuoP; Ning (1993) Chiu (1993/1995) discusses in particular the significance of the occurrence of the element suo in relative clauses such as (4): (4) [Lisi suo mai _ ] de shu Lisi SUO buy DE book ‘the books that Lisi bought.’ Chiu points out that there are various restrictions on the occurrence of suo (which is also always purely optional). First of all, sup may only occur if the relativized noun head corresponds to a gap which is in object position: 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (5) * [_ suo mai shu ] de ren SUO buy book DE person Secondly, sun appears only if there is a genuine gap in the relative clause, and may not occur if a resumptive pronoun is present (Chiu notes that resumptive pronouns in object position are not liked by all speakers, but for those speakers who do tolerate them as objects, suo is not at all possible): (6 ) [Lisi (*suo) renshi tai ] de [nei-ge renji Lisi SUO know him DE that-Cl person ‘the person that Lisi knows’ Chiu then notes certain critical interactions of suo with locality phenomena. When suo is present it is not possible for the gap corresponding to the head noun to occur either inside a Complex NP, a Sentential Subject or a w/z-island containing a wh- adjunct: (7) *[[Lisi suo k a n _ ]z u i heshi ] de shu Lisi SUO read most appropriate DE book intended: ‘the book that it is most appropriate for Lisi to read.’ (8 ) *[[ j suo zu _ k ]d eren i hen duo] de nei-dong fangzik SUO rent DE person very many DE that-cl house intended: ‘the house that the people who rented (it) are many’ (9) *[Lisi xiang-zhidao [Akiu weishenme suo mai-le _ ] de nei-ben shu Lisi wonder Akiu why SUO buy-Asp DE that-CL book intended: ‘the book that Lisi wonders why Akiu bought.’ 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. From this Chiu makes the natural conclusion that relativization does indeed involve movement of an empty operator in Chinese. The effects of this movement may at times be obscured because Chinese might also seem to allow for a resumptive pro element to occur in islands to overcome Subjacency violations (and the equivalents on (7)-(9) are indeed acceptable if suo does not occur). However, when suo appears, the relativization strategy does indeed show itself to be fully island-sensitive. Chiu gives arguments that suo may in fact only occur in structures where movement has taken place and instantiates an Accusative Case Projection (labeled SuoP) whose overt appearance is only triggered when an object has been raised through it to some higher position. In this sense suo is highly similar to the occurrence of object agreement in French which also only appears overtly when there has been some movement of the object (relativization, w/j-movement, passivization or cliticization, as discussed in Kayne 1975). Such an analysis is nicely supported by the fact that suo may not occur in the presence of an overt resumptive pronoun where movement has not taken place— example (6 ). Consequently it may be assumed that relativization in Chinese does indeed involve movement of some element and Chiu suggests that this element is an empty operator moving to SpecCP, just as in English that-relatives.2 In a second recent treatment of relativization in Chinese, Ning (1993) argues for similar conclusions to those in Chiu (1993/1995), but from a rather different set 2 Note that if an object is relativized but suo does not occur, it need not necessarily be concluded that there has been no movement. In French the triggering o f overt object agreement is purely optional even when there has been movement of the object. One can therefore suggest that Chinese may be similar to French and suo is optional when movement o f an object takes place. 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. o f observations. Ning again points out the fact that some cases o f relativization in Chinese might seem able to violate Subjacency/the CED, licitly linking a head-noun to a position within Sentential Subjects, Adjunct Islands, and CNPs. For example, (11) is an apparent example of subjacency violation that relativize from (10). Ning argues that ( 1 1 ) is acceptable only because the gap in ( 1 1 ) can be identified as a (resumptive) pro; therefore there is no movement involved and thus no subjacency violation present in ( 1 1 ). ( 1 0 ) [na-ge xuesheng kai zhe-zhong che] shuoming ta babayouqian. that-Cl student drive this-type-of car show his father rich 'That that student drives this kind of are shows that his father is rich.' ( 1 1 ) [[t kai zhe-zhong che] shuoming ta baba youqia] de na-ge xuesheng t drive this-type-of car show his father rich DE that-CL student '*the student who that (he) drives this type of car shows that his father is rich' Ning then shows that certain gaps may however not be identified as (resumptive) pro elements as this would violate Huang's (1982) Generalized Control Rule (GCR) on the identification and control of pro. This type of gaps may not in fact be located inside any island configuration, as for example in (12). The conclusion Ning draws is that gaps of relativization not licensed as pro by the GCR must indeed result from movement and therefore again that at least a substantial sub-part o f relative clauses in Chinese are formed by movement. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ( 1 2 ) * [[na-ge xuesheng kai t] shuoming ta babayouqian] de na-bu che that-CL student drive t show his father rich DE that-CL car '*the car which that the student drives (it) shows that his father is rich' Ning (1993) also considers the phenomenon of so-called 'gapless' relative clauses in Chinese discussed in Tsai (1992) and elsewhere. Ning argues contra Tsai that Chinese does not in fact have relative clauses which do not contain a gap (either a gap o f movement or a trace resulting from operator-movement), and provides arguments that all relative clauses in Chinese do contain some operator-variable chain. Ning suggests that structures such as (13)-(16) below result from the movement of a null wh-operator which may correspond to one of four universally available values— p/ace, time, manner and reason: (13) [ta xiu che]-de cheku he repair car-DE garage 'the garage where he fixed the car' (14) [ta xiu che]-de wanshang he repair car-DE evening 'the evening when he fixed the car' (15) [ta xiu che]-de fangfa he repair car-DE method 'the method how he fixed the car' (16) [ta xiu che]-de yuanyin he repair car-DE reason 'the reason why he fixed the car' 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ning goes on to show that relations between a head-noun and a clause which do not fall within any o f these four basic types may not license a relative clause structure, for example, neither comitative nor source interpretations o f the head-noun are possible in (17) or (18): (17) *[ta tiaowu]-de guniang he dance DE girl intended: 'the girl he danced with' (18) *[ta lai]-de na-ge xiaozhen he come-DE that-cl town intended: 'the town he came from' There are therefore good reasons for believing that relative clauses in Chinese are licensed by the coindexation of the head-noun with some position in the relative clause, such dependencies arising either via movement or otherwise via the linking of a base-generated pro, and that Chinese relative clauses are not in fact simply licensed by any 'Aboutness Relation' (as e.g. in Kuno 1973). 3.1.2 DE as a Complementizer The next question which needs to be answered is: what might be the status of the element de in relative clause structures? It can be noted that de is certainly obligatory in relative clauses and in this sense might seem to be similar to the occurrence of the element that in English relative clauses where the subject position is relativized and no relative pronoun occurs: 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19) [wo zuotian mai] *(de) shu hen gui I yesterday buy DE book very expensive ‘the book I bought yesterday is very expensive’ (2 0 ) the man *(that) _ helped me In Chiu's (1993/1995) and Ning's (1993) accounts, the relative clause in Chinese is analyzed as a CP with a null operator (often) moving to the Spec position of CP, in which de occurs as a relativizer in the C position, just as that does in English. Such a proposal means that C must naturally be assumed to select its IP-clausal complement to the left, an assumption which has been justified in the literature on the basis of the occurrence of question particles in clause-final position— if one takes such particles to instantiate a Q-morpheme located in C, then C must be clause-final and IP to its left in Chinese: (21) [ [Ni qu-guo Beijing] m a ]? you go-ASP Beijing Q ‘Have you ever been to Beijing?’ Putting all the components of relative clause forms together and making the assumption that Specifiers uniformly branch leftwards in Chinese, one arrives at the structure in (22) representing (3), which is the analysis put forward in Ning (1993) and in somewhat more articulated form also in Chiu (Chiu has NomP in her structure in place of the IP here) : 3 3 See also Audrey Li (1999b) for similar assumptions about the structure of relative clauses in Chinese. 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. shu 'book1 C de Such a structure is very much a direct equivalent o f the standard GB-type analysis of English relative clauses in (2) and has by and large been the regular approach taken to relative clause formation in Chinese for a number of years. 4 In what follows, however, I would like to argue that there are reasons to prefer another type of analysis for relative clause structures in Chinese, one which is based on more recent general proposals concerning phrasal projection made in Kayne (1994). I will attempt to show both that there is empirical evidence in favor of a Kayne-based approach and that such an approach should also be favored for reasons of theoretical simplicity. 3.2 Directionality and C-selection— a Theoretical Problem for Standard Analyses In this section I would like to suggest that a reason to be somewhat ‘suspicious’ of and disprefer a structure such as (22) is that the analysis o f de as a C selecting an IP complement to its left might seem to go against the general directionality o f selection 98 wo zuotian mai tj 'I yesterday bought,^' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in Chinese. Although there has been considerable debate about the head-parameter setting in Chinese and some linguists (notably Li & Thompson 1974) have argued that Chinese displays the beginnings of an O-V order (in ba-construction sentences), the present general consensus of typologists, historical linguists and generativists would appear to be that Chinese neither was a head-final language in the past nor does it show real signs o f changing from a basic head-initial ordering at present (see e.g. Mulder & Sybesma 1992, Sun 1996, Peyraube 1996).5 Both in the lexical domain and in the functional domain one finds that X° heads select their complements in a rightward direction, so for example verbs and prepositions consistently take their objects to the right as in (23) and (24), and auxiliaries which can be taken to occur in 1° similarly have rightward VP complements, as seen in (25): (23) [vp [v mai [mp yi-suo-fangzi]]] buy one-cl-house ‘buy a house’ (24) [P P [P zai [N p wo-jia]]] at my home ‘at my house’ 4 Strictly speaking Ning (1993) is written from an early Minimalist rather than a GB perspective. By including it here I simply mean to contrast it with a Kaynean-style analysis. 5 Though see Audrey Li (1990) and Travis (1984) for interesting discussion of alternate views. 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) [ipta [jhui [vpqu Beijing]]] he will go Beijing ‘He will go to Beijing.’ If one attempts to suggest that the relative clause element de is occurring in a C position to the right o f its complement IP, the directionality of this selection relation would seem to be opposite from what is otherwise attested in Chinese. One piece o f evidence given in support o f the assumption that C is final in Chinese was however the occurrence of question particles such as ma/a in clause- final position, as in (26). It has been suggested that such particles occur in C and that C hence follows IP to its right. Here I would like to argue that this argumentation is not particularly strong for a number of reasons, and that it is more likely that C is initial in Chinese selecting its complement IP in a canonical rightward direction. (26) ni qu ma/a? you go Q 'Are you going?' The first argument against an analysis o f question particles such as ma occurring in C is that similar elements occur in many other languages which are clearly head-initial and which have complementizers occurring to the left of their complement IPs. Two such languages which can be mentioned here are English and Thai (p.c., Andrew Simpson), both very regular SVO head-initial languages. In both English and Thai it is found that question particles occur sentence-finally as illustrated in (27)-(30): 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) You’re going home, eh? (28) She left already, right? (29) wan-nii khun ca hen khaw mai? (Thai) today you will see him Q ‘Will you see him today?’ (30) khaw ca maa meuarai le? (Thai) he will come when Q ‘When is he going to come?’ The occurrence o f these question particles in sentence-final position would seem to be very similar to that in Chinese, yet in English and Thai it is not possible to argue that the particles occur in C° as C° is IP-initial in both languages with clear complementizers occurring in positions preceding IP: (31) John said that [ip Mary left] (32) Daeng book waa [ip Dam mai maa] (Thai) Daeng say C Dam not come ‘Daeng said that Dam didn’t come.’ It therefore seems that one needs to recognize that sentence-final question particles may regularly occur in surface positions which do not correspond to regular complementizer positions and that the position of ma in Chinese does not necessarily indicate the locus o f C in Chinese. A second reason to reject the idea that ma is located in C comes from the observation that ma (and other questions particles such as a) show an important 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. difference from question particles such as ka in Japanese which might be more reasonably located in C .6 Whereas ka productively occurs in all kinds of embedded interrogative clauses such as (33), Chinese ma/a in (34), just like English right/eh in (35) and and Thai particles mai/le in (36), are restricted to occurring only in root/main clauses. If ma/a are taken to be simple instantiations of C, there is no obvious reason why they should not also occur in embedded questions, as embedded clauses obviously do also contain a complementizer/C position. The root/main clause restriction on ma/a then suggests that ma/a is perhaps not in C (but perhaps in a higher matrix head) and therefore that the position of ma/a is not a good indication that C is final in Chinese: (33) Watashi-wa [dare-ga kuru ka] shirimasen I-Top who-Nom come Q not-know ‘I don’t know who is coming.’ 6 It is actually quite possible that ka is not in C in Japanese either, as ka may co-occur with the quotational complementizer to as in (i): (i) Mary-wa [John-ga itsu kuru ka]-to kikimashita (p.c., Yuka Kumagai) Mary-Top John-Nom when come Q C asked ‘Mary asked when John was coming.’ Possibly the standard complementizer domain is split into a number o f heads with quotational and interrogative complementizers occurring in discrete heads in certain languages. Generally when languages allow for multiple instantiations of C in this way however, such functional heads all occur together on one particular side of the clause and hence show a single directionality relation in their c- selection, i.e., all would seem to be heads selecting their complements in the same direction, and one does not seem to find languages in which a quotative complementizer occurs on one side of a clause and an interrogative complementizer on the other. Dutch below is another example of a language with an overtly split-C system: (ii) Ik weet niet of dat hij komt. (p.c., Andrew Simpson) I know not Q C he comes ‘I don’t know whether he is coming.’ 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) *Ta xiang-zhidao [ni qu ma/a] he wonders you go Q intended: ‘He wonders whether you are going.’ (35) *John wonders [you’re going right/eh]. (36) a. *Daeng yaak ruu (waa) [Dam chalaat mai] Daeng want know C Dam clever Q intended: 'Daeng wonders if7whether Dam is clever.' b. *Daeng thaam waa [khrai maa le] Daeng ask C who come Q intended: 'Daeng asked who came.' Original suggestions that ma/a might be in C were made because Chinese seems to lack any obvious equivalent to the Indo-European general-purpose embedding complementizers (such as English that) which might disambiguate where C is located. However, it is not in fact true that Chinese lacks any likely candidate for C. Elements such as if and whether and their counterparts in other languages are co mm only taken to be complementizers located in C, and equivalent elements do exist in Chinese, e.g., ruguo/vaoshi ‘if,' etc. Significantly, such elements are found to occur clause-initially and so would seem to indicate that C in Chinese does indeed select its complement IP in a regular rightward direction: (37) [c p Ruguo/yaoshi [ip ni bu xihuan nei-ge-ren]], ni... if you not like that-cl-person you ‘If you don’t like that person, you..’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Finally, Hwang (1998) notes that the verb to 'say' in Mandarin, Taiwanese and Cantonese is currently following a typical pattern of grammaticalization and becoming reanalyzed as a regular complementizer in certain environments, e.g., following cognitive verbs such as 'think' or informative verbs such as 'tell.' In such instances it is found that the newly grammaticalized complementizers linearly precede their sentential complements, strongly suggesting that the position o f C in Chinese is in fact IP-initial rather than IP-final.7 (38) contains examples of Mandarin shuo 'say' as a complementizer; the same verb in Taiwanese and Cantonese behaves the same. (38) a. taxiangshuo tab u lai le. he think say he not come Asp 'He thinks that he is no longer coming.' b. ta gaosu wo shuo ta bu lai le. he tell me say he not come Asp 'He told me that he is no longer coming.' As a result of the above argumentation, the analysis o f relative clause de as occurring in a C position with a leftward complement seems rather unlikely. All the evidence available suggests that C in Chinese selects for a rightward IP complement, in line with the general direction o f selection in the language. Therefore it might 7 The verb 'say' can also follow a subordinator, which itself can be considered as a C. For example, in (i), the Mandarin verb 'say' is grammaticalized as a complementizer which follows another complementizer 'if.' This can be said to be an example o f multiple instantiations o f C, which is discussed in footnote 5. All instantiations o f C occur together on one side o f the clause and hence show a single directionality relation in their selection. 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. seem that a different account o f the positioning (and possibly syntactic category type) o f de is called for, which may well have further consequences for the general analysis o f internal structure of relative clauses in Chinese. This I now turn to in section 3.3.8 (i) ruguoshuoni bu xiang qu, women jiu bie qu le. if say you not want go we then don't go Asp 'If (that) you don’ t want to go, then let’ s not go.' * There remains the question o f where sentence-final question particles such as ma/a actually occur. If there is independent evidence from the positioning of elements such as ruguo/vaoshi that C is initial in Chinese, and abstracting away from the problem noted that ma/a are restricted to matrix clauses, one possibility might be that the rightward IP complement to a sentence-initial C° occupied by ma/a is raised (leftwards) to SpecCP in question forms giving rise to the ma/a-final surface word- order observed, as schematized in (i): (i) [C P [ ni qu Beijing]; [c ma t, ]] you go Beijing Q ‘Are you going to Beijing?’ An analysis along these lines has been suggested by Dominique Sportiche (1996 class lectures) for French and English intonation questions and accounts for certain facts concerning the licensing o f NPIs. For example, in English it is possible to form yes/no questions both by means of subject- auxiliary inversion (SAI), and with a particular type of sentence-final intonation (a rising pattern): (ii) Did you see John? subject-auxiliary inversion (SAI) (iii) You saw John? rising intonation only However, whereas SAI questions license NPIs, intonation-questions do not: (iv) Did you see anyone? (v) *You saw anyone? Sportiche makes the standard assumption that an NPI is licensed if the Q-morpheme in C is able to c- command the NPI, and this occurs unproblematically in SAI questions such as (iv). Sportiche suggests that a possible explanation why NPI licensing fails in intonation questions is that the entire IP actually undergoes raising to SpecCP, so destroying the c-command relation between the Q- morpheme in C and the NPI in SpecCP. In section 5 of this chapter I return to the general issue of sentence-final particles and suggest that there is actually good evidence for an IP-raising analysis with the Taiwanese S-final particle kong, and therefore that such an analysis might indeed be appropriate for other similar particles in Chinese too. In section 5 it is suggested that generally the surface position of phonetically reduced particle-like elements is not necessarily a reliable indication of any underlying head-complement order and the fret that such elements are phonologically dependent and enclitic may result in significant surface distortions o f underlying linear order. 105 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3 Kayne (1994) and a Uniform Theory o f Relativization Kayne (1994) develops a theory of phrase structure projection, the Linear Correspondence Axiom, which attempts to encode the idea that there can be no ambiguity in the way that terminal/head elements are linearly sequenced. One consequence of the LCA is that rightward adjunction structures may no longer be posited in any syntactic structures. Standard GB-style relative clause analyses in which CPs are taken to be right-adjoined to NPs as in English (1) repeated below therefore have to be re-analyzed by Kayne: ( 1) who helped John man Kayne’s solution to the problem created for relative clauses created by the LCA involves two key components. First of all it is suggested that D selects directly for a CP complement. Secondly it is suggested that the relative clause noun-head actually Note also though that Chinese S-final Q-particles do in fact license NPIs, which might seem to be puzzling given that the hypothetical IP-raising in (v) does not. I believe that an explanation here may relate to the derivational point at which IP-raising occurs. With Taiwanese kong there are good arguments that the IP-raising occurs at PF. If this is the case, it might be expected that such raising would not affect any licensing o f NPIs in the IP. Possibly the IP-raising assumed by Sportiche in English and French may take place not in PF but during the syntactic derivation and for this reason block the licensing of NPIs in these languages. 106 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. undergoes movement to its surface position from some position within the IP unit. The latter suggestion is in part justified by the observation that it allows for idiom chunks to be base-generated together and then split by the process o f relativization, and is a return to a style o f analysis originally put forward for similar reasons by Vergnaud 1985. In forms such as (39) below, an analysis in which the NP headway is base-generated as the object of the verb make and then raised leftwards is fully consistent with the general assumption that complex idioms must be stored in the lexicon as a single unit and introduced into syntactic structure together. Alternative null operator-type analyses of similar examples have been criticized as failing to capture the well-formedness of such idiomatic cases; it is argued that a word such as headway is listed in the lexicon only as part of the idiom make headway and not as an independent noun which could be taken and directly inserted as the head-noun of a relative-clause structure: (39) the [headway]i that we made ti on the project A simple that-relative according to Kayne is formed via movement o f the NP to the SpecCP position as diagrammed in (40): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. [bookji C IP (that) I bought tj The same essential head-initial structure is then also suggested to underlie pre nominal relative clauses in other languages, with the addition of a single extra movement raising the IP constituent to the Specifier of D as in (41): If one were to allow movement to be able to affect C’ as a constituent (although generally it is argued that X ’-level constituents are invisible for the purposes of movement), then one might also expect languages to show the surface order indicated in (42), and this might in fact seem to be what is found in Amharic (p.c., R. Hayward, using English words for ease of exposition): (41) DP (using English words for ease of exposition) I bought tj 1 0 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (42) rrthat n bought t, ] ] k [ d the [ C p bookj [ tk ]]]] Such an analysis essentially allows Kayne to suggest that there is a fully universal structure underlying relative clause forms in all languages, the surface differences attested being simply due to whether (and how much of) the relative clause is moved to SpecDP overtly. Now, concentrating on the possible positions that complementizers can occur in in languages with prenominal relative clauses, it can be seen that they may be found either finally after the head-noun as in (41) after IP-movement, or initially as in (42), and so schematically as in (43): (43) (C) IP D NP (C) (44) (that) [I bought] the [book] (that) If one now takes the pre-nominal relative clauses found in Chinese, e.g., (45), and compares them with the template in (43), paying particular attention to the linear positioning of the element de and the possible positions of complementizers according to Kayne, an interesting discovery is made. It is found that the position which de occurs in does not correspond to any potential complementizer position. Such a conclusion is highly interesting because it is exactly what has been argued for in section 3.2. There it was suggested that it would be quite unnatural to analyze de as a complementizer in C and that some other analysis seemed to be required. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (^5) [ ip w o zuotian mai] de [np shu] I yesterday buy DE book ‘the book I bought yesterday’ In fact, as pointed out in Simpson (1997), it seems that the only likely category that de could conceivably correspond to in such structures is actually D, which is initially somewhat surprising as de would not seem to exhibit the standard patterning of determiner elements, neither having any obvious inherent definiteness value nor co occurring with NPs outside of relative clause and possessor modification environments. However, rather than immediately reject the possibility that de may be taken to be a determiner in D, one might instead reconsider what is understood by such a term and re-examine how the notion o f definiteness may be represented in a DP. This is indeed the path taken in Simpson (1997) and shortly I consider a number of the points raised in that work. Before doing so, it may be helpful to indicate how relative clauses might look in a Kaynean-style approach under the assumption that de is in fact located in D. Examples (46) and (47) represent the sequence of movements necessary to convert a fully head-initial structure into the surface string attested in (45). The first step in (46) is the raising o f the NP which first suggested in Vergnaud (1985) and then suggested in Kayne (1994) to take place in relativization in all languages, while (47) is the further IP-raising absent from languages such as English but present in Amharic: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 'book j1 wo zuotian mai tj 'I yesterday bought tj' (47) wo zuotian mai tj 'I yesterday bought tj' Spec 3.3.1 Determiners, Demonstratives and Definiteness Agreement Referring to work carried out by Greenberg (1978) on African languages, Hopper & Traugott (1993) and Abraham (1997) among others, Simpson (1997,1998a, 1998b) notes that it is a fairly common process for functional elements such as determiners to pass through a number o f stages o f grammaticalization. Generally it is documented that determiners are initially derived from demonstratives after a loss of the deictic force present in such elements. For example, the French determiners le/la have been grammaticalized from the Latin demonstratives ille/illa, German der from ill Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. dieser and English the from the this/that demonstrative series. Greenberg (1978) writes that after further time determiners may then also go into a process of further decay, ultimately losing their definiteness specification with the end result that they either disappear from a language or else remain on with some additional function. As determiners undergo bleaching of their definitess specification, it is common that languages develop some compensatory mechanism to encode the definiteness of a DP. Certain languages may make use o f word order possibilities, while others may employ means such as aspectual marking on the verb to distinguish definite from indefinite DPs (e.g., Abraham 1997 on the development of determiners in German and the use of aspectual systems in Slavic languages). A third common way to compensate for the loss of definiteness specification in determiners is for a language to develop a new series of demonstrative elements in adjective-like positions within the DP. Such demonstratives are then frequently found to co-occur with determiners in a single DP representing a classic example of the typological notion of ‘layering.1 Simpson suggests that this is indeed the case of Chinese de. De is argued to be a determiner which no longer has any intrinsic or overt definiteness value and which consequently may often occur ‘doubled’ by a demonstrative (or other quantifier) specifying the definiteness o f a DP: (48) [wo zuotian mai]-de nei-ben-shu I yesterday buy DE that-cl-book ‘that book which I bought yesterday’ 112 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In more theoretical terms, Simpson suggests that the overt locus of definiteness in a DP need not always be the D position. Following the original ideas in Vergnaud & Zubizarreta (1992) developed also in Szabolsci (1994) and Longobardi (1994), it is noted that many languages would seem to have an EPP-like condition on the D position that it be overtly filled by some element prior to Spell- Out. The element filling the D position need not necessarily have any definiteness specification and could be an ‘expletive’ determiner with no definiteness value; the EPP-like condition on D° is then similar to the clausal EPP which requires that SpecTP be filled, even if only with a semantically empty expletive element. In languages such as English if a demonstrative is generated in the DP, this will have to occur in D. However, in other languages it seems possible to find either a demonstrative in the initial D-position or a determiner in D with a demonstrative occurring lower down in the DP. Such a pattern is found in Spanish, Irish, Greek and a number o f other languages. Example (49) illustrates this alternation with data from Spanish taken from Giusti (1997): (49) a. ]a reaccion alemana esa a las criticas the reaction German that to the criticisms ‘that/the German reaction to the criticisms’ b. esa reaccion alemana a las criticas that reaction German to the criticisms ‘that/the German reaction to the criticisms’ Data such as (49a/b) allow one to draw a number of conclusions. First it is argued in Grosu (1988), Szabolsci (1994), Giusti (1997) and various other works that 113 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. demonstratives are not in fact base-generated in D° but actually in lower DP-intemal adjective-like positions. In many languages they are suggested to undergo raising to an initial position and so appear as if they had been directly inserted into D. Other languages however show that their demonstratives will only raise to the initial position i f no determiner has been inserted there. In Spanish for example, if no determiner is present in the initial position in (49a) the demonstrative is forced to raise to the initial position as in (49b) and it may not remain lower down: (50) *reaccion alemana esa a las criticas reaction German that to the criticisms This has led to the view mentioned above that determiners are often ‘place-holders’ or expletives, and that the definiteness specification of a DP may be primarily encoded by other elements such as demonstratives. Simpson (1997) points out however that it is not sufficient for just any type of D-compatible element to occur in D in the presence of a lower demonstrative, and that generally only the definite determiner is possible, as in the contrast seen in (51): (51) el/*un hombreeste the/a man this This is attributed to the existence of a requirement of ‘definiteness agreement’ or ‘definiteness concord’ between the D-position and lower elements such as demonstratives or other quantifiers. 114 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Consequently it is argued that it is in fact rather natural to find that determiners co-occur with demonstratives within a single DP. Historically this is suggested to be due to a process in which determiners originally grammaticalized from demonstratives may gradually be bleached of an original definiteness value and then tolerate doubling by a secondary set of demonstrative elements which are the primary encoders of definiteness in the DP.9 In languages such as Spanish, Greek and Romanian, the plain determiners may still maintain a definiteness value, though frequently also functioning as expletive place-holders. Chinese, by way of contrast Simpson suggests, is a language in which a determiner element occurs in D which has lost all trace of any definiteness specification, with the result that definiteness concord between the D-position and a lower demonstrative is only covertly encoded. Such a mode of explanation allows one to make certain sense of the idea that de may be an instantiation of D without necessarily having any obvious definiteness value. It therefore provides a way o f reconciling the distribution of elements within the relative clause in Chinese with the structure predicted by a Kaynean approach. Returning to the forms suggested by Kayne for typical prenominal relatives (as in (46) and (47)), one now might ask what motivation could be given for the suggested IP- raising operation shown in (47), repeated immediately below: 9 For discussion o f how demonstratives develop into determiners, see for example Vincent (1997). Vincent notes that demonstratives may also develop into pronouns. Latin ille for example, developed in two ways in modem French; the second syllable le o f ille became modem French masculine singular "the" and the first syllable il became masculine singular "he." 115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. wo zuotian mai tj D CP 'I yesterday bought tj' | Spec 'book j' Pointing out that Chinese is possibly the only language known to exhibit the combination ofpre-nominal relative clauses and a basic head-initial/V-0 order (as in Dryer 1992), Simpson suggests that any explanation o f the IP-raising should be somewhat exceptional in nature and therefore necessarily related to a particular lexical item rather than being a general syntactic property (such as a parameter setting). Simpson proposes that the exceptionality present in Chinese relative clauses relates directly to the element de and suggests that de is an enclitic determiner similar in nature to the enclitic definite determiners found in Romanian, Mokilese and Buginese, attracting an element (in Chinese the IP) to its Specifier position for phonological support. Such a proposal takes its lead from Grosu’s (1988) analysis of the definite determiner in Romanian. Grosu convincingly shows that the definite determiner is an enclitic which attracts an element to the DP-initial position preceding the determiner in order to attach to. The element attracted may be a noun or an adjective if present, as in (52) and (53). Cases such as (52) and (53) with the definite determiner contrast strongly with the use of the indefinite article in 116 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54) which is not an enclitic and does not result in the attraction of either adjectives or nouns: (52) potretj-ul tj unei fete portrait-the a-Gen girl ‘the portrait o f a girl’ (53) frumosj-ul tj baiat nice - the boy ‘the nice boy’ (54) un portret al fete a portrait of-the girl ‘a portrait o f the girl’ So for Chinese the suggestion in Simpson (1997/1998a) which I would also now like to assume, is that de is an enclitic determiner very similar to ul in Romanian, and that de attracts the EP-clause to SpecDP to attach to it, thus giving rise to the surface distortion and the typo logically very unusual prenominal ordering.1 0 As additional support for the cfe-as-determiner analysis, Simpson notes that there are a number o f languages which visibly make use of determiners in the formation of their relative clauses, among them Lhakota, Diegueno, Tzeltal and Hebrew, a cross-linguistic patterning which indicates that it is indeed (relatively) common to find determiners in such positions. 1 0 Following Simpson (1998a), it can be suggested that de is a D° which only selects CP complements, and therefore can not occur simply with an NP as in *de NP/*NP de. In the case o f possessive forms such as Zhangsan de 'Zhangsan's,' Simpson suggests such sequences are not NP-de, but arise from relativization and the raising to Spec DP o f an IP containing a null predicate of possession. 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Finally I would like to point out that there is potentially strong historical support for the contention that de is indeed a determiner element in the D° position. Earlier it was observed that determiners frequently develop from demonstratives when the latter lose their strong deictic force. Given such a process it is now significant to note that there is in fact a direct connection between modem Mandarin de and an earlier demonstrative element. Commonly it is assumed that the modem Mandarin morpheme de developed from an earlier element zlii. which in classical Chinese had a distribution largely parallel to contemporary de (see for example Pulleyblank 1995). Importantly in addition to contexts such as relative clauses in which modem Mandarin de occurs, classical Chinese zhi was also used as a clear demonstrative (example cited from Pulleyblank 1995):1 1 (55) zhi er chong you he zhi these two worm again what know ‘And what do these two worms know?’ (Zhuangzi 1.10) If modem Mandarin de has indeed developed from classical Chinese zhi. and if classical Chinese zhi was a demonstrative, a highly natural analysis of modem Mandarin de is that it is a determiner descended from this earlier demonstrative via the common process o f “definiteness bleaching." The end result of this is that an analysis based on a Kaynean model coupled with an account of de as a bleached enclitic determiner might indeed seem able to 1 1 Pulleyblank notes that zhi could also occur as a pronoun in object positions. As the occurrence o f zhi in (55) is clearly in subject position, Pulleyblank takes zhi to be a demonstrative here, not a pronoun. 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. accommodate the typological oddity o f relative clauses in Chinese in a principled and satisfactory way which is also successful in avoiding the directionality problem faced by earlier GB-style approaches. 3.3.2 Further Evidence fo r the NP-raising Analysis: (1) Language Acquisition: Chiu (1998); (2) Connectivity and Idiom-chunks In addition to providing an analysis of Chinese relative clauses which is in accord with the general direction o f selection dominant in Chinese, there is also interesting language acquisition evidence discussed in Chiu (1998) which supports Vergnaud's (1985) idea (adopted by Kayne 1994) that it is the N-head which undergoes raising rather than an empty operator. Chiu observes that children produce not only relative clauses with regular resumptive pronouns but also relatives with full NPs in place of an extraction-gap (Resumptive NP Relative Clauses--RNP-RCs), as in (56): (56) luotuo chi caomei de caomei camel eat strawberry DE strawberry 'the strawberry which the camel ate' Following Guasti and Shlonsky (1992) in essence, Chiu suggests that the use of an empty 'linking' operator in relativization is a complex strategy which only becomes available to children at a later stage of development. Chiu suggests that children therefore initially form relative clauses via simple movement of an NP from a base generated position within the relative clause (via SpecCP) to a surface position right- adjoined to the relative clause. RNP-RCS such as (56) occur when children 119 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. optionally spell-out both the lower extraction-site copy of the NP as well as the head o f the chain, instead o f just the latter: Because linking operators are maturationally unavailable for children before the age of 6 (cf. Guasti and Shlonsky 1992), children begin forming relatives by directly moving the head noun from within the CP to its surface position, leaving an exact copy at the extraction-site. (Chiu 1998, p. 18) In addition to forms such as (56) and 'regular' relative clause structures such as (57) in which only the head of the chain is phonetically spelled-out, children also produce sequences such as (58). Chiu suggests forms of this Tiead-intemal' type are the result of spelling-out just the lower copy of the relativization chain: (57) Iuotuo chi _ de caomei camel eat DE strawberry 'the strawberry which the camel ate' (58) ganggang nei-ge daxiang tiao de just-now that-Cl elephant jump de 'the elephant which just jumped' Consequently such patterns would indeed seem to provide further support for an N(P)-raising account of relativization and are rather difficult to account for in analyses in which the head-noun is simply base-generated in its surface position following de. If the head-noun were to be base-generated following de and commonly associated with a gap position via an operator-trace/pro linking, one would expect for the argument position in the relative clause to be occupied by either just a trace, a phonologically null pro, or possibly an overt resumptive pronoun, but Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. certainly not a copy o f the head-noun as in (56). The only explanation for the occurrence an overt copy o f the head-noun inside the relative clause would seem to be that this is left by movement of the head-noun itself, not any operator.1 2 1 3 In addition to the above language acquisition evidence for an NP-raising analysis of Chinese relative clauses, there is further evidence from connectivity and 1 2 Possibly one might attempt to suggest that forms such as (56) relate to English non movement structures such as (i) below: (i) ‘the strawberries such that the camel ate the strawberries’ However, such an approach could not be extended to cover cases such as (58) where there is only an NP inside the relative clause— critically there must be an external head for ‘such that’ clauses to be possible. The example in (ii) is an attempted ‘such that’-relative clause without an external head, and is quite unacceptable, even if one adds a pronominal element such as ‘those’: (ii) ??those such that the camel ate the strawberries To the extent that a ‘such that’-relative approach to (56) cannot extend to cover related cases such as (58), a unified Kaynean/Vergnaud-type treatment of all (56-58) suggesting movement o f the head- noun (and some optional deletion) is clearly to be preferred. 1 3 Note that Chiu is o f the opinion that the N(P)-raising strategy suggested to occur in children's language is actually later replaced by empty operator movement in adult Chinese. Presumably Chiu assumes this because forms such as (56) and (58) are not found in adult speech. However, one might alternatively suggest that it is the Spell-Out option which changes in adult speech rather than the fundamental relativization process, i.e., adults would continue to form relative clauses via direct raising o f the N(P) as children do, but then only Spell-Out the head o f the chain. Indeed in languages such as Korean forms both equivalent to head-external structures such as (57) and head- internal relatives such as (58) continue to be produced in adult speech, though William O’Grady interestingly reports (p.c. via Ruth Kempson) that there is a marked change in preference— whereas children seem to produce more head-internal relative clauses such as (58), at around the age of six they start to switch much more regularly to using head-external forms such as (57). The fact that both types of structure continue to exist in adult Korean (and Japanese) might seem to indicate that they result from stylistic (Spell-out) options associated with a single basic relativization strategy rather than being due to two fundamentally quite different strategies associated with different stages of cognitive development. Kayne (1994) also suggests that one might expect for both head-internal and head-external relatives to occur in a single language where the relative clause linearly precedes the head-noun because (in his analysis) there will be no c-command relation between the head-internal position and the external head, and hence either of these (or in theory both) could be spelled-out, the choice being assumed to be largely stylistic (see Kayne 1994 for details). Finally it should be noted that (contra proposals made here) Chiu assumes throughout a basic structure in which the CP relative clause is a left-adjoined to the head-noun/NP and the IP is a leftward complement to de in C°, i.e., the same structure as in Chiu (1993/1995) and Ning (1993)— tree (22). The issues o f directionality and selection and the categorial status o f de discussed here are therefore left unaddressed (not being the focus of Chiu's paper). 121 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. idiom-chunks in favor o f such an account. Connectivity effects occur in English relative clauses such as (59): (59) the picture o f himself that John likes the best (59) contains an anaphor himself which must be bound in order to comply with Principle A of the Binding Theory. This requires that the anaphor must at some point be c-commanded by its antecedent John and as the NP John is located inside the relative clause, this would appear to mean that the NP [picture of himself] must in fact have been base-generated within the relative clause as the object of like (in which position it will be c-commanded by John) and then raised to its surface position. The existence o f connectivity effects such as this has been taken to be a strong argument in favor of a Kayne-style account o f relativization, and was originally discussed in earlier accounts such as Vergnaud (1985). Connectivity has been shown to be dependent upon movement, so that where a movement analysis is blocked, as in the Left Dislocation structure (61) compared to the Topicalization example (60), the connectivity effect disappears: (60) [That picture of himself]i, John liked tj (61) *[That picture of himself], John liked it. Turning now to consider Chinese relative clauses, one finds that connectivity effects are also present in such structures as for example in (62): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (62) Zhangsanj bu xihuan [na-jia chubanshe chuban de] youguan zijij de Zhangsan Neg like that-CL publisher publish DE regarding self DE shu. book 'Zhangsan does not like the books about himself that are published by that publisher.' Such patterns are clearly unexpected in an analysis in which the only element which moves is an empty operator as in (22), as then the anaphor will at no point in the syntactic be c-commanded by its antecedent. In an NP-movement analysis, however, the anaphor ziji will straightforwardly be c-commanded by its antecedent prior to movement o f the NP youguan ziji de shu 'books about himself to Spec of CP. A similar point can be made with respect to the possibility of idiom-chunks in relative clause structures. Vergnaud (1985) notes that parts of an idiom may occur modifying the head noun of a relative clause as already noted in examples such as (39) repeated below: (39) the [headway]; that we made tj on the project Given that the element 'headway' is part of an idiom 'make headway,' it has to be assumed that it is base-generated together with the verb 'make' inside the relative clause and then raised to its surface position. Again, such patterns clearly favor an NP-raising analysis over a null-operator approach which has no obvious way to account for the existence of such forms. Related patterns occur in Chinese relative clause structures as illustrated in (63): 123 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (63) [[takai tj de] daoj] dou hen chenggong. (from Audrey Li 1999b) he open DE knife all very successful T he operations that he performed were all very successful.' In (63), dao 'knife' is part of an idiom kai-dao literally "open-knife" meaning "perform an operation." It would seem that one should therefore assume that it is base-generated together with the verb kai 'open' inside the relative clause and subsequently raised to the head noun position. In an empty operator analysis such as (22), one would not expect for such idiom chunks to be able to occur as head nouns in relative clauses, as in such a structure the head noun cannot be lexically inserted together with material internal to the relative clause. Before concluding that the analysis put forward here is necessarily to be preferred to previous accounts, I will finally consider an account of relativization in Japanese which rejects a straight Kaynean approach to pre-nominal relative clauses and attempt to assess whether such an account might also be appropriate for Chinese. Arguing that the analysis developed here is actually to be preferred, in section 3.3.4 and 3.3.5 I then show how the wider patterning of complex NPs in Chinese adds further support to the Kayne-based analysis. 3.3.3 Relativization in Japanese, Murasugi (1991/1998) Murasugi (1998) takes the original account of Japanese relative clauses which she developed in Murasugi (1991) and examines whether it might be compatible with Kayne’s theory o f Antisymmetry. Her conclusions, following work in Kuno (1973) 124 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and Hoji (1985), are that there are various critical aspects of relativization in Japanese which seem to distinguish Japanese from English-type languages and render a direct translation of Kayne’s proposals inappropriate for Japanese. The first important point that Murasugi makes with regard to relative clauses in Japanese is that they do not seem to need to contain any gap corresponding to the head-noun. Similar remarks appear in a number of works on Japanese relative clauses with this property being first commented on by Kuno (1973). (64) and (65) are typical examples of what are frequently referred to as ‘gapless’ relatives, where it seems highly difficult to re-site the head-noun in any position within the relative clause without adding in a lot o f additional material (which could not be simply deleted in the process of relativization): (64) [John-ga hako-o nutta] omochabako John-Nom box-Acc painted toy-box ‘the toy box that John created by painting a box’ (65) [atama-ga yoku naru] hon head-Nom well become book ‘the book (by reading) which one’s head becomes better’ Murasugi takes the existence o f gapless relatives to be evidence against a straightforward Kaynean analysis of relative clauses in Japanese. An important aspect o f Kayne’s account is that the head-noun is base-generated within the IP in the relative-clause and moved to its surface position; if there arguably is no position Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. within the IP from which it could have originated in gapless relatives, then at least a subset of relative clauses in Japanese must be constructed without any movement. The second point raised by Murasugi, here following Hoji (1985) is that Japanese might not seem to show the important connectivity effects which have been argued to occur in English relative clauses such as (59) in section 3.3.2. Hoji (1985) notes that the occurrence of an anaphor such as iibun modifying the relative clause head-noun in (66) seems to cause the relative clause structure to be unacceptable: (66) *[Johnj-ga_ taipu-shita] [jibunj-no-ronbun] John-Nom type-did self - Gen - paper lit: ‘self’s paper that Johnj typed’ This may then be taken as evidence that the head-noun present in relative clauses in Japanese does not in fact originate within the IP in relative clause structures in Japanese, and that Japanese relative clauses do not involve any Kayne-style movement.1 4 A third related point brought up by Murasugi concerns relative clauses in Japanese and locality. Kuno (1973), Kameshima (1989) and others have noted that the process of relativization in Japanese might not seem subject to Subjacency or the CED. Kuno gives examples of the dependency between the head-noun and a gap position licitly crossing Complex NPs, Sentential Subjects, and Adjunct Clauses: 1 4 Here however there may also be additional non-structural aspects o f the data which contribute to the unacceptability of examples such as (66). Certain other structurally similar examples with iibun modifying the relative clause head-noun are felt to be far more acceptable, e.g: 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (67) [[ _ kawaigatte-ita] inu-ga shinde shimatta] kodomo was-fond-of dog-Nom dying ended-up child ‘the child who the dog (he) was fond of died’ (68) [watakushi-ga _ au koto/no]-ga muzukashii hito I-Nom meet thing-Nom be-difficult person ‘the person who [that I see (him)] is difficult’ (69) [ _ shinda no de] minna-ga kanashinda hito died because all-Nom was-distressed person ‘the person who everyone was upset because (he) died’ Murasugi suggests that this is a further reason to believe that there is no movement involved in relative clause formation in Japanese and that a Kayne-style approach to relative clauses is therefore not suitable for Japanese. Murasugi proposes instead that any apparent gap in a relative clause is actually a base-generated pro controlled by the head-noun which is also base generated in its surface position and not moved from any other position. Referring back to a set of arguments in her 1991 thesis, Murasugi argues that relative clauses in Japanese should actually be taken to be IPs rather than CPs and then suggests the possible structure in (70), noting that this is similar in certain ways to the structure she proposed in her thesis, given in (71): (i) [Johri|-ga _ suki-na] [jibun|-no-shashin] John-Nom likes selfi-Gen-picture ‘the pictures of himself which John likes’ 127 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ...proj D NPj (71) NP IP NP (70) is a potential reflection of the various conclusions/points argued for by Murasugi. If Japanese does not show connectivity effects and if there are fully gapless relatives in Japanese, then one need not assume that the head-noun is base generated within the IP and it may be directly inserted in its surface position. If the lack of Subjacency/CED effects indicate that there is no movement of any kind, including any null operator movement, then there need not be any CP layer of structure present to host an operator in a raised SpecCP position, and the relative clause itself can be argued to be an IP-level constituent. Finally, if the head-noun is not raised to its surface position from within the IP, there is no need to assume that the IP must have been base-generated in a position to the right and lower than the head-noun as in Kayne. Consequently it is possible to suggest that the IP is actually base-generated in its surface position in Spec of DP, this essentially being an updated rather more articulated version of the 1991 base-generation analysis represented in (71). It may be noted that a somewhat similar approach is suggested for Korean relative clauses in Kim (1997), with the relative clause being base-generated in a left-branch preceding the head-noun and no movement o f the head-noun being involved. 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. So, if there are reasons for believing that (70) may be a plausible analysis for pre-nominal relative clauses in Japanese, the question arises as to whether a folly base-generated structure such as (70) should not also be appropriate for pre-nominal relatives in Chinese rather than the Kaynean analysis proposed. I would like to argue that (70) is not suitable for Chinese on the basis of two rather general sets of arguments. First o f all, the structure in (70) arises as a possible analysis of relative clauses in Japanese as a direct result of the conclusion that there would not appear to be any movement involved in the formation of relative clauses in Japanese. This allows one to suggest that the head-noun is base-generated in its surface position rather than moved there and that the relative clause itself is simply an IP with no internal movement to any Spec of CP. In Chinese however it has been argued (in section 3.1.1) that there are in fact good reasons for assuming some kind of movement in relativization (either of an empty operator or the head-noun). In a number of very clear cases, Chiu’s data relating to the patterning with suo and Ning's arguments concerning the Generalized Control Rule convincingly show that relativization in Chinese crucially is island-sensitive when one controls for and blocks the possible use of a resumptive pro strategy.1 5 1 3 Possibly Japanese might also show itself to be island sensitive if one could establish a similar sort o f test and control. That is, maybe the apparent island violations noted in Kuno (1973) are cases where a resumptive pro is made use o f and that there otherwise is indeed some kind of movement similar to Chinese. Possibly if the head-noun were to be a quantificational equivalent to no-one or more than two people one might find the re-appearance o f Subjacency effects (a resumptive pro not being available in such necessarily restrictive relative environments). 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Consequently there must be more to the structure o f relative clauses in Chinese than indicated in (70). Either one may assume that the head-noun undergoes movement in Chinese, as in the Vergnaud-Kaynean analysis proposed earlier, or one could suggest that there is some kind of empty operator movement. Both such accounts would seem to require (70) to be non-trivially modified. If the first option is taken, then the relative clause must originate to the right o f the head noun in a position from which the head-noun may undergo raising (i.e., one cannot assume that the head-noun is lowered from an IP base-generated in Spec of DP as in (70)). (70) will have to be modified into (72): (72) DP Spec CP Spec If the second possibility is taken, it might also seem to lead to an analysis in which the relative clause is base-generated lower than the head-noun and to its right in the structure. When there is null-operator movement rather than movement of the head- noun, the null operator must be assumed to move to a SpecCP operator position (hence the clause must be a CP rather than an IP). As it is a reasonable assumption that operators whose reference is controlled within a sentence should be c- 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. commanded by the element controlling this reference (as e.g. in easy-to-please constructions, topicalization, standard relative clause structures, etc.), this would seem to require that the relative clause CP should be base-generated in a position c- commanded by the head-noun, hence to its right in a modified form of (70), such as (73).1 6 1 7 Either way, the fact that Chinese relative clauses seem to give clear evidence o f movement has for effect that a base-generated structure similar to (70) can arguably not be taken to be appropriate for Chinese. (73) DP Spec D ’ NP CP In fact, as noted earlier, it is not only the Subjacency/CED facts which appear to distinguish Chinese from Japanese, Murasugi’s first argument for a non-movement analysis of relative clauses in Japanese, namely the occurrence of gapless relatives 1 6 If the head-noun is only co-referenced with a pro as in Murasugi’s analysis of Japanese, there is no parallel necessity for a c-command relation between the head-noun and the pro, pronouns in general not needing to be c-commanded by their antecedents. 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. such as (64) and (65), would also seem to be inapplicable for Chinese. Examples such as (64) and (65) when translated into Chinese are quite unacceptable, suggesting that Chinese does not tolerate truly gapless relative clauses, as indeed argued by Ning. Furthermore, as noted earlier in example (62), Chinese also displays connectivity effects which Murasugi takes to be absent from Japanese.1 8 Murasugi assumes that such patterns are indicative o f movement; their presence in Chinese but arguably not in Japanese would further seem to argue against adopting Murasugi's account of relativization in Japanese for Chinese:1 9 2 0 1 7 For the surface order o f the pre-nominal relative clause, CP undergoes raising to precede the head noun. At LF the raised CP will then either reconstruct to the position where the operator in SpecCP is c-commanded by the head-noun, or one might assume that a copy is left in the position from which the CP moves, this satisfying the c-command requirement. 1 8 Such connectivity also indicates that standard GB analyses, in which an operator movement is proposed (Chiu 1993/1995 and Ning 1993 discussed above, for example), are less appropriate than the analyses with the head-noun movement. 1 9 Another interesting difference between relative clauses in Chinese and Japanese which may be noted here concerns the positions o f demonstratives and restrictive/non-restrictive interpretations. In both languages (and also Korean, which patterns like Japanese and different to Chinese, see Kim 1997) it is possible to position demonstratives either before the pre-nominal relative clause or between the relative clause and the head-noun. However, this seems to result in opposite interpretations in Chinese and Japanese. In Chinese if the demonstrative occurs between the relative clause and the head-noun, a regular restrictive interpretation is possible, but such a positioning automatically triggers a won-restrictive interpretation in Japanese. When the demonstrative occurs preceding the relative clause, this may give rise to a non-restrictive interpretation in Chinese (Huang 1982), whereas it normally results in a restrictive interpretation in Japanese (and Korean): (i) [wo zuotian mai]-de zhei-ben-shu I yesterday buy DE this-CL-book 'the book that I bought yesterday' (ii) [watashi-ga kinoo katta] kono hon I-Nom yesterday bought this book this book, which I bought yesterday' (iii) zhei-ben [wo zuotian mai]-de shu this-CL I yesterday buy DE book 'this book, which I bought yesterday' (iv) kono [watashi-ga kinoo katta] hon this I-Nom yesterday bought book 'the book that I bought yesterday’ 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The second broad piece of argumentation against a structure such as (70) and for the Kayne-based approach outlined earlier relates to the syntax of noun complement clause structures in Chinese. Essentially it will be argued that noun complement clause forms have an underlying syntax which is very similar to that of relative clauses, and that as the structure in (70) can be shown to be inappropriate for noun-complement clause units, it may by extension also be taken to be inappropriate for relative clauses. The line of argumentation will take its lead from significant work on the syntax of process nominals presented in Fu (1994) as well as an extension of this in Simpson (1997). After briefly describing the basic intuitions of these two works, I will then indicate how they can be shown to favor the Kayne- based account of relativization rather than any fully base-generated structure such as that in (70). I do not propose to attempt an explanation of this difference here, but just note it as another example o f how apparently similar surface structures in Chinese and Japanese actually seem to have quite different properties on closer examination. Having argued that there is good evidence in favor of a Kaynean-type movement account o f relativization in Chinese, I have suggested that structures such as (46/47) represent the derivation o f regular relative clauses in Chinese. For those cases in which it appears that a resumptive pro can be base-generated in relative clauses in Chinese (and Subjacency violated), I assume that this results from a modification of the basic movement structure in (46/47) below simply with the head noun/NP being base-generated in Spec o f CP and a resumptive pro occurring in some position in IP. This follows the approach commonly adopted towards relative clauses with resumptive pronouns in other languages that these are simply due to base-generation o f the resumptive pronoun and the relative- operator in the same relative clause positions that would otherwise be linked via movement. Such an approach seems more natural than to posit quite different structures for resumptive pronoun cases— if a relative clause movement-type structure such (46/47) is anyway necessary and available for minor modification via base-generation o f a resumptive pronoun, this would seem to be simpler than to assume some different structural form. 133 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3.4 Process Nominals and the Structure ofNoun-Complement Clause CNPs Noun-complement clause structures in Chinese appear very similar to relative clauses. The clausal unit precedes the selecting noun and de is found to occur in between these two elements: (74) [Deng-Xiao-Ping shishi] de xiaoxi Deng Xiao Ping died DE news ‘the news that Deng Xiao Ping died’ Such structures again raise the problem of the direction of selection. Elsewhere in Chinese there is consistent, good evidence that both lexical and functional heads select their complements to the right, i.e., Chinese is a descriptively head-initial language. Here however in the NP/DP the surface patterns seem to show the opposite, i.e., the complement of N seems to be selected to its left. Certain linguists (e.g., Lin 1994 and Tang 1990) have suggested that the bracketed string preceding de in sequences such as (74) is not in fact a complement, despite its connection to the head noun in terms of meaning. In the spirit of Grimshaw (1990) it is suggested that such elements may actually be syntactic adjuncts and so the directionality issue is not a problem after all. However, in a detailed study o f the syntactic properties and structure o f process and result nominals in Chinese, Fu (1994) provides convincing evidence that the elements occurring with the former type of derived nominal are indeed syntactic complements. He shows that when a process reading of these derived nominals is forced by means of duration phrases and the process verb jinxing 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ‘to last,' the object o f the noun head is indeed forced to appear, this being a good indication o f its argumenthood: (75) ta ??(dui zaiqing) de baodao jinxing-le san-ge-xiaoshi he towards disaster DE report last-Asp three-CL-hours ‘His reporting of the disaster lasted three hours.’ (Fu 1994) Fu also shows that with a process reading, only a single de may occur and this has to be placed after all of the arguments associated with the noun. Specifically, the subject argument may not be marked by de although one does find multiple instances of de elsewhere with clearer cases of adjuncts: (76) [Zhangsan (*de) dui zhei-ge-anjian de diaocha] jinxing-le Zhangsan DE towards this-CL-case DE investigation last-Asp liang-ge xiaoshi two hour ‘Zhangsan’s investigation o f this case lasted two hours.’ Fu takes this as evidence for a process o f nominalization in which the head noun (baodao in (75), diaocha in (76), etc.) is inserted as a verb in a regular verbal/clausal constituent (the bracketed string in the examples above), projecting its argument structure in a regular way inside this VP/clause, and then raises rightward over de to an empty nominal head position, so being converted into a nominal category. Simpson (1997) points out that such an account still leaves the directionality problem unaddressed and proposes a modification of Fu’s analysis, suggesting that the verb raises leftward to an N° which selects for the clause in a rightward direction. This Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. first movement is then followed by raising of the clausal remnant to the SpecDP position, with de assumed to be in D°. The derivation of a string such as (77) will then be as in (78) and (79): (77) [tamen dui zhei-ge wenti] de taolun they towards this-CL problem DE discussion ‘their discussion of this problem’ (78) taolunj tamen dui zhei-ge wenti tj DP (79) VPk tamen dui zhei-ge wenti tj N ’ VP Interestingly this analysis actually results in a form closely resembling the potential surface structure for Japanese relative clauses proposed by Murasugi in (70), i.e., there is a clausal element in the SpecDP position and the noun-head is selected in a 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. rightward direction by D° (which has an overt instantiation in Chinese). However, there is an extremely important difference between (79) and the form in (70). In (79) the clause is raised to its surface position from a position in which it is: (a) base generated in a regular structural complement relation with the selecting N°, and (b) this selection relation follows the canonical direction of selection as observed in Chinese. If the VP in (79) and other similar cases is indeed a complement of N, it should not be possible to base-generate it in a SpecDP position, as there is no structural relation of complementation between the N° head and the SpecDP position (in fact there is no standardly definable structural relation between these two positions at all). Consequently a frilly base-generated structure such as (70) would seem to be quite inappropriate for noun-complement clauses in Chinese, and a plausible analysis which accommodates the problem of the unexpected exceptional directionality found in such strings suggests that the complement clause is in fact raised to its surface position. Now, if there is indeed such a productive process of clausal raising to the Specifier of DP’s headed by the enclitic determiner de (and this raising is triggered by the particular properties of de), this would seem to provide natural support for the analysis o f relative clauses outlined earlier. In other words, if the base-generated structure (70) cannot be applied to noun-complement clause strings and a clausal raising analysis appears independently necessary to account for such structures in a theoretically undisruptive way, this same essential analysis should be available for relative clauses as well (the particular lexical properties of de remaining constant). Coupled with the suggestion that an operator present in the 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. relative clause should be c-commanded by the noun/NP which controls it, this would seem to clearly favor the derivation and representations given earlier over the possible alternative analysis in (70).2 1 5.3.5 Other Noun-Complement Clause CNPs Before closing this section, I would finally like to suggest that structures similar to those in (78/79) but without any V-to-N-raising may be appropriate for certain other CNP types in Chinese, CNPs such as those in (80)-(83) below: (80) [ta xue Zhongwen] de yuanyin he study Chinese DE reason ‘the reason he studies Chinese’ (81) [tajiejue wenti] de fangfa he solve problem DE method ‘the way he solved the problem’ 2 1 A general question concerning the account of clausal raising outlined here is why dezszn enclitic needs to attract a clausal constituent rather than, say, the highest XP within that clause. Although I have no great insight to offer here, it can be noted that clitics cross-linguistically do show considerable variation in the hosts which they target. In some languages clitics may attach themselves to a wide variety o f hosts, whereas in other languages clitics may only tolerate a single type o f host (as e.g. the pronominal clitics in French/Italian, which only attach to verbal elements, discussed in Spencer 1991). Here it can be said that de is also rather selective as a clitic and will only attach to verbal projections. In order to implement such an idea technically within a Minimalist approach, one can suggest that the D position is generated with a strong v-feature which triggers raising of some verbal constituent to its Specifier. In the case o f possessive structures such as Zhangsan de shu 'Zhangsan's book' which might seem to be a counter-example here, one might assume Simpson's (1998b) analysis in which possessor structures result from relativization and an empty predicate of possession, as noted earlier in footnote 10. If this is so, then possessor forms will actually have the structure: [D p [ip Zhangsan _ tj ]m de [cp [shu]j tn, ]] and de will still be attracting an IP category. 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (82) [ta chang-chang lai Zhongguo] de shihou he often come China DE time ‘the time he often came to China (83) [wo kanjian ta liang ci ] de difang I see he 2 time DE place ‘the place I saw him twice’ Ning (1993) assumes that such structures are actually relative clauses and suggests that because there is no obvious gap corresponding to the head nouns, relative clauses of this type must be formed by the movement of an empty operator (see section 3.1.1 above). If an empty operator analysis is indeed taken to be necessary for (80)-(83), such examples might be argued to constitute evidence against the general suggestion of the chapter that relative clauses are formed via raising of the NP-‘head’ in Chinese rather than resulting from empty operator movement. If one is forced to concede that an empty operator strategy might occur in this particular subset of cases, it might be objected that such a strategy might possibly also underlie all other cases of relativization (though the connectivity arguments in (62) and acquisition patterns in 3.3.2 would still remain difficult to explain for any hypothetical operator account). In defence of the proposed raising analysis of relative clauses in Chinese, the following points and suggestions can however be made. First of all it can be noted that even if one were forced to concede an empty operator analysis for examples such as (80)-(83), this would actually NOT force the conclusion that ALL relative clauses should necessarily be formed via such empty operator movement and that other cases of relative clauses might not be formed by 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. NP-raising as suggested. It is certainly possible that languages might in fact have more than a unique strategy for relativization. In English, for example, it is commonly assumed that relative clauses may either be created by movement or alternatively be fully base-generated when a resumptive pronoun occurs, and in Hindi Dayal (1995, 1996) suggests that three significantly different strategies of relativization are actually available. Supposing that (80)-(83) were to be formed via empty operator movement, the evidence considered here in the chapter could still be taken to support a second NP-raising analysis of other more regular relative clause structures. Secondly, there is actually a rather simple and natural way to avoid Ning’s conclusion that (80)-(83) are formed via empty operator movement. Rather than automatically assuming with Ning (1993) that such structures are necessarily relative clauses, one can instead assume that they are simple noun-complement clause CNPs similar to ‘the claim [ c p that...]’ CNPs in English and many other languages where a head noun selects a complement clause specifying its content, as in (84)-(86): (84) the claim [ c p that John was a liar] (85) the news [ c p that there was an earthquake] (86) the rumor [cp that she was engaged] In this second widely-available type of CNP it is regularly assumed that the head noun directly selects a complement CP and there is no relativization of an empty operator. Such a noun-complement clause strategy can also be argued to be 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. responsible for direct English equivalents to the Chinese cases in (80)-(83) such as (87)-(90) below: (87) the reason [cp he studies Chinese] ( 8 8 ) the way [ c p he solved the problem] (89) the time [ c p he came to China] (90) the place [ c p he studies Chinese] Note furthermore that although English allows w/z-elements to be used as relative pronouns, such CNP forms frequently resist the occurrence o f any attempted wh relative pronoun indicating that it must be possible for such forms to be created by the noun-complement clause non-relativization strategy: (91) *the way how he solved the problem (92) */??the second when you come here 2 2 Sometimes for certain speakers a wh relative pronoun does seem possible, as in (i-ii): (i) % the reason why Mary likes John (ii) % the time when he met Mary Informants indicate that such forms often seem to be dispreferred in speech, that not ail speakers accept them, and that in many cases overt wh elements are completely rejected by ail speakers (such as for example (91)). I would here like to speculate that all CNPs o f the type seen in (84)-(92) are formed as noun-complement clause structures parallel to [the claim [cp that...]]-forms. When wh elements do occur for certain speakers, they are actually occurring base-generated in C° as selected wA-variants to the default C° that, similar to the fact that C° elements are in other instances also arguably selected by embedding heads, as in (iii) and (v). (iii) John wonders [if/whether Mary will come]. (iv) Bill denied [that he is in love with Mary], (v) Sue asked [for Bill to leave]. Consequently such structures in (i) and (ii) will again not be instances o f relativization. 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In Chinese, as in English, it can therefore be suggested that the clauses which occur with head nouns such as yuanyin ‘reason’ and fangfa ‘way,’ etc. can be considered to be complement clauses specifying the content of the selecting nouns in the same way that CPs occurring with head nouns such as ‘news/rumor/claim’ specify their content. Cases such as (80)-(83) can therefore be analyzed in a way similar to their English counterparts and in a way which is also similar to the Chinese structures (78/79), as in (93/94) below, with yuanyin/fangfa selecting a rightward complement clause which then undergoes raising to SpecDP to support the enclitic de: (93) DP D’ D NP de N CP yuanyin 'reason' ta xue Zhongwen 'he study Chinese' (94) DP Spec D’ [ta xue Zhongwen!m de '[he study Chine yuanyin 'reason' N ’ CP 142 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Cross-linguistically this type o f CNP strategy is subject to certain lexical variation and that not all (types of) head-noun syntactically license a complement clause in such a way. For example, in English ‘way’ licenses a complement CP as in (88), but the semantically similar ‘manner’ does not, as shown in (95). Note that if (88) were to be formed by a process o f relativization with some kind of empty operator rather than being a complement clause structure, it is quite unexpected that such an operator-relative clause possibility should not be available when the head-noun is instead ‘manner.’ Processes of relativization should be fully blind to the lexical properties of the head noun and it should be possible to relativize any head noun if an appropriate operator is available. The fact that (95) is not acceptable therefore clearly suggests (88) is not formed by relativization. In a similar way, whereas English ‘rumor’ and ‘news’ license complement clauses specifying their content (examples 85/86), the nouns ‘sound’ and ‘smell’ by way of contrast do not, as seen in (96/97). This contrasts with languages such as Japanese and Mandarin where equivalents to ‘sound’ and ‘smell’ do indeed license complement clauses specifying their content: (95) *the manner [ c p that John walked] (96) *the sound [cp that Bill hit the drum] (97) *the smell [cp that Mary cooked the garlic] Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (98) [Taroo-ga kitaa-o hiku] oto (Japanese) Taroo-Nom guitar-Acc play sound lit: *‘the sound that Taroo played the guitar’ (99) [Zhangsan da gu] de shengyin (Mandarin) Zhangsan hit drum DE sound lit: *’the sound that Zhangsan hit the drum’ It can therefore be suggested that Chinese is to some extent more permissive than English, and that Chinese allows such noun-complement clause structures with a whole range o f noun types, potentially specifying the content/identity of nouns in terms o f reason, manner, time, place and also other sensory properties.2 3 2 3 Note that there is also another quite simple analysis o f structures such as (82) and (83) in which they can in fact be assumed to be relative clauses resulting from a Kaynean process of NP- raising. Kuno (1973) argues on the basis o f a variety o f data that relative clauses are commonly formed from topic structures and that it is the topic o f a sentence which is relativized as the head- noun/NP o f a relative clause. If this is so, because the head nouns/NPs in structures such as (82) and (83) can occur as topics preceding the bracketed IP as in (i) and (ii) below, it can be suggested that such NPs are relativized to their surface position from an underlying topic position and that such structures are therefore not ‘gapless’ relative clauses: (i) nei-ge-shihou, wo chang-chang Iai Zhonguo that CL time I often come China ‘At that time I often came to China.’ (ii) nei-ge-difang, wo kan-guo ta liang ci that CL place I see-ASP him 2 time ‘In that place I saw him twice.’ An analysis along these lines would consequently seem to allow one to maintain that Kaynean-type NP-raising does occur also in cases where there is perhaps no obvious gap in the relative clause— the extraction gap o f the NP would actually be in the topic position. Note that in such an approach, examples such as (80) and (81) would however still have to be analyzed as being N+ CP complement structures, as the nouns yuanyin ’ reason' and fangfa 'way/method' cannot occur as topics connected to the relative clause IP as in (iii) and (iv): (iii) *nei-ge yuanyin, ta qunian xue zhongwen that-CL reason he last-year study Chinese (iv) *nei-ge fangfa, ta jiejue wenti that-CL method he solve problem 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.4 Taiwanese Tone Sandhi and the Clausal Raising Hypothesis In this last section I would like to present a further set of evidence in favor of the basic account of relative clauses suggested here, based on the phenomenon of tone sandhi in Taiwanese. I will argue that a certain tone sandhi patterning occurring in relative clause structures in the Taiwanese dialect o f Chinese can be given neat explanation if one indeed assumes the approach to relativization proposed here. I will also argue that this patterning remains puzzling and unaccounted for in a GB/Murasugi-type base-generation analysis. In order to make such arguments, it will first be necessary to provide a very brief background sketch of basic tone sandhi phenomena in Taiwanese and then turn to relative clauses in particular. 3.4.1 Tone Sandhi Patterns in Taiwanese Taiwanese is a variety o f Chinese with eight tones:24 The 1st: high-level 5-5, the 2nd: high-falling 5-1, the 3rd: low-falling 2-1, the 4th: low-entering tone (a syllable with a final stop), the 5th: contour-tone 2-1-4, the 6th: high-falling 5-1 (the same as tone 2), the 7th: mid-level 3-3, the 8th: high-entering. In addition to these eight (actually seven) tones there are also syllables which do not carry any tone, this sometimes being referred to as ‘neutral tone’ (NT). In the phenomenon of tone sandhi, the lexically-listed tone of a syllable is generally able to undergo modification according to fully regular rules when preceding some other tone-bearing syllable in the same 2 4 Two o f the ‘eight’ traditionally recognized tones, tone2 and tone6 are actually identical in phonological terms— both are high-falling 5-1. 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. tone sandhi domain. For example, if a syllable with tone 1 precedes another tone- carrying syllable, the tone 1 may change into a tone 7, as illustrated in (100): (100) khi3 pak8kiangl — > khi2 pak8kiangl go Beijing ‘go to Beijing’ Table (101) below shows how the full range of these modifications are made. Note that the changes in tone are not triggered or conditioned by the particular type of tone that the following syllable carries, hence a syllable with tone 1 will change its tone to tone 7 no matter whether the following syllable has tone 1, 2 or 3, etc.; the essential requirement for tone sandhi to apply is that the following syllable have some type of lexical tone rather than just ‘neutral tone.' (101) a. 1st -» 7th b. 2nd -» 1st c. 3rd 2n d d. 4th -> 8th when the syllable ends syllable ends in a glottal stop in p/t/k; -> 2nd when the e. 5th 7th (southern Taiwan); -> 3rd (northern Taiwan) f. 6th -> 1st g- 7th 3rd h. 8th -> 4th when the syllable ends syllable ends in a glottal stop in p/t/k; -> 3rd when the Basically, a high-level tone will become a mid-level tone, a mid-level tone will turn into a low-falling tone— in both cases a ‘down-grading’ o f the tonal value; if a tone 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. occurs at the bottom end o f the tonal system, tone sandhi will convert it back up to the top of the tonal group— e.g., a low-falling tone will become a high-filing tone (and a low-entering tone becomes a high-entering tone, and high-entering tone becomes a low-entering tone). As mentioned just above, tone sandhi may not occur in a syllable if it precedes a syllable which has only neutral tone/no tone. Consequently zau in example (102) may not change its tone-2 when occurring before the toneless element a in (102): (102) zau2 a-NT -> zau2 a-NT run already ‘already left’ Similarly, a syllable may not undergo tone sandhi if it occurs sentence-finally. This is due to the fact that tone sandhi is restricted to apply within certain specific domains. In (103) below, the citation tone-2 of sentence-final ho may not be converted into tone-1 even though followed by a syllable (A-hui) which does carry tone because this latter syllable occurs in a separate sentence. Note that from this point on for simplicity of representation I will indicate tone sandhi change by means of a simple bolded dot following the relevant syllable. Thus if a syllable is followed by a bolded dot, this indicates that it undergoes tone sandhi change, and if a dot is absent, no tone sandhi change is possible. In (103) sentence-final ho is therefore not followed by a dot as no tone sandhi change can occur in sentence-final position: 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (103) A»-sin chin* ho. A*-hui ma* chin* ho. A-sin very good A-hui also very fine ‘I am fine, and my father is also fine.’ Sentence-intemally there would also seem to be other tone sandhi/TS domains relevant for the operation of tonal change, and broadly-speaking every syllable in such a domain will change its tone unless it is the last tone-bearing syllable. Significantly, tone sandhi change in Taiwanese appears to relate to and reveal the underlying syntactic structure in a way which is not found in tone sandhi phenomena in Mandarin, Shanghainese and certain other varieties of Chinese. For present purposes it is important to point out the following three significant generalizations: (104) Generalization A: a head and its complement occur in the same TS domain The presence of an overt complement consistently triggers tone sandhi change on the selecting head, indicating that a head and its complement are in a single TS domain: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (105) (106) (107) a. V -N P object be* [lng*-pun* chhe] buy two-Cl books *buy two books' b. P-NP tui* [goan* lau*pe] to my father 'to my father' c. Aux/I-VP lai will come 'will come' d. Comp-IP na*si* [A*sin m* lai]... if Asin neg-want come 'If Asin is not coming...' Generalization B: a head and its Specifier do not occur in the same TS domain It is found that a head does not trigger tone sandhi change on the final syllable of its Specifier. Consequently the Specifier of a head constitutes an independent TS domain. In (107) below, the final syllable of the subject does not change its tone, despite being followed by the tone-bearing head u ‘have’: [A*-sin] u* lng* chhing* kho A-sin have two thousand dollar ‘ A-sin has two thousand dollars.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In addition to (107) above with the final syllable of a subject in SpecIP failing to undergo tone sandhi, further examples of Specifiers being isolated TS domains are given in (108) and (109) below, where the DP tai-oan-oe ‘Taiwanese’ occurs as either a moved or base-generated topic relating to the object position. In such a Specifier position, its final syllable oe does not undergo any tone sandhi change: (108) A*-sin [tai«oan»oe] be* hiao* kong A-sin Taiwanese not know speak. 'Taiwanese, A-sin can't speak.' (109) [tai«oan«oe] A«-sin be* hiao* kong Taiwanese A-sin not know speak. 'Taiwanese, A-sin can't speak.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (110) Generalization C: adjuncts are self-contained TS domains The final syllable of an adjunct does not undergo tone sandhi even when followed by other tone-bearing syllables. This is illustrated below with the case of a CP adjunct. No tonal change in its final syllable is possible:2 5 (111) [na«si« A«sin m» khil. A»hui ma* be* khi if Asin neg go Ahui also Neg go ’ If Asin is not going, Ahui will also not go.' 3.4.2 Tone Sandhi in Taiwanese Relative Clauses Turning back to relative clauses now, there are certain critical aspects of their tone sandhi patterning in Taiwanese which argue against a standard GB-type account of 2 5 Note that a quite different approach to the description of tone sandhi generalisations is offered in R. Cheng (1968/1973). Rather than stating (Taiwanese-type) Min dialect tone sandhi as a set o f phonological rules which applies to major syntactic units as input (i.e. specifiers, heads, complements and adjuncts), Cheng suggests that such tone sandhi is instead sensitive to the particular categorial type/label of a syntactic constituent, and that tone sandhi is a process which is blocked and fails to apply if a syllable occurs specifically at the end o f an NP, a VP, IP, sentential AdvP or CP. Such a categorial-based approach is however strongly criticized in Chen (1985). Chen vigorously points out that truly productive phonological processes such as tone sandhi are nowhere else found to be directly sensitive to and restricted by particular categorial labels and phonological rules instead seem to be blind to categorial distinctions (for example, one never finds cases of other phonological processes such as vowel harmony, palatalisation, spirantisation etc being restricted by categorial type and therefore only occurring in AdjPs or PPs, or NPs and AdjPs etc, although such processes may be subject to other more general boundary conditions). Chen argues that it would be consequently quite implausible to assume the exceptional existence of rules which do refer to a subset of such labels just in the case o f Min tone sandhi. In contrast to the lack o f such category-specific phonological rules however, Chen notes that cross-linguistically there are many phonological processes which are sensitive to more general syntactic divisions in sentential structure, indicating that phonology (potentially) does recognise distinctions between arguments and adjuncts and other major syntactic relations, and that such more general distinctions are more likely to be relevant for Min tone sandhi. I fully agree with Chen’s argumentation that phonological rules should be assumed to be unable to refer specific category labels and would also like to thank Monik Charette, Moira Yip and Jean-Roger Vergnaud for useful discussion and confirmation o f this point. The chapter consequently continues to assume that Taiwanese tone sandhi is indeed a function o f more general syntactic structure and the generalisations relations suggested in the text and is not a result of categorial labels. For further discussion o f how Min tone sandhi patterns are sensitive to argument/adjunct type syntactic relations, see Chen (1990) and also Chen (2000). 151 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. their structure. A simple relative clause in Taiwanese and the structure that such a form would be naturally assigned under Ning’s (1993) and Chiu’s (1995) analysis of Chinese relative clauses is illustrated in (112) and (113). In (113) the functional element e (=Mandarin de) is analyzed as a relativizing complementizer in C similar to English that (following the Ning/Chiu assumption that Mandarin de is a C°), and the CP relative clause is taken to be left-adjoined to the final NP (again as in the Ning 1993/Chiu 1995 analysis o f Mandarin Chinese): (112) [ipA»sinbe] e» T n p chhe 1 Asin buy Rel book ‘the book Asin bought’ (113) CP NP O: C' chhe 'book' Considering forms such as (112), three fully general and important tone sandhi facts can be noted: A*sin be t, buv t/ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (114) a. The (final syllable of the) head-noun does not undergo tone sandhi. b. The functional element e does undergo tone sandhi. c. The final element in the IP preceding e does not undergo tone sandhi. Taking these properties one by one and reflecting on how they might be accounted for in a structure such as (113), it is found that the first property (114a) (that the final syllable of the head-noun does not undergo tone sandhi) is not difficult for a GB- style analysis to explain. Above in (103) and (106) it has been noted that the final syllable of a sentence and the final syllable o f a Specifier do not undergo tone sandhi. If relative clause forms such as (112) then occur either sentence-finally as objects or as subjects in a specifier position it is expected that the final syllable in the structure will not undergo tone change. Property (114b) however is very difficult for a GB-style account to explain. If one assumes an analysis such as (113), in which e heads a CP left-adjoined to the NP/head-noun, it is unexpected that the functional element e would change its tone. Note here that the Taiwanese element e, unlike Mandarin de, does indeed carry a tone; its lexically-listed citation tone is tone 5, and this undergoes tone sandhi modification when preceding a (tone-bearing) head-noun to tone 7. In (111) it has been noted that CP adjuncts constitute their own isolated tone sandhi domains and tone sandhi does not take place between the final syllable in a CP adjunct and the element it is adjoined to. Consequently, if the relative clause is assumed to be a CP adjunct asin(113)it should not be possible for the head-noun to trigger tone sandhi 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. leftwards onto the the last syllable of the CP adjunct~e, yet e clearly does change its tone. Property (114c) is also quite unexpected in an analysis such as (113). If the IP-clause is taken to be a leftward complement to the functional head e in C°, then this IP and e in C° should constitute a single tone sandhi domain. In (104) it was noted that the presence of an overt complement consistently triggers tone sandhi on the selecting head, and in (105d) it was seen that a C° head and its IP complement constitutes a tone sandhi domain and therefore every syllable but the last syllable in the domain should change its tone. This being so, and given that Taiwanese e is a genuinely tone-bearing syllable, it is expected that the C°-element in (113) should be able to trigger tone sandhi onto the last syllable of the preceding IP-complement~the verb ‘buy,' but this is not possible.2 6 There are consequently good reasons to doubt the correctness of the GB-type analysis of Chinese relative clauses as illustrated in (113). The Kaynean-type 2 6 One might perhaps wonder whether an alternative account of the patterning might be possible, relating it to the paradigm of wh-trace interference in wa/wicr-contraction where the occurrence of an A’-trace between two elements blocks the phonological process of contraction from occurring (i.e., * Who do you wanna win? is unacceptable as the A’-trace of who occurs between want and to). Movement o f an empty operator to Spec of CP might leave behind a trace o f A’-movement which would linearly intervene between the element targetted for tone sandhi application be 'buy' and the following tone-bearing element potentially able to license tone sandhi (e). This A’-trace intervention could then possibly be held responsible for the impossibility of tone sandhi in the verb be. However, tonal change on the final element in the IP is equally impossible when a subject is relativized from within the IP and no trace intervenes between the final IP element and the C° e, as in (0- (i) [ Oj q be* chhe] e* lang buy book e person 'the person who bought books' 154 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. analysis o f relativization in Chinese argued for earlier by way o f contrast is indeed able to explain all o f the tone sandhi patterns observed without any o f the difficulties found in the more traditional analysis in (113). (115) and (116) are representations of the hypothetical derivation of Taiwanese (112) in a Kaynean analysis: (115) (Spec chhei •book’ (116) A»-sin be t j D A»-sin be ti 'A-sin buy tj' 'A-sin buy tj' e» It would therefore seem that it is not the occurrence o f an A’-trace before e in (112/113) which blocks it from licensing tone sandhi (one can also note that A’-traces in fact do not seem to block contraction processes in any clearly uniform way— Sag & Fodor (1995) observe that the w/»-trace resulting from extraction out o f tensed subject positions does not block auxiliary contraction: Who do you think'll/’ s come?). 155 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First o f all, property (114a) that the NP chhe ‘book’ does not show any tone change can be simply assumed to be due to one of the two facts. First it occurs in a Specifier position, Spec of CP, and the final syllables of elements in Spec positions do not undergoing any tone sandhi change as per (106). The second fact is that the NP 'book' is linearly final in the whole DP which itself may occur as an object sentence- finally or as a subject in a Specifier position, and neither o f those positions allow it's final syllable to undergo tone change. Secondly and more significantly, (115/116) are able to account for property (114b), the fact that the functional element e undergoes tone sandhi. In the GB analysis this is a problem because e occurs in a CP adjunct and so should not be visible outside this domain for any tone sandhi operations. In the structure in (115/116) however, no adjunct relation is present, and e is instead a D° selecting a rightward complement. As a head and its complement occur in a single tone sandhi domain and complements consistently cause tone changes in their selecting heads, it is actually very natural that the element following e in its complement CP does trigger tone sandhi on e.2 1 The third property (114c) is also importantly explained, namely that the final element in the B P of the relative clause (the verb be 'buy') does not undergo tone 2 7 Note that if there is no overt NP following e, then e does not undergo tone sandhi, as shown in (i): (i) [IP A*sin be] ej [N P _ ] Asin buy Rel ‘the one(s) Asin bought’ 156 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. sandhi. In the GB analysis this is unexpected because this element is assumed to be in the unmoved complement IP o f the C° head instantiated by e and so the C° should trigger tone sandhi on the preceding element be 'buy.' In (115/116) however, the IP is analyzed as being base-generated as the rightward complement of a phonetically null C° and then raised to its surface position in SpecDP, where e actually instantiates D°. Occurring in a Spec position, it is fully expected that the IP will constitute an isolated tone sandhi domain and that the final syllable in such a constituent (i.e., the verb be 'buy') should not undergo any tonal change. Consequently then, it would seem that there is good evidence from tone sandhi supporting a Kaynean analysis of relative clauses in Chinese of the type outlined in the chapter here. Whereas a more traditional GB approach to relativization in Chinese assuming de/e to be an instantiation of a C° selecting a leftward complement IP is arguably unable to account for the patterns of tonal change revealed in Taiwanese, a uniformly right-branching analysis of relative clauses with de/e taken to be in D° provides a principled and natural account which accords with other general patterns of tone sandhi phenomena productively observed in the language. To the extent that such evidence therefore further bears out the Kaynean analysis already justified on other grounds, it might seem that such an This confirms that the tone sandhi rules apply after the IP has undergone raising to SpecDP as otherwise the presence of the IP following e might be expected to trigger tone sandhi on e. The IP- raising therefore can be assumed to take place in the syntax and not be movement at PF (i.e. after the application o f tone sandhi). 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. analysis is indeed well able to capture the diverse range o f properties associated with relative clauses in Chinese. 3.5 Extensions: Grammaticalized Complementizers in Taiwanese Before concluding this chapter centered mainly on the syntax of relative clauses in Chinese, I would like to suggest that ideas on the interaction of tone sandhi and syntax presented above in section 3.4 can also account for a second interesting phenomenon in Taiwanese— the grammaticalization o f a verb of saying 'kong2' as a complementizer in an unexpected sentence-final position. This section will also attempt to show how sentence-final particles in head-initial languages may quite generally come into being as the result of processes of grammaticalization, and addresses directly the important issue of directionality and surface linear ordering in CP-type constituents in Chinese. Finally, the section will also argue that the paradigms investigated provide evidence for Chomsky’s (1998) idea of Cyclic Spell- Out. Earlier in section 3.2 it was noted that cross-linguistically it is fairly common for certain very general verbs of communication (typically equivalents to English 2 8 It could perhaps be argued that the tone sandhi patterns here might also be explained if a slightly adapted version of the Murasugi-type structure in (70) is adopted with the relative clause base-generated in SpecDP. In such a structure no tone sandhi change would be expected on the final element in the IP preceding e as this would be occurring in a Specifier (SpecDP) and e itself would be expected to undergo tone sandhi change as it would be a D° head preceding an overt NP complement. However, earlier arguments given against such an analysis are still valid here. Because there is evidence o f movement in Chinese relative clauses, it must be assumed that a CP is projected with some element raising to SpecCP; if this is assumed to be an empty operator rather than the head-noun, such an operator will need to be c-commanded by the head-noun and so the CP should be base generated to the right o f the NP in structure (70). Consequently it should not be possible to base- generate the CP relative clause in SpecDP and, as before, a Murasugi-type structure has to be rejected 158 with permission of the copyright owner. Further reproduction prohibited without permission 'say') to undergo grammaticalization as complementizers when they occur after other more specific verbs of communication or cognitive state such as 'yell,' 'answer,' inform,' 'think' or *believe.' Frequently this occurs when a language has serial verb constructions which allow for a sequence o f two verbs of communication (one more specific the second less specific) to become reanalyzed as a sequence o f verb + complementizer, schematically as in (117): (117) Verbl Verb2 Verb(l) Complementizer shout say shout that What is of particular interest and relevance here is the position of the verb ‘to say’ when it becomes grammaticalized as a complementizer. The cross-linguistically common pattern is for the grammaticalized complementizer to occur in the same position that the earlier fully verbal form occurred in. In the many head-initial SVO languages of West Africa and S.-E.Asia which show this type of grammaticalization this means that the new complementizer will occur preceding its clausal complement. In Thai (p.c., Andrew Simpson) for example, the morpheme waa is currently both a verb meaning ‘to say’ as seen in (118) and also grammaticalized as a complementizer preceding its IP complement as in (119). The fact that waa may co occur with verbs of cognition such as khit ‘think’ in (119) no longer with its literal meaning ‘to say’ is evidence that waa has indeed grammaticalized as a complementizer in such positions and is no longer just a verb-in-series. Such an (for Chinese). 159 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. assumption is further supported by the observation that waa may now also occur after nouns as in (120), which shows that waa is no longer a verb 'say': (118) kae waa arai? you say what ‘What did you say?’ (119) khaw book/khit waa Daeng suay he say/think that Daeng be-pretty ‘He says/thinks that Daeng is pretty.’ (120) kham-phaasii waa ‘tham bun dai bun’ proverb that do good get good ‘the proverb (that) ‘If you do good, you will receive goodness.” In West African Ewe (Heine & Reh 1984: p.252) it is found that the verb be ‘say’ grammaticalized as a complementizer no longer occurs with the tense-aspect markings or pronoun prefixes which would otherwise be normal for real verbs in serial verb constructions, again indicating rather clearly that a category change from verb to complementizer has taken place. Similarly in Twi (Lord 1993: p. 176) the verb se ‘say’ occurring as a complementizer also now no longer takes verbal affixes such as negation concord which would otherwise occur with verbs-in-series, confirming as with Ewe and Thai that a category change from verb to complementizer has taken place. In Mandarin Chinese and Cantonese, Hwang (1998) argues that the same type of grammaticalization is taking place, and as Mandarin (121) shows, the verb shuo ‘to say’ occurs following a verb of cognition. As in Thai (119), this element in 160 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (121) no longer has its original verbal meaning of ‘saying’ but instead appears to be functioning as a general embedding complementizer element: (121) Zhangsan xiang shuo Lisi bu lai le (Mandarin) Zhangsan think that Lisi NEG come ASP ‘Zhangsan thinks Lisi is no longer coming.’ Examples such as (121) are important, as they show that where a complementizer/C0 is developing in Chinese, it occurs in a pre-IP position and hence conforms with the otherwise head-initial pattern in Chinese preceding its complement. As noted briefly in section 3.2, where other suggestions have been made that CP is a head-final projection in Chinese, this has been based on the occurrence of sentence-final question particles and the assumption that Chinese has no other regular instantiations of C° equivalent to English ‘that.' Here however one finds that a fairly simple equivalent to English ‘that’ is indeed beginning to occur and significantly it identifies CP as being head-initial and quite regular in its directionality. A similar pattern also occurs in Taiwanese, and one finds that the verb kong occurs following other verbs of communication and verbs of cognition as in (122): (122) A*-hui siong* kong* A*-sin m* lai A-hui think KONG A-sin NEG come ‘A-hui thought that A-sin was not coming.’ Again, as with Thai waa and Mandarin shuo, the fact that kong occurs without its normal verbal meaning of ‘saying’ with verbs of cognition strongly suggests that it 161 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. has grammaticalized away from its original verbal source. This is confirmed by the fact that the verb preceding fcong can take an aspectual suffix, while fcong is in such a position that cannot occur with any aspectual suffixes, suggesting that fcong in these instances has indeed undergone a category change from verb to some other non verbal category and now occurs as a complementizer:2 9 (123) a. goa bo* siong*-koe kong* A*-sin m* lai I haven't think-Asp KONG A-sin NEG come 'I haven't thought that A-sin was not coming.’ b. *goa bo* siong* kong*-koe A*-sin m* lai I haven't think KONG-ASP A-sin NEG come This position preceding the embedded B P in (122) is precisely where one would expect to find fcong occurring as a grammaticalized complementizer, and fcong as a new C° here seems to be fully parallel to Mandarin shuo, Thai waa and equivalents in other SVO serializing languages. However, in addition to forms such as (122), another arguably more interesting pattern is found with fcong, as briefly noted in the introduction. For no immediately clear reason, the same element fcong 2 9 Note that Mandarin (i) below is perfectly acceptable with aspectual -guo attached to ting- shuo Tiear-say.' This indicates that Mandarin shuo in the sequence ting-shuo is still verbal: (i) wo mei ting-shuo-guo ta bu lai. I NEG hear-say-Exp he NEG come 'I didn't heard that he was not coming.' Elsewhere however, Mandarin shuo occurs as a grammaticalized complimentizer similar to Taiwanese hong, and if it occurs following a verb of cognition, such as xiang think,' shuo may not be accompanied by -guo, as seen in (ii) below: (ii) wo xiang-shuo(*-guo) ta bu lai I think-say -Exp he NEG come 'I thought that he was not coming.' 162 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. also seems to occur as a complementizer in clause-^wa/ position, hence following its clausal complement, as in (124) and (125) repeated below: (124) A*hui liau*chun* A*sin si* tai*pak* lang kong* Ahui thought Asin is Taipei person KONG ‘A-hui thought that A-sin is from Taipei.’ (125) goasiong* A*-sin m* lai kong* I think A-sin NEG come KONG ‘I thought that A-sin is not coming.’ As Taiwanese like other varieties of Chinese elsewhere shows evidence of being head-initial (see here the examples in (105)), and kong otherwise does occur as a genuine grammaticalized complementizer in clause-initial pre-EP position (as in (122)), this apparent clause-final V-to-C grammaticalization of kong is rather strange and seems to go against the general headedness specification of the language. It clearly also does not correspond to any serial verb position from which kong could have naturally grammaticalized as a complementizer. In order to explain the puzzle of clause/sentence-final kong, I will shortly suggest that the canonical position of the grammaticalized comeplementizer kong is indeed preceding its IP complement as in (122) and show that there is certain rather clear evidence from tone sandhi patterns indicating that the unexpected exceptional order in (124) and (125) is one which is actually derived, via a process of IP-raising to Spec of CP. 163 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.5.1 Tone Sandhi Patterns with Kons Considering the ordering of C° and IP found in (122), one finds quite regular expected patterns o f tone sandhi. The C° grammaticalized verb kong undergoes tone sandhi in its position preceding the IP complement, this caused by a regular head- complement relation, and the final element in the embedded IP lai does not undergo tone sandhi. This is fully anticipated as sentence-final elements do not undergo tone sandhi (as seen above in (103) and other examples). Turning to (124) and (125), with the unusual ordering of IP-C° in the embedded clause, one now finds two quite unanticipated tone sandhi patterns. The first of these is that the IP-final elmement lang in (124) and lai in (125) do not undergo tone sandhi. If one assumes that the IP is the leftward complement of kong in a final C° position, this should mean that the IP and the C° are in the same tone sandhi domain and it is expected that the head-complement relation should result in tone sandhi occurring between the C° and the element left-adjacent to it in this tone sandhi domain, i.e., the final syllable in the IP, yet this doesn’t happen. The second extraordinary tone sandhi patterning in forms such as (124) and (125) is that the sentence-final element kong does in fact undergo a tone change. This is very much unexpected as no other elements in sentence-final position are known to undergo tone sandhi, the sentence being a self-contained tone sandhi domain as noted earlier when discussing example (103). Furthermore, the grammaticalization of kong might be expected to result in it either maintaining its citation tone2 or simply reverting to a neutral tone/absence of tone as is commonly 164 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. found in other cases o f grammaticalization (e.g., Mandarin, de, le and -zhe, and various functional elements in Taiwanese). However, instead of this, kong undergoes a fully regular tone sandhi change in sentence-final position. Examples such as (124) and (125) need also not be followed by any other sentence for tone sandhi to occur on kong and so it would appear that there is nothing following kong which could trigger its tonal change. Both such patterns can now be argued to have a rather simple explanation. Critically, both of the odd patterns observed in (124) and (125) are exactly parallel to those occurring in “regular” examples such as (122) and (126) below where the complementizer kong occurs preceding its complement IP: (126) A»hui liau*chun« kong* A«sin si* tai*pak* lang Ahui thought KONG Asin is Taipei person ‘A-hui thought that A-sin is from Taipei.’ In (122) and (126), as just noted, the final syllable in the lower IPs, lai and lang respectively, do not undergo tone sandhi (as expected), and kong preceding its IP complement does undergo tone sandhi (again as expected). Comparing (122)/(126) and (124)/(125) it can therefore be seen that precisely the same tone sandhi patterns occur both when kong precedes its complement L P in a regular head-initial C° position and when kong occurs finally in a rather unusual position: permission of the copyright owner. Further reproduction prohibited without permission. (127) a. kong — IP: expected order, expected tone changes (i) final syllable in IP does not undergo tone sandhi (ii) kong does undergo tone sandhi b. IP - kong\ unexpected order, unexpected tone changes (i) final syllable in IP does not undergo tone sandhi (ii) kong does undergo tone sandhi The simple conclusion from such a comparison is that kong in its unusual sentence- final position is behaving for tone sandhi purposes exactly as if it occurred in a regular pre-IP position. Syntactically in order to capture this striking parallelism it can now be suggested that IP-kong forms such as (124)/(125) are actually the result o f an IP-raising operation applying to underlying fully regular kong-W forms before they are converted into IP-kong sequences. Such a pair of assumptions allows for a very straightforward explanation of the otherwise unanticipated tone sandhi facts, as follows. Prior to IP raising, the final element lang/lai in the embedded IP in (124/125) will occur in sentence-final position and kong will occur as a regular C° preceding an IP-complement. If the tone sandhi rules are applied at this derivational point, the result will be (a) that the final syllable in the IP lang/lai does not undergo any tone sandhi change, being in sentence-final position, and (b) that kong does undergo tone sandhi, being in a head-position preceding its IP complement. Observe how the assumption o f IP-raising will explain both the unusual tone sandhi patterns in kong-fvaa\ sentences and the odd sentence-final position occupied by kong and further note that kong occurs as a regular CP-initial complementizer in embedded clauses such as (122)/(126). It might naturally be assumed that the 166 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. hypothesized IP-raising operation applies in the embedded clause in (124)/(125) converting a string such as (122) into (125). Such a derivation is schematically represented in (128): (122) goa siong* kong* A*-sin m* lai I think KONG A-sin NEG come ‘I though that A-sin was not coming.’ ( 1 2 5 ) goa siong* [ c p [iP 2 A*-sin m* lai]i kong* t; ] t _____________ I (128) IPi CP Spec C’ kong However, there is actually good reason to believe that this is not exactly how IP2 and kong become re-positioned relative to each other. Although kong might seem to bear all the hallmarks o f an embedded complementizer grammaticalized from a general verb of communication as in many other languages, further data reveals that kong in fact syntactically embeds not just a lower clause but the entire matrix sentence in which it occurs sentence-finally. The evidence that this is so comes in two forms. First of all, in sentences such as (124) and (125) it is possible to have not only a kong in sentence-final 167 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. position, but also a second kong in a regular grammaticalized embedded Comp position preceding the embedded IP, as in (129) and (130): (129) A*hui liau*chun kong• A*sin si* tai*pak* lang kong• Ahui thought k o n g Asin is Taipei person k o n g ‘A-hui thought that A-sin is from Taipei.’ (130) goa siong* kong• A*-sin m* lai kong• I think k o n g A-sin NEG come KONG ‘I think that A-sin was not coming.’ This indicates that the sentence-final kong does not originate in an embedded C° position, as this position can clearly be filled by a second distinct kong. Consequently, the natural assumption to make is that sentence-final kong is actually in the matrix C° in (124), (125), (129) and (130) and that the entire IPi (i.e., the whole sentence consisting of both clauses IPi and IP2) is raised to the Specifier projected by this matrix C°. Clear confirmation that this is true comes from the fact that it is possible to have a sentence-final kong in single-clause sentences, as in (131)-(133). This indicates that kong here can only possibly be occurring in a matrix Comp as there obviously is no embedded C° in such clausal structures: (131) A*sin m* lai kong* A-sin NEG come KONG ‘A-sin’s not coming.’ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (132) goachahng bo- khi-tai-pak- kong- I yesterday NEG go Taipei KONG ‘Yesterday I didn’t go to Taipei.’ (133) goan- lau-pe si- tai-pak- lang kong- I father be Taipei person KONG ‘My father is from Taipei.’ Furthermore, if one compares (125) which has a single sentence-final kong with (122) where kong occurs preceding the embedded IP, one finds that the interpretation of the two structures is not fully equivalent. Use o f kong in (122) essentially adds nothing extra to the meaning of the sentence, much in the way that the optional addition of the English complementizer ‘that’ adds no extra semantic content when it precedes an embedded clause. Use of sentence-final kong however does add clear extra meaning to the sentences it accompanies, and encodes speaker- related emphatic assertion of the sentence which in English can often be naturally glossed with the expression ‘I’m telling you X!’ (where X = the content of the sentence). This emphatic assertion resulting from the use of S-final kong in (134) below implies the interpretation that: “A-sin has written in his letter saying he is coming, so why do you, the person listening to me (the speaker) think that he will not come?”: (134) A--sine- phoe sia- kong- bin-a-chai beh- lai kong- A-sin GEN letter write KONG- tomorrow want come KONG ‘A-sin's letter wrote that he will come tomorrow.’ 169 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As a result o f the above observations, it can be suggested that kong is indeed a grammaticized C° element, but one which critically occurs in matrix clause positions. Quite possibly this restriction results from kong being licensed by a speaker-centred prepositional attitude (the special emphasis of kong) which can only be encoded in matrix clauses where the speaker is the clear source of the information.3 0 Assuming kong then to be in the matrix C°, the surface forms found in (125) and (131) can actually be argued to have the underlying derivation and structure indicated in (135) and (136):3 1 3 0 Note that is similar to the observation that various propositional attitude adverbs in English and other languages cannot occur in embedded contexts: (i) John said that (*/??frankly) Mary was crazy. Embedding the adverb under a higher clause subject seems to block the speaker’s control o f the propositional attitude expressed by the adverb, a licensing requirement which appears to be necessary for the use o f certain adverbs. 3 1 The analysis of IP-raising suggested here would be further supported if it could be shown that extraction of an element from the IP could not licitly occur, as extraction from within a leftward Spec position (as opposed to from within a complement position) might be expected to violate Subjacency. Unfortunately because IP-final kong is a root/matrix clause C° , such tests cannot be constructed, as there is no higher position in the clause that an element could be legitimately extracted to. Note however that it might be suggested that the unacceptability o f wh elements in kong sentences could be due to Subjacency applying to LF extraction in some way: (i) * A-sin be sia-mih kong? A-sin buy what KONG 'What did A-sin buy?' I would like to suggest though that the wh elements actually cannot co-occur with IP-final kong because kong instantiates a declarative and hence non-interrogative value of C° , kong functioning to emphatically assert the IP. If the C° is non-interrogative it will simply not be able to license wh elements (and the unacceptability o f cases such as (i) will therefore not be due to any LF Subjacency violation). In this regard note furthermore that yes/no question particles can also not co-occur with kong: (i) *A-sin u lai bo kong? A-sin AUX come Q KONG 'Did A-sin come?' Such complementary distribution o f kong and question particles can be taken as indication that kong and interrogative X elements occur as alternative competing instantiations of the same basic C° head position with kong and question particles encoding opposite semantic values— deciarative assertion vs. interrogative +Q. 170 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (135) CP Spec kong• Spec CP (136) Spec C’ kong' (135) and (136) will then allow for the basic explanation o f the tone sandhi patterns already given. Considering (135) which represents the examples considered in (124) and (125), what needed to be accounted for in (124)/(125) were the two significant facts that (a) the final syllable in IP2 lang/lai does not undergo any tone sandhi change, and (b) that sentence-final kong does undergo tone sandhi change. If one assumes that (135) is the underlying structure for (124)/(125) and that the tone sandhi rules apply to (135) before the movement o f IPi (and IP2, etc.) to SpecCPi, these two patterns are simply explained. The final element lang/lai of IP2 will be in 171 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. sentence-final position when tone sandhi changes are applied, and so no tonal change will occur in lang/lai as there is no tone-bearing syllable following it at this point. As for kong in C° of the matrix CPi, it will be followed by its complement IP[ at the point of tone sandhi application and so this will naturally cause tone change on kong. The conclusion that kong is in the matrix C° thus essentially alters nothing in the basic account of the unusual tone sandhi patterns in kong-final sentences, and the suggestion that there is IP-raising in such forms is seen to account both for the odd tone sandhi with kong and its non-canonical sentence-final position.3 2 3.5.2 Grammaticalization of Kong and Motivations for IP-movement I now turn to consider the obvious question of why IP-raising takes place in kong sentences, and also examine a little further how kong has undergone grammaticalization in structural terms. Importantly, because IP-final kong has grammaticalized into the matrix clause C°, the process of its grammaticalization must actually have been somewhat different from the grammaticalization of other verbs of saying as embedded clause complementizers noted in section 3.1.1. In the latter cases, the source of the new sequence of verb and complementizer is a serial verb construction consisting of two verbs, as diagrammed in (31). With IP-final 3 2 Concerning the question of whether other older S-final particles in Taiwanese also result from IP-raising, when one examines these (e.g. aspectual - a (Mandarin le)), one finds that they are now phonetically reduced to the extent that they no longer carry any positive tone which could undergo tone sandhi. Such a lack of possible tone sandhi does not indicate that IP-raising does not occur with these particles, and only has for effect that any (hypothetical) raising can no longer be made visible by possible tone sandhi. Essentially then it is necessary to catch a particle at a particular point in its development in order to be fully confident about its underlying syntax, Taiwanese kong being especially clear and revealing here in still having both an obviously recognisable source as the 172 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. kong however, the most likely explanation of its grammaticalization as a matrix clause C° is that this has occurred when earlier two clause structures containing kong as the higher clause predicate have over time been re-analyzed as single clause structures, as outlined in (137). In such a sequence of development, kong reanalyzed as a C° will come to occur as the C° of the single matrix clause which remains after the collapse of bi-clausal forms into new mono-clausal structures: (137) Stage 1: bi-clause structure, kong a real verb meaning ‘to say’ with an NP subject and a clausal complement: [ N P subject kong [ i p ]] Stage 2: the 2-clause structure re-analyzes as a single clause; kong deverbalizes and loses its NP subject, kong grammaticalizes as a new matrix clause C° [c p [c kong [ip ]]] Stage 3: the IP complement of kong raises to SpecCP (motivation for IP-raising discussed below): [cp [ i p ]i [c kong tj ]] Noting also that it is specifically an emphatic assertion of the first-person speaker which is communicated by the use o f kong, such a first-person restriction can be suggested to have resulted from kong in the original two-clause structure having commonly had a first person subject when used as an emphatic assertive form. As part of the grammaticalization process I suggest that the first person subject verb ‘to say’ and the positive tone which allows it to undergo tone sandhi. 173 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. specification associated with kong emphatic forms may have subsequently become re-analyzed and absorbed directly into the element kong as an inherent restriction on its use. Such a process of re-analysis has indeed been attested elsewhere in similar cases with the grammaticalization of quotative complementizers and the creation of evidential morphemes. Harris & Campbell (1995, p. 169), for example, note that the Georgian quotative complementizer metki can only be used to quote the words of the speaker and point out that metki grammaticalized from an original sequence me vtkvi which literally meant ‘I said (it).'3 3 Similarly, in many American Indian languages evidential suffixes on verbs have grammaticalized from verbs of seeing and hearing following the collapse of two-clause structures into mono-clausal forms in the same way hypothesized for kong. Examining Maricopa, Gordon (1986) notes that the addition of the suffixes - ‘ yuu and - 'a to verbs results in the interpretation that the speaker respectively saw or heard the action described: (138) lima-‘yuu dance-EV ‘He danced (I know because I saw it).’ (139) ashvar-‘a sing-EV ‘He sang (I know because I heard it).’ 3 3 Me is the pronoun ‘I’ and v-tkv-i is the first person singular subject (v-> aorist indicative (- 0 of the verb ‘say.' Note that in the case o f Taiwanese, as Taiwanese subjects can be phonetically null (i.e. pro), there is no necessary phonetically overt reflex/trace of the incorporation o f the first person specification into the reanalyzed kong (unlike in Georgian). 174 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The restriction that it is the speaker who has the visual or aural evidence for the truth o f the proposition simply results from the fact that these suffixes are derived from the first person singular verbal forms of the verbs yuu-k ‘to see’ and av-k ‘to hear’ (the prefix element [ ‘- ] being a first person singular marker). As the morphemes ‘ yuu and 'a are synchronically no longer verbs but clause-final particles, it can be assumed that the first person subject specification has become re-analyzed as an inherent property of these X° heads, restricting their use and resulting in the interpretation that it is specifically the speaker who has the visual/aural evidence for the proposition. In Taiwanese, IP-kong forms are here suggested to have developed from two-clause structures in a similar way, with kong as the higher clause verb undergoing deverbalization and incorporating a first-person speaker-related interpretation from its former syntactic subject.3 4 Assuming this much, I can now outline two possible explanations for the IP- raising which has accompanied grammaticalization of kong as a C° element, one phonological, the other syntactic. The first phonological possibility is that as kong has grammaticalized into a particle-like element, like other particles it has become increasingly more clitic-like and dependent and in need of some kind of phonological support.3 5 Normally in Chinese such support should critically come from an element to the particle’s left, as stress in most varieties of Chinese including 3 4 Speculating a little on why bi-clausal structure might collapse into simplified mono-clausal forms in this way, it can be suggested that this perhaps takes place when there is no longer any pressure to see the content o f the higher clause predicate as instantiating a highlighted discrete event. 3 5 See Bybee, Perkins & Pagliuca (1994, p. 107) for discussion o f the fact that grammaticalization frequently leads to phonetic reduction, causing phonological dependency and cliticization. 175 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Taiwanese is phrase-initial and commonly leads to encliticisation rather than the occurrence of proclitics. One potential explanation of IP-raising with kong is therefore to suggest that the tendency for functional clitic-like elements to attach to their left may directly trigger movement of the IP complement of kong to a position to its left in order to provide kong as an enclitic with phonological support.3 6 A second possible syntactic explanation of the IP re-positioning is to suggest that this movement occurs as the result of the particular informational structure of kong sentences. Recall that in section 3.1.2 it was noted that S-final kong adds to the proposition expressed in its IP complement an assertive interpretation equivalent to English: ‘I’m telling you IP!’ or ‘Why do/would you doubt IP?’ When S-final kong is used, it importantly seems to imply that the hearer may already entertain the proposition expressed in the IP, but perhaps be somewhat doubtful of it for no good reason in the speaker’s opinion. Use of kong by the speaker then expresses the speaker’s strong endorsement of the truth of the proposition, in a way similar to the use o f ‘I’m telling you!’ in English as in (140): (140) He’s gone, I’m telling you! In S-final kong sentences then the proposition encoded in the IP is a possibility which may be entertained as true by both speaker and hearer but with different degrees of certainty. In this sense the IP therefore represents old, topic-like information largely presupposed by the participants in the conversation, and the clear 3 6 See here Grosu (1988) and Giusti (1997) for clear evidence that dependent enclitic definite determiners in Romanian attract elements to D°/SpecDP in order to support them phonologically and 176 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. focus of attention and force of kong sentences lies in the assertion o f the proposition by the speaker via the explicit use of kong. Because of this topic-like property of the IP and the strong focus on the asserting act with kong, an alternative to the encliticisation account o f IP-raising is therefore to suggest that movement of the IP takes place in order to topicalize the IP, placing the IP in sentence-initial topic position and leaving kong in prominent sentence-final position where it is naturally interpreted as being in focus. In such a syntactic analysis the IP-raising in kong sentences would essentially be an operation of defocusing, or ‘p-movement’ in Zubizarreta’s (1998) terms, carried out in order that a secondary element (kong) is cast into focus in a prominent position (sentence-finally here). Both of the above two possible explanations of IP-raising in kong sentences I believe may in fact be plausible as quite general causes o f S-final particle creation in SVO languages. If bi-clausal structures may perhaps more regularly collapse into mono-clausal forms with higher clause predicates grammaticalizing into particles in the way outlined, I suggest that either encliticisation or topic-focus reasons might then in many cases possibly lead to the displacement of IP-like clausal constituents in a leftward direction resulting in the creation of S-final particles. Both phonology and information structure may therefore possibly be genuine forces underlying the frequent occurrence o f particles in S-final position in different instances. Considering the particular case o f kong however, it would seem that an explanation in terms of defocusing is most likely to be the real motivation for the IP-movement, hence that this kind o f attraction for phonological support is indeed attested elsewhere. 177 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. accounting as it does for the particular topic-focus interpretation of £o«g-final forms, and encliticisation/phonology taken as a potential trigger for the movement would seem to miss this link with the meaning of kong sentences. I therefore now assume that in the case o f kong, IP-raising does indeed take place for defocusing reasons and turn to see how such a conclusion interacts with a consideration of the derivational timing of IP-movement. 3.5.3 Evidence for PF Movement? An important part of the IP-raising account of the tone sandhi patterns in kong-final sentences has been the suggestion that tone sandhi changes are made at a particular point in the derivation of such sentences when the IP-complement of kong in C° is still in situ and has not yet been raised to SpecCP. Only if the tone sandhi rules are applied at this point can the unusual patterns be given a principled explanation in line with other tone sandhi patterning in Taiwanese.3 7 Concerning the essential nature of tone sandhi, given that tone sandhi rules alter the phonetic interpretation of an 3 7 The kong paradigm therefore seems to provide a clear argument in favour of a derivational model o f grammar, and would not seem to be easily accounted for in any fully representational approach. In a non-derivational approach, kong sentences would have the (single) representation in (i) with the IP in its surface position relating to a trace/copy following kong: (i) [[IP]; kong tj ] The problem here is that the element following kong in (i) is phonetically null and therefore should not be able to trigger tone sandhi in kong. Elsewhere it is clearly only phonetically overt elements (which furthermore must have non-neutral tone) that can trigger tone sandhi on a preceding element (hence, for example, an object pro does not cause tone sandhi on a verb). Note that it is also not possible to allow for copies of movement (as opposed to base-generated empty categories) to exceptionally cause tone sandhi as the copies left by other types o f movement such as object topicalization do not cause tone sandhi on the elements which precede them. There would therefore seem to be no obvious way to account for the tone sandhi patterns with kong without assuming a derivational approach where the overt IP triggers tone sandhi on kong before raising to its left. 178 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. element and so apply to specifically phonetic features, it is most natural to assume that such rules are indeed phonological rules and consequently apply in the PF component after Spell-Out. This being so, it can be shown that one seems to be led to the further conclusion that the hypothesized IP-raising operation itself significantly has to occur in the PF component too. The critical sequence in the derivation of kong-final sentences is that underlying forms such as (136) repeated below are created in the syntactic component and then prior to IP-raising presented for tone sandhi alternation. Assuming that tone sandhi is a phonological process, under standard assumptions it should only take place after a syntactic sequence has entered the PF component. Now, because the IP-raising operation has to take place after the tone sandhi rules have applied, it seems that one therefore should conclude that the IP-raising also occurs in the PF component and hence is importantly an instance of movement in PF rather than syntactic movement. (136) CP Spec C ’ kong Under such assumptions, the derivation of a kong-fmal sentence such as (46) can be schematized as in (141): 179 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (141) [C p [c kong [ip A-sin m lai] Spell-Out PF~tone sandhi rules apply (changing the tone on kong and maintaining the citation on lai) [c p [ c kong- [ip A-sin m- lai]]] IP-raising [c p [ ip A-sin m- lai]j [c kong- ] tj ]] At first sight, the tone sandhi patterns of Arong-final sentences might therefore seem to offer good support for the general possibility that not all movement operations are necessarily syntactic and that certain occurrences of raising potentially take place in the PF component too. However, further reflection reveals that a serious problem is also introduced by the conclusion that IP-raising occurs after the application o f tone sandhi and hence apparently in PF. Elsewhere in the past where claims have been made that movement occurs in PF, such suggestions have importantly been made when the visible repositioning of certain constituents appears to have no impact on interpretation. In such cases it is suggested that if the relevant movement is assumed to take place only in PF after the derivation has left the syntactic component, its effects will not be present in the structure presented for interpretation at LF, and the fact that the movement is semantically vacuous can be simply explained. Operations of PF movement are therefore clearly expected not to have an impact on the meaning of a sentence and not to be associated with any particular interpretation. Considering 180 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the IP-raising in kong-final sentences, such movement does however appear to be associated with a particular interpretation, and structures in which an IP is raised before kong are regularly interpreted as topic-focus forms with the IP instantiating old information and kong encoding a highlighted assertive focus, similar to other structures generated by operations of defocusing p-movement. Such a connection between IP-movement and the creation o f a particular interpretation therefore suggests that IP-raising should be assumed to take place not in PF but actually during the course of the syntactic derivation in order to be present in the input to LF. The patterns with kong consequently lead to an apparent contradiction. On the one hand there is evidence that IP-raising follows the application of phonological rules and so should be taken to be PF movement, yet on the other hand there are interpretative effects indicating that the movement should in fact be assumed to occur in the syntax. As there is no obvious way of resolving such a paradox in a traditional T- model of grammar, the patterns with kong may therefore seem to suggest that there is actually a rather different interaction between syntax and phonology than assumed such a model, and that a solution to the kong dilemma may perhaps lie in (somehow) allowing phonology greater access to mid-derivational syntactic structures. Discarding the possibility of a PF movement analysis of the kong patterns, I will now see how the conflicting properties o f kong sentences can in fact be naturally reconciled with Chomsky’s (1998) idea of cyclic Spell-Out and that the kong paradigm consequently provides good support for such an approach to the phonology-syntax connection. 181 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.5.4 An Alternative— Cyclic Spell-Out In contrast to earlier GB and Minimalist models, Chomsky (1998) suggests that there is in fact no single point of Spell-Out where the phonetic features of a sentence are fed o ff to PF and phonology, but that sub-parts o f a derivation may be given phonetic interpretation during the course of a single derivation and before a structure is finally completed. A single syntactic derivation is therefore taken to be potentially spelt-out in a number of successive cycles which Chomsky tentatively identifies as CPs and vPs, the “phase” constituents. Once such a phase constituent has been syntactically created, the suggestion is that it may possibly also be spelt-out phonetically before being merged into a higher syntactic unit. The tone sandhi patterns investigated here can be argued to provide good evidence in support of such a cyclic Spell-Out approach and also allow for a better understanding of certain aspects of the process of cyclic Spell-Out.3 8 The critical patterning in kong-fvnal sentences in need of some account is the fact that the IP- movement seems to have to take place after the application of a phonological process, the tone sandhi changes. This led us above to the initial hypothesis that IP- raising perhaps takes place at PF, but such a possibility was then rejected on the grounds that the movement seems to be associated with interpretative effects. If a model incorporating cyclic Spell-Out is however adopted and it is assumed that sub parts of syntactic structure may be given phonetic interpretation mid-way in the 3 8 Thanks go to both Joseph Aoun and David Pesetsky for pointing out how the tone sandhi patterns might be considered evidence for cyclic Spell-Out. 182 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. course of a derivation, a rather simple second explanation for the sequencing of tone sandhi and IP-raising automatically becomes available which does not require the assumption o f movement at PF. Significantly the element kong is taken to occur grammaticalized in C° and hence instantiates the head of a phase-type constituent, CP. It can therefore be suggested that after construction of the phase CP with kong in C° merged with its complement IP to the right (i.e. [cp kong [ip ....]]), this sequence is given phonetic interpretation and spelt-out in PF. Entering PF the lone sandhi rules will apply to the sequence and cause a tone sandhi alteration in kong but not in the final element in the IP, resulting in the surface attested tone sandhi patterns. Following this, the syntactic derivation will then continue, with the IP undergoing raising to a Specifier position to the left of kong. On completion of the full syntactic derivation, the sequence will then be spelt-out and will surface with the linear sequence [[ip...] kong]. In such a cyclic Spell-Out approach the IP-raising therefore occurs as a fully regular syntactic movement, and movement at PF importantly need not be assumed. The apparent paradox that IP-movement must take place in the syntactic component yet after the application of certain phonological rules can consequently be captured rather simply with the idea of cyclic Spell-Out, and to the extent that only such an approach seems able to capture the patterning found with kong, the kong paradigm then clearly offers good support for such a view of syntax and phonology. The Taiwanese tone sandhi patterns also allow for certain further conclusions about the actual process of cyclic Spell-Out and a finer understanding of the nature 183 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. o f phase constituents. A first point concerns the input forms to cyclic Spell-Out. Quite generally, Chomsky (1998) suggests that there is a distinction between Specifiers that are semantically selected by a head and “extra” Specifiers which it is argued are licensed with the categories C, T and v in addition to any selected external argument (EA). Non-selected Specifiers of this second type are taken to host the subject (SpecTP), raised w/i-phrases (SpecCP), and shifted objects (SpecvP). Projections of the “core functional categories” C, T and v are accordingly schematized as in (142), with H being the head, YP its complement, EA a semantically selected Specifier, and XP the extra non-selected Specifier: (142) [ XP [ (EA) H YP]] The outer Specifier XP is furthermore suggested to be a position which is critically visible to syntactic heads which occur higher than a CP or vP phase, allowing for an element in XP to raise to satisfy EPP requirements of a higher head. Elements inside the inner pair of square brackets in (142) are taken to be largely invisible to higher positions due to the opacity of phases (“phase impenetrability”). The outer Specifier is therefore a position which is in a sense importantly not inside the phase proper and not part of the phase’s core. Turning back to Taiwanese and kong-sentences now, it has been suggested that the IP complement of kong raises to a Specifier associated with kong after the sequence kong-lP has been spelt-out. Such a Specifier (SpecCP) is not semantically selected and is therefore of the extra “outer” type just described (XP in (142)). It can therefore now be argued that the input to cyclic Spell-Out may 184 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. quite possibly be the inner core of phases consisting o f the head of a phase, its complement YP and any external argument Specifier (EA), but not necessarily a phase’s outer phase-peripheral Specifier XP. Such a Spec position is perhaps created only after the inner core o f a phase has been sent to Spell-Out. In kong sentences then, the inner core of the phase headed by C° (kong) is constructed resulting in the linear sequence [kong IP/TP] and then this is spelt-out phonologically, critically also undergoing tone sandhi alteration at this point. Following Spell-Out of the inner core of the phase, an outer Specifier position is created and the IP (TP) complement of kong is moved to this position. Finally the full and final syntactic structure is presented to the phonological component again and the linear order [IP/TP kong] is pronounced.3 9 This sequencing is now diagrammed in (143) below (using example (46) again): 3 9 It can be assumed that such an end-of-derivation re-presentation of the completed syntactic form to the phonology will not result in any second application o f tone sandhi rules and that tone sandhi alterations occur only once to any phase. 185 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (143) Syntactic creation of the inner core of phase headed by C° kong: [ kong [ip/tp A-sin m lai ]] Spell-Out of the inner core + application of tone sandhi rules: [ kong* [ip/tp A-sin m* lai ]] Syntactic raising of the output of mid-derivational Spell-Out — > IP/TP raising to outer phase-peripheral Spec of the phase CP: I [c p [ ip/t p A-sin m* lai ], kong* t; ] Final syntactic form is pronounced (as above) Such conclusions about the input forms to cyclic Spell-Out are further strengthened and confirmed by an independent pattern found in English, the interaction of w/j-movement and sentential stress discussed in Bresnan (1971), which largely anticipates the idea of cyclic Spell-Out. Bresnan convincingly shows that w/i-phrases which appear raised in surface forms in fact behave as if they were in situ for purposes of sentential stress assignment. Bresnan notes that whereas sentential stress is normally placed on the final element in a sentence, in w/i-questions and relative clauses it is placed on a raised w/z-phrase, as in (144), with ‘what books’ receiving the sentential (non-contrastive) stress: (144) John asked what BOOKS Helen had written. 186 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Bresnan argues that in order to explain the stress on the w/z-phrase and the lack of stress on the sentence-final verb, sentential stress must be assigned when the wh- element is in situ in sentence-final object position prior to raising to SpecCP. As sentential stress is a phonological rule and this must apply before syntactic raising of the wA-phrase to SpecCP, Bresnan concludes that phonological rules apply to each transformational cycle in syntax before further syntactic operations occur in higher cycles, and that phonology will therefore be interwoven with syntax in a single derivation (i.e. there is cyclic phonological Spell-Out). Here I can point out two significant points relating to the wh data Bresnan presents. First o f all, if sentential stress as a phonological rule is naturally applied to a CP constituent, then importantly it applies to the CP before the SpecCP position is created by raising of the w/z-phrase (i.e. sentential stress applies to the object w/z-phrase in its in situ position before any raising). This therefore seems to result in the same conclusion arrived at on the basis of Taiwanese IP-raising that the mid-derivational input to Spell-Out and phonology is indeed the inner core of a CP phase without its external outer Specifier position.4 0 A second important point results from a comparison o f Bresnan’s patterns with object topicalization in Taiwanese, a construction whose tone sandhi patterns independently require some re-consideration here. As mentioned in footnote 37, and seen in example (109) repeated here below, tone sandhi is not triggered in the verb which precedes the object in the underlying form of an object topicalization sentence (i.e. fcong in (109)): 4 0 As with Taiwanese IP-raising, there are also clear interpretational effects associated with 187 permission of the copyright owner. Further reproduction prohibited without permission. (109) [tai-oan«oe]i [goan* lau»pe] be* hiao* kong q Taiwanese I old-father not know speak. 'Taiwanese, my father can't speak.' If it is assumed that objects are necessarily topicalized to the same SpecCP position that IP-raising targets in Arong-final sentences, this lack of tone sandhi in the verb would be rather surprising. One would expect that the object would first trigger tone sandhi on the preceding verb during cyclic Spell-Out of the CP phase and then undergo raising to the phase’s outer Spec. Because tone sandhi does not however occur in the sentence-final verb it can be suggested that this may then indicate that object topicalization actually does not target SpecCP but some other lower adjoined/Focus-phrase position located in the inner core of the CP phase, and that this will explain the lack of tone sandhi in the verb. Any topicalization/focus-raising to a position lower than C°/SpecCP will critically take place before the CP phase is spelt-out and objects raised and phonetically interpreted in such a higher position will consequently not be able to cause tone sandhi in the lower selecting verb. Good empirical support can also importantly be given for such an explanation o f the lack of verbal tone sandhi with object topicalization. If it is assumed that IP-initial kong is grammaticalized as an embedding C° in subordinate clauses as argued in section 3.1.1, this allows one to test whether object topicalization occurs to a SpecCP position preceding kong in C° or to an adjoined/Focus position following C°. As seen in the contrast in (145) and (146) below, object topicalization can legitimately English w/2-movement indicating that it clearly cannot be analyzed as PF movement and that a cyclic 188 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. occur only to a position below kong in C° and consequently inside the CP’s inner core: (145) A*-sin siong* kong* fhit* pun* chhehl; A*-hui m* be tj A-sin thinks C that CL book A-sin NEG buy ‘A-sin thinks that A-hui doesn’t want to buy that book.’ (146) *A*-sin siong* [hit* pun* chhehji kong* A*-hui m* be tj A-sin thinks that CL book C A-hui NEG buy The lack of parallelism between IP-raising and object topicalization with regard to tone sandhi change on the sentence-final element therefore has a reasonable and simple explanation. It also has an interesting consequence when explored a little further in comparison with English w/z-movement and sentential stress patterns. Note that Bresnan’s w/z-sentential stress patterns could in fact be given a slightly different explanation from the one offered immediately above. Supposing that the input to cyclic Spell-Out could possibly be phases of either CP or vP type, it could be suggested that sentential stress is actually assigned to an object w/z-phrase when vP rather than CP is inputted to cyclic Spell-Out, the object w/z-phrase occurring unraised in vP-final position at such a point and hence in the necessary position to be assigned the relevant stress.4 1 The patterns found with Taiwanese object topicalization now importantly seem to exclude this as a possibility and suggest the conclusion that only CP phases can occur as the input to cyclic Spell-Out. Spell-Out approach is therefore necessary instead. 189 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The reason for this is that if vP phases could occur as the input to cyclic Spell-Out, one would expect (incorrectly) that Taiwanese topicalized objects would indeed be able to trigger tone sandhi on their selecting verbs, as at the hypothetical point of vP cyclic Spell-Out, such objects would occur in situ following the verb in VP.4 2 The fact that tone sandhi does not however occur in the verb in such cases therefore clearly suggests the broad conclusion that phonology has access to mid-derivational syntactic forms only at the clausal level after CPs have been constructed, and does not apply directly to smaller syntactic cycles such as vP phases.4 3 The Taiwanese kong paradigm thus generally both adds interesting positive empirical support for the idea of cyclic Spell-Out itself and also allows one to understand more precisely what may be involved in such a process, indicating that 4 1 Note that Chomsky (1998) suggests that prior to w/z-movement to SpecCP, w/z-phrases may have to raise to SpecvP. However, if such an outer Spec is not created until after the vP phase has been interpreted by cyclic Spell-Out as argued above with CP, then a w/z-object will indeed still be in situ at the point that cyclic Spell-Out may hypothetically apply to a vP. 4 2 Again, as noted in footnote 19, raising of an object to SpecvP and higher positions should only come after the vP is spelt-out. 4 3 Such conclusions also have a further potential consequence for the hypothesis in Chomsky (1998) that w/i-movement to SpecCP occurs cyclically via SpecvP. The latter assumption seems to lead one to expect that when the inner core o f a CP phase is phonetically interpreted by cyclic Spell- Out, an object w/z-phrase will occur raised and phonetically spelt-out in SpecvP, as schematized in (i): ( i ) [cp [ t p Subject [v P ObjectwH [vp V ]]]] This however raises a problem for the assignment of sentential stress to the object, as the object is no longer CP-final and in the appropriate position to receive sentential stress. In order to maintain the claim that w/z-movement does proceed via SpecvP, it will have to be assumed that this movement occurs after Spell-Out of the inner core o f the CP and that a w/z-object will actually be fully in situ when sentential stress applies to the CP phase as in (ii): ( i i ) [ c p [t p Subject [v P [ v p V ObjectW H ]]]] 190 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the input to cyclic Spell-Out is a clausal CP constituent prior to the creation of its external Specifier position.44 3.6 Concluding Remarks This chapter set out to provide an account of the basic architectural properties of relative clauses in Chinese and focused primarily on how the noun-head, the ‘linking’ element de and the relative clause syntactically combine together. Rejecting analyses in which the relative clause is taken to be an exceptional leftward complement to de for the reason that the directionality of selection should (ideally) 4 4 If one assumes a more articulated structure in the left periphery/C-domain, perhaps as in Rizzi (1997), it might be possible to suggest that IP-raising in Taiwanese takes place to a Specifier position which is higher than the Specifier of the projection headed by kong. If this were to be so, one needs to ask to what extent the conclusions reached here might possibly be different. I believe that that the main conclusions would essentially not be much changed, and largely just be re-presented with a somewhat different labelling. The basic thrust of the argumentation has been to suggest that the input to cyclic Spell-Out is a clausal constituent which is actually (just) less than a full CP— a CP lacking an outer Specifier in the terms used here. If one now concludes that kong is perhaps the head of a Mood/QP which encodes the assertive-declarative/interrogative force of the clause (kong occurring in complementary distribution with other Q-morphemes and not allowing for the licensing of w/j-elements, see footnote 37, and if its IP complement perhaps raises to a higher TopP/CP, then the generalisation in essence remains as before but just makes use of different terms: the input to cyclic Spell-Out is a clausal constituent which is somewhat less than a full clause, being a Mood/QP and not a full TopP/CP. I believe however that there is actually a good reason for wishing to maintain the original generalisation in the text. What needs to be captured is the observations in (i) and (ii) below: (i) The XP input to cyclic Spell-Out in Taiwanese is optionally headed by an X° (kong) which encodes declarative force in alternation with interrogative Q-morphemes (and hence occurs in C0 /Q°/Mood°) . (ii) The XP input to cyclic Spell-Out in English cannot be the full XP headed by the X° (C0 /Q°/Mood°) which encodes interrogative force, as this input is formed and spelt- out before w/i-phrases undergo raising to Spec of the XP. If it is reasonable to assume that cross-linguistically there is a uniform input to cyclic Spell-Out, the two generalisations above indicate that this input form must be (at least) an XP headed by a declarative/interrogative head (due to the Taiwanese evidence with kong), but that it cannot be a full XP of this type (due to Bresnan’s English wh patterns). An obvious way of capturing these two generalisations is therefore to maintain the suggestions in the text that the input to cyclic Spell-Out is indeed a CP (or perhaps a Mood/QP) which has not yet projected its external Specifier position. 191 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not be subject to different parameter settings within a single language, a variety of evidence was presented in favor of adopting an analysis o f relativization in Chinese along lines outlined in Kayne (1994), Vergnaud (1985) and Simpson (1997). It was suggested that the apparent exceptionality of relative clauses in Chinese (i.e., the fact that they are prenominal when no other V-O languages seem to have prenominal relative clauses) should be related to exceptional properties o f the lexical item de and that following Simpson (1997) in essence, de should be analyzed as an enclitic determiner triggering leftward IP-raising for phonological support. I then went on to show how such an approach is preferable to base-generated accounts of superficially similar relative clause structures in Japanese and pointed out a variety of differences in the patterning of relative clauses in the two language types which suggest that a fully unitary analysis is not in fact appropriate. Finally I presented a set of evidence relating to tone sandhi phenomena in Taiwanese which added further empirical support to the IP-raising view of relativization. I also developed the idea that such IP-clausal movement may be further widespread in Chinese and in the case of kong- final sentences provide support for Chomsky's (1998) idea of Cyclic Spell-Out. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4. De in Focus Sentences: from D° to T° 4.0 Introduction This chapter sets out to investigate the syntactic and categorial status of the element de found in clause-final position in Mandarin Chinese cleft-type sentences such as (1), and also attempts to account for the alternation found in northern dialects where the object optionally appears positioned after de as in (2): (1) wo shi zuotian mai piao de I BE yesterday buy ticket DE 'It was yesterday that I bought the ticket.' (2) wo shi zuotian mai de piao I BE yesterday buy DE ticket 'It was yesterday that I bought the ticket.' Structures such as (1) have been the subject of a number of pieces of research in recent years, e.g., Chiu (1993), Huang (1982), D. Shi (1994) among many others. There and elsewhere it is noted that shi-de sentences consistently give rise to interpretations similar to English clefts, with the focused element commonly following the copula shi and frequently being an adverb or PP referring to the time or place where some event has occurred, as for example in (3):1 1 Various other options are also attested, though not so frequently, e.g., focussing of a whole IP including its subject when shi precedes the subject, focussing of the entire VP, and focussing of the object (see the above-cited references and section 5 of the present chapter for further details). 193 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (3) ta shi zai Zhongguo xue Yingwen de he BE in China study English DE 'It was in China that he studied English.' Most research on the shi-de construction has centred itself on the focus properties of such sentences and has attempted to offer accounts o f how the focus interpretation may be syntactically encoded. In general this has led to a concentration on the function of shi and various suggestions that LF movement of the focus may be involved.2 Comparatively little attention has however been given to the role and status of the element de in the construction, possibly due to the fact that de may sometimes seem to be optional in its occurrence, and to date there has not been any serious discussion of the alternation illustrated in (1) and (2). Such apparently optional occurrence of the object either before or after de is puzzling as there is no obvious interpretative difference triggering the alternation and purely optional, unmotivated movement should not occur under current Minimalist assumptions. This chapter suggests that a study of the role played by de and the alternation found in examples such as (1) and (2) leads to a better understanding of the shi-de construction and the interesting conclusion that de is currently undergoing a significant re-analysis. It is argued that de is changing category from an original source as a D° element to become a new past tense instantiation of T°, and that the reason for such a shift is in large part the increase o f a past time conversational implicature strongly present in shi-de forms. Syntactically, such D-to-T conversion Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is suggested to be an example o f 'lateral grammaticalization,' a process in which a functional head from one type of syntactic domain may under appropriate circumstances undergo re-interpretation as an equivalent functional head in a second domain, D and T here both being elements which (potentially) assign deictic reference to their complements and therefore having largely corresponding functions in the nominal and clausal domains. The chapter also presents a variety of evidence suggesting that de is actually still ambiguous at present and in different instances may potentially instantiate either tense or a D° head, this having direct effects on a number of syntactic phenomena. Speakers are therefore argued to significantly maintain a dual analysis of de in the current period of change, with different underlying structures being possible depending on the temporal interpretation of de in shi-de sentences. The organization of the chapter is briefly as follows. Section 1 focuses on the alternations in (1) and (2) and argues that the only plausible explanation of the full patterning observed is that de is undergoing movement to the verb. Considering further the interpretation of shi-de forms and the interaction of this with object positioning, it is proposed that de is being re-analyzed as a past tense element, raising as a clitic to the verb from a T° position. Section 2 then considers S-final non-past occurrences of de and suggests that these however occur in CNP structures, showing how such a second dual analysis of de can account for a number of otherwise conflicting patterns in shi-de forms. Section 3 then goes on to present a 2 As for example in Chiu (1993) with LF movement of the focussed XP to Spec-ShiP, 195 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. formal syntactic account of the hypothesized re-analysis of shi-de structures and provides further justification for the route of grammaticalization suggested as well as an examination of the structural position of shi in past tense shi-de sentences. Section 4 considers the D/T relations from a general cross-linguistic perspective and the idea of cross-domain grammaticalization suggested by D-to-T re-analysis. Finally the chapter is closed with an examination of how focus is formally encoded in shi-de sentences. 4.1 Object-Z)e Repositioning in the Shi-de Construction The important alternation illustrated in (1) and (2) above can be schematically represented in simplified form as (4a,b) below, which highlights the part of the structure in need of some explanation, i.e., how the object and the element de come to be repositioned with respect to each other: (4) a. V - Ob - de b. V - de - Ob Concerning the relation of these two orders to each other, given that the order in (a) is found in all Chinese dialects, while that in (b) is more restricted in its occurrence, a first fairly natural assumption that can be made is that the ordering in (b) should somehow be derived from the more basic order (a). It can also be noted that the order in (a) diachronically precedes that in (b), again suggesting that the (b) order has Huang (1982) with shi as an adverb undergoing LF raising together with the focus, and D. Shi (1994) for a similar LF-related idea. 196 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. been derived from the (a) order. Assuming this much, there seem to be three possible ways of analyzing the optional positioning of the object relative to de, as now examined. A first fairly obvious possible way of relating (b) to (a) is to suggest that (b)- type surface forms might result from underlying (a) forms via simple rightwards movement of the object over de to clause-final position, as schematized in (5): (5) V ti de Obi Such a possibility is however unlikely to be right, for a number of reasons. Rightward object extraposition of this kind is cross-linguistically observed to occur predominantly when an object is either heavy or focussed, as e.g. in heavy NP shift in English or finite clausal extraposition in Hindi and German: (6) John gave tj to Mary [everything he possessed]/*itj (7) Er hat tj gesagt, [dass er heute kommen wiirde]i (German) he has said that he today come would 'He said that he would come today.' In shi-de constructions when the object occurs following de, it is however neither necessarily heavy nor focussed and in fact the opposite is actually. In examples such as (2) the post-<ie object cannot indeed be focussed as the focus is here automatically interpreted as being the adverbial element directly following shi. Furthermore a post-cfe object will by preference actually be light rather than heavy for reasons to do with the representation o f old information. Specifically, as the focus in examples 197 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. such (1-3) is taken to be the adverbial/PP element directly following shi, any object present either preceding or following de will be part of the presupposition and so necessarily old information; as such, it will normally be represented by a simple bare (hence light) NP rather than a longer descriptive form of the type which might naturally occur when an NP is introduced for the first time as new information (as for example is often the case in English heavy NP shift). Informants furthermore indicate that when the object is a clausal complement and necessarily somewhat heavy, they in fact strongly prefer for it to be placed before de as in (8) rather than 'extraposed' as in (9):3 (8) ta shi zuotian shuo [ta bu xihuan Mali] de he BE yesterday say he not like Mali DE 'It was yesterday that he said that he didn't like Mary.' (9) ?(?)ta shi zuotian shuo de [ta bu xihuan Mali] he BE yesterday say DE he not like Mali 'It was yesterday that he said that he didn't like Mary.' It therefore seems rather unlikely that (b)-type forms result from any rightwards movement o f the object from its post-verbal base position in (a). This being so, a second possibility to account for the order in (4b) may be to posit that it is actually the verb which is undergoing movement in (4b), raising leftwards to adjoin to the element de which would then be base-generated in some kind of higher functional head as indicated in (10). Such an approach would also have to assume 3 Note that the contrast between (8) and (9) again indicates that the neutral underlying base form should be (4a) rather them (4b). 198 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. that there is movement in (4a) of the whole VP containing the verb and its object as in (11): (10) Vi-de tj Ob (=4b) (11) VPj-de t; (=4a) Such a possibility is however also unlikely to be correct. If de is base-generated in a clausal functional head position, it should clearly occur located above VP-adverbs, yet such elements occur preceding de (as in (2)) and cannot occur following de as shown in (12). This is clearly unexpected if de were indeed to be located in a position in the functional structure dominating the VP. (12) ta shi qunian xue de (*zai xuexiao) Yingwen he BE last year study DE in school English 'It was last year that he studied English.1 If it is therefore concluded that neither movement of the object nor movement of the verb is likely to be responsible for the (b)-type orders, a third logical possibility is that it is actually de itself which is the element changing its position, moving leftwards from a base-generated clause-final position as in (13): (13) V de4 Ob tj Given the additional observation that de prosodically attaches to the right-hand side of the verb in examples such as (2) in the way o f an enclitic element (i.e., it is not possible for there to be any intonational pause between the verb and de), such 199 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. movement could perhaps be suggested to be an instance of clitic- movement/cliticization, movement driven by the needs of an element to attach to a particular type of host for phonological support. Furthermore, just as clitics may over time often show a change in the host they target for attachment, here it could be suggested that the alternation in (4a) and (4b) results from de as an enclitic simply switching to attach to a different, more restricted phonological target. In the more widespread/earlier pattern in (4a) it could be assumed that de in clause-final position unselectively cliticizes to whatever element is present clause-finally (i.e., the verb or a DP object), whereas in (4b) it could be suggested that de is now showing signs of coming to be more selective and deliberately targeting the verb as its host, hence moving over any clause-final object in order to encliticize to the verb. This kind of behavior and a gradual narrowing of the target for attachment is noted to be quite typical of the development of clitics (see e.g., Spencer 1991), such elements often becoming more selective in what they attach to over time and ultimately developing into affixes morphologically attached to just a single type of host. This enclitic hypothesis o f the alternation in (4a/b) which suggests that de in pattern (4b) moves to specifically target the verb as a phonological host can now importantly be shown to be supported in a strong way by the behavior of de in double object constructions (DOCs). In DOCs, in addition to the fully S-final positioning of de, it can be noted that the order in (14/15) is significantly also possible with de preceding both the direct object and the indirect object: 200 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14) NP shi Adv/PP V de IO DO (15) wo shi zuotian gei de tamen san-ben-shu I BE yesterday give DE they 3.CL.book 'It was yesterday that I gave them three books.' Such patterns indicate fairly clearly that sequences where de precedes the direct object as in (4b) (i.e., V de Ob) cannot in fact be assumed to result from simple movement of the object over de as here two elements are found following de, both the direct object and the indirect object. As it is unlikely that there is an operation moving both direct and indirect objects rightwards here (and no motivation for any such movement as noted earlier), it would seem that cases such (15) instead provide strong support for the view that it is instead the element de which is undergoes movement leftwards from a clause-final base position in (4b)-type forms and that this movement of de also specifically targets the verb, in DOCs potentially raising to the verb over both the direct and the indirect objects. Given as noted that de does also exhibit clear enclitic properties in its leftwards prosodic attachment to the verb, the most reasonable explanation of the alternation in (4a/b) can now indeed be concluded to be the encliticization hypothesis, that de moves to cliticize to the verb in shi-de sentences of type (4b) and this is therefore why it ends up linearly preceding the object. Such a hypothetical process o f cliticization to the verb of a clause-final element can also be noted to have a well-documented precedent in Chinese adding further potential plausibility to a clitic analysis of pattern (4b). Many researchers 201 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (e.g., Cao 1987, Li & Shi 1997, Z. Shi 1989, G. Wu 1999 among others) have noted that historically the perfective aspect suffix verbal -le developed from a clause-final full verb liao 'to finish.' Originally liao occurred following the object of the descriptive verb as schematized in (16a). Later on however it underwent reduction and attached itself arguably as a clitic (and later still as a suffix) to the right of the verb and so now occurs between the verb and its object as in (16b). Such a path of development is clearly similar to what is argued to be taking place with S-final de— from an original clause-final position, de becomes attracted to the verb and in cliticizing to the verb moves over the intervening direct object as in (17): (16) a. V Ob liao later — ► b. V-le; Ob t; (17) a. V Ob de in northern dialects now (optionally) — ► b. V-de; Ob tj If it is indeed correct that de is specifically targeting the verb when it undergoes displacement in examples such as (2) and (15), one now needs to ask why this should be happening and what kind of verb-related clitic de could plausibly be. As noted above, when clitics come to target a single host-type rather than just a position and any category filling that position, they are characteristically close to the point at which they may develop further and be re-analyzed as morphological affixes. If this consequently suggests that de may be en route to becoming a new verbal suffix, one needs to consider what type of verbal inflection de might actually be turning into. In the case of liao/-le becoming a verbal suffix, it is widely assumed that -le now 202 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. instantiates the verbal category of perfective aspect, a category which cross- linguistically often occurs affixally attached to the verb. The element de I would now like to argue also instantiates a syntactic category which commonly occurs in affix form on verbs in a wide range o f languages. Specifically, I would like to suggest that various properties of shi-de sentences all point towards the single conclusion that de is currently in the process of becoming a past tense morpheme, and that as such it is undergoing repositioning and developing into a suffix on the verb in a way which is indeed cross-linguistically very common with tense morphemes. A first important point leading towards such a conclusion is that quite generally in shi-de sentences one tends to find a very strong preference to interpret the event described by the predicate as having taken place in the past, even in the absence of any past time adverbials. For example, (18) below only seems to allow for a past time interpretation: (18) wo shi zuo huo-che qu Beijing de I BE sit train go Beijing DE 'It was by train that I went to Beijing.' Secondly one finds that de is in fact also often necessary for a past time interpretation. For example, in (19) below if de is present, then only a past time interpretation is possible, and if de is not included a past time interpretation is actually not available: 203 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19) wo shi gen Zhangsan qu Beijing (de) I BE with Zhangsan go Beijing (DE) with de: 'It was with Zhangsan that I went to Beijing.' without de: 'It IS (indeed) with Zhangsan that I'm going to Beijing.' This indicates that although the occurrence of de might seem to be 'optional,' in fact if a past time interpretation is necessary, then de is actually obligatory with shi, as further illustrated in (20) where the presence of a past time adverbial necessitates a past time reading and this in turn forces the occurrence of de in the structure: (20) ta shi zuotian qu Beijing *(de) he BE yesterday go Beijing DE 'It was yesterday that he went to Beijing.' There is consequently a clear strong connection between the presence of de and past time/tense interpretations. Despite this strong connection however, shi-de sentences with de can in fact have non-past interpretations if this is forced by the use of non past/future adverbials together with the occurrence o f future/modal elements such as hui 'will' or yao 'will.' Note that the latter modal elements are indeed necessary for the non-past reading and it seems that a non-past adverbial on its own is not enough to license a non-past interpretation: (21) ta shi mingtian *(cai hui) qu Beijing de he BE tomorrow only-then-will go Beijing DE 'It is (only) tomorrow that he will go to Beijing.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It can therefore be concluded that the connection of de to past time interpretations instantiates a heavy preference, and that not all occurrences of de necessarily have to be interpreted as past time events. In pragmatic terms, the default tendency for past time interpretation with de has the status of a generalized conversational implicature- -a preference which is clearly strong, but which can still be over-ridden with the deliberate use of certain elements/strategies, such as the occurrence o f non-past adverbials and modals as in (21). A highly significant new piece of evidence bearing directly on the status of de when it occurs in pattern (4b) can now be added to the general patterning observed here. Interestingly it is found that when de precedes the object and is by hypothesis raised and attached to the verb as an enclitic en route to becoming a verbal suffix, it is no longer possible to use modals and adverbs to over-ride the past time implicature of such constructions any more and only a past time interpretation is possible in such a configuration: (22) *ta shi mingtian cai hui qu de Beijing he BE tomorrow only-then-will go DE Beijing Examples such as (22) which instantiate the structure in (4b) here show an important contrast with those in (18)— (21) which represent (4a). The latter cases showed that while a past time interpretation is strongly associated with the use o f de, it still effectively constitutes a preference which can nevertheless be over-ridden with appropriate means. When however de shows signs of really becoming a verbal element and is suggested to undergo re-analysis as an instantiation o f (past) tense, 205 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. moving to cliticize to the verb, at this point it seems that the past time conversational implicature (i.e., the 'preference' for past time interpretation) has actually become strengthened to the extent that it is now part o f the genuine meaning of (verbal enclitic) de and can therefore no longer be over-ridden (i.e., de in such a position can only mean past). The suggested analysis of de in the pre-object pattern of (4b) as a new past tense morpheme therefore seems to be strongly supported. It can also be noted that the occurrence o f syntactic change as the result of the strengthening of the preference for a particular interpretation is a path of development which has been argued to be a common pattern of re-analysis. Hopper & Traugott (1993) in particular suggest that the strengthening of a conversational implicature to the point where it becomes standardized and triggers a formal re-analysis is a mode of syntactic and semantic change which underlies much diachronic change.4 In the case of de, the preference for a past time interpretation can be taken to have finally strengthened to the extent that it has eventually allowed for a re-analysis of de as a genuine instantiation of the category of past tense, such a re-analysis now indeed being manifested in its movement to the verb. 4 A simple example o f this given in Hopper and Traugott concerns the development of the English connective 'since' which originally encoded only a temporal relation between two clauses as in (i): (i) Since I started playing the lottery, I have won four times. Later on a causal relation between the two clauses became a common conversational implicature: (ii) Since Yeltsin came to power, there has been nothing but economic trouble. Currently 'since' may sometimes also be used with only a causal meaning simply because of the long term strengthening of this implicature, as seen in (iii): (iii) Since you know about economics, can you give me a hand? 206 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The development of de as a verbal clitic encoding past time/tense can also be suggested to have been functionally assisted by the natural structuring of information in shi-de forms. As noted earlier, the shi-de construction commonly encodes a clear focus set off against a strongly presupposed background which often consists of the verb and its object as in examples (1-3) and (18-20). As the object is then frequently part of the presupposition and hence old information, there will be a natural tendency for representing it by means of a pronominal element rather than a repetition of a full descriptive NP form. As Chinese furthermore allows for null pronominal objects {pro or topic-operator-bound trace as in Huang 1984), shi-de sentences then frequently occur without any overt object, and also often without any overt shi, as in (23), such forms being preferred to fully-specified examples like (24) which are grammatical but generally felt to be rather awkward/over-specified: (23) (wo) jintian mai-de (I) today buy DE 'I bought it today.' (24) wo shi jintian mai nei-ben-shu de I BE today buy that-CL-book DE 'I bought that book today.' A presupposed/old object may alternatively also occur placed in topic position: A similar story underlies German weil 'because,' which developed from a relative to the English noun 'a while’ (German: eine IVeile); however, unlike 'since,' weil may now no longer be used with a temporal meaning and can only encode causality (see again Hopper & Traugott 1997). 207 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) nei-ben-shu wo jintian mai-de that-CL-book I today buy DE 'That book I bought today.' The clear result in commonly-heard forms such as (23) and (25) is therefore that the verb and de are importantly heard adjacent to each other unseparated by other overt material. Such common verb-cfe adjacency with de prosodically attaching to the verb can be argued to be a clear functional factor favoring the development of de as a specifically verbal clitic and therefore leading to and licensing its attachment to the verb even in the presence of an overt post-verbal object. As a clitic optionally attaching to the verb and being strongly associated with a past time interpretation, the possibility o f de becoming re-analyzed as a past tense morpheme therefore seems to be both natural and perhaps even anticipated as a further stage in its development. In sum then, there are a variety o f good reasons for assuming that the element de in shi-de structures of type (4b) has indeed come to be a new instantiation of (past) tense in northern dialects of Mandarin Chinese. Before we go on to see how this is formally produced in the syntax in section 3, I will first turn to consider what the source of de is in shi-de forms and also rather importantly what the identity of de may be when it occurs in full sentence-final position in the post-object pattern (4a). 4.2 Post-object D e Let us now reconsider what kinds of interpretations arise with de in the different shi- de patterns (4a) and (4b). If de occurs in the pre-object position as in (4b), it has 208 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. been noted that de has a necessary/strict interpretation as past time/tense. If de however occurs in pattern (4a) following the object o f the verb in clause-final position, there is a preference for past time interpretations o f the event described by the predicate, but non-past interpretations are also possible if certain modals and adverbials are used. When de is present with such non-past interpretations, it is obvious now that it cannot be instantiating the category of past tense. Assuming the analysis o f de as a past tense morpheme in pattern (4b) to be correct for the various reasons given, this therefore leads to the conclusion that de is in fact potentially ambiguous in its categorial status in shi-de sentences and in addition to instantiating past tense in certain instances, it must also be able to occur as some other non-tense category in pattern (4a), relating to an underlying syntactic structure which may possibly be quite different from that in past tense occurrences o f de. If we consider now what category and status de might have in these other non-past cases, it can be noted that linguists describing the shi-de construction in previous work (e.g., Chao 1968, Paris 1979, Li & Thompson 1981) have frequently identified de as the same element de which occurs in relative clauses and possessor structures introducing a modification on a following nominal as for example in (26) and (27): (26) [wo zuotian mai] de shu I yesterday buy DE book 'the book I bought yesterday' 209 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (27) Zhangsan de shu Zhangsan DE book 'Zhangsan's book1 This is indeed a reasonable assumption and cross-dialectical evidence within Chinese clearly favors identifying the element de in shi-de forms with 'nominal modifying1 de. In Mandarin Chinese, there are in fact a number of elements having the same pronunciation de— cleft sentence de, relative clause de, possessor/genitive de, potential de, descriptive-clause de, and resultative-clause de. These elements have a variety o f different syntactic properties and do not all derive from the same source, there being essentially two clear groups of de's. Etymologically, the cleft-sentence de, relative clause de, adjectival de, and possessor de are all written in the same character, while the potential de, descriptive-clause de, and resultative-clause de are written differently, suggesting that de in shi-de forms really does relate to nominal modifying de in terms of its origin. The latter group of elements are illustrated in (28): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (28) a. kan-de-jian look-DE-meet 'is able to see' b. ta zou de hen kuai he walk DE very fast 'He walked very fast.' c. ta qi ma qi de hen lei he ride horse ride DE very tired 'He rode a horse until he was very tired.' A survey of non-Mandarin Chinese dialects/languages also significantly reveals that the morphemes all pronounced de in (26), (27), shi-de sentences such as (1) and the group of elements in (28) in Mandarin are consistently split up and grouped in a particular way. In Cantonese, Taiwanese and Shanghainese, the morphemes corresponding to de in shi-de sentences, possessor structures and relative clauses all have a single pronunciation, whereas those corresponding to de in (28) have different pronunciations, as illustrated below in Cantonese (29), Shanghainese (30) and Taiwanese (31): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (29) Cantonese: a. kui hai tingyat maai sue ge. he BE yesterday buy book GE 'It was yesterday that he bought a book.' b. kui mai ge sue he buy GE book 'the book that he bought' c. kui ge sue he GE book 'his book' (30) Cantonese: a. maai-dak-hei buy-DAK-lift 'be able to buy' b. kui ja che ja dak ho faai. he drive car drive DAK very fast 'He drove very fast.' c. kuiduksue duk do/dak ho gwooi. he read book read until/DAK very tired 'He studied until he was very tired.' 212 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (31) Shanghainese: a. i si zonie qi Zeben ge. he BE yesterday go Japan GE 'It was yesterday that he went to Japan.' b. i ma ge su he buy GE book 'the book that he bought' c. i ge su he GE book 'his book' (32) Shanghainese: a. koe-de-ii look-DE-meet 'is able to see' b. i zou de/le lao kua. he walk DE/LE very fast 'He walked very fast.' c. i dzi mo qi de/le lao chili, he ride horse ride DE/LE very tired 'He rode a horse until he was very tired.' 213 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) Taiwanese: a. il si7 chal-hngl khi3 jit8-pun2 e5. he BE yesterday go Japan E 'It was yesterday that he went to Japan.' b. il be2 e5 chhe3 he buy E book 'the book that he bought' c. goa2 e5 chhe3 I E book 'my book' (34) Taiwanese: a. chau2 e7 kin2 run can fast 'be able to buy' b. il khuil chhial khuil kah4 chinl kin2. he drive car drive KAH very fast 'He drove very fast.' c. il thak8 chhe3 thak8 kah4 chinl thiam2. he read book read until very tired 'He studied until he was very tired.' Through all of Cantonese, Shanghainese and Taiwanese, the morpheme occurring in the cleft-focus equivalent to Mandarin shi-de has therefore the same pronunciation as the element used in relative clauses, adjectival structures and possessor structures, and is different from the morpheme occurring in potential verb constructions, the descriptive verb construction and resultative clauses. It can therefore be concluded 214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with reasonable certainty that clause-final cleft de is indeed in its origin nominal modifying de. Returning to consider the underlying structure of shi-de forms now, if the de in such constructions has indeed developed from nominal modifier de, given that this latter element regularly precedes a nominal/N one might expect that there would be some kind of (phonetically null) nominal projected in shi-de forms too. Here it can be noted that in relative clause and possessor structures in Chinese it is not uncommon for the head noun following de to be omitted if it is anaphoric and/or can be recovered from the discourse context, as in (35), indicating that phonetically null/deleted' head nouns are most certainly possible with de: (35) na shi wo zuotian mai de (dongxi) that BE I yesterday buy DE (thing) 'That's what/the thing/the one I bought yesterday.' Shi-de sentences are however unlikely to be simple covert relative clause structures with a deleted/empty head noun, as the interpretation of overtly-headed relative clauses following the copula shi is rather different from shi-de forms and such sequences lack the cleft-like focus interpretation present in shi-de sentences. A second more likely possibility can be suggested to be that shi-de forms are instead more closely related to CNP structures of the 'the fact/news that IP' type where the same de occurs linking a complement clause to a following nominal, as for example in (36): 215 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) [ipta bu hui lai]-de xiaoxi he not will come DE news 'the news that he will not come' In an earlier examination of shi-de sentences, Kitagawa & Ross (1982) indeed propose an analysis along such lines and suggest that there is a phonetically null head-noun following de with an interpretation something like '(the) situation.' Such a proposal is inspired by the observation frequently made in the literature (e.g., in Chao 1968, de Francis 1963) that shi-de sentences always seem strongly linked to the direct discourse setting and function to clarify information relating to some aspect o f the discourse situation which is obvious to both speaker and hearer (for example, explain when or where some obvious, presupposed event has taken place). Kitagawa & Ross suggest that Chinese shi-de sentences (and equivalents in Japanese) therefore have interpretations something like the second gloss given for (1) repeated below: (1) wo shi zuotian mai piao de I BE yesterday buy ticket DE 'It was yesterday that I bought the ticket.' OR: 'As for me, the situation is that I bought the ticket YESTERDAY.' Quite generally, given that nominal-modifying de is indeed the most natural default source to presume for de in the shi-de construction, it is not unnatural to suppose that speakers do in fact also assume the presence o f some kind of semantically very light N following de in the way that Kitagawa & Ross essentially 216 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. suggest. Here it can also be noted that in an equivalent cleft-like construction in Burmese, a 'dummy' head-noun haa is actually physically present in the structure. This element elsewhere occurs as the nominal complement of the demonstrative dii 'this' meaning 'this one,' and is either pronounced as haa or collapsed with dii as the form daa, as shown in (37): (37) a. dii haa 'this (one)' b. daa 'this (one)' In clefts, haa regularly collapses with the non-irrealis morpheme teh to result in taa: (38) dii-nee weh taa paa today buy TAA Pol 'It was today that I bought it.' There is consequently also certain comparative empirical support for the possibility that Chinese shi-de clefts may contain a kind of light N as suggested. An possible alternative account of the source of de in shi-de forms which still recognizes the obvious connection to nominal modifying de but avoids the assumption of any empty N-head following de might instead be to follow Paris (1979), and Li & Thompson (1981) rather than Kitagawa & Ross, and suggest that nominal modifying de in both CNPs and in shi-de forms is essentially just a (clausal) nominalizer. In shi-de sentences it could then be assumed that de simply functions to nominalize the clausal constituent preceding de, and unlike the case of relative clauses and other CNPs such a clausal nominalization might possibly not be syntactically combined with any following (null) head noun. 217 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Either of the above plausible assumptions about the original source of de in shi-de forms can now be noted to have interesting consequences. First of all, it is possible to observe parallels between shi-de forms and the behavior and interpretations of certain other clear nominal/nominalization forms elsewhere, suggesting that the shi-de construction may well be rooted in an original nominal(ization) form and that this may then have possibly influenced the way that shi-de structures have come to be interpreted. Secondly, a nominalization/CNP analysis of current non-past shi-de forms allows for a principled account of various restrictions on the occurrence of w/z-adjuncts and adverbial modification found differently in shi-de forms in past and non-past interpretations and leads to conclusions about differences in the structures which may underlie shi-de sentences. Considering the particular interpretation associated with the use of de, de is essentially argued to be pragmatically appropriate when it marks a structure in which a focused constituent (following shi) is set off against a strongly presupposed background event. In this sense, the use of de can be suggested to provide a 'guarantee' of the occurrence of the background event, clearly signaling that the relevant event either has already taken place in the past or will certainly occur at some point in the future. This function of marking the background event as fully presupposed can now be argued to be largely responsible for the preference for a past time interpretation commonly noted with de. Clearly it is possible to have the greatest confidence and be able to guarantee that an event will occur at some point if it is in fact known that the event has indeed already occurred, and one can be 218 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. confident about the future occurrence of any event with much less certainty. De is therefore naturally found marking and guaranteeing the occurrence of past time events much more frequently than non-past events. As has been noted in section 1 though, the past time interpretation of the predicate when de follows the object (i.e., pattern 4a) still remains a preference which can indeed be over-ridden with appropriate means (non-past time adverbials and modals). From a comparative point o f view, it can now be noted that the important strongly presuppositional property associated with shi-de sentences is also found in certain nominalization forms in English, which seem to heavily imply or guarantee the occurrence of a particular event. Furthermore, as with shi-de forms this 'guarantee' most frequently gives rise to a past time interpretation. For example, in (39) the events represented by the nominalizations are most naturally understood to have taken place in the past: (39) The panel will discuss [the destruction o f the village]/[the killing of the hostages]. However, exactly as with shi-de forms, the past time interpretation would only seem to be a default interpretation, and may be over-ridden with the use of adjectives such as 'planned' and 'scheduled' which function to guarantee the occurrence of the event in the future: (40) The panel will discuss tomorrow's ??(planned) killing of the hostages. (41) They're talking about tomorrow's ?(scheduled) destruction of the bridge. 219 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Given the clear similarities between the interpretational preferences in shi-de sentences and the English nominalizations in (40-41) (and their over-ridability), and given the likelihood that shi-de forms have as their source a nominalization/nominal construction, it is tempting to see the common past time interpretation with shi-de sentences as being potentially related to the existence of an original nominal(ization) syntax, and possible that such a source structure may have to a certain extent been responsible for the past time conversational implicature arising in shi-de sentences. 5 A nominalization/CNP source of shi-de forms can now also be suggested to account for certain restrictions on the distribution and interpretation of adverbs and w/z-adjuncts in shi-de sentences. Broadly one finds two types of restrictions. First of all, w/z-adjuncts such as weishenme 'why' and zenme 'how' cannot occur following shi when there is an intended future interpretation of the predicate, as in (42) and (43): (42) ?*ni shi weishenme cai hui qu Beijing de you BE why only-then-will go Beijing DE (43) *ni shi zenme cai hui qu Beijing de you BE how only-then-will go Beijing DE Secondly, adverbs cannot occur preceding shi and modify the predicate following shi when there is a non-past interpretation of the predicate in shi-de forms, as seen in (44) and (45) below (compared with 21 above): 5 Although the precise connection between nominalization and presupposition certainly requires more formal understanding, one can note that it is nevertheless a phenomenon which surfaces in a wide range o f languages in different ways. For example, in Thai, Japanese and various other languages one finds that predicates which embed /active complements (whose content is hence definite and presupposed) may syntactically nominalize such clausal complements (in Thai with the 220 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (44) *wo mei-tian dou shi hui qu Beijing de I every-day all BE will go Beijing DE (45) *mingtian ta shi (cai) hui qu Beijing de tomorrow he BE (only-then)-will go Beijing DE Such restrictions can be straightforwardly explained if it is assumed that non-past shi-de forms have as their underlying syntactic form an original CNP or nominalization structure. Concerning the latter adverb cases (44) and (45), elsewhere it has often been observed that adverbs which occur external to DPs cannot quantify into DPs or modify events depicted internal to a DP. As a result of this opacity of DPs, in (46) below it is not possible to understand 'yesterday' as referring to the time when Bill betrayed Sue, only to the time of John's discussion: (46) Yesterday John discussed [Bill's betrayal of Sue] Consequently, if the sequence following shi in non-past shi-de sentences is indeed a DP/nominal structure of some kind it is indeed anticipated that adverbs external to this sequence (i.e., preceding shi) should not be able to quantify in to the predicate inside the DP. As for the unacceptability of w/i-adjuncts such as zenme and weishenme following shi in future-type interpretations of the predicate, this is again expected if the sequence following shi in such interpretations is a CNP. Elsewhere CNPs in Chinese do not allow such w/t-adjuncts to occur inside them, as shown in (47/48): nominalizer thii, in Japanese with no), whereas non-factive clausal complements are embedded as 221 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (47) *ta shi [[D P weishenme lai]-de ren]? he BE why come DE person (48) * [ d p ta weishenme/zenme piping Meiguo]-de xiaoxi] bu zhengque? he why/how criticize USA DE news not accurate Such patterns can now be shown to significantly contrast with the occurrence of adverbs and w/j-adjuncts in shi-de forms in pattern (4b) where de encliticizes to the verb and precedes the object, and the predicate accordingly has a past time interpretation. Here it is found that adverbs can in fact occur preceding shi and still quantify over the predicate following shi, as in (49), and that w/2-adjuncts such as weishenme and zenme can occur either following shi or preceding shi and in both cases refer to/question the event in the predicate following shi as in (50): (49) zuotian ta shi gen Zhangsan lai mai-de piao? yesterday he BE with Zhangsan come buy DE ticket 'It was with Zhangsan that he bought the ticket yesterday.' (50) ni qu-nianshi weishenme/zenme qu-de Beijing? you last-year BE why/how go DE Beijing 'How/why was it that you went to Beijing last year?' If the analysis of verbal clitic de in pattern (4b) put forward here is correct, such contrasts have a simple explanation. If de in pattern (4b) is an instantiation of past tense and categorially different from non-past de in pattern (4a), the presence o f de in examples such as (49) and (50) can be suggested to signal the occurrence of a simple simple (non-nominalized) clauses. 222 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. TP rather than a nominal CNP island constituent. As TPs are not islands for wh- adjuncts nor opaque constituents for external adverbial modification, it is quite expected that both w/z-adj uncts and pre-shi adverbs should indeed be able to occur in cases such as (49) and (50). The contrasts between (44-45, 47-48) and (49-50) therefore seem to add good support first of all for the general hypothesis that de may correspond to two distinct syntactic types in its past and non-past interpretations, and secondly for the possibility that significantly different syntactic structures may correspond to these different instantiations of de, in the case of non-past de this being an opaque nominal constituent which is an island for w/z-adjuncts (a CNP or perhaps a clausal nominalization), and in the case of past tense de this being instead some kind of non-opaque, non-island constituent marked by de, quite plausibly a simple TP. The above conclusions now also raise a further important question about the syntactic status of de in the S-final [V Obj de\ pattern (4a) when de has a past time interpretation. Specifically, one would like to know whether de in such cases gives rise to a past time interpretation simply as the result of a pragmatic conversational implicature, or whether de may be formally re-analyzed as a (past) tense morpheme also in these cases as with pattern (4b). Now having considered the patterns with adverbs and w/z-adj uncts we have a potential structural diagnostic for resolving this issue. If past time interpretations of de in pattern (4a) allow for the occurrence of w/z-adj uncts following shi and adverbs preceding shi, then it can be concluded that de in such cases is not associated with the projection of any opaque, island-like 223 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nominal projection, but instead relates to a different TP structure headed by de as an instantiation of past tense. If on the other hand past time interpretations of de in pattern (4a) pattern like future-time interpretations and disallow the occurrence of external adverbs and internal w/z-adj uncts, then one can conclude that past time de in pattern (4a) is indeed associated with the same CNP-type underlying structure and is only interpreted as past by conversational implicature. What one finds, as shown below in (51) and (52), is that w/z-adjuncts and adverbs in (4a) past time interpretations of de significantly show none of the unacceptability of the same elements occurring with non-past de in pattern (4a): (51) zuotian wo shi zai xue-xiao kan-jian tade yesterday I BE in school see he DE 'It was in the school that I saw him yesterday.' (52) ni shi weishenme/zenme lai xue-xiao de? you BE why/how come school DE 'Why/how is it that you came to school?' This clearly suggests the interesting conclusion that de has allowed for a formal re analysis as a past tense element projecting a TP not only in pattern (4b) but also in full S-final position. How this may be possible and what kinds of formal syntactic structures underlie shi-de forms in patterns (4a) and (4b) will now be shortly examined in section 3 below. Summarizing the conclusions made so far, in section 1 I presented evidence and argumentation that de in sequences of [V de Obj] instantiates past tense. I then 224 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. noted that as de in [V Obj de] sequences can have a non-past interpretation this indicates that de in such occurrences must be a different syntactic element from past tense de and therefore that speakers must be assumed to currently maintain a dual analysis of de in shi-de forms. Considering what this latter non-past de might be, it was noted that its most likely source is the 'nominal modifying' de found in CNP (and possessor) type structures. I then saw how the assumption that non-past de might still relate to a nominalization/CNP structure may account for contrasts in the acceptability of certain w/z-adjuncts and adverbs in past and non-past interpretations of de. Finally it has been argued that the lack o f restrictions on adverbs/w/z-adj uncts with past time interpretations of de in the S-final pattern (4a) suggests that de is also instantiating past tense here and that a non-island TP is projected in the structure rather than an opaque nominal category. In section 4 now I will consider a little more closely what kind of underlying syntactic structures are projected in the different shi-de forms and how an original nominal structure may have come to be re-analyzed as a projection of past tense. 4.3 The Syntax of Re-analysis in Shi-de Sentences In attempting to compare the structures underlying past and non-past shi-de forms and how the latter may have undergone re-analysis as a new past tense structure, I will first outline what I assume to be the underlying syntax of non-past shi-de forms. Taking de in non-past shi-de forms to be the nominal modifier de of regular CNPs, I assume as noted in section 2 that non-past shi-de forms contain a CNP structure 225 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. headed by a light N element.6 Concerning the analysis of CNPs in Chinese, chapter 3 of the thesis has defended a Kayne-style (1994) analysis o f relative clauses and other noun-complement clause CNPs in Chinese with the following properties: (53) a. De is a an element o f type D in all CNP type structures. b. De selects a rightward clausal complement, in line with the dominant head-initial direction of complement selection in Chinese. c. In relative clauses the relativized noun/NP raises to SpecCP and then the IP remnant raises higher to SpecDP. d. The motivation underlying this IP movement is that de as an enclitic determiner triggers movement of the IP element to its SpecDP position for phonological support. The derivation of a relative clause structure such as (54) was then argued to be as indicated in (55): (54) [wo zuotian mai]-de shu I yesterday buy DE book 'the book I bought yesterday’ 6 Or alternatively a headless clausal nominalization; here in the text I will generally pursue the former CNP possibility, and note that assuming the latter headless nominalization analysis may perhaps not lead to any very different results. 226 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. book; wo zuotian mai ti 'I yesterday buy tj' b. DP IPk D' wo zuotian mai ti D 'I yesterday buy tj' de .IP In the similar treatment of noun-complement clause CNPs, de in D° was suggested to select a rightward NP (rather than CP) complement, with the N° head of this NP in turn selecting a rightward IP/AspP complement as in (56), representing example (36) repeated below. The element de was furthermore taken to have the same enclitic requirements as in relative clauses, triggering raising of the IP/AspP clausal constituent to SpecDP as in relative clauses: 227 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (36) [ipta bu hui lai]-de xiaoxi he not will come DE news 'the news that he will not come' (56) DP IP/Aspi ta bu hui lai D° NP 'he won’ t come' I I de N' IP/AspP xiaoxi 'news' Turning now to shi-de forms, under the assumption that such structures in non-past interpretations contain a noun-complement clause CNP headed by a phonetically null light N as suggested, the underlying structure and derivation of examples such as (21) can be suggested to be as in (57): (21) ta shi mingtian * (cai hui) qu Beijing de he BE tomorrow only-then-will go Beijing DE 'It is (only) tomorrow that he will go to Beijing.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. V shi 'BE' AspP/IPj mingtian cai hui D qu Beijing \ NP ‘ Asp/IP Taking (57) to be the representation o f non-past shi-de forms, we are now in a position to see how such structures may be re-analyzed into rather different syntactic forms when de comes to instantiate past tense, as well as how further aspects of the syntax of past tense de may be accounted for. The central contention of the analysis argued for in sections 1-2 above is that nominal modifying de undergoes re-analysis as a new instantiation of past tense. In categorial terms it can therefore now be suggested that an original D° head of the CNP taken to occur in the source construction becomes re-analyzed as an element of type T°. Such a re-analysis of the head of the construction will then automatically result in the original DP complement of shi in non-past shi-de forms being reinterpreted as a new TP complement to shi. Concerning the rest of the CNP structure in (57), a further change which can be suggested to follow from the re- 229 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. analysis o f D as T is that the light N head present in (57) is pruned and lost from the underlying structure so that T° comes to directly select for a clausal rather than a nominal complement, as indeed elsewhere with T° elements.7 Such simplification and pruning of the NP projection can be suggested to be possible for two reasons. First, it may be possible because the N° head is a semantically light expletive-type N (similar, for example, to Japanese koto 'thing') assumed to simply fill the regular N- position of the source CNP construction and it is therefore not an element which is really required for any interpretative purposes (and is hence rather easily 'eliminatable'). Secondly, the hypothesized N may be essentially ignored and eliminated during the re-analysis of the DP as a TP as the N is phonetically covert and there is consequently no overt phonetic material which might have to be re- o analyzed as some other category. Such a hypothetical re-analysis of the structure 7 The alternative is to assume that there is no N-head present and there is just a simple clausal nominalization preceding de in shi-de forms as noted earlier. However, the assumption that there is a CNP analysis with a phonetically null N-head has the advantage that it allows for an account of the unacceptability of w/j-adjuncts in non-past interpretations of shi-de forms as discussed in section 2, CNPs being clear islands for w/7-adjuncts elsewhere. As it is not clear that clausal nominalizations would necessarily instantiate islands for the interpretation/licensing of such elements, this would seem to favor the CNP analysis over a bare clausal nominalization approach. 8 Here it can be noted that where a similar light N does have phonetic content in Burmese and a similar kind of re-analysis appears to be occurring, the light N undergoes 'fusion' with the T° element, collapsing as a single morphological form. As pointed out in examples (37) and (38), in cleft forms similar to Chinese shi-de sentences the light N haa 'one/thing' collapses with the preceding T° morpheme teh to result in taa. Significantly, when this happens, the result cannot be optionally separated out into teh and haa (*teh had) and only the fused form taa occurs. This is unlike the behavior of haa elsewhere with D° demonstrative elements where the fusion o f the D° dii and haa is quite optional and allows for either o f the forms in (37) repeated here as (i): (i) a. dii haa 'this (one)’ b. daa 'this (one)’ If a nominal-based cleft construction is indeed undergoing re-analysis in Burmese in a way similar to Chinese, it can be argued that part o f this re-analysis is the elimination of the light N via a process of incorporation of the N into the (new) T° element. Chinese, which does not have the problem of the need to re-interpret any phonetic material associated with a light N can perhaps just eliminate such an element without any prior syntactic N-to-D incorporation being necessary. 230 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. will then result in new TP forms such as (58) occurring as complements to shi in place of DPs, with the clausal complement of T° (arguably an AspP, potentially containing the perfective aspect marker -le) undergoing raising from its base position to SpecTP, essentially as in the original DP structure. Note that here I concentrate on and present the lower de-related part o f the full shi-de structure and later return to consider the upper half of the structure: ( 1 ) wo shi [tp zuotian mai piao de] I BE yesterday buy ticket DE 'It was yesterday that I bought the ticket.' (58) [zuotian mai piao], T '[yesterday bu^ ticket];' J (58) above represents pattern (4a) when de has a past time interpretation. In section 2 I argued that de encodes past tense not only in pattern (4b) but also when it appears following the object as in ( 1 ). De in ( 1 ) and other similar examples can therefore be taken to be occurring base-generated in the new T° position as indicated in (58). Concerning pattern (4b), where de attaches directly to the verb and precedes the object, here I suggest as before that this is the result o f a cliticization operation in A similar kind of fusion may however elsewhere have occurred in classical Chinese resulting in modem Chinese -zhe '(the) one who V's' as in jizhe 'reporter (one who reports/records).' One can speculate that here some nominal element incorporated into the classical Chinese D° zhi (modem Chinese de) changing the vowel quality of zhi and resulting in a new complex fused/collapsed element. 231 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. which de moves to target the verb, and that prior to cliticization and movement de in such cases is base-generated in the same basic T° position as in pattern (4a) and tree (58). The critical optionality between pattern (4a) and pattern (4b) therefore does not relate to any difference in the underlying structure projected, but is taken to relate to de targeting different hosts for cliticization: in the case of (4a), de simply criticizes to the right edge of the AspP in SpecTP; in (4b) however, de alternatively targets the verb as a more specific phonological host and so moves leftwards from its base position to do this. Such an approach to the optionality in the positioning of de which essentially ascribes it to the availability o f different possible targets for cliticization can be suggested to be potentially more appropriate as an analysis of Re positioning in shi-de forms than other pure-syntax-based possibilities. Given that the optional repositioning of de in pattern (4b) appears to result in no real perceptible differences in interpretation (with de encoding past tense in both patterns (4a) and (4b)), it might seem difficult to account for such an alternation in purely syntactic terms, as random, optional syntactic movement having no effects on interpretation is assumed to be unavailable in current minimalist approaches to syntax. Processes of cliticization driven by primarily phonological considerations are however often observed to display optionality of exactly this kind. As one example of this, Spencer (1991, p.372-373) notes that tense clitics in Polish optionally move and attach to a wide variety of phonological hosts: pronouns, complementizers, adverbs, NPs and numerals (these not necessarily occurring in any fixed clausal P2 position either), and hence that there is a considerable degree of optionality in the placement of tense 232 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clitics in this language. Elsewhere studies of the development of clitics have shown that clitics may over time gradually become more selective and switch from targeting general syntactic positions or constituents as hosts to targeting certain more specific syntactic categories, and that during the period of change apparently optional attachment to either original or new host category is possible. In the case of past tense de, it can therefore be suggested that de is currently in just such a transitional stage, tolerating attachment both to an original general syntactic host (AspP) as well as also showing signs of becoming more selective and optionally targeting the specific syntactic category of the verb for phonological support.9 10 9 This of course indicates the conclusion that Chinese is a language which has 'special' clitics, i.e., clitics which undergo movement and repositioning. Such a suggestion has in fact been previously made in Tang (1990) about the element de in descriptive and resultative constructions, which Huang (1988) notes occur in early Chinese in forms parallel to (i) and (ii). Tang (1990:250) suggests that these are formed via movement o f the clitic de over the objects shu and du to the verb: (i) du [-de shu duo] read DE book many 'to read a lot' (ii) zhong [-de du shen] get DE poison deep 'to be seriously poisoned' A similar implication that de in modern-day resultative constructions may be attached to the verb via cliticization from a non-adjacent position in a lower clause is present in work in Yafei Li (1998), who convincingly shows that the post-de 'subject' Lisi in cfe-resultatives embedded by transitive verbs such as in (iii) is actually the object of the higher clause. If it is then assumed that de originates as the complementizer o f the lower clause, it has to be concluded that it reaches its surface position via cliticization movement over the object of kua to the verb kua itself: (iii) Zhangsan kua-de Lisi dou bu xiang gan huor Zhangsan praised-DE Lisi all not want do work 'Zhangsan praised Lisi to the extent that she didn't want to do any work.’ 1 0 Note that the cliticization of past tense de to the verb resulting from a base structure such as (53) will involve cliticization o f de from the head T° to an element (the verb) inside SpecTP. Cliticization o f this type to a position contained within a higher constituent has been observed to occur in a number of instances elsewhere, for example Serbo-Croat where pronominal and auxiliary clitics attach to an element inside the constituent in SpecCP as in (i). Here the clitics mi and je occur encliticized to the demonstrative heading the DP in SpecCP: 233 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A movement-cliticization account of de in pattern (4b) can also be argued to be preferred to an alternative affixation analysis which might suggest that de is now simply base-generated on the verb as a suffix in examples such as (2). Essentially there are two types of argument against the latter possibility. The first is that it is clear that past tense de in pattern (4a) cannot be a verbal suffix as it encliticizes to whichever direct or indirect object DP occurs finally in the AspP preceding it. As it is implausible to imagine that the same past tense morpheme de has two lexical entries, one as a suffix and the other as a clitic, it would seem that a natural uniform analysis of past tense de needs to recognize it as a special clitic which (optionally) moves to the verb from a T° position. A second argument in favor of such a treatment comes from a further consideration of the double object construction (DOC). Earlier it was noted that in addition to a possible S-final position, past tense de can also occur attached to the verb in DOCs, hence both a c/e-final [V IO DO-de] sequence and a ^e-raised [W-dej IO DO tj] form occur as noted in example (15): (15) wo shi zuotian gei de tamen san-ben-shu [V-de IO DO] I BE yesterday give DE they 3.CL.book 'It was yesterday that I gave them three books.' (i) [D P Taj-mi -je pesnik] napisao knjigu that me AUX poet wrote book 'That poet wrote me a book.’ (Spencer 1991) It can also be noted that there is further clear optionality of attachment here and the same clitics may be attached to the right-edge of the DP instead of inside the DP: (ii) [D p Taj pesnik]-m i-je napisao knjigu that poet me AUX wrote book 'That poet wrote me a book.' 234 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Here it can now be noted that in addition to the patterning illustrated in (15), many speakers also allow for de to occur intervening between the indirect object and the direct object as in (52): (59) wo shi zuotian gei tamen de san-ben-shu [V lO-de DO] I BE yesterday give they DE 3.CL.book 'It was yesterday that I gave them three books.' This might initially seem to go against the verbal-encliticization analysis argued for in section 1 , as here de seems to be attaching to the indirect object pronoun not the verb. However, in addition to the clear fact that the order in (15) with de enclitic on the verb does show de to be targeting a verbal host, it can be suggested that the order in (59) results from syntactic incorporation of the indirect object pronoun into the verb prior to de attachment. Following Baker & Hale (1990) and Bresnan & Mchombo (1987) who argue for the general possibility of pronominal incorporation phenomena, it can be suggested that indirect object incorporation in cases such as (59) creates a complex verbal element to which de may subsequently attach as a verbal clitic. 11 As indirect object incorporation may be reasonably assumed to be a syntactic operation, this indicates that that de encliticization/attachment should also be assumed to occur during the syntactic derivation rather than be the product of lexical suffixation, and only such a movement-encliticization analysis of de seems 1 1 The possibility of object pronominal incorporation in Chinese would also seem to account for the fact that a pronoun but not an indefinite object is licensed in structures such as (i) below where both a nominal object and a time adverbial follow the verb: 235 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. able to account for the patterns noted here that de can occur both finally, between the 10 indirect object and direct object, and also attached directly to the verb. Here it can also be noted that such patterns distinguish de from verbal le which can only attach directly to the verb and is unacceptable if positioned after an indirect object as indicated in (60): (60) a. wo gei-le tamen san-ben-shu I give LE they 3.CL.book 'I gave them three books.' b. *wo gei tamen-le san-ben-shu I give they LE 3.CL.book The contrast between (55b) and (54) can be simply accounted for if it is assumed as suggested above that de is a clitic moving in the syntax to attach to the verb together with any incorporated indirect object, and that verbal le is (as commonly assumed) a suffix directly attached to the verb prior to insertion in the syntax and therefore prior to any possible indirect object incorporation to the verb. The cliticization-movement analysis of de's attachment in shi-de sentences can also be suggested to allow for a possible account o f the otherwise unexplained observation that when perfective aspect verbal le occurs in shi-de forms only pattern (4a) and not pattern (4b) is possible, as shown in (61): (i) wo kan-guo ta/*dianying liang-ci. I see-Exp him/movie two-time 'I have seen him/*been to the movies twice.1 236 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (61) a. wo shi zuotian mai-le piao de I BE yesterday buy LE ticket DE 'It was yesterday that I bought the book.' b. *wo shi zuotian mai-le-de piao I BE yesterday buy LE DE ticket Here it can be suggested that the encliticization of de and its attachment for phonological support may be subject to a restriction that de can only target and attach itself to a host which is phonologically 'strong' in the sense of being a syllable which bears a tone and is not destressed. Verbal le is always destressed and has lost any original tone it may have had and can therefore be suggested not to be strong enough to constitute a suitable target for de's cliticization, blocking pattern (4b) and de's movement to the verb. 13 Instead, as pattern (4a) is available as an option, de in such cases will simply cliticize to the right edge of the AspP (which does contain a strong syllable). Such an account of (60a) and (61b) in terms of the suitability of the potential phonological host for encliticization can also be argued to be supported by further patterns with de. First of all it can be noted that the unacceptability of (61b) is actually not specifically due to the fact that it instantiates pattern (4b) with 1 2 In order to account for the possibility that de may occur either directly attached to the verb or to the V+IO complex, one can assume that indirect object incorporation is not always forced to occur. Note also that there are additional restrictions on the acceptability of DOCs combined with de which relate to complex constraints on the relative definiteness values of direct and indirect object. These I do not attempt to investigate here for simple reasons of space. 1 3 The fact that le and de share the same shwa vowel sound may perhaps be another factor contributing to the unacceptability of V-le-de sequences, given that languages frequently disprefer the adjacent positioning o f distinct morphemes which are either identical in sound or have a high degree of similarity (see, for example, Boskovic 1999). 237 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. movement of de over the object; if there is no overt object present at all as in (62), it is similarly unacceptable for de to attach to a verbal base suffixed with le: (62) wo shi zuotian mai(*le)-de I BE yesterday buy (LE) DE 'It was yesterday that I bought (it)' As (62) could in theory be an instantiation of either pattern (4b) or (4a) (i.e., in syntactic terms de in (62) could either be moving to the verb or it could be attaching to the right edge of AspP), this indicates that it is the creation of a sequence of [* V- le-de\ in any way which is unacceptable and that it is not pattern (4b) which is specifically blocked by the occurrence of le on the verb. Clearly something in the surface string [*V-le-de\ (rather than its derivation) must therefore be responsible for the unacceptability of such forms, and here this is suggested to be the weakness of the suffixed base that de as a . clitic is attempting to attach to. Secondly, it can be observed that when de is attached to a base suffixed with the experiential marker guo as in example (63), guo in such cases is regularly be pronounced with its optional tone 4 rather than being destressed into neutral tone as is common elsewhere, supporting the assumption that de needs to cliticize to a phonologically strong syllable: (63) zhei-ben-shu wo shi qu-nian kan-guo-de this-CL-book I BE last year read-ASP-DE 'It was last year that I read this book.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Such patterns then seem to suggest that de can only attach to host element which is phonologically unreduced. Given that clitics are commonly defined as being elements which are syntactically independent but in need o f phonological support, and that clitics also frequently undergo movement to find an appropriate phonological host, the restrictions observed in (61-63) then add additional plausible support for the general encliticization analysis of de's attachment and repositioning. 14 Two further syntactic arguments can now also be added on here as general support for the structure in (58) and the assumption that there is movement of the AspP constituent from complement-of-To position to SpecTP. The first of these requires a reconsideration of how it is that vv/z-adjuncts can legitimately occur in past tense examples such as (47) and (48). In section 2 it was noted that if non-past occurrences of de in shi-de relate to an underlying CNP structure but past time interpretations project instead a TP, then various contrasts in acceptability between past and non-past shi-de structures with w/z-adjuncts and pre-shi adverbs could be reasonably explained, CNPs but not TPs being syntactic islands for the licensing and interpretation of such elements. If however it is now assumed that the AspP 1 4 The patterns noted here might also seem to distinguish the clitic status of de from the suffix status of other elements such as verbal le. In contrast to de, verbal le can in fact easily be attached to a weak destressed syllable, and in (i) the experiential marker guo can be quite naturally pronounced in neutral tone reduced form when followed by le: (i) wo qunian jiu kan-guo-Ie hen-duo-shu I last year then read-ASP-LE ’Last year I read very many books.' This difference between le and de may reasonably follow if le and de are indeed attached at different derivational points as indeed suggested by the patterning noted with DOCs in (59) and (60) above. If le is attached in the lexicon as a suffix and de is attached in the syntax as a clitic, they might well be subject to somewhat different restrictions. Clitics in particular seem to need phonological support from a stronger element present in the structure and so might consequently be more likely than suffixes to require this host to be regularly stressed/not destressed. 239 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constituent in shi-de forms occurs in a specifier position as in (58), such a structure might in fact be expected to constitute an island for the licensing of w/z-adjuncts as confxgurationally it closely resembles a sentential subject structure and such constituents have indeed long been observed to be islands for w/z-adjuncts (see e.g., Huang 1982). Here it can be suggested that in order to allow for the legitimate occurrence o f w/z-adjuncts in past tense shi-de constructions it has to be assumed that at some derivational point the AspP containing such elements must also occur in some other non-specifier position which does not constitute an island for w/z-adjunct licensing. The suggestion that the AspP originates as a rightward complement to T° prior to raising to SpecTP combined with the copy-theory of movement (Chomsky 1995a) or simple reconstruction now offers the possibility of explaining the relevant lack of island effects with w/z-adjuncts here. 13 If it is assumed that the licensing of w/z-elements in Chinese occurs at the level of LF in some way (either via movement or in situ binding), it can be suggested that either the AspP clause reconstructs to its base complement position and any w/z-adjunct present inside is licensed in this reconstructed non-island constituent, or that w/z-adjuncts are licensed as part of the copy left in the complement-of-T° position by movement of the AspP. Consequently it can be argued that it is indeed necessary to assume that the AspP originates in complement position before being raised to the surface SpecTP position, and it cannot be suggested that the AspP is perhaps re-analyzed as being somehow directly base-generated in the leftward SpecTP position. 1 5 For arguments that clausal-type constituents obligatorily reconstruct see Heycock (1995) 240 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. An argument can also be made that the AspP does indeed raise from its underlying complement position to a higher specifier position. An alternative to such an assumption might possibly to suggest that the AspP is instead base-generated as a leftward complement to T° and does not occur in a specifier position at any point in the derivation. Against such an analysis which would clearly be a head-final structure out of alignment with the dominant head-initial direction of selection in Chinese, it can be suggested that the AspP arguably needs to occur in a higher specifier position in order for raising and cliticization of de to the verb to be possible. Elsewhere cliticization has been regularly found to be movement upwards rather than downwards in a tree; if the optional cliticization o f de to the verb is also naturally assumed to follow this strong cross-linguistic generalization, it would seem that the AspP containing the verb targeted by de cannot be in simple complement position as this would clearly involve de undergoing a lowering operation to its target. Rather, it would seem that such a constituent should instead be assumed to be located in a higher position in the tree in order for raising o f de to occur, such as indeed the specifier position suggested. The derivation suggested in (58) with the AspP and also Huang (1993). 241 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. constituent originating as the rightward complement o f T° and then raising higher to the leftward SpecTP position is then generally supported by a range of phenomena. 16 Finally in this section, having motivated and defended the structure and derivation in (58), I will turn to consider how this partial structure combines with the element shi to result in full shi-de focus forms and what the status of shi may be in these larger re-analyzed structures. I will also attempt to offer an answer to a general question raised by the claim of D-to-T re-analysis with de, which is: why is it that de as a new instantiation of past tense does not in fact occur with every verb having a past time interpretation? Approaching these issues, I would like to suggest that recent ideas on the syntax and structure o f tense proposed in Stowell (1996) allow for a modeling o f shi- de sentences which accurately reflects both the syntactic properties and the interpretation of shi-de forms, and it can be argued that past tense shi-de sentences are in fact largely similar to English complex perfect tense forms. An important piece of information concerning the status of shi in shi-de sentences comes from the way that A-not-A questions are composed. As illustrated in (64), significantly it is the element shi which is doubled in such question forms: 1 6 The conclusion that the AspP has to occur in both positions combined with the observation that de can cliticize to either AspP or V° has an additional consequence which needs to be pointed out here. Previously in the CNP structure (52) it was suggested that movement of the AspP is triggered by de's need to encliticize to an element in its Spec position. If past tense de can now undergo cliticization to the V° as well as the AspP, it is difficult to maintain that movement of the AspP prior to cliticization of the de to the verb is itself for cliticization purposes. Here I suggest that movement of the AspP has simply been 'inherited' by the new TP structure as a part of the general process o f re analysis and is now a 'fossilized' automatic movement no longer driven by cliticization requirements but instead formally legitimized by the occurrence o f strong v-features on T° attracting the AspP constituent to SpecTP. For further discussion of how grammars may allow the introduction of categorial features on heads as a means to legitimize movements whose initial motivation has 242 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (64) ni shi-bu-shi zuotian lai de? you SHI-NEG-SHI yesterday come DE 'Is it yesterday that you arrived?' This patterning is potentially important and revealing, as elsewhere it is generally found that the element which is repeated in A-not-A forms is a verb/auxiliary which is interpreted as being finite (i.e., the doubled element is the highest verb/auxiliary in a finite clause). Examples such as (64) therefore seem to suggest that shi is both verbal and also finite not only in non-past shi-de forms where its complement is assumed to be a DP, but also still in past-time interpretations where it has been argued that de heads a TP constituent. 17 This hypothetical co-occurrence of two finite specifications in shi-de structures (with both shi and de) is clearly in need of some explanation and possibly might seem to challenge the past tense analysis of de, suggesting that shi rather than de should be considered to be the head of TP in shi-de sentences. Here I believe that a comparison of shi-de structures with English have - en perfect tense forms such as (65) and in particular Stowell's (1996) analysis of the perfect tense now offers a simple and natural solution to this apparent problem: (65) a. John has eaten all the doughnuts, b. Mary has bought War & Peace. Set within a broad new theory of the syntactic structure o f tense, Stowell (1996) suggests that the English perfect tense form made up o f the verb 'have' together with undergone change, see Simpson (forthcoming). 1 7 See also Huang (1990) for claims that shi is an auxiliary verb in shi-de sentences. 243 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a second inflected verb is in fact a complex tense construction which critically comprises two finite tense forms, rather than a single tense+participial form. Stowell argues that the interpretation of (present) perfect forms such as (65) is that a past event described by the — e«-inflected lexical verb is related to a present speech time encoded in the auxiliary have, and that syntactically both lexical verb and auxiliary project discrete instantiations of tense in the complex structure, present (or sometimes past in past perfect forms) with the auxiliary and past with the lexical verb. The result o f the presence of the two tense forms is that the event described by the lexical verb is interpreted as having taken prior to the speech time (hence in the past) and to have relevance to the (present) speech time. Turning back to Chinese shi-de sentences, such an analysis of the perfect tense can be suggested to be exactly what is necessary to capture the apparent properties and interpretation of shi-de forms. Having seen that both shi and de individually show signs of being +fmite and consequently associated with discrete tense positions, it can now be pointed out that shi-de sentences significantly also have interpretations similar to English perfect forms in depicting past time events which are stressed as having clear relevance to/in the present/speech time. Such aspects of the shi-de construction can be simply captured if it is assumed, following Stowell's analysis of the English perfect, that both shi and de in fact project tense 244 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. positions and TP constituents and that the past tense encoded in de is embedded under a present tense specification with shi as in (6 6 ) below: 18 (6 6 ) TP, Spec T'i [zuotian prok lai], T , [yesterday prok de come], A Such a structure not only licenses the A-not-A patterning found, shi being associated with the structurally higher +finite T° and therefore naturally being the element repeated in A-not-A forms, the embedding o f past-time TP2 under present-time TP, can also be argued to naturally encode the interpretation of strong present relevance of a past event. (6 6 ) furthermore provides an explicit answer to the initial question of how shi and the de-headed TP combine together— shi as a higher verbal/auxiliary element in such structures selects for the TP headed by de. Considering the question of why de as a new instantiation of past tense is not found to occur with every verb having a past time interpretation, the analysis in (6 6 ) can be suggested to offer the relevant answer here and a TP headed by de will 1 8 Note that TP here actually is a simplification o f the tense structure proposed in Stowell's 245 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. essentially only appear if it is indeed selected by the element shi (or a phonetically null equivalent, as shi need not always be overt in shi-de structures). In this sense the occurrence of de in T° will be very similar to that of the English past tense form - en in perfect constructions such as (65), -en only surfacing as an instantiation of past when selected by the auxiliary verb 'have.' Concerning the element shi itself, given its regular association with the creation of a cleft-like focus interpretation, shi will in turn only be selected from the lexicon when a focus interpretation is indeed required, and in this sense will be similar to the common (optional) use of the English auxiliary verb 'do' to cause readings o f emphatic focus in examples such as (67): (67) a. I do like this cake b. Jane does know how write a good letter. The distribution, syntax and interpretation of de as a past tense morpheme can consequently be given a principled account in line with recent assumptions about the structure of complex tense forms and the properties of similar elements in other languages. Having also discussed how other aspects of the syntax of past tense shi- de forms can be explained under the assumption of D-to-T re-analysis and the conversion o f a CNP-type structure into a simple TP constituent, the chapter is now closed with a brief summary and consideration of D-to-T re-analysis as a general hypothetical phenomenon, and (in an appendix to the chapter) with a discussion of how focus may be formally encoded in the shi-de construction. works, but sufficient to represent the basic intuition here. 246 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.4 Summary: D-to-T and Paths of Grammaticalization This chapter set out to establish the categorial status and syntax of clause-final de in the Chinese shi-de construction. Section 1 began by arguing that the optional repositioning of de adjacent to the verb together with the necessary past time interpretation of the verb in such configurations strongly suggests that de instantiates the category of past tense when cliticized to the verb. As non-past interpretations of shi-de forms are however possible with de in post-object position, section 2 concluded that de must also be able to instantiate a category distinct from tense and that speakers consequently allow for a dual analysis of the element de. Taking this non-past de to be nominal-modifying de, section 2 then suggested that non-past shi- de sentences may possibly be assumed to contain a CNP type structure with a phonetically empty head, and that such a hypothetical structure would allow for a natural account of restrictions on w/7-adjunct and adverbial occurrence in non-past interpretations. The lack of similar restrictions with post-object past time interpretations of de then resulted in the conclusion that de in such instances has also undergone re-analysis as the past tense head of a TP and that the optionality in placement of past tense de is essentially due to cliticization of de targeting two potentially different hosts V and AspP, clitics elsewhere frequently showing such optionality in their attachment. Section 3 finally motivated the formal syntax assumed to underlie the re-analysis of de in T°, and argued that this is re-analysis of an original D° element, resulting in the creation o f a complex new tense form similar to the English perfect. In this last (main) section of the chapter, I would like to now 247 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. briefly reflect on certain general aspects of the hypothesized D-to-T re-analysis and on implications of the analysis for the theory of grammaticalization. A first issue which arises from the suggestion that there is D-to-T re-analysis in the shi-de construction is whether it is in fact generally plausible and likely that an element of type D° actually would undergo re-analysis as an element of type T°, or whether such a category change is fully unexpected and therefore rather suspicious despite the empirical evidence supporting the assumption of a D-to-T re-analysis. Here I believe it can in fact be positively concluded that a D-to-T category switch is in fact a rather natural change because the functional roles performed by D° and T° elements in their respective nominal and clausal domains can be argued to be fundamentally very similar, both being deictic elements providing reference to their respective complement constituents. In D-to-T re-analysis it can therefore be suggested that the definite reference-fixing property of a D° element may simply come to be reinterpreted in the locus of temporal reference and definiteness in the clause, the T°-position, and that such a categorial re-analysis is therefore not an unnatural change, a functional element effectively just re-applying its underlying deictic/referential function in/to a new (clausal) type of domain. Reconsidering the particular case of de in the shi-de construction, it was earlier noted that use of this element consistently results in an interpretation in which the past or future occurrence some background event is heavily presupposed and therefore constitutes knowledge which the speaker assumes is shared by the hearer. In this sense the use of de consequently signals and marks the background event encoded in the predicate 248 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. as being definite (i.e., identified, unique, shared knowledge) and so functions in a way which is similar to other D°/determiner elements. In the D-to-T re-analysis argued for in shi-de forms, it can therefore be suggested that this definite reference- marking function present with de is then simply reinterpreted in the clausal domain as applying in a purely temporal manner and de indeed not unnaturally comes to be understood as marking past tense, deictically fixing the reference of its clause as being identified at some point prior to the present/speech time. Assuming that D-to-T re-analysis is then in principle not an implausible type of categorial switch, this raises the natural question of whether it is a re-analysis which might in fact have occurred in other instances besides de. Here there are indeed two other cases which can be mentioned which I believe may also qualify as occurrences of D-to-T conversion, adding further potential support for such a possibility in general. The first of these relates to a language known as Panare (Venezuela) studied in Gildea (1993), where it is found that demonstrative pronouns occur as linking elements between subject DPs and predicate nominals. Gildea points out that the proximal/distal deictic orientation of such linking demonstratives may often be different from and apparently in conflict with that of other demonstrative elements present in predicate nominal constructions. Gildea argues that in such cases there actually is no conflict in spatial deixis and suggests that the linking 'demonstratives' have now in fact been re-analyzed as copulas, with their original proximal and distal deixis value now giving rise to interpretations of future time and past time reference rather than any spatial deixis. If Gildea is correct here, 249 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. and demonstratives are assumed to be elements of type D°, the Panare predicate nominal construction can then similarly be argued to be an instance o f D-to-T re analysis. Another possible case of D-to-T conversion may actually have occurred in English in the apparent re-analysis o f the demonstrative element 'that' as a clausal complementizer. In Pesetsky & Torrego (2000) a wide range of syntactic arguments and evidence have recently been presented for the assumption that the English complementizer 'that' is in fact synchronically first base-generated in T° and then subsequently raised up to C°. If Pesetsky & Torrego are indeed right and 'that' is an instantiation o f T° as well as C°, given that such an element is known to have developed from a demonstrative/D° this offers once again the clear possibility of a further case o f D-to-T re-analysis. The conclusion that D-to-T conversion is a re-analysis which has occurred in Chinese and possibly also elsewhere can now be noted to have consequences for the general theories of categorial re-analysis and the mechanisms which underlie and permit grammaticalization phenomena. In Simpson (1998a/ 1998b), Simpson & Z. Wu (1999, 2000), Roberts & Roussou (1999) and chapters 2 and 6 of the present thesis, it is argued at some length that grammaticalization is a process of re-analysis resulting from movement within a syntactic tree. In all these works it is suggested that if a lexical element a regularly undergoes movement from an original position X to a higher position Y in the functional structure projected over X, then over time the element a may eventually come to be re-analyzed as a simple instantiation of Y rather than X. When such grammaticalization and reinterpretation of the lexical 250 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. element a occurs, a is subsequently base-generated directly in the higher functional head (or alternatively specifier) position Y as indicated in (6 8 ), and the lower position X is instantiated by a new lexical element (P in (6 8 b)): (6 8 ) a. b. This general process of movement-dependent re-analysis and grammaticalization also occurs when a is in fact already a functional (rather than a lexical) element, the result in such cases being that a becomes re-grammaticalized as a higher head/XM A X occurring in the functional structure (see chapters 1 and 6 of the present thesis). With the creation of new tense/auxiliary morphemes, the natural expectation is that such elements will also follow this common path of grammaticalization and arise from the re-analysis of lower light verbal elements raised into the functional structure dominating VP. While such re-analysis does seem to regularly occur as expected (as, for example, in the common development o f future tense verbs from original lexical verbs o f desire such as English will; see also chapter 6 for a further 251 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. case in Chinese), the patterns found with de and the creation of a new past tense morpheme from a functional element originating in the nominal domain now significantly seem to offer good evidence for the general availability of a second quite different path of grammaticalization. This additional route o f categorial re analysis does not result from any movement and re-analysis within a single lexical- functional domain, but instead critically involves the re-analysis o f a functional category from one lexical-functional domain as a functional head in a discrete second type of domain, a kind of 'lateral' cross-domain re- analysis/grammaticalization. Although such re-analysis may (perhaps) be more restricted in its occurrence than the alternative movement-assisted grammaticalization, it can nevertheless now be argued to constitute a genuine possibility when assisted and permitted by properties of the embedding syntactic structure, as indeed seen with de, where a null-headed CNP/clausal nominalization (by hypothesis) allows for reinterpretation as a simple clause headed by de in T°. The shi-de paradigm and the re-analysis of de consequently allow for potentially important insights into general mechanisms of grammaticalization. Finally, D-to-T conversion also allows us to speculate on possible constraints on the general process o f cross-domain re-analysis. In recent years, researchers such as Abney (1987), Szabolcsi (1994) and many others have suggested that there is a significant similarity and parallelism in the internal structure of DPs and clauses with various functional elements in the nominal domain corresponding to similar elements in the clausal domain. If there is indeed such a correspondence relation between 252 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. functional heads in DPs and clauses, the occurrence of D-to-T conversion may suggest that a natural restriction on lateral grammaticalization linking elements in different domains may perhaps be that this is an operation only possible between nominal and clausal functional heads which indeed stand in a formal equivalence relation. Precisely how the various functional heads in nominal and clausal domains do in fact parallel and relate to each other is clearly a matter for much further investigation, however the obvious similarity of D° and T° elements in their reference-providing functions suggests that these two elements at least should be considered to be functional equivalents and that it is this parallelism in role and correspondence relation which has licensed conversion from one domain to the other. 19 4.5 Appendix: on the Focus Interpretation of Shi-de Forms The central interest of this chapter has been on the hypothetical grammaticalization of de as a new tense marker in shi-de forms and how this has had interesting observable effects on the syntactic structure of such constructions. The chapter has therefore concentrated primarily on aspects of the past/non-past tense interpretation in shi-de forms rather than on the focus interpretation which is more frequently the centre of discussion in analyzes of the shi-de construction. For completeness and before closing the chapter I would however briefly like to consider how the 1 9 For further evidence o f the close relation o f D° and T° elements see Lecarme (1996) where it is suggested that demonstratives in Somali give rise to temporal interpretations and may even be coming to encode tense within DPs. 253 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. important focus interpretation may perhaps be syntactically encoded in shi-de forms, and note that there are two fairly obvious possibilities here— either LF movement or (LF) binding. In both Chiu (1993) and D. Shi (1994) it is suggested that the element shi instantiates the X° head of a focus projection which Chiu refers to as ShiP. If one assumes that the selection of shi either results in the projection of a FocusP over VP or that the VP headed by shi is itself interpreted as a focus-licensing projection, the question then arises as to how an element in focus syntactically relates to this projection and how the focused interpretation of an element such as zuotian 'yesterday' in (69) below is licensed in the structure: (69) wo shi zuotian mai piao de I BE yesterday buy ticket DE 'It was yesterday that I bought the ticket.' Frequently it has been observed that it would seem to be the element which is right- adjacent to shi which is the focus in shi-de forms, and hence that gen Zhangsan in (70) cannot encode the sole focus, rather this must be zuotian: (70) wo shi zuotian gen Zhangsan mai piao de I BE yesterday with Zhangsan buy ticket DE 'It was yesterday that I bought the ticket with Zhangsan.' Possibly one might imagine that this necessary interpretation of the right-adjacent element as the focus results from some kind of overt movement of the element to a position related to the focus head shi. However, there is fairly simple evidence that 254 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. this is not so and that the focused element is not related to shi via overt movement. In (71) and (72) for example, it is not possible to overtly move the object of the verb either to a position following shi nor to a position preceding it in order to contrastively focus it (note that (72) is grammatical, but the object is not contrastively focused, rather it is zuotian which is contrastively focused): (71) *wo shi nei-zhang-piao zuotian mai de I BE that-CL-ticket yesterday buy DE (72) wo nei-zhang-piao shi zuotian mai de I that-CL-ticket BE yesterday buy DE 'It was yesterday that I bought the ticket.' If overt movement therefore does not relate a focused XP to shi and FocusP, there are two other plausible ways in which the focus relation might be established and licensed. One possibility is that an element in focus might raise to SpecFocusP at LF in order for this focus relation to be licensed, as essentially suggested in Chiu (1993) and D. Shi (1994). A second possibility is that shi or the head of FocusP might alternatively bind a focused element in its in situ position at LF and license its focus in this way. Both potential approaches would be fully compatible with the structures argued for in this chapter, with shi selecting a rightward DP/TP in (57) and (6 6 ). What both types of account do need to explain in some kind of way is the observation that there appears to be a Spell-Out adjacency requirement holding between shi and the focused XP in shi-de sentences (as noted above in (70)). This might seem to be somewhat surprising given that the licensing of the focus relation 255 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. between shi and the XP is arguably only established at LF either via LF movement or binding.20 However, there are in fact ways of understanding the apparent Spell-Out adjacency requirement in both LF movement and LF binding approaches which avoid the conclusion that it is linear adjacency of shi and the focused XP which is really required at Spell-Out. In order to show this as clearly as possible, I will first provide a clarification of the fuller restrictions on what may be interpreted as a legitimate focus in shi-de forms. With regard to the licensing of focus in shi-de sentences, the following observations need to be explained in any account. First of all, if two adjuncts occur following shi as in (70), it is only the structurally higher adjunct adjacent to shi which can be licensed as the sole contrastive focus in the sentence, as already noted above. Secondly, it appears that only an XP/XM A X constituent can be focussed by shi and not a head X°-level element such as a verb/V°. Consequently in (73) it is not possible to contrastively focus just a verb (either lai 'come' (73iii) or kan 'see' (73iv)). In (73) either the embedded VP2 [ kan Zhangsan] 'see Zhangsan' (73i) or the DP object 'Zhangsan' (73 ii) can constitute the focus in the sentence: 2 0 Given that the focus relation is one of interpretation, it might be difficult to suggest that it is established via binding at any point prior to LF such as Spell-Out/PF. 256 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (73) wo shi [ [vpi lai [vp 2 kan [dp Zhangsan]]] de] I SHI come see Zhangsan DE (i) 'Its to see Zhangsan that I've come (here).' (ii) 'Its Zhangsan that I've come to see.' (iii) NOT: 'Its to SEE Zhangsan that I’ ve come (not to BEAT him).' (iv) NOT: 'I CAME to see Zhangsan (I didn't GO to see him).' The above patterning can essentially be captured if the constraints in (74) and (75) below are assumed as relevant for the determination of focus-licensing in shi-de forms: (74) The licensing of focus in shi-de sentences The focus in a shi-de sentence is the structurally highest overt xM A X YP c-commanded by s/z//Focus°, or the structurally-highest and overt XM A X ZP contained within YP. (75) Maximality restriction on the focus in the shi-de construction It is not possible for the focus to be the entire overt sequence following shi. If we consider (70), the adjunct zuotian adjoined to IP will be structurally higher than gen Zhangsan (which is perhaps adjoined to VP) and so will be selected as the natural focus, c-commanded by shi!the Focus head. In the absence of any time, place or manner adjuncts as in (73), there will be two potential choices for the focus. The sequence [lai kan Zhangsan] is the structurally highest overt constituent c- commanded by shi but this sequence cannot be the focus due to the pragmatic maximality restriction in (75), which essentially seems to require that at least some of the material following shi be old, presuppposed information not in the focus. The 257 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. structurally highest overt XM A X c-commanded by shi which is not the maximal overt string following shi is the embedded VP2 [kan Zhangsan]. This VP is therefore a legitimate candidate for the focus o f the sentence (a YP in (74) which does not violate (75)). (74) also allows for the structurally-highest and overt x M A X ZP contained within such a YP to be a potential focus, and so the object DP Zhangsan is also a legitimate focus in (73). It can therefore be seen that the apparent linear adjacency restriction found in cases such as (70) is in fact just a part of a wider configurational restriction on the availability of a constituent as a possible focus in shi-de forms. Furthermore, there are elements such as DP objects not linearly adjacent to shi which the configurational restrictions in (74) and (75) allow to legitimately function as foci in shi-de. Finally it can be suggested that the structural/hierarchical constraint in (74) together with the more pragmatic maximality condition in (75) may potentially be understood as being restrictions on either movement or perhaps binding relations relating the focused XP to s/zZ/FocusP. If an XP is licensed as the focus of a shi-de sentence as the result of LF movement, it can be suggested that (74) and (75) essentially determine what is structurally visible to shi/Focus0 and so constrain what is potentially available as a target for attraction by the focus head to SpecFocusP. Alternatively, however, if focus is licensed in shi-de forms by a non-movement LF binding relation, it can be argued that (74) and (75) similarly restrict which elements are visible to s/zZ/Focus0 for binding and licensing as an associated focus. Licensing via binding in such a way could then be naturally understood as being similar to the 258 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. non-movement licensing of wh elements in situ by a c-commanding +Q C° as originally proposed in Baker (1970) and more recently in Tsai (1994). 259 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5. Resultative Constructions: Directionality and Re analysis 5.0 Introduction Chapter four discussed in detail the ongoing syntactic change taking place in shi-de structures when a sentence-final de re-positions itself between the verb and its object. In the course of the chapter it was mentioned that such re-positioning of an element from a sentence-final position to one right-adjacent to the verb in fact has a clear historical precedent in Mandarin Chinese and that the element -le/liao is known to have undergone a similar change in position. This surface parallelism in the re positioning of de and -le discussed in chapter four is schematized below in ( 1 ): (1) a. V Object de — > V-de Object b. V Object liao — » V-liao/le Object The present chapter focuses on a third potential instantiation of the same patterning diachronically found in the creation of resultative constructions such as Modem Chinese (2) and (3) below: (2 ) ta xi-ganjing yifu le he wash clean clothes LE 'He washed the clothes clean.' 260 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (3) wo kan-wan shu le I look-finish book LE 'I have finished reading the book.' Forms such as (2) and (3) in which two verbal elements occurring adjacent to each other produce a resultative meaning historically derive from sequences in which the second verbal element actually followed the object of the first verb. 1 The development of the modem resultative construction therefore seems to parallel the patterns noted in (la) and (lb) with a sentence-final element re-positioning itself over the object of the main descriptive verb to be right-adjacent to this verb, as indicated in (4): (4) Vi Object V? — > Vi — V2 Object Chapter five attempts to see how a better understanding of the change in pattern (la) as established in chapter four may now offer useful insights into the earlier change which took place in pattern (4) with resultative constructions, continuing this general 1 Note that since Wei, Jin and Northern and Southern Dynasties there had been other V-V structures consisting in the sequence o f V,-V2 -Object, such as (i) below, where both V, and V2 were originally used as transitive verbs and in these V r V2 sequences occur sharing the same object following both verbs. Such sequences also commonly allowed for the insertion o f a co-ordinator (for example, er 'and') between the two verbs, as in example (ii). Historically such structures never underwent any re-positioning o f the object, unlike the RVC's examined in this chapter, and can be suggested to have originated from rather different co-ordinate structures, as indeed argued in Li and Shi (1998). As a result they will not be considered here in the chapter. Examples below are from Li and Shi (1998). (i) tai-tong daolu, wu you zhangai. open-through road Neg have obstacle 'Open the road and make them free o f obstacles.' (ii) bao zi hou jj er sha zhi. leopard from behind attack and kill it 'The leopard attacked and killed it from behind.' 261 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. theme in chapter six with a detailed re-examination of the similar diachronic re positioning o f le/liao in (lb). In both chapter five and chapter six I set out to answer the following important questions: (5) a. What was the reason for the re-positioning of the sentence- final element to position right-adjacent to the main lexical verb? b. What is the current syntactic status of such constructions? Broadly this chapter will argue that the second verb V2 in resultative constructions essentially has the status of an aspectual suffix combined with the main verb V 1 in the lexicon and licensed/checked by discrete aspectual projections which dominate the VP, contra recent syntactic accounts of resultative V-V sequences in Zou (1993) and Sybesma (1999). In sections 5.2 and 5.3, which concentrate on resultative forms, it is suggested that the movement/re-positioning of Vi to its present position right-adjacent to Vi occurred as part o f a re-analysis process in which V2 became re analyzed as instantiating an aspectual head. Specifically it is proposed that the following steps took place in the re-analysis process. First of all, assuming that V2 may have originally been licensed in some kind of predication relation with the object of Vi, it is suggested that a significant increase in focus on the telic aspectual contribution o f V2 in the construction subsequently led to the re-analysis of V2 as primarily an aspect marker (with certain additional inherent lexical content). Secondly, as a direct result of this re-orientation of the primary contribution of V2 262 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from being one of direct predication of a lexical property onto the object of Vi to being one o f aspectual modification of the predicate, syntactically V2 came to be re analyzed as predicating of the Vj and its object together. Formally this resulted in Vj and its object being re-analyzed as a VP occurring in the Spec of an aspectual projection headed by V2. Thirdly, V2 as Asp0 underwent re-positioning and apparent movement to Vi in a way highly similar to the movement o f de to the verb after de re-analyzed as T°. However, whereas movement of de to the verb occurring (in the VP) in SpecTP is resulting in a new tense suffix on the verb, in the case of resultative V2 re-positioning this created new aspectual suffixes. The chapter then argues that this V2-re-positioning was actually not triggered by any phonological reasons or clitic properties of the V2 but significantly occurred in order to re-align the relevant AspP in accord with the canonical direction of selection in Mandarin and harmonize the location of the AspP with the head-initial parameter setting in Chinese. Quite generally the chapter shows how current changes in shi-de forms and the conclusions reached in chapter four allow for critical insights into the restructuring process when sentence-final elements undergo re-positioning over objects and attach to verbs, using this to arrive at a new and potentially revealing account of Chinese resultative constructions. The chapter also shows how the development of a general process of suffixation in head-initial languages has a natural explanation in terms of headedness and directionality. This I hope to add formal understanding to typological claims that there is language-internal pressure for a single direction of selection and that any examples of non-canonical 263 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. directionality in selection relations will undergo 'correction' over time on account of such pressure. Finally, the chapter adds certain speculations on how alternations in resultative constructions in English and various other languages might be re considered in the light of the suggested analysis o f Chinese. The structure of this chapter is as follows. In section 5.1.1 I introduce the phenomena of resultative verb constructions (henceforth RVC's) in full, and then follow this in 5.1.2-5.1.4 with a critical overview of certain recent lexical and syntactic approaches to RVC's, notably those in Yafei Li (1990), Zou (1994) and Sybesma (1999). In 5.1.5 I raises a number of objections to the syntactic accounts in Zou and Sybesma. In 5.2 and 5.3 I then gradually develop an alternative and fully general analysis of the synchronic and diachronic syntax of RVC's which argues for an aspectual suffix status o f V2 elements and invokes the notion of directionality in functional selection as an explanation for the re-positioning of the V2 element. In 5.4 I give a brief look at the consequences of the analysis developed in other structures. The chapter closes in 5.5 with a speculation on how the analysis suggested for Chinese resultatives can offer insights to English verb-particle constructions. 5.1 Previous Analyses and Current Objections 5.1.1 Resultative Verb Constructions In Li and Thompson (1981), Yafei Li (1990) and many other works V-V sequences such as (2) and (3) above are actually referred to as resultative verb compounds, 264 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. which necessarily infers that they result from a lexical concatenation process. Here I employ instead the more neutral term resultative verb construction/RVC to refer to such strings, in order to avoid any premature classification of their origin as either lexical or syntactic. This term RVC does however refer to the same set of V-V sequences in all major accounts. In Li and Thompson (1981) RVC's are basically defined as consisting of a sequence of two adjacent verbal elements in which the second element 'signals some result of the action or process conveyed by the first element.' (p54). RVC's are then broadly divided into three basic sub-types which can be referred to as simple R VC’ s, directional RVC's and phase RVC's. Simple RVC's are those RVC's which do not fall into the latter two RVC categories and are illustrated in examples (6-9) below: Simple RVC's (6 ) ta zuotian cai xi-ganjing yifu she yesterday then wash-clean clothes 'She just washed the clothes clean yesterday.' (7) Zhangsan yijing pao-lei-le Zhangsan already run-tired-LE 'Zhangsan has already got tired from running.' (8 ) Baoyu xia-shu-le qi2 Baoyu play-lose-LE chess 'Baoyu play chess and lost.' (Yafei Li 1990) 2 Note that until verbal-/? is properly re-examined in chapter 6 I will simply gloss it as LE in the examples. 265 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (9) Lisi da-si-le Baoyu Lisi hit-dead-LE Baoyu ’ Lisi beat Baoyu until Baoyu died.' Directional RVC's are those RVC sequences which involve a V2 which signals some clear directionality to the action denoted by Vj. As noted in Li and Thompson, there may even be more than one directional verb following the Vi action verb, as seen in ( 1 2 ) and (13): Directional RVC's ( 1 0 ) ta song-lai yi-ge-xiangzi he send-come one-CL-case 'He sent over a case.' ( 1 1 ) wo fang-xia wo-de-shubao le I put-descend my-satchel LE 'I put down my satchel.' ( 1 2 ) ta tiao-guo he le he jump-cross river LE 'He jumped across the river.' (13) tamen pao-chu-lai-le they run-exit-come-LE 'They came running out.' Finally, 'phase' RVC's are classed by Li and Thompson as sequences 'in which the second verb expresses something more like the type of action described by the first or the degree to which it is carried out than its result' (p.65). Typically phase RVC's represent simply the completion of an event/action rather than providing 266 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. additional lexical description of the result state that an object might be in after application to it of the first verb. Hence in the simple RVC's (6 ) and (9) the V2 explicitly classifies the object as being respectively 'clean' and 'dead' as a result of the action of the V|, whereas in phase RVC's such as (14-17) below the V2 indicates primarily just completion of the action and no finer description of the end-state o f the object is included: Phase RVC's (14) ta nian-wan shu le he study-end book LE 'He finished studying.' (15) Zhangsan xie-hao-le xin Zhangsan write-good-LE letter 'Zhangsan finished writing the letter.' (16) ta xiang-dao-le wo he think-arrive-LE I 'He thought o f me.' (17) ta mai-dao-le nei-ben-shu he buy-arrive-LE that-CL-book 'He managed to buy that book.' Precisely how all such sequences should be syntactically analyzed poses many challenging and interesting questions and has consequently been the subject o f a large number of theoretical works in recent years. In what follows I now present an overview of two contrastive views of RVC's which can be taken to broadly 267 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. represent two fundamentally different approaches commonly adopted in the literature— the 'lexical' approach and the 'syntactic' approach. 5.1.2 The Lexical Approach— Yafei Li (1990) In a careful, well articulated and interesting piece of work, Yafei Li (1990) provides a lexical compounding analysis of RVC's which largely succeeds in accounting for constraints on the productivity of V-V sequences. Prior to Li's work, no serious formal attempt had been made to model the creation of RVC's in a way which also reflected the many restrictions on their interpretation, and in this regard Yafei Li (1990) constitutes a significant step forward in the analysis of RVC’ s. Li begins by considering examples such as (18) which have been noted to have two possible interpretations as indicated: (18) Baoyu qi-lei-le nei-pi-ma Baoyu ride-tired-LE that-CL-horse (a) 'Baoyu rode the horse and as a result Baoyu got tired.' (b) 'Baoyu rode the horse and as a result the horse got tired.' Li points out that if the two verbs are compounded together in the lexicon as traditionally believed, then technically the combination of qi 'ride' and lei 'tired' might be expected to result in a compound verb with a total of three theta-roles— two theta roles from the transitive verb qi and one from the intransitive stative verb lei. As there are only two argument NPs present in (18) and hence the compound RVC projects only two theta roles, it is clear that any lexical compounding account of 268 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. RVC's needs to explain precisely how the theta-grids of two verbs are combined into a single argument structure. In particular a lexical account needs to explain: (a) how the number of theta roles in the theta grid of a new compound verb may often be less than a simple total of the theta roles associated with each verb, and (b) what constraints there might be on the combination of the theta grids of both verbs when compounded. These two together should account for why it is that only a specific sub-set of the imaginable interpretations in compounding is available. To capture a wide range o f restrictions on the interpretation of RVC's, Li proposes a mechanism of 'theta-identification' in which a theta role from one verb may be identified with a theta role in the second verb's argument structure and consequently assigned to a single argument of the newly compounded V-V form. For example, if Vi qi 'ride' in (18) has an external and an internal theta role labelled [vi 1] and [vi 2 ] and Vt lei 'be tired'just has a single theta role [v2 1 ], then either [vi 1 ] and [v2 1] can be identified and assigned to the same DP resulting in interpretation (a), or [vi 2] and [v2 1] can be identified and assigned to the same DP to give interpretation (b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19) a. VI V2 qi 'ride' lei 'be tired' [vi1] = [V2 1] [v.2] — » X rides and X gets tired: 'the rider became tired' b. VI V2 qi 'ride' lei 'be tired' [v.l] [vi2] = [vz 1] — > X rides Y and Y gets tired: 'the horse was ridden and got tired' When two transitive verbs are combined, and if one restricts one's attention to instances where there are two overt DP arguments licensed by Case in a sentence, it is noted that there are four ways in which the theta grids might potentially be integrated together via theta identification. However, only one of these four potential combinations actually is instantiated in natural language, the one depicted in (2 0 ) and illustrated with example (2 1 ): (20) VI V2 [V11 ] = [V21 ] [VI2] = [V22] (21) Jiaoda de zhuren da-ying-le zhe-yi-zhang Jiaoda DE master hit-win-LE this-one-battle 'Jiaoda’ s master fought and won that battle.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In order to account for the non-occurrence of other potential combinations (and also to explain complex restrictions on further combinations o f transitive and intransitive V 1-V2 pairs), Li proposes a number of constraints on the combinatorial/identification process. These include assuming that VI is the head of the compound in Chinese, that external theta roles have a greater prominence than internal theta roles and that there is also a critical 'prominence-sensitive' ordering in the way that theta identification is carried out. Without going into the full details o f the proposals here, it can be said that generally what emerges is in fact a highly successful lexical account of restrictions on the combination process, and one which offers a formally tight implementation of the view that Vi-V2 RVC formation is indeed an instance of compounding in the lexicon. It should also be noted that Li rejects the possibility that RVC's might be created in the syntax and opts for a Iexical-compounding approach for a rather simple but important reason. Li suggests that if both Vi and V2 were to project discrete VPs in the syntax and be combined by head-movement from V2 ° to V[° then one would have to assume that Vi selects for the VP headed by V2 as a complement.3 However, as there would not seem to be any obvious selection relation holding between Vi and a complement VP headed by V2 in V r V2 pairings (Li maintains), such a syntactic analysis is argued to be unrealistic and unavailable. 3 Li admits that a VP headed by V2 could occur as an optional adjunct adjoined to VP,, but adds that head-movement from V2 ° to V|° should then violate constraints such as Subjacency/the ECP. Consequently, the only syntactic representation one should potentially consider as a competitor to the lexical compounding approach should be one in which V2 indeed heads a VP selected by V, (hence not constituting a barrier to raising). 271 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.1.3 The Syntactic Approach— Zou (1994) A recent example of a syntactic approach which reacts to the lexical proposals in Y. Li (1990) is found in Zou (1994). Zou argues that Li is incorrect when he dismisses the possibility of a selection relation existing between V j and a separate VP headed by V2. Noting the observation made commonly in the literature that V2 consistently encodes a result which is directly caused by the activity of Vi, Zou suggests that there is indeed a semantic relation between V| and V2, and that Vi "implicates" V2 in Vi-V2 pairs. Zou then proposes that this semantic implication relation between the two verbs may be captured by suggesting that V | selects for a second VP depicting the outcome of Vi and that this latter VP is indeed headed by V2. In Zou's words: The event conveyed by the first verb "implicates" the event conveyed by the second VP. Following Chomsky (1985) and Hale and Keyser (1993), this implication relation is then "canonically strucurally realized" as a complementation relation between the first verb and the second VP, as the syntactic embedding corresponds to a semantic composite in which the event denoted by the second VP is a proper part of the event denoted by the first verb. (p.279) An RVC such as (22) with the (b) reading will then have a structure approximately as in (23), the V2 lei 'be-tired' raising overtly to V i zhui 'chase': (22) Baoyu hui zhui-lei Lisi Baoyu will chase-tired Lisi (a) 'Baoyu will chase Lisi and Baoyu will get tired as a result of it.' (b) 'Baoyu will chase Lisi until Lisi gets tired.' 272 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (23) IP Spec I1 VP hui Spec 'will' I Baoyu V zhui Spec 'chase' | 4 Lisi 'tired' In fact, because (22) is ambiguous, there will be other syntactic structures which will correspond to the other interpretations. For example, the (a) interpretation that Baoyu chased Lisi and Baoyu got tired is suggested to result from the structure in (24). (24) IP Spec I' VP hui Spec 'will' I Baoyu, V i lei 'tired' Spec NP zhui :hase' Lisi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note that the underlying linear order o f zhui 'chase' and lei 'be-tired' in (24) is in fact actually the reverse from the order in (23). Zou assumes that the surface linear order of the Vi and the V2 will in all cases be determined by the following independent linearization principle which applies after V2 has raised up to Vi: The verbal morpheme denoting an activity linearly precedes the verbal morpheme denoting the result of such an activity, no matter what their prior order is. (p.280) In both (23) and (24) the relations between the predicates and their arguments is syntactically clear and all theta roles are assigned in standard ways. Where the subject Baoyu is interpreted as receiving a theta role from both Vi and V2 this results from the additional presence in the structure of a pro controlled by the subject of VPi. As VP2 is taken to be selected by V|, the raising of V2 to Vi is also syntactically legitimate. Zou therefore suggests that a syntactic account of RVC's is consequently not only possible once one recognizes the existence o f a selection relation between Vi and V(P)2, but also desirable for the reason that it is well able to represent the thematic relations in Vi-V2 forms in a highly transparent way. Finally Zou adds that a syntactic account is superior to a lexical approach because it can be easily adapted to model certain other 'causative' RVC's which may be problematic for Li's lexical account. In examples such as (25), the surface subject zhei-ping-jiu 'this bottle of wine' has no direct thematic relation with either of the verbal predicates present zui 'get/be drunk' and dao 'fall-over'; consequently it is not obvious how it might be licensed to appear in Li's approach: 274 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (25) Zhei-ping-jiu zui-dao-le Lisi this-bottle-wine drunk-fall-LE Lisi 'Due to this bottle of wine, Lisi got drunk and fell over.' Zou suggests that in (25) there is an additional third VP dominating the regular RVC V1-V2 shell and that zhei-ping-jiu 'this bottle of wane' is base-generated as a causative argument in the Specifier of this VP. The syntactic account proposed by Zou therefore allows for relatively easy extension to cover such cases. 5.1.4 Sybesma (1999), Hoekstra (1992) A second more recent syntactic treatment of RVC's which explores a VP-shell type approach in much greater depth is Sybesma (1999). This interesting work draws its original inspiration in large part from ideas on resultative constructions set out in an imaginative paper by Hoekstra (1992). In the latter paper, Hoekstra isolates a number of canonical properties of resultatives cross-linguistically and uses them to build a novel account of resultatives. Specifically it is noted that resultatives consistently seem to exhibit the following classic properties: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (26) a. The main predicate is a dynamic activity verb: John washed the clothes clean. b. A secondary element (an adjective, particle or PP) predicates a property of the object of the main verb, but never predicates on the subject: Johnk washed [,\p the clothes], clean,,* k . Mary/, looked [ s p the words], up„-k. Suet sang f.w p Mary], to sleep, * t. c. The result state predicated of the object of the main verb serves to make the main activity verb aspectually bounded and telic d. Frequently an intransitive activity verb may be followed by an 'object NP' and a predicate: John laughed himself weak. Bill ran his sneakers threadbare. Putting this information together, Hoekstra suggests that both transitive and intransitive dynamic activity verbs may select for a small clause representing a (result) state predicated of an NP which is the subject of that small clause. Aspectually the small clause functions to provide an endpoint to the activity depicted by the main verb and convert it into a telic predicate, and critically it is this aspectual role which licenses the selection relation between the main verb and the small clause as its complement. Subsequently considering Chinese RVC's, Sybesma (1999) then focuses in on the well-documented property of such constructions that inclusion of the V2 element results in the V[ becoming telic, and procedes to build the Hoekstra approach to 276 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. resultatives into a well-developed account of Chinese RVC’ s. For V 1-V2 sequences such as (27) below, Sybesma suggests that the V2 occurs base-generated in a small clause selected by Vi and that V2 predicates directly onto the NP subject of the small clause from this position.4 At some point prior to Spell-Out the V2 will then raise up to right-adjoin to Vi, as depicted in (28) (where only the VP structure is shown, and the subject of Vi is not included). (27) xi-ganjing yifu wash-clean clothes (28) VP (=SC) wash 'clothes' ganjing 'clean' In support of the small clause analysis Sybesma observes that historically there were indeed linear sequences of [V| NP V2] which parallel the posited underlying structure in (28) and that such [V 1 NP V2] sequences at a later stage became [V 1 V2 NP] sequences. The small clause analysis then suggests that synchronic [V 1 V2 NP] forms still actively derive from a base order which was that of earlier resultatives. 4 Note that the small clause here is a VP (assuming ganjing 'be-clean' to be a stative verb) but that it could in principle be some other categorial type, e.g., a PP, as argued in Hoekstra (1992). 277 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Sybesma also points to the existence o f a similar resultative construction built with the resultative morpheme de (meaning 'to the extent of) where the Wj is again lower than the NP following V), as in (29): (29) ta xi-de yifu hen ganjing. he wash-DE clothes very clean 'He washed the clothes clean.' Sybesma suggests that the NP yifu 'clothes' can be suggested to be in a clear predication relation with V j ganjing 'be-clean' here, and that V 1-V 2 resultatives and [Vi-de NP V2] resultatives are similar in many ways. The occurrence of forms such as (30) with the linear sequence of [Vj NP V2] might then again be seen as suggestive of a V 2-raising analysis of V 1-V2 sequences, with (29) being something like the base underlying order which results in V[-V2 surface forms.5 As in Hoekstra's analysis of resultatives above, Sybesma suggests that the small clause structure in (28) encodes a result state predicated of an NP and that this result state critically functions to make the V| predicate aspectually telic. Such an aspectual relation is also taken to license the selection of the small clause by V\. Note also that an important bi-product of the Hoekstra/Sybesma small clause approach is that there is no direct selection relation between Vi and the NP which might standardly be considered to be its object. For example, in the English resultative: John washed [sc the clothes clean], it is the small clause AdjP which is 5 Later in his analysis Sybesma points out that there are however also significant differences between simple V,-V2 forms and the [Vr de NP V2 ] type of resultative which suggests that they are not derived from a single underlying structure. 278 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the complement selected by the V 1 washed and there is no thematic selection relation between V i washed and the NP clothes. This NP is licensed solely by its predication relation with the small clause predicate clean. The same is true in Chinese (28) and there is no thematic relation between V 1 xi 'wash' and the NP yifu 'clothes' which is base-generated inside the VP-small clause as subject to the predicate ganjing 'clean.' Commenting on the relation between V i and the NP, and following an idea in Kayne (1985), Sybesma suggests that there is only a pragmatic linking between V| and the NP, and that the interpretation of a structure such as (28) is essentially as in (30): (30) interpretation o f (28): wash [clothes (are/become) clean] + There is a washing event. + The state of the clothes being clean aspectually closes the washing event. — > Pragmatic implicature: if the clothes are clean and they are part of a washing event, then they can be assumed to have been washed. — > Pragmatically the NP is inferred to be the object of Vi. Considering the interaction of verbal -le and Vi-V2 sequences, Sybesma makes the interesting and novel suggestion that verbal -le is actually also a resultative predicate and therefore occurs as the head of a small clause. The subject of its small clause in the case o f a Vi-V 2-le sequence will in fact also be a small clause and the proposed underlying (VP) structure and derivation o f (31) is shown in (32): 279 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (31) (ta) xi-ganjing-le yifu (he) wash-clean-LE clothes 'He washed clean the clothes.' (32) VP V' V, SC Xi d -le VP NP V' yifu V2 r ,anjing The V2 ganjing 'be-clean' will raise to -le forming ganjing-le and then this unit will move higher to Vi and result in the surface sequence: xi-ganjing-le yifu. Sybesma's account highlights interesting parallels between the posited underlying structure for RVC's and earlier diachronic sequences and current [Vi-de NP V2] resultatives. Together with Zou's analysis, it furthermore demonstrates that a genuine syntactic alternative to Y. Li's lexical account can indeed be carefully formulated. Finally, Li's principal objection to a syntactic approach that V i cannot be taken to select for a second VP headed by V2 is intelligently responded to: such a selection relation arguably may be motivated on aspectual grounds, VP2/SC serving to provide a telic bound to the action depicted by V[. 280 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.1.5 The Objections to the Syntactic Approach In spite of the attractions mentioned above, I would however like to argue that there are a number of reasons why such a syntactic account is not in fact fully satisfactory and that one might therefore go on to pursue alternative solutions. First of all neither Zou nor Sybesma actually provide any real explanation of the movement of V2 to Vi. so it is not clear why it takes place and why one doesn't instead find sequences with the earlier historical order: [ Vj NP V2 ]. Zou states that: 'the verb-raising from V2 to Vj is motivated by the V-V compound formation.' (p.282) However, compounding is obviously a lexical operation and so critically may not be invoked in a syntactic analysis. Whereas V[-V2 adjacency is a natural consequence of a lexical compounding approach, it really remains significantly unaccounted for in a syntactic treatment of the type outlined by Zou. Sybesma originally treats V 1-V2 forms and Vj-de V2 resultatives such as (31) and (29) as parallel and suggests that a functional head 'Ext(ent)P' present in the structure always must be filled with overt phonetic material. This is argued to be possible in either of two ways— if the morpheme de is present, it occurs in Ext0 and a lower V2 will remain in situ as in (33). If de is not present, the V2 will be forced to raise to provide Ext0 with overt material as in (34): (33) xi [E x tp-de [ VP2 y i f u ganjing]] wash DE clothes clean Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (34) xi [ex ip ganjingi [ VP2 y i f u tj ] ] wash clean clothes The raising in (34) can therefore be taken to be responsible for the surface V[-V2 adjacency. However, later on in the development of the account of RVC's, Sybesma concedes that V 1-V2 forms and Vi-de V2 resultatives have different structures and that an ExtP is not projected in the former RVC type. Consequently the V2 raising is left without any motivation in this account (also as noted in Zou's analysis), and as this is clearly a critical component of the syntactic approach, it is highly unsatisfactory if it cannot in fact be motivated in any way and is only assumed in order to capture the linear sequence o f elements. Given that V2 elements such as ganjing 'clean' are clearly not clitics, the hypothetical raising can obviously not be triggered by phonological reasons either. Secondly, a conceivable problem for both Zou and Sybesma concerns the occurrence of the aspectual morpheme -le following V1-V2 sequences as in (35). (35) Baoyu xi-ganjing-le yifu Baoyu wash-clean-LE clothes 'Baoyu washed clean the clothes.' Zou assumes that -le is an aspectual morpheme located in Infl. This being so, Zou is forced to suggest that V2 first raises up to V t, and then the Vi-V2 complex raises higher to Infl where -le is then attached. However, if the Vi-V2 -le sequence eventually lands in Infl, a VP-type adverbial should follow such a sequence. As 282 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. shown in (36), the potential occurrence o f VP-type adverbials linearly to the left o f verbs carrying -le would strongly argue against overt raising to Infl. (36) (ta) man-manr-de xi-ganjing-le yifu (he) slowly wash-clean-LE clothes 'He slowly washed clean the clothes.' Sybestma states that an explicit goal of his investigation is to provide an account o f the V[-V2-le word order i.e., why one finds this order and not any other such as *[Vi-le-V 2], etc.: The distributional fact I would like to account for is that verb-/e may only occur on matrix verb level in cluster resultatives, in which case it may only follow the cluster as a whole, (p.67) Sybesma thus suggests that -le is a type of V2 element which heads the small clause structure, as shown in (32). Such a treatment of verbal -le and RVC's does seem to avoid the problem noted in example (36) for Zou's syntactic approach. However, ultimately the V r V2-le ordering is not in fact derived in any particularly natural way as a result of the underlying structure posited in the analysis. As seen in (32) above, first the V2 has to raise and left-adjoin to -le and then this unit has to raise and right- adjoin to the V|. There is no clear reason why raising should result in attachment to different sides o f the target raised to (i.e., left-side in one case and right-side in another), and it seems that the choice is made purely in order to mirror the final surface form. Sybesma indeed admits that: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If we were to strictly follow Kayne (1994), an underlying structure like (32) [my numbering] would lead to Y-le-V [i.e., *ganjing-le-xi = V2-le-Vi], It seems to be the case that in order to derive the right surface order we need to stipulate that in the lexicon it is somehow determined and recorded that le is a suffix: it has to come last. (p.76) So it seems that a syntactic analysis which was specifically designed to account for the word order by positing an underlying structure which could naturally result in that word order is actually unable to output the surface form, and one has to resort to a stipulation in order to re-convert the output of the syntactic derivation into the attested linear sequence. Furthermore, the notion 'suffix' cannot in fact be employed here; if an element is classified as a suffix, this means that it attaches in the lexicon and hence not syntactically. In short there is no principled way to ensure that both V2 and -le attach to V) in the syntax and have this attachment mirror the surface order if one assumes an underlying order such as (32). The great potential for a syntactic account to explain the linear sequence of morphemes consequently fails here. It should be also noted that the same ordering problem occurs in Zou's account. As mentioned above, Zou has to evoke an independent linearization principle which will regularly 'correct' a number o f V 1-V2 orders outputted by the syntax which do not correspond to the surface order found (i.e., correct lei-zhui 'tired-chase' to zhui-lei 'chase-tired'). The stated goal of being able to capture the linear ordering with a syntactic approach turns out to be unexpectedly problematic. A third aspect of Sybesma's account which is arguably unattractive (though here depending more on one's theoretical orientation) is that it seems to be building 284 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. functional categories into positions which are actually under and dominated by the lexical predicate VP rather than occurring in the functional superstructure which is commonly taken to dominate a lexical projection. If one takes the clear aspectual nature and contribution of the V2 and also -le in particular to reflect a functional role, one might expect these elements to correspond to functional heads located above rather than below the VP. Nevertheless, in (32) -le is built in as a functional-type head below the lexical predicate V|. Related to this, if one takes the position that V2 and -le are not related to aspectual heads projected higher than the VP, then it is difficult to explain the co-occurrence restrictions which are found between V2 , -le and aspectual heads which do occur higher than VP in aspectuals projections, as e.g., in (37) adapted from Sybesma (1999) (I return to this point later on): (37) wo zhengzai ca-gan-(*le) boli I ASP wipe-dry LE glass 'I am just wiping the glass dry.1 A fourth objection to the proposed small clause analysis of Chinese RVC's concerns the conclusion that there is no direct theta relation between V 1 and the NP which might intuitively seem to be its object, i.e., that there is no theta relation between xi 'wash' and yifu 'clothes' in (31/32). Supposing the NP occurring after the V 1-V2 unit were to relate to V| only via some rather loose pragmatic inferencing, as noted above, rather than stand in a verb-object relation, then one might expect all kinds of interpretations to be possible which do not in fact occur. For example, it is not clear why sentences such as (38) would be ill-formed when compared with (39): 285 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (38) *Lisi ba Zhangsan jiao-xing-le Baoyu Lisi BA Zhangsan call-awake-LE Baoyu Intended: 'Lisi called Zhangsan and Baoyu awoke as a result of this.' (39) Lisi jiao-xing-le Baoyu Lisi call-awake-LE Baoyu 'Lisi called and Baoyu awoke as a result of this.' The small clause hypothesis predicts that (38) should have the following interpretation: 'Baoyu resulted in being awake because of a calling event.' In the absence of ba Zhangsan as in (39) it is suggested that we pragmatically infer that Baoyu was actually the object of the calling event as a natural salient interpretation. However, if a third NP is introduced in a 6<z-NP form or possibly as a sentence-initial topic as in (40) it is expected that this NP could be taken to be the pragmatically inferred object of the scolding action if this could result in a coherent interpretation (for example if Zhangsan was in the same house that Baoyu was sleeping in and Lisi's calling of Zhangsan could wake up Baoyu). The fact that such an interpretation is unavailable strongly suggests that the NP following V 1-V2 is necessarily the object of Vj in a strict syntactic way.6 6 Such a conclusion is in fact also suggested by Yafei Li (1998) for the similar [V,-de NP V2 ] resultative type. Li notes that when V r de is intransitive, it is possible to topicalize an NP in the sequence which follows V|-de as in (i) (i) Zhangsan ku-de [shenme huor]j Lisi dou bu xiang gan tt Zhangsan cried-DE any work Lisi all not want do 'Zhangsan cried to the extent that Lisi didn't want to do any work.' However, when V| is transitive, it is not possible to topicalize an NP in the same way. This is taken to indicate that the NP Lisi is indeed the direct object o f V, and so may not be preceded by a topic from a lower clause: 286 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (40) * Zhangsan ne, Lisi jiao-xing-le Baoyu Zhangsan Top, Lisi call-awake-LE Baoyu Intended: 'Lisi called Zhangsan and Baoyu awoke as a result of this.' A fifth important criticism of the Hoekstra/Sybesma small clause analysis of resultatives which will shortly be returned to again is that very often there is in fact no clear predication relation between the hypothetical small clause head/V2 and the 'object' NP. This is true for both English and Chinese. For example, consider in (41) and (42) below: (41) John looked [sc [n p the reference] up]. (42) Mary beat [sc [n p Peter] stupid ] The small clause analysis predicts the following interpretations for these sentences: (43) interpretation o f (41): The reference is up as a result of a looking event carried out by John. (44) interpretation o f (42): Peter is stupid because of a beating event carried out by Mary. (ii) *Zhangsan kua-de [shenme huor]; Lisi dou bu xiang gan t, Zhangsan praised-DE any work Lisi all not want do Intended: 'Zhangsan praised Lisi to the extent that she didn't want to do any work.' The same conclusion is reached on the basis of prosodic facts. If a pause is created directly after Vi de via the insertion of a particle ya, then this is acceptable only in the intransitive case with ku-de 'cried' and not with the transitive kua-de 'praised,' indicating that the NP Lisi is again the object of a transitive V| and cannot be separated from it. Although there are certain differences between Vr V2 - NP forms and V r de NP V2 resultatives, it is interesting that when the NP occurs between a transitive V( and V2 it clearly does behave like the object of Vh contra Sybesma's analysis o f such forms. 287 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, it is simply meaningless to say: "“The reference is up.,' and the interpretation of (42) is not that Peter is stupid after a beating event. Basically 'up' and 'stupid' only convey the meaning they do through combination with the V) 'look' and 'beat' and there is no isolated predication relation between the particle/adjective and the 'object' NP. The small clause analysis clearly leads one to expect that the 'object' NP and the following particle/adjective should be able to stand in a regular predication (together with a form of the copula to carry tense/agreement) without the need for a higher secondary predicate, yet this is not so. The output does not mirror the actual interpretation of a sentence when added to a higher predicate. For example, (42) does not mean [ John is stupid ] + [this state terminates a beating event]; 'stupid' only has its non-literal meaning 'badly-hurt' when combined with 'beat.' One cannot point to someone after they have been beaten and simply say: "He is stupid now." The same is true of many examples o f V 1-V2 resultatives in Chinese; although (45) is fine, they would not seem to result from combining the parts in (46): (45) wo liu-zhu-le ta I retain-live-LE he 'I stopped him.' (46) [ta zhu-le = he lived] + [this was a result of a retaining event] As in many other cases, zhu in (45) can not have its literal meaning 'to live' and can only have the non-literal meaning of being motionless precisely when combined with another verb. For example, one cannot paraphrase (45) with (47): 288 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (47) wo liu-le ta. *jieguo ta jiu zhu-le. I retain-LE him result he then live-LE 'I retained him. As a result he lived.' Similarly (48) does not mean: [We talked] + [the state that [Christmas arrived] terminated this talking event], that is, it does not mean we talked until Christmas. However, such an interpretation is what is expected from a small clause underlying structure such as (49): (48) women tan-dao-le shengdan-jie we talk-arrive-LE X'mas 'We talked about X'mas.' (49) women tan [ -le shengdan-jie dao] we talk LE X'mas arrive With other Vo elements if one attempts to isolate the object NP and the V2 into a single sentence predication, the result is unacceptable in any reading as in (50) and (51). (50) a. ta mai-zhao-le nei-ben-shu. he buy-catch-LE that-CL-book 'He managed to buy the book.' b. *nei-ben-shu zhao-le. that-CL-book catch-LE '???' 289 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (51) a. taxianzai zhu-jin-le xin-fangzi he now live-enter-LE new-house 'He has now moved into the new house.' b. *xin-fangzi jin-le. new-house enter-LE ’? ? ? ’ With intransitive Vi elements, it is suggested in Sybesma (1999) that the surface subject originates as the subject of the small clause and is predicated of by the \S, shown in (52) and (53). (52) pao [sc ta Iei-Ie] -> ta pao-lei-le. run he tired-LE he run-tired-LE ’ He got tired as a result of running.' (53) wan [sc ta ni-le] -> ta wan-ni-le. play he bored-LE he play-bored-LE 'He got bored of playing.' However in other examples, isolation o f the V2 and such a subject NP often results in a quite different meaning. As shown below, the meaning of the sentence in (54a) does not indicate that the subject originates from the small clause as schematized in (54b), since the result of being afraid is not that the NP women 'we' end up dying. The isolation of the subject and V2 and the subject also often results in a quite unacceptable forms, as shown in (55): 290 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a. women dou pa-si-le we all fear-die-LE ’ We were all scared to death.' b. [women si-le = we died] + [this was a result of a fearing event] a. ta zou-kai-le. *ta kai-le he walk-open-LE he now open-LE 'He walked away. (!He got open as a result o f walking.)' b. *ta kai-le. he open-LE '? ? ? ' The compositional account of resultatives in which the meaning of a small clause encoding a predication relation is added to the meaning of a higher verb is simply incorrect. Either the small clause often has no meaning on its own or its meaning may be quite different from the interpretation found in complex V 1-V2 sequences. It should be noted that the counter-examples presented here are mostly "phase resultatives" of some sort. They still remain as fair counter-examples to Sybesma's small-clause analysis. Following Hoekstra's account for "look the reference up" together with "paint the door red" under the same small-clause analysis, Sybesma also attempts to extend the small clause analysis to the resultatives with phase V2 elments. It is also my belief that these different types of [Vr V2] resultatives should be represented with the same underlying structure. One simple fact is that both literal V2's and phase V2's contribute to the activity with a telic bound and they never co-occur, as shown in (56) and (57): 291 (54) (55) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (56) *ta xi-ganjing-wan yifu le he wash-clean-finish clothes LE intended: 'He finished washing the clothes clean.' (57) *ta xi-wan-ganjing yifu le he wash-finish-clean clothes LE intended: 'He finished washing the clothes clean.' Historical facts also point to the same direction. First the literal V2's appeared in [VI object V2] resultaives during Qin/Han Dynasties, then the phase V2's came to appear in the V2 position o f the same structure during Wei/Jin/Northem and Southern Dynasties (G. Wu 1999), and finally both literal V2's and phase V2's collectively underwent a shift over the object in the Song Dynasty (Li and Y. Shi 1997). The fact that both phase and literal V? elements collectively underwent re positioning to Vj is strong evidence that all VVs came to be licensed and interpreted very much in the same way. A final argument against the small clause syntactic approach again relates to the claim that the V2 in Chinese predicates of the 'object' in a small clause. Basically it is possible to show that there may be a significant difference in (underlying) structure between English resultatives and Chinese V 1-V2 resultatives, contra the assumption made in Sybesma (1999). Although adverbial modification of the secondary predicate in English resultatives is not common, it can occur with time adverbials as shown in (58): 292 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (58) John brushed [the table clean o f dust for five minutes] (and then it got dusty again). In (58) the duration phrase can be understood to modify just the state [the table clean of dust] and not the action o f 'brushing.' In Chinese [Vj-de NP V2 ] resultatives (which is also analyzed to have a small-clause structure by Sybesma 1999), the same kind of modification is also possible: (59) wo qi-de ma lei-le liang-ge-zhongtou I ride-DE horse tired-LE 2-CL-hour 'I rose the horse so that it got tired for two hours.' However, parallel modification is significantly not possible in V|-V2 resultatives: (60) *wo qi-lei-le ma liang-ge-zhongtou I ride-tired-LE horse 2-CL-hour Intended: 'I rose the horse so that it got tired for two hours.' What this would seem to indicate is that adverbial modification of this kind is possible when an NP and a predicate occur in a single predicational-type sequence as in (59) ma lei-le or (58) 'the table clean of dust’ and not possible when the potential predicate V2 occurs bound to another predicate V| in a single unit as in (60). Supposing that V|-V2 resultatives were to have an underlying structure in which there was a small clause-type predicational sequence, one would expect similar adverbial modification to be available. As examples like (60) are however Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. unacceptable, the simple conclusion is that they are not derived from structures encoding a small clause predication.7 5.2 An Alternative Analysis— V2 as Aspect 5.2.1 Telicity and Non-predication Having seen how a syntactic account of RVC's based on VP-shell-type/small clause structures is problematic in a number of potentially serious ways, I will now attempt to develop an alternative account driven largely by the following three key observations concerning RVC's: (61) Lack of necessary predication between Vi and NP 'object' As noted above in 5.1.4. it is not true that there is always a predication relation between the V2 and the post-Vi-V2 'object' NP in RVC's. (62) Telic contribution of V ? As well noted in the literature (e.g., Li & Thompson 1981), the V2 in a V r V2 unit makes a dynamic event telic and aspectually bounded. (63) Historical re-positioning of V? Historically sequences of [Vi Object V2 ] changed to [Vr V2 Object] as the V2 re-positioned itself right-adjacent to Vi I will suggest that these three important properties of Chinese RVC's are intimately connected and combine to result in a single analysis of RVC's. The core claim of the 7 As there is evidence that the NP ma 'horse' in (59) is in object of V! position (see footnote 5), one may assume following Yafei Li (1998) that there is a co-indexed pro subject to the clause/predication headed by lei-le 'tired' and that lei-le predicates onto this NP (which is then 294 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. analysis will be that the V2 in RVC's first underwent re-analysis to become an Aspectual head and then subsequently re-analyzed further as a suffixal instantiation of Aspect which is attached to the preceding V 1. In order to arrive at this, synchronic analysis of RVC's will involve showing how (61-63) are connected to each other and how they have diachronically conspired to produce the V 1-V2 forms of modem Mandarin. Property (62) that the addition of V2 to a dynamic V| results in a telic predicate is so well-established that it here requires no real further qualification. Later on in chapter 6, I will return to consider in more detail how the aspectual domain may be divided into different sub-types, viewpoint and situational aspect (basically following Smith 1997). As for property (63), I will show that V2 is attached to V 1 due to the general pressure for uniformity in the direction of selection combined with frequent occurrence of linearly-adjacent V 1-V2 sequences over a long period of time. Property (61) concerning predication of V2 and the 'object' NP is very important and so can usefully be stressed again. What has been pointed out above in 5.1.4 is that there are many V 1-V2 sequences in which the V2 does not predicate of the following NP in any kind of normal way. It should be remembered that the small clause analysis of resultatives suggests that V2 does indeed predicate directly of the 'object' NP encoding a state which terminates the activity of the Vi. Syntactically and semantically if the V2-object NP relation is one of genuine predication used to interpreted as being co-referential with the object ma. 295 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. represent the end o f some activity, there is no good reason why it should not be possible to isolate this predication from the activity verb and have it stand in a syntactically independent form. However, as illustrated earlier, this is frequently not at all possible, leading to the conclusion that V2 and the object NP do not have to be licensed by standing in any predication relation. Essentially with a whole range of V2 elements, many of which are categorized as phase V2's, there is clearly no direct predication relation between the V? and the object NP and the interpretation of the RVC does not arise from first independently computing the meaning of a small clause and then combining this pragmatically with the action depicted by the main Vj, as is suggested in the Hoekstra/Sybesma approach. Instead it would seem that a large number of commonly occurring V2 's have particular meanings which are built into the meaning of a sentence in some other way. Instructional textbooks tend to list these V2 's as characterizing the whole of particular actions rather than particular object-types, shown in (64) for example: (64) verb literal meaning meaning as V? dao arrive indicates the continuation of an action up a certain point of time cuo be wrong to carry out an action incorrectly zhao *arrive (no longer in completion or attainment of an use as single verb) action 296 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It is also revealing that pedagogical textbooks frequently list the V i types that may typically occur with the V2, this indicating that the close selectional relation which the V2 has is to the type of Vi it is combined with rather then any NP type it might predicate of. The list extract in (65) is from T'ung and Pollard (1982): (65) V2 frequently occurs with: V) -jian (sensory perception) kan (look), ting (listen), peng (collide)... -kai (detachment) da (hit), la (pull)... -zhao (attainment) zhao (search), mai (buy), shui (sleep),... -hui (learned) xue (leam/study) The important conclusion I wish to draw from all of the above is that although it is possible that many V2 elements could be analyzed as predicating of the 'object' NP in RVC's if a small clause analysis were to be adopted, there are also numerous cases where such a predication relation does not and could not exist between the V2 and the object NP. This being the case, at least these V2 elements must be licensed in the Vt-V2 structure in some other way and consequently not in a small clause predication. What I will argue for in future sections is that the mode of licensing which is relevant for the above phase V2's is also actually made use of by all other V2 types as well, even those which might seem to have more literal predicative interpretations. 297 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2.2 Re-analysis and Aspect I will now propose and develop a syntactic account o f the change which suggests that there were two critical cycles of re-analysis in resultative structures responsible for the change, both importantly relating to the category of aspect. In doing so I will show how properties (61) and (62) above are critically relevant to the re-positioning of Vi, i.e., property (63). I will also how the account of shi-de forms in chapter four allows for useful and important insights into structural aspects of the re-analysis that will be argued for. In the present section I intend to concentrate on the first cycle of hypothetical re-analysis and focus on the importance of phase-Vi's in early resultative constructions. Considering the earliest RVC's with [V> Object V2 ] sequences, it would seem that V2 was first instantiated by verbs which did indeed have a close selectional relation with the object and seemed to stand in more of a literal than a 'metaphorical' predication-type relation with it. Such verbs I will henceforth refer to as 'literal V2's and contrast them with phase V2's which do not have any literal predication relation with the object. As G. Wu (1999) notes, phase V2's appear later than literal V2 's in the [Vi Object V2] sequences: As early as in Qin and Han Chinese already developed the "V(0)+complement" structure. The "V(0)+phase complement" structure emerged during Wei, Jin and Northern and Southern Dynasties, (p.22) Given that phase V2's are generally seen as being more functional in application than the literal V2's, this is indeed what one would expect, with more literal lexical 298 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. elements preceding the use of more functional elements in a structure. This emergence of phase V2's has however very important consequences, as will shortly be explained. First though we need to consider the structural status o f literal V2's in the [Vi Object V2] sequences. In section 5.1.4 it was argued that V2 elements in present-day Chinese RVC's are not licensed by any predication relation with the 'object' NP. However, although the various reasons given there showed this to be a valid conclusion for modem Chinese V 1-V2 forms, it might not be unnatural to suggest that fu lly literal early resultative sequences of [V* Object V2] did in fact encode some predicational relation between the Object and the V2. In this early stage the literal V2 was licensed in the resultative structure precisely because of this predicational relation. If this is so, a technical question arises as to how such a predicational relation might have been syntactically encoded. One possibility is to simply assume the small clause structure argued for by Hoekstra and Sybesma, repeated below as (66): (66) Earliest [V1-NP-V2] with literal V2's VP V' V, VP XI NP V' 'wash' 'c l o t h e s ' V2 ganjing 'clean' 299 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, in 5.1.4 it was noted that such a structure has the possibly unattractive aspect that there is no theta relation between the Vi xi 'wash' and the 'object' NP yifu 'clothes' which feels rather counter-intuitive. We have seen that in [Vj-de NP V2] type resultatives where the linear sequencing resembles the early Chinese [V 1 Object V2] RVC's, the post-V 1 NP behaves as if it is an object of the V 1 rather than just in a following predication with the V2. Yafei Li (1998) has suggested that in the modem [Vi-de NP V2] resultatives the overt object NP is linked to the V2 via a co-indexed pro which itself occurs in the predication with V2 (see footnote 5 and footnote 6). We can adopt this suggestion for the early [Vi NP V2] resultatives with literal V2's. Allowing for a co-indexed pro to mediate the relation between the overt NP object and the early literal V2 would then permit a rather different small clause structure as in (67). Here the overt NP being truly the object of the V\ and the small clause being analyzed as an optional adjunct: If it is objected that the object NPj will not c-command the pro and that this is necessary for co-reference, an alternative possibility might be to assume a VP-shell structure such as (68) with the NP 'object' base-generated in the lower SpecVP c- (67) Earliest [V1-NP-V2] with literal V2's pro,. . . ganjing(V 2) 'clean' xi 'wash' yifu, 'clothes' 300 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. commanding the co-rerefential pro in the small clause which is here analyzed as a 0 8 complement. The verb would then be taken to undergo overt raising to v . (68) Earliest [V1-NP-V2] with literal V2’ s vP I v' yifu jV i 'clothes' / prot. . . ganjing 'clean' Without further discussion I will simply assume here that one of these possibilities is correct and that there is indeed an early structure which can accommodate both the overt NP as a genuine object of the V i and a small clause with a pro being predicated of by V2 . A similar structure must indeed be available to account for simple sequences such as (69) in which both an overt object NP occurs and a resultative clause with a co-referential NP subject: 8 A third possibility with a ternary branching structure would be to have both the object NP and the small clause projected off the same complement node, this allowing for c-command between the object NP and a pro in the small clause. 301 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (69) I pushed Johni so that hei fell.9 What should be stressed here as important is the assumption that the most natural analysis of literal V2 in early [Vi NP V2] sequences is that it was indeed initially licensed via predicating o f the object, with this most likely being effected indirectly through the presence of a co-referential pro. Following the establishment of literal V2 in [Vi NP V2], one next finds the introduction of phase V2's in [Vi NP V2 ] sequences and this is highly significant. As argued in 5.1.4 phase V2 elements do not predicate of the object NP in RVC's. Consequently, a key important question is how phase V2's could have come to be 9 Some structure of this type must also be available for resultatives in Korean as reported in Kim and Maling (1997). Case-marking in Korean RVC's is particularly revealing. If the main activity verb is intransitive, the NP 'subject' of the result sequence surfaces in nominative case, as in (i) and (ii): (i) Robin-i paykkop-i ppacki-key wus-ess-ta Robin-NOM belly-NOM come.out-KEY laugh-PAST-DEC 'Robin laughed his belly out.' (ii) Chris-ka palpatak-i talh-key talli-ess-ta Chris-NOM feet-NOM wom-KEY run-PAST-DEC 'Chris ran his feet sore.' However, if the main event verb is transitive, the NP 'subject' of the result predication occurs in accusative case: (iii) Robin-i soy-lul ttukep-key talkwu-ess-ta Robin-NOM metal-ACC hot-KEY heat-PAST-DEC 'Robin heated the metal hot.' Kim and Maling suggest that -key is an inflectional element which assigns nominative case to the subject of the result clause. Because -key is also present in (iii) nominative case must also be available and assigned in (iii). As the overt NP soy-lul 'metal' however surfaces in accusative, it must be assumed that a pro subject occurs in the result clause in (iii) receiving the nominative case, and that the overt NP soy-lul originates in the main clause, being assigned accusative there. It is not possible to suggest that soy-lul somehow originates in the lower result clause and skips the subject position avoiding nominative case to explain the fact that it surfaces with accusative. Subject case-positions cannot simply be skipped over as (iv) below shows: (iv) *John mentioned him, [ (that) t; arrived ] Consequently UG must indeed in some way make available structures in which a transitive verb has a genuine NP object and also allows for the projection of a result clause with a co-referential pro subject. 302 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. licensed in such structures? If the early occurrence of literal Vi’ s in [V i NP V 2] was naturally legitimized via the predication relation they bore with the object NP but phase V2 S are not interpreted in the same way (i.e., as predicating of the object), then some other means o f legitimization must have become available to allow such elements into the early [Vi NP V2] structure. Here I would like to suggest that the emergence of phase-type verbs in the V2 position is the result of a first important stage in re-analysis which re-oriented the resultative construction in a particular way, subsequently allowing in certain V2 elements not instantiating any predication relation with the object NP. What is critical here is property (62), which states that all V2 elements (whether phase or literal) serve to provide a telic aspectual bound to the action depicted by Vj and the object. Prior to the occurrence of phase verbs as V 2 S , all V2 elements can be said to have contributed two discrete properties in early [VI NP V2] resultatives: (70) a. (Literal) V2's lexically encoded the type of result which terminated the event (e.g., the clothes come to be in a clean state after being washed), b. All V2's functionally provided a telic bound to the event of the V,. Intuitively, the important change which occurred in RVC’ s can be suggested to be that after some time V 2 f s could be licensed in resultative constructions by providing simply property (b) and not necessarily property (a) as well. Phase V2's which do not predicate of the object NP and consequently do not describe the result state of the 303 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. object can therefore be said to have become licensed in virtue of providing property (b) alone. If a V2 comes to no longer predicate of an object inside the VP but instead provides instead a more general telic bound to the whole predicate signalling completion of an action, it can be argued that in terms of scope and configurationality the V 2 essentially becomes interpreted as applying its function to the entire VP rather than just a sub-part of it (i.e., just the object). If this is correct, a question which naturally arises now is how could such a potential change effect itself structurally in RVC's? Here I believe is precisely where the account of changes in the shi-de construction developed in chapter four is now potentially very helpful and enlightening and there is indeed a remarkable parallelism in the diachronic development of shi-de forms and RVC's. In order to make this parallelism clear, I will now make a short backtrack through the conclusions of chapter four and remarks on the surface similarity of shi-de constructions and RVC's. As noted at the beginning of this chapter, there is an obvious similarity in the way that shi-de forms and RVC's have shown changes in linear order. In both constructions a sequence of [V Object X] changes into a sequence [V Xj Object t, ] with the element X representing de in shi-de constructions and a V2 in RVC's; in both cases it also seems clear that the element X re-positions itself leftwards, right- adjacent to the main verb, rather than there being any rightward object movement. Considering shi-de forms it was argued that de undergoes re-analysis from category D to category T and that sentences such as (71) have a structure such as (72): 304 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (71) wo shi zuotian lai de I SHI yesterday come DE 'It was yesterday that I came.' (72) TopP Spec Top' ta Top 'he' ShiP Shi’ Shi shi TP VP 'pro yesterday £ ' Where the element de occurs between the verb and its object as in (73) it was argued that the re-analyzed tense head de raises to the verb and cliticizes to it, essentially becoming a tense suffix attached to the verb as is found in many other languages: (73) wo shi zuotian mai-[de]j piao tj I SHI yesterday buy-DE ticket ’ It was yesterday that I bought the ticket.’ If one considers again the structure and derivation in (72), it was suggested that the VP complement o f T° undergoes raising from its base-generated position to the right of de up to the SpecTP position. Such raising was motivated on the grounds that de 305 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is derived from the D° element de and D° de was argued to trigger raising o f an IP to its Spec for phonological support (see chapters 3 and 4). Such a structure now provides exactly the model necessary to understand the V2 change in RVC's. What we have arguably found with de is a structure in which a functional head T° has its sole argument occurring rather exceptionally in its leftward Spec position. The interpretation o f such a structure is that T applies its functional tense value to the VP located in its Specifier position. Turning now to early [Vt NP V2] resultatives, we have argued that the emergence o f phase-type verbs in the V2 position indicates that after time resultative V2 elements could be licensed by simply providing a general telic bounding of the situation described in the predicate VP rather than necessarily also providing a predication onto the object of V[. It was then suggested that this change points to a significant change in scope of the V2 from being a function applied just to the object NP to being a function applied to the whole VP. The general telic completives of this type are therefore interpreted as functional elements which essentially take entire VPs as their arguments. The question we then raised was how it could be possible for such a semantic relation to be syntactically encoded in [Vi Object V2] forms. Now having briefly re-examined the shi-de construction, we have indeed a model and precedent on hand which can suggest what the critical syntactic re-organization in RVC's might have been. I would here like to propose that the first important change which introduced the phase V2* s into the |Vj Object V2] resultatives was that the V2 position instantiated by phase-type verbs became re-interpreted as a functional Aspect head projecting a 306 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. single argument VP in a leftward Specifier position as in (74), precisely on a par The development o f phase-V2's into a predicate-final functional projection is then suggested to be very similar indeed to the ongoing development of de into Tense. Shortly I will turn to examine the actual re-positioning of the phase-V 2 Aspect-head and suggest that this second major change in RVC's also closely resembles the re-positioning of de as Tense. Just as de was argued to re-position itself after its re-analysis as a Tense-head and attach itself leftwards to the main verb o f the predicate, the re-positioning of V2 elements to attach to Vi in RVC's will be suggested to have occurred as a later direct consequence of the first stage of re analysis outlined here, V2 elements essentially becoming re-interpreted as aspectual suffixes in the same way that de is fast becoming a new tense suffix. Precisely how (and also more explicitly why) this second stage of re-structuring in RVC's takes place will be considered in greater detail in a later section. Returning to (74) above and focusing on phase-V 2's, it can be noted that such a structure can indeed be argued to accurately represent the relation between V 2, V[ and the object, and that V2 in fact selects for Vi and the object as its (VP) with de as T° projecting a VP in a leftward Specifier: (74) Early [V1-NP-V2] with phase V2's AspP VP Asp' Vi Object Asp (=phase V2) 307 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. argument.1 0 One should remember that diachronically from what might appear to be a fully open set of possible V2 elements there emerged a more specialized and restricted/closed sub-set of phase V2 's. These phase V2's do not predicate directly of the object but instead are classed as being appropriate for certain classes of situations. Each member of the phase V2 set is interpreted as applying to a different large set of situation-types rather than specific object types and this is precisely why textbooks commonly include lists of the V 1 verbs which are appropriately used with each V2, that is, the V2 indeed selects for the VP and its Vi head rather than having any direct relation to the object of that V 1. Such a selection relation is exactly what is represented in (74)~the V2-Aspect-head selects the VP headed by Vi as its sole argument, projected like an unergative verb in a leftward Spec position. A selection relation of this type also corresponds to the standard relation understood to exist between a functional element and its argument. As phase-V2's are elements which have all generalized their applicability to a wide array of events (completion of the event), this generalization would characterize them as functional elements requiring an argument to apply their function to. Finally, although one finds that there are in fact multiple potential overt instantiations o f essentially the same completive Aspect- 1 0 Note that Korean has Vr V2 sequences and that aspectual completion is frequently expressed by a V2 peli-ia which literally occurs as a verb meaning 'to throw away.' What is important to note is that this V2 has clearly grammaticalized and been re-analyzed as a functional head which is structurally higher than the V[ and its object, just as has been suggested here in Chinese. In the Korean Vr V2 resultative/completive sequence, V! occurs with its object in a structurally subordinate VP marked with the particle -e and the V2 occurs as the higher verb inflected for tense and declarative-marking, as illustrated in (i): (i) ku-nun sihem-ul phokiha-e peli-ess-ta he-TOP exam-ACC give-up-E throw-away-TNS-DEC 'John just gave up on the test.' (Lee 1993) 308 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. head rather than simply a unique morpheme (as might be the case in other functional heads such as Tense), it can be noted that elsewhere there are cases of clear functional elements which have multiple instantiations. One good example here is the nominal classifier system found in Asian languages, where each member of a potentially large classifier set functionally selects for a different range of complement types (and in this sense classifiers are rather similar to the V S aspectual completives, the latter 'classifying' different situations according to telicity). Before closing this section and continuing to examine how literal VS's fit into the picture of change sketched out above, I would briefly like to mention that there is also certain evidence ffom tone sandhi phenomena that phase VS's have a functional rather than a lexical status, as suggested here. In Shanghainese (Selkirk and Shen 1999) one finds that tone sandhi clearly differentiates between lexical and functional elements in an interesting way. Essentially a phenomenon of 'tone-spreading' begins at the first syllable of a lexical word and continues (spreading) throughout all following syllables until a second lexical element is met. Functional elements are distinguished ffom lexical elements in that they do not constitute independent tone sandhi domains and simply undergo the spreading initiated by a preceding lexical syllable (regardless of their syntactic relation to the preceding lexical element). Relevant here is the fact that V2's such as taw (=Mandarin dao 'arrive') in (75) are treated as if they are functional and not lexical elements, undergoing tone-spreading ffom the preceding lexical syllable and not instantiating a new tone sandhi domain: 309 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (75) (MH M H)(LH HL) (M H ) (L H) tsou taw 'noetsiN — > tsou taw 'noetsiN 'walk to Nanjing' In the southern Wu dialect o f Wenzhou reported in Chen (2000) the same basic phenomena occur. Chen presents the following example and notes that the tone of the V 2 dao 'arrive/to' is deleted as with all functional elements and dao receives a tone spread from the preceding lexical syllable. In Chen's notation the material between slash brackets indicates a tone sandhi domain; the inclusion of dao in the tone sandhi domain initiated by the preceding lexical V) element la 'puli' indicates that dao is interpreted as being a functional element rather than a lexical element:1 1 (76) / ba sanlunche / la dao / menkou / BA tricycle pull arrive gate 'Pull the tricycle to the gate.' 5.2.3 Literal V 2 and Re-analysis Above I have largely been concerned with the re-analysis of phase-type V2 elements which do not predicate of the 'object' NP in the [V) object V2] sequences. It was first suggested that early literal V2 elements supplied two discrete properties in resultative constructions: (a) more explicit characterization of the result state of the NP object of the main verb, and (b) a general telic bounding of the predicate. It was proposed that the emergence of V2 elements which do not provide property (a) is indication that at 1 1 Note that if no lexical material occurs to the left of a functional element it may be phrased with the lexical element to its right, this occurring here with ba. 310 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. some stage providing simply property (b) came to be sufficient to be licensed as a V2 in resultatives. These V 2 elements which had no direct relation to the object NP but supplied a more general telic bound to the predicate became re-analyzed as being base-generated in an aspectual head Asp0 with a single selected VP argument projected in its Spec position. The obvious question which now arises concerns the status o f other literal V t !s in [Vi Ob V2] sequences and whether literal V2's occurred in Asp0 or continued to be base-generated and licensed in small clause type predication structures. Note that the suggestion that phase-type V2's occur in Asp0 does not exclude the possibility that literal V2's might indeed have continued to be base-generated and licensed in a predication relation with the object and that there could in fact have been two different resultative structures (for litreal V2's and phase V2's) present in the language. Shortly I will however argue against such a view and suggest that also literal V2f s underwent re-analysis and came to be base-generated in the same Asp0 even in early Chinese, even though it might start out in the small clause structure presented in (67/68). It was an important change in focus on the critical licensing properties of literal V2 which preceded and allowed for the occurrence of phase V 2's. I have pointed out that historically the literal V2's first came to appear in [V1 object V2] resultaives, then the phase V2's was allowed in the V2 position of the same structure (G. Wu 1999), and finally both literal V2's and phase V2's collectively underwent a shift over the object in the Song Dynasty (Li and Y. Shi 1997). I suggest that the fact that both phase and literal V2 elements collectively underwent re-positioning to V| is strong evidence that all V2's came to 311 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. be licensed and interpreted very much in the same way, therefore causing all VVs to undergo a shift over the object from Asp0 to Vj. First I will begin with one of the rather simple reasons for assuming that early literal V2 must have come to instantiate Asp. Considering again how phase-type V2's came to occur as Asp, it was earlier suggested rather informally that literal V2 elements provided two properties— (a) a description of the result state of the object and (b) a telic bound— and that at some stage elements came to be licensed as legitimate W s by simply providing property (b) (i.e., the occurrence of phase VVs). What this change can be argued to indicate a little more formally is that it must have first been possible for literal V2's to be licensed by just providing property (b) before phase V js could then also make use of this less restrictive licensing option and occur in [Vi obj V2] resultatives. One can assume that phase VVs indeed came to be possible in [Vi obj V2] resultatives precisely because earlier literal VVs were optionally licensed in the V2 position solely in virtue of providing property (b). Once literal VVs could be licensed primarily via supplying just property (b), this then allowed for other phase-type elements which did not provide property (a) also to occur as V is. Thus sequentially it can be suggested that phase VVs could really only have been permitted into the [V| obj V2] structure once other earlier literal V is were re-valued in terms of their contribution to the construction. The imagined transition can be schematized as follows: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (77) Step 1: a. Only literal Vi's occur in early [Vi obj V 2] resultatives. b. Literal Vi's provide both properties (a) predication of the object, and (b) addition of a telic bound. c. Literal V2 S are licensed primarily due to property (a). d. At this point V 2 elements which do not predicate of the object (i.e., phase Vi's) do not occur in [V[ obj V2] resultatives, and the literal V2's is predicated of the object. (78) Step 2: Literal Vi's which still provide both properties (a) and (b) come to optionally licensed just in virtue of property (b), in which case the literal Vi's are analyzed as Asp. (79) Step 3: Licensing simply in virtue of property (b) allows in phase-type Vi's which do not supply property (a) at all to also occur in early [V1 obj V2] resultatives. This conclusion of course means that clause-final Asp was in fact first created by an initial re-analysis of literal Vi's into Asp. I will now go on to suggest that this re analysis occurred due to a simple switch in focus in early resultatives. Cross-linguistically, it might seem that both properties (a) and (b) above are regularly present in resultative constructions. Reflecting on their frequent co occurrence one might also initially assume that they are equally weighted in terms of relative importance and that the lexical information provided on the result state of the object is necessarily as important and highly-valued as the bounding effect which occurs. However, it is in fact quite possible that there actually is a critical imbalance 313 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the relationship between the two properties and that one of the two properties is or becomes more highly valued than the other. For example, it is possible that the lexical descriptive property (a) is of more importance in English 'so that' resultative forms and that the focus of interest largely falls on the characterization of the result- state, with the bounding effect (property (b)) being more of a secondary pragmatic bi-product: (80) Mary laughed at John so much that he became embarrassed. In (80) the emphasis is clearly felt to be on describing the type of result state ('embarrassed') rather than focusing simply the completion of an activity (i.e., the termination of laughing). In 'so that' type resultatives it would therefore seem that there is a clear imbalance between property (a) and property (b) and that the latter is of greater importance. In other cases it is not unlikely that an imbalance might swing in the opposite direction and the telic bounding property might become of greater focus than the actual description of the type of result-state. This is precisely what I would like to suggest has occurred in Chinese resultatives. From an earlier stage where the descriptive property (a) was clearly important and necessary for an element to occur as a V2 , later on the emphasis in resultatives switched to the bounding property (b) and simple completion of an event. In this regard, it has often been noted that the V2 in RVC's does indeed carry a special focus in the construction, for example Li & Thompson (1981, p.57) state that "the primary communicative function of an RVC is to comment on whether the result of an action did or did not, 314 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. can or cannot, take place." Thus the emphasis is suggested to be on whether an action is completed, and not so strongly on the characterization o f the type of completion. Furthermore, it is often rather clear from the consideration of a V 1-V2 combination that if there is indeed a focus on Vi then it can only be emphasizing the completion rather than the type of result state. For example, if the V 1-V 2 sequence xi-ganjing (yifu) is used as in (81) below, the emphasis is on completion and not on contrasting the result-state of being clean with other potential result-states. Most speakers do indeed agree that the interpretation of (81) is similar to (82), in the latter case the Vi indicates the end point, instead o f the result state (clean or not), of the washing event. (81) ta ba yifu xi-ganjing-le she BA clothes wash-cIean-LE 'She washed the clothes clean.' (82) ta ba yifu xi-hao-le she BA clothes wash-well-LE 'She finished washing the clothes.' Basically (and this point will be returned to shortly), with a V 1 such as xi 'wash' there is almost no choice in what one could use or expect to hear as a V2 element, and ganjing 'clean' or hao 'well' is indeed neutrally anticipated as the result state V2 rather than any other element. The characterization of the result state by means of ganjing 'clean' is then neither unexpected nor in real contrast with other states which could have resulted from a washing action. It is not its lexical content which is really Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. focused here but simply the telic bound o f completeness which it marks through its occurrence in the V 2 position. When the particular type of V2 is almost predicted by the type of V 1 like this, it is then very clear that the bounding effect is the important emphasized contribution of the V2, and the lexical characterization of the result-state is almost redundant (i.e., there is basically no other state that clothes could naturally be expected to be in after having been washed to completion). The suggestion is then again that an imbalance arose in the relative importance of properties (a) and (b) in Chinese resultatives. Consequently, from being a resultative construction in which the lexical description of the result-state was both clearly very important and necessary (hence only literal V2's predicating of the object could occur), the V2 is suggested to have later come to be valued primarily for its bounding function, with property (b) being the identifying and licensing characteristic of any V2. Formally V2 is therefore suggested to have come to be interpreted as more of a functional element providing primarily a telic bound to the predicate and so being licensed as a VP-operator selecting the entire predicate as its argument. Syntactically, the hypothesized switch of early literal V2's to a predominantly functional role selecting and applying to the VP can then be argued to have given rise to the predicate-final Aspect projection, with the V2's being base generated in the Asp0 head position. Following the creation of the aspect projection, any type of V2 which could provide property (b) could then legitimately occur licensed in resultatives, and this can be argued to have allowed in phase-type V2's which only supply property (b). Possibly there may also have been a period when 316 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. both the new Aspect-functional structure and the older V2 prevocational small clause structure may have continued to co-exist in resultatives, but importantly a good many literal V2's must have been re-interpreted in terms of their focus as providing primarily a bounding function for the creation of the aspect projection, which allows phase V2's to have occurred. 5.2.4 Responses to Apparent Objections Although I believe that there is no other good formal explanation for how non-literal phase VVs could have come to be licensed in resultativs, there might appear to be many objections to any suggestions that literal V^'s might also occur in Asp0. As literal VVs could quite naturally be analyzed as occurring in some kind of predication relation with the object and are indeed assumed to have come from such a source, it could simply be assumed that they continue to be licensed in such structures. Secondly, V js might seem to provide key information characterizing the type of result-state the object comes to be in as a result of the action o f Vi, and it might be natural to assume that lexical descriptive information of this type is presented in some kind o f predication structure rather than with the V2 occurring as a functional element in an aspectual head. Thirdly, the literal V2 elements might seem to be a fully open class with essentially any predicate being potentially available as a V2 result. If this is so, it might again seem to argue against literal V2 status as a functional head, as functional heads are characteristically instantiated by a generally small closed class of elements. However, here I will attempt to show how all of the 317 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. above-mentioned objections can in fact be given satisfactory explanations which strengthen the analysis of all "V ^'s as Asp. One important hypothetical objection to analyzing literal VYs as instantiating a functional head was that literal V2's might seem to be an open class of elements whereas functional elements are normally thought of as a closed set, often with only a single morpheme able to occur as a particular value of a functional head. While this may indeed quite frequently be true and characteristic of functional heads, it is however certainly not always so and there is also nothing a priori which requires that a functional projection be instantiated by a closed and highly reduced set of elements. One example of an assumed functional head which is instantiated by a significantly large membership of overt elements is the classifier position found in eastern Asia, with the languages of Southeast Asia having particularly abundant classifier systems. Note that within classifier systems there are classifiers which may in fact be used with only a single type of noun, this being similar to English 'a pack of lions,' a 'school of dolphins,' a "murder of crows" and "a gaggle of geese" where the classifying elements only used with their respective nouns. A result of this is that classifier systems clearly may potentially be very large with different classifiers corresponding to all the different feature combinations that a particular language chooses to distinguish in its nouns (i.e., +/-flat, +/-long, etc.). In Thai, for example, it is assumed that there may be over 400 distinct classifier elements (p.c., Andrew Simpson). Consequently the hypothetical objection that literal Vj's could not 318 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. instantiate a functional head because they come from a large open-ended class is not in fact so strong a criticism as might be thought. Related to the size of the set of morphemes which instantiate a functional head, it is natural to conclude that the membership of such a set is often small because functional elements may simply have a functional specification and no additional lexical content, and that a single element with such a functional specification will therefore be sufficient to encode the necessary function in the head. However, it has often been observed that when elements grammaticalize as functional elements they do not in fact immediately lose all lexical content and indeed may continue to show even strong traces of earlier lexical content for a considerable time (see e.g., Bybee, Perkins and Pagliuca 1994). As suggested here, what may often happen in grammaticalization is that there occurs a significant shift in emphasis with the continued use of an element. Initially an element has a predominantly lexical descriptive role, which may also give rise to other semi functional pragmatic inferences as an accidental bi-product. Later on this may change to a state in which the functional value of the element becomes most highly valued and the element comes to be predominantly functional. When this happens, the element grammaticalizing does not necessarily lose its lexical content; instead there is simply a switch in focus to the mainly functional role it plays. Consequently there is no a priori reason why there should not exist a set of functional elements, such as the V2 set, which are legitimized as functional heads but also have clearly lexical content as well. 319 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In grammaticalization situations, often one element from a larger set may become more widely used and general in its application (the phenomenon of 'generalization,' Hopper and Traugott 1993), with the result that other member of the set occur less frequently and eventually disappear. One might therefore possibly not expect for such a large V2 set to exist as functional heads without such generalization occurring. In fact, as will be seen in chapter 6, it is quite possibly true that verbal -le may indeed be just such a generalized completive marker occurring with a whole range of verbs, and therefore that there has been generalization in the completive aspect group. If this is in fact so, one might expect, incorrectly, that the other V2 elements would no longer be used as instantiations of the aspect head. However, I think there is actually a good explanation for why this has not taken place. We have noted that there is commonly felt to be emphasis on the completion of the event in a resultative construction. If this emphasis requires further stressing, it will only be possible to do this with a Vj which has the phonetic strength to carry stress and which has not undergone the phonetic reduction that has occurred with -le. Therefore the continued use of a range of V2's alongside the more general completive -le may well be due to phonological reasons and the focus properties of the construction.1 2 Furthermore, if one turns again to classifier systems such as that in 1 2 Arguably this phenomenon may show through when one considers the position of the verbal complex. If no object follows the V r V2 pair (as e.g., in a 6a-construction) and the V2 is then sentence-final, the V2 will normally carry a focus stress. In such a situation, it is more common to find a completive V2 with the phonetic strength to carry this stress rather than the general completive - le. Elsewhere when an object follows the V r V2 pair and the object can absorb any sentence-final stress -le occurs quite freely (and verbal completive -le is possibly more common than other V2 completives in such non-emphatic positions): 320 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Mandarin Chinese, one finds that alongside a large set of classifiers which are appropriate for subsets of nouns, there is also a very general classifier ge. Consequently it is indeed possible for a generalized functional element to co-exist with a second set of more specialized alternative forms, and the existence of a generalized functional head does not necessarily mean that other more specialized instantiations will automatically cease being used. Before leaving the general topic o f functional heads being commonly taken to correspond to closed sets of elements, I would like to make one more relevant point. There is possibly a tacit common assumption that because literal V j elements provide simple descriptions of the result state, the set of such literal V2 elements is most probably fully open. Then it should be possible for any descriptive verb to be employed as a V2, making the set of elements hypothetically associated with a functional head obviously very large indeed. Such an assumption is actually not so clearly true when one comes to probe a little further, and it has been commented on (i) wo ba shu kan-wan-Ie/??kan-le I BA book look-finish-LE/look-LE 'I finished reading the book.' (ii) nei-ben-shu wo kan-wan-Ie/?kan-Ie that-CL-book I look-finish-LE/look-LE 'I finished reading the book.' (iii) wo kan-le liang-ben-shu le I look-LE 2-CL-book LE 'I read two books.' Note interestingly that it is technically incorrect to classify the relevant stress as being sentence-final because it actually occurs on the element preceding sentence-/e. If one assumes that sentence -le is a sentential operator and therefore in some structurally high position, the stress can therefore more accurately be classed as occurring on the most deeply-embedded element— the object in (iii) and the V2 in (i) and (ii). Such a conclusion clearly raises interesting questions about the position o f sentence -le and when the stressing rule applies (i.e., is the sequence to the left of sentence -le possibly raised there to Spec-/e from a rightward complement position, and does the stress rule apply to the sequence before raising?). 321 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in various works that result-state predicates are often more restricted than generally thought (see e.g., Ludeling 1997). Essentially what one often finds in resultative constructions is that there is a strong preference for V2-type result-state predicates to represent canonical results, or results which are frequently expected given the content of V[. For example, as was noted above, with a Vi such as xi 'wash' the type of result ganjing 'clean' is heavily predicted and expected, and it is difficult to imagine what other kind of predicate could occur as V2. Elsewhere one finds very naturally co-occurring V 1-V2 pairs such as those listed in (83): (83) kuo-da expand-big da-ying fight-win qi-lei ride-tired yong-guan use X until one is used to X In all cases the V2 represents a largely canonical or expected result. In certain cases there may be a small number of alternate possibilities; for example, alongside da- ying 'fight-win' there is also da-bai 'fight-lose,' but by and large the V2 is frequently predicted to a considerable extent by the V 1 and actually not so wide open as one might otherwise imagine. Such restrictions have been noted to occur in resultatives in English, German and many other languages. For example, Ludeling (1997) points out that only certain result-state predicates seem possible in a considerable number o f cases, as illustrated in (84-88) below: 322 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (84) a. He laughed himself insane. b. *He laughed me happy. c. *He cried me sad. (85) He ran me tired/*exhausted. (86) Sally painted the door red/*beautiful. (87) We laughed the speaker off the stage/*embarrassed. (88) Robert ran clear of the door/away/*exhausted. In many cases it might seem that the result described by the secondary predicate is best when it is the most general canonical description o f what might be expected as a result given the action of the main verb. For example, in (85) 'tired' is fine where 'exhausted' is unacceptable; as 'exhausted' is simply a more explicit characterization of being 'tired,' it seems that the simplest and most general descriptive form is preferred over more explicit less common forms.1 3 In (86) too, the action of painting leads one to expect a color result state and it is found that a non-color result state such as 'beautiful' is unacceptable. Elsewhere the result states have the feeling of being idiomatic and also not open, as for example in (84) and (87). What this may be taken to indicate is that the set of result state predicates is possibly not so wide open as might be initially assumed. The preference for 1 3 The preference for canonical commonly occurring general forms such as 'tired' rather than 'exhausted' in resultative constructions focused on completion can be linked up with a universal common preference for generality found in grammaticalization. In Bybee et al (1994) and elsewhere it is noted that it is standardly more general verbs such as 'come' or 'go' which tend to grammaticalize as functional tense and aspect elements rather than semantically related but more specific manner of motion verbs such as 'amble,* 'wander,' 'stroll,' etc. The more general forms allow more easily for focus to fall on the particular abstract tense/aspect function they encode whereas more specific descriptive verbs have stronger lexical content which distracts away from functional uses. 323 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. canonical results encoded by general and commonly-occurring predicates also suggests as before that the primary function of result state predicates in many resultative constructions is indeed to supply a telic bound to an event rather than to focus on the description of the result state. Where a non-canonical result state predicate occurs this will tend to attract the focus away from the emphasis on completion and so there will be two competing focal points, resulting in the oddity felt in many RVC's. The idiomatic nature o f many resultatives also suggests that the result state may be so commonly expected from application of the Vj that the result state and the V i become a quasi-lexical unit. Similar patterns may arguably also be discerned in Chinese, and not all V 1-V2 pairs are felt to be well-formed. For example, (89-91) below are all felt to be either quite unacceptable or rather odd: (89) ??ta xiao-le-le wo he laugh-happy-LE me intended: 'He laughed and as a result I became happy.' (90) ?*ta (shuo gushi) shuo-ku-le wo he say story say-cry-LE me intended: 'He told a story which made me cry.' (91) *ta ku-beishang-le wo he cry-sad-LE me intended: 'He cried so that I became sad.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (92) *ta xi-ganganjingjing yifu he wash-clean clothes intended 'He washed clean the clothes.1 Note that I have suggested that the focus is on the completion o f the event in V 1-V2 resultatives rather than on the description of the result state, and that this imbalance in favor of property (b) is what has resulted in V2 elements being interpreted as Asp. I also suggested that this imbalance might not necessarily be present in other resultative constructions which could quite possibly value property (a) more highly than property (b). This would seem to be true of the English 'so-that' resultative already mentioned, and also Chinese [ Vi-de NP V2 ] forms. Because the V2 must encode a genuine independent description of the result state, one finds that only genuinely predicative V2 elements may occur in such forms. As expected, English resultative particles and Chinese phase V fs which do not predicate of the object may not occur in such resultatives: (93) a. *1 looked (at) the number so that it was up. b. I looked the number up. (94) a. *1 threw him so that he was out. b. I threw him out. 325 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (95) a. *wo kan-de shu wan-le I look-DE book finish-LE intended: 'I finished reading the books.' b. wo kan-wan-le shu le I look-finish-LE book LE 'I finished reading the books.’ (96) a. *wo zhao-de shu dao le I search-DE book arrive LE intended: 'I searched and found the books.' b. wo zhao-dao-le shu le I search-arrive-LE book LE 'I searched and found the books.' (97) a. *wo mai-de shu zhao-le I buy-DE book arrive-LE intended: 'I managed to buy the book.' b. wo mai-zhao-le shu le I buy-arrive-LE book LE 'I managed to buy the book.' This ungrammaticality frequently found when one attempts to convert a V1-V2 RVC into a corresponding [ Vj-de NP Vt ] resultative strongly confirms both that phase V2's do not predicate of the object in V 1-V2 RVC's (as they cannot be predicated of the same NP in [ Vi-de NP V2 ] resultatives), and that phase "V ^'s are consequently not licensed in V 1-V2 RVC's in virtue of any predication relation with the object but rather for their pure bounding effect. 326 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The ungrammatical examples involving literal VYs in (89-91) can further support the suggestion that literal VYs also instantiate Asp, when compared to their [Vi-de NP V2] counterparts. [Vi-de NP V2] resultatives consequently require a genuine predication relation onto the NP preceding V2 and so may be considered to require and possibly highly value property (a). This contrasts with V 1-V 2 RVC's which focus property (b) and do not require property (a) with V 2 S. Given this potential difference in focus and orientation between the two resultative constructions, one might wonder if those literal V j elements which do not occur well in V|-V2 RVC's might possibly be better in [ V r de NP V 2 ] resultatives, and quite interestingly this turns out to be so in an obvious number of cases. (98-100) are [Vi de NP V2 ] equivalents to the unacceptable/odd V 1-V2 RVC's in (89-91) and are all fully well-formed. This suggests that literal W is in V 1-V2 RVC's do not form a predication relation with the object. Linguists such as Sybesma have argued that literal VVs in V 1-V2 RVC's are derived from the [Vi-de NP V2] resultatives, and therefore assumed a predication relation for V 1-V2 RVC's. As we can see from the comparison between (89-91) and (98-100), the derivation from (98-100) to (89-91) can not be held true and neither can the predication relation in V 1-V2 RVC's. (98) ta xiao-de wo hen le he laugh-DE I very happy 'He laughed and as a result I became happy.' R e p ro d u c e d with permission of the copyright owner. Further reproduction prohibited without permission. (99) ta (shuo gushi) shuo-de wo ku-le he say story say-DE I cry-LE 'He told a story which made me cry.' (100) ta ku-de wo hen beishang he cry-DE I very sad 'He cried so that I became sad.’ (101) ta xi-de yifu ganganjingjing he wash-DE clothes clean 'He washed so that the clothes are very clean.' What all the above might seem to show is that V2 elements may actually be more restricted than imagined in V r V2 RVC's and that this is not because of any inherent semantic incompatibility between a particular result state and a main event verb V 1 (as these V?'s may co-occur in a different type of resultative with the same Vi's), but rather because of the focus of the V 1-V2 RVC. V 1-V2 RVC's are primarily oriented around bounding the activity rather than describing the result state and hence frequently predictable, canonical result type predicates are naturally preferred because they result in the least distraction away from the focus on completion. Consequently the potential objection to a primarily functional analysis of V2 elements that they constitute a fully open random class is again actually not so strong and V2's turn out to be often surprisingly more predictable than imagined.1 4 1 3 1 4 A further observation that also indicates that V2 's may frequently be highly predictable and hence not fully open is that it is very difficult to stress a V2 and understand this as contrastive. This is because it is simply often difficult to imagine what the V2 might contrast with, i.e., what other possible outcomes application of the V, to the object could result in: 328 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (i) ?*wo ba yifu xi-GANJING-Ie, bushi HONG-Ie I BA clothes wash-clean-LE, not red-Ie ?*'I washed the clothes CLEAN, not RED.' Furthermore, in German resultatives (see LOdeling 1997) focus-topicalization patterns show that Vi- equivalent type elements are often not really new information but actually rather predictable. In (ii) below it is seen that the adjective kaputt 'broken' can appear fronted in a simple sentence licensed by encoding new information in the focus-topic position: (ii) Kaputt ist die Tilr broken is the door 'The door is broken.’ However, when kaputt occurs in the resultative adjective-verb pairing kaputt-schlagen 'hit-broken' it can only occur in the fronted position if it is emphasized by an additional intensifier such as ganz 'quite' or vollig 'completely' as shown in (iii). This indicates that when it occurs in a resultative the adjective does not encode information about a new state in the same way that it does in (ii) and is instead licensed primarily as a telic bounding element. Only if further intensification is added to the adjective can it therefore be licensed in the focus-topic position: (iii) *(Ganz/v8llig) kaputt hat Max die TQr geschlagen. quite/completely broken has Max the door hit 'Max completely smashed the door.' 1 5 Having argued at length that V2 elements are licensed in RVC's in virtue of providing the general bounding property (70b), I have not discussed how any lexical descriptive content they may have comes to be interpreted. There are a couple of points and possible suggestions which can be made here. First of all, it has been noted that many functional elements do also have clear lexical content; if these elements are however licensed as primarily functional and therefore base-generated in functional heads, it must be assumed that any lexical content they have may be built into the structure from their position o f origin in the functional head. Such an assumption may now also be made for V2 elements argued to be base-generated in Asp and the lexical content o f literal V2 's may be interpreted from within the functional head Asp. Note also that V2 -Asp is taken to select for the VP and hence the type of situation represented by the VP. The frequent interpretation that the object in the VP has the result state property represented by V2 may then be a function o f V2 selecting for the situation in which that object is affected (e.g., a 'washing' event is one which can be selected by a V2 - Asp 'clean,' and the object of the washing may then naturally be interpreted as having such a result state property). In some way the selection relation here is similar to that existing between classifiers and nouns; just as a classifier is sensitive to certain features of a range of nouns (e.g., +round), V2 's may classify situation types according to their potential natural outcomes, and in both classifier and V2-Asp case this translates into a selection relation. There is also a possible similarity here to the idea of 'pragmatic association' proposed in Kayne (1985) and taken up in Sybesma (1999) in a different way from the proposal here. It can be suggested that if there is a washing event involving an object and this is terminated by some clean state, then it is just pragmatically natural to associate the object with that clean state. Another alternative might be to make use of Borer's (1994) idea that objects check case in the Spec of a telic aspectual head and suggest that at LF the object raises to an outer SpecAsp to check its case and at the same time is predicated of by the V2 in Asp. This would be similar to analyses of fougA-movement in which the surface subject is raised from object position and predicated o f by the entire roug/i-predicate when in its raised position (see e.g., Brody 1994): (i) John; [is easy to please q] 329 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. So, to briefly summarize, in this section I have argued that a significant change occurred in the underlying structure o f early Vi Ob VS sequences. From an original form in which V2 occurred licensed as a genuine secondary predicate, the V2 underwent re-interpretation as instantiating an aspectual head which encodes a telic bound on the event depicted by the V| and its object. Such a change was suggested to have been the result of a change in focus o f the resultative construction. An increase in the focus on the pure bounding property of V2 later allowed phase VVs with no direct predicational relation with the object NP to also occur in early V 1 Ob V2 sequences. The syntactic realization of this change was argued to be that V S became base-generated in a discrete Asp-head selecting VP as its argument projected in a leftward Specifier position. This has a clear parallel in the structural re-analysis of de as Tense also projecting its predicate argument in a leftward Specifier. Finally I reflected on certain potential objections to the claim that V2 instantiates Asp and attempted to show that a functional analysis of V2 elements can in fact be maintained. Flaving established the V2-as-Asp proposal and showed how it is able to offer certain insight into the early development of resultatives, I will now proceed to examine a second significant change in Chinese resultatives: the re-positioning of V2 to a position right-adjacent to V |. I will offer an account of how and why this important re-structuring took place and suggest that the change should be similarly attributed to the functional-aspectual properties instantiated by V2. Because the V2 -Asp is licensed in virtue of its functional-aspectual property (70b), such predication would not be forced to occur and phase VVs would not have to predicate of the object (in contrast to small clause analyses in which the V2 is licensed precisely as a predicate and therefore must predicate of the object in order to be legitimized). As we will shortly see, the situation is clearer in V|-V2 pairs 330 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.3 V 2 Re-positioning and Directionality According to G. Wu (1999) and Li and Shi (1997), V) N P 0bject V2 resultatives largely changed into V 1-V2 NPobject forms around the time of the Song dynasty (960-1279), possibly having first appeared during the Tang dynasty (Mei 1981). Certainly V 1-V2 Ob sequences are taken to be the only acceptable forms from the time of the Song dynasty onwards. The linear re-positioning of V2 over the object to become right- adjacent to Vi is represented again in (102): (102) V, N P obje« V2 -> V, V2 NPobject First of all it should be noted that the later forms do not seem to have resulted from earlier sequences via any Heavy NP Shift type operation which post-poses heavy objects. It is fairly clear that even light object NPs occurred after V2 in the new order and so the change is much more likely to have resulted from a re-positioning of the V2 and possibly movement of the V2 to Vj. The critical question in need of answering is of course what actually triggered this re-positioning and what this might then be able to tell us about the synchronic structure of V|-V2 forms?1 6 Both the after the second stage of re-analysis and so I will not discuss this further here. 1 6 As noted earlier, there are certain obvious parallels between Chinese RVC's and English verb-particle forms. In the latter there are currently two orders available, which might seem to reflect the two historical orders in Chinese. In (i) the particle follows the object, whereas in (ii) the particle occurs preceding the object and adjacent to the verb: (i) John found the information out. [vp V Object Prt] similar to early Chinese [V| Obj V2] (ii) John found out the information. [vp V Prt Object] similar to later Chinese [V! V2 Obj] 331 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. syntactic accounts o f Chinese RVC's we have reviewed assume that there is indeed movement of V2 to V| still actively occurring in the underlying syntax, and hence that the historical source is still present in the base-generated structure of RVC's; however neither account offers any understanding of what might actually cause the movement of V2 to Vj. In Y. Li's lexical approach by way o f contrast, reference is simply not made to the earlier [V 1 Ob V 2] forms and there is consequently no attempt to explain how syntactically created [V 1 Ob V2] sequences might later come to be formed as V 1-V2 compounds in the lexicon. The development of the modem [V| Vi object] forms from [Vi Ob V2] sequences therefore remains as an interesting puzzle. 5.3.1 Arguments Against a Phonological Explanation One possible mode o f explanation for the change to V 1-V2 forms which one initially might entertain could be that the re-positioning may have occurred for phonological reasons; this I will also argue against. Earlier in the introduction it was noted that the change in RVC's and with verbal -le/liao might seem to resemble the change in shi de forms in that an initial sequence of a main verb, an object and some sentence-final element X (where X= -de, liao, V?) undergoes re-ordering to [V X Object]. With de it was pointed out that the change in position arguably resulted from -de as a clitic targeting the verb as its new phonological host following earlier syntactic re-analysis It is pointed out in Johnson (1991) that the variant in (ii) is also not produced by any Heavy NP Shift operation as the object can be significantly lighter than in regular Heavy NP Shift examples. For example, simple NPs can normally not be extraposed in potential HNPS forms such as (iiia), yet bare NPs do easily occur following particles in verb-particle sequences: 332 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. as a tense element. Considering the V2 S in RVC's however, it is not possible to argue that such elements moved to attach to the main verb Vi in any parallel way because the V2 series do not show obvious signs o f being clitics. Both phase V2's such as wan 'finish,' dao 'attain,' zhu 'hold,' and non-phase V2 S such as ganjing 'clean,' lei 'tired,' etc. seem to occur quite unreduced in phonological shape when occurring in V 1-V2 pairs.1 7 It would therefore be somewhat implausible to attempt to treat such elements as clitics. Furthermore, as will be shown in a later section, there is also good evidence that verbal -le only underwent reduction from its full pronunciation as liao to its present -le form after it had moved adjacent to the Vi and so this re-positioning can similarly not have been caused by any cliticization targeting V 1. There is also a second way in which one might attempt to explain the V2 re positioning in phonological terms. Although it is unlikely that clitic-like phonological dependency in the V2 alone was responsible for V2 movement to Vi, there is a possibility that some rather more subtle phonological deficiency in the V/ could have resulted in the attraction of V 2 to the V[. It has long been noted that Chinese early on developed an increased tendency for bi-syllabic compound-type elements in places where mono-syllabic forms had previously often occurred, and frequently given explanations for this increase in bi-syllabicity significantly relate it to some other process of other phonological decay. A number of researchers (e.g., (iii) a. *John painted tj red the doorj. b. John painted tj red all the doors which face northwards,. 1 7 Directional V2 's do however lose their tone when occurring in RVC's. 333 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Karlgren 1923, Wang 1980, Norman 1988) have pointed to the fact that Chinese underwent a large-scale simplification in its syllable structure from being (C)(C)V(C)(C) to (C)V(C) and noted that the immediate decrease in the number of contrasting syllables available led to an increase in ambiguity among mono-syllabic words. It is then suggested that in order to compensate for this increased ambiguity, many earlier monosyllabic words came to be bi-syllabic, with the addition of a second syllable serving to disambiguate otherwise similar mono-syllabic words. A more recent and interesting treatment of the increase in bi-syllabic words along the same basic lines is found in Feng (1998). Feng agrees that the increase in bi-syllabic words in Old Chinese was most likely due to the simplification of the syllable structure, but argues against the view that it was increased ambiguity which caused the growth of bi-syllabicism. Instead it is argued that simplification of the syllable structure led to the occurrence of a changed prosodic structure in which new mono syllabic elements may no longer have been heavy enough to form minimal independent prosodic units/feet. Feng claims that consequently the new prosodic system required that a prosodic foot be formed by not just one but two syllables, this leading directly to the increase in new bi-syllabic words. Either the more traditional view relating to ambiguity or that proposed by Feng could clearly be used to attempt to account for the change in RVC's and the re positioning of V2 adjacent to V[. It has also been suggested that mono-syllabic V] verbs may have become either too potentially ambiguous or have been prosodically too light to stand on their own and that re-positioning o f the V2 adjacent to V| would 334 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. have served as a means to overcome this deficiency, creating unambiguous phono logically legitimate bi-syllabic resultative compounds (as proposed in Frank Li (in prep). Note that similarly ambiguous or phonologically weak mono-syllabic V2 elements would also have arguably benefited from such a re-positioning. Despite the possibility o f a phonological account of the creation o f Vi-V2 forms however, there are several reasons to be suspicious of such a mode of explanation. First of all, if main verbs essentially became either too ambiguous or prosodically too light to stand alone and unsupported, it is not clear why a resultative V2 and not other elements should have necessarily been selected as the means to overcome such a phonological deficiency. If mono-syllabic main verbs became too light or ambiguous on their own, one might expect that the NP object o f the verb could have been re-phrased together with the verb to create a prosodic unit with enough weight to stand alone. It is certainly true that a significant number of verb- object compounds do exist in Chinese so one might expect verb-object re-phrasing to have been the most naturally available solution to overcome any phonological weakness in a monosyllabic verb. The presence of the object with the main verb would have also arguably overcome any ambiguity present. Another option which could have been made use of would have been to add a second verb with a similar meaning to the first resulting in sequences. Such a 'strengthening/disambiguating' strategy may seem to have been made use of with nouns (e.g., dao-lu 'way-path') so could have been a natural means to strengthen/disambiguate verbs as well: such as da-ji 'hit-strike' and guan-kan 'view-watch,' etc. Nevertheless one finds that instead 335 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of these options it is a specifically resultative V2 which becomes added to the main verb and the obvious potential for the object of a verb to strengthen it seems to have been curiously ignored, which is very puzzling. If one reflects on the fact that [V] Object V 2] sequences became [Vi V2 Object], it is not at all clear why the presence of the object as a second mono-syllable adjacent to the Vi could not have been enough to support it and the Vj specifically required the re-positioning of the V2 to provide it an extra syllable. One would certainly not expect any phonological deficiency to be so semantically selective. Furthermore if monosyllabic elements became too weak to stand as legitimate prosodic units in Old Chinese and this triggered the creation of V 1-V2 forms, one would expect to find the necessary presence o f bi-syllabic units throughout the language; however, singular pronouns and a large number of adverbs did not obligatorily add on second supporting syllables. As the suggested dramatic change in phonological structure should have affected all phonological units in the same blind way, the particular increase of V 1-V2 resultative forms tends to stand out as somewhat odd. Note also that it is indeed phonologically legitimate for mono syllabic verbs to stand licensed in sequences of [V1 Object V2] in Chinese if they have a non-resultative purpose interpretation, as shown in (103). This again indicates that there is nothing ill-formed in such strings from a phonological point of view: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (103) (wo qu) [Vp [vi mai] [o b j dongxi] [v2 chi ]] I go buy things eat 'I am going to buy something to eat.' A third reason to be wary o f any claim that a change in the prosodic structure directly triggered the re-positioning of V2 to Vi is that dramatic phonological change of this type (i.e., movement of V2 over the object to Vi) in general tends to follow on from some underlying syntactic change rather than precede and bring about new syntactic forms/changes (see e.g., Harris and Campbell 1995). So if there are phonological changes in V 1-V2-NP resultatives, they would have in principle occurred after the repositiong of V2 ; the phonological changes do not trigger the repositioning of V2. Finally, and possibly most significantly, there is a telling and important time difference between the suggested rise of bi-syllabicity in Chinese and the formation of Vi-V2 resultatives. Packard (1998), Feng (1998) and others report that the noted shift from mono-syllabic to bi-syllabic words began at some time during the Zhou dynasty around 1000-700 BC, and that this turned into a sharp increase in bi-syllabic words during the Han dynasty (206 BC-220 AD). Feng argues that the prosodic changes responsible for the increase in bi-syllabicity also largely occurred around the Han period. This being so, if V 1-V 2 forms did result from the need for mono-syllabic main verbs like other mono-syllabic elements to be supported by a second syllable, one would clearly expect that V 1-V2 resultatives would have been formed at the time when other mono-syllabic elements became bi-syllabic~i.e., during the Han period. 337 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, it is well-reported that the critical change from [V] Object V2] to [V 1-V2 Object] resultatives only came about in the Tang dynasty (618-907 AD) and that [V(- V2 Object] only fully replaced [Vi Object V2] sequences in the Song dynasty (960- 1279 AD) which is well over 700 years later than the Han period and the time of the suggested change in prosodic/syllable structure. This seems to cast seriously in doubt any possibility that early phonological changes leading to necessary bi- syllabism was the trigger for the change from [Yi Object V2 ] to [Vi-V2 Object]. It seems that [V 1 Object V2] forms in fact existed quite legitimately for many centuries after the general switch to bi-syllabic forms in the Han period, and therefore that the change to V]-V2 resultatives arguably must have some other non-phonological trigger and explanation. 5.3.2 V 2 Elements as Verbal Suffixes One important aspect o f the approach towards RVC's being developed here has been to highlight and focus in on the obvious parallels with ongoing changes in the shi-de construction. In both RVC's and shi-de forms a sentence-final element (V \/de) re positions itself over the object, targeting the main verb, and settles in a position right-adjacent to the verb as in (104): (104) a. Vi Ob V2 -> V, V2 Ob b. V Ob de V de Ob Such clear and striking similarity in the change in both constructions suggests that some common re-structuring process is indeed likely to be responsible for both the 338 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. earlier RVC change and the present re-analysis in shi-de forms. Considering how we approached the latter construction in chapter four, we essentially first noted the positional change with de in northern dialects of Mandarin and then worked our way on to various claims about restructuring in shi-de forms. Observing from a range of data that de appears to be developing into a verbal suffix of some kind, we reflected on the kind of verbal suffix this was likely to be, and subsequently came to the conclusion that de is undergoing re-analysis as a past tense morpheme suffixed to the verb. Supposing now that the linear re-positioning in RVC's and shi-de forms does indeed reflect a common restructuring process and that sentence-final V2 elements in RVC's targeted the verb in the same essential way that de is currently doing, the natural conclusion to make then would be that resultative V2 elements in the Tang/Song dynasty underwent re-analysis as verbal suffixes. Having furthermore spent time clearly motivating an earlier re-analysis of V2 elements as heads of an Aspect phrase, the parallel with de might really seem to become very clear and in both cases a functional head with a verb-related meaning would be re-analyzing as a verbal suffix, (past) tense in the case of de, completive aspect in the case of V2 elements in RVC's. It is also cross-linguistically common for tense and aspect to be encoded via suffixes on a verb, and so the suggestion that V2 elements which were re-analyzed as Aspect ultimately came to be encoded as verbal suffixes would seem to be rather natural from a very general cross-linguistic perspective too. 339 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. However, despite the intuition that suffixation of V2 as Aspect is both generally plausible as an ultimate description of V 1-V2 pairs and that such a change is also specifically motivated by the parallels with de, there is nevertheless still much to explain here. First of all and as mentioned earlier, V2 elements are phonologically not obviously reduced or weakened, and so in this regard not so immediately like suffixes. Secondly, it is also not fully obvious that re-positioning of the V2 to Vi was actually the result of movement o f the V2 to V|, as might be initially assumed. Thirdly, although movement of de to the verb is possibly consistent with de becoming a tense suffix, it does not really explain this movement. It should be remembered that evidence was presented indicating that sequentially de first undergoes re-analysis in its sentence-final position and then raises up to the verb; having successfully re-analyzed as tense, it is not entirely clear why de should then subsequently need to raise to the verb. Noting that de is a clitic element, it could be suggested that this simply reflects a re-alignment of what its target host element is specified to be. That is, as an enclitic de has to target some phonological host for support, and if it encodes a classically verbal property such as tense then it may be natural for it to specify the verb as its target-host. Such a clitic-based explanation is however unavailable for the V2 re-positioning for the noted reason that V2 elements are not obviously clitic-like. Quite generally then there is a significant difficulty in understanding why V2 should have raised to Vi. Remember that the analysis developed so far has argued that V2 elements were first re-analyzed as instantiating a sentence-final Aspect head, and critically it was argued that as such aspectual heads 340 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the V 2-Asp elements became operators which apply their function to entire VP predicates. Such scope over the VP was therefore taken to be encoded by V2-Asp selecting VP as its (sole) argument and projecting this in its Specifier position. Given therefore that such a configuration with V2 in the head of AspP naturally corresponds to the scope and selection relation between V 2 and the VP, the obvious question in need of an answer is why V2 would re-position itself and adjoin instead to the V 1-head. Such a re-adjustment might seem to represent a decrease in its scope, with V2 c-commanding and having scope only over V 1 rather than the whole VP. A similar criticism could be possibly made o f the encliticization of de to V°— tense is an operator which takes a predicate as its argument and this is naturally represented in the structure with de in T° and the VP in its Specifier. If de re positions to adjoin to V° this might again seem to represent a decrease in its scope to just the V°, so there is clearly much to explain here. 5.3.3 V2 Elements Do Not Undergo Raising to V/ In what follows I will argue that the surface re-positioning of V2 to Vi did indeed directly result from the aspectual properties of V2 as Asp and does also represent V2- Asp turning into a verbal suffix. I will however claim that this re-positioning did not occur as the result of any movement of V2 to Vi but arose due to natural V 1-V2 adjacency. It will be suggested that due to focus on the V2 representing completion the object in such RVC's is frequently old information and therefore either fronted to a pre-verbal position or null as a pro, resulting in V 2 being linearly adjacent to V|. 341 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Such natural adjacency is then argued to lead later generations of speakers to re analyze the V2-Asp in Vj-V2 pairs as an aspectual suffix attached to Vi in the lexicon. The critical motivation for such a re-analysis is then suggested to be directionality— re-analyzing V2-Asp as a suffix results in speakers constructing an analysis in which the suffix V2-Asp is licensed at LF by a higher AspP generated to the left o f VP, with VP as its complement. Such an analysis then effectively serves to re-align the AspP in synchronization with the general head-initial direction of selection in Mandarin Chinese. Prior to this the earlier sentence-final AspP is shown to have grammaticalized in a way which was in disharmony with the canonical direction of functional selection, and re-interpretation of V2 -Asp as a suffix represents a natural correction made by speakers having a default head-initial parameter setting in their language. An important assumption made in the approach here is that there is constant pressure towards uniformity in the directionality o f selection in a language and that languages are indeed canonically either head-initial or head-final. Such a view of directionality is essentially present from the time o f Greenberg (1963) through other early typological work in Lehmann (1973), Vennemann (1974), Hawkins (1979) and continued on in Chomsky (1981) and other more recent work. At least among the earlier typologists, the view is also commonly held that if a language somehow exhibits a structure which goes against the basic directionality of selection in a language, there will be a general pressure for the language to re-align such a structure in harmony with the canonical direction o f selection. Vennemann (1973) 342 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. talks of a Principle o f Natural Serialization as a force which operates to restore consistency in languages which become 'inconsistent' (i.e., not uniformly head-initial or head-final) as the result of external influence or disruptive language-internal change. Lehmann (1973) also argues that if inconsistency creeps into a language then there will be pressure for this inconsistency to be corrected, and Hawkins (1979) proposes that a Principle of Cross-Category Harmony will similarly lead to languages re-aligning any non-uniform directionality. Commenting on the above proposals, Mallinson and Blake (1981, p.417) point out that correction of this type may be explained in terms of natural economy, and that it leads to a reduction in category/word-specific rules: "A disharmonic language requires more category- particular ordering rules and thus the grammar of such a language is more complex." In more recent work, Audrey Li (1990) and Travis (1984) seek to explain deviations from expected orderings by considering the possible interactions o f case and theta theory, while Kayne (1994) suggests that apparent ordering inconsistencies are in fact surface distortions of an underlying universal SVO/head-initial order which result from often large-scale movement operations. Descriptively, much research has shown that languages do tend to follow either a consistent head-initial or head- final ordering and that there would seem to be internal pressure for all categories to conform to the basic directionality present in a language.1 8 1 8 An interesting case here is Amharic as reported by Comrie (1989). Amharic having originated as a classically head-initial Semitic language, migration of the Amharic-speaking people to a head-final Cushitic-speaking area resulted in change in many of the basic orderings, due to the external influence o f the Cushitic languages. Pressure for uniformity has now led Amharic to become head-final in just about all its category types, preposition-NP ordering being the sole residue of the earlier head-initial order. 343 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The sentence-final Aspect projection argued to have arisen in Chinese as a natural result of V 2 elements being re-analyzed as instantiating completive aspect can now be suggested to be in potential conflict with the general directionality of functional/complement selection in Chinese. When the sentence-final AspP was initially proposed, we took as a model the structure assumed for de in the TP after re analysis, and suggested that the argument VP selected by V2-Asp is projected in the SpecAspP position. The parallel pointed to with TP headed by de was that the predicate complement of T°-de becomes raised into SpecTP. It therefore seemed to be a reasonable suggestion that the predicate complement of V2-Asp might similarly be simply occurring in a Spec position, likening this to in essence to unergative intransitive verbs which are taken to project their sole argument NP in a Specifier position. However, I would now like to argue that such a suggestion needs to be re assessed and that the label 'Specifier' here is actually rather deceptive. Consider again the basic structure which has been proposed for AspP: (105) AspP What needs to be reflected on is the way in which the term 'Specifier' is understood and the way in which a Specifier is licensed in the projection of a head. Canonically it might seem that a Specifier has the following general properties: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (106) Specifier properties a. a Specifier is an XP position b. the XP in the Spec position is either directly selected by the Y°-head o f YP if base-generated there, or agrees with the Y°- head if moved to the Spec position. c. the Specifier closes off the YP d. the Specifier is built into the YP after the complement of Y° is combined with Y°. Complements, have the general properties in (107): (107) Complement properties a. a complement is an XP b. the XP is directly selected by the Y° c. the complement is combined with its selecting head Y° as the first branching sister of Y° Abstracting away from the issue directionality in the projection of Specifiers and complements, it might seem that in one sense a Specifier is only really different from a complement in virtue of being built into a projection critically after the complement of the head has been combined with the head. Both Specifiers and complements are maximal projections commonly selected by the head of a projection; the Specifier simply happens to be inserted into the structure at a point later than the complement. While such an ordering relation does indeed result in a clear hierarchical difference when both complements and specifiers are projected as in (108) (assuming a Spec-initial complement-final pattern for illustration), if a head were not to project any complement, the difference between a Specifier and a 345 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. complement becomes less obvious. If indeed one follows assumptions commonly- made in recent work that a node is not projected unless it is genuinely branching (i.e., there should be no vacuous non-branching nodes (see Chomsky 1995a and references therein)), then configurationally the difference between complements and Specifiers reduces further, as illustrated in (109): therefore never have a complement branching off Asp' to its right. If the Asp' node should therefore not be projected (due the no vacuous branching nodes constraint), (105) should instead actually be represented as (110): never projects any complement XP off an intermediate Asp' level, that the leftward (108) X P Specifier complement (109) a. YP XP Specifier b. YP complement Now, considering the structure for AspP in (105), Asp0 is a head which is suggested to have a Specifier which hosts the single argument of Asp0 (the VP) and will ( 110) VP Asp0 I would now like to suggest that in such a structure, where the head Asp0 importantly 346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. branching XP (i.e., the VP) can no longer be determined as a Specifier but is instead itself interpreted as a leftward-branching complement. In other words, one might suppose that a Specifier is a relation which may possibly depend on the potential existence of a complement XP, so that an element can only be functionally determined as a Specifier if there may also occur a hierarchically lower complement XP. Here in the AspP projection, the VP is essentially so similar to a complement of Asp0 that although we have been referring to it as a Specifier, in fact in the structure in (110) it gets actually interpreted as a complement. Note that the structure in (108) and (109) had two particular models: (a) de with VP in SpecTP, and (b) unergative intransitive verbs. In case (a) the VP was certainly originally raised to SpecTP from regular comeplement position and so T° did have a complement (therefore VP could be determined as a legitimate Spec for a time). In case (b) it might be assumed that unergative verbs have a structure such as (111a), assuming English as the relevant language here. If a no vacuous non-branching nodes constraint is adopted, this should then be represented as (11 lb): In (111b) this should mean that the NP cannot be determined as a Specifier and should be interpreted as a complement, making it an exceptional head-final structure 347 (111) a. NP V' b. VP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. in a language such as English. However, in recent work (e.g., Hale & Keyser 1993) it has been argued that unergative intransitive verbs do in fact have complement NPs in addition to Specifier subjects. One may either assume that unergative verbs have null cognate object type complements so that an example such as: 'John sighed' is interpreted as: ’ John sighed a sigh.,1 or one may suggest that ’ John sighed.’ is possibly derived via incorporation of the N° head of an object NP ’ sigh’ into V°, represented as (112a) and (112b) respectively: (112) a. VP NP V’ John V° NP I I sighed (a sigh) b. ^ V P ^ NP V’ John V° NP nigh In either case there would be no head-final structure as the subject NP could be legitimately determined as a Specifier due to the existence of the complement NP. Returning to Chinese (110) now, I would like to maintain that such a structure is indeed interpreted as being head-final, with the VP failing to be determined as a Specifier and instead being interpreted as a leftward projected 348 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. complement. Having argued in chapter three that Chinese is essentially a head-initial language, and having noted in this section that languages will tend to be either head- initial or head-final wherever possible, the grammaticalized AspP clearly goes against the general headedness and directionality o f complement selection in Chinese. Assuming there to be pressure towards a uniform directionality of headedness, one might naturally imagine that structures such as (110) would then be under pressure to be re-analyzed in a way conforming to the head-initial parameter if the opportunity for this arose. I will now attempt to argue that such directionality considerations are exactly what motivated the re-positioning of V2 to V * and show how re-alignment of the structure in a head-initial way was indeed able to take place. In a Minimalist (Chomsky 1995b) approach to morpho-syntax, X°-level functional morphemes which correspond to syntactic head positions are taken to be base-generated either as independent word-level elements actually in the relevant head positions or alternatively as affixes attached to other (generally) lexical X°-leveI elements. If an element is base-generated as an affix on a lexical head but instantiates a functional category (such as for example tense), it is assumed that the element (or the formal features it carries) will undergo raising from the base generated lexical head position to a relevant higher functional head position to be licensed at some point in the derivation. Concretely this allows for three possibilities, illustrated in (113-115). Considering here the example of a functional morpheme instantiating tense and how it occurs and is licensed in syntactic structure, if the tense morpheme is a free standing word, it may be base-generated separately 349 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from the verb and directly in the tense head T °a s in (1 1 3 ). It will therefore be licensed directly in this base-generated position. (113) is modeled on an SVO language with overt subject raising: (113) TP Spec subject, tense A second possibility is if the tense morpheme is an affix, it will be base-generated attached to the verb in the lexicon and then inserted together with the verb in V°. If the licensing of tense has to be effected in the overt pre-Spell-Out syntax in a language, this will then trigger overt raising of the verb-tense combination to T° as in (114): (114) TP Spec r verb-tnsk NP NP object 350 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The third possibility occurs when the tense morpheme is base-generated as an affix on the verb but licensing o f tense does not have to be effected in the overt syntax. Here it is assumed that the verb+tense affix (or the formal features present in the tense morpheme) will undergo covert post-Spell-Out raising to T° to be licensed/feature-checked at some point prior to LF. (115) represents the Spell-Out structure of this possibility with the verb and its tense affix still in situ in V°. Raising of the verb-tns combination to T° will occur later at LF: (115) subjecti T verb-tns Returning now to Chinese RVC's, the analysis of V2-Asp elements as occurring base-generated in a sentence-final Asp0 essentially corresponds to the situation in (113)~the functional morpheme occurs as a free-standing word directly in the relevant licensing functional head, and this has been argued to lead to interpretation as a head-final structure in disharmony with the head-initial parameter- setting in Chinese. Supposing however it were to be possible for V2-Asp elements to be somehow re-interpreted as aspectual suffixes attached to the V i in the lexicon, it could then be assumed that the V|-Asp unit is inserted directly into V°. Due to the occrrence of the Vi-Asp unit to the right o f VP-type adverbials such as manmarde 351 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 'slowly' (see example in (31)), ), it would then be assumed that the aspectual suffix is licensed later at LF, this corresponding in essence to the third possibility above where tense is licensed by LF-raising of the verb-tns unit to T° in (115). (116) ta man-manr-de xi-ganjing-Ie yifu he slowly wash-clean-LE clothes 'He slowly washed clean the clothes.' Significantly it then becomes possible to assume that the AspP projection which will license the aspectual suffix at LF in fact occurs as a fully regular head-initial functional projection selecting its VP argument in the canonical rightward direction. This is represented in the Spell-Out structure (117). It is assumed that the V|-Asp unit (or the formal features of the Asp-suffix) will then at LF undergo raising up to the licensing Asp°-head (and maybe also higher to T°), this now being represented by a broken line in (117). TP is included in the structure in order to place the subject; I return to consider the status and instantiation of TP and other higher functional projections in chapter 6. Note that I also do not include verbal -le here as this is examined in depth in chapter 6. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subject, T AspP Asp' VP Asp NP ‘ NP Vi-Asp(v2) Object (117) if instantiated with lexical items would result in (119) corresponding to the simple string in (118): (118) (yaoshi) wo kan-wan shu,... (if) I look-finish book 'If I finish reading the book,...' (119) Asp' VP Asp NP NP shu 'book' jcan-wan 'read-finish' 353 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Importantly then, re-analysis of V2 -Asp as a suffix attached to Vi in the lexicon and licensed via LF raising to Asp0 would critically allow for the grammaticalized sentence-final AspP to be re-interpreted as a regular harmonic head-initial structure. I will now suggest that certain aspects of the resultative construction conspiring with general properties of Chinese in fact made such a re analysis of the V2-Asp as a suffix very easy to arrive at. Earlier it was suggested that an important aspect of the earlier re-analysis of V2 as completive aspect occurring in Asp0 was that the focus in RVC's came to be squarely centered on the completion of the event represented by V2. Such emphasis on the V2 has been frequently pointed out (e.g., Li & Thompson 1981) and, as noted earlier, may well have arisen naturally due to the sentence-final position of V2 in early Chinese, sentence-final position attracting a natural focus. Now, if the V2 element does indeed encode the focus in RVC's, this will have a consequence that the object of the verb will normally not be taken to be in focus. This will in turn mean that the object in RVC's will frequently be old and given information rather than represent a newly-introduced NP (which, being new information, would constitute a focus). Consequently, commonly being old information, the object of the verb will often be represented either pronominally and hence normally by a null pro element in Chinese, or as a topic fronted to a sentence-initial position. It is also suggested in Li and Shi (1998) that many V1-0-V2 occurrences in the Song dynasty were later replaced by verb-copying, ba-preposing of the object and object topicalization in 354 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. order to prevent the intervention of the object in between VI and V2. We can represented these instances abstractly in (120). (120) a. Subject Vi (pro) V2 b. Subject VI OV1-V2 c. Subject ba-O VI-V2 d. Topicj Subject Vi (tj) V2 The result of such representation of the object in earlier [Vi Object V2] sequences will be that the Vi and the V2 are effectively heard linearly adjacent to each other. It can now be suggested that such frequently occurring linear adjacency of Vi and V2 would naturally allow for V2 to be re-analyzed as a suffix attached to V ( rather than being taken as base-generated in an independent head position. Consequently, suitable conditions for a re-analysis of V2 in a way which would automatically permit a head-initial re-alignment of AspP were indeed arguably quite naturally available in the language. I would therefore like to suggest that general pressure for uniformity in the direction of selection combined with frequent occurrence of linearly-adjacent V r V2 led speakers in the late-Tang and Song periods to re-analyze V2 -Asp as a suffix on Vi in the way outlined. These speakers may then have formally reinforced the re-analysis by themselves producing new RVC's with objects occurring in situ in post-V2 positions. In other words, speakers during the period of change can be suggested to have had frequent input of forms with objects either in pre-verbal position or phonetically null and from such data re-analyzed the V|-V2 A sp linear sequences as verb-suffix units occurring in V°. Following this, such speakers 355 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with a re-analyzed interpretation of V2-A sp as a suffix would have then deduced that if an object were to be both overt and in situ, then it should occur following the Vj-Asp unit. Such a hypothetical process of re-analysis would then eventually have led the new generations o f speakers to produce forms with overt objects following Vj-ViAsp units.1 9 Note that in such an analysis one does not in fact need to assume that any movement of V2 over the object to Vj ever actually occurred and the whole sequence of change is simply the result of different generations of speakers re-analyzing what they hear in rather different ways. Whereas it was noted that the 'movement' of V2 to Vi largely remains unexplained in other syntactic accounts of RVC's, here with the suggestions made above and following careful justification of each step of the argumentation it has finally been possible to reach a rather natural explanation of the critical re-positioning of V2. The suggested account therefore achieves the important joint goals set out earlier in section 5.1 o f explaining both the historical re positioning of V2 and the current syntactic status of V|-V2 RVC's, and doing this in a 1 9 It should be noted that (rather unsurprisingly) reference to certain resultative V2 elements as suffixes has in fact been made elsewhere, though largely without much justification. For example, Smith (1997) simply states that: 'Mandarin marks the distinction between telic and atelic verb constellations with verb suffixes known as Resultative Complements.’ (p.43). There is no further qualification of use o f the term suffix, nor any discussion of whether it is intended to refer to just phases or more literal V2's as well. In Starosta, Kuiper, Ng & Wu (1998) there is also suggestion that certain phase-type V2 elements may be considered to be suffixes on the grounds that they are not elements which encode literal predication on any NP (precisely as argued here). However, there is no proposal that any unifying property (such as telicity/aspect) might possibly link these elements together as a single functional group and so the conclusion is simply that any non-predicating V2 should by default be classed as a suffix. The critical question o f why literal V2 's might pattern syntactically in the same way as phase V2 ’ s is also not touched upon. Here the attempt has been made to develop a unified treatment of literal and phase V2 's both from a historic and a synchronic point of view and provide a principled account of precisely why the modem structures may exist as they do. For further general discussion of the possibility that V2 elements in the V,-V2 -Object sequences may be aspectual verbs, see Wang (1990) and Mei (1994). 356 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. way which links the two together in a principled and natural way. The analysis now developed has taken the three properties of RVC's identified as critical in section 5.2.1, notably (a) the lack of necessary predication between V2 and the NP 'object,' (b) the telic bounding property of V2, and (c) the historical re-positioning of V2, and suggested how these interact in important ways to result in the present day Vi-V2 resultative. It has been suggested that there were in fact two stages to the re-analysis process. First, increased focus on the completive function of V2 led to its re-analysis as instantiating a sentence-final Aspect head and allowed in non-predicative phase V2 elements. Later, V2 was re-analyzed as an aspectual suffix attached to V| in the lexicon and raised to Asp0 for licensing in the covert syntax, this being fully in line with recent Minimalist assumptions concerning morpho-syntax. In such an analysis with V 2A sp added to Vi in the lexicon and inserted together as a unit in V°, the V[-V2 pairs are essentially formed rather like complex- predicates, as a single verbal element combining the properties of two discrete morphemes, the main activity verb and the aspectual V 2 element. Consequently it is now fairly straightforward to suggest that any predicational properties which the V2A sp might have in addition to its critical aspectual function may be simply added to those o f the main V|. If there are lexical descriptive properties inherent in the V2A sp which apply to the object (as with 'literal' V2 S), then these can be applied to the object in a natural way from within V°. Given that there are possibly a range of restrictions on the ways that V 1-V2 pairs may be interpreted, I would now like to suggest that Yafei Li's (1990) rather successful lexical account of the way the 357 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. predicational structure of V[ and V2 is merged together may now be adopted and taken up into the present analysis which takes V2 to be primarily a functional aspectual element with possible additional lexical content/contribution. As noted earlier and stressed several times, there are numerous V2A sp elements which are purely functional and it was suggested that the V2A sp group is syntactically licensed in virtue of its functional contribution. Consequently V2A sp elements will not necessarily have to contribute any descriptive properties in \ W 2A sp pairs— any lexical/descriptive properties which are present may be interpreted subject to the set of constraints presented in Yafei Li (1990), but all of the V2A sp group will be formally licensed by the aspectual contribution.2 0 5.4 Related Consequences to Other Structures From a general point of view, it has been suggested that in certain circumstances structures might in fact naturally grammaticalize in a way which is ultimately in conflict with the general direction of selection in a language. It was also suggested that in such cases there will be pressure for these structures to be re-analyzed in a way which does conform with head-parameter setting in the language. As part of the 2 0 Note that Wang (1990) and Mei (1994) also suggest that wherever a V2 element from an earlier Vl-Obj-V2 form occurs re-positioned before the object in a new V,-V2 -Obj sequence, this V2 element is an aspectual morpheme. The present analysis attempts to provide a principled explanation for exactly how such a reanalysis may have taken place. In view of the fact that V2 elements could occur in V r Obj-V2 sequences without there being any predication relation between V2 and the object, such patterns strongly suggest that V2 elements grammaticalized as functional aspect heads in final position prior to re-positioning and suffixation to V). The latter re-positioning of V2 in the present analysis occurred as a direct result of the re-analysis of V2 as a functional category in final position. In this sense the surface reflex of re-analysis, i.e., suffixation o f V2 to Vf, consequently occurs after the underlying syntactic re-analysis. Such a sequencing of surface reflex following initial re-analysis 358 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. argumentation here, it was proposed that elements may become functionally determined as Specifiers only if there is a hierarchically lower complement present. In the absence of such a complement the 'Specifier' may instead become functionally determined as a complement itself. If this element then branches in a direction which does not coincide with the general direction of complement selection, the result will be that the structure constitutes an exception to the general headedness directionality in a language and will be under pressure to undergo re-analysis. Reflecting on this idea of functional determination of the Specifier relation, one might now reconsider certain other cases which could seem to be counter-examples to such a proposal. Although we noted that unergative verbs may be analyzed as having null cognate objects (or to arise via N-incorporation into V° and hence derivationally have a complement position present), one might wonder about the structures projected with relative clause de and the Taiwanese complementizer kong; in both such cases there is a large predicate-like element in the 'Specifier' of a functional head (D° and C° respectively) and so they might seem to be similar to the AspP case where the predicate VP occurs in 'Spec'-AspP. However, in relative clauses the D° head does indeed have a complement--the CP— and so the leftward IP is in fact legitimately licensed as occurring in a Specifier position. In the case of kong, the IP predicate in its Spec is raised to this position from regular rightward complement position, so again the Spec position is legitimately licensed (in virtue of the existence of the lower complement position). What the potential constraint might is suggested to be quite common in diachronic change (see Harris and Campbell 1995). 359 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. seem instead to class as exceptional and therefore under pressure to be re-analyzed is structures in Chinese in which a predicate element is base-generated in the 'Spec' of a functional head which has no other complement. Here one might think of sentential -le— if the predicate argument o f -le is base-generated to its left and -le has no other complement then the predicate will be determined as a left-branching complement rather than a Specifier. However, in footnote 14 it was pointed out that there may be prosodic stress-related evidence indicating that the predicate is actually raised to its surface position from an underlying position as a rightward complement, and so this case might also not be an exceptional left-branching structure. One last non-Chinese case I would like to consider briefly here is that of N° elements, and I believe that the hypothetical constraint on Specifier determination might possibly offer some insight into a hitherto unsolved puzzle. Specifically one might wonder whether it is possible for nouns to project a specifier external argument but not a complement. In a language such as English where Specifiers branch to the left and complements branch to the right this would result in an exceptional structure— the 'Specifier' would be determined as a left-branching complement. In this regard, Grimshaw (1990) suggests that only process nominals have a genuine argument structure with external 'subject' arguments and internal complements, and furthermore that the projection of this argument structure is largely optional. What is of particular interest and relevance here is the observation Grimshaw makes that the external subject argument of process nominals may only be projected if the internal complement argument also co-occurs and that whereas (121) 360 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is fine, (122) with 'John' being interpreted as the agent of the 'examination' is not. Note that the adjective 'continual' is used to ensure the process nominal interpretation: (121) John's continual examination of the patients impresses everyone. (122) (*)John's continual examination impresses everyone. (*) = ungrammtical if 'John' is Agent, only acceptable if 'John' is Patient This interesting fact about the interpretation of process nominals has arguably not been explained in any non-stipulative way since Grimshaw's original observations. However, given what has been suggested above it is actually predicted that (122) should be exceptional with 'John' being interpreted as the external argument of the noun. For 'John' to be interpreted as the external argument it would have to be base generated in Spec of NP (and then raised to Spec of DP for case). However, if the complement patient NP 'patients' is not projected, the result will be that the subject in NP Spec position will be determined as an exceptional leftward complement. This exceptionality may then explain why (122) is commonly felt to be bad. What is also of interest is that there is in fact a legitimate (although not immediately obvious) interpretation of (122). If'John' is interpreted not as the external agent but instead as the patient complement of the 'examination,' (122) is relatively acceptable (although perhaps not an immediately obvious interpretation). This is quite possibly a consequence of the 'Spec' position functionally being determined as a complement position in the absence of a true rightward branching complement~the NP base- 361 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. generated in this position actually does get interpreted as if it were the complement of the N°-head. Finally, another general aspect of the account has been to show how suffixation may come about in an SVO type language. Typologically suffixation is extremely common in head-final SOV-type languages and much less frequently found in SVO type languages (see e.g., Greenberg 1963, Dryer 1992). This imbalance is generally explained by the way suffixes arise naturally in SOV-type languages. Most commonly verbal suffixes are grammaticized functional morphemes which originated as independent verbs embedding a clausal complement (which they applied their particular function to). For example, causative suffixes often arise when an embedding verb encoding causation such as 'make/do' reduces and becomes dependent on the main lexical verb in its clausal complement. Such a grammaticalization process is both common and natural in SOV languages for the reason that the lexical verb o f the subordinate clause and the functional verb of the higher clause occur linearly adjacent to each other, as seen in (123) (using English words for simplicity; see Mallinson and Blake 1981 for details). (123) is intended to mean: 'John made Mary leave.': Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Subject Subject2 John Mary leave made Such adjacency between the lexical verb of the lower clause and the functional embedding verb o f the higher clause together with the phonological reduction frequently accompanying grammaticalization of a functional element naturally results in the Vi becoming phonologically dependent on the verb in VP2 and re analyzed as a suffix attached to the verb VP2 . With SVO-type languages there is no such adjacency between a matrix verb and the verb o f its complement clause21, as seen in (124): Subjecti Subject2 VP2 John made Mary leave 2 1 Unless of course the Subjecti comes to be positioned after V2 in a PP as in certain Romance languages in causatives: (i) Jean a fait manger les pommes a Marie. (French) Jean has made eat the apples to Marie 'Jean made Marie eat the apples.' (personal communication, Lina Choueiri) 363 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Consequently the conditions for natural re-analysis of a higher embedding functional type verb as a suffix attached to a lower lexical verb are absent from SVO type languages. The existence of limited suffixation in such languages therefore needs some explanation. Here a commonly adopted approach has been to suggest that suffixation in SVO languages must actually have arisen at a time when the language in question was SOV in its basic word order, and the existence of suffixation in SVO languages is even suggested to be evidence that such languages must have been SOV at some earlier point~see e.g., Givon (1971) on Bantu. However, while such a theory may be justified for west European languages with earlier SOV forms and their origins in Latin, it is very contentious for languages such as the Bantu group which otherwise show no real convincing signs of ever having been SOV. One clear result of the current chapter has been to show precisely how and why suffixation might in fact also arise in SVO type languages, and the analysis developed therefore now allows for potential insights into directionality and suffixation in other head- initial languages. 5.5 English Resultatives— a Speculation Finally before continuing on in next chapter to consider the nature of verbal -le and its relation to resultative sequences, I would like to add certain thoughts about the structure of resultatives such as (125) in English: (125) John ate the apple up. [V-Obj-Prt] 364 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. English resultatives appear to show a number o f similarities with Chinese RVC's. First of all there are resultatives in which the element following the object of main verb is an adjective which seems to predicate rather literally of the object in some way, as in (126), and there are also resultatives in which a particle occurs and there is no literal predication relation between this element and the object NP, as illustrated in (127). This split resembles the distinction noted between literal and phase/non literal V2 elements in Chinese RVC's: (126) a. John washed the plates clean. The plates are now clean, b. Mary painted the door purple. The door is now purple. (127) [V-Obj-Prt]: a. John ate the apple up. *The apple is now up. b. John looked the article over. *The article is now over. c. He thought the problem through. T h e problem is now through. d. He dried the socks out. *The socks are out. Secondly it has been noted that there are two distinct orders possible with verb- particle resultatives; alongside the [V Object Prt] order in (127) it is possible to use a [V Prt Object] order as in (128): (128) [V-Prt-Obj]: a. John ate up the apple. b. John looked over the article. c. He thought through the problem. d. He dried out the socks. 365 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. These two orders possible with V-Prt pairings seem to resemble the past and present orderings o f Vi and V2 in Chinese RVC's, with the [V Object Prt] appearing like the original [V| Object V2] Chinese type and the [V Prt Object] alternate possibly resembling current Chinese [V 1-V2 Object] forms, as schematized in (129) and (130). (129) a. Old Chinese V, OB V2 b. English (i) V OB Prt (130) a. Mod. Chinese V, V2 OB b. English (ii) V Prt OB It should be noted that this second [V Prt Object] ordering is perfectly natural with verb-particle pairs, but often rather odd and unacceptable with verb + literal adjective sequences: (131) ??He painted purple the door. (132) ??/*He rode tired the horse. For this reason I will here restrict my attention to verb + particle pairs, which seem to show more of the patterning found in Chinese RVC's. Such English verb-particle sequences have indeed been studied by numerous investigators, e.g., Curme (1914), Kennedy (1920), Visser (1963), Kayne (1984), Johnson (1991), Hoekstra (1992), den Dikken (1995) and Collins and Thrainsson (1996) among many others, and a variety of different analyses have been proposed (which I will not review here for obvious reasons of space). Now that the present chapter has developed an analysis of Chinese RVC's, and given the apparent similarities with English verb-particle 366 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. resultatives, it is natural to wonder whether the account of Chinese might allow for some further new insight into English verb-particle sequences. Supposing that English verb-particle forms were indeed to be syntactically equivalent to Chinese RVC's, then the particle would essentially be the equivalent of the Chinese V2A sp - The [V Object Prt] sequences might be analyzed as having an underlying structure in which the particle instantiates a completive aspect head Asp0 with the [verb + object] as a VP projected in Spec o f AspP. Other [V Prt Object] forms if directly paralleling the analysis of Chinese might then have resulted from re analysis o f the particle as an aspectual suffix attached to the main verb and licensed by a higher AspP via LF raising, precisely as outlined for Chinese. Such an analysis cannot however be correct English, I believe, for two simple reasons. First of all, supposing the particle were to be interpreted as an aspectual suffix in [V Prt Obj] forms, this would result in a morphological sequence in (133): (133) [ verb-stem + tense-suffrx + aspect-suffix] look- -ed over (the article) dry- -ed out (the socks) Such a sequencing would seem to go against the cross-linguistic patterning commonly observed that aspectual morphemes occur closer to the verb stem than tense morphemes. This cross-linguistic patterning corresponds to the assumption that tense projections are structurally higher than aspect projections and that the highest functional projections relate to affixes which are farthest away from the verb stem (essentially Baker's 1985 Mirror Principle). If Tense is universally ordered 367 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. higher than Aspect in syntactic structure and if the Mirror Principle (or a Minimalist equivalent) is a universally valid constraint then one should always find the affix order in (134): (134) expected order following the Mirror Principle *look-over-ed = V-Asp-Tns *dry-out-ed = V-Asp-Tns Secondly, conversion of [Vi Object V2 A sp] forms into [Vi-V2 A sp Object] sequences in Chinese is suggested to have resulted from Vi V2 adjacency arising naturally through frequent topicalization of an old/definite object and the occurrence of pro in object position. As English does not obviously have pro in regular object position nor topicalization to the same frequent extent as in Chinese, it might seem that this natural route o f change would not be open to English verb-particle structures. If this is so, it is natural to ask where the English alternation comes from and how it might relate to Chinese RVC's, as there are clear surface similarities. Here I suggest that a potential solution is to be found in considering an additional piece of extremely interesting and revealing historical information which is curiously ignored in most recent analyses of verb-particle constructions. Abundant evidence from Old English recorded in Curme (1914), Kennedy (1920) and Visser (1963-1973) indicates that there was in fact much earlier a third alternative ordering of verb, particle and object, in which the particle actually precedes the verb as in (135): 368 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (135) English "type (iii)" resultatives: particle - verb - object In fact a fourth variant o f this was also apparently possible, with the object preceding both the particle and the verb, as in (136): (136) Type (iii) variant object - particle - verb Thus there were equivalents to forms such as: (137) a. X out dried the-clothes b. X the-clothes out dried (138) a. X up ate the-food b. X the-food up ate Certain researchers such as Curme (1914) and Kennedy (1920) suggest that such forms were in fact more common than equivalents in which the object and the particle follow the verb (127) and (128). They also argue that these forms with the particle critically preceding the verb were earlier than the forms with the particle following the verb. Such an early order in which the particle occurs before the verb is interesting and revealing. If the particle occurs to the left of the verb it could be suggested that the particle is actually in some functional head higher than the VP, hence quite naturally in a (completive) Aspect head. Such a possibility is indeed quite likely to be correct for verb-particle resultatives which occur in the Kartvelian languages as 369 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. reported by Harris and Campbell (1995). There it is noted that in Svan there are particles which occur either following the verb as in (139), or alternatively preceding it as in (140). Interestingly when the particles precede the verb they are clearly not prefixed or cliticized to the verb for two reasons. First of all an adverb such as ud 'again' may intervene between the particle and the verb as in (141): (139) acad sga s/he.go inside 'S/he went inside.' (140) sga acad in s/he.go 'S/he went inside.' (141) sga ud etqarix yerbats in again they.implore God.DAT '. .. again they implore God.' Secondly, the particles may occur standing alone and unsupported as (affirmative) answers to yes/no questions, as in (142) with the particle ka 'out': (142) ka acad ma? — ► answer: ka out s/he.go Q out 'Did s/he go out?' 'Yes.' This suggests that such elements are neither raised from some lower position and cliticized to the front of the verb, nor base-generated as verbal prefixes, as if they were clitics/prefixes they should not be able to stand alone without the verb or be 370 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. separated from the verb by adverbs. The patterning rather suggests that such elements are instead free standing heads occurring base-generated in some functional head dominating the VP, hence naturally in an Aspect head. Furthermore, it is a common characteristic across a wide range of languages (including Chinese, Thai, Vietnamese, etc.) that yes/no questions are answered in the affirmative with the highest verbal element in a sentence (hence a modal if present, otherwise the main clause lexical verb, see Simpson (forthcoming)). This might again seem to indicate that the particles in Svan and other Kartvelian languages are indeed independent heads in a verb-related functional projection dominating VP - as the hierarchically highest 'verbal' elements present they are used as answer-forms instead of the main verb which is lower down in the structure. Returning back to English, if it is true that forms with the particle preceding the verb are indeed older and were more frequent than present-day forms in which the particle (and the object) follows the verb, then it is reasonable to suggest that the present-day forms may well have derived from the earlier particle-verb sequences. (143b) and (143c) should then be both derived from (137a) in some way: (143) a. (Obj) Prt V (Obj) = (the-clothes) out dried (the-clothes) b. V Obj Prt = dried the-clothes out c. V Prt Obj = dried out the-clothes That is, type (iii) forms schematized in (144) historically developed into the present type (i) and type (ii). 371 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (144) AspP VP out V Ob I I dry the-clothes Various possibilities suggest themselves. Certainly it would seem that one needs to allow for object-shift to play a role (as in fact assumed in Johnson (1991), as it is clear that the object may occur in two independent positions in (143a). In addition to this one might perhaps then make the assumption that the particle is indeed in Asp0, and not just in (143a) but in (143b) and (143c) as well. In (143a) both the particle and the verb would be in their base-generated positions in Asp0 and V° respectively. (143c) would then be derived from an underlying sequence [Prt V Obj] by raising the verb to some higher position above completive Asp as in (145). (143b) would result from similar raising of the verb combined with object-shift to Spec-AspP, as in (146): (145) V| [a sP Prt [vph Obj]] = driedj out tj the-clothes t I (146) Vj [ a s P p Objk [ a s P Prt [vptj tk ]]] = driedj the-cIothesk out tt tk t T i However, if the particle is in fact in Asp0 , it might be objected here that movement of the verb to a higher head-position should possibly violate the Head Movement Constraint (HMC), being head-movement from a lower V° over a filled Asp0 to some higher X°. An alternative might therefore be to propose that the particle is actually an aspectual-prefix attached to the verb in the lexicon and inserted into V°. In a way 372 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. similar to Johnson (1991) it can then be suggested that when the verb raises it strands the aspectual particle in V° and then moves on higher. This would not result in any HMC violation and would essentially be a treatment of particle-stranding similar to that commonly adopted for particle-verb separation in German V2 clauses. As shown in the following, whereas the German particle and verb occur together in V° in (147), when the verb raises to C° in V2 clauses, it strands the particle in V° (or possibly some other higher head), as in (148): (147) [cp Johann [c will [IP die Tur aufmachen ]]] Johann wants the door open-make 'Johann wants to open the door.' (148) [ c p Johann [ c machti [ip die Tur auf-tj ]]] Johann makes the door open 'Johann opens the door.' In such an approach to English, the particle as a stranded aspectual prefix would then undergo raising to Asp at LF for licensing/feature-checking. A further variant of such an approach would be to suggest that the verb + particle prefix first raise together to Asp0 and the particle is stranded there when the verb moves to a higher head. Either way, understanding the particle as an aspectual prefix added in the lexicon in a way similar to the analysis of Chinese allows for a plausible account of resultative verb-particle structures which has historical support and is able to make good sense of the third type of patterning noted where the particle precedes the verb. 373 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This third patterning is curiously ignored in recent treatments of particle-verb forms and has led to an emphasis on analyses in which the particle is taken to predicate o f the object, e.g., Hoekstra (1992), Collins and Thrainsson (1996). As noted in (127) there are clear problems with such predication-based analyses, problems which an alternative aspect-based approach is able to avoid. Furthermore, if historically later [V Object Prt] sequences such as [dry the-clothes out] are actually derived via verb- movement and object-shift from earlier structures in which the particle was a verbal prefix, then syntactically there is no predication relation between the particle and the NP 'the clothes'— the NP in such sequences is in a derived hence non-theta position and is consequently not the subject of any (small) clause type predication. The 'predication' structure seen in the linear sequence of NP followed by particle in [ dry [n p the clothes] out ] is actually a derivational illusion resulting from a combination o f raising operations and is not a primitive theta relation. Finally, the conclusion that particle elements in both English and the Kartvelian languages (among others) are either verbal affixes or base-generated in a pre-verbal Asp-head would seem to add support to suggestions here that similar resultative V2 elements in Chinese are also affixal instantiations of completive Asp. Further careful historical research is certainly now required, but generally the analysis of resultative elements as being 2 2 The aspect-based approach proposed here is most similar to the account argued for in Johnson (1991). However, Johnson treats the particles as unclassified lexical suffixes on the verb and makes no connection to aspectual characteristics. Treatment o f the particles as suffixes in English can also be suggested to run into a number o f difficulties concerning the generation and licensing of the tense suffix which appears attached inside the particle (see Collins and Thrainsson 1996 for relevant criticism here). 374 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. potentially re-analyzed instantiations of Aspect would seem to open up a number of interesting cross-linguistic possibilities. 5.6 Summary Before continuing on to examine the diachronic and synchronic patterning of verbal -le and how it may or may not relate to the changes which have taken place in RVC's. I now provide a short summary of the main conclusions of this chapter. The chapter began as an investigation of the apparent diachronic re positioning of V2 elements to a position right-adjacent to the main verb Vj and how an understanding of this historical change might shed better light on the current syntactic status of V 1-V2 resultative sequences. Reviewing first previous syntactic approaches to V 1-V2 composition, I pointed out a number of unsatisfactory aspects in analyses which assume that V2 elements must synchronically undergo raising to the Vi and instead attempted to develop an alternative approach, critically linking the diachronic word order change to an increase in focus on the telic contribution of V2 and the lack of a necessary predication relation existing between V2 and the object of Vj. Suggesting that V2 elements became re-analyzed as instantiating a sentence-final aspectual head, this allowed for an explanation of why it was that phase-type V2 elements could occur licensed in such positions with no necessary predicational relation to the object of Vi. Subsequently considering the actual re-ordering of V2 elements right-adjacent to the Vi, it was argued that this actually did not result from any movement of V2 but instead occurred as an instance of re-structuring when both 375 permission of the copyright owner. Further reproduction prohibited without permission. the fronting o f objects and the occurrence of objects in null pronominal form would have given rise to natural instances of V| V2 adjacency. In such re-structuring it was suggested that the V2 aspectual head became re-analyzed as an aspectual suffix taken to be combined with Vj in the lexicon. Trying then to make sense of why such a re analysis might have actually taken place, it was ultimately argued that re-structuring of this type would have allowed for a non-canonical head-final structure to be naturally re-anaiyzed as head-initial, and hence that directionality can be suggested to have been the critical force behind the second stage in the re-analysis and the physical 're-positioning' of V2 adjacent to V|. Finally, it was noted that the occurrence of suffixation in head-initial languages is typologically much less common than in head-final languages for certain fairly simple reasons relating to the linear ordering o f elements in SOV-type languages. A significant general conclusion of the chapter has therefore also been to provide a principled explanation of just how and why less frequent instances of suffixation in head-initial languages might in fact possibly develop. permission of the copyright owner. Further reproduction prohibited without permission. 6. Verbal — Lei Aspect and Tense 6.0 Introduction In this chapter I turn to verbal -le and consider how the re-positioning of -le/liao relative to the main verb and its object relates to parallel changes observed in resultative constructions. Suggesting first that -le/liao historically underwent structural re-analysis as completive aspect similar to other V 2 elements in RVCs, I then concentrate on the synchronic status of verbal -le and argue that a particular process of upwards grammaticalization in the functional structure dominating VP has resulted in verbal -le currently being a morpheme which may actually instantiate three discrete functional heads— completive aspect, perfective aspect and also more controversially past tense. The chapter attempts to show that Smith's (1997) binary theory of situational and viewpoint aspect provides an insightful model for understanding the roles played by different functional suffixes on the verb, and argues for the possibility that functional morphemes undergoing change may naturally instantiate more than just a single functional head in any extended functional sequence. In general, the central conclusion of the chapter that affixal elements undergo grammaticalization and re-analysis in a way which significantly parallels the movement-dependent re-analysis of free-standing morphemes is shown to provide strong evidence in favor o f the Minimalist hypothesis that affixes are licensed via raising to a higher functional head, and that this raising may take place either overtly or covertly at the level of LF. 377 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1 The Re-positioning o f Le/Liao Historically it is well-documented that modem day verbal -le in fact originated as the full verb liao meaning 'to finish' in a sentence-final position and then later re positioned itself right-adjacent to the main descriptive verb. Where the main verb is transitive with an object NP, this re-positioning is particularly obvious, as illustrated in (1): (1) a. V Object Iiao/le — > b. V liao/le Object The change from (la) to (lb) is reported to have taken place primarily during the Song dynasty (960-1279) (see for example Mei 1981, Z. Shi 1988, and G. Wu 1999). The obvious question, as with V2 re-positioning in RVCs, is why such re-positioning should have taken place. One possibility which might naturally be entertained is that the change occurred for phonological reasons. It is clear that modem day verbal -le is phonologically much reduced from its earlier full form liao and is now fully dependent on the verb to its left. It might therefore be suggested that liao first reduced to a clitic-form le and then raised to the verb to encliticize to it as a suitable host element. However, such a possibility can in fact be rather quickly dismissed. In a recent very informative paper G. Wu (1999) shows that there is good evidence that liao first re-positioned itself right-adjacent to the verb and only much later underwent reduction to le. Among other evidence for this conclusion is a particularly telling argument from Korean. In the Korean textbook of Mandarin 378 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chinese called the Chunggan Nogoltae written in 1795, the element corresponding to modem day verbal le is transcribed as having a pronounciation equivalent to liao not le. Verbal le then still had its original shape liao at least until the end of the 18th century, five hundred years after it underwent its positional change. Consequently such re-positioning cannot be ascribed to any phonological clitic-like properties of le triggering movement of le to the verb. Fully in line with suggestions made in a number of works (e.g., Mei 1981, Z. Shi 1988, G. Wu 1999, Sybesma 1999), I would instead like to assume that the re positioning of liao/le was actually just one instance of the general phenomenon of V2 re-positioning which occurred in resultative constructions. It has been frequently noted that the change in liao's position essentially coincided with the re-positioning of V2 elements in RVCs. The latter is suggested to have occurred around the time of the Song dynasty (probably preceded by its beginning emergence in the Tang dynasty), which is precisely when liao also underwent its positional change. G. Wu (1999) writes that: The "V(0)+phase complement" structure emerged during the Wei, Jin and the Northern and Southern Dynasties, and changed into the "V+phase complement (O)" structure around Song, (p.22) It is therefore commonly suggested that liao re-positioning may be thought of having been strongly influenced by V2 re-positioning in RVCs and might even be considered to be a further occurrence of the general change in RVC word order. G. Wu (1999) continues: 379 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. . . . the grammaticalization of liao coincides with the development of the "verb resultative complement" structure and the formation of resultative verb compounds in the language. In other words, the grammaticalization o f liao is actually a part of the process, (p.23) The original element liao did indeed also have the meaning o f a typical resultative phase-type V2 'to finish' and can therefore be taken to have signaled simple completion much in the same way as other V2 elements such as wan 'to finish' do. I would therefore now like to simply assume that liao originally occurred in the sentence-final completive Asp0 and then later underwent 're-positioning' to be right- adjacent to the verb for the same reasons suggested to underlie the re-positioning of RVC V2 elements in general, i.e., liao became re-analyzed as an aspectual suffix on the V 1 in order that the head-final AspP could be re-interpreted as being head-initial instead. Only much later on liao would then have phonologically reduced to its present pronunciation as le. This much concerning verbal -le is not so new in terms of the basic underlying conclusion that liao/le was originally just one of the general phase-V2 group in RVCs. The mechanism which led to its re-positioning is also taken to be essentially the same as that which occurred with other V 2f s. What is now of greater interest with verbal le is the possibility that le has in fact undergone further significant development from being just a simple completive aspect marker to encode other higher functions as well. I will therefore spend the rest of this chapter considering what kind o f formal syntactic structures might correspond to the interpretations instantiated by modem day verbal le and how diachronic development 380 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. might have also given rise to such structures, i.e., what mechanisms might lead to the changes observed. 6.1.1 The Current Status o f Verbal L e_ and Completive Aspect Given the fact that liao had the typical completive-type meaning of other V2 elements and that it underwent re-positioning at the same basic period as other V2's it would seem to be fairly natural and straightforward to treat it as a simple V2 as suggested here and also by various other investigators. In the present approach proposed in chapter 5, this consequently means assuming that it was first generated in the sentence-final completive AspO and then together with the other V2's re analyzed as a verbal suffix licensed by a higher 're-aligned' completive AspP at LF, as demonstrated in (2): (2) TP subject, T Asp0 VP A NP V’ t, V NP Vi-Asp(V2) Object 381 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Now, although it can be maintained that such an analysis has good motivation as a diachronic account o f the origin and re-positioning o f liao/le, there is also evidence indicating that synchronically verbal le in fact may not have exactly the same status as other members of the V2 set and that le may therefore have undergone some further syntactic change since its initial re-analysis. One rather simple but strong piece of evidence that le is currently different from other V 2 aspect suffixes is that le may occur as a suffix in addition to a second regular V2 suffix of either 'literal' or phase type, as shown in (3) and (4) below. Such examples contrast with (5) and (6) where le occurs just with a bare Vj: (3) ta xi-ganjing-le suoyoude yifu he wash-clean-LE all clothes 'He washed all the clothes clean.' (4) wo yijing kan-wan-le san-ben-shu le1 I already look-finish-LE 3-CL-book LE 'I already finished reading 3 books.' (5) wo yijing chi-le fan-le I already eat-LE rice-LE 'I've already eaten.' (6) ta zuotian qu-le Beijing he yesterday go-LE Beijing 'He went to Beijing yesterday.' 1 I do not attempt to provide any account o f sentence-final -le here. It is included in many of the examples simply for naturalness. 382 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In (3) and (4) it is clear that ganjing 'be-clean' and wan 'be-finished' are in the regular V2 position signaling completion of the action and can be therefore taken to instantiate completive aspect, raising to and being licensed by Asp0 at LF. This being so, the le which occurs outside the V 2A sp must be taken to encode some other function here. A second reason to believe that le is at least in some o f its occurrences not licensed as completive aspect is that le is also frequently found with achievement predicates which represent instantaneous events with no extension over time. Typical V 2 completive verbs/suffixes occur with V[ elements which represent activities which do occur over a period of time; the V ^A sp suffix marks the end-point of the activity and signals completion o f the action. In contrast to activities (such as 'run,' 'wash,' 'read,' etc.) or accomplishments (such as 'draw a circle,' 'walk to school,' etc.), achievement predicates such as 'arrive,' 'recognize someone' and 'die' are commonly interpreted as referring to events which are instantaneous and do not take any time. Consequently such predicates are normally quite unnatural with expressions of completion which require that the predicate express an action which naturally extends over a period o f time before it is indeed completed. Examples in (7)-(9) below have achievement predicates whereas those in (10)-(11) have an activity and an accomplishment predicate:2 2 If a hearer can interpret predicates such as 'recognize X,' and 'die' in a non-canonical way and create a special context in which they can be understood as requiring certain time, then sometimes such predicates can be acceptable. Native speakers indicate that this is particularly difficult with 'arrive' and only marginally possible with 'recognise' in a game show type context. 383 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (7) */??John finished recognizing Mary. (8) */??Mary finished arriving. (9) */??Bill finished dying. (10) John finished reading. (11) John finished drawing the circle. In Chinese one finds that le is highly natural with all of the achievement class and no special context is required for use of le with achievement verbs, i.e., they are still interpreted as representing instantaneous actions/events. This indicates that le in such instances cannot be expressing completion but some other function. (12) and (13) are examples with clear achievement predicates: (12) ta ba-dian-shi-wu-fen jiu dao-le huo-che-zhan he 8-hour-15-minute then arrive-LE train-station 'He arrived at the train station at 8.15.' (13) shou-shang-de ren dang-zhong, huran si-le liang-ge-ren receive-injury-DE people among suddenly die-LE 2-CL-people 'Among the injured, suddenly another two people died.' It should be noted that such predicates are completely unacceptable with other general V2 A 5 p completive suffixes such as wan 'finish' confirming that they are interpreted as instantaneous actions and that they are consequently incompatible with a V2 which expresses (general) completion: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (14) *ta zuotian si-wan-le he yesterday die-finish-LE literally: ??'He has finished dying yesterday.' (15) *tamen xianzai dao-wan-le huo-che-zhan they now arrive-finish-LE train-station literally: *'They have now finished arriving at the train station.' It can therefore be concluded that le may express some function which is clearly distinct from simple completion. The question to be answered now of course is what exactly is this function of lei In order to approach the issue I will now present a brief overview of Smith's two-tiered theory of aspect and show how it provides a rather natural answer to the question of le's role in examples such as (3)/(4) and ( 12)/( 13). Anticipating the results somewhat, it will be suggested that le here encodes perfective aspect as is in fact suggested in various other descriptive approaches such as Li & Thompson (1981), though the notion of perfective aspect which will be adopted here is somewhat different from that assumed in certain typological works such as Bybee et al (1994). Syntactically it will be suggested that this perfective aspect is structurally distinct from completive aspect, with both types of aspect being represented by discrete functional heads/projections. Incorporating certain insights in Sybesma (1999) and others it will subsequently be argued that le is in fact able to encode both completive and perfective aspect with a number of verbs and that it is the natural grammaticalization of completive le with these verbs to a structurally higher aspect head as a result of LF movement which allows for its current generalized use as a perfective. 385 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1.2 Smith (1997): Two Different Types of Aspect Smith (1997) convincingly argues that the commonly used term 'aspect' in fact refers to two quite different types o f properties, which she then describes as 'situational aspect' and 'viewpoint aspect.' Situational aspect is suggested to be an inherent property of predicates and refers to the basic types of situations represented by a predicate— the Aktionsart o f a predicate in more traditional terminology. As proposed in Vendler (1957) and frequently assumed in other works, situations described by predicates may essentially be classified according to three parameters: (a) whether they are aspectually bounded/telic or not, (b) whether they are instantaneous or have extension over time, and (c) whether they are dynamic or not. The cross-classification of such properties is commonly taken to result four basic situation types, as in (16): (16) Situation Types Situation type example properties stative be happy -dynamic, -instantaneous, -telic activity run, talk +dynamic, -instantaneous, -telic accomplishment draw a circle, run to X +dynamic, -instantaneous, +telic achievement arrive, recognize X +dynamic, +instantaneous, +telic Viewpoint aspect, by way of contrast, is suggested to encode how one views any of the above situation types on a particular occasion. Following earlier work such as Comrie (1976) it is noted that it is possible for a single basic situation type to be linguistically presented in different ways, this corresponding to different perspectives 386 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. or 'viewpoints' on the event/situation. If one assumes that events and situations potentially may have initial points, end points and internal stages as represented in (17) with a sequence o f points, viewpoint aspect is suggested to allow for focusing on these properties in two common ways:3 (17) Potential Event Type Initial Point.............. (internal stages)...............End Point Imperfective viewpoint aspect is suggested to focus just the internal parts/stages of a situation, and does not include either any initial or final endpoint in its focus/presentation. In English, the be...-ing form is the standard imperfective aspect; this focuses attention on the internal progression of the situation rather than any initial point or endpoint it may have. Note that viewpoint aspect is a type of aspect which is distinct from the classification of an event in terms of its situation aspect. The imperfective may combine with either an unbounded activity situation type as in (18a) or a bounded accomplishment as in (18b). However, because the imperfective form by definition focuses the internal stages o f a situation, it cannot normally be combined with achievement situations which are +instantaneous and have no internal stages, as seen in (17c). Also, in English, the be...-ing imperfective does not combine with stative situations because the be...-ing imperfective requires a situation which is +dynamic: 3 Smith (1997) also argues for a third type of viewpoint aspect which is referred to as 'neutral' viewpoint aspect. As the existence o f such a third basic viewpoint type is somewhat open to question, I here concentrate on the more traditional binary contrast o f perfective and imperfective (viewpoint) aspects. 387 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (18) Imperfective viewpoint aspect a. John is/was running. imperfective + activity imperfective + accomplishment imperfective + achievement imperfective + stative b. Mary is/was drawing a circle. c. *John was recognizing Mary. d. *Mary is/was being pretty. Note also that both the imperfective viewpoint aspect and the situation aspect of any event/situation are properties which are independent of the tense property associated with that event/situation, and that tense simply functions to locate an event/situation in time. Because o f this independence, the tense associated with an event/situation may clearly be varied while maintaining the viewpoint aspect constant as imperfective, as seen above in (18a/b). In contrast to imperfective viewpoint aspect, perfective viewpoint aspect is argued to focus the whole of a situation and therefore critically include both the initial and the final endpoint of a situation in this focus. In many languages such perfective viewpoint aspect would actually not seem to be expressed by any distinct overt morpheme on the verb; in English. For example, there is no identifiable perfective counterpart to the imperfective be...-ing form; rather, it may be suggested that perfective has zero morphological expression and that a verb may come to be interpreted as perfective when it is not marked positively for imperfective viewpoint. Concretely, when a simple past tense form of a verb in -ed occurs, it expresses perfective viewpoint in addition to past time location as it importantly contrasts with the combination o f past and imperfective which results in was V-ing: 388 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (19) past + perfective + situation/verb -ed + 0 + walk to the park — > (John) walked to the park (20) past + imperfective + situation/verb past + be...-ing + walk to the park — > (John) was walking to the park. As is the case with imperfective viewpoint aspect, perfective aspect is in essence independent of the situation aspect of a predicate and may therefore be combined with events encoding different situation aspects. For example, the following are all taken to encode perfective rather than imperfective aspect (largely due to the contrastive absence of be...-ing): (21) John walked to the park. perfective + accomplishment (22) John arrived at the park. perfective + achievement (23) John walked in the park. perfective + activity Consequently it may be argued that there are (at least) three distinct formal properties potentially associated with any predicate— (a) its situation type/aspect, (b) the viewpoint aspect used to present that situation, and (c) its temporal location, i.e., the tense used to locate the situation relative to other events/situations. These three properties might also seem to occur in a natural hierarchical ordering. Compositionally, a situation/event will first be identified and defined in terms o f its situation aspect; then the situation will be presented in a particular way, either as a whole with perfective aspect, or with a focus on its internal structure with imperfective aspect. Finally the situation/event so interpreted may be located in time 389 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. via the use of tense. Such a natural compositional hierarchy among the three properties is also found to be reflected in the linear ordering o f tense and aspect elements in many languages. In head-initial languages where tense and aspect markers are instantiated by independent free-standing morphemes rather than by suffixes, it is commonly noted that tense morphemes precede im/perfective-type morphemes, which in turn precede morphemes relating to situation-type aspect (see Cinque 1999 for a wide range o f data and relevant discussion). If one assumes that such unbound morphemes are located in (functional) head positions, in head-initial languages this leads to the simple conclusion that tense/TP hierarchically dominates perfective aspect which in turn dominates situation aspect, as schematically illustrated in (24): (24) AspP I =im/pcrfective Aspi AspP 2=situation A structure such as (24) also naturally encodes the selection relation between viewpoint aspect and situation aspect. For example, above in (18) it was noted that the English imperfective viewpoint aspect be...-'mg combines only with predicates 390 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. which have the aspectual situation type of either activity or accomplishment. (Viewpoint) Aspi° can therefore be taken to select for the situation type represented by (situation) AspP2 - The main semantic difference between the viewpoint types argued for by Smith concerns how much of a situation that the viewpoints 'make visible' in any particular instance. As noted, perfectives focus a situation in its entirety and so include the initial and final endpoints o f the situation, whereas imperfectives only focus on the internal stages of an event and so exclude the endpoints from the focus. Smith points out that such a difference in the presentation o f an event can be clearly observed when one attempts to qualify the description of im/perfective events with denials and continuative statements. For example, it is possible to combine imperfective aspect with a telic sitiuation (and past tense) to describe a certain situation and then deny the completion o f that situation, as in (25): (25) Mary was walking to school, but she didn't actually get there. This contrasts with the combination o f perfective aspect with a telic situation and past tense to describe the same essential situation. Such a combination does not allow for any denial of the completion of the event: (26) *Mary walked to school, but she didn't actually get there. The contrast in (25) and (26) can be given the following explanation. In (26) the use of a perfective viewpoint (i.e., lack of be...-ing form) makes both the initial and final endpoints of the event (properties of its situation aspect) naturally 'visible' in the 391 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. description. When past tense is applied to a telic event whose endpoints are visible and focused by the perfective viewpoint, this consequently results in an interpretation in which both the initial and the final endpoint are interpreted as having been realized. Because the final endpoint of having arrived at the school is therefore understood as having occurred, it is impossible to follow this with a denial that it occurred. In (25) however, due to presentation of the same telic situation with imperfective viewpoint aspect, there is only focus on the internal stages of the event, and while it can be deduced that the initial point must have occurred in order for there to be these internal stages, there is no necessity that the final endpoint also be realized. Semantically the endpoint is simply not made visible in the presentation of a telic situation with an imperfective viewpoint, and consequently it is possible to deny that such an endpoint actually is realized. It should be carefully noted that the relevant factor here is not tense, as (25) and (26) are both given past tense (nor is it situation aspect, as the situation is identical and telic in both cases); it is critically just the viewpoint choice which is responsible for the difference in acceptability. Such simple contrasts again illustrate the conclusion that tense, viewpoint aspect and situation aspect are indeed three essentially independent properties. The same basic patterning is seen again with attempted statements of continuation. (27) John was writing the letter an hour ago, and in fact he still is writing it. 392 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (28) * John wrote the letter an hour ago, and in fact he still is writing it. In (27) use of the imperfective does not make the endpoint visible and so it can be asserted that the endpoint effectively has not occurred. In (28) however, the perfective viewpoint does make the endpoint visible and so when combined with past tense results in an interpretation that the letter-writing was completed and therefore cannot be still continuing. Given that examples such as (26) and (28) with past tense and perfective viewpoint applying to a telic predicate cannot be followed by any denials of completion, and given also that past tense applied to the same telic situation presented with imperfective aspect does allow for a denial of completion, one can assert the following conclusions, in line with Smith (1997): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (29) a. realized completion is part of the genuine meaning of past tense applied to perfective aspect presenting a telic situation and is not simply a pragmatic inference. b. perfective viewpoint is critical in effecting the meaning of realized completion, because past tense and a telic predicate otherwise with imperfective aspect do not result in any necessary interpretation of completion. c. perfective viewpoint does indeed 'exist' in English although not obviously instantiated here by an overt morpheme distinct from past tense. d. it is necessary to recognize three independent properties which interact with each other in different combinations— tense, viewpoint aspect and situation aspect. Finally it can be noted that Smith (1997) argues strongly that both types of viewpoint and situation aspect are in fact always present in some form in the linguistic representation of events. It is suggested that the lower situation aspect properties of a predicate can only be semantically interpreted if some kind of viewpoint aspect is applied to make them visible and that this requirement results in the occurrence of viewpoint aspect being necessarily encoded in some way in the descriptions of all events. 394 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1.3 Verbal -Le and Perfectivity The above discussion of Smith's well-reasoned two-tiered theory of aspect now allows for a clearer understanding of the status of verbal -le in Vi-V2 -le forms such as (3) and (4) repeated below: (3) ta xi-ganjing-le suoyoude yifu he wash-clean-LE all clothes 'He washed all the clothes clean.' (4) wo yijing kan-wan-le san-ben-shu le I already look-finish-LE 3-CL-book LE 'I already finished reading 3 books.' It was noted that because there is already a V2 A sp suffix instantiating completive aspect on the main verb in addition to verbal -le, then verbal -le in such instances cannot itself be taken to be completive aspect. Verbal -le was also seen to occur with achievement verbs with no extension over time and no process to complete, which again disallows an analysis -le as completive aspect with such verbs. The question therefore naturally arose as to what kind of functional head the verbal -le suffix might correspond to in Vi-V2-le forms and with achievement verbs. If one now adopts Smith's two-tiered model of aspect, a simple answer to this question suggests itself, and verbal -le can be assumed to instantiate perfective viewpoint aspect here. It can be suggested that corresponding to Smith's viewpoint and situation aspect syntactically there are indeed two independent functional projections encoding aspect precisely as illustrated in the tree in (24), and that 395 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. whereas Chinese V2-completive suffixes are licensed by the lower A s p 2-Situaiion head, le in Vi-V2-le forms will be licensed as a perfective suffix by the higher Aspi -viewpoint head. Such a general assumption not only coincides with traditional views that verbal -le is a perfective marker, as for example expressed in Li & Thompson (1981) and many other works, but there is also simple evidence from certain co-occurrence restrictions which indicates that -le is here functioning as an instantiation of perfective aspect. Specifically, the element zhengzai is rather clearly a marker of (progressive) imperfective aspect and therefore arguably occurs in the higher Aspi. v ie w p o in t head. Significantly it is not possible for zhengzai to co-occur with a Vi-V 2-le sequence even though zhengzai may appear with V 1-V2 forms, as seen in (30), (31) and (32): (30) wo zhengzai ca boli ne I PROG wipe glass NE4 'I am just wiping/cleaning the glass (at the moment).' (31) wo zhengzai ca-gan boli ne (Sybesma 1999) I PROG wipe-dry glass NE 'I am just wiping dry the glass.' (32) *wo zhengzai ca-gan-le boli ne I PROG wipe-dry-LE glass NE In this case, verbal -le and zhengzai are thus in complementary distribution. Assuming zhengzai to be in Asp 1.v iew p o in t, this co-occurrence restriction is naturally 4 The element ne is a sentence-final emphatic particle which frequently co-occurs with (zheng-)zai. I do not offer an account o f its syntax here. 396 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. explained if -le instantiates perfective aspect and needs to be licensed by the same A spi-view point head; the clash of im/perfective values in zhengzai and -le will simply result in -le failing to be licensed here. Note that in the current approach it is indeed expected that zhengzai and a V 2 element such as gan 'dry' will be able to co-occur. Zhengzai is argued to be in Aspi0 instantiating imperfective viewpoint aspect, while the V2 gan is taken to be a suffix encoding completive aspect and hence licensed by the lower situation aspect head Asp2°. The two elements are hence not competing for the same aspectual head and so one might indeed expect that they could co-occur. The same is basically true of English where it is possible to combine imperfective aspect be...-ing with a meaning of completion/a completive situation:5 5 Stylistically the combination of a Vr V2 sequence with zhengzai may perhaps be avoided for the reason that both the V2 and zhengzai encode a separate focus— use of a V2 element focuses the meaning of completion and use o f an overt imperfective marker such as zhengzai focuses the progressive nature of the action. This may result in two independent foci in a single clause/sentence which may be felt to be somewhat confusing in terms of presentation of the information (and hence avoided). However, such strings are not ungrammatical. What may not occur is the combination of zhengzai with a Vr V2 sequence where the V2 is a phase-type V2 such as wan: (i) *wo zhengzai kan-wan shu ne I PROG look-finish book NE intended: 'I am finishing reading the book.' This I believe is because kan-wan is taken to be an instantaneous event with no internal or prior stages/extension over time available for modification by zhengzai. Being imperfective aspect describing the progression o f an action, zhengzai importantly requires an event which does have internal stages potentially available. As V2 's such as wan are themselves instantaneous achievement predicates and so may disallow any focusing on any internal stages of an activity depicted by the V|, instead encoding the single final stage of completion o f the event. Note that the instantaneous/durative distinction is one which is standardly assumed to be part of situation aspect/a predicate's Aktionsart rather than relate to higher viewpoint-type aspect, and predicates are distinguished as being either instantaneous (e.g., achievements) or durative (e.g., accomplishments, activities). Consequently it is natural that wan and other V2 's which might encode instantaneity should be licensed by Asp2 .sjn iatio n rather than A sp i.v jeW p0jIll. 397 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (33) John is finishing reading the book. (34) He is looking the reference up/washing the dishes clean. Another reason for believing that the Vi and -le are respectively completive and perfective aspect aside from their interpretation concerns their relative ordering on the Vi as suffixes. In section 6.1.2 it was noted that in terms o f relative scope, viewpoint aspect is naturally higher than situation aspect— viewpoint aspect applies to a predicate which already has a certain aspectual value as +/-telic, +/-durative established by the situation aspect and functions to focus in on some particular aspectual property o f the predicate (making its endpoints visible, focusing its internal stages, etc.). Such a hierarchical ordering should normally be reflected in syntactic structure and also here significantly in the ordering of morphemes on a lexical stem. For example, (as pointed out above in 6.1.2) it has been noted that when functional categories such as tense, mood and aspect are instantiated by free-standing lexical elements, these elements seem to consistently occur in a common cross-linguistic ordering and that interestingly this is precisely the inverse ordering found when the same functional types are represented by suffixes on a lexical stem (i.e., Baker's 1985 Mirror Principle). Specifically Cinque (1999) observes that in head-initial Finally it should be noted that the unacceptability of examples such as (32) with -le and zhengzai cannot be ruled out by any similar considerations. In (32) -le is clearly not occurring in the V2 position and it is therefore not possible to suggest that -le is unacceptable because it makes the predicate +instantaneous (thereby being in conflict with the +durative requirements of zhengzai). As pointed out, the +/-instantaneous property is encoded in the situation aspect and consequently by V2; - le in Vi-V2 -le sequences is however not in the V2 situation aspect position and therefore effects some other higher function (perfectivity, most obviously). 398 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. languages there is a broad ranking of free-standing verb-associated functional elements as in (35): (35) free-standing morphemes: epistemic-modality — tense — root-modality — aspect - V The hierarchical ranking is actually far more refined in Cinque (1999), but (35) can be taken to translate into a functional structure approximately as in (36): epistemic modals tense root modals aspect VP What is significant here is that if the elements representing such functional categories are not free morphemes but suffixes instead, they are consistnetly observed to occur bound to the verb in the inverse linear ordering, as in (37): (37) Suffixes: V — aspect - root-modality - tense - epistemic-modality In Government and Binding theory (GB) this ordering of suffixes in a sequence which is the opposite of the linear ordering of corresponding free-standing functional equivalents has been referred to as the Mirror Principle (Baker 1985). It is suggested 399 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. that the (suffix) ordering arises due to the movement o f a verb (or other lexical category) through the functional heads which dominate it, sequentially attaching the suffixes which are base-generated there. For example, supposing that a verb occurs with suffixes corresponding to aspect and tense and these suffixes are labeled S i and S2 in Asp0 and T°, the verb will first move to Asp0 and attach Si and then subsequently raise to T° to attach S2. The final ordering will be V-S1-S2 which is the mirror image of the linear ordering of the relevant functional heads: (38) Movement 1 — > V-Si TP Spec NP 400 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (39) Movement 2 — > V-S1-S2 TP Spec r < ^ • 0 Asp Asp V P V' NP ti In the Minimalist Program, for a variety of reasons it is assumed that inflectional affixes are actually attached to their stems in the morphological component prior to syntactic insertion in a tree and that they are licensed (checked) via subsequently raising to the relevant functional heads during the course of the syntactic derivation. The Mirror Principle effects are accounted for by assuming that there is an inherent ordering to the checking of all suffixal elements and that inner suffixes adjacent to the lexical stem must be checked before outer suffixes. Given a base-generated sequence of a verb and two suffixes V-S 1-S 2, suffix Si will then have to check against a functional head before the S2 suffix can be licensed. Consequently the functional head corresponding to Si will necessarily need to be lower in the tree than the functional head which licenses S2. Concerning these two frameworks, in chapter 5 I have argued against the overt raising of V° for the attachment of -le. Recall that in (40) overt raising o f V° to -le in Infl/Asp fails to explain why VP adjunct precedes V-le after the overt raising of the verb. Therefore it seems that the Minimalist 401 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Program in which -le could be assumed to be a verbal suffix checked at LF would seem to be a naturally more desirable choice to take. (40) ta man-manr-de xi-ganjing-le yifu he slowly wash-clean-LE clothes 'He slowly washed clean the clothes.' Given the basic Mirror Principle patterning and considering Chinese Vi-V 2-le sequences, one would now expect that the V2 element as an inner suffix Si should correspond to a licensing functional head which would be projected lower in the structure than the outer S2 suffix -le, and this is indeed exactly what it entailed by the relative hierarchical ordering suggested to exist between viewpoint and situational aspect. Above it was noted that viewpoint aspect applies to a predicate which has already been determined for its situation aspect, and that viewpoint aspect is therefore hierarchically superior to situation aspect. In terms of syntactic structure, viewpoint aspect and situation aspect as functional projections in a head-initial language are therefore expected to projected themselves as in (41) (=24): 402 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ( 4 1 ) AspP I=im/perfective Asp i AspP 2=situation In any sequence of two suffixes relating to viewpoint and situation aspect on a single verb, the inner aspectual suffix should therefore be licensed against the lower Asp2°S itu a tio n and the outer aspectual suffix should be licensed by the higher Asp M m /perfective- In Vi-V 2-le sequences this is exactly what is suggested to be the case~the inner 'V2 suffix is argued to be an instantiation of completive (situation) aspect and so should be licensed against Asp2°, and the outer -le suffix is suggested to be perfective aspect and so should be licensed against the higher Aspi0 head. The relative ordering of V2 and -le as suffixes on the Vi therefore seems to be further good confirmation that -le in V i-V2-le sequences is indeed occurring as perfective aspect. A final relevant point supporting treatment of -le as perfective aspect which has often been made in the literature is that verb -le and the negative form mei(-you) 'Neg-have' would seem to be in simple complementary distribution as seen in (42): 403 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (42) wo mei-you xie(*-le) xin I NEG-have write(*-LE) letter 'I didn't write the letter.' Commonly it has been assumed that mei-you is the combination of negation and perfective aspect (see e.g. Li & Thompson 1981). In a Chomskean approach one may therefore assume that the auxiliary element you is in a perfective aspect head. The complementary distribution of mei(-you) and verbal -le consequently has a natural explanation. Assuming that the perfective aspect head maximally checks a single set of perfective aspect features, if you and -le co-occur and both carry perfective aspect features, one of the sets of features will remain unchecked and cause the derivation to crash. In negative perfective aspect sentences the perfective aspect head is occupied by you (or alternatively a phonologically null equivalent) and the features are checked in-situ ; it is therefore not possible to have a second instantiation of the same aspect type (i.e., -le).6 7 There are therefore a variety of good reasons indicating both that -le is representing perfective aspect in the cases reviewed and that formally -le may be 6 The situation in Chinese here is similar to ungrammatical English sentences such as (i): (i) *John did not walked to school. Parallel to the patterns with aspect in Chinese, it would seem that tense/T° can not check more than a single set o f Tense features. 7 Note that sentence-final -le can co-occur with mei-you 'have not,' as shown in (i): (i) w ohoulaijiu mei-you zai huiqu Beijing le. I later then Neg-have again return Beijing LE 'I never returned to Beijing after that.' As with many others, I assume that sentence-final -le is a morpheme distinct from verbal -le and not an instantiation of perfect aspect. In many varieties of Chinese, such as Cantonese, the equivalent morphemes to verbal -le and sentence-final -le are indeed pronounced in quite different ways, indicating that they are different morphemes. 404 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. licensed as a functional suffix by raising (at LF) to a higher Asp i-p erfectiv e head position which dominates Asp2C o m p ietiv e- 6.1.4 Grammaticalization and the Dual Status o f Verbal -Le Section 6.1 began with the suggestion that verbal -le historically was a member of the V2 set and grammaticalized as an instantiation of completive aspect in sentence- final position just like all the other V 2 elements. This analysis of -le as a simple V2 was supported by its literal completive meaning and by the fact that it underwent re positioning to the Vj at essentially the same time that other V2 elements re positioned themselves adjacent to Vj. Now however it has been suggested that verbal -le should be taken to instantiate perfective aspect rather than completive situation aspect in a number of rather clear synchronic cases. The question now is obviously how to reconcile these two rather different views of -le. I would like to suggest that the element liao/le did indeed originate as an instantiation of completive aspect as indeed proposed in 6.1, but that since this time - le has in fact undergone further grammaticalization and re-analysis as perfective aspect, and that both this diachronic development of -le and its current sometimes ambivalent status can be neatly and naturally captured in a development of the movement-and-re-analysis approach to grammaticalization introduced in chapter two. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.1.4,1 P_^aJlel jp ^m m atic^ .X°r heads and affixes It may be remembered that the analysis of ge in chapter two made use of an idea (originally suggested in Simpson 1998) that grammaticalization may often result from a combination of movement and re-analysis within the functional structure dominating a lexical element. In the case of ge it was argued that ge originated in Cl° but then later became re-analyzed as an instantiation of D° following continued raising from Cl° to D°. The process which leads to grammaticalization and category change is then taken to be a sequence of movement to a particular functional head/Spec and then subsequent re-analysis relative to/in that position. Essentially there are three basic stages to the process: (43) Stages of grammaticalization (for X°-head elements)8 Stage 1: an element a is base-generated and remains throughout the derivation in its position of origin, X° Stage 2: a is base-generated in X° but then raises up to a second functional head position Y° as a is associated with the interpretation of position Y° Stage 3: a becomes re-analyzed as being base-generated in Y° and no longer is interpreted as instantiating the properties of X°; consequently a new element p is/can be base-generated in X° Trees (44)-(46) show the process of change from stage 1 to stage 3. In stage 1 the element a is simply base-generated in X° and remains there. In stage 2 a is base generated in X° but also raises up to the functional head Y° and is licensed/checked 406 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. also against Y°. In stage 3 a is re-analyzed as being directly base-generated in Y° and has effectively grammaticalized as an element which only instantiates a head o f type Y°. A direct consequence of a being re-analyzed as being base-generated in Y° is that X° may be filled with a new discrete element, here p. (44) Stage 1 a (45) Stage 2 8 XP elements may go through a similar route of grammaticalization, as Simpson (1998) suggests for French pas (see chapter two). 407 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (46) Stage 3 YP X° I P Such an approach to grammaticalization was initially developed to account for category changes occurring with free-standing heads and phrases which also undergo overt positional changes. For example, French pas was noted to have undergone raising from canonical object position and then later grammaticalized as a simple marker of negation (Simpson 1998b), the English modal set is commonly taken to occur base-generated in a position higher than lexical verbs having earlier undergone raising from the V° position and grammaticalized in a higher functional head (Lightfoot 1979), and Chinese ge was similarly argued to have raised from Cl° to D° and then grammaticalized as a D°. If it is assumed that the same essential modes of interpretation should be equally open to both morphologically free and phono logically dependent elements, i.e., affixes, then it might be imagined that the basic underlying process of re-analysis outlined above should also in theory be available with affixal elements, although somewhat different in its surface execution due to the phonological free/bound distinction. Inflectional affixes are indeed interpreted as instantiating functional heads and hence correspond to a genuinely real 408 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. part o f the syntactic structure projected in a tree. It is simply due to their phonologically dependent nature that affixes first require attachment to some host before they can be raised and matched against a relevant syntactic head. Given then that the free/bound distinction is therefore really just a phonological property and is not assumed to correspond to any kind o f fundamental semantic difference (i.e., tense as a suffix will not be interpreted differently from tense instantiated by a free standing morpheme), it would be natural to assume that the re-interpretation of an element as a different functional type should not be restricted to just free-standing morphemes but also occur with morphologically bound elements. Below I therefore now outline exactly how a process of grammaticalization with affixal elements might formally be understood to take place, and then suggest that just such a sequence of change and re-analysis has in fact occurred with Chinese verbal -le. In a Minimalist approach to inflectional morphology, affixes are assumed to be licensed not immediately when inserted on their lexical host (e.g. in V°) but only later on in the derivation when they are raised to the relevant licensing functional head. In this sense they become 'active' only when matched against a particular functional head during the course of movement through the functional structure dominating a lexical element. Grammaticalization and re-analysis of affixes will below essentially be suggested to occur when a particular affix first comes to be 'active' and licensed against a series of two distinct functional heads, and then is interpreted as being actively licensed only against the higher of these heads. The movement and re-analysis sequence found with free morphemes will then also occur 409 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. with affixes, but with the difference that because affixes are phonologically parasitic on a secondary host throughout the derivation, they may not be simply base generated independently in a different higher position after an occurrence of re analysis but will instead be simply re-interpreted as being 'active' and engaged in licensing relations with functional heads during a different progressively higher portion of raising through the functional structure. In more detail, the following sequence o f steps can be suggested to take place in affixal re-analysis. Stage 1 represented in (47) is the simple case of a suffix a attached to a lexical host S raising with S to be licensed against a functional head X°. The functional head X° essentially represents the interpretation given to a (e.g. +past tense), and at an abstract level of analysis it is assumed that a set of features corresponding to the meaning o f X° (i.e., +past) is lexically added to or built into the overt entity a . The morpheme a is therefore the physical host for a feature-set (i.e., meaning) which corresponds to a parallel specification in the functional head X°. S and a are base-generated together in the head-position Z° and then raise together up to X°. The lower bracketed (x) following a is a visual specification of the type of functional head that a needs to be matched/checked against. Similarly on the functional head X° there is a lower bracketed a indicating that X° corresponds to the meaning taken to be instantiated by a. In (47) S-a (X ) simply raises up to X° and the features carried by a are checked and licensed against those on X°: 410 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (47) Stage 1 In stage 2 with free-standing morphemes such as ge and French pas it was argued that ge/pas both fulfil two functions. Ge in stage 2 instantiates both the CI° head and the D° head, and pas is both an object DP and an emphatic marker in negative contexts. With affixes it can be assumed that this same kind of multi-functionality may also sometimes occur; in stage 2 of affixal change it can therefore be suggested that a single affix a is able to act as the physical lexical host for two different feature sets. This will essentially result from a being understood to be 'active' and engaged in a checking relation not only against X° but also a higher functional head Y° as it raises up to Y° with the lexical stem S. One may suppose that prior to this change the relevant features on Y° would have been checked either by (features on) a phonologically null affix attached to S-a or via the base-generation of a phonologically null head with a feature-set directly in Y° (i.e., base-generation of a free-standing head in the Y° position). From an initial situation (stage 1) in which 4 1 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. features on Y° are checked by an element with zero phonological realization and a is interpreted as checking only features on X°, in stage 2 a comes to be interpreted as being in a licensing relation with both X° and Y°. This dual function stage 2 is represented in (48). As seen, the suffix a is followed by a lower bracket enclosing an ordered sequence of x and y indicating the functional heads which it is understood as checking against; X° and Y° also both have lower-bracketed specifications for a encoding the fact that these heads are interpreted as having meanings licensed by the presence of the element a on S: (48) Stage 2 YP S-O. (X ,y) In stage 3 with free-standing morphemes, ge and pas were argued to have undergone re-analysis as instantiating only the higher position raised to (D° and SpecFocP respectively). In affixal terms, stage 3 will correspond to a situation in which the suffix a in the model here is re-analyzed as only representing the interpretation of the higher head Y°. Formally, features encoding the meaning o f only the functional 412 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. head Y° will be lexically added to a, and during the course of raising the suffix a will only be in an active licensing relation with the higher head Y° and not X°. Re analysis of free-standing morphemes in a higher position in stage 3 was seen to allow for the original base-generated site to be occupied by another different element (a classifer in the case of ge, and an object DP with pas). When a suffix such as a undergoes stage 3 re-analysis as only relating to a higher head, this should allow for a new affixal element to be base-generated in suffixal slot previously occupied by a, or alternatively a phonologically zero affixal element is understood as carrying the relevant features relating to X°. The above is all represented in the stage 3 tree (49). Note that the suffix a now only has a lower bracket specification for Y°, and a new suffix P (or a phonologically zero suffix 0 ) carries the features which will be licensed against X°: (49) Stage 3 YP Z I S-P(x/0(X)-Ot(y) 413 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note furthermore that the above sequence of change and re-analysis may be taken to apply equally to either overt movement of an element from Z° to X° to Y° or to covert LF raising. In Chomsky (1995) it is suggested that raising at LF only involves purely formal feature-sets, and other semantic-type features are stranded. This perhaps might lead one to assume that LF movement would simply raise an unstructured cluster of features rather than a structured object such as the hypothetical form [S-(3(X / 0 ( X )-a (y )]- However, there is good reason to believe that what raises at LF cannot in fact be fully unstructured. If ordering and hierarchical structure were absent from the LF-raised feature-sets which correspond to a lexical stem and its affixes, one would not expect to find Mirror Principle type affixal ordering phenomena in languages where movement occurs only at LF. Specifically, supposing a verb were to have tense and aspect suffixes but only raised to Asp0 and T° at LF, if the LF feature-set were to be fully unordered then one might expect that the tense and aspect suffixes would not need to occur in any particular order on the verbal stem and that they would simply be checked when the feature-set raised through the appropriate heads. However, Mirror Principle suffixal ordering phenomena occur in the same way both in languages with overt movement and languages where the raising occurs only at LF. In the case of a verb with tense and aspect suffixes, it is found that aspect suffixes come closer to the verbal stem than tense both when raising of the verb is overt and when it is covert (Baker's 1985 Mirror Principle). What this indicates clearly is that LF raising actually involves an element with as much and effectively the same internal structure as equivalent forms 414 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. which are raised overtly. In the case of the stage 3 sequence in (49) then, the hypothetical form [S-P(X / 0 ( X )-a (y > ] will essentially raise with the same internal structuring whether this raising is overt or at LF. Finally a word should be said about the hierarchical direction of affixal re analysis. If such a process basically parallels the grammaticalization of free-standing elements, then one would expect that re-analysis should follow the path of movement and only go in an upward direction; specifically one would expect that an affix originally encoding the meaning o f a functional head X° would possibly allow for re analysis as relating to a higher head Y° , but not allow for downward change and re analysis as relating to a functional head W° hierarchically lower than X°. Careful research on affixal change is obviously first necessary to establish that there is indeed an 'upward' direction to re-analysis with phonologically bound elements, but imagining this to be a likely conclusion one might suggest the following to account for the patterning. It may be suggested that an affix essentially becomes 'active' when it and its host raise up to the head against which the affix is licensed and that before such a point the affix is effectively completely ignored. Having once become 'active' though, it may be possible for the affix to be interpreted as also engaging in a licensing relation with some subsequently higher functional head, especially if this head is otherwise interpreted as being licensed by a phonologically zero affix/form. Re-analysis would then be possible only in an upward direction after an affix had become active in licensing terms, and never in a downward direction - prior to becoming active the affix will simply not be visible for any possibility of re- 415 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. interpretation. In such a way then affixal re-analysis would also be dependent on the upward direction of movement in a tree. Turning back to Chinese, I would like to suggest that the hypothetical three- stage process of affixal re-analysis proposed here now allows for a simple and natural explanation of the diachronic patterning and synchronic status of verbal -le. In sections 6.1.1 above it was argued that there are good reasons to believe that verbal -le originated as an instantiation of completive aspect; liao/le both originated as a verb meaning 'to finish' and hence had the typical lexical meaning of other completive phase Wjs, and liao/le underwent re-positioning adjacent to the verb at the same basic time that other V is are taken to have re-positioned themselves. In section 6.1.3 however, other evidence has indicated that synchronically verbal -le appears in many situations to be rather clearly instantiating perfective viewpoint aspect— notably when it occurs in Vi-V2-/e forms and also with achievement predicates. The view o f affixal grammaticalization and re-analysis proposed above now allows one to reconcile these apparently conflicting views. I would like to suggest that -le as a suffix has over time gone through a process of upwards re- analysis, beginning as an instantiation of completive aspect/Asp2° and then becoming re-interpreted as perfective aspect licensed against the structurally higher Aspi°. This hypothetical process o f grammaticalization is illustrated in (50-52) below. In stage 1 of the process -le is taken to be a simple instantiation of completive aspect licensed against Asp2° after raising at LF as in (50) (using the verb mai 'to sell' as a verbal stem): 416 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (50) Stage 1 A s p i P (im/perfectivc) Aspf Asp! A S P 2P (completive) Asp2 (le ) mai-le (A sp 2) In stage 2, -le is re-interpreted as instantiating not only completive aspect but also perfective aspect, being active in a checking relation not only with Asp2° but also the higher Aspi°, as in (51): (51) Stage 2 Asp 1P (im/pcrfectivc) A S P 2P (completive) Asp 1 (le) Asp2 (le ) mai-le (Asp2, Asp i ) 41 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In stage 3 of the process, -le comes to be re-analyzed fully as potentially instantiating only the higher aspect type Asp i p e rfe c tiv e - This then allows for a new suffixa1 instantiation of the lower Asp2C O m p ie tiv e to be inserted in the completive aspect position immediately right-adjacent to the verbal stem, as shown in (52) (using the completive suffix wan 'be finished'; alternatively completive aspect features might be carried by a suffix with zero phonological realization 0 (A sp 2)): ( 5 2 ) Stage 3 AspiP(im/perfective) Such a process of upward affixal grammaticalization within the functional structure is able to provide a rather simple account of both the origins of le/liao and its current status in Vi-V2-le forms. Having begun as completive aspect, -le is taken to have grammaticalized further up the functional structure to be re-interpreted as perfective aspect, allowing for its current co-occurrence with clear completive aspect suffixes such as wan 'finish' etc. Aspi A s p i (ie) Asp2P(completive) Asp2' Asp2 (iva n / 0 ) VP ▲ V' V mai-0(AsP 2/wan(Asp2)-le(Asp\) 418 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6--l.-4,2Cu!Tentdistincti_onbe^e^ diachronic development The suggested diachronic development of -le modeled in (50-52) might also seem to be supported by certain interesting patterns pointed out in Sybesma (1999) and Lti (1980). In the latter work it is suggested that there are 28 verbs in Modem Chinese which verbal -le combines with as a V2 resultative type element similar to other phases.9 Examples both in Lu (1980) and Sybesma (1999) with verbs such as mai 'sell,' giian 'close,' wang 'forget,' and chi 'eat' show that there are distinct contrasts in the patterning of these verbs and other verbs not in the group of 28. Basically the verbs from this 28-verb group can combine with -le without indicating perfectiveness, although a completive reading of the action verb is intended. For example, it is possible to combine many of the group of 28 with -le and a higher modality verb such as xiang 'think of (doing X),' keyi 'be allowed to' or yao 'want to,' whereas other verbs do not allow this. The examples below are mostly taken from Lu and Sybesma: (53) ni bu ai ting, keyi guan-le shouyinji. you NEG like listen, can close-LE radio 'If you don't want to listen, you can turn the radio off.' 9 Those 28 verbs listed in LQ (1980) are: wang 'forget,' diu 'throw/get ride o f,' guan 'close, shut,' he 'drink,' chi 'eat,' yan 'swallow,' tun 'swallow,' po 'splash,' sa 'spill,' reng 'throw/get rid of,' fang 'release,' tu 'scribble,' mo 'wipe,' ca 'wipe,' peng 'bump,' za 'break,' shuai 'throw,' ke 'crack,' zhuang 'hit,' cai 'step on,' shang 'injure,' sha 'kill,' zai 'kill,' qie 'cut,' chong 'flush,' mai 'sell,' huan 'return,' hui 'destroy.' 419 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (54) wo xiang mingtian mai-le nei-liang-che I want tomorrow sell-LE that-CL-car 'I'm thinking of selling off that car tomorrow.' In (53) and (54) the verbs guan 'close' and mai 'sell' are members of Lii's group of 28. In (55) and (56) it is seen that verbs such as xie 'write' and kan 'look at' not in this group do not allow combination with similar modality verbs and -le: (55) *wo xiang/yao xie-le yi-feng-xin I want/want write-LE one-CL-letter intended: 'I want to write a letter.' (56) *ni yao zhidao, keyi kan-le zhei-ben-shu you want know can look-LE this-CL-book intended: 'If you want to know (this), you can read this book.' Sybesma also notes that verbal -le may combine with verbs such as wang 'forget,' a member of the group of 28, and the habitual adverb zongshi resulting in a completion type reading, but that this kind of combination is again not possible with verbs from outside the group of 28, as seen in (57-58): (57) wo zongshi wang-le ni-de-mingzi I always forget-LE you-DE-name 'I always forget your name.' (58) *wo zongshi kan-le ta xie de shu I always look-LE he write DE book intended: 'I always read the books he writes.' 420 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note that -le does not have a perfective meaning in (57), and that (57) describes a situation which is not closed with visible endpoints. Finally, Sybesma shows that members of the group of 28 may actually co occur with the negative perfecive element mei-you 'Neg-have,' which is not at all possible with verbs from outside the group of 28: (59) wo hai mei-you mai-le nei-liang-che I still NEG-AUX sell-LE that-CL-car 'I still didn't sell off that car.' (60) *wo hai mei-you xie-le nei-feng-xin I still NEG-AUX write-LE that-CL-letter intended: 'I still didn't write that letter.' From all this patterning Sybesma concludes that there must indeed be two distinct verbal -le's, one which occurs as a phase V2 element with a restricted group of verbs (the group of 28), and a second element which is fully general in its occurrence. The first of these Sybesma calls 'Endpoint' -le and the second 'Realization' -le. Both are treated as small clause predicates base-generated as complements to the main verb V). (61) corresponds to a surface string mai-le che 'sell-LE car' with -le occurring as a phase V2 signaling simple endpoint with one the group of 28 verbs: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (61) Sybesma's Endpoint -le VP I V’ V, SC mai NP V 2/ / e (endpoim) 'sell; che 'car' (62) is the structure suggested by Sybesma to underlie the use of Realization -le and a surface sequence such as xie-wan-le shu 'write-finish-LE book.' Note that here the V2 wan has to raise to left-adjoin to -le and then the sequence wan-le has to raise and right-adjoin to xie in order for the word order to match the surface string: (62) Sybesma's Realization -le VP I V' V, SC xie ' 'write' V SC '(realization) NP shu 'book' wan 'finish' For a variety of reasons given earlier in chapter 5 I decided to explore an alternative solution to the small clause analysis of RVCs and -le. With regard to 'Realization' -le two more brief criticisms may now be added. First o f all the small clause analysis 422 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. really appears to build functional categories into positions under the lexical VP and selected by the main predicate which is intuitively rather strange. Functional categories (such as for example 'realization') are normally assumed to be functions which take lexical objects as their arguments rather than the reverse (i.e., lexical elements selecting functions). Secondly the structure in (62) in which realization -le predicates of a (second) small clause as its subject leads one to expect that it should be possible for -le to stand alone with this small clause in an isolated predication relation. However, whereas this may sometimes be possible, and (64) corresponding to the hypothetical lower small clause part of (63) is indeed a well-formed string, in other cases this is not at all acceptable, as (66) relating to (65) shows: (63) ta xi-ganjing-le yifu he wash-clean-LE clothes 'He washed the clothes clean.' (64) yifu xianzai ganjing-le clothes now clean-LE 'The clothes are now clean.' (65) ta kan-wan-le shu he read-finish-LE book 'He finished reaind the book.' (66) *shu xianzai wan-le book now finish-LE 'The book is now finished.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It is not clear if the -le in (64) and (66) is a verbal -le or a sentence-final -le, although the contrast in small clause possibility clearly exists. A clearer example against verbal -le heading a small clause presents itself in (67). For (67), Sybesma has to assume a structure such as (68) to account for fang-zai-le 'put-at-LE' string, which in turn wrongly predicts that the small clause headed by -le can be separated from Vi. As shown in (69), V2 zai can clearly be separated from V) in (69a). However, (69b) indicates that when -le is present, zai-le has to co-occur with V) as in (67), otherwise the sentence is ungrammatical. (67) ta ba shu fang-zai-le zhuozi-shang he BA book put-be-LE table-top 'He put the book on the table.' 10 V ^(realization) shu V2 'be-at' zhuozi-shang 'table-top' 1 0 Note that sentences such as (67) are not acceptable with -le in all dialects of Mandarin, but can nevertheless frequently be found in mainland Chinese writing. No dialect (to my knowledge) however allows for sentences such as (69b). 424 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (69) a. shu zai zhuozi-shang book be-at table-top 'The book is on the table.' b. *shu zai-le zhuozi-shang book be-at-LE table-top This would therefore seem to confirm that when -le occurs with a Vi and a V2 as in (63), (66) and (67), it is not hierarchically lower than the Vi, predicating of a small clause as its subject (as per (62)), but rather a higher function applying to the whole [V1-V2 NP] as suggested in the perfective aspect analysis put forward here. The observations made concerning Lii's group o f 28 verbs can now be interpreted in a potentially revealing way in the present attempted modeling of the grammaticalization of verbal -le. Sybesma's interesting insight (following Lii) is that -le with a restricted set of verbs might seem to function in Modem Mandarin like an instantiation of the V2 group. If this is true, and -le can indeed simply instantiate completive aspect with certain verbs, it would seem to substantiate the assumption made on other grounds that verbal -le indeed originated as a completive aspect suffix. Typically, as mentioned in chapter 5, each member o f the V2 group would seem to be lexically selective and combine with a limited number o f V 1 elements. If there is an element -le which only combines with a particular set of 28 verbs resulting in a completive meaning, this would seem to identify -le in these cases as really being a canonical (completive aspect) V2 element. Such a conclusion concerning the synchronic patterning of -le with a subset o f verbs now arguably leads to a strengthening of the grammaticalization account. It can be suggested that 425 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. after re-positioning adjacent to the main verb Vi, -le originally was a completive aspect suffix in stage 1 o f the re-analysis process represented in (50). At this point in time -le would have only occurred with a restricted set of Vi elements just as other V2 completives did. Later on -le can be argued to have entered into stage 2 of the grammaticalization process shown in (51) potentially instantiating both completive aspect and perfective aspect. Here I think it is natural to assume that the possibility of a stage 2 interpretation of -le with certain verbs (i.e., as instantiating completive aspect as well as perfective aspect) would not have necessarily immediately excluded a stage 1 interpretation o f -le as instantiating only completive aspect. In other words it may be assumed that -le was only optionally interpreted as instantiating both completive and perfective aspect and that both stage 1 and stage 2 type interpretations could actually be available during a single time period (formally, any perfective aspect features present could be added either to -le, or to some other phonologically null affix so that -le would be interpreted as only instantiating completive aspect). Such an assumption essentially reflects the intuition that certain types of diachronic change are not effected catastrophically at a single point in time but instead occur over extended periods of time. The arrival at stage 2 with -le optionally being interpreted as instantiating perfective aspect/carrying perfective aspect features is however highly significant. It can be argued that once perfective aspect features could be added to -le with a certain set o f verbs, then this option might have also became quite generally available and - le could have come to be interpreted as encoding perfectivity with a full and wide 426 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. range of verbs. In the V2 completive aspect (suffix) position, -le would have contrasted with a large number of other V2 elements and so it is not unreasonable to accept that it would be lexically restricted, each V2 selecting a certain set of situations. However, when in stage 2 optionally instantiating perfective aspect, -le would not have stood in contrast with other (overt) instantiations of perfective and so it is consequently very natural that it should have become a fully general marker of perfectivity. Furthermore, although completive aspect Asp2° selects directly for a predicate VP and hence a complement with clear descriptive content, perfective aspect Aspi0 selects for an ASP2P which is essentially just a telic situation and hence far more abstract. It is therefore not surprising that lexical restrictions should occur in the selection relation between the instantiaion of Asp2C o m p ie tiv c and VP but not between Asp 1 p erfectiv e and Asp2P and that -le as a perfective marker should be quite general in its applicability. This point at which -le may be taken to instantiate just perfective aspect with verbs which do not allow -le as a simple completive is stage 3 of the process - -le is available as a pure instantiation of the higher head Asp 1 p e rfe c tiv e - The hypothetical sequence of change is partially summarized in (70): 427 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (70) a. liao occurs as completive aspect sentence-finally b. liao re-positions adjacent to the verb as a suffix and reduces to -le still as completive aspect = stage 1; -le is (assumed to be) lexically restricted as other V 2c o m p ie iiv e elements are c. -le can optionally be interpreted as instantiating perfective aspect as well as completive aspect; this possibility corresponds to stage 2. A stage 1 interpretation of -le with a restricted set of verbs as just instantiating completive aspect is also still possible d. the possibility that -le can instantiate perfective aspect initiated with the restricted set of verbs becomes fully general, arguably due to the lack of contrast of perfective -le with other parallel perfective morphemes. This represents stage 3 of the process. It is important to stress here that although stage 1-3 of the re-analysis process are taken to be sequentially ordered, the occurrence of a later stage of the process does not necessarily exclude an earlier stage still being available as an interpretation. Thus in order for stage 3 to be reached it is assumed that there was first the possibility of a stage 1 interpretation and later a stage 2 interpretation. However, the possibility of a stage 3 interpretation does not mean that a stage 2 interpretation should necessarily no longer be available. The stage 1-3 interpretations for -le are listed in (71) below: 428 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (71) Stage 1 interpretation: -le instantiates only completive aspect Stage 2 interpretation: -le instantiates completive and perfective aspects Stage 3 interpretation: -le instantiates only perfective aspect What these stages represent are possibilities~the possibility of associating -le with a certain type of interpretation. In terms of formal features, stage 2 represents the possibility of adding perfective aspect features to -le in addition to completive aspect features, and -le being interpreted as being potentially active in a checking relation with both the lower Asp2co m p ictiv e and the higher Asp i p erfectiv e- However, in a situation of ongoing re-analysis over time it can be imagined that once a possibility (a new interpretation) becomes available, it does not have to be used in every instance; concretely in terms of the change with -le, the arrival o f the possibility of adding perfective features to -le (i.e., interpreting -le as perfective) does not mean that every time that -le might be used perfective features would have to be added to it in addition to completive aspect features, and not every occurrence of -le would have to be interpreted as perfective in addition to being completive. Similarly the possibility of a stage 3 interpretation o f -le as instantiating just perfective aspect would not necessarily mean that a stage 2 interpretation should no longer be available with verbs from the group which tolerate -le as instantiating completive aspect. What this is intended to suggest in general is that re-analysis of an element of type A as possibly instantiating a second function B does not have to result in the loss of the original A-type interpretation, even though this may in fact even frequently occur over time. In the case of Mandarin -le I would like to suggest that 429 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the patterning highlighted by Sybesma (1999) showing critical differences between Lu's group of 28 verbs and other predicates when combined with -le indicates that le synchronically may still indeed be interpreted with stage 1 status when occurring with members o f the group of 28 verbs. When appearing with all other verbs -le occurs interpreted with its stage 3 meaning of perfectivity alone. Finally there may be evidence that -le can also be interpreted as instantiating both completive and perfective aspect with verbs in the group of 28, i.e., stage 2. First of all it is clear that -le can occur with verbs in this group instantiating just perfective aspect, i.e., stage 3. In (72) below there is an overt completive V2 element occurring in addition to -le and so -le must here be instantiating perfective rather than completive aspect: (72) ta chi-wan-le san-tiao-yu he eat-finish-LE 3-CL-fish 'He ate three fish.' When -le occurs without a second distinct completive V2 as in (73) the interpretation is effectively the same as in (72) and all three fish are understood to have been consumed: (73) ta chi-le san-tiao-yu he eat-LE 3-CL-fish 'He ate three fish.' Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. If completion of the eating of all three fish results from the use of completive aspect, it may consequently be suggested that -le in (73) also instantiates completive aspect and hence has a stage 2 interpretation. Summarizing then, it has been argued that -le has developed over time from an original status as a marker of completive aspect similar to other V2 elements to also instantiate perfective viewpoint aspect. Synchronically it has been suggested that -le actually has an ambivalent current status and that there is evidence indicating that it may still occur instantiating either stage 1, 2, or 3 of the hypothesized route of diachronic change, this being represented in the table below: Current potential interpretations of — le Stage 1 interpretation: -le occurs with a restricted group of 28 verbs instantiating just completive aspect Stage 2 interpretation: -le occurs with a restricted group of 28 verbs instantiating both completive aspect and perfective aspect Stage 3 interpretation: -le occurs as a general instantiation of perfective aspect Quite generally in this section I have argued for the significant conclusion that affixes are subject to the same type of (further) grammaticalization and re analysis that occurs with free-standing morphemes. It was argued that as with the re analysis of free-standing morphemes, this process of change takes place in a way which importantly parallels the direction of syntactic movement upward in a tree, -le as a completive aspect suffix coming to be re-interpreted as a perfective aspect suffix licensed by an AspiP which is higher in the functional structure than the completive 431 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. aspect head Asp2°. If this conclusion is correct and affixal re-analysis does mirror the re-analysis of free morphemes in relating an element to successively higher functional projections, it is clearly important for two reasons. First of all, it would seem to provide strong evidence in support of the Minimalist assumption that affixes are licensed via movement to higher functional heads— specifically, if affixal re analysis shows parallels to the grammaticalization of free morphemes and the latter occurs as a result of movement to a higher functional position and re-analysis in this position, parallel affixal re-analysis can then also be assumed to result from movement and re-analysis relative to a licensing functional head. Secondly, the particular case of -le would further seem to substantiate the Minimalist hypothesis that movement for the licensing of functional morphemes/heads (i.e., feature- checking) may frequently take place covertly at LF. As -le does not appear to occur overtly raised out of V°, it must be assumed that the movement to Asp2° and Aspi° must occur at some later derivational point hence LF. Considerations of language change and affixal re-analysis such as outlined here can thus be suggested to provide important evidence and arguments in favor o f critical Minimalist assumptions concerning the licensing o f functional interpretations in syntax. 6.2 Verbal -Le and Tense In this last section of the chapter I would like to explore the rather contentious possibility that verbal -le is currently coming to instantiate past tense in Modem 432 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chinese in addition to its other specifications.1 1 Such a speculation has been specifically argued against by linguists such as Li & Thompson (1981) and there would furthermore seem to be a quite wider tacit assumption that -le cannot be tense. However, here I will attempt to show that not only are there good reasons to support an analysis of -le as plausibly instantiating tense, there are also quite reasonable counter-arguments to Li & Thompson's objections to such an analysis. The structure of the section is as follows. First I will outline the evidence and patterning which suggests that -le may be coming to instantiate past tense. Then I will discuss the relation of aspect to tense, in particular focusing on the notion of perfectivity and argue that the categories tense and perfective must be taken to correspond to discrete syntactic heads, which may however be phonetically instantiated by a single overt morpheme. Subsequently showing how -le can rather naturally be analyzed as representing also tense in the model of grammaticalization proposed above, I will then go on to re-examine Li & Thompson's reasons for rejecting an analysis of -le as tense and point out how they are actually not as strong as might be initially assumed. Ultimately it will be concluded that even though certain aspects of the patterning of - le may remain in need of further explanation, a close re-consideration of all the 1 1 Chiu (1993) assumes that TP is a universal projection in all languages and in Chinese -le is a past tense marker that heads this universally projected TP in Chinese. See also Tsang (1978), Rohsenow (1978), Z. Shi (1990), Ross (1995) and Hsieh (1998) for other attempted suggestions that - le might be a relative past tense marker (vs. absolute tense marker: see the discussions in Comrie 1985). The particular focus and interests o f this chapter/section are (a) to attempt to show how the development of -le as a past tense marker is a natural consequence o f a certain formal approach to grammaticalization, (b) to show how the development of -le as a past tense marker can be understood better once a more explicit analysis o f Aspect is assumed (as argued for above), and (c) to provide clear and principled attempted explanations o f the wide variety of counter-arguments frequently given against any analysis o f -le as a past tense marker. 433 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. evidence and assumptions concerning the interaction of tense and aspect does indeed support the contention that verbal -le is now coming to functionally represent tense. The evidence in favor o f an analysis taking -le to instantiate past tense are really very simple. First of all it is well-observed that the use of verbal -le very frequently co-incides with an interpretation of the predicate as having taken place in the past. This simple fact is noted in almost all descriptions of the patterning of -le and in analyses of le's distribution in experimental studies. For example, Smith (1997) refers to Chang's (1986) investigation of -le's occurrences in newspaper articles as noting that -le was frequently found to be used as a past realis marker. In another experimental study, Spanos (1979) reports that subjects used -le when the context made it 'necessary to explicitly state the realization of an action.' (p.81). This characterization of a 'realized action' locates an event as having occurred and would seem to be a description of past tense simply using other terms. As noted earlier, Sybesma (1999) also refers to (one type of) -le as 'Realization -le' and this again is arguably just a camouflaged description of past tense. It is therefore quite uncontentious and a well-established observation that -le does indeed consistently occur in past time contexts, this clearly suggesting and supporting an analyis that it is used and interpreted as a past tense marker. A second simple observation indicating that -le is coming to instantiate past tense is the fact that in non-subordinate contexts only a past time interpretation is available when -le is used (with one potential exception to which I later return). For 434 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. example, (74) below can only be interpreted as having taken place in the past and neither a present tense nor a future reading is available: (74) wo zai Beijing canguan-le Gugong I in Beijing visit-LE Imperial Palace 'I visited the Imperial Palace in Beijing.' Not: 'I am visiting/will visit/will have visited the Imperial Palace in Beijing.' In (75) and (76) it is also shown that it is not possible to over-ride the past time reading by providing suitable future, present-time or habitual adverbs: (75) *wo mingtian (hui/cai-yao) zai Beijing canguan-le Gugong I tomorrow (will/then-will) in Beijing visit-LE Imperial Palace intended: 'Tomorrow I will visit/will have visited the Imperial Palace.' (76) *wo mei-tian/xianzai gen Zhangsan canguan-le Gugong I every day/now with Zhangsan visit-LE Imperial Palace intended: 'Every day I visit/I am now visiting the Imperial Palace with Zhangsan.' The fact that it is not possible to over-ride the past time interpretation occurring with -le here would seem to indicate that past time interpretation is not simply a coversational implicature which might arise when -le occurs as perfective aspect but significantly now actually part of the meaning o f -le in these contexts. One also finds that -le might sometimes seems necessary to signal a past time interpretation even though contextually a past time interpretation might appear to be obvious: 435 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (77) A: ni zuotian wanshang zuo-le shenme? you yesterday evening do-LE what 'What did you do last night?' B: wo shang*(-le) Fawen ke I attend LE French class 'I attended French classes.'1 2 There is then simple reason to believe that -le may be more than just a marker of perfective aspect but also have an association with the higher functional category of tense, specifically instantiating past tense in a way which cannot be cancelled as an implicature. Further plausibility to the contention that -le instantiates tense comes from a reflection on the likelihood o f such an interpretation. If it is supposed that -le originated as a marker of completive aspect and then grammaticalized to instantiate perfective aspect, it would be a highly natural third development for it to possibly re analyze further as an instantiation o f tense. Not only do completives and perfectives commonly give rise to past time implicatures which may naturally strengthen over time to become an encoded part of the meaning of a morpheme, formally it can be shown that such a re-analysis would also be very natural in the model of 1 2 Note that if a phrase such as zai xuexiao 'at school' is added into the structure of (77b), as in (i), it might seem that -le is not in fact necessary. (i) wo *(zai xuexiao) shang Fawen ke I be-at school attend French class 'I was at school attending French classes.' As an answer to (77A), the phrase zai xuexiao 'be at shool' is obligatory and the interpretation of (i) is as glossed, and not '1 attended French classes at school.' I suggest that in such cases zai xuexiao 'at school’ is in fact the first part of a serial verb construction with zai being a verb 'to be at.' It has frequently been noted (for example, Li and Thompson 1981, Tung and Pollard 1997) that -le combines with serial verb constructions in ways which are different from clauses with a single verb. The proper study o f serial verb constructions is beyond the scope of the present chapter. 436 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. grammaticalization proposed here. Le as a suffix could simply be re-interpreted as being active in a checking relation not only with A s p i ° Perfective but also with the T ° head which occurs above AspjP, i.e., -le would be interpreted as instantiating an additional higher head in the functional structure which the verb raises to at LF. If there are then rather straightforward indications of an association between - le and tense suggesting that -le may instantiate past, one should obviously ask why it is that this has not been previously assumed and why a tense-/<? link has even been explicitly denied. In order to do this I will first briefly re-examine the notions of tense and aspect, emphasizing that they are formally distinct properties and sketching out a hypothetical analysis of -le as tense and then turn to look at the specific objections Li & Thompson (1981) raise to the assumption that -le is tense. 6.2.1 Tense, Aspect and Perfectivity Section 6.1.2 presented an overview of Smith's two-tiered view of aspect which justifies the division of aspect into two clear major types - situation aspect and viewpoint aspect. Situation aspect corresponds to Vendler's (1967) notion of Aktionsart and the classification of a predicate according to its telic, durative and dynamic properties. Viewpoint aspect was then argued to apply to a predicate which already has a situation aspect characterization and present it from a particular perspective (viewpoint), essentially either as imperfective and focusing on the internal stages of an event or perfective and focusing on the event as a whole by making both its endpoints visible. These functions of aspect were then shown to be 437 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. independent of tense. In (78) below the situation type is +telic as the predicate is naturally bounded, and the viewpoint type encoded by the be...-ing form is imperfective, this resulting in a focusing of the internal stages of the event. It is also significantly seen that tense is an independent function which can be varied quite separately from both the situation aspect and the imperfective viewpoint— either a present, future or past tense specification can co-occur in the sentence: (78) Mary is/was/will be drawing a circle. Commenting on the tense/aspect distinction, Smith (1997, p.98) observes that: "Temporal location (i.e., tense) and aspect are complementary temporal systems. The former locates a situation in time, while the latter specifies the internal temporal structure of the situation." Assuming therefore that there is indeed good justification for taking tense and viewpoint aspect to correspond to formally distinct functions, and arguing furthermore that completive aspect is licensed by a discrete functional head potentially encoding situation aspect, the expanded structure in (79) was adopted for a head-initial language such as Chinese. In (79) each distinct function is represented by a distinct functional projection/head. In the course of the analysis, various overt instantiations of both types o f aspect heads were argued for, and although tense was not discussed there are also obvious free-standing candidate morphemes which may be taken to independently fill the T° position, e.g. hui 'will': Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (79) TP r AspP l=im/perfectivc Asp,’ Aspi0 s itu a tio n Asp2' Asp2 VP Considering in particular perfective now, following Smith (1997) its functional role with imperfective, such a role is assumed to be formally quite distinct from the addition of tense to a structure. A somewhat different view o f perfective is to be found in Bybee, Perkins, & Pagliuca (1994). These authors define perfectives as signaling a situation which is temporally bounded (p.54), but also seem to assume that perfective is often a complex category combining both aspect and past tense. Perfective is described as standing in contrast with imperfective aspect, but not being distinct from or combining with any simple past tense morpheme. In other words, whereas an imperfective aspect morpheme may combine with a discrete simple past tense morpheme to result in a past imperfective interpretation, commonly it is found that perfective morphemes do not combine with the same past tense morpheme to indicate past perfective; instead they signal past time 'inherently' without other overt support. Cross-linguistically this may perhaps often be true and there may well only 439 has indeed been taken to be a focusing o f the endpoints of a telic situation, and as Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. be a single overt morpheme resulting in the interpretation o f 'past perfective.’ However, it is important to recognize that there are nevertheless two distinct formal properties involved here~the property o f perfective aspect contrasting with imperfective and the property o f past time reference contrasting with present/future. As such properties are semantically quite distinct and also represented by discrete morphemes when the aspectual specification is imperfective, it is only natural to assume that there are indeed two functional heads/projections present in the syntactic structure (as in (79)) even when the two properties of past and perfective are phonetically encoded in a single overt morpheme. The observation which Bybee et al (1994) make that it is common across languages for there to be a distinct morpheme encoding past tense with imperfective aspect markers, but not so with perfective morphemes is however interesting, and indicates that perfective morphemes potentially may also signal a past time specification in some way. Structurally it must be assumed that distinct tense and perfective aspect functional heads must be present, but in terms of overt morphology it is actually a single element (the perfective morpheme) which may encode both the relevant properties. This is in essence what I have already argued in the case with -le and aspect, namely that the single overt morpheme -le may sometimes carry the specifications of two functional heads, completive and perfective aspect. In a development of the same theme in this section I will now go on to suggest that -le may also instantiate past tense in addition to perfective aspect. 440 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. First however, I would like to note that a developmental connection between aspect markers such as -le and the encoding of past time reference is indeed a common cross-linguistic phenomenon. Bybee et al (1994) observe that simple past and perfective markers which both signal past time reference frequently derive from completives and resultatives, i.e., elements of the V2 type in Chinese RVCs and hence also -le in its origins. Harris (1982) and Bybee and Dahl (1989) similarly discuss how resultative constructions often evolve into perfectives and pure past tense markers, referring to languages such as French, Italian, Dutch, German and Turkish among others. Harris and Campbell (1995) also point to the Mayan language Cakchiquel as another instance in which a completive aspect marker has been re-analyzed as a past tense morpheme. This common process of change is generally assumed to result from a natural increase in the strength of a past time association with resultatives and completives. Use o f a resultative to make reference to a result state is most common if the action leading to the result state has already occurred; consequently the use of resultatives often infers a past action (so does completives). Such a natural past time inference then may over time become strengthened to the point of becoming part of the genuine meaning of a morpheme. Writing about Chinese in particular, Smith (1997) writes that: 'Temporal location is often conveyed by a perfective or imperfective viewpoint. There is a conventional association of the imperfective with the Present and the perfective with Past. In the absence of other information, including adverbials, the viewpoints are taken to convey these times.' (p.279). Smith therefore assumes that perfective markers such 441 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. as -le also give rise to natural inferences of past time interpretation in Chinese as in other languages. While the suggestion that resultatives and perfectives frequently infer past time is both natural and easy to accept, it is important to note that there comes a time during language change when a conversational implicature strengthens to the point of becoming a real part of the meaning associated with a morpheme so that this meaning can no longer be denied or over-ridden. Smith explicitly argues that the past time interpretation found with perfective markers in Chinese is still essentially a conversational implicature (a 'conventional association') which is made: 'in the absence o f other information, including adverbials. . . However, as shown in (75) and (76) repeated below, in fact this is not true, and it is actually not possible to use adverbials to over-ride the past time inference with -le: (75) *wo mingtian (hui/cai-yao) zai Beijing canguan-le Gugong I tomorrow (will/then-will) in Beijing visit-LE Imperial Palace intended: 'Tomorrow I will visit/will have visited the Imperial Palace.' (76) *wo mei-tian/xianzai gen Zhangsan canguan-le Gugong I every day/now with Zhangsan visit-LE Imperial Palace intended: 'Every day I visit/I am now visiting the Imperial Palace with Zhangsan.' What this then now strongly suggests is that diachronically -le has indeed followed a very common and natural route o f change originating as a completive and later turning into a perfective marker, but that synchronically -le now also encodes past 442 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. time not simply as a pragmatic inference but genuinely as an instantiation of tense/T°. Considered from the point of view o f the more formal model of grammaticalization developed in 6.1.4, such a change to a situation in which tense is taken to be encoded on -le is also very natural. Attempting to account for the earlier aspectual change where -le hypothetically instantiated completive aspect with a restricted set o f verbs to the present situation where -le occurs as a general perfective marker it was suggested that this resulted from -le raising at LF with the rest o f the verbal complex and being re-interpreted as instantiating not only the lower Asp2° head (completive aspect) but also as being licensed against the higher Aspi° head (perfective). Here with regard to tense it can be suggested that -le simply becomes interpreted as being actively engaged in a licensing/checking relation with a higher head still, the T° head which selects for the perfective AspiP. As noted above, a past time/tense interpretation of a perfective/completive morpheme is a natural pragmatic inference resulting from the inherent meaning of completion/resultatives, and so the re-interpretation of perfective -le as instantiating not only perfective aspect but also the structurally higher past tense would also be a very normal extension o f its original interpretation. Formally the following stages of re-analysis can be suggested to result in the present synchronic situation. At an initial stage it may be assumed (as in 6.1.4) that - le is base-generated with completive/perfective aspect features and that tense features are inserted on a phonologically null suffix 0(Tns) as in (80) (or alternatively 443 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. base-generated on a free-standing phonologically zero morpheme inserted directly into T°). (80) considers a case where the verb stem attaches both completive and perfective aspect encoded by -le (raising through the functional structure as indicated at LF): (80) TP T' 'T - 0 I i (0) Stage 1 AspP | =im/perfective Asp,’ Aspi (|e ) AspP 2=complcUve Asp2' Asp2 (le) mCli-le(/\Sp2, Asp l )"0 (T n s) In stage 2 -le comes to have an association with past time reference, but this has the status of an implicature which can be cancelled given appropriate means (adverbials, etc.)- This association of -le with past time in a way which allows for cancellation can be described somewhat formally as signifying that -le at such a stage is optionally understood as representing past tense. In terms of feature-theory, the 'implicature period' may be suggested to correspond to a stage 2 in which (past) tense features are only optionally added to -le and checked against T°. If synchronically 444 with permission of the copyright owner. Further reproduction prohibited without permission. there is evidence that past tense/past time reference has in fact now become a part of the meaning of -le which cannot be cancelled by adverbials, this can be taken to indicate that past tense features will now always be added to -le for subsequent Ultimately then, the hypothesis that further grammaticalization of -le has led to -le coming to instantiate past tense would seem to have much to support it. First of all, cross-linguistically the re-analysis of resultatives and completives similar to -le into higher functional types encoding past time reference is well-attested, and such a change is easily understood to be the simple result of a pragmatic inference naturally available with such elements becoming standardized as part of the inherent meaning checking against T° as schematically represented in (81) (the 'optional' period Stage 2 simply corresponds to the choice of either the derivation in (80) or (81)): (81) TP Stage 3 AspP I=im/perfective Aspi' AspP 2=completive Asp2' Asp2 °(le). VP V’ ITlC ti-le(Asp2, Aspi, Tns) 445 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of these morphemes over time. Secondly, there is clear synchronic evidence that -le may often only have a past time interpretation and such an interpretation cannot be over-ridden via the use of any adverbs. Thirdly, the hypothesized development of -le from perfective into tense is also highly natural in the modeling of grammaticalization argued for here, where categorial re-analysis results from upward movement within a tree. Lower affixal elements come to be re-analyzed as optionally instantiating higher functional heads when raised up to such positions in larger feature-sets. In the case of -le, it is naturally interpreted as potentially being active in a checking relation not only with the lower heads Asp2° and Aspi° but also with the higher T° head which selects for AspjP. In what follows I re-examine the classic objections to an analysis of -le as tense, and argue that in all cases the objections are either misconceived or alternatively allow for other explanations. 6.2.2 Possible Objections to an Analysis o f -Le as Tense In Li & Thompson's (1981) chapter 6: '-le Does Not Mean Past Tense.' (pp. 213-215) there are four essential arguments against the assumption that -le instantiates past tense. To these can be added two other obvious reasons why one might not initially think of taking -le as a morpheme encoding tense. 446 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Li & Thompson's claim that -le does not mean past tense is based on the following observations.1 3 First of all, it is noted that even if an event is interpreted as having taken place in the past it is often not necessary for -le to occur. If -le were to be a past-tense marker it is therefore inferred that one might expect it to be necessary in all references to events occurring in the past. (82) and (83) are two o f the examples given: (82) wo zao zhidao you yi-dian bu dui. I early know be a-little not right 'I knew a long time ago that something was wrong.' (83) zuotian ye-li wo meng-jian wo muqin. yesterday night-in I dream-meet I mother 'Last night I dreamed about my mother.' Secondly, it is noted that -le need not signal past time in certain subordinate clauses such as (84): (84) wo chi-le fan zai zou. I eat-LE rice then go 'I'll go after I eat.' Thirdly, Li & Thompson point out that -le does not encode any past time reference when it occurs in imperative sentences such as (85): 1 3 Note that the order o f presentation of these points is slightly changed from that in Li & Thompson. This is done simply in order that they can be sequentially addressed as problems in the most natural and logical order. 447 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (85) he-le ta drink-LE it 'Drink it!' Finally, it is shown that there is even one instance in which -le may apparently occur in a matrix clause with a non-past meaning: (86) mingtian wojiu kaichu-le ta tomorrow I then expel-LE him 'I'll expel him tomorrow!' In addition to Li & Thompson's arguments it can be added that there is a widely-held assumption articulated in Bybee et al (1994) that for a morpheme to be considered a marker of past tense it should be fully general in its potential application and combine with verbal elements of all types and aspects. With regard to such a criteria, it can be noted that -le is commonly taken not to be able to combine with a whole range of simple verbs such as shuo 'say,' jueding 'decide,' xiwang 'hope (to),' etc. This may therefore be taken to suggest that it is not in fact a simple marker of past tense: (87) Ta shuo(*-le) Zhangsan mai-le yi-liang xin che le he say LE Zhangsan buy-LE 1 -CL new car LE 'He said that Zhangsan has bought a new car.' (88) Ta jueding(*-le) mai yi-ge-fangzi he decide(-LE) buy 1 -CL-house 'He decided to buy a house.' 448 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Finally, as well as not being able to combine with a certain sub-set of quite regular activity and achievement verbs, -le is also argued (e.g., in Smith 1997 and many other works) to be unavailable as a simple marker of past with all of the stative verb group. It is commonly stated that although -le may in fact occur with stative verbs, this does not result in a simple interpretation of a past event/state but rather in an inchoative interpretation. For example, (89) is not interpreted as 'Zhangsan was ill.' (a simple past interpretation) but as 'Zhangsan became ill.' (inchoative). Such a patterning is again seen as a reason not to consider -le as past tense: (89) Zhangsan bing-le. Zhangsan be-sick-LE 'Zhangsan got sick.' There might therefore seem to be a range of potentially good reasons not to consider verbal -le as tense in fact, despite the evidence and arguments presented in 6.2.1 supporting such an account. In the following I will however argue against these objections. 6.2.3 Objections to the Objections I will now procede to re-examine the above objections one by one and attempt to show that they can in fact be quite satisfactorily answered while still maintaining the suggestion that -le has indeed come to be an instantiation of tense. 449 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 .2.3.1.0 The first of Li & Thompson's objections I will re-consider is their observation that -le often need not occur in situations where there is a past time interpretation, as in (82) and (83) above. It is inferred that if -le were to be a genuine marker o f past tense then it would be expected to have to occur here and in all instances o f past time reference. Such a position would seem to be influenced by considerations of the synchronic state of English and various other languages, where past tense must always be marked in past time situations. From a cross-linguistic and a diachronic point of view however, it is certainly not true that languages always encode tense explicitly wherever this may be possible. Bybee et al (1994) note that one common property of grammaticizing morphemes is "frequency increase," and that from a stage where new functional elements are only used when obligatory in otherwise potentially ambiguous envimoments, they later come to be simply applied wherever their meaning is compatible with the general context. An example they give is that the English simple past tense -ed is (now) used not only where it supplies new information that a situation occurred in the past, but also in contexts where it is very clear from other elements such as adverbs and previous occurrences of past tense that the interpretation must be past time. The use of past tense in Modem English is then frequently rather redundant. However, similar obligatory use of functional morphemes in every compatible situation is by no means fully common. Bybee et al note for example that the past habitual form used to need not be used in every case 450 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. where there is a past habitual interpretation, and that (91) has the same interpretation as (90) even without the overt past habitual form: (90) When she lived with him, she used to sing to him every day. (91) When she lived with him, she sang to him every day. Bybee et al also mention Cheyenne as an example o f a language where none of the tense markers are obligatory (p.98). It is therefore not true that a functional morpheme must appear in every compatible context in order to be classed as a functional element. Such rather redundant overuse is simply a possible stage in the development of grammatical elements, and functional morphemes which are only used when really necessary are by no means any less functional than those which are used wherever possible. Relevant and interesting here is also the diachronic development of the English past tense form in -ed which arguably came into use as an optional encoding of an emphatic past tense form (see Bybee et al p. 150). The original source of the - ed suffix was in fact the Old English free-standing auxiliary dyde (Modem English did). Old English being verb-final the auxiliary frequently occurred following the main verb and dyde over time naturally became the reduced suffixal element -ed, as abstractly diagrammed in (92) (see also section 5.4 in chapter 5 of this dissertation on suffix-creation in SOV languages): (92) John to London walk did. — > John to London walk-ed. 451 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Diachronically then it can be argued that the English past tense morpheme would have originally occurred when the speaker also intended a certain emphasis and would not have occurred in every past time environment. This I believe is quite possibly also true with the use of Chinese -le. Basically it might seem that -le occurs in two particular types of situation. Studies such as Spanos (1979) reported in Smith (1997) have agreed with other investigations that -le is used when the context makes it 'necessary to explicitly state the realization of a given action.' (p.81), and that speakers used -le commonly when past time reference was not clear from the context. Le therefore frequently occurs when past time reference must be made clear and explicit— arguably a disambiguating function. Elsewhere it might appear that -le is often used when the speaker wishes to emphasize an action. In Chang's (1986) study reported in Smith (1997) it is argued that -le occurs predominantly as 'an explicit marker for the peak event in a discourse segment.' (p.265). Smith (1997) also observes that when-type clauses in Chinese such as in (93) function to foreground the event in the main clause and that this usually means that the verb in the main clause needs to carry a viewpoint morpheme such as -le: (93) Ta zai Beijing de shihou xue-le Hanyu he be-in Beijing DE time study-LE Chinese 'When he was in Beijing he learned Chinese.' If the w/zett-clause is the background, then this means that the verb carrying the viewpoint morpheme is highlighted/focused. T'ung & Pollard (1997) additionally note that -le 'is not necessary when describing circumstances (i.e., the background to 452 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. an event) or relating sequences o f events (if it is used, it breaks the sequence into separate steps).' (p. 144). It therefore might seem that speakers use -le when the event depicted requires a certain emphasis, and hence that -le is actually rather similar to (the development of) the English emphatic past dyde/-ed. Largely following the conclusions of other researchers, the distribution of -le can consequently be suggested to be largely dictated by two main factors - -le is used when the context requires disambiguation as past, and -le is also used to mark an emphatic past event.1 4 Given the apparent optionality of -le it is rather natural that the use of overt phonetic material should in fact result in an emphasized interpretation. Similar patterns are elsewhere found where a language has both null pronominals (pro) and overt pronouns; the overt forms are basically only used when there is either ambiguity or in order to emphasize the referent. Elsewhere where no emphasis or disambiguation is required, overt forms are avoided, this being reflected in Chomsky's (1981) suggestion of an Avoid Pronoun Principle. With optional elements such as -le one may assume that a parallel and general Avoid Overt Realization Principle also leads speakers to only use -le when necessary (or for emphasis). Ultimately then, the optionality of -le does not argue against its status as a tense element in any convincing way. There are both other languages in which tense 1 4 It is commonly noted that verbal -le must also appear when the object of the verb is quantified and specific. This can be considered to be a sub-case o f the focusing use of -le. If an object is explicitly quantified in object position, then both the quantified object and the predicate are arguably in focus. 453 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is clearly only given optional phonetic realization and also simple reasons why one can understand the avoidance of overt forms if this is possible in a language. 6.2.3.2 Lack of .general ity;_ with certmn simple verbs An understanding of the potentially emphatic properties of -le now also allows for an explanation of some of the lack of generality noted with -le above. Whereas Li & Thompson point out that -le sometimes need not occur in past time descriptions, it was also noted that -le apparently may not occur with certain simple verbs such as shuo 'say, jueding 'decide,' zhidao 'know' and dasuan 'decide to' as in (87-88) repeated below together with an example with zhidao (94): (87) Ta shuo(*-le) Zhangsan mai-le yi-liang xin che le he say LE Zhangsan buy-LE 1 -CL new car LE 'He said that Zhangsan has bought a new car.' (88) Ta jueding(*-le) mai yi-ge-fangzi he decide(-LE) buy 1 -CL-house 'He decided to buy a house.' (94) Ta zhidao(*-le) Mali yao he Zhangsan jie-hun le he know LE Mali want with Zhangsan marry LE 'He knew that Mali wanted to get married with Zhangsan.' Here I believe there is a very simple explanation of the unacceptability o f these examples. With all o f these rather general verbs it is their following clausal complement which encodes the new and hence focused information in the sentence. This being so it is quite inappropriate for -le to be used on the embedding verb as -le 454 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is essentially licensed as a means to emphasize and focus a verb. Were the matrix verb to carry an emphatic focus with -le this would result in two unconnected foci in the sentence— the first is the emphasis on the act of saying, deciding etc., and the second the natural focus encoded by the new information of the complement clause. The likelihood o f such a mode of explanation being correct can in fact easily be checked. Whereas clausal complements to verbs of saying, knowing and deciding tend to be consistently interpreted as new information, if the complement of these verbs is instead a DP headed by an anaphoric demonstrative such as zhei/nei 'this/that' it is easily interpretable as old information not in focus. Interestingly as soon as the objects of zhidao and jueding etc. are nominal and old information it is found that -le is perfectly acceptable on the verb as shown in (95) and (96): (95) ta huran zhidao-le nei-jian-shi. he suddenly know-LE that-Cl-thing 'Suddenly he knew it/found it out.' (96) nei-jian-shi ta jueding-le hen jiu/liang-tian le that-Cl-thing he decide-LE very Iong/2 day LE 'He decided that very long ago/two days ago.' Furthermore, while shuo 'say' is a very general verb of oral communication and hence a verb which will naturally focus its complement clause as new information, if an adverbial is added to shuo which allows for emphasis on the verb it is found that shuo can occur both with -le and a clausal complement, as in (97) (note that you 455 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 'again' is also used to try to force an interpretation in which the complement clause must be old information and emphasis is on shuo): (97) Ta you da-sheng-de shuo-le ta bu qu he again big-voice-ly say-LE he not go 'He then again shouted that he wasn't going.' Consequently, the apparent 'non-generality' of -le with certain verbs actually can be explained in a rather simple way and cannot be taken to provide evidence against a past tense analysis of -le (on the grounds of lack of expected generality). Once certain discourse factors are controlled for it is found that the verbs in question are in fact able to carry -le signaling past tense and that there is no genuine lack of generality with these predicates. 6.2.3.3 _ . Lack of generality : with, statives The next potential objection to an analysis of -le as tense which I would like to consider here is another very broad case of lack of generality— the apparent inability of -le to occur signaling simple past tense/time with the stative verb group. Commonly it is argued that only a special inchoative interpretation is available when stative verbs combine with -le and not an interpretation corresponding to a simple past usage. As noted above, (89) repeated below is not interpreted as 'Zhangsan was ill.' (a simple past interpretation) but as 'Zhangsan became ill.' (inchoative): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (89) Zhangsan bing-Ie. Zhangsan be-sick-LE 'Zhangsan got sick.' In order to tackle this 'problem,' we need to understand what is meant by inchoativity as a linguistic term and whether referring to a particular interpretation as inchaotive would necessarily classify it as anything significantly different from 'regular' past interpretations. In section 6.2.1 above it has essentially been suggested that because of the way that it has developed from perfective (and earlier completive aspect) marker, verbal -le synchronically has a dual nature and instantiates both past tense and perfective aspect. Such a hypothetical dual specification as past and perfective will below be argued to allow for a natural account of the apparent inchoative restrictions on the interpretation of -le with statives. First of all though it can be noted that despite more traditional morphological views (such as the Item and Arrangement approach) that a single morpheme corresponds to just a single function, there are indeed good reasons to believe that a single morpheme may fulfill multiple functions (as in fact assumed in the Word and Paradigm approach to morphology).1 5 Such dual functionality is also very commonly found in the interaction of tense and aspect functions. Smith (1997) for example notes the case of English used to which encodes both past tense and habitual aspect and can nowadays only occur in the past tense— thus a tense and an aspectual specification are bound up inseparably in a single morpheme: 457 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (98) John used to/*uses to like Mary. The French Imparfait verb form is another case given where there is an inseparable combination of tense and (imperfective) aspect. As shown in (99), French Imparfait suffix, like English "used to," only occurs in the past tense (Smith 1997). To express a present counterpart o f a progressive event, a different strategy is employed as shown in (99b), rather than using an imperfective verb form. (99) a. Jean lisait un roman Jean read.past.3rd sg.Imp a novel 'Jean was reading a novel.' b. Jean est en train de lire. Jean is in train of read 'Jean is (in train of) reading a novel.' Also aspectual restrictions on the use of English simple present tense are noted in Smith. When present tense in English does not occur with imperfective aspect (i.e., be...-ing), it receives an automatic interpretation as habitual aspect, indicating again that tense and a type o f aspectual interpretation are bound up together in a single overt morpheme (as well as 3rd .sing. agreement): (100) John plays tennis. Consequently it is fairly clear that diachronic change may frequently lead to situations where tense and aspectual properties co-occur in a single overt element. 1 5 For both models, see the discussions in Matthews (1974). 458 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The assumption that Chinese -le encodes the dual functions of past tense and perfective aspect (and sometimes also hypothetically completive aspect) now provides the basis for an explanation of the patterning with -le and stative verbs. Comrie (1976) points out that there is in fact a common connection between perfective aspect and inchoativity with stative verbs: 'In many languages that have a distinction between perfective and imperfective, the perfective forms of some verbs, in particular o f some stative verbs, can in fact be used to indicate the beginning of a situation (digressive meaning).' (p. 19). What needs to be asked now is whether inchoativity really is a primitive type of aspect different from simple perfective, or whether (as Comrie hints) that it is possibly quite a predictable bi-product of perfective aspect combining with a certain class of verbs. Supposing that inchoative were to be a primitive different type of viewpoint aspect contrasting with perfective and imperfective, one could possibly attempt to explain the inchoative interpretation of -le with stative verbs in the following way. It could be suggested that because perfective aspect -le originally derives from a predicate meaning 'to finish' it may still retain something of its earlier meaning and still require/select for a complement which is +dynamic. For example, in English the word 'finished,' when used as a perfective marker, still requires an agentive subject and a +dynamic predicate: (101) a. John finished reading. b. *John finished being sleepy. 459 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. It is actually quite common cross-Iinguistically for aspectual verbs derived from predicates having an original interpretation o f the 'to finish' type to show similar restrictions to Chinese -le and give rise to inchoative readings when combined with stative verbs (see here Bybee et al 1994). If such predicates originally had agentive subjects in control of dynamic actions, then it is possible that they would not readily combine with non-dynamic states.1 6 Then a type of aspect different from perfective would have to occur with stative predicates, and inchoative aspect might then be this different type. However, such an approach would seem to require the assumption that -le could represent two types of different aspect contrasting in the viewpoint category, which seems to be somewhat implausible. Furthermore, -le is derived from a sentence-final -liao which takes the whole sentence as its argument, so it is not so obvious that liao originally did have anything like an agentive subject and a necessary dynamic complement, unlike other aspectual elements such as English 'finished' in (101). Smith (1997) seems to suggest that inchoativity is really independent of perfectivity and instead part of situation aspect: 1 6 See here Bybee et al (1994, p.76): 'In the early stages it would not be normal for constructions with 'finish' or anteriors from be or have auxiliaries to be used with stative predicates. They are compatible only with dynamic predicates, and it is the meaning they develop with dynamic predicates that is transferred in their use with stative predicates.' 460 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The perfective is not available to statives in Chinese, Russian and Navajo. These languages have no perfective sentences with stative verb constellations and the interpretation of a basic-level stative situation type. Stative verb constellations do allow the perfective viewpoint when they undergo a shift in situation type. They appear as inchoatives in derived telic sentences. As such they present a change into the state which the verb constellation lexically denotes. In situation type they are either Achievements or Accomplishments, depending on the feature of duration, (p.70) However, this also does not seem to be right for two clear reasons. First of all if stative verb + -le combinations are intransitive telic achievements or accomplishment predicates, they should be expected to exhibit unaccusative syntax. In Chinese achievement/accomplishment verbs are either inherently unaccusative/telic, for example si 'die1 or become telic and unaccusative in virtue of a V 2 telic bound. All such unaccusatives allow indefinite subject inversion, whereas unergative intransitives with simple perfective -le do not. As noted in Sybesma (1999), ku 'cry' in (102) is not one of the 28 verbs which allow -le as a completive V2 and -le must therefore be simply perfective when combining with ku. Perfectivity alone does not license indefinite subject inversion whereas the telicity encoded by a V2 corresponding to Asp2C o m p ie tiv e does (examples (102b) and (103 b) are adapted from Sybesma 1999): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (102) a. dang-shi henduo ren ku-le that-time many people cry-LE 'Many people cried then.1 b. *dang-shi ku-le henduo ren that-time cry-LE many people intended: 'Many people cried then.' (103) a. dang-shi henduo ren ku-lei-le that-time many people cry-tired-LE 'Many people cried themselves tired then.' b. dang-shi ku-lei-ie henduo ren that-time cry-tired-LE many people 'Many people cried themselves tired then.' (104) a. dang-shi henduo ren si-le that time many people die-LE 'At that time many people died.' b. dang-shi si-le henduo ren that time die-LE many people 'At that time many people died.' If -le occurring with a stative verb resulted in a telic achievement or accomplishment, then it is expected that stative verbs and -le should allow for indefinite subjects to invert. This is however not the case: 462 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (105) a. dang-shi henduo ren bing-le that-time many people be-ill-LE intended: 'At that time many people got ill.' b. * dang-shi bing-le henduo ren that-time be-ill-LE many people intended: 'At that time many people got ill.' This would seem to indicate both that -le is not occurring in the V2 position here making the stative verb telic, and that inchoativity does not correspond to any telic shift to an achievement/accomplishment type occurring in the situation aspect. Secondly, if -le occurring with stative verbs were to represent some further type of aspectual primitive either in complementary distribution with simple perfectivity or somehow in the completive V2 position, one would expect that it should be able to occur in a sentence with future reference such as (106), but again this is not at all possible: (106) *wo mingnian yiding hui bing-le san tian I next-year certainly will be-ill-LE three day intended: 'Next year I will certainly be sick for three days.' As shown in the intended gloss, there is a perfectly good anticipated interpretation that should be available if inchoativity were to be an independent type of aspect, and the notion of inchoativity should be in theory compatible with future time reference. The fact that sentences such as (106) are unacceptable can be taken to indicate instead that -le in fact is simply perfective aspect here combining also with past 463 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. tense, and it is the past tense specification which is incompatible with the future oriented adverb. Consequently, if 'inchoative -le' is not taken to be a primitive independent type of aspect contrasting with perfective, the inchoative interpretation still needs some explanation. I would like to suggest that this can in fact be understood quite naturally when one reflects on what happens when a perfective such as -le is applied to stative verbs. Following Smith (1997), the role of perfective viewpoint is taken to be the focusing o f endpoints present in a telic predicate. As statives are atelic and clearly have no obvious endpoints, when perfective is combined with a stative verb it can only signal realization of the state, not completion of the state (which would require a final endpoint). As simple realization logically implies that there must be some beginning point but no explicit end, the realization interpretation comes to be understood as simple inchoativity. An alternative to such an account is to suggest that because perfective -le will always require a predicate which is telic to apply to, the element -le which occurs with stative verbs is actually not perfective verbal -le. Instead it may be suggested that what occurs with stative verbs is actually sentence-final -le. In all of the examples commonly given in the literature -le with intransitive stative verbs comes in sentence-final position and there has not been any attempt to control for whether this might really be sentence -le instead. As sentence -le signals that the general situation of a predicate is ongoing and relevant to the current moment, an interpretation of inchoativity with stative verbs might indeed be rather natural, 464 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. indicating that the state of the predicate has begun and is relevant to the speech time and is not yet closed. In support of such a possibility is the data in (107) and (108) which seem to indicate that when a stative verb is transitive, it cannot in fact occur with verbal -le but instead allows for an interpretation of inchoativity with sentence - le: (107) *xianzai ta zai-le Beijing now he be-in-LE Beijing intended: 'Now he is in Beijing.1 (108) xianzaita zai Beijing le now he be-in Beijing LE *Now he is in Beijing.’ However the inchoative interpretation arises, either as a default via perfective aspect just signaling realization with no indicated endpoint or alternatively through sentence -le signaling open-ended relevance to the current situation, the conclusion is that inchoativity here does not correspond to any primitive alternative aspectual type necessarily excluding perfective aspect. There is also simple but highly significant evidence indicating that verbal -le actually can in fact encode a straightforward interpretation of perfectivity and simple past tense with stative verbs, despite the common assumption that only inchoative interpretations are possible. Compare (89) to (109). A durative time phrase is added on to the combination of a stative verb and -le in (109), whose word order makes it clear that this is a verbal -le and not a sentence-final -le. The resulting interpretation is of simple past time: 465 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (89) Zhangsan bing-le. (inchoative) Zhangsan be-sick-LE 'Zhangsan got sick.' (109) ta (zai Beijing de shihou) bing-le san tian he (in Beijing DE time) be-ill-LE 3 day 'He was sick for three days (when he was in Beijing).' It clearly shows that stative verbs do indeed combine quite regularly with verbal -le to give a basic past time perfective reading, despite the frequent denial that this is possible. Above it was suggested that stative verbs might perhaps not easily combine with perfectivity because the perfective's function is to focus the endpoints of a situation and stative predicates have no inherent endpoints. Functionally what the addition of the duration phrase to a stative verb + -le does is simply to provide an explicit temporal bound to the predicate. Once this is added in to the structure, the combination of stative verb and -le is perfectly well-formed with a past perfective interpretation just like other predicate types.1 7 What is also interesting is that the 'inchoative' interpretation supposed to result from the particular combination of verbal -le and stative verbs is actually not restricted just to this class of predicate and also arguably occurs with activity verbs when they occur without a telic bound. The interpretation of (110) and (111) with an activity verb occurring alone with -le, 1 7 Note also that it is not possible to suggest that this is somehow a 'disguised' inchoative reading. The interpretation is of a simple extension in the past just as with other predicate types with an extension over time in the past. Furthermore inchoativity is essentially instantaneous and picks out a single time point like an achievement, the coming into being of some state. Consequently, as with instantaneous achievements, inchoative interpretations do not combine well with durative adverbials: (i) */??John has become/started to be sick for three days. 466 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. paralleling the common examples of a stative verb alone with -le (as in (89)), would also seem to have primarily an inchoative reading signaling an entry into the activity:1 8 (110) Ta ku-le he cry-LE 'He started to cry.' (111) Tashui-le he sleep-LE 'He started to sleep (he has gone to bed).' Again, once some kind of telic bound is added in, the interpretation is of simple past perfective: (112) Ta pao-le/shui-le liang-ge-zhongtou he run-LE/sleep-LE 2-Cl-hour 'He ran/slept for two hours.' The ultimate conclusion concerning stative verbs and -le is therefore that there is no fundamental difference between stative verbs and other predicates in the possibility 1 8 Note that informants indicate that the most salient interpretation in the combination of past perfective and activity verbs which are not explicitly bounded in English is also that of inchoativity and entry into the activity, as e.g. in: (i) John ran. (ii) Mary ate. This inchaotive interpretation is very clear and almost forced when a single time point is added in as in (iii) and (iv): (iii) When I looked at him, John ran. (iv) Mary ate at two/Mrs. Smith's command. Although there would seem to be inchoative interpretations here, it is not obvious that one would want to say that the occurrence o f past perfective is any different here from that in non-inchoative readings. 467 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of being combined with past perfective -le, and that any inchoative reading is simply due to the lack of any telic bound, statives behaving here just like activity verbs which otherwise do combine with past perfective -le. Consequently, as in the previous cases considered (i.e., the potential combination of shuo, jueding, zhidao etc. with -le), careful investigation shows that ultimately there is no necessary restriction on the generality o f past perfective -le, and that potential lack of generality can therefore not be used as an objection to the analysis of -le as instantiating past. 6-.2-.3.4 _ . Subordinate, clauses I will now move on to consider the three other remaining objections to a past tense analysis o f -le raised by Li & Thompson. The first of these was that verbal -le regularly occurs embedded in examples such as (84) without necessarily signaling past time: (84) wo chi-le fan zai zou. I eat-LE rice then go 'I'll go after I eat.' If -le here may occur without having a past time interpretation this might seem to falsify the contention that verbal -le has come to instantiate past tense and suggest that such a claim should be abandoned. However, given the constant past time interpretation arising with -le in environments other than this subordinate clause case and the two other specific contexts to be re-considered below (imperatives and main 468 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. clauses with jiu 'then') I believe it is worth attempting to see if such cases might perhaps have some alternate explanation. I believe that a possible solution to the problem involves focusing on two particular properties o f this non-past usage of -le. The first is that this non-past use of -le is confined to a certain type of subordinate clause/constituent. The second is the general suggestion that -le is a morpheme undergoing diachronic change and re analysis, i.e., it is claimed to have developed from a completive to a perfective marker, and now to be an instantiation of past tense. Now, much diachronic research has shown that morpho-syntactic change and re-analysis is often not effected at the same time in all potentially available environments and that changes very frequently occur in main/matrix clauses before they later spread to other subordinate contexts. Harris & Campbell (1995) attribute this 'discovery' originally to Biener (1922a/b) and note that it is an assumption both widely held among historical linguists and substantiated by much research work. Harris & Campbell note that it is believed: subordinate clauses are less subject to syntactic change than are main clauses because they exhibit a more restricted range of morpho- syntactic trappings due to their backgrounding function in discourse (Givon 1971, 1984; Hopper and Thompson 1984). The general idea involved is the belief that change starts in main clauses and may or may not ultimately come to affect subordinate clauses, but that it does not begin in subordinate clauses, later reaching main clauses, (p.27) In a footnote (footnote 8, p.382) it is also added that: '. . . in general, subordinate clauses do contain fewer morphosyntactic contrasts than main clauses.' Making use of this general insight that morpho-syntactic changes occur first in main clauses and then only later spread to subordinate clauses— or possibly even remain confined to 469 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. main clauses and do not get incorporated into subordinate environments— it could here be quite reasonably suggested that such a main clause/subordinate clause developmental distinction is behind the non-past interpretation o f -le in (84). Having argued that -le has undergone re-analysis into an instantiation o f past tense as well as perfective, it could now be suggested that while this change has indeed occurred in regular main clauses and even many subordinate clause environments, it has not yet spread to the subordinate clause type found in examples such as (84), and consequently -le simply instantiates perfective in such an environment. Such a suggestion would be fully in line with the view mentioned above that morpho- syntactic changes may only gradually spread from matrix clauses to other subordinate environments. It would also be supported by the assumption made here that -le instantiates perfective as well as past in most environments— the particular subordinate context found in (84) would simply be a case where -le has re-analyzed as far as being perfective but not undergone the further change to be interpreted as past tense as well (i.e., has not reached Stage 3 of the process outlined in (81)). One particularly relevant example of a morpho-syntactic change which occurred in main clause contexts but not in all subordinate clauses is the diachronic re-analysis of the classical Japanese aspect system into tense. From an earlier stage in which Japanese is taken to have had only a contrastive system of aspect in both main and embedded clauses there arose the Modem Japanese system of tense, created when various of the early aspect markers became re-interpreted as tense elements (see e.g. Takeuchi 1998). What is interesting to note is that while the element -ta has been 470 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. grammaticalized as clearly instantiating past tense in most main and embedded clause environments, in relative clauses it is still possible for -ta to occur with just perfective aspect interpretation and actually refer to a future context. Example (113) is from Nakamura (1994): (113) [ashita ichiban hayaku kital hito-ni kore-o ageru tomorrow most early came person-Dat this-Acc give 'I will give this to the person who comes (lit. came) first tomorrow.' In Japanese it is fully accepted that -ta is a past tense marker, but there remains this one subordinate context in which it can still be interpreted as perfective. Such a situation is precisely what may be argued to be found also in Chinese— although there has been a general re-analysis of -le as both past and perfective, in the particular subordinate clause case (84) this change has not yet occurred and -le remains just perfective. A developmental approach such as that outlined above therefore allows one to maintain the claim that -le has indeed essentially undergone re-analysis as past as claimed and is also supported by the occurrence of similar historical changes in other languages. A possible formal interpretation of such a hypothesis is to suggest that -le is not necessarily interpreted as instantiating past tense in (84) because there actually is no legitimate tense position in the subordinate clause in (84) which could license -le as past. It can be suggested that the subordinate clause containing chi-le 'eat-LE' is either just an AspP rather than a TP, or that it is a necessarily non-finite TP. In either case, because no potentially finite T° occurs, -le will not be interpreted as (past) tense 471 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. but instead be licensed as simply perfective aspect (i.e., -le will only be necessarily interpreted as past when there is the opportunity for it to be licensed as tense in a finite T° position). In support o f the assumption that the subordinate clause in forms such as (84) is essentially without any tense specification, one can note that the temporal orientation of this clause is completely determined by the time/tense interpretation of the main clause predicate. Thus in (114) where the matrix verb zou 'to leave1 in combination with hui 'will' necessarily refers to a point in the future, then the subordinate clause action must also be understood as occurring in the future. If, however, the matrix verb zou 'to leave' is understood to refer to a past time as in (115), then the verb in the subordinate clause chi(-le) 'eat-LE' must also be understood as having past time reference. (114) wo chi-le fan jiu hui zou. I eat-LE meal then will leave 'I'll leave after I've eaten.' (115) wo chi-le fan jiu zou-le. I eat-LE meal then leave-LE 'I left right after I ate.' Structures of (84)/(114) and (115) therefore have the fully dependent status of English subordinate -ing clauses such as (116) and (117) and are best translated into English with clauses of this type: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (116) Having eaten, I will leave. (117) Having eaten, I left. In (116/117) just as in (84) the action of the subordinate clause must be interpreted as occurring at the same past or future time as the action of the main clause. Such a dependency can be suggested to result either from the complete absence of a T°/TP in the subordinate clause in (84) (and therefore no possibility of an independent time specification), or perhaps from a T° head being present but obligatorily filled by some [-finite] specification essentially controlled by the main clause tense specification (i.e., a PRO-like non-finite tense, as e.g. suggested in Stowell 1996). Consequently the assumption that the subordinate clause in structures such as (84) does not contain a +finite T° allows for a straightforward account of why -le may occur without necessarily causing a past time interpretation. I have suggested here that -le obligatorily instantiates +past only when there is a potentially +finite T° in a clause and it is actually possible for -le to be interpreted as past. If a clause either does not contain any T° position at all, or this T° can only be interpreted as - finite and necessarily dependent/controlled by a higher clause tense, then there simply is no possibility for -le to be interpreted as past and it will only be licensed as perfective aspect. It will therefore only be where there is a genuine +finite T° present in the structure that the possibility for -le to be interpreted as past actually arises and is now made obligatory. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Note that an important component of such an explanation of (84) is the suggestion made earlier in section 6.1.3 that overt lexical morphemes such as -le are essentially just lexical hosts for the semantic features which really encode meaning in any structure. It was suggested that semantic/functional features are critically added to elements such as -le in the lexicon and that the lexical host and the features which it carries are subsequently inserted into syntactic structure together. Such a view essentially assumes that overt lexical morphemes may not inherently instantiate a meaning (such as +past for example), but instead acquire this meaning via a compositional process in the lexicon where functional/semantic features are added to the morpheme, and the morpheme is effectively just a physical carrier/host for the relevant features. Most frequently this process of combining features and a particular host will be almost automatic as certain morphemes are understood as having fairly constant meanings. However, during the course of language change there may be periods in which a certain set of features is only optionally added to a particular morpheme in the lexicon and this morpheme may consequently only optionally be understood to instantiate a certain meaning. Here in the case of -le it has been argued that -le has come to instantiate not only perfective aspect (and possibly sometimes completive aspect) but also past tense. What this is taken to mean in formal terms is that real functional-semantic features encoding +perfective and +past are regularly combined with -le in the lexicon resulting in its constant +perfective +past interpretation, and the morpheme -le has been constantly been described as 'instantiating' past/perfective aspect. In such an approach in which 474 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. functional-semantic features and their physical hosts/carriers are combined together before insertion into syntactic structure and the functional-semantic features are assumed not to be a fully inherent part of the lexical host, it can be argued that the relevant features are combined with their natural specified hosts only wherever this is possible. In the case o f -le this means that +past features will be combined with -le only where there is a possibility for -le to instantiate a past tense meaning, hence only where there is a genuine +flnite T° also present in the sub-part o f the numeration linked to a particular structure (e.g., a subordinate clause). Such a view essentially sees overt lexical morphemes as carriers o f certain meanings rather than actually inherently communicating those meanings. Naturally in most cases there will be a very close even automatic association of a morpheme with a certain meaning, but the approach also allows for change to occur and for a morpheme to only optionally instantiate a certain meaning when it is used. In support of such a general view and as a further clear example of the active combinatorial process suggested to take place between lexical hosts and feature-sets, I would like to mention briefly the interesting case of yes/no question morphemes in Egyptian Arabic. As noted in Wahba (1984) and Demirdache (1991) in Egyptian Arabic yes/no questions are signaled by the occurrence of a pronoun in sentence-initial Comp position as seen in (118a/b) and (119a/b) taken from Wahba:1 9 1 9 The same strategy also occurs with w/j-questions. 1 only include examples of yes/no questions as these are enough to illustrate the point to be made. 475 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (118) a. Mona ablit il-talamiiz Mona met the-students 'Mona met the students.' b. hiyya Mona ablit il-talamiiz ? she Mona met the-students 'Did Mona meet the students?' (119) a. il-talamiiz ablu Mona the-students met Mona 'The students met Mona.' b. humma il-talamiiz ablu Mona? they the-students met Mona 'Did the students meet Mona.' As these pronouns clearly specify the questionhood of the sentences they occur in, it can be assumed that in questions they carry a +Q feature specification into Comp (otherwise Comp/C° would not be interpreted as +interrogative). Significantly in other non-question contexts such elements occur fully regularly as simple non-clitic pronouns and do not give rise to any +interrogative interpretation. What this consequently indicates rather straightforwardly is that in questions some additional +Q feature specification is added to the pronouns as lexical hosts before they are inserted into the syntactic structure, and the pronouns simply serve as specified hosts for the +Q feature set on certain (+interrogative) occasions. Furthermore it can be noted that this combination of +Q features with a pronoun in questions must be a productive and active process. Supposing there were to be just a single pronoun type used to signal a yes/no question, it could be argued that such an element might have 4 7 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. been grammaticalized with +Q features as a distinct entry in the lexicon. However, from (118b) and (119b) it can be seen that the pronoun occurring as the +Q morpheme actually varies according to the subject of the sentence. If the subject is feminine singular, the pronoun used will be hiyya 'she' (example (118b), whereas if the subject is masculine plural as in (119b) this will trigger a different pronoun question marker humma 'they.' This suggests then that there is indeed an active combinatorial process in the lexicon precisely as suggested, and certain specified lexical hosts may have additional functional-semantic features actively added to them when this is required to encode a particular interpretation. In Egyptian Arabic +Q features are optionally added to a pronoun in the lexicon when a +interrogative interpretation is required. Turning back to -le then, it can be argued again that past tense features are formally distinct from -le and not an inherent encoding but combined with -le as a physical host wherever this is possible. In a numeration in which there is no +Finite T° which could license past tense features carried by -le, the past tense features will simply not be added on to -le in the lexicon and -le will just instantiate perfective aspect. Ultimately then, careful reflection and re-consideration does indeed allow for a principled account of the non-past interpretation of -le in the subordinate clause of cases such as (84), and such an account is significantly fully compatible with the basic assertion that -le is in other instances a genuine instantiation of past tense. A ll Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. £.2.3.5.. Imperatives The next o f Li & Thompson's objections to an analysis o f -le as tense to be briefly re considered here is the occurrence of -le in imperative sentences such as (85). As there is no past time interpretation in (85) it is suggested that -le cannot be past tense: (85) he-le ta drink-LE it 'Drink it!' Concerning such cases, I will adopt the basic approach outlined for the subordinate clause case above and suggest that -le cannot be licensed as past tense here in (85) because imperative sentences are reduced clausal structures and there simply is no +Finite T° present in the structure to license -le as past tense. In fact, it can be suggested that there is also no (Im)perfective aspect projection present in imperative sentences. Closer investigation of the occurrence of -le in commands reveals that verbal -le can only combine with Lii's (1980) 28 verbs (see footnote 8). These 28 verbs are argued previously to be the only verbs that can combine with -le to give rise to a purely resultative (completive) interpretation. This suggests that the verbal le which occurs in imperative sentences is in fact only the instantiation of inner completive aspect and not outer perfective aspect, as otherwise -le should be able to occur with the whole range o f verbs which elsewhere allow for perfective -le. As examples (120) and (121) show, this is however not possible. 478 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (120) *qu-le Beijing go-LE Beijing 'Go to Beijing.' (121) *xue-le zhei-ge zi study-LE this-CL character 'Learn this character!' Consequently, if no T° node and no Aspviewpoint node occurs in imperative structures, past tense features will not be added to -le when taken from the lexicon, and -le will not be interpreted as past tense (or perfective aspect). Note that such an assertion can be suggested to be generally plausible, as cross-linguistically tense does not seem to occur in any imperative sentences. 6.2,3.6 Matrixjiu sentences The final case noted by Li & Thompson where -le occurs without necessarily resulting in a past time interpretation is (86) repeated below: (86) mingtian wo jiu kaichu-le ta tomorrow I then expel-LE him 'I'll expel him tomorrow!' Whereas with other cases of matrix clause use -le seems to automatically result in a past time interpretation, here (86) naturally refers to a future event and so might seem to falsify the hypothesis that -le encodes +past wherever this is possible. As (86) is obviously a non-imperative matrix clause it must be assumed that a +Finite T° 479 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. is indeed present in the structure and it is not possible to suggest that any subordinate or reduced clausal status is responsible for the possible non-past interpretation. (86) would therefore seem to constitute a serious potential counter-example to the -le-as- past hypothesis. Considering rather carefully what allows for the non-past interpretation in (86), I believe that a plausible and interesting explanation can in fact be offered for this othenvise exceptional non-past occurrence of -le in a non-subordinate clause. Essentially it is the critical addition of the element jiu 'then' in (86) which facilitates the non-past interpretation, as a future-oriented adverb such as mingtian is otherwise not sufficient to make available a future time interpretation, as was shown earlier in (75): (75) *wo mingtian (hui) zai Beijing canguan-Ie Gugong I tomorrow (will) in Beijing visit-LE Imperial Palace intended: 'Tomorrow I will visit/will have visited the Imperial Palace.' There is consequently something in the element jiu which is highly relevant here. Focusing on the importance of jiu in allowing for the non-past interpretation, I believe that an account of the patterning can now be given which significantly supports the central contention that -le both commonly instantiates tense and that such an interpretation is formally licensed via LF movement to a higher T° position. Quite simply I would like to suggest that (86) is very much like a covert instance of the db-support phenomena which occurs in English in the presence of negation, here the blocking element beingy'/w rather than negation. 480 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In English it is currently assumed (e.g., Chomsky 1993, 1995) that finite verbs are base-generated together with a tense specification/suffix and that tense features are licensed/checked only at LF when the verbal complex/its feature-set raises up to T° . It is also well-observed that sentential negation may not combine with a finite verb as illustrated in (122), and that db-support is necessary to save such a combination, as shown in (123): (122) * John not walked home/* John walked not home. (123) John did not walk home. Structures such as (122) are assumed to be ungrammatical because the presence of negation as an X°-head intervening between T° and V° prevents the tensed verb/its tense features from raising up to be checked in T° at LF, such a movement hypothetically violating the Head Movement Constraint. In such instances the tense features are therefore base-generated directly on the dummy element do in T° and LF movement occurs. Turning to examples such as (86), the same mode of explanation can be offered for the fact that the presence of -le does not necessarily give rise to an interpretation o f past tense in such cases. Remember that it has been argued that past tense features are added to -le wherever possible and licensed/checked against a +finite T° via LF raising. If it is now suggested that the element jiu is a head element occurring between T° and V°, it is expected that jiu might indeed block the LF raising o f the verb and -le to tense. I would like to suggest that this is indeed what 481 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. happens in (86) and that when -le cannot raise up to T° a (phonologically covert) tense specification is independently base-generated in T° precisely as in cases of English i/o-support in the presence of negation.2 0 Because such a tense specification will not be licensed by -le, it is therefore possible for this tense to be either past or non-past, accounting for the possible future orientation found in examples such as (86). Le itself, when it cannot be licensed as past tense, will then only have perfective features added to it prior to insertion and raise up to AspiP (jiu can therefore be assumed to be higher than AspjP but lower than TP). In this sense -le will be effectively used and interpreted in the same way as in cases such as (84), i.e., just as perfective.2 1 In further support of such a possibility, there is also interesting evidence from tone sandhi patterns in Taiwanese showing that the direct equivalent morpheme to Mandarin jiu syntactically is indeed an X°-head element rather than an XP in specifier position. As argued in chapter three, X°-heads which followed by tonic 2 0 I do not attempt to go into the challenging question of why the base-generation of features directly in T° is not always/elsewhere used and only occurs as a means to save a derivation which would otherwise crash. I will simply assume that whatever explanation allows for such a possibility to occur in English with t/o-support may also apply in Chinese with a covert counter-part to do. 2 1 Such an approach to -le leads one to expect that it should be possible for -le to occur as simple perfective aspect if another element instantiates the T° position. This prediction is borne out and one finds that -le may indeed co-occur with a future modal such as hui 'will' or yao 'will/want to' in T°, as in (i): (i) women mingtian hui/yao kaichu le ta. we tomorrow will fire LE him 'We will fire him tomorrow.' One also further expects that other heads lower than tense and higher than Aspv,C w p o im may block the possibility for -le to be licensed as past tense and therefore allow for -le to be interpreted as just perfective aspect. As example (ii) below shows, the head ye 'also' patterns as expected and allows for -le to legitimately occur as simple perfective aspect in non-past interpretations: 482 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. elements undergo tone sandhi and elements in specifier positions do not exhibit any tone changes. Regarding the element toh (Mandarin jiu), critically this element does undergo tone sandhi clearly suggesting that it is a syntactic head rather than a specifier.2 2 This is shown in (124) where a bolded dot following a syllable indicates that the syllable undergoes tone sandhi: (124) Goa bin»-a»-chai toh# khi« chhoe* lin* lau»su. I tomorrow then go look-for your teacher 'I'll go see your teacher tomorrow.' Such a patterning would consequently seem to add good support to the hypothesis that Mandarin jiu may well block LF verb-raising due to having an X°-head status. Example in (124) also has further potentially revealing information. Earlier it has been noted that simple adverbs such as mingtian 'tomorrow' may not combine with -le on the verb, as seen in (125): (ii) women mingtian ye kaichu le ta! we tomorrow also fire LE him 'We'll also fire him tomorrow.’ 2 2 If the analysis in Taiwanese tone sandhi proposed in chapter 3 is on the right track, simple adverbs in Taiwanese will have to be assumed to be in the head position. This is different from Cinque (1999), in which the adverbs are Specifiers of related functional heads. Since Specifiers are maximal projections and some adverbs are consistently singe-worded and can not be further modified, I do not see any good reason that adverbs should be all assumed to be maximal projections (which are potentially consisted of specifiers and complements) in Specifier position. I believe certain adverbs are projected directly in the head position, although the others are Specifiers (and probably some of them could even have either option). I regret that I can not be clearer about this. It should be noted that if the tone sandhi phenomenon in Taiwanese can be taken to be structure-sensitive as proposed, probably it can be a test for distinguishing X°-adverbs and XP-adverbs. 483 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (125) * Zhangsan mingtian qu-le Beijing Zhangsan tomorrow go-LE Beijing intended: 'Zhangsan will go to Beijing tomorrow.' The simple explanation for this was that -le instantiates past tense here and this is incompatible with the future time phrase mingtian. Now, supposing that mingtian were to be an X°-head element similar to jiu and that mingtian could occur located structurally lower than T°, one might expect that it would also block LF raising of the verb and therefore allow for -le to occur licensed as simply perfective in the same way that jiu does. This would then in theory allow for tense features to be base generated on a covert <sfo-equi valent in T° and (125) might be expected to be acceptable with a future time orientation. Although it is not fully clear where mingtian might be base-generated, it certainly is possible for mingtian to occur linearly following hui 'will,' which itself might indicate the location o f T°, as in (126): (126) wo hui mingtian qu Beijing I will tomorrow go Beijing 'I will go to Beijing tomorrow.' If it is therefore perhaps possible for mingtian 'tomorrow' to be base-generated lower than T° one might anticipate that sentences such as (125) should be acceptable with a future time reference, as suggested. However, interestingly the example (124) shows that phrases such as 'tomorrow' in Taiwanese (bin-a-chai) behave as specifiers rather than heads and their final syllable does not undergo tone sandhi in sharp contrast to 484 R e p ro d u c e d with permission of the copyright owner. Further reproduction prohibited without permission. toh, the equivalent to jiu.2 3 If it is not unreasonable to imagine that the phrase signaling 'tomorrow' in Mandarin and Taiwanese might have a similar syntactic status, then this automatically allows for an account of why jiu but not mingtian allows -le to co-occur with it and why it may have an exceptionally non-past interpretation here, a patterning which otherwise remains quite mysterious. Suggesting then that a r/o-support type treatment of -le in sentences with a matrix jiu is indeed able to provide an account of examples such as (86) which is also fully in line with the general analysis of -le and its licensing, ultimately it turns out that all of the objections to an analysis of -le as instantiating past tense have now been successfully answered. Consequently it does in fact seem possible to maintain the central contention of section 6.2.1 and conclude that the morpheme -le has developed from a marker of completive aspect not only into perfective aspect but subsequently has also come to instantiate past tense wherever such an interpretation is possible in a structure. Although such a suggestion might initially have seemed somewhat speculative, close examination of the arguments against a tense analysis of -le and a consideration of the environments in which -le seems to occur without a 2 3 The equivalents of time phrases such as 'tomorrow,' 'today,' 'yesterday,' etc. in Chinese can be easily taken to be maximal projections (most likely nominal projections), and probably they can occur in specifier position. For example, they can occur in subject and object position, as shown in (i): (i) a. mingtian yiding shi meihaode yi-tian. Tomorrow certainly be wonderful one-tay. 'Tomorrow must be a wonderful day.' b. buyao huinian zuotian. yao ba xiwang fang-zai iintian. Don't commemorate yesterday should BA hope put-at today. 'Don't just commemorate yesterday; you should place your hope in today.' 485 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. past interpretation has actually led to a confirmation of the -/e-as-tense hypothesis. Such a hypothesis has however been shown to be more complex than just a simple equation o f -le with tense and it has been suggested that -le is a morpheme which may indeed encode more than just a single interpretation. In the course of the investigation and in order to provide a principled account of the full complexities of the patterning, the analysis has arrived at a number o f interesting conclusions. One of these is the important assumption that lexical morphemes act as simple physical hosts for functional-semantic features and that lexical morphemes are actively combined with such features in the lexicon. While this combination operation may quite frequently be semi-automatic, critically the assumption that features and morphemes are formally distinct objects allowed for the suggestion that in periods of diachronic change the combination operation may be either optional or even blocked by other interfering factors. Such a possibility then permitted a fully explicit account of a wide range of non-random constraints on the distribution and interpretation of - le. One part of this included an explanation of the interaction of the adverbial element jiu with -le, and it was seen that a plausible comparison with r/o-support and negation quite naturally led to an analysis which seems to fully support the assumption that -le and other suffixes are indeed licensed via covert LF raising to a relevant licensing head. Finally, it can be noted that evidence from Egyptian Arabic presented in (118)-(l 19) has added strong support for the idea that the combination Although it is not clear where mingtian 'tomorrow' is generated in (125), the contrast between time phrases such as mingtian 'tomorrow' and short adverbs such as jiu 'then' is clear. Time phrases are maximal projections and simple adverbs (most likely) are X°-elements, and thus the former does not exhibit tone change in the final syllable while the latter undergo tone change (in the final syllable). 486 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of functional features and lexical host material is indeed an active and productive process. Before closing here it can be pointed out that this has a significant wider relevance. In chapter 5 it was argued that lexical V 2 elements are combined with Vi main verbs as aspectual suffixes, and while there were many good arguments in favor of such an approach, it still might be objected that potential V 2 elements such as ganjing 'clean' would seem to be lexical descriptive morphemes rather than grammaticized functional suffixes. The idea supported by the Egyptian Arabic patterns that functional features are added to lexical hosts converting them into primarily functional types is now able to make better intuitive sense of the claim that V2 elements are used as functional suffixes. Just as pronouns in Egyptian Arabic occur as host elements for +Q features and so take on the role of question-markers, it can be argued that lexical V2 elements with clear descriptive properties are actively combined with completive aspect features in the lexicon and that it is this productive combination process which then allows the range of V2 elements to occur licensed as functional suffixes. 6.3 Summary I close the chapter here with a short summary o f the main points argued for. The chapter began as an attempt to understand the origin and development of verbal -le. Largely because of the completive meaning of liao/le and its re-positioning adjacent to the verb at a time when other V2 elements also became right-adjacent to the verb, it was concluded that liao/le initially instantiated completive aspect. It was then 487 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. noted that -le often appears to have a different function in Modem Chinese and occurs together with other overt V2-completive elements. In order to account for this, I adopted Smith's (1997) two-tiered approach to aspect and suggested that since its initial grammaticalization, -le has developed from being a completive aspect marker to instantiate structurally higher perfective viewpoint aspect, a view which accords with common synchronic perceptions of -le. I then attempted to provide a formal modeling of this hypothetical change of -le from completive aspect to perfective aspect and argued for a development of the approach to grammaticalization introduced in Simpson (1998) and assumed in chapter 2 for ge. Whereas this approach was initially intended to account for the re-analysis of free standing morphemes, the current chapter has argued that suffixes are significantly also subject to highly similar (further) grammaticalization and re-analysis, and that the model of grammaticalization initiated for free morphemes can be naturally developed to account for this. Importantly it was argued that grammaticalization and re-analysis in both free and bound morpheme cases is critically movement- dependent and results from raising of an element upwards in the functional structure dominating a lexical projection to successively highly positions. In the case of the hypothetical re-analysis of -le this was argued to have the further significance that it strongly supports the Minimalist view that functional affixes are base-generated together with their lexical host and then licensed via movement to a functional head, such movement furthermore often being covert and taking place at LF (as with -le). Following this, in section 6.2.1, I attempted to account for other aspects of the 488 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. interpretation of -le and made the somewhat contentious speculation that -le has now developed further to instantiate past tense as well as perfective aspect. While such a possibility has been explicitly rejected in the literature, it was argued that this is largely due to the rather narrow view that overt morphemes must stand in a fully rigid one-to-one correspondence relation with a single meaning/function (such as past). Here it was suggested instead that functional meanings/interpretations are associated with morphemes as the result of an active combination process linking functional features and lexical hosts. In such a process it is possible both for a single morpheme-host to be combined with more than a single functional interpretation/feature (and therefore instantiate multiple functions), and it is also possible for the association of a particular function with a particular morpheme to be either optional or blocked in certain circumstances. Such assumptions and a careful re-consideration of the range of objections to an analysis of -le as tense then ultimately allowed for a full account of the distribution and interpretation of verbal - le and a principled defense of the hypothesis that -le instantiates tense. The chapter concludes that this past tense interpretation is significantly not random but actually determined by clear syntactic factors. 489 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. B ib l io g r a p h y Abney, Steven P. 1987. The English noun phrase in its sentential aspect. Doctoral dissertation, MIT. Abraham, Werner. 1997. The interdependence of case, aspect and referentiality in the history of German: the case of the verbal genitive. In Parameters o f Morphosyntactic Change, ed. Ans van Kemenade and Nigel Vincent, 29-61. Cambridge: Cambridge University Press. Allen, R. L. 1966. The verb system of present-day American English. The Hague: Mouton. Baker, C. L. 1970. Notes on description of English questions: the role of an abstract question morpheme. Foundations o f Language 6: 197-219. Baker, Mark. 1985. The mirror principle and morpho-syntactic explanation. Linguistic Inquiry 16: 373-415. Baker, Mark & Kenneth Hale. 1990. Relativized minimality and pronoun incorporation. Linguistic Inquiry 21: 289-297. Biener, Clemens. 1922a. Zur Methode der Untersuchungen tiber deutsche Wortstellung [On the methods of inverstigation of word placement in German]. Zeitschrift fur deutsches Altertum und deutsche Literatur [Journal o f German Antiquity and German Literature] 59: 127-44. Biener, Clemens. 1922b. Wie ist die neuhochdeutsche Regel iiber die Stellung des Verbums entstanden [How did the new high German rule of verb placement arise]? Zeitschrift fur deutsches Altertum und deutsche Literatur [Journal o f German Antiquity and German Literature] 59: 165-79 Boskovic, Zeljko. 1999. What is special about multiple wh-fronting? Paper presented at NELS 1999, Rutgers University. Borer, Hagit. 1994. The projection of arguments. In University o f Massachusetts occasional papers in linguistics 17, eds. E. Benedicto and J. Runner, 19-48. Bresnan, Joan. 1971. Sentence stress and syntactic transformations. Language XLVII.2: 257-281. Reprinted in Syntactic argumentation, eds. D. J. Napoli & E. Rando (1979), 233-257. Georgetown University Press. Bresnan, Judith and Samuel Mchombo. 1987. Topic, pronoun, and agreement in Chichewa. Language 63:741 -782. 4 9 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Brody, Michael. 1994. Lexico-logicalform. Cambridge, Mass.: MIT Press. Bybee, Joan, Perkins Revere and William Pagliuca. 1994. The evolution o f grammar. Chicago: University of Chicago Press. Chang, Vincent. 1986. The particle le in Chinese narrative discourse: an integrative description. Doctoral dissertation, University o f Florida. Cao, Guangshun. 1987. Yuqici LE yuanliu qianshuo [Preliminary disscussion on the origin of the particle LE], Yuwen Yanjiu 2: 10-15. Chao, Y.-R. 1968. A grammar o f spoken Chinese. Berkeley: University of California Press. Chen, Matthew. 1985. The syntax of Xiamen tone sandhi. Phonology Yearbook 2, eds. Colin Ewen and John Anderson, 109-150. New York/Cambridge: Cambridge University Press. Chen, Matthew. 1990. What must phonology know about syntax? In The phonology-syntax connection, eds. Sharon Inkelas and Draga Zee, 19-46. Chicago: University of Chicago Press. Chen, Matthew. 2000. Tone sandhi. New York/Cambridge: Cambridge University Press. Cheng, Lisa Lai-Shen and Rint Sybesma. 1998. Bare and not-so-bare nouns and the structure of NP. Ms., UCI/HIL/Leiden (to appear in Linguistic Inquiry). Cheng, R. (1968) Tone sandhi in Taiwanese. Linguistics 41:19-42. Cheng, R. (1973) Some notes on tone sandhi in Taiwanese. Linguistics 100: 5-25. Chiu, Bonnie. 1993. The inflectional structure of Mandarin Chinese. Doctoral dissertation, UCLA. Chiu, Bonnie. 1995. An object clitic projection in Mandarin Chinese. Journal o f East Asian Linguistics 4.2: 77-117. Chiu, Bonnie. 1998. Relative clauses in child Chinese. Ms., National Taiwan University. Chomsky, Noam. 1977. On wh-movement. In Formal syntax, eds. P. Culicover, T. Wasow and A. Akmajian, 71-132. New York: Academic Press. Chomsky, Noam. 1981. Lectures on government and binding: the Pisa lectures. Dordrecht: Foris. 491 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chomsky, Noam. 1985. Knowledge o f language: its nature, origin and use. New York: Praeger. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The view from building 20: essays in linguistics in honor ofSylvain Bromberger, 1-52. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1995a. Bare Phrase Structure. In Government & Binding Theory and the Minimalist Program, ed. Gert Webelhuth, 383-439. Cambridge, Mass.: Blackwell. Chomsky, Noam. 1995b. The minimalist program. Cambridge, Mass.: MIT Press. Chomsky, Noam. 1998. Minimalist inquiries: the framework. Ms., MIT. Cinque, Guglielmo 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford University Press. Collins, Chris and H. skuldur Thrainsson. 1996. VP-intemal structure and object- shift in Icelandic. Linguistic Inquiry 27: 391-444. Comrie, Bernard. 1976. Aspect. Cambridge, Mass.: Cambridge University Press. Comrie, Bernard. 1985. Tense. Cambridge, Mass.: Cambridge University Press. Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago: University o f Chicago Press. Curme, George. 1914. The development of verbal compounds in Germanic. In Beitrage zur Geschichte der deutschen Sprache und Literatur [Research into the history o f German language and literature] 39, 320-61. Davis, Henry & Lisa Matthewson. 1997. Determiners, tense, and the entity/event parallel. Paper presented at NELS, Toronto. Dayal, Veneeta. 1995. Quantification in correlatives. In Quantification in Natural Language, eds. Emmon Bach, Eloise Jelinek, Angelika Kratzer and Barbara Partee, 179-205. Dordrecht: Kluwer. Dayal, Veneeta. 1996. Locality in WH quantification: questions and relative clauses in Hindi. In Studies in Linguistics and Philosophy. Dordrecht: Kluwer. de Francis, John. 1963. Beginning Chinese. New Haven, Connecticut: Yale University Press. 492 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Demirdache, Hamida. 1991. Resumptive chains in restrictive relatives, appositives and dislocation structures. Doctoral dissertation, MIT. Den Dikken, Marcel. 1995. Particles. Oxford: Oxford University Press. Diesing, Molly. 1992. Indefinites. Cambridge: MIT Press. Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 81-138. Feng, Shengli. 1998. Prosodic structure and compound words in Classical Chinese. In New approaches to Chinese word formation, ed. Jerome Packard, 197-260. Berlin: Mouton de Gruyer. Fu, Jingqi. 1994. On deriving Chinese derived nominals: evidence for V-to-N raising. Doctoral dissertation, U. o f Massachusetts, Amherst. Gamillsheg, Emst. 1957. Historische franzosische Syntax [Hostorial French Syntax]. Tubingen: Niemeyer. Gildea, Spike. 1993. The development of tense markers from demonstrative pronouns in Panare (Cariban). Studies in Language 17-1: 53-73. Giusti, G. 1997. The categorial status of determiners. In The new comparative syntax, ed. L. Haegeman, 95-123 . London/New York: Longman. Givon, Talmy. 1971. Historical syntax and synchronic morphology. In Papers from the 7th Regional Meeting o f the Chicago Lingustic Society, 394-415. Givon, Talmy. 1984. Syntax: a functional-typological introduction, vol.I. Amsterdam: John Benjamins. Gordon, Lynn. 1986. The development of evidentials in Maricopa. In Evidentiality: the linguistic encoding o f epistemology, ed. W. Chafe, 75-88. Norwood, New Jersey: Ablex. Greenberg, Joseph. 1963. Some universals of language with particular reference to the order o f meaningful elements. In Universals o f language, ed. Joseph Greenberg, 73-113. Cambridge, Mass: MIT Press. Greenberg, Joseph. 1978. How does a language acquire gender markers? In Universals o f human language, vol. 3: Word structure, ed. Joseph Greenberg, 47-83. Stanford: Stanford University Press. Grimshaw, Jane. 1990. Argument structure. Cambridge, Mass.: MIT Press. 493 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Grosu, Alex. 1988. On the distribution of genitive phrases in Roumanian. Linguistics 26: 931-949. Guasti, M.T. & U. Shlonsky. 1992. The acquisition o f French relative clauses reconsidered. Paper presented at the Annual Boston University Conference on Language Development, Boston. Hale, Kenneth & Jay Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In The view from building 20, eds. K. Hale & J. Keyser, 53-110. Cambridge, Mass.: MIT Press. Harris, Alice. 1982. Georgian and the unaccusative hypothesis. Language 58: 290- 306. Harris, Alice & Lyle Campbell. 1995. Historical syntax in a cross-linguistic perspective. Cambridge: Cambridge University Press. Hawkins, J. A. 1979. hnplicational universals as predictors of word order change. Language 55: 618-648. Heine, B. & M. Reh. 1984. Grammaticalization and reanalysis in African languages. Hamburg: Helmut Buske Verlag. Heycock, Caroline. 1995. Asymmetries in reconstruction. Linguistic Inquiry 26:547-570. Hoekstra, Teun. 1992. Aspect and theta theory. In Thematic structure, its role in grammar, ed. I. Roca, 145-174. Berlin: Foris/Mouton. Hoji, Hajime. 1985. Logical Form Constraints and Configurational Structures in Japanese. Doctoral dissertation, University of Washington. Hopper, Paul and Elizabeth Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul and Sandra Thompson. 1984. The discourse basis for lexical categories in universal grammar. Language 60: 703-52 Hsieh, Miao-ling. 1998. Tense interpretations in Chinese and the related issues. To appear in On the form al way to Chinese language, eds. Chen-sheng Luther Liu and Sze-wing Tang. Stanford: Center for the Study of Language and Information. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Doctoral dissertation, MIT. 494 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Huang, C.-T. James. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15: 531-574. Huang, C.-T. James. 1988. Wo pao de kuai and Chinese phrase structure. Language 64: 274-311. Huang, C.-T. James. 1990. Shuo ‘shi’ he ‘you’ [On ‘Be’ and ‘Have’ in Chinese]. The Bulletin o f the Institute o f History and Philology 59-1: 43-64. Taipei, Taiwan: Academia Sinica. Huang, C.-T. James. 1993. Reconstruction and the structure of VP: some theoretical consequences. Linguistic Inquiry 24:103-13 8. Hwang, Jya-Lin. 1998. A comparative study on the grammaticalization o f saying verbs in Chinese. Paper presented at the 7th Annual Meeting o f the International Association of Chinese Linguistics/the 10th North American Conference on Chinese Linguistics. Stanford University, June 1998. Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9, 577-636. Kameshima, N. 1989. The syntax of restrictive and non-restrictive relative clauses in Japanese. Doctoral dissertation, University o f Wisconsin, Madison. Karlgren, Bernhard. 1923. Analytic dictionary o f Chinese and Sino-Japanese. Paris: Librairie Orientaliste Paul Geunther. Kayne, Richard. 1975. French syntax: the transformational cycle. Cambridge, Mass.: The MIT Press. Kayne, Richard. 1985. Principles of particle constructions. In Grammatical representation, eds. J. Gueron, H.-G. Obenauer and J.Y.PolIock, 101-140. Dordrecht: Foris. Kayne, R. 1994. The antisymmetry o f syntax. Cambridge, Mass.: MIT Press. Kennedy, Arthur. 1920. The Modern English verb-adverb combination. Language and Literature, vol.l, n o .l.. Stanford University Publications. Kim, Soonwon and Joan Maling. 1997. A cross-linguistic perspective on resultative formation. In Texas Linguistic Forum 38 The Syntax and Semantics o f Predication, eds. Blight, Ralph and Michelle Moosally, 189-205. Austin, Texas: Dept, of Linguistics, University of Texas at Austin. Kim, Young-Kook. 1997. Agreement phrases in DP. In University College London (UCL) working papers in linguistics. London: UCL. 495 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Kitagawa, Chisato and Claudia Ross. 1982. Pronominal modification in Chinese and Japanese. Linguistic Analysis 9: 119-53. Kuno, Susumu. 1973. The structure o f the Japanese language. Cambridge, Mass.: MIT Press. Lapointe, Steven and Sarah Nielsen. 1996. A reconsideration of type III gerunds in Korean. In Japanese/Korean Linguistics 5, eds. Noriko Akatsuka, Shoichi Iwasaki and Susan Strauss, 305-320. Stanford: Center for the Study of Language and Information. Lecarme, Jacqueline. 1996. Tense in the nominal system: the Somali DP. In Studies in Afroasiatic Grammar, eds. J. Lecarme, J. Lowenstamm and U. Shlonsky. The Hague: Holland Academic Graphics. Lee, Sookhee. 1993. The syntax of serialization in Korean In Japanese/Korean Linguistics 2, ed. Patricia Clancy, 447-463. Stanford: Center for the Study of Language and Information. Lehmann, W.P. 1973. Historical linguistics: an introduction. New York: Holt, Rinehart & Winston. Li, Charles and Yuzhi Shi. 1997. Lun Hanyu tibiaoji dansheng de jizhi [On the mechanism of the origin of Chinese aspect markers]. Zhongwen Yuwen 14: 83-96. Li, Charles and Yuzhi Shi. 1998. Hanyu dong-bu jiegou de fazhan yu jufa jiegou de shanbian [Development of the 'V-complement' construction and changes of syntactic structures]. In Studies in Chinese Linguistics 2, 83-100. Beijing: Beijing Language and Culture University Press. Li, Charles and Sandra A. Thompson. 1974. An explanation of word order change: S VO-> SO V. Foundations o f Language 12:201-214. Li, Charles and Sandra Thompson. 1981. Mandarin Chinese: a functional reference grammar. Berkeley: University o f California Press. Li, Fang. 1993. "Shu+ge+bu" jiegou de tantao [On "verb-ge-complement" construction]. Ms., Beijing University. Li, Frank, (in prep.) The rise and development o f verb compounding in Chinese: A historical and grammaticalization perspective with cross-linguistic implications. LinCom Europa. 4 9 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Li, Yafei. 1990. On V-V compounds in Chinese. Natural Language and Linguistic Theory 8: 177-207. Li, Yafei. 1998. Chinese resultative constructions and the uniformity of theta assignment hypothesis. In New approaches to Chinese word formation, ed. Jerome Packard, 285-310. Berlin: Mouton de Gruyer. Li, Yen-hui Audrey. 1990. Order and constituency in Mandarin Chinese. Dordrecht: Kluwer. Li, Yen-hui Audrey. 1998. Nominal projections in Mandarin Chinese. In Proceedings o f the 9th North American Conference on Chinese Linguistics, ed. Hua Lin, 139-155. Los Angeles: GSIL, University of Southern California. Li, Yen-hui Audrey. 1999a. Plural in a classifier language. Journal o f East Asian Linguistics 8: 75-99. Li, Yen-hui Audrey. 1999b. Word order, structure and relativization. Ms., University of Southern California. Lightfoot, David 1979. Principles o f diachronic syntax. Cambridge: Cambridge University Press. Lin, Jo-wang. 1993. Object expletives in Chinese. Paper presented in North American Conference on Chinese Linguistics. University of Delaware. Lin, Jo-wang. 1994. Noun phrase structure in Mandarin Chinese: DP or NP. Paper presented at the Third International Conference on Linguistics and Language in China. Hsinchu, Taiwan: Ching-hua University. Liu, Hsiao-Mei. 1993. Fanchou moxing lilun yu Hanyu shuliangci tixi [Categorial model theory and Chinese numeral-measure system]. Ms., Fuda Linguistics Institute, Taipei. Longobardi, G. 1994. Reference and proper names: a theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609-65. Lord, Carol. 1993. Historical change in serial verb constructions. Amsterdam: John Benjamins. Lu, Shu-Xiang. 1980. Xiandai Hanyu babai ci [Eight hundred words o f Modern Chinese]. Beijing: Shangwu. 497 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lii, Shu-Xiang. 1984. Ge zi de yingyong fanwei, fix lun danweici qian yi zi de tuoluo [The applicable scope of ge and the deletion of yi before the measure word]. Hanyu yufa lunwenji [Papers on Chinese grammar], Beijing: Shangwu. Liideling, Anke. 1997. Strange resultatives in German: new evidence for a sematnic treatment. Texas Linguistic Forum 38 The Syntax and Semantics o f Predication, eds. Ralph Blight and Michelle Moosally, 223-235. Austin, Texas: Dept, of Linguistics, University of Texas at Austin. Mallinson, Graham and Barry Blake. 1981. Language typology. Amsterdam/New York/Oxford: North Holland Publishing Company. Matthews, P. H. 1974. Morphology: an introduction to the theory o f word- structure. Cambridge: Cambridge University Press. Matthews, Stephen and Virginia Yip. 1994. Cantonese: a comprehensive grammar. London: Routledge. Mei, T.-L. 1981. Xiandai Hanyu wancheng mao jushi he ciwei 'le' de laiyuan [The origin of the perfective sentences of Modem Chinese and the suffix ‘le’]. Yuyan Yanjiu 1: 65-77. Mei, T.-L. 1994. Tangdai Songdai gongtongyu de yufa he xiandao fangyan de yufa [Syntax of the common language in Tang and Song Dynasties and syntax of Modem dialects]. Zhongguo Jingnei Yuyan j i Yuyanxue 2: 61-97. Mulder, Rene and Rint Sybesma. 1992. Chinese is a VO language. Natural Language and Linguistic Theory 10: 439-476. Murasugi, Keiko. 1991. Noun phrases in Japanese and English: a study in syntax, leamability and acquisition. Doctoral dissertation, University of Connecticut. Murasugi, Keiko. 1998. An antisymmetry analysis of Japanese relative clauses. Ms., Kinjo Gakuin University. Nakamura, Akira. 1994. Some aspects of temporal interpretation in Japanese. In Formal approaches to Japanese linguistics I: MIT Working Papers in Linguistics 24, eds. Masatoshi Koizumi and Hiroyuki Ura, 231-246. Cambridge, Mass.: MIT Press. Ning, Chunyan. 1993. The overt syntax of relativization and topicalization. Doctoral dissertation, University of California at Irvine. Norman, Jerry. 1988. Chinese. Cambridge: Cambridge University Press. 498 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Osawa, Fuyu. 1998. The emergence o f the D-system and the demise of morphological case in English. University College London (UCL) Working Papers in Linguistics 10, eds. John Harris and Corinne Iten, 467-98. Packard, Jerome. 1998. Introductioa In New approaches to Chinese word formation, ed. Jerome Packard, 1-34. Berlin: Mouton de Gruyer. Paris, Marie-Claude. 1979. Nominalization in Mandarin Chinese. Doctoral dissertation, Paris 7. Pesetsky, David & Esther Torrego. 2000. T-to-C movement: causes and consequences. To appear Ken Hale: A Life in Language, ed. Micheal Kenstowicz. Cambridge, Mass.: MIT Press. Peyraube, Alain. 1996. Recent issues in Chinese historical syntax. In New Horizons in Chinese Linguistics, eds. C.-T. James Huang and Y.-H. Audrey Li, 161-213. Dordrecht: Kluwer. Pulleyblank, Edward. 1995. Outline of Classical Chinese Grammar. Vancouver: UBC Press. Ritter, E. 1991. Two functional categories in noun phrases: evidence from modem Hebrew. In Syntax and Semantics 26, ed. S. Rothstein, 37-62. New York: Academic Press. Rizzi, L. 1997. The fine structure of the left periphery. In Elements o f grammar, ed. L. Haegeman, 281-337. Dordrecht: Kluwer. Roberts, Ian & Anna Roussou. 1999. A Formal Approach to “Grammaticalization.” Ms., University of Stuttgart and University o f Cyprus. Rohsenow, John. 1978. Syntax and semantics o f the perfect in Mandarin Chinese. Doctoral dissertation, University of Michigan. Ross, Claudia. 1995. Temporal and aspectual reference in Mandarin Chinese. Journal o f Chinese Linguistics 23.1: 87-135. Safir, K. 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry 17: 663-90. Sag, Ivan and Janet Fodor. 1995. Extraction without traces. In WCCFL Proceedings, eds. R. Aranovich, W. Byrne, S. Preuss and M. Senturia, 365- 384. Stanford: Center for the Study o f Language and Information. 499 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Selkirk, Elisabeth and Tong Shen. 1990. Prosodic domains in Shanghai Chinese. In The phonology-syntax connection, eds. Sharon Inkelas and Draga Zee, 313- 338. Chicago: University of Chicago Press. Shi, Dingxu. 1994. The nature of Chinese emphatic sentences. Journal o f East Asian Linguistics 3: 81-101. Shi, Ziqiang. 1988. The present and past of the particle le in Mandarin Chinese. Doctoral dissertation, University of Pennsylvania. Shi, Ziqiang. 1989. The grammaticalization of the particle le in Mandarin Chinese. Language Variation and Change 1: 99-114. Shi, Ziqiang. 1990. Decomposition o f perfectivity and inchoativity and the meaning of the particle le in Mandarin Chinese. Journal o f Chinese Linguistics 18: 95-123. Simpson, Andrew. 1997. On the status of'modifying' de and the structure of the Chinese DP. Ms., SOAS, London. To appear in On the formal way to Chinese language, eds. Chen-sheng Luther Liu and Sze-wing Tang. Stanford: Center for the Study of Language and Information. Simpson, Andrew. 1998a. Definiteness agreement and the Chinese DP. In Preceedings o f International Symposium o f Chinese Linguistics and Language. Hsinchu, Taiwan: Ching-hua University. Simpson, Andrew. 1998b. Empty determiners and nominalization in Chinese, Japanese and Korean. Paper presented in Symposium on the Syntax of East Asian Lanuges, University o f Southern California, Los Angeles. Simpson, Andrew, forthcoming. Focus, presupposition, and light-predicate raising in East and South-East Asia. To appear in Journal o f East Asian Linguistics. Simpson, Andrew & Zoe Wu. 1999. The grammaticalization of formal nouns and nominalizers in Chinese, Japanese and Korean. Paper presented at the Sheffield Conference on Language Change in Japan and East Asia. To appear in Language Change in East Asia, ed. N. Tranter. Curzon: London. Simpson, Andrew & Zoe Wu. 2000. Agreement, shells and focus. Paper presented at the 19th WCCFL. UCLA, Los Angeles. Smith, Carlota. 1997. The parameter o f aspect. Dordrecht: Kluwer. 500 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Spanos, George. 1979. Contemporary Chinese use of le: a survey and a pragmatic proposal. Jounral o f the Chinese Language Teachers Association 14.1, 36- 70, 14.2,47-102. Spencer, Andrew. 1991. Morphological theory. Cambridge: Blackwell/Cambridge University Press. Starosta, Stanley, Koenraad Kuiper, Siew-ai Ng and Zhi-qian Wu. 1998. On defining the Chinese compound word: headeness in Chinese compounding and Chinese VR compounds. In New approaches to Chinese word formation, ed. Jerome Packard, 347-370. Berlin: Mouton de Gruyer. Stowell, Tim. 1996. The syntax of tense. Ms., UCLA Sun, Chunfen. 1996. Word order change and grammaticalization in the history o f Chinese. Stanford: Stanford University Press. Sybesma, Rint. 1999. The Mandarin VP. Dordrecht: Kluwer. Szabolsci, Anna. 1994. The noun phrase. In Syntax and Semantics 27: The structure o f Hungarian, ed. K. Kiss, 179-274. New York: Academic Press. Takeuchi, Lone. 1998. The history and structure of Japanese: from Yamatokotoba to Nihongo. Ms., SOAS, London. Tang, Chih-chen Jane. 1990. Chinese phrase structure and the extended X’-theory. Doctoral dissertation, Cornell University. Tenny, Carol. 1987. Grammaticalizing aspect and affectedness. Doctoral dissertation, MIT. Travis, Lisa. 1984. Parameters and effects on word order variation. Doctoral dissertation, MIT. Tsai, W r-T. D. 1992. On the absence of island effects. Ms., MIT. Tsai, W-T. D. 1994. On economizing the theory of A-bar dependencies. Doctoral dissertation, MIT. Tsang, Chui Lim. 1981. A semantic study of modal auxiliary verbs in Chinese. Doctoral dissertation, Stanford University. T’ung, Ping-cheng and D.E. Pollard. 1997. Colloquial Chinese. London: Routledge. 501 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Vendler, Zeno. 1967. Verbs and times. In Linguistics in Philosophy. 97-121. Ithaca, NY: Cornell University Press. Revised from Zeno Vendler (1957), Verbs and times. The Philosophical Review 66: 143-60. Vennemann, Theo. 1974. Analogy in generative grammar: the origin o f word order. In Proceedings o f the 11th international congress o f linguists 2, ed. Luigi Heilmann, Societa editrice II Mulino, 79-83. Bologna: Bologna. Vergnaud, Jean-Roger. 1985. Dependances et niveauxde representation en syntaxe [Dependencies and Levels o f Representation in Syntax]. Amsterdam: Benjamins. Vergnaud, Jean-Roger and Maria Luisa Zubizarreta. 1992. The definite determiner and the inalienable construction in French and English. Linguistic Inquiry 23: 595-652. Vincent, Nigel. 1997. The emergence of the D-system in Romance. In Parameters o f morphosyntactic change, eds. Ans van Kemenade & Nigel Vincent, 149- 169. Cambridge: Cambrige University Press. Visser, F. T. 1963-1973. An historical syntax o f the English language. 3 vols. Leiden: Brill. Wahba, Wafaa. 1984. Wh-constructions in Egyptian Arabic. Doctoral dissertation, University of Illinois at Urbana-Champaign. Wang, Li. 1980. Hanyu shigao [An outline o f Chinese language history]. Beijing: China Book Company. Wang, Li. 1990. Hanyu yufa shi [History of Chinese Syntax]. In Wang Li Wenji [Collective works o f Wang Li] Qingdao: Shandong Jiaoyu Chubanshe. Wu, Guo. 1999. The origin of the Mandarin sentence-final particle le. Ms., Leiden University. Yoon, J. 1990. Korean nominalizations, lexicalism, and morphosyntactic interfaces. Ms., University of Illinois, Urbana-Champaign. Zou, Ke. 1994. Resultative V-V compounds in Chinese. In The morphology-syntax connection: MIT Working Papers in Linguistics 22, eds. Heidi Harley and Colin Phillips, 271-290. Cambridge: MIT Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Zhuangzi, cited according to Zhuangzi Yinde [A concordance to Chuang-tzu]. 1956. Cambridge, Mass.: Harvard University Press. Reprinted from Harvard- Yenching Institute Sinological Index Series, supplement no. 20. Peiping: Yenching University Press, 1947. Zubizarreta, Maria Luisa. 1998. Prosody, focus, and word order. Cambridge, Mass.: The MIT Press. Zwicky, Arnold. 1977. On clitics. Bloomington: Indiana University Linguistics Club. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Beyond words and phrases: A unified theory of predicate composition
PDF
Ellipsis constructions in Chinese
PDF
Form and meaning: Negation and question in Chinese
PDF
Asymmetry of scope taking in wh -questions
PDF
A comparative study of focus constructions
PDF
Aspects of questions in Japanese and their theoretical implications.
PDF
Children's scope of indefinite objects
PDF
Discourse functional units: A re-examination of discourse markers with particular reference to Spanish
PDF
Issues in the syntax of resumption: Restrictive relatives in Lebanese Arabic
PDF
Locatives in Chinese and Japanese: distribution, case assignment and c-command
PDF
Educational development aid and the role of language: A case study of AIT -Danida and the Royal University of Agriculture in Cambodia
PDF
An anaphoric approach to clitic position in Spanish
PDF
EPR analysis of soluble and membrane -bound annexin A2
PDF
The grammaticalization of present and past in Basque
PDF
Empires of the habanera: Cuba in the cultural imaginary of Catalonia
PDF
A geochemical and hydrological assessment of groundwater in the Portuguese Bend landslide, California
PDF
Impostors, antichrists, and satanists: Textual strategies of the central and eastern European fin de siecle
PDF
Generals and particulars in Thucydides
PDF
Counter -matching in nested case -control studies: Design and analytic issues
PDF
Authenticity, appropriation and voice: One primary classroom's writing experiences
Asset Metadata
Creator
Wu, Xiu-Zhi Zoe
(author)
Core Title
Grammaticalization and the development of functional categories in Chinese
School
Graduate School
Degree
Doctor of Philosophy
Degree Program
Linguistics
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
language, linguistics,OAI-PMH Harvest
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Li, Audrey (
committee chair
), Aoun, Joseph (
committee member
), Hoji, Hajime (
committee member
), Vergnaud, Jean-Roger (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c16-140031
Unique identifier
UC11329254
Identifier
3041544.pdf (filename),usctheses-c16-140031 (legacy record id)
Legacy Identifier
3041544-0.pdf
Dmrecord
140031
Document Type
Dissertation
Rights
Wu, Xiu-Zhi Zoe
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
language, linguistics